GBP_2012_06_AnnRpt_2011 by fRI Research

FOOTHILLS RESEARCH INSTITUTE GRIZZLY BEAR PROGRAM 2011 ANNUAL REPORT

Prepared and edited by Gordon B. Stenhouse and Karen Graham April 2012

DISCLAIMER This report presents preliminary findings from the 2011 research program within the Foothills Research Institute (FRI) Grizzly Bear Program. It must be stressed that these data are preliminary in nature and all findings must be interpreted with caution. Opinions presented are those of the authors and collaborating scientists and are subject to revision based on the ongoing findings over the course of these studies.

Suggested citation for information within this report: Karine Pigeon.. 2012. Den Site Selection for Grizzly Bears In The Rocky Mountains and Foothills of Alberta, Canada: A Spatially Explicit And Hierarchical Approach. In: G. Stenhouse and K. Graham (eds). Foothills Research Institute Grizzly Bear Program 2011 Annual Report. Hinton, Alberta. This is an interim report not to be cited without the express written consent of the senior author.

ACKNOWLEDGEMENTS A program of this scope and magnitude would not be possible without the dedication, hard work and support of a large number of people. We would like to thank the capture crew members: Bernie Goski, Dave Hobson, Terry Larsen, John Saunders and Steve Wotton and veterinarians Marc Cattet, Nigel Caulkett and Larissa Ozeki. Exemplary flying (and field!) skills were provided by John Saunders and Steve Wotton of Peregrine Helicopters of Hinton and fixed wing pilots Sean Biesbroek and Jason Harvey of Wildlife Observation Air Services. Appreciation is also extended to the summer field crew members for the various projects we had on the go: Karine Pigeon, Etienne Cardinal, Amy Stenhouse, Paul Stenhouse, Sarah Rovang, Kirby England, Lindsay Dewart and Evan Vandervalk whose hard work and enthusiasm ensured a successful field season. Special thanks to Tracy McKay for all her great work in doing a bit of just about everything that a diverse project like the bear program can dish out. Alberta Sustainable Resource Development provided logistical support and camp accommodations. The staff at the Hinton Training Centre provided a great deal of assistance in many ways including food and lodging for collaborators and summer staff during short stays. The Grizzly Bear Program Activity Team and the Board of Directors of the Foothills Research Institute provided valuable support and assistance that allowed the research to proceed in order to address management needs. Thanks to Debbie Mucha and Darren Wiens for their help in all areas relating to GIS, and a special thanks to Julie Duval for all her invaluable help with data management and GIS. A word of praise goes out to Sean Kinney, Joan Simonton and Fran Hanington for their efforts with communication support and Franâ&#x20AC;Śthanks for editing this annual report! Thank you to Jennifer Hancock, and especially Denise Lebel for an excellent job in managing the administrative details of this program and keeping all the crews in line. Definitely not an easy task! Thank you to Jerome Cranston for his ongoing GIS work (and much much more!) and to John Boulanger who provided superb statistical advice throughout the year. Dr. David Paetkau at Wildlife Genetics International completed the lab work on all DNA hair samples and Matsonâ&#x20AC;&#x2122;s Lab conducted our tooth aging. This program would not have been possible without the many program sponsors (See Appendix 1).

TABLE OF CONTENTS CHAPTER 1: SUMMARY OF 2010 CAPTURE PROGRAM .................................................... 1 Introduction................................................................................................................................... 1 Study Areas.................................................................................................................................... 1 Grizzly Bear Captures .................................................................................................................... 1 Sex and Age Characteristics .......................................................................................................... 4 GPS Radio-Telemetry Data ............................................................................................................ 4 Grizzly Bear Health Evaluation ...................................................................................................... 4

CHAPTER 2: GRIZZLY BEAR RESPONSE TO OIL AND GAS ACTIVITIES IN ALBERTA ............... 5 EXECUTIVE SUMMARY ........................................................................................................................ 5 BACKGROUND, STUDY AREA, AND RESEARCH OBJECTIVES ............................................................... 7 Background.................................................................................................................................... 7 Study Area ..................................................................................................................................... 7 Research objectives ..................................................................................................................... 10 TOPIC A: GRIZZLY BEAR USE OF HABITAT CONTAINING OIL AND GAS DEVELOPMENT AND ACTIVITIES......................................................................................................................................... 12 Abstract ....................................................................................................................................... 12 Introduction................................................................................................................................. 12 Methods ...................................................................................................................................... 14 Grizzly bear telemetry data: ................................................................................................ 14 Oil and gas features and activity: ........................................................................................ 15 Resource selection functions: ............................................................................................. 16 Analysis ........................................................................................................................................ 16 Results ......................................................................................................................................... 18 Discussion .................................................................................................................................... 25 TOPIC B: GRIZZLY BEAR USE OR AVOIDANCE OF OIL AND GAS WELLSITES ..................................... 32 Introduction................................................................................................................................. 32 Methods: ..................................................................................................................................... 33 Bear selection (use) or avoidance of the wellpad and ZOI .................................................. 33 Availability of bear foods at wellsites .................................................................................. 38 Results ......................................................................................................................................... 38 Bear selection (use) or avoidance of the wellpad and ZOI .................................................. 38 Availability of bear foods at wellsites .................................................................................. 42 Discussion .................................................................................................................................... 44 TOPIC C: GRIZZLY BEAR MOVEMENT IN RELATION TO OIL AND GAS ACTIVITIES ............................ 50 Introduction................................................................................................................................. 50 Methods ...................................................................................................................................... 51 Grizzly bear telemetry data ................................................................................................. 51 Oil and gas features ............................................................................................................. 51 Disturbance data ................................................................................................................. 52 Resource Selection Functions .............................................................................................. 52

Disturbance Buffers ............................................................................................................. 52 Velocity Analysis .................................................................................................................. 52 Latent Areas......................................................................................................................... 53 Telemetry Interpolation ...................................................................................................... 55 Results ......................................................................................................................................... 55 Discussion .................................................................................................................................... 60 TOPIC D: MORTALITY RISK RELATED TO OIL AND GAS DEVELOPMENT ........................................... 65 Introduction................................................................................................................................. 65 Methods ...................................................................................................................................... 66 Mortality data ...................................................................................................................... 66 Mortality risk surfaces ......................................................................................................... 66 Mortality risk datasets ......................................................................................................... 68 Landcover:.......................................................................................................................................................... 69

Comparison of predicted mortality risk values to reported mortality locations ................ 69 Results ......................................................................................................................................... 70 Discussion .................................................................................................................................... 72 CONCLUSIONS AND RESEARCH APPLICATIONS ................................................................................ 74 Research Applications and Future Direction ............................................................................... 75

CHAPTER 3: DEN SITE SELECTION FOR GRIZZLY BEARS IN THE ROCKY MOUNTAINS AND FOOTHILLS OF ALBERTA, CANADA: A SPATIALLY EXPLICIT AND HIERARCHICAL APPROACH ..................................................................................................................................... 79 Introduction................................................................................................................................. 79 Methods ...................................................................................................................................... 81 Study Area ........................................................................................................................... 81 Home-range scale den site selection................................................................................... 82 Grizzly bear den location data ............................................................................................. 83 Predictor variables............................................................................................................... 83 Scales of den site selection.................................................................................................. 84 Statistical Analyses: ............................................................................................................. 85 Results ......................................................................................................................................... 86 Home-range scale den site selection................................................................................... 86 Discussion .................................................................................................................................... 92 Disturbance scale ................................................................................................................ 92 Data Limitations................................................................................................................... 93 Management implications and recommendations ............................................................. 93

CHAPTER 4: FORECASTING FOOD AVAILABILITY AND ASSESSING GRIZZLY BEAR RESPONSE TO HARVESTING MATURE STANDS OF LODGEPOLE PINE: IMPLICATIONS FOR MOUNTAIN PINE BEETLE MANAGEMENT IN ALBERTA .................................................................... 111 GENERAL INTRODUCTION ......................................................................................................... 111 Study Area ................................................................................................................................. 113 TOPIC A: FOOD SUPPLY .................................................................................................................. 114 Methods .................................................................................................................................... 114 Grizzly Bear Food Resources ............................................................................................. 114 Vegetation Sampling and Environmental Characteristics ................................................. 114

iii

Predictor variables............................................................................................................. 115 Species Occurrence and Abundance ................................................................................. 116 Spatial Harvesting Sequence ............................................................................................. 117 Results ....................................................................................................................................... 118 Grizzly Bear Food Occurrence ........................................................................................... 118 Grizzly Bear Food Abundance............................................................................................ 118 Grizzly Bear Food Supply ................................................................................................... 120 Discussion .................................................................................................................................. 121 Grizzly Bear Food Occurrence in the Upper Foothills ....................................................... 121 Grizzly Bear Food Abundance in the Upper Foothills ........................................................ 121 Grizzly Bear Food Supply in the Upper Foothills ............................................................... 124 TOPIC B: GRIZZLY BEAR RESPONSE ................................................................................................. 136 Methods .................................................................................................................................... 136 Grizzly Bear GPS Locations and Home Range Area ........................................................... 136 Grizzly Bear Habitat ........................................................................................................... 136 Grizzly Bear Use and Selection of Habitat at the Population Level ................................... 138 Results ....................................................................................................................................... 138 Discussion .................................................................................................................................. 140 Conclusions................................................................................................................................ 142

CHAPTER 5: POTENTIAL IMPACTS OF CLIMATE CHANGE ON CRITICAL GRIZZLY BEAR FOOD RESOURCES IN ALBERTA.............................................................................................. 154 Project Description .................................................................................................................... 154 Progress To Date ....................................................................................................................... 155 Summer 2010 .................................................................................................................... 155 Fall 2010 ............................................................................................................................ 155 Summer 2011 .................................................................................................................... 155 Winter 2011-12 ................................................................................................................. 156 Spring 2012 ........................................................................................................................ 156 Deliverables Update .................................................................................................................. 158

CHAPTER 6: INTEGRATED SUITE OF GIS FORECASTING TOOLS FOR GRIZZLY BEAR HABITAT PLANNING .................................................................................................................. 160 Summary ................................................................................................................................... 160 Habitat Models .......................................................................................................................... 161 Processing Steps ........................................................................................................................ 163 User Input: Area Of Interest .............................................................................................. 164 User Input: Population unit ............................................................................................... 164 User input: New cutblocks ................................................................................................ 164 User Input: New pipelines ................................................................................................. 165 User Input: New roads....................................................................................................... 165 User input: Deletions ......................................................................................................... 165 User input: forecast age .................................................................................................... 165 User input: Output filename ............................................................................................. 167 Secondary inputs ............................................................................................................... 167 Regeneration of RSF models ..................................................................................................... 167 Regeneration of Risk model ...................................................................................................... 168

Regeneration of Habitat States model ...................................................................................... 168

CHAPTER 7: ESTIMATION OF DENSITY AND POPULATION TREND FOR THE YELLOWHEAD 2011 DNA GRID ........................................................................................................... 170 Introduction............................................................................................................................... 170 Methods .................................................................................................................................... 171 Study Area ......................................................................................................................... 171 2011 Survey Efforts ........................................................................................................... 171 Multiple Data Source Analysis ........................................................................................... 171 Estimation of Density ........................................................................................................ 172 Estimation of trend (2004-2011) ....................................................................................... 173 Results ....................................................................................................................................... 173 Discussion .................................................................................................................................. 174

CHAPTER 8: FUTURE OF ALBERTAâ&#x20AC;&#x2122;S FORESTS FOR GRIZZLY BEARS: CLIMATE AND LANDSCAPE SCENARIOS AND SIMULATIONS................................................................ 180 EXECUTIVE SUMMARY .................................................................................................................... 180 WORK PACKAGE 1: UPDATE LANDSCAPE CHANGE AND CONDITION PRODUCTS. 2008 TO 2010 . 180 2008-2011 STAARCH data and processing ................................................................................ 181 Results ....................................................................................................................................... 182 WORK PACKAGE 2: CLIMATE AND DEVELOPMENT OF LANDSCAPE LAYERS .................................. 184 Deviation of Data: ..................................................................................................................... 184 WORK PACKAGE 3: PHENOLOGY CAMERAS ................................................................................... 188 Methods .................................................................................................................................... 188 Field validation, phenophase codes and root nutrition data ............................................ 189 Image Analysis: .................................................................................................................. 190 Results ....................................................................................................................................... 191 Landscape forage quality................................................................................................... 192 Conclusions................................................................................................................................ 192 WORK PACKAGE 4: LANDIS SIMULATION SUMMARY .................................................................... 194

CHAPTER 9: DESIGN OF DNA-BASED OCCUPANCY SAMPLING AND POPULATION ESTIMATION FOR GRIZZLY BEARS IN NORTHERN ALBERTA .......................................... 200 Introduction............................................................................................................................... 200 Methods .................................................................................................................................... 201 Assessment of previous inventory projects to estimate detection probabilities of hair snag sites.................................................................................................................................... 201 The analysis of live-capture efforts to define a range of likely occupancy probabilities for the study area.................................................................................................................... 202 Estimation of sampling effort needed to obtain occupancy estimates of adequate precision. ........................................................................................................................... 202 Further consideration of sampling designs to obtain approximate estimates of population size ..................................................................................................................................... 202 Results ....................................................................................................................................... 202 Estimation of HS detection rate ........................................................................................ 202 Estimation of occupancy from snare data......................................................................... 204

Designs needed to obtain adequate occupancy estimates ............................................... 205 Grid cell size needed for occupancy estimation ................................................................ 207 Sampling requirements to allow estimation of population size ....................................... 209 Optimized study area size design ...................................................................................... 211 Budgets for proposed designs ........................................................................................... 213 Discussion .................................................................................................................................. 214 Recommendations .................................................................................................................... 215

CHAPTER 10: HABITAT SELECTION FOR REGENERATING CLEAR-CUT FORESTS BY BROWN BEARS: A COMPARATIVE ANALYSIS BETWEEN ALBERTA AND SWEDEN ........................ 220 Executive Summary ................................................................................................................... 220 Introduction............................................................................................................................... 221 Material and methods ............................................................................................................... 223 Study areas ........................................................................................................................ 223 Alberta ............................................................................................................................................................. 223 Sweden ............................................................................................................................................................ 223

Study area population characteristics ............................................................................... 224 Selection of bears for analysis ........................................................................................... 225 Animal location (habitat use) and home range definitions ............................................... 225 Characterizing habitat classes for 3rd order habitat selection........................................... 226 Characterizing age, size, and shape of clear-cut patches for 4th order selection .............. 226 Statistical methods ............................................................................................................ 227 3rd order habitat selection (habitat types): ..................................................................................................... 227 4th order habitat selection (within clear-cut patches): ................................................................................... 228

Evaluating support for major hypotheses ......................................................................... 229 Results ....................................................................................................................................... 229 3rd order habitat selection (habitat types) ........................................................................ 229 4th order habitat selection (within clear-cut patches) ....................................................... 231 Discussion .................................................................................................................................. 232 Regenerating forest hypothesis ........................................................................................ 232 Landscape characteristics hypothesis ............................................................................... 232 Functional response hypothesis ........................................................................................ 233 Conclusions................................................................................................................................ 234

CHAPTER 11: ADVANCEMENTS IN THE DEVELOPMENT OF HAIR CORTISOL CONCENTRATION AS AN INDICATOR OF LONG-TERM STRESS IN BROWN BEARS .......... 249 Executive Summary ................................................................................................................... 250 Introduction............................................................................................................................... 250 Materials and Methods ............................................................................................................. 252 Sources of brown bear hair ............................................................................................... 252 Laboratory analysis of hair cortisol concentration ............................................................ 254 Statistical analysis .............................................................................................................. 255 Results ....................................................................................................................................... 257 Discussion .................................................................................................................................. 266 Life history effects on HCC ................................................................................................ 267 Body condition effects on HCC .......................................................................................... 268 Capture effects on HCC ..................................................................................................... 269 HCC as a conservation tool ................................................................................................ 271

Conclusions ........................................................................................................................ 272

APPENDIX 1:LIST OF PROGRAM PARTNERS (1999 – 2010) ............................................ 280 APPENDIX 2: LIST OF PROGRAM PARTNERS (2011) ...................................................... 281 APPENDIX 3: LIST OF REPORTS, THESES AND PUBLISHED PAPERS................................. 282

vii

LIST OF TABLES CHAPTER 1: SUMMARY OF 2010 CAPTURE PROGRAM TABLE 1. GRIZZLY BEAR CAPTURES ............................................................................................................... 3

CHAPTER 2: GRIZZLY BEAR RESPONSE TO OIL AND GAS ACTIVITIES IN ALBERTA TABLE 1: GRIZZLY BEAR DEMOGRAPHICS FOR POINTS FROM ALL TIMES OF THE DAY. VALUES INDICATE NUMBER OF BEARS, WITH NUMBER OF POINTS IN BRACKETS..................................................................................... 15 TABLE 2: GRIZZLY BEAR DEMOGRAPHICS FOR POINTS SEPARATED INTO DAY AND NIGHT POINTS, WITH CREPUSCULAR POINTS REMOVED. VALUES INDICATE NUMBER OF BEARS, WITH NUMBER OF POINTS IN BRACKETS. ............... 15 TABLE 3: THE LARGEST PROPORTIONAL RESPONSE TO OIL AND GAS FEATURES BY EACH BEAR CLASS IN EACH SEASON, PARTITIONED BY DAY AND NIGHT. THE MEAN OF THE OBSERVED DISTANCES TO EACH FEATURE WAS CALCULATED FOR INDIVIDUAL BEARS AND STATISTICALLY COMPARED TO EXPECTED DISTANCES. NO SIGNIFICANT DIFFERENCE WAS CONSIDERED NEUTRAL............................................................................................................... 20 TABLE 4: MEAN OBSERVED AND EXPECTED DISTANCES TO THE OIL AND GAS FEATURES DURING DAY AND NIGHT. * INDICATES SIGNIFICANT DIFFERENCE BETWEEN OBSERVED DAY AND NIGHT MEAN DISTANCES AT Α = 0.05 USING A MANN-WHITNEY U-TEST............................................................................................................... 21 TABLE 5: COMPARISON BETWEEN THE MEAN AGE OF SELECTED AND NOT-SELECTED WELLSITES. FOR EACH INDIVIDUAL BEAR, SELECTED WELLSITES WERE DETERMINED BY COMPARING THE NUMBER OF OBSERVED TELEMETRY POINTS WITHIN 100M OF A WELL AND THE NUMBER OF RANDOMIZED POINTS WITHIN THAT SAME BUFFER. NOTSELECTED WELLSITES WERE THOSE REMAINING WITHIN THE HOME RANGE. RESULTS WERE TABULATED BY AGE/SEX CLASS. SIGNIFICANCE(P≤0.05 *) WAS DETERMINED USING THE NORMALIZED WELLSITE AGE, WHICH WAS CALCULATED FOR EACH BEAR BY SUBTRACTING THE MEAN WELLSITE AGE FROM THE INDIVIDUAL VALUES AND DIVIDING BY THE STANDARD DEVIATION. ....................................................................................... 25 TABLE 6: AREAS OF HOME RANGES AND HABITAT ZONES. ............................................................................... 36 TABLE 7: TOTAL LOCATIONS WITHIN HOME RANGES AND LOCATIONS WITHIN HABITAT ZONES. EXPECTED NUMBERS OF LOCATIONS WERE CALCULATED BASED ON TOTAL LOCATIONS FOR EACH BEAR AND THE PROPORTIONAL AREA FOR EACH ZONE (SEE TABLE 6). ................................................................................................................ 39 TABLE 8: SELECTION RATIOS BY INDIVIDUAL BEAR. (NS = NO SIGNIFICANT RELATIONSHIP DETECTED). ................... 40 TABLE 9: SELECTION RATIOS FOR INDIVIDUAL FEMALES................................................................................... 42 TABLE 10: SELECTION RATIOS FOR INDIVIDUAL MALES.................................................................................... 42 TABLE 11. AVERAGE VELOCITY (M/HR) OBSERVED FOR GRIZZLY BEAR TELEMETRY LOCATIONS IN UNDISTURBED (DISTURBANCE = 0) AND DISTURBED AREAS. O&G REPRESENTS CUMULATIVE DISTURBANCE ASSOCIATED WITH WELLS, ROADS, AND PIPELINES. ALL REPRESENTS CUMULATIVE DISTURBANCES ASSOCIATED WITH WELLS, ROADS, PIPELINES, AND CUTBLOCKS. * INDICATES SIGNIFICANT DIFFERENCE AS CALCULATED WITH A K-S TEST............ 57 TABLE 12. AVERAGE VALUES OF MEASURES OF ANTHROPOGENIC DISTURBANCES (NUMBER OF WELLS, METRES OF ROADS ANDPIPELINES, AND KM2 OF CUTBLOCKS) WITHIN 500M OF GRIZZLY BEAR LATENT POINTS (L) AND NONLATENT POINTS (NL). LATENT POINTS WERE COMPARED TO THE NON-LATENT POINTS WITH A K-S TEST; Z SCORES ARE LISTED. WITH * INDICATING A SIGNIFICANT DIFFERENCE. ....................................................... 59 TABLE 13: TOTAL REPORTED MORTALITIES IN THE KAKWA STUDY AREA FROM 1999 – 2011, BY MORTALITY CAUSE.

.................................................................................................................................................... 70 TABLE 14: MEAN PREDICTED MORTALITY RISK VALUES AT LOCATIONS OF REPORTED MORTALITIES, 2005-2010. .... 72

viii

CHAPTER 3: DEN SITE SELECTION FOR GRIZZLY BEARS IN THE ROCKY MOUNTAINS AND FOOTHILLS OF ALBERTA, CANADA: A SPATIALLY EXPLICIT AND HIERARCHICAL APPROACH TABLE 1. COEFFICIENTS FROM SAMPLING INTENSITY ANALYSE USING CONDITIONAL LOGISTIC REGRESSIONS FOR ROAD, TRUCK TRAIL, AND WELL SITE DENSITIES AT 25, 50, 100, 150, 250, AND 500 RANDOM LOCATIONS FOR 27 DENS IN THE GRANDE-CACHE MANAGEMENT UNIT AREA........................................................................ 83 TABLE 2. SPATIAL ATTRIBUTES USED TO ASSESS DEN SITE SELECTION AT THE HOME-RANGE SCALE IN THE GRANDECACHE AND YELLOWHEAD POPULATION UNITS, ALBERTA, CANADA. KM2 (KM2). ....................................... 84 TABLE 3. SAMPLING SCALE COEFFICIENTS AND AIC VALUES OF CONDITIONAL LOGISTIC REGRESSIONS FOR PROPORTIONS OF CONIFER STANDS AND HIGH-RANKED DIET-BASED HABITAT PRODUCTIVITY INDEX VALUES FOR SPRING (MAY 01 – MAY 14) AND FALL (SEPT21 - SEPT FOOD-31) AT 0.03-KM, 1-KM, 2.4-KM, AND 6.8-KM SCALES FOR DENS IN THE GRANDE-CACHE POPULATION UNIT (N = 27). BEST AIC MODELS FOR EACH CATEGORY ARE IN BOLD. .................................................................................................................................. 86 TABLE 4. COEFFICIENTS AND AIC VALUES FROM SCALE SAMPLING ANALYSIS USING CONDITIONAL LOGISTIC REGRESSIONS FOR ROAD, TRUCK TRAIL, AND WELL SITE DENSITIES AT 0.15-KM (_015), 0.3-KM (_03), 0.5-KM (_05), 1-KM (_1), AND 2-KM (_2) SCALES FOR DENS IN THE GRANDE-CACHE POPULATION UNIT (N = 27). PSTRDD; PRIMARY, SECONDARY ROAD, AND TRUCK TRAIL DENSITY, PSRDD; PRIMARY AND SECONDARY ROAD DENSITY, TRDD; TRUCK TRAIL DENSITY, WD; WELL SITE DENSITY. BEST AIC MODELS FOR EACH CATEGORY ARE IN BOLD. ........................................................................................................................................ 87 TABLE 5. MODEL SELECTION FOR FOOD, TOPOGRAPHY, HABITAT, AND ANTHROPOGENIC VARIABLES ON THE SELECTION OF DEN SITES IN THE GRANDE-CACHE POPULATION UNIT. ABBREVIATIONS: K, NUMBER OF PARAMETERS; AICC, AIKAKE INFORMATION CRITERION ADJUSTED FOR SMALL SAMPLE SIZE; Ω, AKAIKE WEIGHT; ∆ AICC, DIFFERENCE IN AICC AMONGST MODELS. BEST MODELS ARE IN BOLD. ....................................................................... 88 TABLE 6. PARAMETER ESTIMATES AND STANDARD ERRORS FOR THE TOP COMPETING MODELS IN EACH OF THE FOUR CATEGORIES OF LANDSCAPE SPATIAL ATTRIBUTES FOR THE GRANDE-CACHE POPULATION UNIT. THESE MODELS ND WERE INCLUDED IN THE 2 -LEVEL SELECTION PROCESS. ABBREVIATIONS: Β, COEFFICIENT. ........................... 89 TABLE 7. MODEL SELECTION FOR FOOD, TOPOGRAPHY, HABITAT, AND ANTHROPOGENIC VARIABLES ON THE SELECTION OF DEN SITES IN THE YELLOWHEAD POPULATION UNIT. ABBREVIATIONS: K, NUMBER OF PARAMETERS; AICC, AIKAKE INFORMATION CRITERION ADJUSTED FOR SMALL SAMPLE SIZE; Ω, AKAIKE WEIGHT; ∆ AICC, DIFFERENCE IN AICC AMONGST MODELS. BEST MODELS ARE IN BOLD. ....................................................................... 89 TABLE 8. PARAMETER ESTIMATES AND STANDARD ERRORS FOR THE TOP COMPETING MODELS IN EACH OF THE FOUR CATEGORIES OF LANDSCAPE SPATIAL ATTRIBUTES FOR THE YELLOWHEAD POPULATION UNIT. THESE MODELS ND WERE INCLUDED IN THE 2 -LEVEL SELECTION PROCESS. ABBREVIATIONS: Β, COEFFICIENT. ........................... 90 TABLE 9. FINAL MODEL SELECTION FROM BEST MODELS OF FOOD, TOPOGRAPHY, HABITAT, AND ANTHROPOGENIC CATEGORIES AND MIX HYPOTHESES FOR THE SELECTION OF DEN SITES IN THE GRANDE-CACHE POPULATION UNIT. ABBREVIATIONS: K, NUMBER OF PARAMETERS; AICC, AIKAKE INFORMATION CRITERION ADJUSTED FOR SMALL SAMPLE SIZE; Ω, AKAIKE WEIGHT; ∆ AICC, DIFFERENCE IN AICC AMONGST MODELS. BEST MODELS ARE IN BOLD. .................................................................................................................................................... 91 TABLE 10. FINAL MODEL SELECTION FROM BEST MODELS OF FOOD, TOPOGRAPHY, HABITAT, AND ANTHROPOGENIC CATEGORIES AND MIX HYPOTHESES FOR THE SELECTION OF DEN SITES IN THE YELLOWHEAD POPULATION UNIT. ABBREVIATIONS: K, NUMBER OF PARAMETERS; AICC, AIKAKE INFORMATION CRITERION ADJUSTED FOR SMALL SAMPLE SIZE; Ω, AKAIKE WEIGHT; ∆ AICC, DIFFERENCE IN AICC AMONGST MODELS. BEST MODEL IS IN BOLD. . 91 TABLE 11. PARAMETER ESTIMATES AND STANDARD ERRORS FOR THE BEST MODELS OF THE 2ND-LEVEL SELECTION PROCESS IN THE GRANDE-CACHE AND YELLOWHEAD POPULATION UNITS. ABBREVIATIONS: Β, COEFFICIENT. .. 92

CHAPTER 4: FORECASTING FOOD AVAILABILITY AND ASSESSING GRIZZLY BEAR RESPONSE TO HARVESTING MATURE STANDS OF LODGEPOLE PINE: IMPLICATIONS FOR MOUNTAIN PINE BEETLE MANAGEMENT IN ALBERTA TABLE 1. MEAN DRY WEIGHT OF FORBS (G/STEM) AND FRESH WEIGHT OF FRUITS (G/BERRY), SAMPLE SIZE AND STAND ERROR (S.E.), PERCENT DRY MATTER (D.M.) PER BERRY, AND PERCENT DRY MATTER DIGESTIBLE. MEANS AND STAND ERRORS WERE DERIVED FROM THE PLOT LEVEL AVERAGE WEIGHT (G/STEM OR G/BERRY) OF EACH SPECIES........................................................................................................................................ 126 TABLE 2. ENVIRONMENTAL VARIABLES USED TO PREDICT THE DISTRIBUTION AND ABUNDANCE OF GRIZZLY BEAR FOODS IN PINE FOREST OF THE UPPER FOOTHILLS NATURAL SUBREGION WITHIN WEYERHAEUSER GRANDE PRAIRIES FOREST MANAGEMENT AGREEMENT IN NORTH-WEST CENTRAL ALBERTA, CANADA. ................................ 127 TABLE 3. GRIZZLY BEAR FOOD OCCURRENCE PROBABILITIES IN PINE, HARVESTED (CUT), AND NON-HARVESTED (UNCUT) INCLUDING ODDS RATIOS (OR), LIKELIHOOD RATIO TESTS (LR Χ2), MODEL SIGNIFICANCE (P), AND 2 PSEUDO R FROM LOGISTIC REGRESSION ANALYSIS. THE PRESENCE OF BEAR FOODS WAS DETERMINED FROM MEANDER SEARCHES WITHIN A 30X30M SAMPLING AREA AT 147 CUT AND 142 UNCUT FIELD PLOTS IN PINE. .................................................................................................................................................. 128 TABLE 4. SAMPLE SIZE (N), NULL AND FINAL MODEL AICC SCORES, NUMBER OF MODEL PARAMETERS (K), AND PERCENT DEVIANCE EXPLAINED (D2) FROM ZERO-INFLATED NEGATIVE BINOMIAL REGRESSION MODELS DESCRIBING THE DISTRIBUTION (INFLATE) AND ABUNDANCE (COUNT) OF SHRUBS, FRUITS, AND FORBS. INFLATE REPRESENTS THE MODEL FIT PRIORI TO ADDING COVARIATES TO THE COUNT PORTION OF THE MODEL. ......... 129 TABLE 5. FEMALE GRIZZLY BEAR POPULATION LEVEL USE, SELECTION, AND AVAILABILITY OF HABITAT BY SEASON, OPERABILITY, AND STAND TYPES. ...................................................................................................... 144 TABLE 6. FEMALE GRIZZLY BEAR POPULATION LEVEL USE, SELECTION, AND AVAILABILITY OF HABITAT BY SEASON AND AGE CLASS. ................................................................................................................................... 145 TABLE 7. FEMALE GRIZZLY BEAR POPULATION LEVEL USE, SELECTION, AND AVAILABILITY OF HABITAT BY SEASON, ELEVATION, AND AGE CLASS. ........................................................................................................... 146

CHAPTER 7: ESTIMATION OF DENSITY AND POPULATION TREND FOR THE YELLOWHEAD 2011 DNA GRID TABLE 1. ESTIMATES OF AVERAGE POPULATION SIZE AND DENSITY FOR THE 2004 AND 2011 DNA SURVEYS ........ 176 TABLE 2. MODEL SELECTION OF THE HUGGINS DENSITY MODEL FOR THE 2011 GRID......................................... 176 TABLE 3. MODEL AVERAGED ESTIMATES OF SESSION-SPECIFIC DETECTION RATES FROM THE 2011 GRID ............... 176 TABLE 4. MODEL-AVERAGED ESTIMATES OF DEMOGRAPHY AND SUPERPOPULATION SIZE FROM THE PRADEL MODEL ANALYSIS . .................................................................................................................................... 177

CHAPTER 8: FUTURE OF ALBERTA’S FORESTS FOR GRIZZLY BEARS: CLIMATE AND LANDSCAPE SCENARIOS AND SIMULATIONS TABLE 1: ACQUISITION DATES OF THE 9 LANDSAT-5 SCENES USED FOR THE STAARCH DISTURBANCE DETECTION .. 181 TABLE 2 SUMMARY OF FIRE PREDICTION VARIABLES BASED ON 100 AND 200KM RADII AVERAGES AROUND HINTON. .................................................................................................................................................. 194 TABLE 3 SUMMARY OF AREA BURNED BASED ON VARIABLES PROVIDED IN TABLE 2 ........................................... 196

CHAPTER 9: DESIGN OF DNA-BASED OCCUPANCY SAMPLING AND POPULATION ESTIMATION FOR GRIZZLY BEARS IN NORTHERN ALBERTA TABLE 1: HABITAT VARIABLES USED TO DESCRIBE HABITAT-RELATIVE OCCUPANCY RELATIONSHIPS AND FACTORS INFLUENCING DETECTION PROBABILITIES FOR DNA HAIR-SNAG LOCATIONS IN ALBERTA. VARIABLES WERE SUMMARIZED AT THE SITE AND LANDSCAPE (10K BUFFER) SCALE........................................................... 201 TABLE 2: MODEL SELECTION FOR OCCUPANCY MODEL OF FIXED SITE DATA FROM THE YELLOWHEAD (2004) AND GRAND CACHE (2008) DNA INVENTORY PROJECTS. ........................................................................... 203 TABLE 3: PARAMETER ESTIMATES FROM THE MOST SUPPORTED OCCUPANCY MODEL ALONG WITH ODDS RATIOS OF SLOPE ESTIMATE PARAMETERS ......................................................................................................... 204 TABLE 4: MODEL AVERAGED ESTIMATES OF OCCUPANCY AND DETECTION PROBABILITY (FROM MODELS IN TABLE 2) .................................................................................................................................................. 204 TABLE 5: THE APPROXIMATE NUMBER OF CELLS NEEDED TO SAMPLE THE CHINCHAGA STUDY AREA 19,824 KM2... 208 TABLE 6: THE BENEFITS AND RISKS WITH VARIOUS PROPOSED DESIGNS FOR THE NW ALBERTA CHINCHAGA STUDY AREA. SEE APPENDIX 1 FOR BUDGET DETAILS. ................................................................................... 216

CHAPTER 10: HABITAT SELECTION FOR REGENERATING CLEAR-CUT FORESTS BY BROWN BEARS: A COMPARATIVE ANALYSIS BETWEEN ALBERTA AND SWEDEN TABLE 1. AREA (KM2) AND PERCENT COMPOSITION OF LAND COVER CLASSES FOR THE ALBERTA, CANADA AND SWEDEN STUDY AREAS. .................................................................................................................. 238 TABLE 2. BROWN BEAR LIFE-HISTORY PARAMETERS IN CANADA (ALBERTA) AND SWEDEN (SEE ALSO ZEDROSSER ET AL., 2012). THE NUMBER OF INDIVIDUALS A PARAMETER ESTIMATE IS BASED ON IS GIVEN IN PARENTHESES. ...... 239 TABLE 3. DESCRIPTION OF SECONDARY HYPOTHESES TESTED FOR 3RD AND 4TH ORDER HABITAT SELECTION OF CLEARCUT PATCHES IN ALBERTA AND SWEDEN BY BROWN BEARS. .................................................................. 240 TABLE 4. AVERAGE (SAMPLE SIZE OF BEARS IN PARENTHESES) SEASONAL (MAY – OCTOBER) 4TH ORDER HABITAT SELECTION OF CLEAR-CUT PATCH CHARACTERISTICS (E.G., AGE AND SIZE OF PATCH) USED BY BROWN BEARS IN ALBERTA AND SWEDEN. AVERAGE COEFFICIENTS REPORTED BY SEX-OFFSPRING CLASS AND STUDY SITE (AB = ALBERTA; SW = SWEDEN) WITH SIGNIFICANT (P<0.05) DIFFERENCES FROM ZERO INDICATED BY A SUPERSCRIPT § SYMBOL. ................................................................................................................................... 241

CHAPTER 11: ADVANCEMENTS IN THE DEVELOPMENT OF HAIR CORTISOL CONCENTRATION AS AN INDICATOR OF LONG-TERM STRESS IN BROWN BEARS TABLE 1. SOURCE, SAMPLING SITUATION, AND SUPPLEMENTARY DATA FOR BROWN BEAR HAIR SAMPLES PROVIDED TO THIS STUDY FOR THE DETERMINATION OF HAIR CORTISOL CONCENTRATION AS A MARKER OF LONG-TERM STRESS. .................................................................................................................................................. 254 TABLE 2A. COMPARISON OF MEAN HAIR CORTISOL CONCENTRATIONS BETWEEN FEMALE AND MALE BROWN BEARS FROM SEVERAL SOURCES. ............................................................................................................... 260 TABLE 2B. COMPARISON OF MEDIAN HAIR CORTISOL CONCENTRATIONS BETWEEN FEMALE AND MALE BROWN BEARS FROM SEVERAL SOURCES. ............................................................................................................... 260 TABLE 3. MODEL COEFFICIENTS DESCRIBING VARIATION IN THE HAIR CORTISOL CONCENTRATIONS OF BROWN BEARS SAMPLED IN SWEDEN AND ALBERTA BASED ON ANALYSES USING THE SAME POTENTIAL EXPLANATORY VARIABLES.................................................................................................................................... 261 TABLE 4. MODEL COEFFICIENTS DESCRIBING VARIATION IN THE HAIR CORTISOL CONCENTRATIONS OF BROWN BEARS SAMPLED IN ALBERTA BASED ON ANALYSIS USING THE FULL SUITE OF POTENTIAL EXPLANATORY VARIABLES. .. 263

TABLE 5. MEDIAN HAIR CORTISOL CONCENTRATION (HCC) AND PREDICTED BODY CONDITION STATUS OF ALBERTA GRIZZLY BEAR MANAGEMENT UNITS (BMU) BASED ON THE HCC OF HAIR SAMPLES COLLECTED FROM 340 GRIZZLY BEARS DURING HAIR-SNAG GENETIC INVENTORIES CONDUCTED IN THE PROVINCE FROM 2004-2008. .................................................................................................................................................. 265

xii

LIST OF FIGURES CHAPTER 1: SUMMARY OF 2010 CAPTURE PROGRAM FIGURE 1: STUDY AREA BOUNDARIES FOR THE 2011 CAPTURE SEASON .............................................................. 2

CHAPTER 2: GRIZZLY BEAR RESPONSE TO OIL AND GAS ACTIVITIES IN ALBERTA FIGURE 1: KAKWA STUDY AREA IN ALBERTA................................................................................................... 8 FIGURE 2: STUDY AREA WITH CURRENT DEVELOPMENTS. ................................................................................. 8 FIGURE 3: GRIZZLY BEAR HABITAT SELECTION IN THE SPRING FORAGING SEASON FOR ALL AGE CLASSES (F-ADULT FEMALES, M- ADULT MALES, SF – SUBADULT FEMALES, SM-SUBADULT MALES). OBSERVED BEAR LOCATIONS WERE COMPARED STATISTICALLY TO EXPECTED LOCATIONS AND THE PERCENTAGE OF BEARS EITHER SIGNIFICANTLY FARTHER OR CLOSER THAN EXPECTED AT RANDOM TALLIED. LOCATIONS THAT WERE NOT SIGNIFICANTLY DIFFERENT FROM RANDOM WERE CONSIDERED NEUTRAL. .................................................. 19 FIGURE 4: GRIZZLY BEAR HABITAT SELECTION IN THE SUMMER FORAGING SEASON FOR ALL AGE CLASSES (F-ADULT FEMALES, M- ADULT MALES, SF – SUBADULT FEMALES, SM-SUBADULT MALES). OBSERVED BEAR LOCATIONS WERE COMPARED STATISTICALLY TO EXPECTED LOCATIONS AND THE PERCENTAGE OF BEARS EITHER SIGNIFICANTLY FARTHER OR CLOSER THAN EXPECTED AT RANDOM TALLIED. LOCATIONS THAT WERE NOT SIGNIFICANTLY DIFFERENT FROM RANDOM WERE CONSIDERED NEUTRAL. .................................................. 22 FIGURE 5: GRIZZLY BEAR HABITAT SELECTION IN THE FALL FORAGING SEASON FOR ALL AGE CLASSES (F-ADULT FEMALES, M- ADULT MALES, SF – SUBADULT FEMALES, SM-SUBADULT MALES). OBSERVED BEAR LOCATIONS WERE COMPARED STATISTICALLY TO EXPECTED LOCATIONS AND THE PERCENTAGE OF BEARS EITHER SIGNIFICANTLY FARTHER OR CLOSER THAN EXPECTED AT RANDOM TALLIED. LOCATIONS THAT WERE NOT SIGNIFICANTLY DIFFERENT FROM RANDOM WERE CONSIDERED NEUTRAL. ...................................................................... 23 FIGURE 6. THE OBSERVED MEAN DISTANCES FOR ALL BEAR GROUPS TO THE VARIOUS OIL AND GAS FEATURES DURING A) SPRING; B) SUMMER; C) FALL. ..................................................................................................... 24 FIGURE 7: WELLSITE WITH 100M AND 500M BUFFER ZONES. ......................................................................... 34 FIGURE 8: PERCENT OF WELLPAD VEGETATED VERSUS AGE OF WELLPAD (YEARS SINCE INITIAL CLEARING). .............. 43 FIGURE 9: TOTAL ABUNDANCE OF DANDELION, CLOVER AND EQUISETUM VERSUS WELLPAD AGE (YEAR SINCE INITIAL CLEARING). .................................................................................................................................... 43 FIGURE 10: WELLPAD WITH 100M AND 500M BUFFER. POINTS ON THE WELLPAD INCLUDE LOCATIONS FOR G229 (2006), G238 (2007), G265 (2007), AND G254 (2008 AND 2009)................................................... 45 FIGURE 11. THE DETERMINATION OF A PPA ELLIPSE TO DEFINE LATENT AREAS. .................................................. 54 FIGURE 12. A COMPARISON OF AVERAGE RSF VALUES FOR LATENT AND NON-LATENT AREAS. ERROR BARS ARE ONE STANDARD DEVIATION ...................................................................................................................... 56 FIGURE 13: REPORTED GRIZZLY BEAR MORTALITIES BY YEAR, KAKWA STUDY AREA.............................................. 70 FIGURE 14: PREDICTED MORTALITY RISK IN THE KAKWA STUDY AREA BY YEAR. PERCENT INCREASES SHOWN FOR EACH YEAR. ............................................................................................................................................ 71 FIGURE 15: MORTALITY RISK VALUES IN THE KAKWA STUDY AREA FOR 2010, INCLUDING REPORTED MORTALITIES FOR

1999-2011................................................................................................................................... 71

CHAPTER 3: DEN SITE SELECTION FOR GRIZZLY BEARS IN THE ROCKY MOUNTAINS AND FOOTHILLS OF ALBERTA, CANADA: A SPATIALLY EXPLICIT AND HIERARCHICAL APPROACH FIGURE 1. MAP OF THE GRANDE-CACHE (NORTH) AND YELLOWHEAD (SOUTH) POPULATION UNITS SHOWING ELEVATION GRADIENT. ..................................................................................................................... 98 FIGURE 2. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE GRANDE-CACHE POPULATION UNIT USING THE BEST MODEL IN THE FOOD CATEGORY OF LANDSCAPE ATTRIBUTE.............................................. 99

xiii

FIGURE 3. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE GRANDE-CACHE POPULATION UNIT USING THE BEST MODEL IN THE TOPOGRAPHY CATEGORY OF LANDSCAPE ATTRIBUTE. ................................ 100 FIGURE 4. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE GRANDE-CACHE POPULATION UNIT USING THE BEST MODEL IN THE HABITAT CATEGORY OF LANDSCAPE ATTRIBUTE. ....................................... 101 FIGURE 5. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE GRANDE-CACHE POPULATION UNIT USING THE BEST MODEL IN THE ANTHROPOGENIC CATEGORY OF LANDSCAPE ATTRIBUTE............................ 102 FIGURE 6. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE GRANDE-CACHE POPULATION UNIT USING THE MOST SUPPORTED MULTIVARIATE MODEL. ......................................................................... 103 FIGURE 7. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE YELLOWHEAD POPULATION UNIT USING THE BEST MODEL FROM THE FOOD CATEGORY OF LANDSCAPE ATTRIBUTES. .................................... 104 FIGURE 8. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE YELLOWHEAD POPULATION UNIT USING THE BEST MODEL FROM THE TOPOGRAPHY CATEGORY OF LANDSCAPE ATTRIBUTES. ......................... 105 FIGURE 9. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE YELLOWHEAD POPULATION UNIT USING THE BEST MODEL FROM THE HABITAT CATEGORY OF LANDSCAPE ATTRIBUTES. ................................ 106 FIGURE 10. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE YELLOWHEAD POPULATION UNIT USING THE BEST MODEL FROM THE ANTHROPOGENIC CATEGORY OF LANDSCAPE ATTRIBUTES. .................... 107 FIGURE 11. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN THE YELLOWHEAD POPULATION UNIT USING THE MOST SUPPORTED MULTIVARIATE MODEL. ......................................................................... 108 FIGURE 12. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN A FOOTHILLS WATERSHED OF THE GRANDE-CACHE POPULATION UNIT USING THE MOST SUPPORTED MULTIVARIATE MODEL.......................... 109 FIGURE 13. MAP OF THE RELATIVE PROBABILITY OF DEN SITE OCCURRENCE IN A MOUNTAIN WATERSHED OF THE YELLOWHEAD POPULATION UNIT USING THE MOST SUPPORTED MULTIVARIATE MODEL. ............................ 110

CHAPTER 4: FORECASTING FOOD AVAILABILITY AND ASSESSING GRIZZLY BEAR RESPONSE TO HARVESTING MATURE STANDS OF LODGEPOLE PINE: IMPLICATIONS FOR MOUNTAIN PINE BEETLE MANAGEMENT IN ALBERTA FIGURE 1. RELATIONSHIP BETWEEN MEAN DRY DIGESTIBLE MATTER (G/900M2) AND STAND AGE (AGE20; 20 YEAR INTERVALS) FOR EQUISETUM SPP. AND H. LANATUM IN LEADING (SOLID) AND NON-LEADING (LONG DASH) STANDS OF LODGEPOLE PINE AND IN 2008 (LONG DASH - DOT) AND 2009 (DOT) FOR S. AMPLEXIFOLIUS. REFER TO THE SECONDARY Y-AXIS FOR EQUISETUM SPP. ................................................................................ 130 FIGURE 2. A) RELATIONSHIP BETWEEN MEAN COVER (% COVER/900M2); B) MEAN DRY DIGESTIBLE MATTER (G/900M2) OF FRUIT IN 2008; C) MEAN DRY DIGESTIBLE MATTER (G/900M2) OF FRUIT IN 2009 AND STAND AGE (AGE20; 20 YEAR INTERVAL) FOR FIVE FRUIT PRODUCING SHRUB SPECIES IN LEADING (SOLID) AND NONLEADING LODGEPOLE PINE (LONG DASH). .......................................................................................... 131 FIGURE 3. A) RELATIONSHIP BETWEEN MEAN SHRUB COVER (%/900M2); AND B) DRY DIGESTIBLE MATTER (G/900M2) OF FRUIT AS A FUNCTION OF ELEVATION (ELEV) FOR FIVE FRUIT PRODUCING SHRUB SPECIES IN LODGEPOLE PINE. .......................................................................................................................... 132 FIGURE 4. TOTAL HARVESTED AREA (HECTARES) OF LEADING LODGEPOLE PINE OVER TWELVE HARVESTING PERIODS BY STAND AGE (AGE20) WITHIN THE UPPER FOOTHILLS OF WEYERHAEUSER GRANDE PRAIRIE FOREST MANAGEMENT AGREEMENT IN NORTH-WEST CENTRAL ALBERTA, CANADA............................................. 133 FIGURE 5. CHANGES IN DRY DIGESTIBLE MATTER (DDM; TONNES) OF EQUISETUM SPP, H. LANATUM, S. AMPLEXIFOLIUS (2008 LONG DASH; 2009 DOT), AND TOTAL FORBS OVER 12 HARVESTING PERIODS WITHIN THE UPPER FOOTHILLS. REFER TO THE SECONDARY Y-AXIS FOR TOTAL FORBS AND EQUISETUM SPP................... 134 FIGURE 6. CHANGES IN A) % COVER OF SHRUBS; B) DRY DIGESTIBLE MATTER (DDM; TONNES) OF FRUIT UNDER OPTIMAL CONDITIONS (2008); C) DDM OF FRUIT UNDER SUB-OPTIMAL (2009) OVER 12 HARVESTING PERIODS FOR FIVE SHRUB SPECIES IN THE UPPER FOOTHILLS. REFER TO THE SECONDARY Y-AXIS FOR TOTAL ESTIMATES. .................................................................................................................................................. 135

xiv

CHAPTER 5: POTENTIAL IMPACTS OF CLIMATE CHANGE ON CRITICAL GRIZZLY BEAR FOOD RESOURCES IN ALBERTA FIGURE 1. GRIZZLY BEAR FOOD-PLANT PHENOLOGY OBSERVATION NETWORK .................................................. 157

CHAPTER 6: INTEGRATED SUITE OF GIS FORECASTING TOOLS FOR GRIZZLY BEAR HABITAT PLANNING FIGURE 1: HABITAT MODEL EXTENT ......................................................................................................... 161 FIGURE 2: HABITAT STATES MATRIX ......................................................................................................... 162 FIGURE 3: ARCTOOLBOX SCRIPT AND DIALOG FOR GBTOOLS ......................................................................... 164 FIGURE 4: CROWN CLOSURE VS. CUTBLOCK AGE.......................................................................................... 166 FIGURE 5: HABITAT STATES MODEL .......................................................................................................... 168

CHAPTER 7: ESTIMATION OF DENSITY AND POPULATION TREND FOR THE YELLOWHEAD 2011 DNA GRID FIGURE 1. YELLOWHEAD GRIZZLY BEAR POPULATION UNIT ILLUSTRATING TERRAIN, PROTECTED AREAS (JASPER NATIONAL PARK HAS DOTTED PATTERN, WHILE PROVINCIAL PARKS ARE STIPPLED), GRIZZLY BEAR CONSERVATION AREAS (CORE OR SECONDARY) AND THE LOCATION OF 2004 AND 2011 GRIDS AND BEAR DETECTIONS. ....... 177 FIGURE 2. DNA HAIR-SNAG SAMPLING (MONITORING) LOCATIONS ESTABLISHED DURING 2011 IN THE HINTON, CADOMIN, AND ROBB AREAS. SPATIAL SAMPLING INTENSITY WAS TWO SITES PER 50-KM2 HEXAGON. BACKGROUND MAP REPRESENTS THE MULTI-SEASONAL HABITAT SELECTION MODEL (NIELSEN ET AL. 2009). 178 FIGURE 3. GRIZZLY BEAR MONITORING SITES AND DETECTIONS OF GRIZZLY BEARS DURING THE 2011 HAIR-SNAG DNA SURVEY. BACKGROUND MAP REPRESENTS THE MULTI-SEASONAL HABITAT SELECTION MODEL (NIELSEN ET AL. 2009). ........................................................................................................................................ 179

CHAPTER 8: FUTURE OF ALBERTA’S FORESTS FOR GRIZZLY BEARS: CLIMATE AND LANDSCAPE SCENARIOS AND SIMULATIONS FIGURE 1. TEMPORAL DISTRIBUTION OF DISTURBANCE EVENTS. .................................................................... 182 FIGURE 2. OVERVIEW OF THE STAARCH DISTURBANCE EVENTS FOR 2008-2011. ........................................... 183 FIGURE 3 EXAMPLES OF GENERATED CLIMATE DATA OVER ALBERTA. ANNUAL PRECIPITATION(A) IN 2020, 2080 AND MAXIMUM JULY TEMPERATURE (B) IN 2020 AND 2080 USING 3 CLIMATE SCENARIOS (A1B, A2 AND B1). . 187 FIGURE 4 EXAMPLE IMAGES OF A) RGB IMAGE, B) 2G-RB INDEX, C) GREENUP DATE, D) END OF SEASON ............ 190 FIGURE 5: DATES OF MAYOR PHENOLOGICAL EVENTS AS OBSERVED IN THE FIELD AND AS DERIVED FROM THE CAMERA SPECTRAL RESPONSE (R2 = 0.89, P < 0.001, N=16)........................................................................... 191 FIGURE 6 AVERAGE CRUDE PROTEIN CONTENT OF HEDYSARUM ALPINUM ROOTS SAMPLED IN THREE LOCATIONS IN WEST-CENTRAL ALBERTA, CANADA IN DIFFERENT PHENOLOGICAL STAGES. ERROR BARS ARE 1 STANDARD ERROR ABOVE AND BELOW THE MEAN ........................................................................................................ 192 FIGURE 7 HISTOGRAMS OF GREENUP, FLOWERING, END OF SEASON FOR H.ALPINUM FROM THE CARDINAL DIVIDEWIDE CAMERA.............................................................................................................................. 193 FIGURE 8: SAMPLING OF FIRE AROUND HINTON, ALBERTA. ........................................................................... 195 FIGURE 9. AGE CLASS HISTOGRAM (HA) SHOWING DISTRIBUTION OF AGE CLASSES BASED ON 4 SCENARIOS SUMMARIZED IN TABLE 2................................................................................................................ 196 FIGURE 10: INITIAL SIMULATIONS OF STAND AGE BASED ON A RANGE OF FOREST FIRE SCENARIOS DESCRIBED IN TABLE 2. ............................................................................................................................................... 197

CHAPTER 9: DESIGN OF DNA-BASED OCCUPANCY SAMPLING AND POPULATION ESTIMATION FOR GRIZZLY BEARS IN NORTHERN ALBERTA FIGURE 1: LEVELS OF OCCUPANCY ASSUMING AS A FUNCTION OF NIGHTLY CAPTURE PROBABILITIES OF LIVE CAPTURE SITES ASSUMING A NAÏVE OCCUPANCY RATE OF 0.00975 OF SNARE SETS. ............................................... 205 FIGURE 2: THE ESTIMATED NUMBER OF SITES AND SESSIONS NEEDED TO OBTAIN OCCUPANCY ESTIMATES WITH A COEFFICIENT OF VARIATION OF 20% AS A FUNCTION OF DETECTION PROBABILITIES (P) AND OCCUPANCY (Ψ). .................................................................................................................................................. 206 FIGURE 3: ESTIMATED COEFFICIENT OF VARIATION OF OCCUPANCY ESTIMATES AS A FUNCTION OF OCCUPANCY AND ASSOCIATED DETECTION PROBABILITIES (LINES) FOR A STUDY WITH 2 AND 4 SAMPLING SESSIONS AND 200 SAMPLING SITES. ........................................................................................................................... 207 FIGURE 4: FOOD MODEL BASED PREDICTIONS OF BEAR HABITAT IN THE CHINCHAGA (JUNE 1-15). ..................... 209 FIGURE 5: PROPOSED SAMPLING GRID WITH CELLS THAT WOULD RECEIVE 2 SITES OUTLINED IN BLUE. THIS AMOUNTS TO 98 SITES. OTHER CELLS WOULD RECEIVE 1 SITE. ............................................................................ 210 FIGURE 6 PROPOSED OPTIMIZED SAMPLING GRID WITH CORE AREA CELLS THAT WOULD RECEIVE 2 SITES OUTLINED IN BLUE (61 CELLS), CELLS THAT WOULD RECEIVE 1 SITE OUTLINED IN GREEN (35 CELLS), AND 20X20 KM CELLS THAT WOULD RECEIVE 1 SITE (23 CELLS). .......................................................................................... 212 FIGURE 7: ESTIMATED COEFFICIENT OF VARIATION OF OCCUPANCY ESTIMATES AS A FUNCTION OF OCCUPANCY AND ASSOCIATED DETECTION PROBABILITIES (LINES) FOR THE REDUCED AREA STUDY DESIGN WITH 4 SAMPLING SESSIONS AND 180 SAMPLING SITES. ................................................................................................ 213 FIGURE 8: THE CHINCHAGA STUDY AREA WITH OPTIMIZED SAMPLING AREAS AND ROAD ACCESS AREAS DENOTED. EACH ROAD IS BUFFERED BY 1 KILOMETER. ........................................................................................ 214

CHAPTER 10: HABITAT SELECTION FOR REGENERATING CLEAR-CUT FORESTS BY BROWN BEARS: A COMPARATIVE ANALYSIS BETWEEN ALBERTA AND SWEDEN FIGURE 1. LOCATIONS OF TWO BROWN BEAR STUDY AREAS IN WEST-CENTRAL ALBERTA, CANADA AND CENTRAL SWEDEN. ..................................................................................................................................... 243 FIGURE 2. EXAMPLE OF PATCH EDGE DELINEATION (A.) AND SHAPE INDEX (B.) FOR CLEAR-CUTS IN WEST-CENTRAL ALBERTA, CANADA. ....................................................................................................................... 244 FIGURE 3. PATTERNS OF HABITAT SELECTION BY SEX-OFFSPRING CLASS AND STUDY AREA FOR 3 DIFFERENT POSTHARVEST AGE CLASSES (A. 0-9 YEARS; B. 10-25 YEARS; C. 26-60 YEARS). SELECTION COEFFICIENTS (SE) ARE BASED ON AVERAGED (INVERSE VARIANCE WEIGHTS) ANIMAL-SPECIFIC RSFS........................................... 245 FIGURE 4. AKAIKE WEIGHTS (WI) FROM AICC SCORES RANKING SUPPORT AMONG 10 A PRIORI MODELS DESCRIBING 3RD ORDER HABITAT SELECTION FOR CLEAR-CUTS BY BROWN BEARS IN ALBERTA AND SWEDEN BASED ON GENERAL AGE CLASSES (A. 0–9 YEARS; B. 10–25 YEARS; C. 26–60 YEARS). SEE TABLE 3 FOR DESCRIPTION OF HYPOTHESES. ................................................................................................................................ 246 FIGURE 5. MULTI-SEASONAL (AVERAGE) PREDICTED HABITAT SELECTION RESPONSES TO AGE AND SIZE OF CLEAR-CUTS BY STUDY AREA (ALBERTA VS. SWEDEN) AND SEX-OFFSPRING CLASS. ...................................................... 247 FIGURE 6. AKAIKE WEIGHTS (WI) FROM AICC SCORES RANKING SUPPORT AMONG 10 A PRIORI MODELS DESCRIBING 4TH ORDER HABITAT SELECTION WITHIN CLEAR-CUTS BY BROWN BEARS IN ALBERTA AND SWEDEN BASED ON AGE (A.), SIZE (B.), AND AGE X SIZE INTERACTIONS (C.). SEE TABLE 3 FOR DESCRIPTIONS OF HYPOTHESES. ......... 248

CHAPTER 11: ADVANCEMENTS IN THE DEVELOPMENT OF HAIR CORTISOL CONCENTRATION AS AN INDICATOR OF LONG-TERM STRESS IN BROWN BEARS FIGURE 1. ALBERTA GRIZZLY BEAR MANAGEMENT AREAS SAMPLED BY BARBWIRE HAIR SNAG FROM 2004-2008 FOR DNA INVENTORIES TO ESTIMATE POPULATION SIZE AND DENSITY. ......................................................... 257

xvi

FIGURE 2A. COMPARISON OF MEAN HAIR CORTISOL CONCENTRATION BETWEEN DIFFERENT SOURCES OF BROWN BEAR HAIR AND DIFFERENT METHODS OF HAIR COLLECTION. SAMPLE SIZES ARE PROVIDED IN PARENTHESES ALONG THE ABSCISSA. LETTERS ‘A’-‘D’ ARE USED TO INDICATE SIGNIFICANT DIFFERENCES (P ≤ 0.05) BETWEEN MEANS WHERE A > B > C > D BASED ON TUKEY’S B-TEST FOR A POST-HOC COMPARISON OF MEANS. ...................... 258 FIGURE 2B. COMPARISON OF MEDIAN HAIR CORTISOL CONCENTRATION BETWEEN DIFFERENT SOURCES OF BROWN BEAR HAIR AND DIFFERENT METHODS OF HAIR COLLECTION. SAMPLE SIZES ARE PROVIDED IN PARENTHESES ALONG THE ABSCISSA. LETTERS ‘A’ AND ‘B’ ARE USED TO INDICATE SIGNIFICANT DIFFERENCES (P ≤ 0.05) BETWEEN MEDIANS WHERE A > B BASED ON ALL PAIR-WISE COMPARISONS. ............................................ 259 FIGURE 3. THE EFFECTS OF A) BODY CONDITION, B) CAPTURE METHOD, AND C) RADIOTELEMETRY DEVICE ON THE PREDICTED HAIR CORTISOL CONCENTRATION FOR AN ADULT BROWN BEAR DURING DIFFERENT CAPTURE SEASONS. PREDICTED MEAN VALUES AND 95% CONFIDENCE INTERVALS (HATCHED LINES) WERE ESTIMATED FROM THE MODEL PRESENTED IN TABLE 4 AND ADJUSTED FOR AN 8-YEAR OLD BEAR ................................ 264 FIGURE 4. PERFORMANCE CURVES FOR THE APPLICATION OF HAIR CORTISOL CONCENTRATION (HCC) AS AN INDICATOR OF BODY CONDITION STATUS IN ALBERTA BROWN BEARS. THE SELECTION OF AN HCC THRESHOLD VALUE OF 2.05 PG/MG PROVIDES HIGH SENSITIVITY (91.7%) TO DETECT BEARS IN POOR BODY CONDITION, BUT LOW SPECIFICITY (44.2%) TO DETECT BEARS IN NORMAL BODY CONDITION. ................................................... 266

xvii

Chapter 1-Capture Summary

CHAPTER 1: SUMMARY OF 2011 CAPTURE PROGRAM Karen Graham Foothills Research Institute

Introduction The 2011 Foothills Research Instituteâ&#x20AC;&#x2122;s (FRI) Grizzly Bear Program (GBP) focused its capture and collaring efforts on grizzly bears located within ongoing oil and gas activities. The purpose was to collect grizzly bear location and movement data to examine habitat use with respect to resource extraction activities and denning. We continued to collect important information on health parameters of all grizzly bears handled during our operations. In 2011 we also provided capture and handling assistance to a study being conducted by Alberta Tourism, Parks and Recreation in association with the University of Alberta (U of A) looking at human activities and grizzly bears.

Study Areas We captured and sampled grizzly bears in two distinct study areas in 2011 (Figure 1). Six grizzly bears were captured between Grande Prairie and Grande Cache, known as the Kakwa study area. Eight grizzly bears were captured south of Hinton and east of Jasper National Park, known as the U of A study area.

Grizzly Bear Captures The capture crew consisted of biologists and veterinarians with experience in grizzly bear capture. Grizzly bears were captured via helidarting or culvert traps fitted with satellite trap alarm systems that were placed along existing forest access roads. Field capture efforts began in early May in the Kakwa study area. The crew was based out of the Kakwa Tower camp. A limited amount of helicopter darting was employed which targeted specific bears for recapture and collar replacement but most captures were from culvert traps.

Chapter 1-Capture Summary

Figure 1: Study area boundaries for the 2011 capture season

Capture efforts for the U of A study area began in late May. Again, helicopter darting was targeted at specific bears for recapture and collar replacement. New bears were captured in culvert traps. No fall captures occurred. One bear we captured had a tattoo visible but illegible. We suspect this bear may be G016. DNA was collected and will be used for the correct identification.

Chapter 1-Capture Summary

Table 1. Grizzly bear captures

name

Date

Recapture

Sex

Age Class Population Unit

Capture Method

G004

05-Jun-11

yes

adult

Yellowhead

heli dart

G016?

15-Jun-11

yes

adult

Yellowhead

heli dart

G023

05-Jun-11

yes

adult

Yellowhead

heli dart

G037

28-Jun-11

yes

adult

Yellowhead

heli dart

G111

21-Jun-11

yes

adult

Yellowhead

heli dart

G115

21-Jun-11

yes

adult

Yellowhead

heli dart

G118

22-Jun-11

yes

adult

Yellowhead

culvert trap

G119

12-Jun-11

subadult Yellowhead

snare

G251

16-May-11

yes

adult

Grande Cache

heli dart

G254

21-May-11

yes

adult

Grande Cache

heli dart

G270

12-May-11

yes

adult

Grande Cache

heli dart

G273

18-May-11

adult

Grande Cache

culvert trap

G274

18-May-11

adult

Grande Cache

culvert trap

G276

24-May-11

subadult Grande Cache

culvert trap

We anaesthetized grizzly bears using a combination of xylazine and Telazol administered by remote drug delivery, e.g., dart rifle or jab pole. Once immobilized, grizzly bears were weighed, and measured (chest girth, zoological length, and straight-line length). Samples were collected (blood, hair, skin biopsy, and tooth). Radio-collar and ear tag transmitters were attached. A transponder (microchip) was also inserted beneath the skin for future identification purposes. Vital functions and blood-oxygen levels were monitored throughout the handling period. Following handling, we administered atipamezole to reverse the effects of anaesthesia and monitored the grizzly bears until they showed imminent signs of recovery. We re-checked all bears within 24 hours of capture to ensure they had recovered fully from immobilization. All details of capture operations conformed to national standards on the capture and handling of ursids as well as provincial standards. In total, we captured 14 grizzly bears in our 2011 field season (Table 1). Six bears were caught in the Kakwa area, and 8 were captured in the U of A study area south of Hinton, Alberta. No black bears were handled this field season and no other non-target species were captured. No capture related mortalites occurred during the 2011 field season.

Chapter 1-Capture Summary

Sex and Age Characteristics Of the 14 grizzly bears captured 12 (86%) were adults, 2 (14%) were sub-adults, 5 (36%) were males, 9 (64%) were females (Table 1). Adult females were captured most often (57%), followed by adult males (29%) and subadult females and subadult males (7% each). No cubs of the year were caught.

GPS Radio-Telemetry Data We deployed a Global Positioning System (GPS) radio-collar and (VHF) ear-tag transmitter on captured bears. All radio-collars have an integrated remote release mechanism in addition to a rot-off system as a backup in case of electronic failure. Thirteen radio-collars were deployed and consisted of Follow-it Tellus II or I, Follow-it Satellite collars and Telemetry Solutions Satellite collars. Follow-it collars collect locations on the following schedule: ď&#x201A;ˇ ď&#x201A;ˇ

April 1 to November 31 - 1 location/hours. December 1 to March 31 - 1 location/2 hours

Telemetry Solution collars collected locations either hourly or every 2 hours. We conducted data upload flights for all Tellus-collared bears in the Kakwa study area once in the summer and fall. We collected over 100,000 GPS location points from the Tellus and Satellite collars combined.

Grizzly Bear Health Evaluation We gathered health information from all 14 grizzly bears as part of our research activities. The data from these bears include data on physical and physiological measurements recorded at capture as well as results from subsequent laboratory analyses of blood serum, skin, and hair. All health data for 2011 are now entered into our project health database.

Chapter 2-Response to Oil and Gas Activities

CHAPTER 2: GRIZZLY BEAR RESPONSE TO OIL AND GAS ACTIVITIES IN ALBERTA Final Report Prepared For:

Alberta Upstream Petroleum Research Fund Edited by Tracy McKay2 and Gordon Stenhouse2 Research Team: Karen Laberee1, Benjamin Stewart1, Trisalyn Nelson1, Tracy McKay2, Gordon Stenhouse2, Amit Saxena3, Ellinor SahlĂŠn4, Jed Long1, Darren Wiens2, Jerome Cranston5 1

Spatial Pattern and Analysis Laboratory, University of Victoria, British Columbia. Foothills Research Institute Grizzly Bear Program, Hinton, Alberta 3 Devon Canada Corporation, Calgary, Alberta 4 PhD student, Department of Wildlife, Fish and Environmental Studies, SLU, Sweden. 5 Arctos Ecological Services, Hinton, Alberta. 2

EXECUTIVE SUMMARY The Kakwa region in west-central Alberta has ongoing recreational use, timber harvesting activities, and extensive oil and gas activities, and it is expected that resource extraction in this area will increase in the near future. The Kakwa also contains important habitat for grizzly bears, a species designated as threatened in Alberta. In 2010 and 2011, the Foothills Research Institute Grizzly Bear Program (FRIGBP) received funding from the Alberta Upstream Petroleum Research Fund (AUPRF) and other program partners to investigate how grizzly bears are responding to oil and gas activities within the Kakwa study area. Although previous publications addressed the exploration and development phases of oil and gas extraction, to our knowledge, there are currently no published data regarding grizzly bear response to ongoing oil and gas operations in western Canada. The overall goal of this two year research project is to assist the oil and gas industry and resource managers in assessing, managing and mitigating the potential impacts of development on grizzly bears and their habitat. Our investigation of bear response to oil and gas activities was focused to address the main topics below: Topic A: Grizzly bear use of habitat containing oil and gas development and activities 5

Chapter 2-Response to Oil and Gas Activities Topic B: Grizzly bear use or avoidance of oil and gas wellsites Topic C: Grizzly bear movement in relation to oil and gas activities, and Topic D: Grizzly bear mortality risk in relation to oil and gas development. We investigated whether bears were closer, no different, or further from oil and gas features than expected based on a conditional randomization analysis. Our results show that grizzly bears use habitat with oil and gas activities differently than what was expected at random. Bears in our study generally did not avoid habitat containing oil and gas features during spring. During summer, males used habitat with oil and gas development less than expected, and fall was the season with the most changes in large scale habitat use in response to oil and gas features. At a smaller spatial scale, we analyzed selection ratios to determine if bears were selecting for the wellpad area, and we completed vegetation assessments at wellsites. Our results show that the majority of bears selected for wellpads, and most bears did not avoid the 500m zone surrounding wellsites. Use of the wellpad is likely related to the growth of bear foods on the wellpad and along the edges of the wellsites. To investigate grizzly bear movement rates in response to oil and gas features, we used movement velocities and a novel spatial analysis method to look at areas of slow movement (latent areas). There were no obvious increases in grizzly bear movement rates in direct response to the presence of wellsites in our analysis; however, our results indicate faster movement when roads or pipelines were present. To evaluate mortality rates and predicted mortality risk, we analyzed current mortality datasets and applied mortality risk models to create annual mortality risk surfaces specific to the Kakwa study area. Human-caused mortalities were similar to those observed in other areas of Alberta. Our estimated annual mortality rate was 1.7%. Mortality risk steadily increased throughout 2004 to 2010. The mean mortality risk ranged widely between home ranges. Comparison of reported mortality locations with predicted mortality risk values indicates that the current mortality risk model is a good predictor of mortality risk for the Kakwa study area. This research addresses the current knowledge gap regarding how grizzly bears respond to oil and gas operations. Results and management recommendations presented in this project may be applied to the Kakwa region and across other areas of oil and gas development in North America to enhance and improve land management in grizzly bear range.

Chapter 2-Response to Oil and Gas Activities

BACKGROUND, STUDY AREA, AND RESEARCH OBJECTIVES Tracy McKay and Gordon Stenhouse Background Alberta has an abundance of petroleum reserves that are currently being explored, developed and utilized. The Kakwa region in west-central Alberta (Figures 1 and 2) has ongoing timber harvesting activities as well as extensive oil and gas exploration and development. As additional shale gas opportunities emerge in this area, it is expected that development will increase in the near future. The Kakwa region also contains important habitat for grizzly bears, a species designated as threatened in Alberta. This region contains designated core grizzly bear conservation areas as defined by the provincial government (Alberta Sustainable Resource Development [ASRD]). In both 2010 and 2011, the Foothills Research Institute Grizzly Bear Program (FRIGBP) received funding from the Alberta Upstream Petroleum Research Fund (AUPRF) and other funding partners to investigate how grizzly bears are responding to oil and gas activities within the Kakwa study area. The overall goal of this two year research project was to assist the oil and gas industry and resource managers in assessing, managing and mitigating the potential impacts of energy sector development on grizzly bears and their habitat.

Study Area The study area was comprised of 8,334 square kilometres within a diverse, multi-use landscape in west-central Alberta, Canada (Figures 1 and 2). The area includes high elevation snow, rock, and ice in the west along with increasing anthropogenic disturbance in the low elevation foothills to the east. Elevation ranges from 549m to 2446m, annual precipitation varies from 550mm to 1050mm, and mean daily temperatures range from 4.7 to 11.3째C (Natural Regions Committee 2006). Over half of the area is conifer or coniferdominated mixed forest. Resource extraction industries have been active in the area for a number of decades (White et al. 2011), with most disturbances in the area arising from the forest industry and oil and gas development (Schneider 2002). Approximately 76% of the forested land base in the Kakwa region is managed for timber harvesting.

Chapter 2-Response to Oil and Gas Activities

Figure 1: Kakwa study area in Alberta.

Figure 2: Study area with current developments.

Chapter 2-Response to Oil and Gas Activities Previous studies of grizzly bear response to industrial development have included significant research regarding roads and forestry operations. A number of previous studies have also investigated the response of bears to oil and gas exploration and development, including seismic surveys, exploratory drilling, construction of facilities and roads, and human-bear conflicts at camps and facilities (Harding & Nagy, 1980; Schallenberger, 1980; Tietje & Ruff, 1983; Reynolds et al., 1986; McLellan & Shackleton, 1989; Follmann, 1990; Amstrup, 1993). Additional work has investigated grizzly bear landscape use in response to seismic cutlines (Linke, 2005), and the use of edge habitat along roads and pipelines (Stewart, 2011). Preliminary investigation of grizzly bear response to oil and gas wellsites in the Kakwa study area of Alberta suggested that bears do not avoid wellsites, but may select for the area around the wells (SahlĂŠn, 2010). Although previous publications addressed the exploration and development phases of oil and gas extraction, to our knowledge, there are currently no published data regarding grizzly bear response to ongoing oil and gas operations in western Canada. The Kakwa region has a relatively high density of oil and gas development and activities, and the density of oil and gas features is increasing each year (Appendix A). As oil and gas development continues to expand throughout grizzly bear habitat, it is important to understand how these activities may impact grizzly bears.

Chapter 2-Response to Oil and Gas Activities

Research objectives Our investigation of bear response to oil and gas activities was focused on addressing the main potential impacts of oil and gas development: effects on habitat use (broad scale and fine scale), influence on movement patterns, and changes to mortality risk. The specific research objectives were: 1. Grizzly bear use of habitat containing oil and gas development and activities: a. Do grizzly bears use habitat containing oil and gas development (active wellsites, inactive wellsites, roads, and pipelines) more, less, or no differently than expected in each season? b. Do grizzly bears use habitat containing oil and gas development (active wellsites, inactive wellsites, roads, and pipelines) more, less, or no differently than expected during day and night (in each season)? c. Does grizzly bear response change during the life cycle of oil and gas operations; is selection of wellsites affected by the age of the wellsite (years since initial drilling)? 2.

Grizzly bear use or avoidance of oil and gas wellsites: a. Do grizzly bears show selection (use) or avoidance of the wellpad itself and/or the zone immediately surrounding the wellpad? b. Are grizzly bear foods available on wellpads?

3. Grizzly bear movement in relation to oil and gas activities: a. Are there differences in grizzly bear movement rates (velocities) related to the presence or absence of disturbance features in the vicinity? b. Are the densities of disturbance features different in areas of slow movement as compared to the rest of the home range? 4. Grizzly bear mortality risk in relation to oil and gas development: a. What are the numbers of reported mortalities and their causes in the Kakwa region? b. What is the predicted annual mortality risk in the Kakwa study area, and how does annual mortality risk change from 2005 to 2010? c. How does mortality risk vary between grizzly bear home ranges, and from year to year for individual bears? d. How do known mortality locations in the Kakwa compare to predicted risk values from the existing mortality risk model?

This research addresses the current knowledge gap regarding how grizzly bears respond to oil and gas operations. Results from this project may be applied in the Kakwa region and other areas of western North America to enhance and improve land management in grizzly bear range.

Chapter 2-Response to Oil and Gas Activities

Literature cited Amstrup, S. C. 1993. Human disturbances of denning polar bears in Alaska. Arctic 46:246250. Follmann, E. H. and J. L. Hechte. 1990. Bears and pipeline construction in Alaska. Arctic 43:103-109. Harding, L. and J. A. Nagy. 1980. Responses of grizzly bears to hydrocarbon exploration on Richards Island, Northwest Territories, Canada. Bears: Their Biology and Management 4:277-280. Linke, J., S. E. Franklin, F. Huettmann, and G. B. Stenhouse. 2005. Seismic cutlines, changing landscape metrics and grizzly bear landscape use in Alberta. Landscape Ecology 20:811-826. McLellan, B. N. and D. M. Shackleton. 1989. Grizzly bears and resource-extraction industries: habitat displacement in responce to seismic exploration, timber harvesting and road maintenance. Journal of Applied Ecology371-380. Reynolds, P. E., H. V. Reynolds, and E. H. Follmann. 1986. Responses of grizzly bears to seismic surveys in northern Alaska. International Conference on Bear Research and Management 6:169-175. Sahlen, E. 2010. Do grizzly bears use or avoid wellsites in west-central Alberta, Canada? MSc Thesis, Swedish University of Agricultural Sciences, Faculty of Foresty, Department of Wildlife, Fish and Environmental Studies. Schallenberger, A. 1980. Review of oil and gas exploitation impacts on grizzly bears. BearsTheir Biology and Management 4:271-276. Stewart, B. P. S. Quantifying grizzly bear habitat selection in a human disturbed landscape . 2011. University of Victoria. Tietje, W. D. and R. L. Ruff. 1983. Responses of black bears to oil development in Alberta. Wildlife Society Bulletin 11:99-112.

Chapter 2-Response to Oil and Gas Activities

TOPIC A: GRIZZLY BEAR USE OF HABITAT CONTAINING OIL AND GAS DEVELOPMENT AND ACTIVITIES Karen Laberee, Benjamin Stewart, Trisalyn Nelson, Tracy McKay, Gordon Stenhouse, and Amit Saxena

Abstract Oil and gas development is increasing in areas of Alberta that are also home to threatened grizzly bear (Ursus arctos) populations. The effects of forestry operations and roads on grizzly bear habitat use have been well studied. However, researchers are only beginning to specifically examine possible impacts of the oil and gas industry. In this chapter, we characterize the spatial patterns of grizzly bear habitat use in the Kakwa area of Alberta, Canada in relation to oil and gas development and operations. Datasets were obtained for roads, pipelines, and wellsites. Wells were categorized as active or inactive. Grizzly bear telemetry data collected from 2005â&#x20AC;&#x201C;2010 were partitioned by year, foraging season, and age/sex classes, and individual home ranges were created. Data were also analyzed by time of day (day/night). A series of resource selection functions (RSFs) were created for each season of each year. Telemetry locations were randomized conditional to home ranges and RSFs to create expected grizzly bear locations. To determine if bear locations were closer, further, or no different than expected from oil and gas features, the distances from observed locations to oil and gas features were compared to the distance distribution of randomized locations. In general, bears were closer than expected to all features during spring. Adult males were farther than expected to all features during the summer season. Active wellsites were avoided by all groups in the fall. Out of the wellsites available within their home ranges, bears generally selected for older wellsites (11-14 years postdevelopment).

Introduction In Alberta, Canada it is estimated that fewer than 700 individual grizzly bears (Ursus arctos) remain (Alberta Sustainable Resource Development 2010). As such, the species was designated as Threatened by the provincial government in 2010 (Clark & Slocombe 2011). It is thought that the current abundance and distribution of this species in Alberta can be directly attributed to excessive human-caused mortality through overhunting, poaching, management removal, and defense of life and property over the last century (Weaver et al. 1996; Garshelis et al. 2005). Ongoing activities related to natural resource extraction create an abundance of new linear features (i.e. roads, seismic lines, and pipelines), increasing levels of human access into grizzly bear habitat. Human access into grizzly bear habitat becomes a management concern, as grizzly bear mortality rates have been shown to rise with increasing road density (Benn & Herrero, 2002; Nielsen et al. 2004a). Understanding grizzly bear habitat use and response to natural resource extraction is essential for effective conservation and recovery efforts in Alberta. Telemetry data collection has been widely used to determine the movements of grizzly bears and to 12

Chapter 2-Response to Oil and Gas Activities understand their habitat use (Nielsen 2004). As large omnivores, grizzly bears must devote considerable time to foraging over large areas (Schwartz et al. 2003). Grizzly bears are known to change their diet to correspond with seasonal changes in food availability (Munro et al. 2006). Knowledge of grizzly bear habitats, landscape conditions and anthropogenic features has contributed to the creation of resource selection functions (RSFs) that spatially describe grizzly bear habitat quality and predict seasonal bear occurrence (Nielsen 2007). An RSF describes the probability of resource use across the landscape (Manly 2002), and many studies have utilized RSFs to analyze the impact of human disturbance on habitat use, including studies on elk (Edge et al. 1987), southern mule deer (Bowyer and Bleich 1984), and grizzly bears (Nielsen 2005; Ciarniello et al. 2007). Many large carnivores have been shown to alter their use of habitat under various anthropogenic pressures, ranging from industrial development to tourism (Olson et al. 1998; Boydston et al. 2003). Previous studies have shown that grizzly bear habitat selection is impacted by human activity (McLellan 1990; Gibeau et al. 2002; Roever et al. 2008; Graham et al. 2010). Changes to habitat use may be spatial, such as selecting or avoiding an area; temporal, using the area at different times of the day or year; or a combination of both. In Scandinavia, where most of the landscape is affected by the presence of humans, it was observed that bears did not avoid habitat with human presence, but used the habitat at night when human activity was lower (Martin et al. 2010). This spatial-temporal avoidance of humans has also been previously observed with wolves in Alberta (Hebblewhite & Merrill 2008) and grizzly bears in Alaska (Rode et al. 2006). Behavioural adaptations to human activity have been shown to manifest differently between male and female grizzly bears (Gibeau et al. 2002) and between adults and subadults (Gibeau et al. 2002; Mueller et al. 2004). For example, adult male grizzly bears have been shown to use high quality roadside habitat at night, while females used this period of human inactivity to forage near other features such as human settlements (Gibeau et al. 2002). In an analysis of road crossings in west-central Alberta, Graham et al. (2010) reported that males had lower crossing indices than females during daylight hours. Further, a study by Mueller et al. (2004) on the selection of habitat containing anthropogenic features showed spatial-temporal differences in selection for subadults, and suggested that their avoidance of the more dominant adult males superseded avoidance of humans. Oil and gas development and related activities are expanding in many areas of Alberta that are designated as core and secondary conservation zones for grizzly bears. The drilling of wellsites represents a period of concentrated intense human activity on the landscape. Unlike research related to roads and forestry, limited research has been completed regarding the specific effects of oil and gas activities on grizzly bear habitat selection (Linke et al. 2005; Sahlen 2010). Knowing how grizzly bears in this region react to disturbances is crucial to outlining successful conservation strategies (Weaver et al. 1996). While there is evidence that grizzly bears show selection for anthropogenic disturbances, such as forestry cutblocks (Elgmork and Kaasa 1992; Nielsen et al. 2004b; Stewart et al. 2012), this selection increases the risk of human-bear interactions, and results in increased mortality risk for 13

Chapter 2-Response to Oil and Gas Activities bears (Nielsen et al. 2004b). Alternatively, avoidance of anthropogenic features and activities can result in functional habitat loss (Dyer et al. 2001). As is the case with many human disturbances, the effects of oil and gas development can be examined through two lenses: increased human presence on the landscape and the alteration of the landscape itself. The objective of our research was to determine if grizzly bear habitat use is affected spatially and/or temporally by oil and gas development and activities. To achieve this we framed our research to answer the following questions: a. Do grizzly bears use habitat containing oil and gas development (active wellsites, inactive wellsites, roads, and pipelines) more, less, or no differently than expected in each season? b. Do grizzly bears use habitat containing oil and gas development (active wellsites, inactive wellsites, roads, and pipelines) more, less, or no differently than expected during day and night (in each season)? c. Does grizzly bear response change during the life cycle of oil and gas operations; is selection of wellsites affected by the age of the wellsite (years since initial drilling)? To meet our objectives, we took a spatial pattern approach to quantifying the impact of oil and gas activity on grizzly bear habitat use. Based on theory in spatial ecology (Fortin & Dale, 2005) and spatial statistics (Getis and Boots, 1978; Haining, 2003), we built our research on the notion that spatial patterns are an expression of spatial processes. When it is not possible to measure spatial processes directly, characterizing spatial patterns is a mechanism for characterizing the nature of processes on the landscape. Here, we used spatial patterns of grizzly bears on the landscape in reference to the distribution of oil and gas related disturbance as a mechanism for characterizing how oil and gas activities have impacted grizzly bear habitat use.

Methods Grizzly bear telemetry data: Telemetry data were collected for 33 grizzly bears from 2005â&#x20AC;&#x201C;2010. Aerial darting, leg-hold snaring, and culvert traps were used to capture grizzly bears following capture protocols accepted by the Canadian Council of Animal Care for the safe handling of bears (animal use protocol number 20010016) (Stenhouse & Munro 2000). Captured bears were fitted with a Televilt/Followit brand GPS collar (Lindesburg, Sweden), which collected grizzly bear locations once per hour. Data from collars were collected remotely using monthly VHF data upload flights from fixed winged aircraft. Telemetry data were cleaned for accuracy using positional dilution of precision (PDOP), which evaluates three-dimensional accuracy of locations (D'Eon & Delparte 2005). While the removal of GPS locations with dubious accuracy has been addressed in habitat selection studies (Frair et al. 2004), we decided to eliminate spatially inaccurate locations in this study, given our analysis methods focus on Euclidean distance. Of the 116,611 grizzly 14

Chapter 2-Response to Oil and Gas Activities bear points analyzed, 2984 (3%) of the points had a PDOP > 10 and were excluded from the analysis. Grizzly bear telemetry data were separated into seasonal categories following established feeding patterns for our study area (Munro et al. 2006; S. Nielsen, pers. comm.). Seasons were defined as spring (1 May to 15 June), summer (16 June to 31 July), and fall (1 August to 15 October) (Nielsen 2005). Grizzly bear data were partitioned into sets of telemetry points for individual bears by separate years and seasons. The full dataset was also divided into daytime and nighttime locations based on sunrise and sunset calculators (pers. comm. Julie Duval, FRI). Data falling into crepuscular periods (twilight morning and twilight evening), were removed in order to restrict analysis to distinct periods of daylight and darkness presumed to correspond with periods of human activity. Data were grouped based on age/sex classes, with bears 3 to 4 years of age considered as subadults and those 5 and greater considered as adults. Datasets with less than 50 GPS locations were excluded due to the effects of small sample sizes on home range calculations (Seaman & Powell 1996). The demographic breakdown of grizzly bear data points is shown in Table 1 (all points) and Table 2 (by day and night). Table 1: Grizzly bear demographics for points from all times of the day. Values indicate number of bears, with number of points in brackets. Spring Summer Fall Males 8(5421) 7(10981) 5(7075) Females 11(10686) 12(16566) 12(16884) Subadult Males 5(2604) 7(10957) 6(11255) Subadult Females 5(3401) 5(4166) 5(5308) Table 2: Grizzly bear demographics for points separated into day and night points, with crepuscular points removed. Values indicate number of bears, with number of points in brackets. Spring Summer Fall Day Night Day Night Day Night Male 8(3924) 7(1398) 7(8193) 5(2698) 5(4132) 5(2943) Female 11(7972) 11(2689) 12(12279) 12(4244) 12(9128) 12(7756) Subadult Male 5(1943) 3(592) 7(8292) 6(2649) 6(6418) 6(4837) Subadult Female 4(2479) 4(860) 5(3030) 4(1005) 5(2957) 5(2351)

Oil and gas features and activity: Disturbance datasets were obtained describing oil and gas wellsites, pipelines, and roads. Wellsite and pipeline data were provided to us from a database maintained by Alberta Energy; spatial data and attributes were originally obtained directly from oil and gas operators. Roads in this area are the product of oil and gas development as well as forestry activities. The roads dataset was originally obtained from Alberta Sustainable Resource Development, and subsequently updated by our research team through heads-up digitizing using medium and high resolution imagery. Road data were attributed with a year of 15

Chapter 2-Response to Oil and Gas Activities disturbance (if constructed after 1994), pipeline data were attributed with date of construction, and wellsite data were attributed with date of initial clearing and drilling, first date of active pumping, and reclamation/deactivation dates. Wellsites were further divided into active and inactive wells based on dates of original drilling and end of production or abandonment. Activity status for wells was specific for each season, and a status of â&#x20AC;&#x153;activeâ&#x20AC;? indicates that a well was drilled and actively pumping during that seasonal time period. Resource selection functions: Grizzly bear habitat quality was inferred from resource selection function (RSF) values. A series of seasonal RSFs were generated in order to condition the randomization process described below. An RSF is estimated through a comparison of observed resource use to availability of resource units (Boyce et al. 2002). We created our seasonal RSF following Nielsen et al. (2009). Individual RSFs were created for each foraging season (spring, summer, and fall) and each year of our study area (2005â&#x20AC;&#x201C;2010), creating a total of 18 RSFs. The RSF model predicts probability of resource use based on land cover, crown closure, species composition (conifer canopy), compound topographic index, distance to streams, and distance to forest clearings (Nielsen et al. 2009). Models were created from a random sample of 90% of the telemetry data, with the remaining set aside for model evaluation. RSF accuracy was evaluated using the remaining 10% of the training data using Spearman rank coefficients. The results from 2007 showed p-values of 0.005, 0.0013, and 0.0010 for the spring, summer, and fall, respectively, indicating sufficient accuracy for use in further analysis.

Analysis To determine if grizzly bears were closer, further, or no different than expected to oil and gas features, the Euclidean distance from each telemetry point to the nearest oil and gas feature of each type was calculated. For each bear, the average observed distance was compared to an expectation. Average observed and randomized distances were also compared and results of individual bears summarized for seasonal/age/sex classes. Comparisons of observed and expected (randomized) average distance to oil and gas features were repeated for telemetry data partitioned by time of day. Sets of individual grizzly bear locations (single bear, in a single season, in a single year) were compared to available wellsites within their home range to determine if wellsites that were selected were significantly different in age compared to those that were not selected. 1. Randomization of grizzly bear data Randomization (Monte Carlo) approaches were used to generate the expectation for comparison (i.e., Nelson & Boots 2005). For each grizzly bear, in each season, telemetry locations were randomized and the average distance from randomized telemetry locations to oil and gas features calculated. Randomization was performed 99 times, enabling statistical tests with a critical value of 0.01.

Chapter 2-Response to Oil and Gas Activities Intuitively, we know that grizzly bear habitat use is not random, and comparisons to complete spatial randomness will yield unreasonable results. Therefore, randomization was conditioned by limiting locations to the grizzly bear home range. Home ranges were generated for each bear using kernel density estimation (KDE). The 95 th percentile by volume contour was applied to a KDE surface generated from telemetry points using a Gaussian kernel (Seaman and Powell 1996; Borger et al. 2006). KDE was chosen for generating home ranges because it is a well accepted method in wildlife management (Laver and Kelly 2008) and does not tend to overestimate home range area (Powell 2000). The bandwidth for the KDE was calculated using least-squares cross-validation (LSCV). For each individual bear, LSCV values ranged from 500 m to 1000 m. Even within a home range we do not expect habitat to be used randomly. Therefore, we determined the frequency distribution of RSF values at locations where use was observed, as determined by bear locations. The randomized locations were determined based on the observed frequency distribution of RSF usage. For instance, if 15% of observed grizzly bear locations were found to have RSF values greater than eight, then 15% of the randomized bear locations were allocated to RSF values greater than eight. 2. Comparison of observed and expected distances to oil and gas features Each set of grizzly bear locations (individual bear, in a single year, in a single season) was labelled as either closer, neutral, or farther than expected from each type of oil and gas feature based on comparison of average observed and expected distances to oil and gas features (Îą = 0.05). For bears with multiple years of data, each bear-year was considered as an individual dataset. Once the statistical comparisons were made for each individual bear, results were summarized by season, sex, and age class. For example, the percentage of adult female bears with an average minimum distance statistically closer, neutral, or further than expected was calculated for each feature. 3. Day/night distance comparison In order to determine if oil and gas features were influencing temporal use of habitat, day and night KDE home ranges were constructed for each bear, for each season, for each year. For each season, each set of daytime grizzly bear locations (individual bear, in a single year) and each set of nighttime locations (individual bear, in a single year) was labelled as either closer, neutral, or farther than expected from each type of oil and gas feature (Îą = 0.05), based on comparison of average observed and expected distances to oil and gas features. Results were summarized by sex/age class. In addition, the mean observed distance to each of the disturbances was calculated for all of the bear locations by age/sex class for each season for day and night. For comparison, the mean expected distances (based on conditional randomization) were also calculated. The observed distances to each disturbance type were compared between day and night using a Mann-Whitney u-test (Îą = 0.05) to determine if there were significant differences between daytime and nighttime distances.

Chapter 2-Response to Oil and Gas Activities 4. Wellsite age comparison To investigate whether the selection of habitat containing wellsites was impacted by the age (years since initial drilling) of the wells, individual grizzly bear home ranges were used to identify the wellsites within the home range (available wellsites) for each bear. For this analysis, active and inactive wellsites were combined. Each wellsite within each grizzly bearâ&#x20AC;&#x2122;s home range was buffered by 100m. The wellsite was defined as selected if the density of observed telemetry points inside the buffer was significantly higher than the density of expected points from the randomization process. For each individual bear, a list of selected and not-selected wellsites was generated, and the mean wellsite ages were calculated. Given the variation in wellsite ages amongst home ranges, the normalized ages1 of the wellsites were calculated for each group of bears and seasons in order to determine statistical difference between selected and not-selected wellsites. The normalized ages were compared using a one-sided Mann-Whitney U-test to determine if selected wells were significantly older than not-selected wells. To typify those wells being selected, the mean ages of selected and not-selected wells were calculated by bear age/sex class and season.

Results For each age/sex class and season, distance results are described as: 1) closer, neutral, or farther than expected from oil and gas features (all points, based on comparison to expected points generated in the randomization) 2) closer, neutral, or farther than expected, points separated by day and night, and 3) numerical (mean) distances to features (in metres). During spring, habitat containing oil and gas features (pipeline, roads, active and inactive wellsites) was generally not avoided by grizzly bears. The majority of adult females were significantly closer than expected to all features, while most adult males were closer than expected to pipelines, roads, and inactive wellsites (Figure 3). All (100%) subadult males were closer than expected to pipelines, roads, and active wellsites, and most subadult females (80%) were closer than expected to both active and inactive wellsites. When the day/night use of habitat was examined separately, the majority of subadult males and females were closer than expected to all features at night (Table 3). In contrast, nighttime locations for adult males were not closer than expected to pipelines or wellsites, and only 43% of bears were closer to roads at night (Table 4). However, when comparing daytime to nighttime, mean distances at night were significantly smaller than the daytime distances for all disturbance features for adult males (Table 3). In other words, although adult males were not closer than expected to disturbance features during night, they were closer during nighttime than during the day.

Normalized wellsite age = (wellsite age) â&#x20AC;&#x201C; (mean of available wellsites) / (standard deviation of wellsite ages)

Chapter 2-Response to Oil and Gas Activities

Figure 3: Grizzly bear habitat selection in the spring foraging season for all age classes (F-adult females, M- adult males, SF â&#x20AC;&#x201C; subadult females, SM-subadult males). Observed bear locations were compared statistically to expected locations and the percentage of bears either significantly farther or closer than expected at random tallied. Locations that were not significantly different from random were considered neutral.

Chapter 2-Response to Oil and Gas Activities

Table 3: The largest proportional response to oil and gas features by each bear class in each season, partitioned by day and night. The mean of the observed distances to each feature was calculated for individual bears and statistically compared to expected distances. No significant difference was considered neutral. Subadult Male Female Day Night Day Night Day Night Largest Mean Response (C=% Bears Closer; N=Neutral; F=Farther) 44% C 63% C 80% N 80% C 42% C 57% N 44% C 50% N 60% N 60% C 50% N 43% C&N 50% C 63% C 60% N 60% C 50% N 57% N 50% C 56% C 60% C 60% C 58% N 43% F Female Spring Pipeline Roads Active Wells Inactive Wells Summer Pipeline Roads Active Wells Inactive Wells Fall Pipeline Roads Active Wells Inactive Wells

Subadult Male Day

Night

40% F&N 80% C 60% N 80% C

67% C 100% C 67% C 67% C

42% C 53% C 42% C 42% C

58% C 58% C 47% N 47% N

50% C 67% C 50% F 50% F

60% C 80% C 40% N 60% C

62% F 54% F 62% F 46% F

75% F 75% F 75% F 50% F

78% C 78% C 78% C 56% C

86% C 71% C 57% C 43% C&N

53% F 47% F 58% F 42% F

56% C 56% C 44% C 50% C

80% F 60% C 60% F 80% F

60% F 60% C 60% F 60% N

50% C 50% C 50% C 50% C&F

56% C 78% C 56% N 44% C

71% C 57% C 57% F 86% C

57% C 78% C 43% F&N 71% C

Chapter 2-Response to Oil and Gas Activities

Table 4: Mean observed and expected distances to the oil and gas features during day and night. * Indicates significant difference between observed day and night mean distances at Îą = 0.05 using a Mann-Whitney U-test.

Female Day Night Pipeline

Subadult Female Day Night

Spring Observe Expected d Observe Expected d Observe Expected d Observe Expected d Summer Observe Expected d Observe Expected d Observe Expected d Observe Expected d Fall Observe Expected d Observe Expected d Observe Expected d Observe Expected d

Male Day

Night

Subadult Male Day Night

985m 999m 1431m 1418m 1432m *959m 2008m 1904m 1042 1056m 1596m 1626m 1444m 1019m 1983m 2029m Roads 522m 518m 587m 585m 907m *695m 689m 668m m 594m 586m 664m 682m 885m 719m 787m 802m Active wells 1527 1496m 2273m 2314m 1986m *1379m 2253m 2170m 1640 1652m 2538m 2608m 1979m 1505m 2335m 2398m m Inactive wells 1500 1515m 2090m 2117m 1984m *1682m 2041m 2095m m 1647 1699m 2278m 2338m 2098m 1715m 2187m 2256m m m Pipeline 1292 *1231m 1507m 1613m 2507m *2236m 1759m *1480m 1244 1257m 1697m 1888m 1902m 1566m 1798m 1755m m Roads 599m *518m 488m 510m 1889m *1711m 907m *769m m 629m 650m 708m 758m 1372m 1173m 860m 803m Active wells 2009 *1932m 2672m 2662m 3517m *3225m 2557m *2426m 2004 1949m 2916m 2967m 2407m 2392m 2338m 2318m m Inactive wells 1525 *1472m 2260m *2161m 3351m *2968m 2096m *2063m m 1571 1610m 2285m 2452m 2782m 2392m 2065m 2041m m m Pipeline 1352 *1305m 2176m *2031m 2424m *2113m 3111m #*2951 1270 1260m 1909m 1734m 2360m 1974m 1543m 1490m m m Roads 655m *602m 616m 635m 1195m *1018m 935m *881m m 682m 685m 699m 722m 1392m 1239m 1132m 1008m Active wells 2239 *2053m 3126m *2795m 3346m *3069m 3280m 3263m 2028 1942m 2549m 2231m 2951m 2538m 2941m 2773m m Inactive wells 1605 *1539m 2475m *2244m 2476m 2247m 2150m *2074m m 1616 1612m 2282m 2110m 2809m 2517m 2548m 2443m m # One subadult male bear (G229) was observed on m average 4 km from pipelines, day and night, over two separate years leading to the skewness of the data.

Chapter 2-Response to Oil and Gas Activities In summer, fewer bears were closer than expected to the oil and gas features than in spring (Figure 4). This trend was most pronounced in the adult males; the majority were farther than expected to all disturbance features. About half of adult females were closer than expected to all disturbances (42-53%), marking a decrease from the previous season. The majority of subadult males (67-78%) were still closer than expected to all features, and 83% of subadult females were closer to roads. When data were analyzed separately for day/night, no distinct differences in response were observed (Table 3). For example, 53% of adult females were closer than expected to roads during the day, and this only increased to 58% of bears at night. When comparing mean observed distances during daytime to distances at night, mean distances to all features were significantly smaller at night compared to day for adult females, adult males, and subadult males (Table 4). For subadult females, the mean observed distance to inactive wells was also significantly smaller at night compared to day (Table 4).

Figure 4: Grizzly bear habitat selection in the summer foraging season for all age classes (F-adult females, M- adult males, SF â&#x20AC;&#x201C; subadult females, SM-subadult males). Observed bear locations were compared statistically to expected locations and the percentage of bears either significantly farther or closer than expected at random tallied. Locations that were not significantly different from random were considered neutral.

In the fall, the most striking result is the consistency with which all bear classes are closer than expected to roads, and farther than expected to active wells (Figure 5). Seventy eight percent of adult males were closer to roads than expected in the fall, a dramatic increase from 8% in summer. When daytime and nighttime use of habitat use were separated, this response was relatively unchanged, as the majority of adult males, subadult males, and 22

Chapter 2-Response to Oil and Gas Activities subadult females were closer than expected to roads during both day and night (Table 3). However, day/night differences were observed for adult females; the majority of females were farther than expected from all features during the day, and closer than expected at night (Table 3). In addition, the mean observed distance to all features was significantly smaller during night compared to day for adult females (Table 4). In the case of subadult males and pipelines, while the mean distances to pipeline was almost 3 km (Table 4), the majority of subadult males were closer than expected to pipelines (Table 3). This was likely due to one bear (G229) with an average distance of 4 km from pipelines during two different fall seasons.

Figure 5: Grizzly bear habitat selection in the fall foraging season for all age classes (F-adult females, M- adult males, SF â&#x20AC;&#x201C; subadult females, SM-subadult males). Observed bear locations were compared statistically to expected locations and the percentage of bears either significantly farther or closer than expected at random tallied. Locations that were not significantly different from random were considered neutral.

Plotting the mean observed distances for each of the age/sex classes allows us to visualize trends in relative proximity for each sex/age class of bears to each disturbance feature (Figure 6). For all groups, the smallest distances to oil and gas features were for roads, over all seasons, during both day and night. Active wells had the largest distances for all groups in each season (Figure 6). Adult females generally had the smallest distances to all features across all seasons and times of day, although subadult females had slightly smaller distances to roads during the day in summer and fall (Figure 6; Table 4). As the seasons progressed, there was an increase in distances to active wells for adult females, from around 1.5 km in the spring to 2.0 km in summer and 2.2 km in fall. The largest change in 23

Chapter 2-Response to Oil and Gas Activities distance to wellsites from spring to summer was observed with adult males for daytime locations; mean distances increased from 2 km to 3.5 km for active wells and from 2 km to 3.3 km for inactive wells. 4000

A )

3500

Distance (m )

3000 2500

Pipelines

2000

Road

1500

Active Wells Inactive Wells

1000 500 0 Ad Fem Day

Ad Male Ad Male Sub Day Night Fem Day

Sub Fem Night

Sub Male Day

Sub Male Night

4000

B )

Ad Fem Night

3500

Distance (m)

3000

Pipelines

2500

Roads

2000

Active Wells

1500

Inactive Wells

1000 500 0 Ad Fem Day

Ad Fem Night

Ad Male Ad Male Sub Day Night Fem Day

Sub Fem Night

Sub Male Day

Sub Male Night

4000

3000

Distance (m )

3500

2500

Pipelines

2000

Roads

1500

Active Wells

1000

Inactive Wells

500 0 Ad Fem Day

Ad Fem Night

Ad Male Ad Male Sub Day Night Fem Day

Sub Fem Night

Sub Male Day

Sub Male Night

Figure 6. The observed mean distances for all bear groups to the various oil and gas features during A) Spring; B) Summer; C) Fall.

Results from the comparison between the age of selected versus not-selected wellsites are presented in Table 5. For all classes, the normalized age of the selected wellsites was significantly higher than the age of the not-selected wellsites, indicating that bears are generally selecting for older wellsites out of those that are available to them within their home range. Seasonally, the oldest wellsites were selected for in the summer, with the average age of selected wellsites ranging from 12.4 years for subadult males, to 13.6 years for those selected by adult females.

Chapter 2-Response to Oil and Gas Activities Table 5: Comparison between the mean age of selected and not-selected wellsites. For each individual bear, selected wellsites were determined by comparing the number of observed telemetry points within 100m of a well and the number of randomized points within that same buffer. Not-selected wellsites were those remaining within the home range. Results were tabulated by age/sex class. Significance(pâ&#x2030;¤0.05 *) was determined using the normalized wellsite age, which was calculated for each bear by subtracting the mean wellsite age from the individual values and dividing by the standard deviation. Mean Age of Mean Age of Mean Age of Bear Class Selected Wells Not-selected All Wells (Years) Wells (Years) (Years) Spring Adult Female 11.3 10.4 10.5 Adult Male 11.8 10.2 10.3 Subadult Female 13.6 18.0 17.7 Subadult Male *12.2 8.5 8.6 Summer Adult Female *13.6 10.2 10.7 Adult Male *13.2 9.9 10.0 Subadult Female 13.2 13.3 13.3 Subadult Male *12.4 8.4 8.5 Fall Adult Female *12.7 10.7 11.0 Adult Male 11.5 11.8 11.7 Subadult Female 12.4 14.0 13.8 Subadult Male *13.1 9.4 9.5

Discussion Selection of habitat can be thought of as a trade-off between acquiring resources, finding mates and minimizing mortality risk (Frid & Dill 2002). The requirement for vast amounts of habitat in an area of increasing resource extraction pressures may put grizzly bear populations at greater risk. While human disturbance increases human/wildlife interactions and leads to greater mortality risk, changes to the landscape from resource extraction industries can also create new foraging opportunities for grizzly bears (Nielsen et al. 2004b; Stewart et al. 2012). Given increasing oil and gas activities in Alberta grizzly bear habitat, it was our objective to examine the spatial-temporal effects that existing oil and gas activities might have on grizzly bear habitat use. We used the locations of features associated with the oil and gas industry as a surrogate for human presence and influence in grizzly bear habitat. The expected distances used for comparison were based on RSF values in order to rule out the effects of habitat quality and allow clear interpretation of results as a response to the oil and gas features on the landscape. In the spring, there was little avoidance of habitat containing oil and gas development and activities by all groups of bears. Bears also had the shortest mean observed distances to all 25

Chapter 2-Response to Oil and Gas Activities features in spring. Drilling of wellsites in the Kakwa region occurs throughout the year, however, the highest frequency of drilling takes place in the winter, with moderate levels of drilling in summer and fall (Appendix 2). Wells are infrequently drilled during spring, as the saturated ground is too soft for heavy equipment. Drilling represents a period of intense human activity on the landscape, and its absence during spring could explain why bears are closer to oil and gas features during this time compared to other seasons. Daytime and nighttime are assumed to correspond with times of human activity and inactivity, respectively. Spring was observed to have the fewest day/night differences, supporting the notion that human activity has the least effect on bear habitat selection during this season. Adult males were the sex/age class with the largest distances to roads in each foraging season, similar to the results of McLellan & Shackleton (1988), Chruszcz et al. (2003) and Graham et al. (2010). In contrast, an opposite result was found by Gibeau et al. (2002), who observed females farthest from roads, while males were the closest group at night. The observed summer distances for adult males in our study were >500m larger than expected distances, indicating an avoidance of roads by adult males in summer. Adult males were also farther than expected to the other oil and gas features, suggesting a general avoidance of all anthropogenic features during the summer. It is not clear why adult males are often observed farther away from roads and other anthropogenic features than bears of other sex/age classes, and this phenomenon has been identified as an area that requires further investigation (Graham et al. 2010). However, it has been suggested that due to the sexual dimorphism seen in grizzly bears, some of the differences in habitat use patterns observed in adult males could be due to their different nutritional requirements (Rode et al. 2006). Understanding impacts that the oil and gas industry may have on adult females is paramount for effective conservation planning (Gibeau et al. 2002; Mueller et al. 2004; Nielsen et al. 2006). In our study, adult females were most consistent in their response to the various oil and gas features across the seasons. Adult females were observed closest to all oil and gas features as compared to other sex/age classes. However, during fall they were farther than expected from active wells and, to a lesser extent, pipelines, suggesting a spatial avoidance of oil and gas development during this period. When data were partitioned by day and night, adult females were also found farther than expected during the day and closer than expected at night for all features, suggesting a temporal avoidance of human activity. Gibeau et al. (2002) suggested that adult females were the most riskaverse, and, consistent with our findings, found that they were closer to human features during the â&#x20AC;&#x153;human inactiveâ&#x20AC;? period. In summer, subadult females were observed to be the closest to roads out of all sex/age classes during both day and night, consistent with results reported in Mueller et al. (2004). It is also worth noting that there were no significant day/night differences in distances to roads for subadult females. Similarly, Mueller et al. (2004) observed that subadult females had the least difference in day/night use of roads. Similarly, subadult males were consistently closer than expected to almost all disturbances across seasons. Subadult males have been shown to use habitat close to human disturbances such as large highways, 26

Chapter 2-Response to Oil and Gas Activities especially during night (Gibeau et al. 2002). The subadult results in our day/night analysis need to be interpreted with caution, due to the low number of individuals, and our findings on this question should be considered preliminary. Disturbance by oil and gas industries includes direct alteration of the landscape through clearing the forest to create roads, pipelines, or well pads. Such clearing of the forest increases sunlight to the area, creating locations where bear foods can grow (Nielsen et al. 2004c). However, oil and gas disturbance also involves human presence, including high levels of noise and activity during drilling, and periodic human activities during operations. To differentiate between the effects of habitat change versus the impacts of human presence on the landscape, we examined active wellsites separately from inactive wellsites. Differences between the two categories of wellsites were most pronounced in the fall season, when more bears were closer than expected to inactive wellsites compared to active wells. These results cannot simply be explained by the timing of drilling operations (Appendix 2), as the frequency of drilling during summer and fall is very similar. Maintenance visits to the active wellsites are also not likely to be different between the two seasons, while human presence at inactive wellsites is rare. Data collected during field sampling at wellsites within our study area indicated that a number of berry species are abundant along the edges of wellsites in the Kakwa area (Foothills Research Institute [FRI], unpublished data). A possible explanation for the preference for habitat containing inactive wellsites in fall is that the inactive wells provide bears with a predictable source of berries in locations with relatively low levels of human activity. The alteration of the landscape in this case could provide the bears with a foraging opportunity without the associated human activity. During summer and fall, the mean age of wellsites that were selected by bears was significantly older than for wellsites that were not selected, and the oldest wellsites were selected during summer. Field data indicate that the proportion of a wellsite covered by dandelion (Taraxacum officinale), clover (Trifolium spp), and horsetails (Equisetum spp) increases after initial clearing of the wellpad, then begins to level off after approximately 10 years (FRI, unpublished data). These foods form a major part of a grizzly bear’s diet in the summer (Nielsen et al. 2004c; Munro et al. 2006). The mean age of selected wellsites in our study was greater than 10 years (from 11.3 to 13.6 years) across all age/sex classes and seasons, suggesting that bears are selecting for wellsites with an abundance of these important foods.

Literature cited Alberta Sustainable Resource Development (ASRD) 2010. Status of the grizzly bear (Ursus arctos) in Alberta: Update 2010. Wildlife Status Report No. 37. Edmonton, AB, 44p. Benn, B., and Herrero, S. 2002. Grizzly bear mortality and human access in Banff and Yoho National Parks 1971–1988. Ursus 13: 213–221. Blanchard, B.M, and Knight, R.R. 1991. Movement of Yellowstone grizzly bears. Biological Conservation 58: 41–67. 27

Chapter 2-Response to Oil and Gas Activities Borger, L., Franconi, N., Midhele, G., Gantz, A., Meschi, F., Manica, A., Lovari, S., and Coulson, T. 2006. Effects of sampling regime on the mean and variance of home range size estimates. Journal of Animal Ecology 75: 1393–1405. Bowyer, R.T., and Bleich, V.C. 1984. Effects of cattle grazing on selected habitats of southern mule deer. California Fish and Game 70: 240–247. Boyce, M.S., Vernier, P.R., Nielsen, S.E., and Schmiegelow, F.K.A.. 2002. Evaluating resource selection functions. Ecological Modelling 157: 281–300. Boydston, E.E., Kapheim, K.M., Watts, H.E., Sykman, M., and Holekamp, K.E. 2003. Altered behaviour in spotted hyenas associated with increased human activity. Animal Conservation 6: 207–219. Carroll, C., Noss, R.F., and Paquet, P.C. 2001. Carnivores as focal species for conservation planning in the Rocky Mountain Region. Ecological Applications. 11: 961–980. Chruszcz, B., Clevenger, A.P., Gunson, K.E., and Gibeau, M.L. 2003. Relationships among grizzly bears, highways, and habitat in the Banff-Bow Valley, Alberta, Canada. Canadian Journal of Zoology 81: 1378–1391. Ciarniello, L.M., Boyce, M.S., Heard, D.C., and Seip, D.R. 2007. Components of grizzly bear habitat selection: density, habitats, roads, and mortality risk. The Journal of Wildlife Management 71: 1446–1457. Clark, D.A., and Slocombe, D.S. 2011. Grizzly bear conservation effort in the Foothills Model Forest: appraisal of a collaborative ecosystem management effort. Policy Science 4: 1–11. Cressie, N.A.C. 1993. Statistics for Spatial Data. Wiley, Toronto. D'Eon, R., and Delparte, D. 2005. Effects of radio-collar position and orientation on GPS radio-collar performance, and the implications of PDOP in data screening. Journal of Applied Ecology 42:383–388. Dyer, S.J., O’Neill, J.P., Wasel, S.M., and Boutin, S. 2001. Avoidance of industrial development by woodland caribou. The Journal of Wildlife Management 65: 531– 542. Edge, W.D., Marcum, C.L. and Olson-Edge, S.L. 1987. Summer habitat selection by elk in western Montana: a multivariate approach. The Journal of Wildlife Management 51: 844–851. Elgmork, K., and Kaasa, J. 1992. Food habits and foraging of the brown bear Ursus arctos in central south Norway. Ecography 15(1) 101–110. Fortin, M.J., and Dale, M.R.T. 2005. Spatial Analysis: A Guide for Ecologists. Cambridge University Press, Cambridge. 365p. Frair, J.L., Nielsen, S.E., Merrill, E.H., Lele, S.R., Boyce, M.S., Munro, R.H.M., Stenhouse, G.B., and Beyer, H.L. 2004. Removing GPS collar bias in habitat selection studies. Journal of Applied Ecology, 41(2): 201–212. 28

Chapter 2-Response to Oil and Gas Activities Frid, A., and Dill, L. 2002. Human-caused disturbance stimuli as a form of predation risk. Conservation Ecology 6(1): art.11 (http://www.consecol.org/vol6/iss1/art11). Garshelis, D.L., Gibeau, M.L., and Herrero, S. 2005. Grizzly bear demographics in and around Banff National Park and Kananaskis Country, Alberta. J Wildl Manage 69(1): 277–297. Getis, A., and Boots, B., 1978. Point Pattern Analysis. Sage Publications Inc., Newbury Park, 93p. Graham, K., Boulanger, J., Duval, J., and Stenhouse, G. 2010. Spatial and temporal use of roads by grizzly bears in west-central Alberta. Ursus 21:43–56. Gibeau, M.L., Clevenger, A.P., Herrero, S., and Wierzchowski, J. 2002. Grizzly bear response to human development and activities in the Bow River Watershed, Alberta, Canada. Biological Conservation 103: 227–236. Haining, R., 2003. Spatial Data Analysis: Theory and Practice. Cambridge University Press, Cambridge, 432p. Hebblewhite, M., and Merrill, E. 2008. Modelling wildlife-human relationships for social species with mixed-effects resource selection models. Journal of Applied Ecology 45: 834–844. Laver, P.N., and Kelly, M.J. 2008. A critical review of home range studies. The Journal of Wildlife Management 72: 290–298. Linke, J.E., Franklin, S.E., Huettmann, F., and Stenhouse, G.B. 2005. Seismic cutlines, changing landscape metrics and grizzly bear landscape use in Alberta. Landscape Ecology 20: 811–826. Manly, B.F.J. 2002. Resource selection by animals: statistical design and analysis for field studies. Springer, London. Martin, J., Basille, M., Van Moorter, B., Kindberg, J., Allainé, D., and Swenson, J.E. 2010. Coping with human disturbance: spatial and temporal tactics of the brown bear (Ursus arctos). Canadian Journal of Zoology 88: 875–883. McLellan, B.N. 1990. Relationship between human industrial activity and grizzly bears. Bears: Their Biology and Management 8: 57–64. McLellan, B.N., and Shackleton, D.M. 1988. Grizzly bears and resource-extraction industries: The effects of roads on behaviour, habitat use and demography. The Journal of Applied Ecology. 25(2): 451–460. Mueller, C., Herrero, S., and Gibeau, M.L. 2004. Distribution of subadult grizzly bears in relation to human development in the Bow River Watershed, Alberta. Bears: Their Biology and Management 15:35-47. Munro, R.H.M., Nielsen, S.E., Stenhouse, G.B., and Boyce, M.S. 2006. Seasonal and diel patterns of grizzly bear diet and activity in west-central Alberta. Journal of Mammalogy 87: 1112–1121. 29

Chapter 2-Response to Oil and Gas Activities Natural Regions Committee 2006. Natural Regions and Subregions of Alberta. Compiled D.J. Downing and W.W. Pettapiece. Government of Alberta. Pub. No. T/852.

Nellemann, C., Støen, O.-G., Kindberg, J., Swenson, J. E., Vistnes, I., Ericsson, G., Katajisto, J., Kaltenborn, B. P., Martin, J., and Ordiz, A. 2007. Terrain use by an expanding brown bear population in relation to age, recreational resorts and human settlements. Biological Conservation 138(1-2): 157–165. Nelson, T.A., and Boots, B. 2005. Identifying insect infestation hot spots: an approach using conditional spatial randomization. Journal of Geographical Systems 7(3-4): 291–311 Nielsen, S.E. 2004. Habitat ecology, conservation and projected population viability of grizzly bears (Ursus arctos L.) in west-central Alberta, Canada. Dissertation. Department of Biological Studies, University of Alberta, Edmonton, Alberta, Canada. Nielsen, S.E. 2007. Seasonal grizzly bear habitat for six population units of Alberta, Canada. In: Stenhouse, G. and K. Graham (eds). Foothills Model Forest Grizzly Bear Research Program 2006 Annual Report. Hinton, Alberta. 87 pp. Nielsen, S.E., Herrero, S., Boyce, M.S., Mace, R.D., Benn, B., Gibeau, M.L., and Jevons, S. 2004a. Modelling the spatial distribution of human-caused grizzly bear mortalities in the Central Rockies Ecosystem of Canada. Biological Conservation 120:101–113. Nielsen, S.E., Boyce, M.S., and Stenhouse, G.B. 2004b. Grizzly bears and forestry I. Selection of clearcuts by grizzly bears in west-central Alberta, Canada. Forest Ecology and Management 199(1): 51–65. Nielsen, S.E., Munro, R.H.M., Bainbridge, E.L., Stenhouse, G.B., and Boyce, M.S. 2004c. Grizzly bears and forestry II. Distribution of grizzly bear foods in clearcuts in westcentral Alberta, Canada. Forest Ecology and Management 199(1): 67–82. Nielsen, S.E., Stenhouse, G.B., and Boyce, M.S. 2006. A habitat-based framework for grizzly bear conservation in Alberta. Biological Conservation 130:217–229. Nielsen, S.E., Cranston, J. and Stenhouse, G.B. 2009. Identification of priority areas for grizzly bear conservation and recovery in Alberta, Canada. Journal of Conservation Planning 5: 38–60. Nielsen, S.E., McDermid, G., Stenhouse, G.B., and Boyce, M.S. 2010. Dynamic wildlife habitat models: Seasonal foods and mortality risk predict occupancy-abundance and habitat selection in grizzly bears. Biological Conservation. 143: 1623–1634. Olson, T.L., Squibb, R.C., and Gilbert, B.K. 1998. Brown bear diurnal activity and human use: a comparison of two salmon streams. Ursus 10: 547–555. Powell, R.A. 2000. Animal home ranges and territories and home range estimators. Pages 65–110 in L. Boitani, andT. K. Fuller, editors. Research Techniques in Animal Ecology Controversies and Consequences. Columbia University Press,New York. Rode, K.D., Farley, S.D. and Robbins, C.T. 2006. Behavioral responses of brown bears mediate nutritional effects of experimentally introduced tourism. Biological Conservation 133: 70–80. 30

Chapter 2-Response to Oil and Gas Activities Roever, C.L., Boyce, M.S. and Stenhouse, G.B. 2008. Grizzly bears and forestry II: Grizzly bear habitat selection and conflicts with road placement. Forest Ecology and Management 256: 1262–1269. Ross, P.I. 2002. Update COSEWIC status report on the grizzly bear Ursus arctos in Canada, Committee on the Status of Endangered Wildlife in Canada. Ottawa. 99 p. Sahlen, E. 2010. Do grizzly bears use or avoid well-sites in west-central Alberta, Canada? Thesis. Swedish University of Agricultural Sciences, Umea, Sweden. Schneider, R.R. 2002. Alternative Futures: Alberta's Boreal Forest at the Crossroads. The Federation of Alberta Naturalists, Edmonton, Alberta. Schwartz, C.C., Miller, S.D., and Haroldson, M.A. 2003. Grizzly bear. In: Feldhamer, G.A., Thompson, B.C., and Chapman, J.A.(eds.).Wild Mammals of North America: Biology, Management, and Conservation. Second Ed., Johns Hopskins University Press, Baltimore, MD. Pp: 556–586. Ch. 26. Schwartz, C.C., Cain, S.L., Podruzny, S., Cherry, S., and Frattaroli, L. 2010. Contrasting activity patterns of sympatric and allopatric black and grizzly bears. Journal of Wildlife Management 74(8): 1628–1638. Seaman, D.E., and Powell, R.A. 1996. An evaluation of the accuracy of kernel density estimators for home range analysis. Ecology 77: 2075–2085. Stenhouse, G.B., and Munro, R.H.M. 2000. Foothills Model Forest Grizzly Bear Research Program: 1999 Annual Report. Stewart, B.P., Nelson, T.A., Wulder, M.A., Nielsen, S.E., and Stenhouse, G.B. (Accepted) Impact of disturbance characteristics and age on grizzly bear habitat selection. Applied Geography (Accepted Dec 2011, JAPG-D-11-00311). Weaver, J.L., Paquet, P.C., and Ruggiero, L.F. 1996. Resilience and conservation of large carnivores in the Rocky Mountains. Conservation Biology 10(4): 964–976. White, J.C., Wulder, M.A., Gomez, C., and Stenhouse, G. 2011. A history of habitat dynamics: Characterizing 35 years of stand replaceing disturbance. Canadian Journal of Remote Sensing. 37(2): 234-251.

Chapter 2-Response to Oil and Gas Activities

TOPIC B: GRIZZLY BEAR USE OR AVOIDANCE OF OIL AND GAS WELLSITES Tracy McKay, Ellinor Sahlén, Gordon Stenhouse, and Karen Graham

Introduction The Kakwa region in west-central Alberta currently has extensive oil and gas development and activities, with a current wellsite density of approximately 0.43 wells/km2. As additional oil and gas development opportunities emerge in this region, it is expected that the number of wellsites will increase in the future. Wellsites result in habitat alteration along with human presence and noise at the wellsite. Previous studies of grizzly bear response to industrial development has largely focused on roads and forestry operations, and research investigating bear response to oil and gas activities has mainly focused on impacts of exploration and development (Harding & Nagy, 1980; Schallenberger, 1980; Tietje & Ruff, 1983; Reynolds et al., 1986; McLellan & Shackleton, 1989; Follmann, 1990; Amstrup, 1993). To our knowledge, there are currently no published data regarding grizzly bear response to wellsite operations in Alberta. Information regarding habitat use around oil and gas wellsites could assist in grizzly bear management in these areas. Anecdotally, grizzly bears are known to visit wellpads, as indicated by sightings of bears at wellsites and observations of scat and tracks. Previous FRIGBP research in the Kakwa area suggested that some bears selected for areas near wellsites (Sahlén 2010, Labaree et al. 2012). Results from Topic A of this study (Labaree et al. 2012) indicated that bears were closer than expected to wellsites, but the overall response was captured at a large spatial scale; mean distances to wellsites were in the range of 1500m to 3000m. Sahlén’s (2010) analysis included habitat selection in 5 buffer zones within 500m of the wellsite, with the smallest buffer at 224m from the wellsite centre; the remaining home range was not included in the selection analysis. Results from these previous analyses provided important initial information regarding selection around wellsites, but the scale of the analyses could not indicate use or avoidance of the wellpad itself, or the relative use of these zones in the context of home ranges. The purpose of this investigation was to expand upon and refine the previous analysis of wellsite selection in this area (Sahlén 2010). Our analysis used an updated wellsite dataset with improved spatial and temporal accuracy, and an additional 6 grizzly bears in the location dataset. We focused specifically on the wellpad and surrounding zone, within the context of individual home ranges. Previous research has suggested that a zone of influence exists around anthropogenic disturbances; a zone where the effects of human activity extend out from the disturbance feature (Gibeau et al. 1996, Forman 2000). Other grizzly bear studies have applied a 500m “zone of influence” (ZOI) around human developments (Mace et al. 1996, Berland 2008, Sahlén 2010), and a 500m zone was investigated in this study. Based on the assumption that some bears may select for wellpads, we also set out to investigate whether selection could be based on availability of 32

Chapter 2-Response to Oil and Gas Activities bear foods at wellsites, including the presence/abundance of bear foods on the wellpads or along the edges. Therefore, we framed our research to answer the following questions: a. Do grizzly bears show selection (use) or avoidance of the wellpad itself and/or selection or avoidance of the zone immediately surrounding the wellpad? b. Are grizzly bear foods available on wellpads? To meet these objectives, we analyzed selection ratios in three habitat zones: 1) the wellpad, 2) the surrounding zone, and 3) the remaining home range. In addition, we visited wellsites to assess vegetation cover and approximate bear food abundance.

Methods: Bear selection (use) or avoidance of the wellpad and ZOI Telemetry data were collected for 37 grizzly bears in the Kakwa region during 2005–2011. Aerial darting, leg-hold snaring, and culvert traps were used to capture grizzly bears following capture protocols accepted by the Canadian Council of Animal Care for the safe handling of bears (animal use protocol number 20010016) (Stenhouse & Munro 2000). Captured bears were fitted with a Televilt/Followit brand GPS collar (Lindesburg, Sweden) programmed to collect hourly locations. Location points were limited to those with a dilution of precision (DOP) ≤6 (Ganksopp & Johnson, 2007), as a high level of spatial accuracy was required for the small scale of this analysis. Only non-denning points that fell within calculated home ranges (95% Kernel) were included in the analysis. We restricted our dataset to bears with ≥ 90% of their annual home range area within the study area boundary. At the small spatial scale of this analysis, there were not enough data to separate results by season. However, seasonality could affect habitat selection patterns, due to the seasonal availability of bear foods. Nielsen (2004) determined hypophagia (spring), early hyperphagia (summer), and late hyperphagia (fall) foraging seasons for grizzly bears in west-central Alberta. To prevent data from being skewed when data were missing for an entire season , we only used data for those bears with GPS collar locations across all three foraging seasons (hypophagia, 1 May 1st to June 15th; early hyperphagia, June 16th to July 31st; and late hyperphagia, August 1st to October 15th). Data were also investigated to determine if selection ratio results were correlated with the number of location points. Using a Geographic Information System (GIS), a 100m radius buffer was generated around the centre of each wellsite (point data) to create an area that incorporated the cleared wellpad along with the edge habitat immediately adjacent to the wellpad, but essentially limited to the wellsite (Figure 7). For the “zone of influence”, we generated a 500m buffer around wellsites. The area of the 100m buffer was erased from the 500m buffer to create distinct (separate) buffer zones. 33

Chapter 2-Response to Oil and Gas Activities

Figure 7: Wellsite with 100m and 500m buffer zones.

Annual home ranges for each bear were calculated based on 95% kernels using the program ABODE (Laver 2005) in ArcGis. Fixed biweight kernels were calculated using a volume contouring method. We used a least squares smoothing factor (Silverman 1986) and a grid cell of 300m. Data were standardized using the unit variance. We only included grizzly bears with location data for the entire non-denning period to ensure the home range estimate was representative of the full active period (Arthur & Schwartz 1999, Belant and Follmann 2002, Girard 2002). We used the seasons from Nielsen (2004) and the following number of locations per season based on Belant and Follmann (2002) to ensure annual home range estimates were based on locations that spanned all three foraging seasons ; spring: > 30 locations; summer: > 49 locations and fall: > 49 locations. The 100m and 500m buffer zones were subtracted from the total annual home range area for each bear to obtain a third zone incorporating all home range area >500m from wellsites. Habitat selection was investigated for the three resulting habitat zones: 1. Wellsite area and edge habitat (100m buffer zone) 2. 500m zone of influence (ZOI) 3. Home range zone >500m from wellsites Grizzly bear collar locations from 2005 to 2011 were intersected with these three zones, and the number of points within each area was calculated for each bear for each year (bear-year). Locations that intersected with wellsite buffers (both 100m and 500m) were carried forward to the selection analysis if the bear location date was later than the initial drilling date for the corresponding well, to ensure that the wellsite was present at the time of the bear location.

Chapter 2-Response to Oil and Gas Activities The total areas (square kilometres) for each habitat zone for each bear-year are shown in Table 6. As wellsites are drilled throughout the year, the number of wellsites within each bearâ&#x20AC;&#x2122;s home range changed; therefore, the total area of each habitat zone changed from spring to fall. For our analysis, it was not feasible to calculate a weekly or monthly area for each habitat zone; analysis was completed by year. However, to provide the best annual estimates, we examined the dates of wellsite construction over the 7 years of our study, and determined that approximately half of the wellsites were constructed before July 15th each year. Therefore, this date was chosen as the cutoff for area calculations to approximate the mean area of wellsite buffers for each bear-year. We used selection ratios to determine whether grizzly bears were using each habitat zone more, less, or no differently than expected. Selection ratios compare observed habitat use to expected habitat use based on the relative availability of each habitat zone. We calculated selection ratios for each zone as in Manly et al. (2002), where selection ratio = the proportion of the sample of used resource units that are in each category the proportion of available resource units that are in that category.

For our analysis, the proportion used (observed) was calculated from the number of locations that intersected each zone divided by the total number of locations within the home range (Table 7). The proportion of available resource units was based on the proportional area of each zone, calculated from the area of each zone divided by the total area of the home range. For individual male and female bears with multiple years of collar locations (i.e. >1 bearyear), data were grouped by bear for calculation of selection ratios. However, due to the fact that reproductive status was different each year for each female bear, data were analyzed separately by bear-year for comparison of differences in selection ratios between females with or without young. To determine if patterns of habitat selection were significantly different than expected, the chi-squared goodness of fit test was calculated for each set of selection ratios. The chisquared test evaluates the null hypothesis that animals are randomly selecting habitat in proportion to availability (Manly et al. 2002). In order for the chi-squared test to be valid, the dataset must meet the assumption that expected and observed frequencies are not less than five (Manly et al. 2002). In this case, results from datasets with location counts (expected or observed) of less than five in any of the habitat zones were interpreted as non-significant.

Chapter 2-Response to Oil and Gas Activities Table 6: Areas of home ranges and habitat zones.

Bear

Total home range area (km2)

Total area of 100m buffers (km2)

Proportional area 100m buffer zone

Total area of 500m ZOIs (km2)

Proportional Total area Proportional area zone >500m from area 500m ZOI >500m from wellsites (km2) wellsites

G223 G224 G230 G236 G238 G253 G254 G257 G258 G260 G262 G264 G265 G269 G270 G271

607.39 629.46 656.42 527.93 886.47 267.31 244.65 4149.43 423.70 1275.35 3257.35 586.04 186.29 363.21 1840.05 319.38

6.52 3.39 1.24 8.38 6.74 1.12 0.97 33.42 5.47 15.69 41.44 2.15 0.75 1.43 22.31 3.00

0.01 0.01 0.00 0.02 0.01 0.00 0.00 0.01 0.01 0.01 0.01 0.00 0.00 0.00 0.01 0.01

137.10 76.23 27.75 153.27 142.82 24.94 19.87 700.91 115.96 310.78 840.62 48.47 16.21 29.77 433.04 61.05

0.23 0.12 0.04 0.29 0.16 0.09 0.08 0.17 0.27 0.24 0.26 0.08 0.09 0.08 0.24 0.19

463.77 549.83 627.43 366.27 736.91 241.25 223.81 3415.10 302.28 948.88 2375.29 535.42 169.33 332.01 1384.70 255.33

0.76 0.87 0.96 0.69 0.83 0.90 0.91 0.82 0.71 0.74 0.73 0.91 0.91 0.91 0.75 0.80

Chapter 2-Response to Oil and Gas Activities If the chi-squared test for the set of ratios is significant, the calculation of confidence intervals for each selection ratio indicates further whether or not the selection ratio for each habitat zone is significantly different from expected selection. We calculated 95% confidence intervals based on the standard error of the selection ratio using the following formula (Manly et al. 2002): , where oi = the proportion of the sample of used resource units that are in each category, u+ = the total number of resource units sampled, and πi = the proportion of available resource units that are in each category.

In our analysis, if the chi-squared test for the set of selection ratios was significant (p<0.05) and the confidence interval for the selection ratio did not include 1, the selection ratio was considered significantly different from expected selection. An additional approximate test for significance of the selection ratio estimate1 was calculated and compared to the chisquared distribution for one degree of freedom, as in Manly et al. (2002); these statistics were used as an additional test for significance of the selection ratios. We compared selection ratios between males and females, females with/without young, and daytime/nighttime. Females were considered to have young if they were accompanied with cubs of the year or yearlings. If reproductive status could not be confirmed, females were classified as without young. Day and night time locations were defined based on sunrise and sunset calculators (pers. comm. Julie Duval, FRI), and data falling into crepuscular periods (twilight morning and twilight evening), were included with night locations. Selection ratio results were tested for normal distribution using the Kolmogorov-Smirnov and Shapiro-Wilk tests in SPSS, as well as visual inspection of histograms. For datasets that were not normally distributed, non-parametric statistical tests were used for analysis. Selection ratios were compared between males and females (by bear) and between females with and without young (by bear-year) using the MannWhitney U test (α=0.05) in SPSS. Day/night data for 100m buffer locations (wellpad) were compared using a paired Wilcoxon signed-rank test (α=0.05), and day/night data for ZOI and the home range>500m from wellsites were compared using a paired t-test.

Approximate significance (p) = {(wi - 1) / se(wi)}² 37

Chapter 2-Response to Oil and Gas Activities

Availability of bear foods at wellsites Bear food availability estimates at wellsites in the Kakwa study area were collected in August 2011. Study sites were selected to include both categories of wellsite status (active versus inactive) and a range of wellsite ages (based on years since initial drilling). We collected general data regarding the overall characteristics of the wellsite, the percentage of the wellpad covered with vegetation, percent barren ground, and percent water. The approximate abundance of bear foods on the wellpad and along the four wellsite edges were estimated using the following cover classes:      

+ (present, <1%) 1 (<5%) 2 (5-25%) 3 (26-50%) 4 (51-75%) 5 (76-100%)

Any bear activity (scat, digging, anting, tracks) detected on the wellpad or along wellsite edges was also recorded.

Results Bear selection (use) or avoidance of the wellpad and ZOI Our final analysis included location datasets for 16 individual bears across seven years of data collection (2005 through 2011), and a total of 30 bear-years. The sample included 10 adult females (19 bear-years), 4 adult males (7 bear-years), and 2 subadults (4 bear-years). Selection ratio results were not correlated with the number of location points in each dataset, indicating that results for our final sample of bears were not biased by the variation in sample size for locations. Total number of locations and locations in each zone for each bear are presented in Table 7.

Chapter 2-Response to Oil and Gas Activities Table 7: Total locations within home ranges and locations within habitat zones. Expected numbers of locations were calculated based on total locations for each bear and the proportional area for each zone (see Table 6). Total Years of Bear locations data G223 G224 G230 G236 G238 G253 G254 G257 G258 G260 G262 G264 G265 G269 G270 G271

6191 3146 1267 2097 3673 1943 3340 6979 6440 15810 3251 949 1954 2607 778 1620

2 2 1 1 2 1 2 3 2 5 3 1 1 2 1 1

Observed Observed Expected Proportion Observed Expected Proportion 100m locations 100m 100m 500m ZOI 500m ZOI 500m buffer >500m from locations locations locations locations locations locations wellsites 72 77 9 14 31 20 98 107 268 594 52 10 22 106 20 63

66 17 2 33 28 8 13 56 83 194 41 3 8 10 9 15

0.0116 0.0245 0.0071 0.0067 0.0084 0.0103 0.0293 0.0153 0.0416 0.0376 0.0160 0.0105 0.0113 0.0407 0.0257 0.0389

1340 765 69 507 503 223 364 1268 1770 4971 628 118 399 365 273 361

1397 381 54 609 592 181 271 1179 1763 3853 839 78 170 214 183 310

0.2164 0.2432 0.0545 0.2418 0.1369 0.1148 0.1090 0.1817 0.2748 0.3144 0.1932 0.1243 0.2042 0.1400 0.3509 0.2228

4779 2304 1189 1576 3139 1700 2878 5604 4402 10245 2571 821 1533 2136 485 1196

Expected locations >500m from wellsites

Proportion locations >500m from wellsites

4727 2748 1211 1455 3053 1754 3056 5744 4594 11763 2371 867 1776 2383 585 1295

0.7719 0.7324 0.9384 0.7515 0.8546 0.8749 0.8617 0.8030 0.6835 0.6480 0.7908 0.8651 0.7845 0.8193 0.6234 0.7383

Selection ratios by bear for each habitat zone are presented in Table 8. In the 100m buffer zone, ten of sixteen bears had ratios significantly greater than 1.00, indicating selection of the wellpad region (Table 8). One bear had a selection ratio significantly less than 1.00, indicating avoidance of wellpads, while selection ratios for 3 bears were not significantly different from expected. Two bears (G230 and G264) had counts of less than five for expected locations in the 100m zone (Table 7); therefore, selection ratio results for these bears fail to meet the assumptions of the chi-squared goodness of fit test. Confidence intervals and significance estimates strongly suggest selection of the 100m zone in both cases; however, a conservative interpretation of results classifies the response of these two bears as not significant. The mean selection ratio for the 100m zone was 3.28 (SD = 2.52, range = 0.42-10.31) (Table 8).

Chapter 2-Response to Oil and Gas Activities Table 8: Selection ratios by individual bear. (NS = no significant relationship detected). 100m buffer zone Chi-squared Selection Bear probability Ratio SE

500m zone of influence

95% CI Pattern of Significance Selection Lower Upper habitat use estimate Ratio SE

Home range >500m from wellsites

Pattern of 95% CI habitat Significance Selection Lower Upper use estimate Ratio SE

Pattern of 95% CI habitat Significance Lower Upper use estimate

G223

0.1838

1.08 0.13

0.83

1.33 NS

0.5122

0.96 0.02

0.91

1.00 NS

0.0764

1.01 0.007

1.00

1.02 NS

0.1160

G224

0.0000

4.54 0.51

3.54

5.54 Selection

0.0000

2.01 0.06

1.88

2.13 Selection

0.0000

0.84 0.009

0.82

0.86 Less

0.0000

G230

0.0000

3.76 1.25

1.31

6.21 NS*

0.0270

1.29 0.15

0.99

1.58 NS*

0.0559

0.98 0.007

0.97

1.00 NS*

0.0100

G236

0.0000

0.42 0.11

0.20

0.64 Avoidance 0.0000

0.83 0.03

0.77

0.90 Avoidance 0.0000

1.08 0.014

1.06

1.11 More

0.0000

G238

0.0003

1.11 0.20

0.72

1.50 NS

0.5780

0.85 0.04

0.78

0.92 Avoidance 0.0000

1.03 0.007

1.01

1.04 More

0.0001

G253

0.0000

2.46 0.55

1.39

3.53 Selection

0.0077

1.23 0.08

1.08

1.38 Selection

0.0030

0.97 0.008

0.95

0.99 Less

0.0002

G254

0.0000

7.41 0.74

5.97

8.86 Selection

0.0000

1.34 0.07

1.21

1.47 Selection

0.0000

0.94 0.007

0.93

0.95 Less

0.0000

G257

0.0000

1.90 0.18

1.55

2.26 Selection

0.0000

1.08 0.03

1.02

1.13 Selection

0.0057

0.98 0.006

0.96

0.99 Less

0.0000

G258

0.0000

3.23 0.19

2.85

3.60 Selection

0.0000

1.00 0.02

0.96

1.04 NS

0.8344

0.96 0.008

0.94

0.97 Less

0.0000

G260

0.0000

3.05 0.12

2.81

3.29 Selection

0.0000

1.29 0.02

1.26

1.32 Selection

0.0000

0.87 0.005

0.86

0.88 Less

0.0000

G262

0.0000

1.26 0.17

0.92

1.60 NS

0.1367

0.75 0.03

0.70

0.80 Avoidance 0.0000

1.08 0.010

1.07

1.10 More

0.0000

G264

0.0000

2.88 0.90

1.10

4.65 NS*

0.0381

1.50 0.13

1.25

1.76 NS*

0.0001

0.95 0.012

0.92

0.97 NS*

0.0000

G265

0.0000

2.81 0.59

1.64

3.97 Selection

0.0024

2.35 0.10

2.14

2.55 Selection

0.0000

0.86 0.010

0.84

0.88 Less

0.0000

G269

0.0000

10.31 0.98

8.39 12.24 Selection

0.0000

1.71 0.08

1.55

1.87 Selection

0.0000

0.90 0.008

0.88

0.91 Less

0.0000

G270

0.0000

2.12 0.47

1.20

3.04 Selection

0.0167

1.49 0.07

1.35

1.63 Selection

0.0000

0.83 0.023

0.78

0.87 Less

0.0000

G271

0.0000 4.14 0.51 Mean 3.28 SD 2.52

3.14

5.15 Selection

0.0000 1.17 0.05 Mean 1.30 SD 0.44

1.06

1.27 Selection

0.0022 0.92 0.014 Mean 0.95 SD 0.08

0.90

0.95 Less

0.0000

*These datasets had <5 expected locations and failed to meet the assumptions of the chi-squared goodness of fit test.

Chapter 2-Response to Oil and Gas Activities Within the 500m ZOI, selection was significantly greater than expected for nine bears, significantly less than expected for three bears, and not significant for 4 bears. Overall, selection was not as strong for the 500m zone compared with the 100m zone, with a mean selection ratio of 1.30 (SD=0.44, range = 0.75-2.35). Results were consistent across the two zones, and no contradictory results were observed; in other words, there were no results that indicated selection of the 100m zone with avoidance of the surrounding 500m zone. For nine of the ten bears that selected for the 100m zone, results also indicated selection for the 500m ZOI, while the remaining bear had neutral results in the 500m zone. Similarly, the bear that showed avoidance of the 100m zone also avoided the 500m zone (Table 8). There were no cases of selection of the 500m zone without selection of the 100m zone. The habitat zone >500m from wellsites makes up a much larger area than the two wellsite buffer zones, and is comprised of the remaining home range for each bear. Therefore, results cannot be interpreted as “selection” or “avoidance” in this zone, as it is not possible for the bear to “avoid” its home range. However, we would expect that more or less use of the smaller zones would be reflected in shifts in use of the remaining home range. Selection ratios for this zone are interpreted as habitat use that is “less than expected”, “more than expected”, or neutral, and these results should correspond with results observed for the 100m and 500m zones. For example, for the 10 bears that showed selection of the 100m and/or 500m zones, the use of the remaining home range was less than expected. Similarly, for 3 bears that showed avoidance of the 100m and/or 500m zones, use of the zone >500m from wellsites was more than expected (selection ratios significantly greater than 1.00). Results were consistent across the three zones, with no contradictory results. The remaining 3 bears did not have significant results. Overall, use of the home range >500m from wellsites was slightly less than expected, with a mean selection ratio of 0.95. Comparatively, the mean selection ratio for the 100m zone (3.28) was more than twice as high as the mean SR for the 500m zone (1.30), and about three times the mean SR for the remaining home range (0.95). Although the mean selection ratio was slightly larger for females (3.69) than for males (2.38) within the 100m zone (Tables 9 and 10), no significant differences between males and females were detected for any of the three habitat zones (Mann-Whitney U test, p values >0.20). Similarly, no significant differences were detected for selection ratios between females with young and females without (p values >0.32). When daytime and nighttime selection ratios were compared for females, the mean nighttime selection ratio was greater than daytime, but no significant differences were detected for the 100m zone (Wilcoxon signed-rank test, p=0.441). However, nighttime selection ratios for females were significantly higher than daytime selection ratios in the 500m ZOI (Paired t-test, p=0.010), and accordingly, use of the zone >500m from wellsites was significantly lower during the day (Paired t-test, p=0.011). For males, there were no significant day/night differences in selection of the 100m zone (Wilcoxon signed-rank test, p=0.465), 500m zone (p=0.068), or home range >500m from wellsites (p=0.068). 41

Chapter 2-Response to Oil and Gas Activities Table 9: Selection ratios for individual females.

Bear G223 G224 G236 G238 G253 G254 G258 G260 G265 G269 G271

100m 1.08 4.54 0.42 1.11 2.46 7.41 3.23 3.05 2.81 10.31 4.14

500m 0.96 2.01 0.83 0.85 1.23 1.34 1.00 1.29 2.35 1.71 1.17

>500m 1.01 0.84 1.08 1.03 0.97 0.94 0.96 0.87 0.86 0.90 0.92

Mean SD

3.69 2.93

1.34 0.49

0.94 0.08

Table 10: Selection ratios for individual males.

Bear G230 G257 G262 G264 G270

100m 3.763991 1.90387 1.257392 2.875052 2.120415

500m 1.288362 1.075604 0.748522 1.503425 1.491013

>500m 0.981791 0.975639 1.084509 0.946906 0.828393

Mean SD

2.38 0.96

1.22 0.32

0.96 0.09

Availability of bear foods at wellsites We visited 16 wellsites in the Kakwa study area, including 9 active wellsites and 7 inactive wells. Bear foods were present on the wellpads at 14 of 16 wellpads. The two wellpads where bear foods were absent had both been cleared within the previous year and were 100% barren of vegetation. There was a consistent relationship between age of the wellsite (years since clearing) and percent of the wellpad that was vegetated, for both active and inactive wellsites (R2=0.87) (Figure 8).

Percent of wellpad vegetated

Chapter 2-Response to Oil and Gas Activities 100 80 60 Active

Inactive

20 0 0

Age (years)

Figure 8: Percent of wellpad vegetated versus age of wellpad (years since initial clearing).

Total abundance (cover estimation) of clover, dandelion and equisetum

A similar relationship was observed between age of wellsite and relative abundance of bear foods known as initial colonizers of disturbed areas, including dandelion (Taraxacum officinale), clover (Trifolium spp), and horsetails (Equisetum spp) (Figure 9). There was no significant difference (unpaired t-test) in the approximate abundance of the dandelion/clover/Equisetum group between active and inactive wellsites. Clover was the most abundant bear food on wellpads, (modal value = the 5-25% cover class), followed by Equisetum and dandelion. Alfalfa (Medicago sp.), strawberry (Fragaria sp.), Rubus species, and anthills were also common on wellpads. 10 8 6 Active

Inactive

2 0 0

Years since wellpad clearing

Figure 9: Total abundance of dandelion, clover and Equisetum versus wellpad age (year since initial clearing).

Berry species were relatively abundant along the edges of all 16 wellsites, ranging in abundance (total of all berry species) from cover class 2 (5-25%) to cover class 3 (26-50%). Berry species included Vaccinium species, Rubus sp., Viburnum sp., and Ribes sp. Vaccinium species included Vaccinium myrtilloides, V. membranaceum, V. vitis-idaea, and 43

Chapter 2-Response to Oil and Gas Activities V. caespitosum. The abundance of all Vaccinium species (grouped) ranged from cover class 1 (<5%) to cover class 3 (26-50%). There was no apparent relationship between the abundance of berries along wellsite edges and the number of years since clearing.

Discussion More than half (10 of 16, or 62.5%) of the bears in our study selected for wellpads, and only one bear showed avoidance. These results are in contrast to the avoidance of wellsites reported for other species including mule deer (Sawyer et al. 2009), caribou (Dyer et al. 2001), and elk (Powell 2003). SahlĂŠn (2010) previously reported that grizzly bears in the Kakwa area generally selected for areas <224m from the wellsite centre. Results from Topic A of this study (Labaree et al. 2012) indicated that bears were generally closer than expected to wellsites during spring, while the response in other seasons depended on sex/age class. Grizzly bears have been reported to select for other anthropogenic disturbances such as forestry cutblocks and road clearings. In the foothills of west-central Alberta, Nielsen et al. (2004a) reported that grizzly bears selected for harvested areas more than expected during the summer, Roever et al. (2008a) showed that grizzly bears selected habitats close to roads in spring and early summer, and Graham et al. (2010) found that females with cubs were within 200m of roads more than expected in spring. Stewart (2011) also reported grizzly bear use of edge habitat created by cutblocks, roads, and pipelines. When forests are cleared to construct oil and gas wellsites, the existing vegetation and top soil is removed, piled up, and stored next to the well pad. For reclaimed wellsites, guidelines include planting species which are representative of that natural sub-region, ecosite, and plant community to obtain â&#x20AC;&#x153;equivalent land capabilityâ&#x20AC;? (Alberta Environment, 2010). However, during production (usually over a 20 year period), these areas are not usually replanted. After initial clearing, early colonizing plant species begin to grow, and edge habitat is created where the openings meet the surrounding forest. Therefore, changes to the landscape due to resource extraction can create potential habitat for grizzly bears (Nielsen et al. 2004a, Stewart 2011). Results from our study indicate that a number of bear foods grow on wellpads and along the edges of wellsite clearings. Dandelion, clover, and Equisetum spp are frequent colonizers of disturbed areas, and these species were abundant on wellpads. Similarly, Roever et al. (2008b) reported that roadsides had a higher frequency of Equisetum spp., Taraxicum officinale, and Trifolium spp than forest habitats. These plants are an important part of the diet for grizzly bears in the foothills of west-central Alberta (Munro et al. 2006), and have also been identified as significant food items in the Kakwa region (Larsen & Pigeon 2006). Berry species were also relatively abundant along the wellsite edges, including important fall food items such as Vaccinium species. Although most research reports the avoidance of wellsites by wildlife, some species have been reported to select for oil and gas features when a valuable resource is associated with the feature, such as deer selection of saline seepage at gas wells in West Virginia (Campbell et al. 2004).

Chapter 2-Response to Oil and Gas Activities Our analysis did not account for effects of additional factors affecting selection at a broader scale, such as habitat quality, or the presence of adjacent cutblocks, roads, and pipelines. However, these factors are unlikely to have an effect at the small scale of the 100m buffer zone, which incorporated only the wellpad and immediate edge. At the small spatial scale of our analysis, the selection patterns can be interpreted as use of the wellpad. Our results indicated a stronger selection for the wellpad than for the surrounding area; overall, bears were at least twice as likely to select for the wellpad than for the 500m zone, and about three times as likely to use wellpads compared to the rest of the home range. SahlĂŠn (2010) also found that bears generally had the highest selection ratios in the zone <224m from the wellsite centre compared to additional buffers within 500m of the wellsite, although results only included bear locations within 500m of the well site, and selection was not compared to rest of home range. Anecdotal evidence from the Kakwa region includes sightings of bears foraging on wellpads, as well as observations of bear scat containing clover and Equisetum. Results from this study suggest that bears are attracted to wellsites by the abundance of food available on the wellpads and along the edges. On visual inspection of location points in GIS, we observed that selected wellpads were commonly visited by more than one bear over multiple days and years (Figure 10). One individual bear that showed strong selection for wellsites had 68 locations at one wellpad, across 10 different days and three different months, suggesting that the bear returned to the site repeatedly during that year.

Figure 10: Wellpad with 100m and 500m buffer. Points on the wellpad include locations for G229 (2006), G238 (2007), G265 (2007), and G254 (2008 and 2009).

There was individual variation in selection patterns among individuals, with some bears showing very strong selection for wellpads (e.g. SR=10.31), while one bear had a pattern of avoidance. Variation in selection ratios was not explained by sex class or reproductive status. These results are in contrast to other studies that have reported differences in habitat use and response to anthropogenic features between age-sex classes (Roever et al. 45

Chapter 2-Response to Oil and Gas Activities 2008a, Graham et al. 2010, Stewart 2011), including differences in habitat use between males and females in response to wellsites, pipelines, and roads (Labaree et al. 2012). Females with unknown reproductive status were grouped with females without young in this analysis; the potential presence of young with some of these females may have affected results. The relatively small sample size in this study may have prevented detection of significant differences; however, it is possible that differences in selection were related to individual foraging patterns rather than sex/age class or reproductive status. Differences in behaviour and feeding patterns between individual bears have been previously reported (Stenhouse, unpublished data). While not all bears showed selection for the 500m zone in our study, the vast majority of bears (13 of 16, or 81%) did not avoid this â&#x20AC;&#x153;zone of influenceâ&#x20AC;? surrounding the wellsite. Previous studies of grizzly bears have designated a ZOI around human disturbances (Mace et al. 1996, Benn 1998, Berland 2008). Benn (1998) reported that 89% of human-caused mortalities occurred within 500m from roads. Ungulates have also been observed to have a zone of influence or avoidance of wellsites of about 1000m (Hebblewhite 2011), but no previously published research has specifically investigated a ZOI for bears and wellsites. A zone of influence or avoidance does not appear to apply to wellsites in our study region. Interestingly, in our study, all bears that selected for the 500m zone also selected for the wellpad itself. The 500m zone is a larger area than the wellpad, and other factors may be affecting habitat selection in this zone. Habitat quality, along with the presence of cutblocks, roads, and pipelines as well as the presence of other bears within the 500m zone were not accounted for in this study. Therefore, while the lack of avoidance in this zone is conclusive, selection for this zone must be interpreted with caution. No significant differences were detected between day and night time selection of the wellpads, but nighttime selection was significantly higher than daytime selection for the 500m zone of influence. Due to relatively small sample sizes at the small spatial scale of the wellpad, we were unable to separate active and inactive wellsites in our analysis; therefore, potential response to human activity directly at the wellpads was not investigated. SahlĂŠn (2010) found significant higher selection ratios in the 224m wellsite buffer at night, and Labaree et al (2012) reported that bears were generally closer to oil and gas features at night. Previous research also indicates that bears used cutblocks more during the crepuscular and nocturnal periods (Nielsen et al., 2004a). It has been suggested that bears may avoid human presence by shifting their use of habitat near disturbances to periods of lower human activity (Aune & Kasworm 1989, Schwartz et al. 2010). Bears in our study may have been bedding down in the surrounding zones, or may have been avoiding human activity in these zones by using these areas at night. The 500m zone surrounding each wellsite in our study contains roads used to access the wellsite, and may also contain additional wellsites and other human developments (pipelines, cutblocks). Human activity levels could be higher in the 500m zone compared to the wellpad, which may explain why these areas are used less during the day. The influence of roads was not investigated in this study, but in part 1 of this project, results also indicated that bears were further from roads during the day versus the night (Labaree et al. 2012a). As previously 46

Chapter 2-Response to Oil and Gas Activities discussed, selection in this 500m zone surrounding the wellsite may be influenced by a number of additional factors, and must be interpreted with caution.

Literature Cited Alberta Environment. 2010. 2010 Reclamation Criteria for Wellsites and Associated Facilities for Forested lands. Alberta Environment, Edmonton, Alberta. 99 pp. Amstrup, S. C. 1993. Human disturbances of denning polar bears in Alaska. Arctic 46, 246250. Arthur, S. M. & Schwartz, C. C. 1999. Effects of sample size on accuracy and precision of brown bear home range models. Ursus 11, 139-149. Aune, K. & Kasworm, W. 1989, Montana Department of Fish, Wildlife, and Parks. Belant, J. L. & Follmann, E. H. 2002. Sampling considerations for American black and brown bear home range and habitat use. Ursus 13, 299-315. Benn, B. 1998. Grizzly bear mortality in the central rockies ecosystem, Canada. Eastern Slopes Grizzly Bear Project 1, 151. Berland, A., Nelson T., Stenhouse, G., Graham, K., & Cranston, J. 2008. The impact of landscape disturbance on grizzly bear habitat use in the Foothills Model Forest, Alberta Canada. Forest Ecology and Management 256, :1875-1883. Campbell, T. S., Laseter, B. R., Ford, W. A., & Miller, K. V. 2004. From the field: Unusual white-tailed deer movements to a gas well in the central Appalachians. Wildlife Society Bulletin 32, 983-986. Follman, E. H., Dieterich, R. A., & Hechtel, J. L. 1980. Recommended carnivore control program for the Northwest Alaskan Pipleine Project including a review of humancarnivore encounter problems and animal deterrent methodology. 113 pp. Forman, R. T. & Deblinger, R. D. 2000. The ecological road-effect zone of a Massachusetts (U.S.A.) suburban highway. Conservation Biology 14, 36-46. Ganskopp, D. C. & Johnson, D. D. 2007. GPS error in studies addressing animal movements and activities. Rangeland Ecology Management 60, 350-358. Gibeau, M. L., Herrero, S., Kansas, J. L., & Benn, B. 1996. Grizzly bear population and habitat status in Banff National Park . A Report to the Banff Bow Valley Task Force 1, 62. Girard, I., Ouellet, J., Courtois, R., Dussault, C., & Breton, L. 2002. Effects of sampling effort based on GPS telemetry on home-range size estimations. Journal of Wildlife Management 66, 1290-1300. Graham, K., Boulanger, J., Duval, J., & Stenhouse, G. 2010. Spatial and temporal use of roads by grizzly bears in west-central Alberta. Ursus 21, 43-56. Harding, L. & Nagy, J. A. 1980. Responses of grizzly bears to hydrocarbon exploaration on 47

Chapter 2-Response to Oil and Gas Activities Richards Island, Northwest Teerritories, Canada. Bears-Their Biology and Management 4, 277-280. Hebblewhite, M.2011. Effects of energy development on ungulates. In Energy Development and Wildlife Conservation in Western North America, ed. D. E. Naugle, Island Press, Washington D.C., U.S.A. Larsen T. and Karine Pigeon 2006, The diet of grizzly bears (Ursus arctos) in North-Central Alberta, Canada, part of a new food-based model approach to mapping grizzly bear habitat In: Stenhouse, G. and K. Graham (eds). Foothills Model Forest Grizzly Bear Research Program 2006 Annual Report. Hinton, Alberta. 87 pp. Laver, P. N. & Kelly, M. J. 2008. A critical review of home range studies. The Journal of Wildlife Management 72. Mace, R. D., Waller, J. S., Manly, T. L., Lyon, L. J., & Zuuring, H. 1996. Relationships among grizzly bears, roads and habitat in the Swan Mountains, Montana. Journal of Applied Ecology 1395-1404. Manly, B. F. J., McDonald, L. L., Thomas, D. L., McDonald, T. L., & Erickson, W. P. 2002. Resource Selection by Animals: Statistical Design and Analysis for Field Studies, 2nd edn. Boston, Massachusetts, USA. McLellan, B. N. & Shackleton, D. M. 1988. Grizzly bears and resource-extraction industries: effects of roads on behaviour, habitat use and demography. Journal of Applied Ecology 25, 451-460. Munro, R. H. M., Nielsen, S. E., Price, M. H., Stenhouse, G. B., & Boyce, M. S. 2006. Seasonal and diel patterns of grizzly bear diet and activity in west-central Alberta. Journal of Mammalogy 87, 1112-1121. Nielsen, S. E. 2004. Habitat ecology, Conservation, and Projected Population Viability of Grizzly Bears (Ursus arctos L.) in West-Central Alberta, Canada PhD Thesis. Department of Biological Sciences, University of Alberta, Edmonton, Alberta, Canada. Nielsen, S. E., Boyce, M. S., & Stenhouse, G. B. 2004a. Grizzly bears and forestry I. Selection of clearcuts by grizzly bears in west-central Alberta, Canada. Forest Ecology and Management 199, 51â&#x20AC;&#x201C;65. Nielsen, S. E., Munro, R. H. M., Bainbridge, E. L., Stenhouse, G. B., & Boyce, M. S. 2004b. Grizzly bears and forestry II. Distribution of grizzly bear foods in the clearcuts of west-central Alberta, Canada. Forest Ecology and Management 199, 67â&#x20AC;&#x201C;82. Powell, J. 2003, Distribution,habitat use patterns, and elk response to human disturbance in the Jack Marrow Hills, Wyoming University of Wyoming. Reynolds, P. E., Reynolds, H. V., & Follmann, E. H. 1986. Responses of grizzly bears to seismic surveys in northern Alaska. International Conference on Bear Research and Management 6, 169-175. 48

Chapter 2-Response to Oil and Gas Activities Roever, C. L., Boyce, M. S., & Stenhouse, G. B. 2008a. Grizzly bears and forestry II: Grizzly bear habitat selection and conflicts with road placement. Forest Ecology and Management 256, 1253-1261. Roever, C. L., Boyce, M. S., & Stenhouse, G. B. 2008b. Grizzly bears and forestry I: Road vegetation and placement as an attractant to grizzly bears. Forest Ecology and Management 256, 1253-1261. Sahlen, E. 2010. Do grizzly bears use or avoid wellsites in west-central Alberta, Canada? MSc Thesis, Swedish University of Agricultural Sciences, Faculty of Foresty, Department of Wildlife, Fish and Environmental Studies. Schallenberger, A. 1980. Review of oil and gas exploitation impacts on grizzly bears. BearsTheir Biology and Management 4, 271-276. Schwartz, C. C., Cain, S. L., Podruzny, S., Cherry, S., & Frattaroli, L. 2010. Contrasting activity patterns of sympatric and allopatric black and grizzly bears. Journal of Wildlife Management 74, 1628-1638. Silverman, B. W. 1986. Density Estimation for Statistics and Data Analysis Chapman and Hall, London, UK. Stenhouse, G.B., and Munro, R.H.M. 2000. Foothills Model Forest Grizzly Bear Research Program: 1999 Annual Report. Stewart, B. P. S. 2011, Quantifying grizzly bear habitat selection in a human disturbed landscape University of Victoria. Tietje, W. D. & Ruff, R. L. 1983. Responses of black bears to oil development in Alberta. Wildlife Society Bulletin 11, 99-112.

Chapter 2-Response to Oil and Gas Activities

TOPIC C: GRIZZLY BEAR MOVEMENT IN RELATION TO OIL AND GAS ACTIVITIES Karen Laberee, Benjamin Stewart, Trisalyn Nelson, Tracy McKay, Jed Long, and Gordon Stenhouse2

Introduction In response to increasing scientific and public concern over the low population estimates of grizzly bears (Ursus arctos) in Alberta, the species was designated as Threatened by the provincial government in 2010 (Clark and Slocombe 2011). The Alberta populations are characterized by especially low fecundity and low population densities, which further hinders their recovery (Nielsen et al. 2009). Loss and fragmentation of grizzly bear habitat resulting from human disturbance is an additional concern (Ross 2002). The Kakwa region of Alberta, Canada is home to ongoing activities related to the extraction of natural resources. Recent research included in Topic A of this report (Laberee et al. 2012) has indicated that bears in this area were closer than expected to some oil and gas related disturbances in some seasons, but further than expected from other features. Bears have also been shown to use cutblocks in this area (Nielsen et al. 2004; Berland et al. 2008; Stewart et al. 2012). In addition to studies that link grizzly bear locations to different features on the landscape, studies of movement can give us an indication of how the landscape is being used by wildlife (Dickson et al. 2005; Chetkiewicz et al. 2006). Anthropogenic disturbance can result in changes in movement patterns. Patterns in the timing of daily movement of grizzly bears have been reported to change under increasing anthropogenic pressures (Schwartz et al. 2010). In a movement study of another large carnivore, the cougar (Puma concolor), the authors reported faster velocities through urban areas (Dickson et al. 2005). Results from Topic A of this study (Laberee et al. 2012) provide information regarding whether grizzly bears use habitat containing oil and gas features, but behaviours or movement patterns associated with this habitat use are unknown. The objective of our research was to determine if resource extraction activities influence the movement of grizzly bears in the Kakwa region of Alberta, Canada. To achieve this we framed our research to answer the following questions: a. Are there differences in grizzly bear movement rates (velocities) related to the presence or absence of disturbance features in the vicinity? b. Are the densities of disturbance features different in areas of slow movement as compared to the rest of the home range? We took two approaches to characterizing movement. The first movement measure is a standard measure of velocity or the speed travelled, calculated as a ratio of the distance 50

Chapter 2-Response to Oil and Gas Activities and time between two consecutive telemetry points. The second involves the use of a novel method adapted from time geography and ecological movement models (Morales et al. 2004; Long and Nelson, 2012) to identify areas of slow movement, defined as â&#x20AC;&#x153;latent areasâ&#x20AC;?. The latent area is a specific geographic area where bears are observed to spend the largest amount of consecutive time with minimal distance traveled. For both objectives, we related movement to the various disturbance features (wells, roads, pipelines, and cutblocks) within 500m of bear location points. Grizzly bears are sexually dimorphic, with the much larger males requiring maximum foraging opportunities (Rode et al. 2006). Males are known to travel more widely than their female counterparts (Proctor et al. 2004); therefore, we expect that males should have different movement patterns from females. As well, the predominance of various foraging activities (herbivory, root digging, frugivory, insect foraging, and ungulate kills) changes seasonally (Munro et al. 2006), and we expect that movement patterns should also vary by season. As such, we also expect to observe distinct seasonal differences in movement rates and the characteristics of latent areas. Therefore, we examined adult female and male grizzly bear movements separately, for three distinct foraging seasons (spring, summer, and fall).

Methods Grizzly bear telemetry data Telemetry data were collected for 33 grizzly bears in west-central Alberta, Canada from 2005-2010. Grizzly bear capture followed the accepted protocols of the Canadian council of animal care for the safe handling of bears (animal use protocol number 20010016) (Stenhouse and Munro 2000). Bears were captured using a combination of aerial darting, leg-hold snaring and culvert traps, and bears were fitted with a Televilt/Followit brand GPS collar (Lindesburg, Sweden). Telemetry data were cleaned for accuracy using positional dilution of precision (PDOP), which evaluates three-dimensional accuracy of locations. Telemetry locations with a PDOP value greater than 10 were eliminated (D'Eon and Delparte 2005). Grizzly bear telemetry data were separated into seasonal categories following established feeding patterns for our study area (Munro et al. 2006). Seasons were defined as spring (1 May until 15 June), summer (16 June to 31 July), and fall (1 August to 15 October) (Nielsen 2004). Oil and gas features Disturbance datasets were obtained describing oil and gas wellsites, pipelines, and roads. Wellsite and pipeline data were provided to us from a database maintained by Alberta Energy; spatial data and attributes were originally obtained directly from oil and gas operators. The roads dataset was originally obtained from Alberta Sustainable Development, and subsequently updated by our research team through heads-up digitizing using medium and high resolution imagery. Roads in this area not exclusively used by the oil and gas industry, but for the purpose of this study, they are considered 51

Chapter 2-Response to Oil and Gas Activities to be a feature of this industry. Roads data were attributed with a year of disturbance (if constructed after 1994), pipeline data were attributed with date of construction, and wellsite data were attributed with date of initial clearing and drilling. Disturbance data A series of stand replacing forest disturbances were created through image differencing of a series of satellite image pairs from 1973 â&#x20AC;&#x201C; 2008 , based on Landsat imagery (see White et al.2011). This dataset detects all natural and anthropogenic stand-replacing disturbances; however, we were concerned specifically with cutblocks. Following the change classification routines of Stewart et al. (2009), non-harvest disturbances were eliminated. The cutblock data were then integrated into the oil and gas disturbance database. Resource Selection Functions A resource selection function (RSF) is estimated through comparisons of observed resource use to availability of resource units (Boyce et al. 2002). We created our seasonal RSFs following Nielsen et al. (2009). Individual RSFs were created for each foraging season (spring, summer, and fall periods, as defined for grizzly bear telemetry data) and each year of our study area (2005â&#x20AC;&#x201C;2010), creating a total of 18 RSFs. The RSF model predicts probability of resource use based on land cover, crown closure, species composition (conifer canopy), compound topographic index, distance to streams, and distance to forest clearings (Nielsen et al. 2009). Models were created from a random sample of 90% of the telemetry data, with the remaining set aside for model evaluation. RSF accuracy was evaluated using the remaining 10% of the training data using Spearman rank coefficients. The results from 2007 showed p-values of 0.005, 0.0013, and 0.0010 for the spring, summer, and fall, respectively, indicating sufficient accuracy for use in further analysis. Disturbance Buffers A 500m buffer was generated around each bear location point. Within each buffer, the number of wells, length of roads and pipelines (metres), and area of cutblocks (km 2) was determined. Location points were partitioned into two categories: 1) those that had zero disturbances within 500m (disturbance absent), and 2) those that had greater than zero disturbances within 500m (disturbance present). Velocity Analysis The Euclidean distances between telemetry fixes were calculated. A simple measurement of displacement velocity (m/hr) was calculated between every two consecutive GPS locations for adult male and female grizzly bears in each season. A Kolmogorov-Smirnov (K-S) test was used to compare velocities at points with zero disturbances (disturbance absent) to those with greater than zero disturbances (disturbance present) within the 500m buffer. Velocities were examined separately for each disturbance category (wells, roads, pipelines, and cutblocks). Data were also grouped by oil and gas related disturbances (wells, roads, and pipelines), and with all disturbances included (wells, roads, pipelines and cutblocks). 52

Chapter 2-Response to Oil and Gas Activities Latent Areas Following Long and Nelson (2012), we calculated grizzly bear home ranges using the potential path area (PPA) home range method. The PPA home range is the potential area accessible to the animal, given its sequence of telemetry fixes and a parameter of movement ability termed vmax, related to the animal’s maximum travelling speed. For each telemetry fix, an ellipse is created between that location and the next fix in the telemetry dataset. The size of this ellipse is governed by the vmax parameter and the time difference between the fixes (Figure 11). This ellipse encompasses the entire area the bear could have traversed based on its movement ability (as defined by vmax). Combining the n-1 ellipses (from a dataset of n telemetry fixes) results in the PPA home range (Long and Nelson, 2012). We developed a novel method of analyzing grizzly bear telemetry data in order to quantify “latency” in grizzly bear movement paths. Our new latency measure utilizes the PPA home range described above to determine the time periods and areas where individual grizzly bears spend the longest continuous period of time in each season. As a concept, latency includes those behaviours where the grizzly bear exhibits little or no movement. Using GPS telemetry data, we can observe latency behaviour by identifying spatial clusters of consecutive fixes. The key distinction with latency compared to other cluster-based concepts (e.g. core areas) is that the fixes must be consecutive in order to be considered latent. A natural choice for identifying latency in telemetry data is to delineate a latent area around a telemetry fix, and count the number of subsequent consecutive fixes that occur within this area. The count statistic, denoted Li, can be computed for each fix in the telemetry dataset. The simplest method for calculating the area-of-latency for each telemetry fix would be to compute a spatial buffer (with a fixed width) around each fix, and compute Li. However, this would require the objective selection of a buffer width. Rather than a spatial buffer, we chose to use the PPA ellipse (described above) as the latent area for each telemetry fix. Use of the PPA ellipse is advantageous in two ways: first, it does not include any inaccessible areas in the latent area, given the known movement trajectory; and second, it degenerates to a circle (or spatial buffer) when the bear is completely stationary. The maximum value of Li was identified and used in subsequent analysis. Based on the derivation above, this latent period consists of a subset of consecutive fixes from the telemetry dataset. We extracted those fixes that made up the latency period; for example, if max Li = 24, along with fix i we extract the 24 subsequent fixes following fix i. We then computed the PPA home range of the latent fixes only, which provides a spatial representation of the latent area. The latent area is a subset of the home range, much in the same way a core area is a subset of the home range. Each latent area in each season was often used by a bear for a period of several days.

Chapter 2-Response to Oil and Gas Activities

Figure 11. The determination of a PPA ellipse to define latent areas.

The home range of each individual grizzly bear was divided into two regions: latent areas and non-latent areas. Based on whether or not a location fell within a latent or 54

Chapter 2-Response to Oil and Gas Activities non-latent area, each location point was then classified as latent or non-latent. To provide more detailed information than presence/absence, the number of wellsites, length of roads and pipelines (metres), and area of cutblocks (km2) present within the 500m buffer areas were calculated for latent and non-latent points. The number (or length or area) of disturbance features was compared between latent and non-latent points using a Kolmogorov-Smirnov test. RSFs predict grizzly bear occurrence based on environmental covariates, and RSF values are often used to infer habitat quality. The average RSF values for the latent areas and non-latent areas were also calculated for each sex and season, and RSF values were compared between the latent and non-latent areas to determine if habitat quality affected latency results. Telemetry Interpolation As is inevitable with GPS-based telemetry data, missing fixes can occur due to physical obstructions and/or technical malfunctions (Rempel et al. 1995). Missing locations can result in biases in subsequent analyses (Frair et al. 2004), and are problematic when considering the temporal trajectory of telemetry data. In this study, we are interested in quantifying latency, defined as stationary behaviour determined from GPS telemetry data. Latency is directly related to the temporal structure of telemetry data, and missing fixes can disrupt this structure. In order to improve the analysis of latent areas/time periods, we estimated records for those fixes that are missing. To account for missing fixes, a linear interpolation technique was used. The result is a telemetry dataset where fixes (observed and estimated) are taken at a regular time interval. We chose a linear interpolation technique over more complex alternatives (e.g., Wentz et al. 2003, Tremblay et al. 2006). When the GPS-signal was disrupted for an extended period of time (i.e., > 4 fixes), we analyzed the bear trajectory separately on either side of the disruption. This procedure avoids introducing biases associated with unreliable interpolation of numerous consecutive fixes.

Results Mean grizzly bear velocity results related to the presence or absence of features (wellsites, roads, pipelines, and cutblocks) within the 500m disturbance buffers are presented in Table 11. For wellsites, significant differences in velocities were detected in fall, but results differed between the sexes. For females, velocities were higher in the presence of wells, while for males, velocities were lower when wellsites were present. No significant effects were observed in spring or summer. At locations where roads were present within the 500m buffer, grizzly bear movement velocities were significantly greater than movement velocities at locations where roads were absent. These results were consistent across males/females and all seasons other than females in the fall. Grizzly bear velocities were also significantly higher for all sexes and seasons when pipelines were present. When oil and gas related disturbances were considered as a group 55

Chapter 2-Response to Oil and Gas Activities (wellsites, roads, and pipelines), significantly higher velocities were found for all bear groups if any of the oil and gas disturbances were present, compared to locations where none of these features were present within 500m. The effects of cutblocks varied with bear group and season. Females in spring had lower velocities at points where cutblocks were present within 500m, while velocities in the summer were lower if cutblocks were absent. In the summer, males had significantly higher velocities when cutblocks were present. All other comparisons with cutblocks showed no significant differences in velocities. When all disturbance features were grouped (wellsites, roads, pipelines, and forest cutblocks), velocities were higher in the presence of any disturbance compared to velocities at locations with no disturbances within 500m. Significant differences were seen in summer and fall for females, and spring and summer for males. The average RSF values partitioned by sex and season are compared graphically in Figure 12. RSF values were not significantly different between latent and non-latent areas, in any seasons, or between males and females. 10 9

Average RSF Value

8 7 6 Latent Area

Non-latent Area

4 3 2 1 0 Adult Female Spring

Adult Female Summer

Adult Adult Male Female Fall Spring

Adult Male Summer

Adult Male Fall

Figure 12. A comparison of average RSF values for latent and non-latent areas. Error bars are one standard deviation.

Chapter 2-Response to Oil and Gas Activities

Table 11. Average velocity (m/hr) observed for grizzly bear telemetry locations in undisturbed (disturbance = 0) and disturbed areas. O&G represents cumulative disturbance associated with wells, roads, and pipelines. All represents cumulative disturbances associated with wells, roads, pipelines, and cutblocks. * Indicates significant difference as calculated with a K-S test. Female Spring

Female Summer

Female Fall

Male Spring

Male Summer

Male Fall

Disturbance

Wells

266

222

315

395

258

*261

572

681

463

524

397

*357

Roads

240

*260

297

*321

261

*257

524

*699

436

*546

335

*531

Pipelines

242

*262

300

*324

242

*283

539

*721

453

*529

373

*499

Cutblocks

264

*244

233

*299

254

238

557

629

459

*491

437

388

O&G

228

*264

286

*325

246

*266

502

*702

435

*529

356

*467

All

278

247

291

*315

224

*269

469

*671

427

*517

345

453

Chapter 2-Response to Oil and Gas Activities The number of wellsites within 500m of latent location points was significantly greater than for non-latent points for adult females in spring (Table 12). For all other groups, the number of wells within 500m was not significantly different between latent and nonlatent points. In the spring, there was a significantly shorter length of roads within 500m of latent points for adult males and females, but only female latent points had significantly less roads in summer (Table 12). However, in fall, latent points for both adult males and females had a significantly greater length of roads within 500m as compared to non-latent points. Response to pipelines varied by sex and season (Table 12). In the spring, there was a significantly greater length of pipeline within 500m of adult female latent points, while in the summer there was significantly less pipeline within 500m of adult male latent points. In fall, both adult bear groups had significantly more pipeline within 500m of latent telemetry points. While the area of cutblocks within 500m of latent telemetry points was significantly different than non-latent points for all comparisons, results were not consistent (Table 12). Males in spring, females in summer, and both sexes in fall had significantly less harvested area within 500m of their latent points.

Chapter 2-Response to Oil and Gas Activities

Table 12. Average values of measures of anthropogenic disturbances (number of wells, metres of roads andpipelines, and km2 of cutblocks) within 500m of grizzly bear latent points (L) and non-latent points (NL). Latent points were compared to the non-latent points with a K-S test; z scores are listed. with * indicating a significant difference.

Disturbance

Spring L

Summer NL

Fall

Wells (number)

Females 0.53

0.35

*3.210 0.39

0.35

0.939

0.24

0.31

1.728

Males

0.19

1.027

0.16

1.780

0.08

0.15

1.375

Roads (m)

Females 484.9 555.9 *2.503 451.3 491.3 *2.882 768.2 531.5 *9.006 Males

0.18

0.07

476.1 552.0 *2.226 279.2 271.5 1.926

770.5 383.9 *4.492

Pipelines (m)

Females 993.9 664.2 *2.513 623.8 614.1 1.037

818.8 635.1 *6.172

Cutblocks (km2)

Females 0.130 0.108 *4.436 0.094 0.109 *3.076 0.080 0.120 *4.526

Males

613.2 773.4 1.624

200.9 358.4 *2.594 902.9 332.8 *11.277

0.043 0.096 *3.270 0.074 0.071 *3.246 0.045 0.069 *6.597

Chapter 2-Response to Oil and Gas Activities

Discussion We used the relationships between consecutive grizzly bear telemetry points to characterize movement patterns, using velocities of travel as well as areas and points of slow movement (latency). Information regarding movement patterns complements knowledge about habitat use, providing more information to assist in management for grizzly bears in areas of oil and gas development. RSF values were similar between the latent areas and the non-latent areas. This indicates that use of latent areas by grizzly bears was not explained by the landscape characteristics (environmental covariates) incorporated in the RSF values (land cover, crown closure, species composition, compound topographic index, distance to streams, and distance to forest clearings). Therefore, use of latent areas may be related to other influences on habitat use, such as nearby disturbances (Chetkiewicz et al. 2006). The presence of wellsites was not related to higher velocities for males in any season, or for females in either spring or summer. In the spring, latent points for females also had significantly more wellsites than non-latent locations, with no differences for summer or fall. These results suggest that grizzly bear movement rates are not generally faster in the presence of wellsites. Results from Topic B of this report suggest that wellsites may provide valuable food sources for bears (McKay et al. 2012). Foods available on wellsites (clover, equisetum, and dandelion) are known to be important spring and summer foods for bears (Nielsen et al. 2004c; Larsen & Pigeon 2006, Munro et al. 2006), and slow movement rates (latency) in the presence of wellsites is consistent with foraging at these sites. In contrast, roads and pipelines were associated with higher velocities for both males and females across almost all seasons, but the average length of roads and pipelines within 500m of latent points varied by season. In other words, higher movement velocities were consistently associated with roads and pipelines, but in some seasons, latent (low movement) points had more kilometres of roads or pipelines nearby. This apparent contradiction in results between the velocity analysis and latency analysis indicates that velocity of travel may not be directly determined by the total length of roads or pipelines within 500m of a bear. Our previous work (Labaree et al. 2012) found somewhat variable response to roads and pipelines, but both males and females were generally closer to both roads and pipelines in spring and fall. In a study of grizzly bear response to roads in the foothills of westcentral Alberta, Graham et al. (2010) also reported variable results across sex/age classes and seasons; some groups were close to roads more than expected, while others were close to roads less than expected. In research completed in Topic A of this report, bears were closer to roads than to any other type of oil and gas related feature, in spite of a greater density of pipelines in the area. The consistently higher velocities in the presence of roads and pipelines could indicate faster movement through these areas to avoid human presence, or it could indicate use of these linear features as travel 60

Chapter 2-Response to Oil and Gas Activities corridors. At the spatial scale of our analysis (linear features within 500m), it is not possible to determine whether the bears were on the linear features themselves. Research on other large carnivores has shown that the highest movement rates are in areas of human disturbance (Dickson et al. 2004). The response of animals to disturbance, including changes in behaviour or displacement, can cause energetic losses (Bradshaw et al. 1998, Cassirer et al. 1992, Frid & Dill 2002, White et al. 1995). While some wildlife species appear to avoid linear developments, previous research has also shown that wolves (and other species) use linear corridors (roads, seismic lines, trails, railway lines) as travel routes (Thurber et al. 1994, Musiani et al. 1998, James and Stuart-Smith 2000, Whittington et al. 2005). Nielsen et al. (2004) modeled the spatial distribution of human-caused grizzly bear mortalities in the Central Rockies, and found that grizzly bear mortalities were more likely to occur closer to linear human use features. If bears in our study are moving faster through areas with roads and pipelines to avoid human activity, there could be energetic consequences resulting from the faster movement rates. On the other hand, if bears are using roads and pipelines as travel corridors, there could be consequences for survival. Collectively, the group of oil and gas related disturbances (wellsites, roads, and pipelines) were always associated with significantly higher average velocities. However, based on the analyses of separate disturbance features, roads and pipelines are associated with higher velocities, and are likely affecting the grouped results. All wellsites have an associated road access, but roads do not always have wellsites in the proximity. These associations may complicate interpretation of results; however, distinctly different patterns were observed with wellsites (no increase in velocity) as compared to roads (increased velocity). Therefore, it seems likely that the effects of wellsites and roads were separated out in our analysis. Velocities associated with cutblocks were not consistent across groups and seasons. During summer, both males and females had significantly greater average velocities when cutblocks were present; however, the area of cutblocks within 500m of latent points was greater during summer for males. Compared to velocities in the presence of other disturbance features, average velocities in the presence of cutblocks were the lowest. These results suggest that bears do not generally have higher movement rates in the presence of forestry cutblocks. Bears have been shown to select for cutblocks due to food resources they provide, particularly in the summer (Nielsen et al. 2004; Berland et al. 2008), and lower movement rates in these areas are consistent with bears using these features. A lesser area of cutblocks was associated with latent points during fall, consistent with the decrease in the use of cutblocks in fall previously reported in the literature (Nielsen et al. 2004).

Chapter 2-Response to Oil and Gas Activities

Literature Cited Berland, A., Nelson, T.A., Stenhouse, G.B., Graham, K., and Cranston, J. 2008. The impact of landscape disturbance on grizzly bear habitat use in the Foothills Model Forest, Alberta, Canada. Forest Ecology and Management. 226: 1875-1883. Boyce, M.S., Vernier, P.R., Nielsen, S.E., and Schmiegelow, F.K.A.. 2002. Evaluating resource selection functions. Ecological Modelling 157: 281–300. Chetkiewicz, C.-L.B., St. Clair, C.C., and Boyce, M.S. 2006. Corridors for conservation: integrating pattern and process. Annual Review of Ecology, Evolution and Systematics. 37: 317-42. Clark, D.A., and Slocombe, D.S. 2011. Grizzly bear conservation effort in the Foothills Model Forest: appraisal of a collaborative ecosystem management effort. Policy Science 4: 1–11. D'Eon, R., and Delparte, D. 2005. Effects of radio-collar position and orientation on GPS radio-collar performance, and the implications of PDOP in data screening. Journal of Applied Ecology 42:383–388. Dickson, B.G., Jenness, J.S., and Beier, P. 2005. Influence of vegetation, topography, and roads on cougar movement in Southern California. Journal of Wildlife Management. 69(1): 264-276. Frair, J.L., Nielsen, S.E., Merrill, E.H., Lele, S.R., Boyce, M.S., Munro, R.H.M., Stenhouse, G.B., and Beyer, H.L. 2004. Removing GPS collar bias in habitat selection studies. Journal of Applied Ecology 41:201-212. Franklin, S.E., Stenhouse, G.B., Hansen, M.J., Popplewell, C.C., Dechka, J.A., and Peddle, D.R.2001. An Integrated Decision Tree Approach (IDTA) to mapping landcover using satellite remote sensing in support of grizzly bear habitat analysis in the Alberta Yellowhead ecosystem. Canadian Journal of Remote Sensing 27(6): 579592. Hunter, A. 2007. Sensor-based Animal Tracking. Thesis. University of Calgary, Calgary, Alberta. Laberee, K., Stewart, B.P., Nelson, T.A., McKay, T., Stenhouse, G.B., and Saxena, A. (In review). Grizzly bear response to oil and gas development and activities in Alberta. Larsen T. and Karine Pigeon 2006, The diet of grizzly bears (Ursus arctos) in NorthCentral Alberta, Canada, part of a new food-based model approach to mapping grizzly bear habitat In: Stenhouse, G. and K. Graham (eds). Foothills Model Forest Grizzly Bear Research Program 2006 Annual Report. Hinton, Alberta. 87 pp. Long, J.A., and Nelson, T.A. 2012. Time geography and wildlife home range delineation. Journal of Wildlife Management 76(2): 407-413.

Chapter 2-Response to Oil and Gas Activities Morales, J.M., D.T. Haydon, J. Frair, K.E. Holsinger, J.M. Fryxell. 2004. Extracting more out of relocation data: building movement models as mixtures of random walks. Ecology 85(9): 2436-2445. Munro, R.H.M., Nielsen, S.E., Stenhouse, G.B., and Boyce, M.S. 2006. Seasonal and diel patterns of grizzly bear diet and activity in west-central Alberta. Journal of Mammalogy 87: 1112–1121. Nielsen, S.E. 2005. Habitat ecology, conservation and projected population viability of grizzly bears (Ursus arctos L.) in west-central Alberta, Canada. Dissertation. Department of Biological Studies, University of Alberta, Edmonton, Alberta, Canada. Nielsen, S.E., Boyce, M.S., and Stenhouse, G.B. 2004. Grizzly bears and forestry I. Selection of clearcuts by grizzly bears in west-central Alberta, Canada. Forest Ecology and Management 199(1):51-65. Nielsen, S.E., Cranston, J. and Stenhouse, G.B. 2009. Identification of priority areas for grizzly bear conservation and recovery in Alberta, Canada. Journal of Conservation Planning 5: 38–60. Proctor, M.F., McLellan, B.N., Strobeck, C., and Barclay, R.M.R.. 2004. Gender-specific dispersal distances of grizzly bears estimated by genetic analysis. Canadian Journal of Zoology. 82: 1108-1118. Rempel, R.S., Rodgers, A.R., and Abraham, K.F.1995. Performance of a GPS animal location system under boreal forest canopy. Journal of Wildlife Management 59: 543-551. Rode, K.D., Farley, S.D. and Robbins, C.T. 2006. Behavioral responses of brown bears mediate nutritional effects of experimentally introduced tourism. Biological Conservation 133: 70–80. Ross, P.I. 2002. Update COSEWIC status report on the grizzly bear Ursus arctos in Canada, Committee on the Status of Endangered Wildlife in Canada. Ottawa. 99 p. Schneider, R.R. 2002. Alternative Futures: Alberta's Boreal Forest at the Crossroads. The Federation of Alberta Naturalists, Edmonton, Alberta. Schwartz, C.C., Cain, S.L., Podruzny, S., Cherry, S., and Frattaroli, L. 2010. Contrasting activity patterns of sympatric and allopatric black and grizzly bears. Journal of Wildlife Management 74(8): 1628–1638. Stenhouse, G.B., and Munro, R.H.M. 2000. Foothills Model Forest Grizzly Bear Research Program: 1999 Annual Report. Stewart, B.P., Wulder, M.A., McDermid, G.J., and Nelson, T. 2009. Disturbance capture and attribution through the integration of Landsat and IRS-1C imagery. Canadian Journal of Remote Sensing 35(6): 523-533. 63

Chapter 2-Response to Oil and Gas Activities Stewart, B.P., Nelson, T.A., Wulder, M.A., Nielsen, S.E., and Stenhouse, G.B. (In Press) Impact of disturbance characteristics and age on grizzly bear habitat selection. Applied Geography (Accepted December 2011). Tremblay, Y., Shaffer, S. A., Fowler, S. L., Kuhn, C. E., McDonald, B. I., Weise, M. J., Bost, C. A., Weimerskirch, H., Crocker, D. E., Goebel, M. E. and Costa, D. P. (2006). Interpolation of animal tracking data in a fluid environment. The Journal of Experimental Biology 209, 128-140. Wentz, E.A., Campbell, A.F., and Houston, R. 2003. A comparison of two methods to create tracks of moving objects: linear weighted distance and constrained random walk. International Journal of Geographical Information Science 17: 623645. White, J.C., Wulder, M.A., Gomez, C., and Stenhouse, G. 2011. A history of habitat dynamics: Characterizing 35 years of stand replacing disturbance. Canadian Journal of Remote Sensing. 37(2): 234-251.

Chapter 2-Response to Oil and Gas Activities

TOPIC D: MORTALITY RISK RELATED TO OIL AND GAS DEVELOPMENT Tracy McKay, Darren Wiens, Jerome Cranston, and Gordon Stenhouse

Introduction Based on the results from this report, it appears that grizzly bears have variable patterns of habitat use around oil and gas developments. Although some sex/age classes avoided developments in some seasons, our results did not indicate a strong pattern of avoidance of oil and gas features across all sex/age classes and seasons. In some cases, our results suggested selection or use of these features, potentially for food resources or movement. The primary limiting factor for grizzly bears in Alberta is human-caused mortality (FestaBianchet 2010). In other regions of Alberta, it has been shown that increased development may create habitat for grizzly bears (Nielsen et al. 2004a), but development can also result in increased grizzly bear mortality risk due to human access into grizzly bear habitat (Nielsen et al. 2004b). Human-caused mortalities are common in grizzly bears, and the mortalities are related to linear access features (Benn 1998, McLellan et al. 1999; Benn and Herrero 2002; Nielsen et al. 2004b). Benn (1998) reported that 89% of human-caused grizzly bear mortalities in the Central Rockies Ecosystem of Alberta (1972-1996) occurred within â&#x20AC;&#x153;zones of influenceâ&#x20AC;? along roads and trails. As development increases in grizzly bear habitat, bears may further increase their individual mortality risk by selecting and using development features as attractive sinks (Nielsen 2007); their presence near these features increases their risk of bear-human interactions. With increased mortality risk as a potential consequence of grizzly bear habitat use around oil and gas developments, specific information regarding grizzly bear mortalities and mortality risk in the Kakwa is important for management. Our objectives were to examine data on known grizzly bear mortalities in the Kakwa as well as to investigate the changes in mortality risk that result from ongoing development. An additional objective was to compare predicted mortality risk values with the locations of actual mortality events, to determine if our current models are accurately predicting risk for this area. Based on these objectives, we developed the following research questions: a. What are the numbers of reported mortalities and their causes in the Kakwa region? b. What is the predicted annual mortality risk in the Kakwa study area, and how does annual mortality risk change from 2005 to 2010? c. How does mortality risk vary between grizzly bear home ranges, and from year to year for individual bears? d. How do known mortality locations in the Kakwa compare to predicted risk values from the existing mortality risk model?

Chapter 2-Response to Oil and Gas Activities To meet our objectives, we analyzed current mortality datasets and applied updated datasets using mortality risk models to create annual mortality risk surfaces specific to the Kakwa study area.

Methods Mortality data Sources of reported mortalities included Alberta Fish and Wildlife and the Foothills Research Institute Grizzly Bear Program (FRIGBP), and data used for the analysis were maintained in a FRIGBP database. Mortalities were mainly detected through registration of legal hunting (previous to 2006), bears found incidentally, collared bears found using radio telemetry, and â&#x20AC;&#x153;problemâ&#x20AC;? bears killed through management actions. Data were up to date as of December 2011. Mortality data included the location at which the bears were found (usually the presumed site of death), mortality location accuracy, approximate date of death, and mortality cause. Based on information available at the time of the mortality, causes were categorized as follows: legally hunted, illegally hunted, legal defense of life or property, management actions, research related, mistaken for black bear, road kill, natural causes, rail kill, and unknown. Although the database contained reported mortalities from as early as 1972, early data were summarized by population unit, and did not consistently include mortality locations until 1999. Therefore, only mortalities within the Kakwa study area that occurred during or after 1999 were included in this analysis. Mortalities were summarized by year (1999 through 2010) and by mortality cause. For comparison to mortality data from other regions in Alberta, the number of mortalities per year per 1000 square kilometres was calculated using the area (8334km2) of the study area. Mortality events in our dataset include only those that have been reported to Alberta Fish and Wildlife or to the GBP. In a review of mortality data from Alberta, British Columbia, and the western United States, McLellan et al. (1999) estimated that without the mortality information gained from radio-collared bears, management agencies would have been aware of only about 60% of human-caused deaths. To account for unreported mortality events, we increased our estimate of human-caused mortalities by 40% (Festa-Bianchet 2010). To estimate the mortality rate for the Kakwa study area, we used the corrected humancaused mortality number along with a population density estimate. A DNA inventory study in 2008 estimated the population density in the non-protected area of the Grande Cache population unit (including the Kakwa study area) at approximately 16.18 bears per 1000km2 (Alberta Grizzly Bear Inventory Team 2008). Mortality risk surfaces Mortality risk surfaces were calculated using the existing FRIGBP mortality risk model, incorporating datasets updated and improved in 2011. The mortality risk model is a spatial model that describes the relative probability of human-caused grizzly bear mortality as a function of landscape variables, developed in May 2007 by Dr. Scott Nielsen. The model is based on multivariate logistic regression analysis of a sample of 297 anthropogenic grizzly 66

Chapter 2-Response to Oil and Gas Activities bear mortalities within the Central Rockies Ecosystem (CRE), dating from 1971 to 2002. In this model, mortality risk is a function of 9 variables: 1. 2. 3. 4. 5. 6. 7. 8. 9.

cost distance to road (distance to road, incorporating terrain) distance to stream distance to forest edge terrain ruggedness index (TRI) proportion of upland tree upland tree squared proportion of protected area (federal and provincial parks and protected areas) cost distance to trail x proportion protected cost distance to trail x proportion White Zone (agricultural zone)

All nine variables are derived from six base layers: landcover, TRI, roads/trails, White Zone (agricultural land), streams, and protected areas. Risk values are calculated for each pixel, and values range from 0 (no risk) to 10 (high risk). Mortality risk surfaces are currently included in the GBP deliverables provided to program partners each year. These risk surfaces were based on the data available at the time the annual deliverables were published. Datasets obtained for the habitat use and movement analyses in Topics A to C of this report are more spatially and temporally accurate than previously available data. Improvements in spatial accuracy translates to a more accurate depiction of mortality risk on the landscape, and improvements in temporal accuracy means that development features can be included in risk models at the appropriate time. Our objective was to focus specifically on mortality risk in the Kakwa area, using the most detailed and up-to-date datasets possible, including updated landcover data, road layers, and cutblock polygons, in addition to an improved and updated oil and gas feature dataset with additional attribute data. Use of these datasets for the period of our analysis (2004 to 2010) also produces more consistent and directly comparable datasets across years, improving predictability of the risk surfaces and allowing for direct year to year comparisons for this region. Starting with 2004 as the baseline risk surface for the model, we added annual changes to the landscape each year as â&#x20AC;&#x153;newâ&#x20AC;? openings and roads to generate annual mortality risk surfaces for 2005 through 2010. The distance to road and distance to forest edge variables change with each addition of new openings or roads. Adding these annual changes to the baseline year allowed for a direct year to year comparison of the change in mortality risk. The mean mortality risk within each individual annual home range was calculated for comparison of variability in risk across the study area and between sex/age classes. Mean mortality risk values by individual home range were tested for normal distribution using the Kolmogorov-Smirnov and Shapiro-Wilk tests in SPSS, as well as visual inspection of histograms. The majority of datasets were not normally distributed; therefore, nonparametric statistical tests were used for analysis. Mean mortality risk values (by annual home range) were compared between males, females, and subadults using a Kruskal-Wallis test. Correlation between age of individual bears and mean mortality risk values was 67

Chapter 2-Response to Oil and Gas Activities investigated for females and males using a Spearman’s rank-order correlation for nonparametric samples. Mortality risk datasets Roads: Roads in the Kakwa region are the product of oil and gas development as well as forestry activities. The roads dataset was originally obtained from Alberta Sustainable Resource Development, and subsequently updated by our research team through heads-up digitizing using medium and high resolution imagery. An additional detailed update was completed in January 2012 for roads data in the Kakwa study area. Road segments were attributed with a year of disturbance, based on remote sensing imagery from July or August of each year. Roads were presumed to be on the landscape before July 31st of the year of disturbance, and that year was applied to determine when a road segment was added to the mortality risk surfaces. Road segments directly influence the “cost distance to roads” parameter in the mortality risk model, and the coefficient for this parameter is positive, meaning that risk increases with proximity to roads. Forestry cutblocks: Cutblock data are maintained by the FRIGBP. In January 2011, cutblocks were updated for 2009 to 2011 for the Kakwa study area, and using improved available imagery, our GIS team filled in areas that were previously missing for 2005 and 2007. Oil and gas features: Wellsite, pipeline, and facilities data were provided from a database maintained by Alberta Energy; spatial data and attributes were originally obtained directly from oil and gas operators. Wellsites, pipelines, and facilities create new openings and linear features that result in changes in mortality risk model input parameters. Original wellsite data were point data, and conversion to polygons was required to include these developments as openings (areas) on the landscape. The Foothills Research Institute has access to the provincial Digital Integrated Dispositions (DIDs) database, which includes polygons of mineral surface leases (MSLs) and other dispositions. If a wellsite corresponded to a disposition polygon in the DIDs dataset, the associated polygon was used in the analysis as the forest opening. If there was no corresponding MSL data in the DIDS dataset, the wellsites were buffered by 70m, creating an area of approximately 1.5 hectares, equal to the average area of wellsite MSLs in the DIDS dataset. Original pipeline data were line data, and were not buffered, as they are considered as linear features (trails) in the risk model. Addition of pipeline segments influences the “cost distance to trail” parameter in the model. Wellsite data were attributed with a date of initial clearing and drilling, and pipeline data were attributed with a date of construction. Risk models were generated annually, but inclusion of all features added during each year would overestimate the features (and risk) on the landscape during the non-denning (active) period for bears. The mid-date for the non-denning period (averaged across all sex-age classes) was July 28th, and imagery used for updating roads was also from July or August of each year. Based on these factors, July 31st was selected as the cutoff date for developments included in annual risk models, to approximate an average value of developments present, and to keep data consistent with 68

Chapter 2-Response to Oil and Gas Activities roads data. In other words, any wellsites or pipelines added to the landscape after July 31 st were not included in risk calculations for that year. Data for oil and gas facilities were in the form of point data, representing the centroid of the section of land on which they were built, and data were not attributed with dates of construction. Based on the large number (1372) of facilities in our dataset, we investigated methods of including these features in calculation of mortality risk surfaces. However, the low spatial accuracy and lack of temporal data for these features was problematic. On visual inspection using remote imagery, most facilities were also small and/or were contained within the footprints of other oil and gas features (i.e. pipelines and wellsites). Therefore, only large facilities (gas plants, central treatment plants, and waste plants) were included in the analysis. Areas were included as disposition polygons obtained from the DIDS dataset. Landcover: Landcover is a raster-based GIS layer, identifying 10 land cover classes derived from remote sensing imagery (Landsat TM7) at 30m resolution (pixel size). We used a 2004 landcover layer as the baseline for the mortality risk surfaces. This layer was updated for 2004 disturbances using improved imagery available in 2012. Comparison of predicted mortality risk values to reported mortality locations The original mortality risk model was derived and validated using data from the central Rocky Mountains of Alberta. Our dataset was not large enough to statistically validate the model for the Kakwa study area. However, the reported mortalities in our dataset provided an opportunity to investigate whether the risk model accurately predicts where mortalities take place in the Kakwa region. Depending on how a mortality event was reported, location accuracy varied from within 100m to a general location within a township and range section (accuracy ~7000m). We created buffers around mortality locations based on the reported location accuracy, and used the buffer areas to calculate the mean mortality risk value within the buffer area, based on the mortality risk surface generated for the year of each mortality.

Chapter 2-Response to Oil and Gas Activities

Results Within the Kakwa study area, there were 21 reported grizzly bear mortalities during 19992011, corresponding to an average of 1.75 mortalities per year, or 0.21 mortalities per year per 1000km2. Total reported mortalities for each year are included in Figure 13, and mortalities categorized by mortality cause are included in Table 13. The number of annual reported mortalities decreased from 1999 (6 mortalities) to 2010 (1). The most common cause of mortality during 1999-2010 was legal hunting (11 mortalities), followed by illegal hunting (3). However, following suspension of the legal grizzly bear hunt in 2006, the most common causes of mortality were illegal hunting (2) and road kill (2).

Annual reported mortalities

7 6 5 4 3 2 1 0 1999 2000 2003 2005 2006 2007 2008 2009 2010 Year

Figure 13: Reported grizzly bear mortalities by year, Kakwa study area. Table 13: Total reported mortalities in the Kakwa study area from 1999 â&#x20AC;&#x201C; 2011, by mortality cause.

Mortality cause Illegally hunted Legal defense of life or property Legally hunted Research Road kill Unknown Total

Number of mortalities 3 2 11 1 2 2 21

Overall, 19 of the 21 reported mortalities (90%) were human-caused, with the remaining 2 resulting from unknown causes. Applying a 40% correction for unreported mortalities, we estimate human-caused mortalities at 27, approximately 2.2 human-caused mortalities per 70

Chapter 2-Response to Oil and Gas Activities year, or 0.27 mortalities per year per 1000km2. Using this estimate of 0.27 human-caused mortalities per year per 1000km2 and the regional population density estimate of 16.18 bears per 1000km2 (Alberta Grizzly Bear Inventory Team 2008), we estimate human-caused mortality rates in the Kakwa study area at approximately 1.7%.

Predicted mortality risk

During 2004 to 2010, the average mortality risk across the Kakwa study area increased from 6.21 to 6.36, an increase of 2.4% (Figure 14). The largest increases in risk were between 2004 to 2005 and 2005 to 2006. 6.40 6.35 0.84

6.30

0.23

0.13

0.11 0.06

1.01

6.25 6.20 6.15 6.10 2004

2005

2006

2007

2008

2009

2010

Year

Figure 14: Predicted mortality risk in the Kakwa study area by year. Percent increases shown for each year.

A map of mortality risk in the Kakwa area for 2010 is presented in Figure 15.

Figure 15: Mortality risk values in the Kakwa study area for 2010, including reported mortalities for 1999-2011.

Chapter 2-Response to Oil and Gas Activities There was a wide range of mortality risk values between individual home ranges, with mean mortality risk ranging from 2.79 to 8.27. Mortality risk was highest in the northeast region of the study area, and lowest in the southwest, due to differing levels of development in these regions (Figure 15). No significant differences were detected in mean mortality risk between home ranges of adult males, adult females, or subadults. Age of individual bears was not correlated with mean home range mortality risk. Predicted mortality risk values compared to observed mortality locations are included in Table 14. The overall mean predicted mortality risk value at mortality locations was 7.76 (range = 4.77 to 9.32). All mortality locations except for one had predicted risk values of greater than 7 (Figure 14, Table 14). Table 14: Mean predicted mortality risk values at locations of reported mortalities, 2005-2010.

Year 2005 2006 2007 2008 2009 2009 2010

Mortality cause Legally hunted Research Illegally hunted Unknown Road kill Illegally hunted Road kill

Mean risk value 7.14 9.32 7.56 8.20 9.03 8.31 4.77

Discussion Average annual mortalities per 1000 km2 (since 1999) were slightly higher in the Kakwa study area (0.21) compared to the Yellowhead population unit (0.15) and the Grande Cache population unit overall (0.17). The proportion of reported mortalities that were humancaused was very similar to results previously calculated for 1999-2008 for the entire province of Alberta (87% human-caused, FRI unpublished data). In moderate (nonoptimal) habitats, it is estimated that the sustainable human caused mortality rate for a grizzly bear population is approximately 2.8% (McLoughlin 2003). Our estimated mortality rate for 1999 to 2010 was 1.7%, which falls below this value. It appears that current mortality rates are sustainable by the grizzly bear population in the Kakwa area, but future increases could make mortality rates unsustainable for this population. The mean predicted mortality risk within the entire study area was moderate. Mortality risk steadily increased throughout the study period, although at a relatively low rate. The most pronounced increase in mortality risk occurred between 2004/2005 and 2005/2006, which corresponds with the largest number of wellsites, pipelines and roads added to the landscape in a single year during our study period (see Appendix A). There were no significant differences in risk within home ranges across sex/age classes, indicating that 72

Chapter 2-Response to Oil and Gas Activities different sex-age classes are not likely selecting areas with different risk. There were no indications of patterns of survival related to predicted mortality risk, as age was not correlated with a lower risk value. The mean predicted mortality risk ranged widely between home ranges. Bears living in the southwestern portion of the study range are subject to much lower development and correspondingly lower mortality risk as compared to bears in the northeastern range of the study area. As development increases, mortality risk will increase further, and without mitigation measures, it is likely that mortality rates could increase in these relatively undeveloped areas. Comparison of reported mortality locations with predicted mortality risk values indicates that the current model is a good predictor of mortality risk for the Kakwa study area. Compared to the possible values of 0 to 10, all mortality locations but one had relatively high (>7) predicted mortality risk values. These results indicate that applying current models to predict mortality risk in the Kakwa is a useful tool for grizzly bear management in this area. Literature cited Alberta Grizzly Bear Inventory Team 2009. Grizzly bear population and density estimates for the 2008 DNA inventory of the Grande Cache Bear Management Area (BMA 2). Report prepared for Alberta Sustainable Resource Development, Fish and Wildlife Division. 39 pp. Benn, B. 1998. Grizzly bear mortality in the central rockies ecosystem, Canada. Eastern Slopes Grizzly Bear Project 1, 151. Benn, B. & Herrero, S. 2002. Grizzly bear mortality and human access in Banff and Yoho National Parks, 1971-98. Ursus 13, 213-221. Festa-Bianchet, M. 2010. Status of the grizzly bear (Ursus arctos) in Alberta. Alberta Sustainable Resource Development, Alberta Conservation Association, Alberta Wildlife Status Report No.37 (Upadate 2010) 43pp. McLellan, B. N., Hovey, F. W., Mace, R. D., Woods, J. D., Carney, D. W., Gibeau, M. L., Wakkinen, W. L., & Kasworm, W. F. 1999. Rates and causes of grizzly bear mortlity in the interior mountains of British Columbia, Alberta, Montana, Washinton, and Idaho. Journal of Wildlife Management 63, 911-920. McLoughlin, C. A. 2003, University of Alberta, Edmonton. Nielsen, S. E., Boyce, M. S., & Stenhouse, G. B. 2004a. Grizzly bears and forestry I. Selection of clearcuts by grizzly bears in west-central Alberta, Canada. Forest Ecology and Management 199, 51â&#x20AC;&#x201C;65. Nielsen, S. E., Herrero, S., Boyce, M. S., Mace, R. D., Benn, B., Gibeau, M. L., & Jevons, S. 2004b. Modelling the spatial distribution of human-caused grizzly bear mortalities in the Central Rockies ecosystem of Canada. Biological Conservation 120, 101-113.

Chapter 2-Response to Oil and Gas Activities

CONCLUSIONS AND RESEARCH APPLICATIONS Gordon Stenhouse Alberta is the largest producer of oil and natural gas in Canada. The landscapes where these resources are found are also home to a threatened population of grizzly bears, a species with both cultural and ecological significance. Our research addresses the current knowledge gap regarding how grizzly bears respond to oil and gas operations. Results and management recommendations from this project in the Kakwa region of west central Alberta may be applied in similar habitats in Alberta, and in other areas of oil and gas development in North America to enhance and improve land and resource management in grizzly bear habitat. Our results show that grizzly bears used habitat containing oil and gas activities (wellsites, roads, and pipelines) differently than what would be expected from habitat values (RSFs). Differences in habitat use with oil and gas disturbance features varied spatially (by season and by sex/age class), and temporally (day versus night). Bears in our study generally did not avoid habitat containing oil and gas features during spring. This may be due to lower levels of human activity during this time, as the least amount of drilling takes place in the spring. During summer, males used habitat with oil and gas development less than expected, and fall was the season with the most changes in large scale habitat use in response to oil and gas features, with all bears avoiding habitat containing active wellsites. Our results also suggested a temporal shift of habitat use by adult females in fall, with less use of habitat containing oil and gas features during the day compared to nighttime. Patterns of habitat use vary; however, it appears that there may be some displacement of male bears from habitat containing oil and gas features during the summer. At a smaller spatial scale, our results show that the majority of bears selected for wellpads, and most bears did not avoid the 500m zone surrounding wellsites. Therefore, bears continued to use habitat surrounding wellpads, and were not displaced. Bears were likely attracted to the abundance of bear foods growing on wellpads and along forest edges surrounding the wellsites; these bear foods are more abundant on older wells (>10 years since initial clearing). The results of our analysis of grizzly bear habitat use at and around oil and gas wells provides strong evidence that future cumulative effects assessments of these types of industrial activities should not use a habitat withdrawal approach to remove â&#x20AC;&#x153;available habitatâ&#x20AC;? adjacent to current or proposed well sites. There were no obvious increases in grizzly bear movement rates in response to wellsites in our analysis. Our results indicate faster movement when roads or pipelines were present, indicating either a response to disturbance (displacement) or possible use of these linear features as movement corridors. At the spatial scale of the analysis (linear features present/absent within 500m), it was not possible to determine whether the bears were on the linear features themselves. The FRIGBP has received funding from the AUPRF to initiate a study in 2012 involving an investigation of grizzly bear use of pipelines as 74

Chapter 2-Response to Oil and Gas Activities movement corridors, and comparison to movement patterns along roads. Results from the 2012 research will specifically investigate movement along these linear features, and will build on current results to provide information for management and mitigation of oil and gas development in grizzly bear habitat. Human-caused mortalities were similar to those observed in other areas of Alberta. Our estimated mortality rate for 1999 to 2010 was 1.7%, below the estimated sustainable human caused mortality rate for grizzly bear populations in moderate habitats (2.8%). The mean predicted mortality risk within the entire study area was moderate, and mortality risk steadily increased throughout the study period. Comparison of reported mortality locations with predicted mortality risk values indicates that the current model is a good predictor of mortality risk for the Kakwa study area. While clearings such as wellpads may provide new foraging habitat or movement corridors for grizzly bears, bear use of human features such as wellsites also increases the potential for bear-human interactions. It appears that bears are using wellsites for access to food, but it is unknown if wellpads represent an important source of foraging habitat within our study area. Grizzly bear mortality risk is known to increase near human features; therefore, wellsites may result in an increased risk of human-caused mortality for bears foraging on the wellpads. As development in the Kakwa region increases, mortality risk will further increase. Increases in grizzly bear mortality will have a negative impact on provincial recovery efforts.

Research Applications and Future Direction The primary limiting factor for grizzly bears in Alberta is human-caused mortality, and reduction in human-caused mortality is included in the Alberta Grizzly Bear Recovery Plan (Alberta Grizzly Bear Recovery Team 2008). Mortality risk is directly related to specific characteristics of industrial developments, such as linear access and the accompanying increase in human activity, as well as the â&#x20AC;&#x153;sightabilityâ&#x20AC;? of bears in the clearings created by human use features. These characteristics provide opportunities for mitigation. Mitigation measures that reduce linear access to grizzly bear habitat and/or reduce sightability of grizzly bears near openings created by development activities in this region would reduce the probability of human-caused grizzly bear mortalities. The FRIGBP is currently developing a GIS-based application that will allow resource managers to select particular linear features within a specific area for decommission to maximize the benefits for grizzly bears with the lowest kilometers of roads reclaimed. The FRIGBP is committed to presenting results from this study at a stakeholder meeting with oil and gas operators in 2012, in order to exchange information and to discuss additional on-the-ground mitigation opportunities that could be applied to oil and gas operations in the Kakwa region. Dialogue with oil and gas operators will provide an opportunity to apply the information gained in this study in combination with industry expertise, in order to develop practical applications in daily operations. A copy of this report has been sent to the FRIGBP oil and gas advisor for this project (Amit Saxena, Devon 75

Chapter 2-Response to Oil and Gas Activities Canada), with a request for management recommendations. Our research team members are also available to present this research at an AUPRF or PTAC forum in 2012, and we will submit an annual AUPRF newsletter article for consideration. Any recommendations that result from workshops, presentations, or communications with oil and gas operators may be distributed to the AUPRF as an addendum to this report. We will also provide AUPRF with our 2012 FRI Grizzly Bear Program annual report and program deliverables, available in July 2012. Our research team will strive to present the results of this work at international scientific symposiums and conferences within the next year (International Conference on Bear Research and Management, Alberta Wildlife Society Convention).The support of the AUPRF will be recognized in all papers, posters and presentations related to this research. We intend to publish at least two peer reviewed journal publications from this research, which may include: 1) a publication on large scale grizzly bear habitat use in response to oil and gas developments and activities in Alberta; 2) a publication on small scale (wellsite) grizzly bear habitat use in regions of oil and gas developments and activities in Alberta; and 3) a publication on grizzly bear movement patterns in response to oil and gas developments and activities in Alberta. These papers will be submitted for journal review within 12 months of submission of this report, and the AUPRF will receive copies of both journal submissions and final publications. Results from this project serve as baseline information to fill the knowledge gap regarding oil and gas development and grizzly bears. This research can be directly applied to oil and gas operations in grizzly bear habitat. Future studies will continue to address this knowledge gap, and provide more information to better manage development to assist recovery of this threatened species.

Acknowledgements This project would not have been possible without the assistance and support of a number of people. Many thanks to Tom Churchill of Alberta Energy for supplying the oil and gas datasets, and for his endless patience and assistance with the data. Thanks also to Karen Graham for assistance with fieldwork, and for her valuable advice and consultation throughout this project. Appreciation is also extended to Darren Wiens, Julie Duval, and Jerome Cranston for their GIS and data management expertise.

Chapter 2-Response to Oil and Gas Activities

Appendix: Oil and gas features: new features added per year, total numbers, and densities. Wellsite density in the Kakwa study area, and new wellsites added per year. New Total Density Year wells number (wellsites/km2) 2004 baseline 2101 0.25 2005 336 2437 0.29 2006 309 2746 0.33 2007 165 2911 0.35 2008 197 3108 0.37 2009 167 3275 0.39 2010 199 3474 0.42 2011 122 3596 0.43 Density of pipelines in the Kakwa study area, and kilometres of new pipelines per year. New Total Density Year pipelines length (km (km) (km) pipe/km2) 2004 baseline 3538 0.42 2005 756 4294 0.52 2006 681 4975 0.60 2007 504 5480 0.66 2008 364 5844 0.70 2009 273 6116 0.73 2010 171 6288 0.75 Density of roads in the Kakwa study area, and kilometres of new roads per year. Note: new data were not available for 2010. New Total roads length Density (km Year (km) (km) roads/km2) baseline 4590 2004 0.551 2005 74 4664 0.560 2006 523 5187 0.622 2007 77 5264 0.632 2008 71 5335 0.640 2009 8 5343 0.641

Chapter 2-Response to Oil and Gas Activities

Frequency of drilling operations through the year. Seasons indicated correspond with grizzly bear foraging seasons for the study area.

Chapter 3-Den Site Selection

CHAPTER 3: DEN SITE SELECTION FOR GRIZZLY BEARS IN THE ROCKY MOUNTAINS AND FOOTHILLS OF ALBERTA, CANADA: A SPATIALLY EXPLICIT AND HIERARCHICAL APPROACH Final Report Prepared for the Alberta Ecotrust Pigeon, Karine1, Scott Nielsen3, Etienne Cardinal2, Gordon Stenhouse2, and Steve D. Côté1, 1

Département de biologie and Centre d’études nordiques, Université Laval, 1045 Av. de la Médecine, Québec, Québec, Canada, G1V 0A6 2 Foothills Research Institute Grizzly Bear Program (FRIGBP), Box 6330, Hinton, Alberta, Canada, T7V 1X6 3 University of Alberta Renewable Resources

Introduction As of 2010, grizzly bear (Ursus arctos) populations are considered to be a threatened species in Alberta, Canada (Alberta Sustainable Resource Development, 2010). Grizzly bears, once a common species across the Canadian prairies, have been confined to the western part of the country since the early 1900’s because of colonization pressures (McLellan 1998; Weaver et al. 1996). In 2002, the Alberta Grizzly Bear Recovery Team recommended that the species be listed as threatened because of small population sizes, low reproductive rates, low immigration from adjacent populations, and high levels of human-caused mortality in the province (Alberta Grizzly Bear Recovery Plan, 2008). In recent years, the ongoing growth of industrial activities resulting in more roads, cut blocks, well sites, and human-use of the landscape has amplified the potential for human-bear conflicts in the province. Unlike other large mammals under management in Alberta, grizzly bears hibernate each winter. Hibernation is commonly viewed as an adaptive strategy to avoid harsh environmental conditions and the associated lack of food and water resources (Bartels et al. 1998; Folk et al. 1980; Geiser and Ruf 1995; Harding 1976) By entering torpor and ceasing all but necessary metabolic activities, hibernators are able to greatly reduce their energetic requirements thereby effectively coping with long-term suboptimal environmental conditions (Bartels et al. 1998; Geiser and Ruf 1995; Munro et al. 2005; Thomas et al. 1990). Unlike most hibernators, bears do not show the typical periodic state 79

Chapter 3-Den Site Selection of torpor defined by a pronounced reduction in responsive capacity and metabolic rate (Folk et al. 1980). Even though bears reduce their metabolic rate from 35 to 75% during hibernation (Watts and Jonkel 1988), their body temperature is only slightly lowered and they can easily be aroused while in dens (Graber 1990; Linnell et al. 2000; Tietje and Ruff 1980). Because of this, the denning period is a sensitive period during which disturbances can have significant fitness costs (Linnell et al. 2000; Swenson et al. 1997). Even though bears hibernate in order to reduce their energy expenditure, Farley and Robbins (1995) observed that during hibernation, lactating female grizzlies lost nearly twice as much weight as non-lactating females. These large energy costs imply severe consequences from potential winter disturbances and den abandonment. Winter is a critical period during which bears have to cope with unfavorable environmental conditions and the choice or availability of an appropriate den site may be a crucial factor in the survival and reproductive success of hibernating individuals. This is especially true of females who give birth and begin lactation. In a study looking at 194 denning events, Swenson et al. (1997) found that pregnant females that were disturbed while denning had an increase in cub mortality when compared to females that were not disturbed. Several instances of winter den emergence have been documented with human disturbance being the primary cause of den abandonment (e.g. Tietje and Ruff, 1980; Swenson et al., 1997; Linnell et al., 2000). Currently, land and forest management planning in grizzly bear habitat in Alberta does not include any mitigation measures to minimize possible impacts on denning habitat. We believe this is due to our current inability to adequately understand and predict where grizzly bears may den in boreal forest environments. For hibernators and other species utilizing den sites, the selection of adequate den locations can occur at different spatial scales depending on individual requirements for reproduction, survival, or other activities (Henner et al. 2004; Squires et al. 2008). Dens may be concentrated in specific areas limited by factors such as landscape-level human disturbances, habitat patch size, food availability, microclimate, and the physical characteristics of the den such as soil stability and insulation. Looking at the selection of den sites by Canada lynx (Lynx canadensis), Squires et al. (2008) determined that Canada lynx in western Montana, USA, selected den sites by a hierarchical process involving at least three different spatial scales. At the den site level (11-m radius circle around the den) individuals selected higher horizontal cover and log volume than expected, while at the den area scale (100-m radius) larger diameter trees were selected, and finally at the den environment scale (1-km radius) individuals selected drainage-like areas, farther from forest edges than randomly expected. Henner et al. (2004) also observed that raccoons selected different landscape or site attributes as a function of spatial scales while establishing their dens. These findings affirm the importance of considering how a multitude of requirements ranging across several spatial scales can affect the selection of appropriate den sites. In order to accurately manage species utilizing dens, efforts should therefore be made to understand den site selection across several spatial scales. Several factors have been shown to be important in the selection of den sites by grizzly bears. A number of studies (e.g. Vroom 1980; McLoughlin et al. 2002) located grizzly bear 80

Chapter 3-Den Site Selection dens on leeward slopes within their respective study areas in Alberta and the Northwest Territories. This suggests that at the den-site scale, bears may select for increased den insulation because snow accumulates on leeward slopes. Baldwin and Bender (2010) also found that black bears denning in Rocky Mountain National Park, Colorado, USA, chose site characteristics consistent with greater snow depth. In a study conducted in central-eastern British-Columbia, Canada, Ciarniello et al. (2005) found that at the landscape level, grizzly bears selected den sites at mid-to-upper elevations, in older-aged forest stands located further from roads than randomly expected. Goldstein et al. (2010) looked at landscapelevel selection of den sites in the Kenai Peninsula, Alaska, and found that bears selected for sites away from human disturbances while selecting for steep slopes and high elevations. These findings suggest that a number of spatial scales are likely important in the den site selection process for grizzly bears and that different requirements may act at various spatial scales. Because a major portion of industrial activities occurring within Albertaâ&#x20AC;&#x2122;s forest take place during the winter, understanding grizzly bear denning behavior and den site requirements in the forested landscape will help minimize potential den site disturbances and reduce inherent human-bear conflicts. Our objective was to assess the relative probability of grizzly bear den site occurrence in the new grizzly bear conservation zones for two population units within Alberta (Alberta Sustainable Resource Development, 2010). The results of this research will delineate prefered den site areas within the conservation zones and could be used as a conservation and management tool for grizzly bear management conservation and recovery in the area. We quantify the importance of human disturbance, topography, and habitat including spring and fall food availability on a hierarchical den site selection process by grizzly bears in our study area. We hypothesized that bears would select sites (1) on steep, high elevation, leeward slopes, (2) with a greater than expected availability of spring and fall foods, and (3) in undisturbed contiguous conifer stands.

Methods Study Area We studied grizzly bear den site selection from individual adult grizzly bears of both genders captured and collared between 1999 and 2010 in the core conservation areas of the Grande-Cache and Yellowhead provincial grizzly bear population units (Bear Management Areas (BMA); Alberta Sustainable Resource Development, 2010; Figure 1). Both population units include portions of protected and unprotected mountainous terrain as well as rolling foothills extensively altered by anthropogenic activities mainly associated with forestry, coal mining, and oil and gas exploration. The Grande-Cache population unit (GCPU) extends west from the British-Columbia / Alberta border of the Kakwa Wildland Provincial Park towards Highway 40 North (54Â° 30' N / 119Â° 6' W). Altitude varies from 543m to 2440m asl. with upper and lower foothills dominating the landscape. This area is primarily composed of lodgepole pine (Pinus contorta), black spruce (Picea mariana), and white spruce (Picea glauca), while balsam fir (Abies balsamea), balsam poplar (Populus balsamifera) and trembling aspen (Populus tremuloides) occur less frequently. Forestry and 81

Chapter 3-Den Site Selection oil and gas exploration are the main anthropogenic activities in the area. The Yellowhead population unit (YPU) extends west from the Jasper National Park boundary to the towns of Cadomin and Hinton, and also includes Whitehorse Wildland Provincial Park (52째 48' N / 117째 7' W). Altitude varies from 712m to 3680m asl. and is dominated by mountainous terrain. Dominant tree species are lodgepole pine (Pinus contorta), subalpine fir (Abies lasiocarpa), white spruce (Picea glauca), and Englemann Spruce (Picea englemanni) at the higher elevations. Coal mining and forest harvesting are the dominant anthropogenic activities in this area. These areas also experience a significant amount of human use related to recreational activities. Home-range scale den site selection We investigated the selection of den sites at the home-range scale for 17 male and 42 female grizzly bears 4-years or older using GIS-derived spatial variables. Because we were interested in mapping potential grizzly bear den sites at the population level and because subdividing individuals would diminish the statistical power of our analyses, we did not consider potential differences among sex and reproductive status. We compared individual den sites to 150 random locations within the fall home range of respective individuals. We chose to restrict random locations to areas with a minimum slope of 7째 because it is well known that bears do not den on flat ground (e.g. Baldwin and Bender 2008; Elfstrom et al. 2008; Goldstein et al. 2010; Vroom et al. 1980), and because the minimum GIS-derived slope associated with a den site for our dataset was 7째. We tested the sensitivity of coefficients to different sampling intensities by varying the number of random locations associated with each den event for a subset of independent variables using the GCPU dataset. Elevation, slope, density of moderately to heavily used roads and truck trails, and well site densities were used to investigate variation in coefficients for categories of random locations ranging from 25 to 500. We chose 150 random locations for all subsequent analyses because (1) the coefficients did not vary according to the number of random locations (Table 1) and (2) using more than 150 locations oversampled the available area.

Chapter 3-Den Site Selection Table 1. Coefficients from sampling intensity analyse using conditional logistic regressions for road, truck trail, and well site densities at 25, 50, 100, 150, 250, and 500 random locations for 27 dens in the Grande-Cache management unit area.

Grizzly bear den location data From 1999 to 2010, we captured and collared individuals using a combination of techniques involving aerial darting, leg hold snares, and culvert traps. See Cattet et al. (2003) for a full description of techniques. Captured individuals were fitted with ATS (Advance Telemetry Systems) or Televilt Global Positioning System (GPS) collars. We located dens using data acquired from GPS-collars during the denning period. We travelled to and physically marked the location of 49 dens out of the 79 used in this analysis with a hand-held GPS unit and used the centroid (geographic center) of the GPS locations acquired during the denning period to define the location of the remaining 30 den sites. Predictor variables We developed models (maps) describing the relative probability of occurrence of den sites based on hypotheses from four categories of landscape spatial attributes. Categories are referred to as (1) topography, (2) food, (3) habitat, and (4) anthropogenic. Variables are described in Table 2. Diet-based habitat productivity values were extracted from spring and fall food-based habitat models from Nielsen et al. (2010) and represent proportions of high-ranked seasonally important foods predicted at a site based on knowledge of seasonal diets and species distribution models for individual bear foods (Nielsen et al. 2010). We used the periods from May 01 â&#x20AC;&#x201C; May 14 (spring) and Sept 21 - Sept 31 to represent spring and fall food-based habitats respectively. Slope, hillshade and compound topographic index (CTI) variables were derived from a 30m-scale digital elevation model (DEM). A Snow Load Index (SLI) was developed to account for the influence of predominant winds and 83

Chapter 3-Den Site Selection elevation on the snow loading potential of slopes in the study area. Because predominant winds are generally from the west in the area, we modified and standardized between 0 and 1 a hillshade model that represented maximum values (i.e., 1) for East-facing slopes, minimum values (i.e., 0) for West-facing slopes, and intermediate values occurring between East and West slopes. We then standardized a DEM between 0 and 1 based on elevation (0 for low elevation and 1 for high elevation) and used the product of the standardized elevation and east hillshading models to represent our SLI index. East-facing, high elevation slopes therefore represented a maximum SLI index, while West-facing, low elevation slopes represented minimum SLI index values. For the GCPU, we removed well sites that had been reclaimed or abandoned (non-active) prior to the year of denning and calculated well site densities using only active sites but because of data limitations, we calculated well site densities for the YPU using data that potentially includes both active and inactive well sites. Table 2. Spatial attributes used to assess den site selection at the home-range scale in the Grande-Cache and Yellowhead population units, Alberta, Canada. Km2 (Km2).

Scales of den site selection As we had little previous knowledge of the scales at which bears might demonstrate a response to various degrees of human disturbance while in dens, we tested the sensitivity of the response variable (presence of den sites) to roads and truck trails, and active well site densities for sampling distances ranging from 0.15-km to 2-km using the GCPU dataset. Roads were categorized as either primary (heavy-use) or secondary (moderate use) roads, or truck trails (low use) and sampling distances were tested on all roads (PSTRDD), primary and secondary roads only (PSRDD), and truck trails only (TRDD; Table 2). We chose to 84

Chapter 3-Den Site Selection conduct all subsequent analyses at a 0.3-km scale because (1) coefficients obtained from road and truck trail densities at all scales were negative and significant and the 0.3-km scale had the lowest AIC score, and (2) for active well site densities, the only significant scale was the 0.3-km scale which also had the lowest AIC score (Table 4). Unlike road and well site densities which can be used as a measure of potential human disturbance while in dens, the selection of habitat-based variables reflect a selection process that does not relate to disturbances that occur during the denning period. Because of this, we extracted spring and fall diet-based habitat productivity values (Nielsen et al. 2010) and proportion of mature conifer stands at four biologically relevant spatial scales. The scales we chose refer to the mean daily distance traveled by female grizzly bears in our study area (6.8-km), the average daily linear distance travelled by female grizzly bears in the flathead valley, BC (Apps et al. 2004), the 0.3-km human-disturbance scale detected from our road and well site scale analyses, and a 1-km scale used as an intermediate between the behaviour-based and human-disturbance scales mentioned above. We again chose to conduct subsequent analyses at the 0.3-km scale because (1) coefficients obtained from spring diet-based habitat productivity values were significant at all scales but the 0.3-km scale had the lowest AIC score, (2) fall diet-based habitat productivity values were significant at the 0.3-km and 1-km scale with the lowest AIC score again associated with the 0.3-km scale, (3) the proportion of conifer habitat was also only significant at the 0.3-km scale and that scale had the lowest AIC score (Table 4). Statistical Analyses: We used a two-step information theoretic model selection approach with multiple working hypotheses based on the Akaike Information Criterion for small sample sizes (AICC) to obtain relative probabilities of den occurrence within each population unit. We first tested a suite of candidate models within the four categories of landscape attributes (topography, food, habitat, and anthropogenic). We used a conditional logistic regression approach (PROC PHREG, SAS 9.3, SAS Institute Inc., 2011) and developed candidate models based on biologically plausible hypotheses of factors potentially driving den site selection. Ranking these models according to their delta AICC (ΔAICC) values and Akaike weights (ωί), we used the best models obtained from each of the four categories, combined them with additional models built from mixing variables across categories and performed model selection on this new set of candidate models. We again ranked the final set of candidate models based on delta AICC and Akaike weights to compare model performance (Burnham and Anderson, 2002). The model with the lowest delta AICC value and the highest Akaike weight best described the probability of den occurrence within each population unit.

Chapter 3-Den Site Selection

Results Home-range scale den site selection Scales of den site selection - Sampling scales showed that no bear in the GCPU denned within 0.15-km of any roads or truck trails (Table 4). While no dens were situated within 0.3-km of any moderate to heavy-use road, some individuals dug dens between 0.15 to 0.3-km of truck trails. Bears in this population unit also avoided areas within 0.3-km of well sites. Any level of well site density beyond this 0.3-km area did not influence the presence of dens (Table 4). The 0.3-km scale was also the best scale amongst models of proportions of conifer stands and spring and fall food-based habitat productivity index (Table 3.) Table 3. Sampling scale coefficients and AIC values of conditional logistic regressions for proportions of conifer stands and high-ranked diet-based habitat productivity index values for spring (May 01 â&#x20AC;&#x201C; May 14) and fall (Sept21 - Sept food-31) at 0.03-km, 1-km, 2.4-km, and 6.8-km scales for dens in the Grande-Cache population unit (n = 27). Best AIC models for each category are in bold.

Chapter 3-Den Site Selection Table 4. Coefficients and AIC values from scale sampling analysis using conditional logistic regressions for road, truck trail, and well site densities at 0.15-km (_015), 0.3-km (_03), 0.5-km (_05), 1-km (_1), and 2-km (_2) scales for dens in the Grande-Cache population unit (n = 27). PSTRDD; Primary, secondary road, and truck trail density, PSRDD; Primary and secondary road density, TRDD; Truck trail density, WD; Well site density. Best AIC models for each category are in bold.

Univariate home-range scale den selection models - We investigated den site selection for 27 (5 males and 22 females) and 52 (12 males and 40 females) individuals in the GCPU and YPU respectively for variables (models) within each of the four hypothesized factors. Dens in the GCPU were predominantly in the foothills while dens in the YPU were found in both foothills and mountainous terrain. To our knowledge, no den was re-used during the study period. The weights of evidence for models within the four categories of landscape attributes were generally much higher for the first or top two-ranked models in both population units with the exception of anthropogenic models in the YPU (Tables 5 and 7). The selection of landscape attributes was similar between the two population units. The main differences were the importance of the fall habitat productivity index, elevation, and well site densities for the YPU (Tables 5 and 7). Compound topographic index had a greater importance in the GCPU than in the YPU; the evidence ratios between the best model and 87

Chapter 3-Den Site Selection the model including CTI for both population units were 1.3 and 3.5 respectively. Parameters chosen for the final set of candidate models are shown in Table 6 for the GCPU and Table 8 for the YPU.

Table 5. Model selection for food, topography, habitat, and anthropogenic variables on the selection of den sites in the Grande-Cache population unit. Abbreviations: k, number of parameters; AICc, Aikake Information Criterion adjusted for small sample size; Ď&#x2030;, Akaike weight; â&#x2C6;&#x2020; AICc, difference in AICc amongst models. Best models are in Bold.

Chapter 3-Den Site Selection Table 6. Parameter estimates and standard errors for the top competing models in each of the four categories of landscape spatial attributes for the Grande-Cache population unit. These models were included in the 2nd-level selection process. Abbreviations: Î˛, coefficient.

Table 7. Model selection for food, topography, habitat, and anthropogenic variables on the selection of den sites in the Yellowhead population unit. Abbreviations: k, number of parameters; AICc, Aikake Information Criterion adjusted for small sample size; Ď&#x2030;, Akaike weight; â&#x2C6;&#x2020; AICc, difference in AICc amongst models. Best models are in Bold.

Chapter 3-Den Site Selection

Table 8. Parameter estimates and standard errors for the top competing models in each of the four categories of landscape spatial attributes for the Yellowhead population unit. These models were included in the 2nd-level selection process. Abbreviations: Î˛, coefficient.

Multivariate home-range scale den site selection model - Of the 79 dens investigated, 63 were located in conifer forest while 3 were located in mixed forest, 10 in herbaceous/alpine and 4 in shrub habitat. Overall, bears selected den sites on steep slopes, away from roads and rivers and in areas with a greater percentage of conifer than randomly expected (Table 11). For the GCPU, most of the model support (Ď&#x2030;i= 93%) was among four models (Table 9) with weights ranging from 40% to 11% . These four models included slope, road and truck trail densities and distances, and proportion of conifer as covariates. For the YPU, the most supported model had a weight of 80% and was superior to any other models (Table 10).

Chapter 3-Den Site Selection

Table 9. Final model selection from best models of food, topography, habitat, and anthropogenic categories and mix hypotheses for the selection of den sites in the Grande-Cache population unit. Abbreviations: k, number of parameters; AICc, Aikake Information Criterion adjusted for small sample size; ω, Akaike weight; ∆ AICc, difference in AICc amongst models. Best models are in Bold.

Table 10. Final model selection from best models of food, topography, habitat, and anthropogenic categories and mix hypotheses for the selection of den sites in the Yellowhead population unit. Abbreviations: k, number of parameters; AICc, Aikake Information Criterion adjusted for small sample size; ω, Akaike weight; ∆ AICc, difference in AICc amongst models. Best model is in Bold.

Chapter 3-Den Site Selection Table 11. Parameter estimates and standard errors for the best models of the 2nd-level selection process in the Grande-Cache and Yellowhead population units. Abbreviations: Î˛, coefficient.

Relative probability of den site model (map)- Using coefficients from the most supported multivariate models of den site selection; we predicted the relative probability of den site occurrence (i.e., a resource selection function, RSF) for both population units (Figures 6 and 11). We also built RSFs from the most supported univariate models within each of the four landscape attribute categories (GCPU; Figures 2 to 5 and YPU; 7 to 10). Because harvest activities and the construction of roads and well sites occur at scales much smaller than the scale of grizzly bear population units, we also provide examples of RSFs for single watersheds in the GCPU and YPU (Figure 12 and 13).

Discussion As we hypothesised, bears selected den sites in conifer stands with a greater than expected availability of spring foods, on steep slopes and away from anthropogenic factors. Bears that had access to higher elevation sites also preferred dens on high slopes. Our snow loading model did not however, influence den site selection for bears in either population units. Disturbance scale Considering the pervasive nature of anthropogenic features on the landscape, there is surprisingly little specific distances or densities reported as threshold values from which anthropogenic effects limit ecological and biological processes (Leu et al. 2008). Most studies looking at the effects of anthropogenic activities on wildlife behaviour report a reduction in space use and utilization within proximity to human features but threshold values are rarely investigated (see Forman and Alexander 1998 for exceptions).

Chapter 3-Den Site Selection Still, in a literature review of five publications on European and North American brown bear denning behaviour, Linnell et al. (2000) found that bears selected dens at mean distances of 0.5 km- to 1.7-km from roads and 1.4 km- to 3.5-km from habitations. In a study looking at den abandonment caused by human disturbance, Swenson et al. (1997) suggested that human activities be avoided in an area of 0.1-km to 1-km surrounding den sites. Petram et al. (2004) looked at brown bear den site selection in Solvenia and found that no caves suitable for denning were used within 0.5-km of villages. Our results clearly indicate that bears avoid denning within 0.3-km of roads and well sites (Table 4). During the active period, a number of studies suggest that grizzly bears avoid areas around moderately to heavily-used roads at distances ranging from 0.1-km to 3-km (Apps et al. 2004; Mace et al. 1996; McLellan and Shackleton 1988 etc.). In west-central Alberta, however, bears have been found to select habitats closer to roads because of associated habitats and food attractants (Graham et al. 2010; Roever et al. 2008a; Roever et al 2008b; Roever et al. 2010). Schwartz et al. (2010) defined secure habitats in the Greater Yellowstone Ecosystem as any area of at least 10 acres located at a distance greater than 0.5-km from any road. They used 0.25-km2 and 2.6-km2 moving window scales to calculate road densities. Considering that even within Banff National Park, 91% of known adult grizzly bear mortalities recorded between 1971 and 1998 occurred within 0.5-km of roads or 0.2-km of high-use trails (Benn and Herrero 2002; Nielsen et al. 2004), the importance of anthropogenic linear features on the survival of grizzly bears should not be underestimated. Data Limitations Gathering the data needed to understand and evaluate grizzly bear den site selection is a challenging and long term effort. Because most of the bears in the GCPU were in the foothills, we believe that the RSF scores for this bear management area may overemphasize the importance of the mountains as suitable denning habitats in the region. In the foothills, roads are the main factors contributing to low RSF scores (Figure 5 and 6) and because no roads were present within the mountains, most mountainous habitats were assigned high RSF scores in the GCPU. Pooling data from both population units to generate a single RSF model would average the selection between mountains and foothills and reduce the over-emphasized importance of mountains in the GCPU. Well site data that includes yearly status updates (active vs. non-active wells) would also improve the strength of the analysis in the YPU. Management implications and recommendations Our findings provide data that can be used to develop guidelines to minimize human-bear interactions and the potential impacts of land-use activities on occupied and potential grizzly bear denning habitat. The integration of key landscape attributes in a GIS-based model can assist land managers in making sound management decisions that includes the recognition and preservation of grizzly bear denning habitats. Based on the results presented in this report and our experience investigating grizzly bear den sites over an 1193

Chapter 3-Den Site Selection year period we offer the following management recommendations along the eastern slopes of Alberta: 1. We recommend minimizing and/or reducing the construction of high traffic volume roads and well sites in areas of high relative probability of den site occurrence within core conservation units of both the Yellowhead and Grande Cache grizzly bear population units. 2. When road removal or reclamation is occurring in core and secondary grizzly bear conservation areas, the impacts on denning habitat should be factored into management actions. 3. If forest harvesting, road or well site construction is planned to occur during the winter in high RSF denning habitat, pre-construction surveys should be conducted by qualified staff to determine if active denning is occurring in the planned construction area. These surveys should be conducted in late November following a recent snow fall. 4. Since grizzly bears select mature conifer stands to dig dens, we also recommend limiting the area of cutblocks within high quality denning habitats in core grizzly bear conservation areas. From an operational perspective, forest harvesting designs could also take into account the identified landscape characteristics needed for grizzly bear denning. 5. Based on the results of this research and the analysis presented in this report, a new GIS based application should be developed by the FRI Grizzly Bear Program and become part of the GB Tools package provided by this research team. This new Den Habitat calculator would allow the calculation of planned development activities on predicted grizzly bear denning habitat. 6. A provincial database should be established to collect den site information on all confirmed bear dens in grizzly bear conservation areas in Alberta. This data will be important to continue the development of new predictive models for denning habitat in all conservation units.

Acknowledgements We thank Alberta Ecotrust, Alberta Student Temporary Employment Program, Alberta Sustainable Resource Development, Alberta Conservation Association (ACA), Canada Summer Jobs, Conoco Phillips, Devon Energy, Encana Corporation, Husky Energy, Weyerhaeuser Company, and Y2Y Sarah Baker Memorial fund for the research funds that made this project possible. We also thank L’Université Laval, le Centre d’Étude Nordique, and NSERC scholarships for support provided to the senior author. We thank J. Duval, D. Weins, and J. Cranston for GIS support, Denis Talbot and Anne-Sophie Julien for help with statistical analyses, R. Théorêt-Gosselin, R. Strong, A. Auger, E. Rogers, C. Curle, T. Larsen, P. Stenhouse, and A. Stenhouse for collecting field data, and John Saunders and Steve Wotton at Peregrine helicopters. K. Graham provided helpful comments and suggestions that improved the manuscript. 94

Chapter 3-Den Site Selection

References Alberta Grizzly Bear Recovery Plan 2008-2013. 2008. Alberta Sustainable Resource Development, Fish and Wildlife Division, Alberta Species at Risk Recovery Plan No. 15. Edmonton, AB. 68 pp. Alberta Sustainable Resource Development 2010. Grizzly bear conservation in Alberta, 2010 management activities and recovery implementation. http://srd.alberta.ca/FishWildlife/WildlifeManagement/BearManagement/GrizzlyB ears/Default.aspx. Apps C.D., McLellan B.N., Woods J.G., and M.F. Proctor. 2004. Estimating grizzly bear distribution and abundance relative to habitat and human influence. Journal of Wildlife Management 68: 138-152. Baldwin R.A. and L.C. Bender. 2008. Den-Site Characteristics of Black Bears in Rocky Mountain National Park, Colorado. Journal of Wildlife Management 72: 1717-1724. â&#x20AC;&#x201D;. 2010. Development of Equations Predictive of Size and Condition for Black Bears in Rocky Mountain National Park, Colorado. American Midland Naturalist 164: 44-51. Bartels W., Law B.S, and F. Geiser. 1998. Daily torpor and energetics in a tropical mammal, the northern blossom-bat Macroglossus minimus (Megachiroptera). Journal of Comparative Physiology B-Biochemical Systemic and Environmental Physiology 168: 233-239. Benn B., and S. Herrero. 2002. Grizzly bear mortality and human access in Banff and Yoho National Parks. Ursus 13: 213-221. Burnham, K.P. and D.R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Springer-Verlag, New York, pp.484. Cattet M.R.L., Christison K., Caulkett N.A., and G.B. Stenhouse. 2003. Physiologic responses of grizzly bears to different methods of capture. Journal of Wildlife Diseases 39: 649-654. Ciarniello L.M., Boyce M.S., Heard D.C., and D.R. Seip. 2005. Denning behavior and den site selection of grizzly bears along the Parsnip River, British Columbia, Canada. Ursus 16: 47-58. Elfstrom M., Swenson J.E., and J.P. Ball. 2008. Selection of denning habitats by Scandinavian brown bears Ursus arctos. Wildlife Biology 14: 176-187. FAO. 2010. Global forest resource assessment 2010. FAO Forestry Paper 163: 1-378. Farley S.D. and C.T. Robbins. 1995. Lactation, hibernation, and mass dynamics of American black bears and grizzly bears. Canadian Journal of Zoology-Revue Canadienne De Zoologie 73: 2216-2222. Folk G.E., Hunt J.M., and M.A. Folk. 1980. Further evidence for hibernation of bears. Paper presented at 4th International Conference on Bear Research and Management, Kalispell, MT. Forman R.T.T. and L.E. Alexander. 1998. Roads and their major ecological effects. Annual Review of Ecology and Systematics 29: 207-231. Geiser F. And T. Ruf. 1995. Hibernation versus daily torpor in mammals and birds â&#x20AC;&#x201C; physiological variables and classification of torpor patterns. Physiological Zoology 68: 935-966. 95

Chapter 3-Den Site Selection Goldstein M.I., Poe A.J., Suring L.H., Nielson R.M., and T.L. McDonald. 2010. Brown Bear Den Habitat and Winter Recreation in South-Central Alaska. Journal of Wildlife Management 74: 35-42. Graber D.M. 1990. Winter behavior of black bears in the Sierra-Nevada, California. Ursus 8:269-272. Graham K., Boulanger J., Duval J., and G. Stenhouse. 2010. Spatial and temporal use of roads by grizzly bears in west-central Alberta. Ursus 21: 43-56. Harding L.E. 1976. Den-Site Characteristics of Arctic Coastal Grizzly Bears (Ursus-Arctos L) on Richards Island, Northwest Territories, Canada. Canadian Journal of ZoologyRevue Canadienne De Zoologie 54: 1357-1363. Henner C.M., Chamberlain M.J., Leopold B.D., and L.W. Burger. 2004. A multi-resolution assessment of raccoon den selection. Journal of Wildlife Management 68: 179-187. Leu M., Hanser S.E., and S.T. Knick. 2008. The human footprint in the west: A large-scale analysis of anthropogenic impacts. Ecological Applications 18: 1119-1139. Linnell J.D.C., Swenson J.E., Andersen R., and B. Barnes. 2000. How vulnerable are denning bears to disturbance? Wildlife Society Bulletin 28: 400-413. Mace R.D., Waller J.S., Manley T.L., Lyon L.J., and H. Zuuring. 1996. Relationships among grizzly bears, roads and habitat in the Swan Mountains, Montana. Journal of Applied Ecology 33: 1395-1404. McLellan B.N. 1998. Maintaining viability of brown bears along the southern fringe of their distribution. Ursus 10: 607-611. McLellan B.N. and D.M. Shackleton. 1988. Grizzly Bears and Resource-Extraction Industries - Effects of Roads on Behavior, Habitat Use and Demography. Journal of Applied Ecology 25: 451-460. McLoughlin P.D., Cluff H.D., and F. Messier. 2002. Denning ecology of barren-ground grizzly bears in the central Arctic. Journal of Mammalogy 83: 188-198. Munro D., Thomas D.W., and M.M. Humphries. 2005. Torpor patterns of hibernating eastern chipmunks Tamias striatus vary in response to the size and fatty acid composition of food hoards. Journal of Animal Ecology 74: 692-700. Nielsen S.E., McDermid G., Stenhouse G.B., and M.S. Boyce. 2010. Dynamic wildlife habitat models: Seasonal foods and mortality risk predict occupancy-abundance and habitat selection in grizzly bears. Biological Conservation 143: 1623-1634. Nielsen S.E., Herrero S., Boyce M.S., Mace R.D., Benn B., Gibeau M.L., and S. Jevons. 2004. Modelling the spatial distribution of human-ccaused grizzly bear mortalities in the Central Rockies ecosystem of Canada. Biological Conservation 120: 101-113. Petram W., Knauer F., and P. Kaczensky. 2004. Human influence on the choice of winter dens by European brown bears in Slovenia. Biological Conservation 119: 129-136. Roever C.L., Boyce M.S., and G.B. Stenhouse. 2008a. Grizzly bears and forestry - I: Road vegetation and placement as an attractant to grizzly bears. Forest Ecology and Management 256: 1253-1261. â&#x20AC;&#x201D;. 2008b. Grizzly bears and forestry - II: Grizzly bear habitat selection and conflicts with road placement. Forest Ecology and Management 256: 1262-1269. â&#x20AC;&#x201D;. 2010. Grizzly bear movements relative to roads: application of step selection functions. Ecography 33: 1113-1122. 96

Chapter 3-Den Site Selection SAS Institute Inc. 2011. SAS/STATÂŽ 9.3 Userâ&#x20AC;&#x2122;s Guide. Cary, NC:SAS Institute Inc. Schwartz C.C., Haroldson M.A., and G.C. White. 2010. Hazards affecting grizzly bear survival in the Greater Yellowstone Ecosystem. Journal of Wildlife Management 74: 654667. Squires J.R., DeCesare N.J., Kolbe J.A., and L.F. Ruggiero. 2008. Hierarchical den selection of Canada lynx in western Montana. Journal of Wildlife Management 72: 1497-1506. Swenson J.E., Sandegren F., Brunberg S., and P. Wabakken. 1997. Winter den abandonment by brown bears Ursus arctos: causes and consequences. Wildlife Biology 3: 35-38. Thomas D.W., Dorais M., and J.M. Bergeron. 1990. Winter energy budgets and cost of arousals for hibernating little brown bats, Myotis lucifugus. Journal of Mammalogy 71: 475-479. Tietje W.D., and R.L. Ruff. 1980. Denning behavior of black bears in boreal forest of alberta. Journal of Wildlife Management 44: 858-870. Vroom G.W., Herrero S., and R.T. Ogilvie. 1980. The ecology of winter den sites of grizzly bears in Banff National Park, Alberta. Ursus 4: 321-330. Watts P.D., and C. Jonkel. 1988. Energetic cost of winter dormancy in grizzly bear. Journal of Wildlife Management 52: 654-656. Weaver J.L., Paquet, P.C., and L.F. Ruggiero. 1996. Resilience and conservation of large carnivores in the Rocky Mountains. Conservation Biology 10: 964-976.

Chapter 3-Den Site Selection

Figure 1. Map of the Grande Cache (North) and Yellowhead (South) population units showing elevation gradient.

Chapter 3-Den Site Selection

Figure 2. Map of the relative probability of den site occurrence in the Grande Cache population unit using the best model in the Food category of landscape attribute.

Chapter 3-Den Site Selection

Figure 3. Map of the relative probability of den site occurrence in the Grande Cache population unit using the best model in the Topography category of landscape attribute.

100

Chapter 3-Den Site Selection

Figure 4. Map of the relative probability of den site occurrence in the Grande Cache population unit using the best model in the Habitat category of landscape attribute.

101

Chapter 3-Den Site Selection

Figure 5. Map of the relative probability of den site occurrence in the Grande Cache population unit using the best model in the Anthropogenic category of landscape attribute.

102

Chapter 3-Den Site Selection

Figure 6. Map of the relative probability of den site occurrence in the Grande Cache population unit using the most supported multivariate model.

103

Chapter 3-Den Site Selection

Figure 7. Map of the relative probability of den site occurrence in the Yellowhead population unit using the best model from the Food category of landscape attributes.

104

Chapter 3-Den Site Selection

Figure 8. Map of the relative probability of den site occurrence in the Yellowhead population unit using the best model from the Topography category of landscape attributes.

105

Chapter 3-Den Site Selection

Figure 9. Map of the relative probability of den site occurrence in the Yellowhead population unit using the best model from the Habitat category of landscape attributes.

106

Chapter 3-Den Site Selection

Figure 10. Map of the relative probability of den site occurrence in the Yellowhead population unit using the best model from the Anthropogenic category of landscape attributes.

107

Chapter 3-Den Site Selection

Figure 11. Map of the relative probability of den site occurrence in the Yellowhead population unit using the most supported multivariate model.

108

Chapter 3-Den Site Selection

Figure 12. Map of the relative probability of den site occurrence in a foothills watershed of the Grande-Cache population unit using the most supported multivariate model.

109

Chapter 3-Den Site Selection

Figure 13. Map of the relative probability of den site occurrence in a mountain watershed of the Yellowhead population unit using the most supported multivariate model. .

110

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting

larsenterry@gmail.com, 2University of Alberta

GENERAL INTRODUCTION In western North America, the mountain pine beetle (Dendroctonus ponderosae: hereafter MPB) is considered the most destructive insect of mature pine forests (Sanfranyik et al. 2010). Since the mid 1990â&#x20AC;&#x2122;s in British Columbia, epidemic outbreaks of MPB had impacted over 13 million hectares of pine forests by 2008 (Wulder et al. 2010), an area ten times larger than any previously recorded outbreak (Sanfranyik et al. 2010). In 2006, as MPB expanded across the interior of British Columbia, widespread infestations of MPB were also detected in northwest central Alberta. This raised concerns that MPB populations could spread across the eastern slopes of the Rocky Mountains and Canadaâ&#x20AC;&#x2122;s boreal forest (Anonymous 2007). In response, Alberta Sustainable Resource Development initiated a long-term pine management strategy (Anonymous 2007). The intent of the pine management strategy is to aggressively change the age class structure of the landscape in an effort to reduce the susceptibility of lodgepole pine (Pinus contorta) stands to MPB attack (Anonymous 2007). Because of fire suppression efforts in Alberta over the past century, even-aged stands of mature lodgepole pine, an age class highly susceptible to MPB infestation, are extensive particularly within the Foothills Natural Region (Anonymous 2007, Day 1972). The pine strategy directed Forest Management Agreement (FMA) holders within public and operable lands to increase the annual allowable cut and surge harvest 75% of the most susceptible pine stands over a 20 year period (Anonymous 2007). Accelerated harvesting of pine is a significant departure from standard even-flow timber supply management. The implications of the surge cut, both immediate and in the future, to habitat considered essential to the conservation and recovery of a threatened species like the grizzly bear (Ursus arctos) is a key concern in Alberta. Considerable research indicates that forest harvesting can negatively affect grizzly bear survival through the development of roads (Nielsen et al. 2004b, Nielsen 2006, Nielsen et al. 2008). While challenging to manage, the risks of roads to grizzly bears can be controlled by

111

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting limiting human access and activity (Nielsen et al. 2009). However, less is known about how clearcut harvesting will alter habitat conditions for grizzly bears. Depending on local stand conditions (i.e. tree species composition, elevation) (Nielsen et al. 2004c, Visscher and Merrill 2009), clearcuts are thought to contain a wider array of plant based food resources for grizzly bears than mature forests while total fruit production may not differ (Nielsen et al. 2004c). However, past models of grizzly bear food supply as a function of forest age in pine forest have either been qualitative, based on presence/absence data, and/or limited to community level descriptions of vegetation form (shrub vs. herbaceous) (Nielsen et al. 2004c, Bainbridge and Strong 2005, Visscher and Merrill 2009). Although clearcut logging generally results in increased graminoid, forb, and shrub cover for many species, destruction of rhizomes can reduce shrub growth substantially for others (Bainbridge and Strong 2005, Zager et al. 1983). Conversely, unimpeded tree growth after harvest may result in more shrubs but less fruit production, as fruit production is influenced by fire history in many plants (Hamer and Herrero 1987). To our knowledge, no studies have quantified grizzly bear food abundance as a function of stand age in lodgepole pine forests and in relation to local stand conditions. Given that shrub abundance and fruit production may not be that tightly linked (Noyce and Coy 1990, Reynolds-Hogland 2006, Suring et al. 2009), it is important to quantitatively evaluate how Albertaâ&#x20AC;&#x2122;s MPB strategy may affect grizzly bear foods now and in the future. Previous studies that have investigated the response of grizzly bears to clearcut forest harvesting suggest that some populations avoid (Zager et al. 1983, McLellan and Hovey, 2001) while others select for clearcuts (Mace et al. 1996, Nielsen et al. 2004a). This observed difference is hypothesized to be related to the availability of natural disturbances that create openings such as fires and avalanches (Nielsen et al. 2004a). These studies also suggest that selection varies tremendously among grizzly bears reflecting individual behavior or variation in local food resource distribution and abundance related to plant phenology and environmental gradients (i.e. elevation). However, relatively few studies have examined the seasonal selection of clearcuts by grizzly bears of different age classes (Nielsen et al. 2004a) specific to pine dominant stands (Mattson 1997). And, none have investigated age associated differences in selection according to local environmental gradients such as elevation. Because food resources in pine forests may vary seasonally according to differences in these stand conditions (La Roi and Hnatiuk 1980, Strong 2002), it is important to assess the response of grizzly bears to clearcut harvesting in operable pine habitats where MPB management is expected to occur in Alberta. Habitat selection analysis is commonly used in wildlife management and conservation applications. Selection is based on the premise that disproportionate use of habitat (% use / % available) by animals reflects a behavioral consequence of individuals actively selecting for essential resources (habitat) fundamental to their survival and reproduction (Boyce 1999). However, the measure of selection is sensitive to the proportion of available habitat (Beyer et al. 2010). When habitats are relatively common, animals may show avoidance even though use is high, whereas when habitats are relatively rare habitats may be selected when use is comparatively low. This may influence inference and misguide management of wildlife habitat and consequently animal populations (Long

112

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting et al. 2009). Given that mature forests are abundant on the eastern slopes, independent measures of use and selection may provide complimentary information when assessing grizzly bear response to pine forest harvesting and making management recommendations. Here we quantify changes in food supply and assess grizzly bear response at the population level to harvesting mature stands of lodgepole pine in north-west central Alberta where MPB management is underway. The two main topics and objectives of this research include: Topic A. Food Supply: Our specific objectives were 1) identify the most common and abundant bear foods in the Upper Foothills; 2) determine what effect overstory pine composition, stand age, and environmental gradients had on the abundance of shrubs, the abundance of fruits, and the abundance of forbs; 3) forecast total food biomass as a function of changes of pine forest age over the proposed spatial harvesting sequence for MPB while accounting for local environmental gradients. Topic B. Grizzly Bear Response: Our specific objectives were to describe the seasonal use and selection of forested habitat by female grizzly bears categorized as 1) operable and inoperable; 2) pine and other forested stand types in the operable land base; 3) pine forest age classes in the operable land base; and 4) pine forest age classes by elevation in the operable land base.

Study Area The study area encompasses the southern boundary of Weyerhaeuser Grande Prairie Forest Management Agreement (FMA) in northwest-central Alberta, Canada (119° 13’W and 54° 32’N). Variation in climate, topography, and vegetation follow a prominent elevation gradient from the south west to the north east. The higher elevation and more rugged conifer dominated forests of the Subalpine and Upper Foothills transitions to lower elevation gently rolling terrain characteristic of the Lower Foothills and Central Mixedwood Natural Sub Regions. The Subalpine is a mixture of lodgepole pine, engelmann spruce (Picea engelmanni), and subalpine fir (Abies lasiocarpa) (Natural Regions Committee 2006). Lodgepole pine dominates the Upper Foothills often with black spruce (Picea Marianna), while white spruce (Picea glauca) occurs along river valleys (Natural Regions Committee 2006). Mixed and deciduous stands are confined to south and west aspects at the lowest elevations (Natural Regions Committee 2006). The Lower Foothills are diverse with pure and mixed stands of trembling aspen (Populous tremuloides), poplar (Populous balsamifera), white birch (Betula papyrifera), lodgepole pine, black and white spruce, balsam fir (Abies balsamifera), and tamarack (Larix laricina) (Natural Regions Committee 2006). In the Upper Foothills, Alberta Vegetation Inventory (AVI) suggests the area is dominated by closed canopy conifer forests (95%) with few natural openings or burns (<1%). The remaining 4% is anthropogenic footprint including roads, pipelines, seismic lines, and well sites. The majority of stands contain pine (91%) and of those, 61% is pine leading. Since the early 1970s, forests have been intensively managed for commercial timber production creating a mosaic of seral stages, with an age distribution skewed towards mature (>80 years) and young (<35 years) age classes. Topographic relief is rolling and steeply sloping with elevations ranging from 950 to 1750m (Natural

113

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting Regions Committee 2006). Average annual precipitation varies from 452.5 to 992.9 millimetres with average annual temperature from -1.6 to 2.5 °C (Natural Regions Committee 2006). The short growing season is cool and wet, and winters are cold with substantial amounts of precipitation in the form of snow (Natural Regions Committee 2006).

TOPIC A: FOOD SUPPLY Methods Grizzly Bear Food Resources Grizzly bear diet in the Rocky Mountain Ecosystem of Canada is well documented (Hamer and Herrero 1987, Hobson et al. 2000, Larsen and Pigeon 2006, McLellan and Hovey 1995, Munro et al. 2006). Based on scat analysis, the diet of bears and presumably the availability of specific plants appears to differ by latitude and altitude, thus, we quantified plant-based foods known to be used seasonally by grizzly bears in west-central and north-west central Alberta (Larsen and Pigeon 2006, Munro et al. 2006, Nielsen et al. 2004c). Seasonal shifts in the diet of grizzly bears typically follow changes in food availability associated with plant phenology (Munro et al. 2006). After den emergence and prior to green-up, grizzly bears dig sweet-vetch roots (Hedysarum alpinium, Hedysarum boreale). With the onset of green-up, grizzly bears forage on horsetails (Equisetum spp.), cow parsnip (Heracleum lanatum), clover (Trifolium spp.), dandelion (Taraxacum officinale), and clasping-leaved twisted stalk (Streptopus amplexifolius). With the ripening of fruits, grizzly bears consume buffaloberry (Shepherdia canadensis), black huckleberry (Vaccinium membranaceum), lingonberry (Vaccinium vitis-idaea), velvet-leaved blueberry (Vaccinium myrtilloides), dwarf blueberry (Vaccinium caespitosum), bearberry (Arctostaphylos uva-ursi), raspberry (Rubus idaeus), and crowberry (Empetrum nigrum). Grasses and sedges are often consumed by grizzly bears, but we did not consider these because no specific species have been identified in previous dietary studies.

Vegetation Sampling and Environmental Characteristics We used a Geographic Information System (GIS) forest inventory database (net land base) provided by Weyerhaeuser to determine the extent of pine forests and to define our sampling area within the Upper Foothills. The net land base was the spatial input for the revised Timber Supply Analysis (TSA) for MPB management amended in 2009 (Anonymous 2006a). We excluded stands where age and overstory composition were not known or classified as inoperable (land use, water buffer, steep slopes). Remaining upland stands of pine were stratified according to age of origin using a 5 year interval for harvested (1-35 years) and 30 year interval for non-harvested (≥38 years). Our goal was to sample more intensively in harvested areas because we expected greater variation in food abundance in young stands. We used Hawth’s Tools (Beyer 2004) to generate random location coordinates within the stratified area, with the restriction that our sampling was within 1 km of roads because of accessibility. We established 30 x 30m vegetation plots, orientated south to north, at random location coordinates in harvested (n=145) and non-harvested pine (n=104) from June 17 to September 1,

114

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting 2008 (n=136) and from June 28 to October 1, 2009 (n=113). We chose a 30 x 30m (900 m2) sampling area to match the grain of our GIS raster grids used as explanatory variables. Plot centre was located as the nearest Universal Transverse Mercator (UTM) coordinate with the lowest GPS error that did not exceed 10m. If any of the plot area fell within 30m of a harvested area boundary or anthropogenic land use feature (road, trail, seismic line, or well-site) the plot was moved in a random cardinal direction perpendicular to the edge or feature until there was no overlap. A 30m distance was used to avoid possible edge effects (Lopez et al. 2006, Redding et al. 2003, Harper and Macdonald 2002). Plots were also moved 30m in a random cardinal direction if ten percent or greater of the plot area in harvest blocks contained overstory retention trees. Overstory tree species composition was determined by using an ocular estimate of percent cover values in classes (0, 1-5, 6-10, and 11-20...90-100) for each tree species that formed the canopy. Tree species composition was determined by taking the median percent cover value for each tree species averaged at the plot level. For ten plots in clearcuts the overstory composition prior to harvesting was determined from AVI. Within the 30 x 30m sampling area, we established five 1 x 30m transects (south to north), evenly spaced along the southern plot boundary at 0.5, 7.5, 15, 22.5, and 29.5m. Using the same spacing, we placed five 1 x 1m quadrats along each transect. In quadrats, we counted individual stems of T. officinale and Equisetum spp. We also counted berries and estimated percent cover of E. nigrum, V. membranaceum, V. myrtilloides, V. caespitosum, V. vitis-idea, Arctostaphylos uva-ursi, and Trifolium spp. Cover was estimated occularly using single percent values from 1 to 20 and in 5 percent increments above 20. In transects, we counted individual Hedysarum spp. plants, stems of S. amplexifolius, and H. lanatum and stems and berries of S. canadensis, R. idaeus, and Aralia nudicaulis. In addition, we performed meander searches within the 30 x 30m area recording the presence/absence of forb and shrub species. During the 2009 sampling season, we clipped the stems of forbs at ground level and collected ripe berries from a maximum of 5 random quadrats and 5 random transect sections (7.5m). Ripe fruit was weighed in the field using a 10g PESOLA速 scale. Vegetation was collected, dried at room temperature, and weighed in the field using an OHAUS速 Adventurer SL digital scale. At the plot level, we calculated mean fresh weight of fruit (g/berry) and dry biomass of forbs (g/stem) for each of the species sampled. For each species, mean weight was converted to mean digestible dry matter (hereafter DDM) (Table 1). We used DDM rather than dry biomass because 1) the digestibility of food types by bears can differ and 2) DDM is linearly related to digestible energy (KJ/g) in most species (Partridge et al. 2001, Pritchard and Robbins 1990, Welch et al. 1997). Our DDM estimates of forbs may be biased high because we did not dry samples in an oven at a specific temperature over a period of time.

Predictor variables We used GIS raster grids to represent site specific environmental conditions at each plot and across the extent of pine stands within the Upper Foothills. From a 30m pixel digital elevation model (DMTI 2003), we created aspect and slope grids using the Spatial Analyst Extension in ArcGIS 9.2

115

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting (ESRIÂŽ Redlands, CA). Aspect and slope grids were then used as inputs to calculate a Solar Radiation Index (SRI; Keating et al. 2007) and a Compound Topographic Index (CTI). The solar radiation index simultaneously combines aspect, slope, and latitude (Keating et al. 2007). CTI is an index of wetness and has been correlated with soil attributes such as horizon depth, percentage of silt, organic matter content, and phosphorous (Evans 2002 and references within). We used elevation as a surrogate variable for climate because it was correlated (>0.7) with climate normals including: annual moisture index, degree days below zero, length of the frost free period, growing season precipitation, mean annual temperature and precipitation, and summer moisture index. Many of these variables had very high variability with clear outliers. Thus, to minimize the impacts of this high variability, continuous environmental variables (CTI, SRI, Elev) were transformed into equal frequency groups (Table 2). A binary variable was used to represent leading (1) or non-leading (0) stands of pine and whether the sampling year was 2008 (0) or 2009 (1). Leading pine was defined as greater than or equal to 60 percent overstory canopy pine composition. From hemispherical canopy photos (unpublished data) averaged for each plot (n=5), canopy closure in pine stands younger than 20 years was 3% (range 0-40) with 63 of 78 sites having a value of zero. The average canopy cover for harvested stands between 21-35 years of age was 43% (range 0-68) and 53% (range 34-65) in non-harvested stands. We represented stand age by three variables: 1) a binary variable to distinguish between harvested (1) and non-harvested (0) (hereafter CutVsUncut); 2) a continuous 20 year interval (hereafter Age20); and 3) a categorical variable that split age into four levels: A) 1-20 years; B) 2135 years; C) 36-80 years; and D) â&#x2030;Ľ80 years (herafter AgeCat). AgeCat represents A) harvested stands and relative open canopy conditions; B) harvested stands and relative closed canopy conditions; C) non-harvested stands and rare at the landscape level; and D) non-harvested stands, abundant at landscape level and the target of MPB management. The categorical classification of age was similar to the seral stages used by Weyerhaeuser (Early 1-10; Intermediate 11-40; Mature 41-80; Late >80). Our rationale for considering these different definitions was to test whether a specific age variable was a better predictor of plant distribution/abundance. From a sampling perspective, the 37-80 year age class was relatively uncommon, especially stands between 37-59 years. Of the 249 plots we sampled in 2008 (136) and 2009 (113), 188 were pine leading stands of which 108 were clearcuts. Only 14 plots were in the 37-80 year age class.

Species Occurrence and Abundance We calculated the occurrence probability of forbs and shrubs in pine, harvested, and non-harvested stands from meander searches of species presence/absence. We rank ordered bear foods by type (forbs or shrubs) according to the probability of occurrence in pine. We used logistic regression (StataCorp 2009) to predict the effect of CutVsUncut on the probability of bear foods occurring. We summed counts of forbs, shrubs, and fruits at the plot level for each species. Low counts in 8 of the 16 bear foods precluded modelling abundance. Sampling variance was high for the other 8 species, thus, to minimize the effect of outliers on model predictions, we visually inspected scatter plots of counts versus predictors removing obvious outliers. We used zero-inflated negative binomial regression (StataCorp 2009) with robust estimates of variance to quantify the effect of

116

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting predictors on the mean relative abundance of the response variables (Table 2). We used a Vuong test statistic (Vuong 1989) and a model fitting procedure (Long and Freese 2005) to confirm that ZINB was appropriate. The zero-inflated mixture model combines the probability distributions of two latent groups, a point mass at zero and a negative binomial distribution (Martin et al. 2005, Nielsen et al. 2005). First, the always zero or inflation group (presence/absence) is modelled as a logistic function with an outcome of 0 and a probability of 1, while the non-zero or count group (abundance) may have a zero count, but has a non-zero probability of the count being positive (Nielsen et al. 2005). The model simultaneously considers covariates influencing both distribution and abundance, while final predictions are the product of the two (Martin et al. 2005, Nielsen et al. 2005). Because plants are sessile and sampling occurred during the growing season, we assumed that absences represent â&#x20AC;&#x2DC;true zerosâ&#x20AC;&#x2122; and the result of ecological process rather than imperfect detection due to observer error or plant phenology (Martin et al. 2005). In our first step and beginning with the inflation group of response variables, we fit models with individual parameters for sampling year and each age variable including a non-linear effect for Age20 (squared term). In separate models we introduced year and each age variable including interactions between year and age. In our second step, we tested whether the top model from step 1 was better than individual models representing pine composition, environment (Elev, CTI, and SRI) or all possible combinations of these variables including non-linear effects for environment (squared term). Possible interaction effects between elevation and sampling year as well as pine composition and sampling year were observed, so we included these effects as separate models in our candidate set. We repeated these steps for the count group, but included the best fit model from the inflation group. Prior to model fitting, variance inflation factors (VIF) were used to assess multicollinearity for each combination of predictors excluding interaction effects. VIF was considered influential when greater than 10 or when the mean of all VIFs was greater than 1 (Nielsen et al. 2005). In our last step, we retained model parameters when slopes significantly differed from zero (Îą <= 0.1). Percent deviance (D2) was calculated from the null and full models for each latent group to quantify model fit.

Spatial Harvesting Sequence In GIS we used a spatial harvest sequence provided by Weyerhaeuser to determine the age of pine stands in period 1 (2006) and at each decadal time step from periods 2 through 12. Each harvesting period represents 5 years. We assumed that pine composition remained constant and that stands did not die (age=1) after 300 years for three reasons. First, leading stands were assumed to transition without change within the TSA (Anonymous 2006b). Second, deciduous leading stands were rare and would have little influence on food abundance estimates unless a dramatic conversion to deciduous forest occurs. Third, our goal was to quantify potential changes in food abundance that was representative of harvesting rather than stand senescence. We consider the first 20 years of the projection to be a likely scenario since many of the blocks proposed for harvest were pre-selected. On the other hand, the next forty years is less likely because the TSA did not limit harvesting within caribou management zones, thus, our projections beyond 2018 and the end of the surge cut are more speculative. However, it does represent a change in the age distribution

117

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting of pine and the legacy of MPB surge harvesting. GIS layers from each time step that described pine composition, stand age, and environmental conditions for each 30x30m pixel were used to predict total abundance (counts) of forbs, shrubs, and fruits from the best fitting models. Final model predictions for shrubs, fruits, and forbs were scaled to a 900m2 area by multiplying predicted counts by a factor of 36 for species collected in quadrats (25 m2) and 6 for those in transects (150 m2). We then converted count predictions of forbs and fruits to DDM (g/900m2). Because it is difficult to count individual stems of Vaccinium spp., we used % cover estimates from the quadrats as our measure of abundance. Cover estimates were treated as counts to allow the use of zero-inflated models. Unreported analyses using a logit link in a generalized linear model (StataCorp 2009) provided similar predictions. Vaccinium spp. abundance was scaled by taking the mean predicted % cover and dividing by 90000 (100 units of cover/1m2) to get abundance per 900m2. Using this logic, we assumed one R. idaeus stem was equivalent to 1 unit of % cover and scaled abundance accordingly.

Results Grizzly Bear Food Occurrence The rank order occurrence probability of forbs was Equisetum > S. amplexifolius >H. lanatum > T. officinale > Trifolium spp. > Hedysarum spp (Table 3). For fruit producing shrub species, the rank probability of occurrence was V. vitis-idaea > V. myrtilloides > V. membranaceum > R. idaeus > V. caespitosum > S. canadensis > A. nudicaulis > A. urva-ursi > E. nigrum (Table 3). Based on logistic regression analysis, the forbs H. lanatum and Trifolium spp. were significantly more likely to occur in harvested pine, while Equisetum spp. and S. amplexifolius did not differ. Of the shrub species, R. idaeus and A. uva-ursi were significantly more likely to occur in harvested and V. vitis-idaea, V. membranaceum, V. caespitosum, and E. nigrum were more likely to occur in non-harvested (Table 3). CutVs.Uncut had no effect on S. canadensis or A. nudicaulis (Table 3).

Grizzly Bear Food Abundance ZINB models describing the distribution and abundance of forbs, shrubs, and fruits explained between 1.4 and 11.8% of the total variance (Table 4). Below we report on percent changes in mean DDM (g/900m2). Pine composition was a predictor in 11 of 13 response variables but did not explain variation in S. amplexifolius or V. caespitosum shrubs. Compared to leading pine, DDM of Equisetum (120%), H. lanatum (234%), and R. idaeus shrub (128%) and fruit (83%) increased in non-leading stands. Conversely, DDM of V. myrtilloides shrub (46%) and fruit (46%), V. membranaceum shrub (61%) and fruit (82%), and V. vitis-idaea shrub (13%) and fruit (47%) decreased in non-leading stands. There was a significant interaction between pine composition and year in V. caespitosum as fruit was only abundant in pine leading stands in 2008. Stand age was a predictor in 11 of 13 response variables but had no effect on V. myrtilloides or V. caespitosum shrubs. Compared to mature stands (>80 years), within the first 20 years post harvest (open canopy) DDM of Equisetum spp. (239%), H. lanatum (155%), R. idaeus shrub (983%) and fruit

118

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting (700%), V. myrtilloides fruit (1172%), and V. caespitosum fruit (266%) increased (Figures 1 and 2). With succession over the next twenty years (closed canopy), Equisetum spp. (46%), R. idaeus shrub (72%), V. myrtilloides fruit (85%), and V.caespitosum fruit (22%) decreased, while H. lanatum increased by 137% (Figures 1 and 2). As stands continued to age, these species were predicted to decrease. V. membranaceum shrub (82%) and fruit (86%) and V. vitis-idaea shrub (61%) and fruit (82%) decreased substantially after clearcutting (Figure 2). Shrubs and fruits were predicted to increase linearly with age, however, V. membranaceum fruit reached an asymptote in mature stands and V. vitis-idaea fruit peaked between 100-120 years, while shrubs continued to increase (Figure 2). In older clearcuts (21-40 years), V. membranaceum and V. vitis-idaea fruit increased from young clearcuts (1-20 yrs) to 68 and 61% of the mean of mature stands and to 32 and 16% less than the mean of mature stands in those from 41-80 years (Figure 2). Elev, CTI, and SRI were predictors in 9, 11, and 2 of 13 response variables, respectively. In 4 of the 5 fruiting species, the relationship between DDM to Elev was consistent between shrub and fruit models. V. myrtilloides was most abundant at low and intermediate Elev (Figure 3). R. idaeus decreased with an increase in Elev, while V. caespitosum and V. membranaceum increased (Figure 3). V. vitis-idaea shrubs were most abundant at intermediate Elev, but this relationship was not corroborated by the fruit model (Figure 3). Comparing the two most abundant fruiting species, dominance shifted from V. myrtilloides to V. membranaceum at an Elev of about 1204m based on the shrub (mean of 4th Elev interval) model or 1228m (upper limit of 4th Elev interval) according to the fruit model (Figure 3). There was a significant interaction between pine composition and Elev in V. membranaceum shrub and fruit models. In non-leading stands, the slope of the line was steeper as an increasing function of Elev. CTI consistently predicted patterns of shrub and fruit abundance in 4 of the 5 species. With an increase in CTI, Equisetum spp., V. myrtilloides, V. caespitosum, and V. vitis-idaea increased. Conversely, H. lanatum, S. amplexifolius, and R. idaeus decreased. The relationship between V. membranaceum shrub abundance and CTI was non-linear and peaked at intermediate levels, while fruit showed no response. Both Equisetum spp. and V. caespitosum fruit decreased with an increase in SRI. Sampling year was a predictor in 5 of 13 response variables but had no effect on Equisetum spp., H. lanatum, shrubs, or R. idaeus fruit. Vaccinum spp. fruit was most abundant in 2008 (Figure 2). In 2009, the mean dropped by 63, 47, 52, and 69% in V. myrtilloides, V. caespitosum, V. membranceum, and V. vitis-idaea, respectively (Figure 2). For V. myrtilloides, the effect of year on abundance was more pronounced because of the interaction between year and AgeCat. From 2008 to 2009 in clearcuts 1-20 years, the percent change in V. myrtilloides fruit was 96% (Figure 2). We also found a significant interaction between year and CutVs.Uncut in S. amplexifolius. In 2008, S. amplexifolius was more abundant in harvested and in 2009 this species was more abundant in nonharvested pine (Figure 1). The rank order of abundance of forbs was Equisetum spp. > H. lanatum (77%) > S. amplexifolius (98%); shrubs was V. membranceum > V. vitis-idaea (50%) > V. myrtilloides (59%) > R. idaeus (87%) > V. caespitosum (91%); fruits in 2008 was V. myrtilloides > V. membranceum (5%) > V. vitis-idaea

119

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting (72%) > R. idaeus (78%) > V. caespitosum (92%); and fruits in 2009 was V. membranceum > R. idaeus (53%) >V. myrtilloides (55%) > V. vitis-idaea (81%) > V. caespitosum (93%).

Grizzly Bear Food Supply Stands of leading pine accounted for 68% of the simulation area. In period 1, leading pine was skewed towards mature (>80 years; 65%) and harvested (1-40 years; 28%) age classes (Figure 4). Taking the average between periods 2 and 4 to represent the end of the surge cut in period 3, mature decreased to 43% from period 1, while young clearcuts (1-20 years) went from 17 to 33%, and older clearcuts (21-40 years) from 11 to 16% (Figure 4). With the progression of harvesting and succession following the surge cut, in period 12 about 21% was mature, 17% was young clearcut, 18% was older clearcut, and 44% was between 41-80 years of age (Figure 4). Non-leading stands were also harvested in the spatial sequence and the change in the age distribution was similar but without any noticeable increase in any one age class. At the landscape level in period 1, the rank order abundance of forbs was Equisetum spp (930.4t). > H. lanatum (134.4t) > S. amplexifolius (2008=36.8t; 2009=41.0t) (Figure 5). Of the fruiting species, the rank order abundance of shrubs was V. membranceum (0.029%) > V. myrtilloides (0.026%) > V. vitis-idaea (0.024%) > V. caespitosum (0.006%) > R. idaeus (0.001%); fruits in 2008 was V. membranceum (59.0t) > V. myrtilloides (20.9t) > V. vitis-idaea (20.3t) > R. idaeus (6.4t) > V. caespitosum (3.1t); and fruits in 2009 was V. membranaceum (28.6t) > R. idaeus (6.4t) > V. vitisidaea (6.3t) > V. myrtilloides (5.8t) > V. caespitosum (1.3t) (Figure 6). At the end of the surge cut in period 3, DDM of the forbs Equisetum spp. (26%) and H. lanatum (15%) increased. Shrub cover increased in R. idaeus (62%), did not change in V. myrtilloides and V. caespitosum, and decreased in V. membranaceum (11%) and V. vitis-idaea (8%). Based on optimal conditions (2008), DDM of S. amplexifolius (5%) and fruits of R. idaeus (42%), V. myrtilloides (67%), and V. caespitosum (21%) increased, while V. membranaceum (17%) and V. vitis-idaea (23%) decreased. In sub-optimal conditions (2009), the percent change in abundance was nearly equivalent for each species with the exception that S. amplexifolius decreased by 5% and V. myrtilloides only increased by 25%. Post surge and until the end of period 12, Equisetum spp. and H. lanatum both decreased but were 12 and 19% more abundant than in period 1. However, in periods 6 and 8 H. lanatum was exceedingly abundant compared to other harvesting periods. Shrub cover also decreased post surge, however, R. idaeus was 28% above, while V. membranaceum and V. vitis-idaea were 11 and 8% below pre-harvest estimates. Given optimal conditions (2008), DDM of S. amplexifolius decreased but was 3% more abundant than pre-harvest abundance. DDM of fruit also decreased, but R. idaeus (32%), V. myrtilloides (30%), and V. caespitosum (27%) were more abundant than in period 1, whereas V. membranceum (23%) and V. vitis-idaea (34%) were less abundant. Again, percent change in DDM was similar for each species in sub-optimal condition (2009) with the exception being that S. amplexifolius was 3% less abundant and V. myrtilloides fruit 4% more abundant compared to period 1.

120

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting After the surge cut, total DDM of forbs increased by 24% and shrub cover decreased by 5%. According to model estimates in 2008, total DDM of fruit increased by 2%, but decreased by 4% based on estimates in 2009. Post surge, total DDM of forbs remained high in period 4 with a declining trend from periods 6 to 12, although there was a noticeable increase in period 8. Shrub cover increased from period 3 to 4, but declined precipitously until the end of period 12. Total DDM of fruit was similar in period 4, decreased sharply in period 5, and remained stable until the end of period 12. After 12 harvesting periods, total DDM of forbs was predicted to be 13% higher than in period 1. Total shrub cover and DDM of fruit given conditions in 2008 and 2009 were predicted to be 8, 10, and 13% lower than abundance prior to MPB harvesting.

Discussion Grizzly Bear Food Occurrence in the Upper Foothills Of the 15 bear foods we examined, 7 were relatively uncommon and associated with disturbance features or were outside common environmental conditions in the Upper Foothills. Similar to Nielsen et al. (2004c), the exotic species T. officinale and Trifolium spp. were more likely to occur in clearcuts, but are even more common adjacent to roads where seeding is used for erosion control (Roever al. 2008). Hedysarum spp. appears to be limited to lower elevation disturbed sites or riparian areas that have high soil wetness which explains the lack of occurrence in the Upper Foothills (Nielsen et al. 2010a). The remaining fruit producing shrub species were less common than Vaccinium spp., yet associated with lodgepole pine forests under different climatic and edaphic conditions (La Roi and Hnatiuk 1980, Strong 2002). In a montane environment, La Roi and Hnatiuk (1980) found that S. canadensis was associated with a xeric moisture regime, dominant below 1250m with cover peaking between 1100-1300m, and became rarer with increasing elevation. Similarly, A. uva-ursi was associated with xeric conditions and maximum cover was below 1150m where it occurs with S. canadensis (La Roi and Hnatiuk 1980). Nielsen et al. (2010b) found that S. canadensis was more abundant in the Lower Foothills. E. nigrum was associated with higher elevation pine stands and appears to be a secondary species with other Vaccinium spp. (La Roi and Hnatiuk 1980). Our results in the Upper Foothills show that using food models from different regions may be problematic. Even though lodgepole pine is widely distributed across Albertaâ&#x20AC;&#x2122;s Rocky Mountain region, there are distinct compositional changes in understory shrub dominance relative to other areas of the province. We suspect that local and regional gradients in moisture, edaphic conditions, and productivity (Wang et al. 2004) driven by elevation and latitude may explain variation in bear food distribution and abundance in lodgepole pine stands across the eastern slopes of Alberta (La Roi and Hnatiuk, Strong 2002). Regional inventories of bear foods in lodgepole pine are required to examine the consequences of MPB management at the provincial level.

Grizzly Bear Food Abundance in the Upper Foothills Our results suggest that when the overstory composition was less dominated by pine, shrub cover and fruit production decreased in the Vaccinium spp. while there was an increase in Equisetum spp., H. lanatum, and R. idaeus shrubs and fruits. Other studies have also found Vaccinium spp.

121

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting shrub cover and fruit production decreases when other conifers and deciduous species are more prolific (Pelchat and Ruff 1986, Ihalainen and Pukkala 2001). Presumably, Vaccinium spp. thrive in nutrient poor conditions with few competitors or where shading from other conifers is lacking (Bainbridge and Strong 2005, Ihalainen and Pukkala 2001, Kardell 1980, Timoshok 2000). Even in pure aspen, ericaceous shrubs occur where soil fertility is poor as they are outcompeted by other vascular plants when soil fertility is high (Chen et al. 2004, Hart and Chen 2006). An increase in site productivity associated with nutrient load, light, or tree density could explain the increase in Equisetum spp., H. lanatum, and R. idaeus, which are commonly associated with aspen or spruce (Beaudry et al. 2001, Kojima 1996, Lautenschlager 1999). Within the first 20 years post harvest, aggressive early seral pioneers (Equisetum spp., H. lanatum, and R. idaeus) that rapidly colonize available space via seeds, root suckers, or deep rhizomes were abundant (Bainbridge and Strong 2005, Beasleigh and Yarranton 1974, Hendrix 1984, Lautenschlager 1999). As the overstory canopy closed, R. idaeus shrubs decreased as expected (Lautenschlager 1999, Kardell 1980) yet fruit production remained high (Whitney 1982). Equisetum spp. also decreased, but was more abundant than non-harvested stands, a trend that has been observed with other herbaceous species (Visscher and Merrill 2009). An increase in H. lanatum suggested that species may be better able to balance water demands and reduce the chance of frost damage given its large leaf area in closed canopy forest (Young 1985). Our data suggest these forbs will eventually decrease to pre-disturbance levels as pine continues to age since most understory growth occurs within the first 30 years of clearcutting when resources (nutrients, light, moisture) are most available (Bainbridge and Strong 2005). Vaccinium spp. can be severely impacted by disturbance from fire or clearcutting because they propagate from rhizomes (Haeussler et al. 1999, Moola and Mallik 1998, Minore 1984, Nielsen et al. 2004c). However, following stand initiation from clearcut harvesting, V. myrtilloides and V. caespitosum spp. had recovered to pre-disturbance levels within 20 years of succession in our study area. Because fruit production was high, shrubs appeared to be locally abundant and not suppressed by competing vegetation. As with most ericaceous shrubs, their performance likely depends on the duration and intensity of available light (Moola and Mallik 1998). Although shrub cover may not be synonymous with fruit yield, less damage to rhizomes (Haeussler et al. 1999) could enhance fruit production in V. myrtilloides and V. caespitosum beyond what we predicted. Clearcutting severally reduced shrub cover and fruit production in V. membranaceum and V. vitisidaea and post disturbance recovery tended to be slow (Bainbridge and Strong 2005, Coxson and Marsh 2001, Minore 1979). However, we found a positive relationship with age. Recovery to predisturbance abundance is likely given that shading should not inhibit vegetative growth (Anzinger 2002, Hall and Shay 1981). Our finding that V. vitis-idaea fruit production peaked in pine stands older than 80 years (50% canopy cover) was consistent with Kardell (1980), yet fruit yields can be ample in young managed stands (Ihalainen and Pukkala 2001). In association with stand replacing wildfire and recurrent surface fire, V. membranceum fruit production tends to be highest under an open canopy (Anzinger 2002, Martin 1983, Minore 1979, Minore 1983). In certain years, fruits may be relatively abundant in mature stands (Kerns et al. 2004). In a similar species, V. globulare,

122

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting Martin (1983) found that shrub cover and fruit was low to nonexistent in scarified clearcuts less than 25 years old. Conversely, in stands burned by wildfire 25-60 years ago with 3% canopy closure, fruit production can be considerably higher compared to stands older than 60 years with 30% canopy closure and similar shrub cover (Martin 1983). With mechanical harvesting, fruit production tends to be higher than mature forests when stands are partially cut or broadcast burned (Anzinger 2002, Martin 1983). Alternatively, canopy removal without mechanical damage to shrubs appears to be optimal for V. membranaceum (Minore 1984). We found elevation to be an important factor that influenced shrub cover and fruit production at the species level. Considering the two most abundant fruiting species, at 1204-1228m there was a shift in dominance from low elevation V. myrtilloides to higher elevation V. membranceum. This corresponds with patterns found in other studies (La Roi and Hnatiuk 1980, Strong 2002). Elevation was a particularly important driver of V. membranaceum abundance, a pattern that also appears along the eastern slopes and areas of the USA. In Oregon, Anzinger (2002) found that a 450m increase in elevation (1200m-1600) corresponded to a one unit increase in V. membranaceum fruit production class, while in Montana, V. globulare fruit was limited at the extremes of a wide elevation (820-2120m) gradient, but production was pronounced between 1300-1850m (Martin 1983). Strong (2002) identified V. membranaceum as a dominant understory component in a lodgepole pine community type that occurs between 1050 and 1600m elevation along the eastern slopes. Plausible mechanisms controlling this phenomenon could be related to deeper winter snow packs and more growing season precipitation that could minimize spring frost damage (Minore and Smart 1978), promote plant growth (Anzinger 2002), or reduce interspecific competition (Martin 1983). Drought is often implicated in a lack of berry production (Hamer and Herrero 1987, Krebs et al. 2009). In 2009, we observed less precipitation during the growing season (June and July) that appeared to impact productivity in the Vaccinium spp. Annual precipitation maps showed that summer precipitation levels were between 90-130% from normal in 2008 and 70-90% from normal in 2009 (Alberta Environment 2012). Although simultaneous fruit failures are rare (McLellan and Hovey 1995), differences in annual fruit yield are well known in the Vaccinium spp. (Ihalainen et al. 2003, Kardell 1980, Martin 1983, Minore and Dubrasich 1978). Variation in R. idaeus fruit occurs (Noyce and Coy 1990) but crops may be more consistent from year to year (Costello and Sage 1998, Kardell 1980, Noyce and Coy 1990). The specific mechanisms controlling fruit production are poorly understood, yet temporal fluctuations in temperature and precipitation during and prior to the growing season and in previous years may alter plant bud growth and phenology (Bokhorst et al. 2008, Eriksson and Ehrlen 1991, Krebs et al. 2009, Selas 1990). For many species, meteorological factors affect berry production in the same way (Wallenius 1999). Spring frost, cold summer and winter temperatures, insufficient snow and rainfall, and windy conditions during pollination can influence crop yields (Howatt 2005). S. amplexifolius has an affinity for low soil moisture. We found an abundance shift from harvested to non-harvested stands in 2009, reflected in the strong contrast in soil moisture between habitats (Hart and Chen 2006, Redding et al. 2003).

123

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting Grizzly Bear Food Supply in the Upper Foothills Under optimal conditions for fruit production (2008), we predicted an overall negative change in food supply based on proposed plans to surge harvest mature (>80 years) stands of lodgepole pine. At the end of the surge cut in 2018, lodgepole pine habitats became increasingly more productive with a 25% increase in forbs and 2% gain in fruit production largely because of Equisetum, H. lanatum, R. idaeus, V. myrtilloides that were associated with young and old clearcuts. However, this will be a short term gain as habitat productivity will decline near the end of the 60 year planning period with the progression of harvesting and succession of the large cohort of young pine. Although forbs will be more abundant than pre-harvest conditions, fruit production will decline by 10% based on our model predictions. The initial increase in R. idaeus and V. myrtilloides may temporarily offset the decline of V. membranceum and V. vitis-idaea post surge cut. Over the long-term, the gain in V. myrtilloides alone would be enough to compensate for the loss of V. vitisidaea, but it would not make up for the reduction in V. membranceum. The benefits of the surge cut and increase in food supply for grizzly bear occurred mainly below 1228m of elevation. This was because V. myrtilloides was abundant below whereas V. membranceum was abundant above this point. Because recovery of shrubs to a reproductive phase is slow with secondary succession following clearcutting (Anzinger 2002, Kardell 1980, Martin 1983), harvesting primarily above 1228m will likely lead to a decline in V. membranceum over time. Overall, an increase in habitat productivity associated with forbs does not outweigh the potential negative effects of reduced fruit availability that may be linked to reproduction and density of interior grizzly bear populations (McLellan 2011). MPB harvesting is in its infancy so there is the potential to mitigate possible negative effects on grizzly bear food supply. Because an increase in forbs is imminent with harvesting more emphasis should be placed on managing habitat to optimize fruit production. The availability of energy dense fruit compared to other food types is hypothesized to be particularly important for females to gain body fat that could influence reproduction, cub survival, and have a disproportionate effect on population processes that influence density (Farley and Robbins 1995, Fellicetti et al. 2003, Robbins et al. 2007, Rogers 1976, McLellan 2011, Ferguson and McLoughlin 2000). To optimize food resourses for grizzly bears, the ideal MPB management scenario and mitigation measure would be to limit the harvesting of mature stands above 1228m. In doing so, the two dominant species, Vaccinium spp., V. membranceum and V. mytilloides, that can be major dietary components of grizzly bear would be maximized and at the same time would promote V. vitis-idaea (Larsen and Pigeon 2006, McLellan and Hovey 1995, Munro et al. 2006). In the spring, Munro et al. (2006) found that bears may forage on over wintered V. vitis-idaea as has been observed in other studies (Nilsen 2002). Not only would retaining mature stands above 1228m likely benefit grizzly bears, management would also have the opportunity to enhance V. membranceum fruit production by creating canopy openings (Minore 1984) or introducing surface fires (Trusler and Johnson 2008). Given the threat of MPB in the study area, retaining large areas of mature pine may not be feasible. As an alternative, â&#x20AC;&#x2DC;goodâ&#x20AC;&#x2122; patches within stands could be identified and retained in the pre-block layout and prior to clearcutting. However, from an operational perspective this may not be

124

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting realistic. Therefore, if harvesting continues as projected, broadcast burning clearcuts above 1228m may be necessary to improve fruit production in V. membranceum (Anzinger 2002, Martin 1983), although this technique may not be that effective (Minore 1979). In this case, maintaining more young clearcuts below 1228m by annually harvesting a proportion of older clearcuts and/or the 4180 age class may be another solution. Not only would this preclude the need for future surge cutting, but it would also mitigate the decline in forbs and fruiting species associated with an increase in the 41-80 year age class that according to our model predictions are most deficient in food resources. Spatially and temporally optimizing total potential fruit producing habitat may be prudent given the uncertainty of annual weather events that could impact Vaccinium spp. fruit yields.

125

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting Table 1. Mean dry weight of forbs (g/stem) and fresh weight of fruits (g/berry), sample size and stand error (S.E.), percent dry matter (D.M.) per berry, and percent dry matter digestible. Means and stand errors were derived from the plot level average weight (g/stem or g/berry) of each species.

126

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting Table 2. Environmental variables used to predict the distribution and abundance of grizzly bear foods in pine forest of the Upper Foothills Natural Subregion within Weyerhaeuser Grande Prairies Forest Management Agreement in north-west central Alberta, Canada.

127

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting Table 3. Grizzly bear food occurrence probabilities in pine, harvested (cut), and non-harvested (uncut) including odds ratios (OR), likelihood ratio tests (LR Ď&#x2021;2), model significance (p), and pseudo R2 from logistic regression analysis. The presence of bear foods was determined from meander searches within a 30x30m sampling area at 147 cut and 142 uncut field plots in pine.

Hedysarum spp. includes Hedysarum alpinum and Hedysarum boreale. Equisetum spp. represents Equisetum arvense, Equisetum pratense, and Equisetum sylvaticum.

128

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting Table 4. Sample size (n), null and final model AICc scores, number of model parameters (k), and percent deviance explained (D2) from zero-inflated negative binomial regression models describing the distribution (inflate) and abundance (count) of shrubs, fruits, and forbs. Inflate represents the model fit priori to adding covariates to the count portion of the model.

Percent deviance (D ) for the inflate model was represented as the reduction in the log likelihood from the 2 intercept only (null) model (k=3). D for the count model was determined by the reduction in the log likelihood from the â&#x20AC;&#x2DC;bestâ&#x20AC;&#x2122; inflate model.

129

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting

500

2500

450 2000

350

DDM (g/900m )

400

300

1500

250 200

1000

150 100

500

50 0 1-20

21-40

41-60

61-80

81-100

101-120 121-140

0 >140

Age20 (years) Equisetum spp.

S. amplexifolius

H. lanatum

Figure 1. Relationship between mean dry digestible matter (g/900m2) and stand age (Age20; 20 year intervals) for Equisetum spp. and H. lanatum in leading (solid) and non-leading (long dash) stands of lodgepole pine and in 2008 (long dash - dot) and 2009 (dot) for S. amplexifolius. Refer to the secondary y-axis for Equisetum spp.

130

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting

0.08

R. idaeus

0.07

V. myrtilloides

0.06

V. caespitosum

0.05

V. membranaceum

0.04

V. vitis-idaea

Cover (%/900m )

0.03 0.02 0.01 0 1-20

21-40

41-60

61-80

81-100

101-120

121-140

>140

101-120

121-140

>140

101-120

121-140

>140

Age20 (years)

B) 60

DDM (900m )

50 40 30 20 10 0 1-20

21-40

41-60

61-80

81-100

Age20 (years)

C) 60

DDM (900m )

50 40 30 20 10 0 1-20

21-40

41-60

61-80

81-100

Age20 (years)

Figure 2. A) Relationship between mean cover (% cover/900m2); B) mean dry digestible matter (g/900m2) of fruit in 2008; C) mean dry digestible matter (g/900m2) of fruit in 2009 and stand age (Age20; 20 year interval) for five fruit producing shrub species in leading (solid) and non-leading lodgepole pine (long dash).

131

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting A) 0.06

Cover (%/900m )

0.05 0.04 0.03 0.02 0.01 0 1

Elev

DDM (g/900m )

50 40 30 20 10 0 1

4 Elev

R. idaeus

V. myrtilloides

V. caespitosum

V. membranaceum

V. vitis-idaea

Figure 3. A) Relationship between mean shrub cover (%/900m2); and B) dry digestible matter (g/900m2) of fruit as a function of elevation (Elev) for five fruit producing shrub species in lodgepole pine.

132

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting

Figure 4. Total harvested area (hectares) of leading lodgepole pine over twelve harvesting periods by stand age (Age20) within the Upper Foothills of Weyerhaeuser Grande Prairie Forest Management Agreement in north-west central Alberta, Canada.

133

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting

Figure 5. Changes in dry digestible matter (DDM; Tonnes) of Equisetum spp, H. lanatum, S. amplexifolius (2008 long dash; 2009 dot), and total forbs over 12 harvesting periods within the Upper Foothills. Refer to the secondary y-axis for total forbs and Equisetum spp.

134

Chapter 4-Grizzly Bear Food Supply and Response to Forest Harvesting A)

DDM (Tonnes)

B) 60

120

110

100

60 1

Harvesting Period

DDM (Tonnes)

C) 60

0 1

Harvesting Period R. idaeus V. membranaceum

V. myrtilloides V. vitis-idaea

V. caespitosum Total Fruit

Figure 6. Changes in A) % cover of shrubs; B) dry digestible matter (DDM; Tonnes) of fruit under optimal conditions (2008); C) DDM of fruit under sub-optimal (2009) over 12 harvesting periods for five shrub species in the Upper Foothills. Refer to the secondary y-axis for total estimates.

135

Chapter 4-Bear Foods and Response to MPB management

TOPIC B: GRIZZLY BEAR RESPONSE Methods Grizzly Bear GPS Locations and Home Range Area Female grizzly bears (n=12) were captured between 2007 and 2010 by the Foothills Research Institute’s Grizzly Bear Program using helicopter darting, leg hold snares, or culvert traps. All bears were fit with a Televilt Tellus 2D brand GPS radio-collar (Televilt® TVP Positioning AB, Lindesberg, Sweden) programmed to collect locations at an hourly fix interval. We grouped GPS locations by individual bear, year, and seasonal period (hereafter, home range). Seasonal periods correspond to major dietary changes in grizzly bears based on scat analysis (Munro et al. 2006) and have been used in other studies of bear habitat selection. The spring (May 1 until June 15) represented hypophagia (low food intake), whereas the summer (June 16 to 31 July) and the fall (August 1 to October 15) represented early hyperphagia (high food intake) and late hyperphagia (Nielsen et al. 2009). Late hyperphagia characterizes the time of the year when energy intake and fat deposition is maximized (Robbins et al. 2007). The availability of seasonal fruit is hypothesized to be critical for females since the ability to gain body fat is essential to successful hibernation and cub production (Robbins et al. 2007). Using 47809 GPS locations, we calculated a utilization distribution (UD) for each grizzly bear home range (min=154, max=1244, median=828, mean=887) using fixed kernel analysis (Worton 1989). A UD is a relative probability surface that sums to 1 and describes both the spatial boundary (x-y axis) and intensity of use (z-axis) of GPS locations. Because the spatial distribution of our GPS data was partially clumped, we used the plug-in method to select the appropriate bandwidth (smoothing parameter) for each UD (Gitzen 2006). Smoothing parameter values and UD’s were calculated using the KS package (Duong 2011) and the statistical software R (R Development Core Team 2011) following the methods of Robinson et al. (2010). We use the Percent Volume Contour tool available in Hawth’s Tools (Beyer 2004) for ArcGIS 9.2 to extract the 95% volume from UD raster grids (30m pixel) for each home range, which corresponds to an area that, on average, 95% of GPS locations are found within. We used the home range area (mean=163km2, SD=121.6) to define available habitat (Johnson 1980).

Grizzly Bear Habitat We used a GIS forest inventory polygon database (net land base) provided by Weyerhaeuser to stratify grizzly bear habitat in our study area. For each year of grizzly bear data, we intersected the corresponding home ranges with the net land base and removed polygons where overstory tree species composition and age were not known (2%; naturally non-vegetated, slivers) including all polygons classified as anthropogenic (4%; roads, well sites, pipelines, seismic lines and other land use dispositions), or non-forested (2%). Forested stands were either operable (82%) or inoperable with the latter distinguished by water buffers (7%) or steep slopes (5%). Because stand age and overstory composition was based on updated inventory data for 2009, we back casted harvested stand attributes to pre-harvest values for home ranges circa 2007 and 2008 where applicable. We stratified forested stands according to operable and inoperable, leading pine (>60% overstory

136

Chapter 4-Bear Foods and Response to MPB management composition) and other forest types, and by age class: A) 1-20 years; B) 21-36 years; C) 37-80 years; D) >80 years. The age classes represent A) young clearcuts and relative open canopy conditions; B) old clearcuts and relative closed canopy conditions; C) non-harvested and rare on the landscape; D) non-harvested mature, common on the landscape, and the focus of MPB management. We defined the harvested area boundary (edge) as the polygon line that separated harvested from adjacent non-harvested or harvested stands according to age class. We calculated the area of each habitat strata by home range using Hawth’s Tools (Beyer 2004). Grizzly Bear Habitat Use, Availability, and Selection Of the 47892 GPS locations used to define grizzly bear home ranges, 8815 overlapped with polygons classified as anthropogenic and 1113 with non-vegetated. We intersected the remaining 37879 locations that overlapped forested grizzly bear habitat. About 68 and 13% of locations occurred within the Upper and Lower Foothills Natural Regions with the remaining 19% in the Subalpine Natural Sub Regions. Locations were sub divided into three datasets according to our defined habitat strata: 1) operable (31664) and inoperable stands (3643 riparian buffers; 2572 steep slopes); 2) operable stands only and distinguished by leading pine (19720) and other forest types (11994); and 3) stands that were leading pine only and distinguished by age class. We generated 3 random locations (available) within the area of leading pine by home range for each grizzly bear location (19720) using Hawth’s Tools (Beyer 2004). Stand age and distance to the nearest edge was determined from the habitat stratification, while elevation was obtained from a 30m pixel Digital Elevation Model (DMTI 2003). Because the 37-80 year age class was relatively rare, locations in non-harvested were grouped into a single age class (≥ 37 years). We used a binary category to distinguish locations that were 70m from the nearest edge from locations in the interior as our fourth dataset. We used 70m since this cut off point roughly corresponds to edge effects related to forest understory composition and structure (up to 60m) (Harper and MacDonald 2002) and the lower end of GPS error (10-28m) commonly reported (Friar et al. 2010) . We used a binary category to distinguish locations that were above or below 1228m of elevation as our fifth dataset. This cut off point represents a change in fruit abundance between the two dominant Vaccinium spp. in the system (I. Food Supply). The mean (1223m) or median (1210m) number of bear use locations by elevation (Min=770, Max=1771) in leading pine generally corresponds to this cut off point. We calculated the proportion of bear locations and electivity (Ivlev 1961) in each habitat class by home range as our measures of use and selection. Electivity ([E i = (oi - pii)/(oi + pii)]; where i indicates the i th habitat, oi is the use proportion of habitat i, and πi is the availability of habitat i) is equivalent to the selection ratio used in RSF models (Manly et al. 2002) (wi = oi / pii), however, values of electivity are constrained between -1 and +1 whereas selection ratios are bound between - / + infinity. Therefore, electivity may offer an advantage when averaging across the population because the mean and confidence intervals should be less sensitive to high and low values (Jacobs 1974). Interpretation follows that when a habitat is selected electivity values are positive, negative denotes habitat avoidance, and zero represents random use of habitat (Strauss 1979). Availability

137

Chapter 4-Bear Foods and Response to MPB management was either calculated as the proportion of habitat area within the home range (datasets 1-3) or the proportion of random locations (datasets 4-5). Because we retained the individual animal as the sampling unit, our analysis followed a design 3 approach to habitat selection (Thomas and Taylor 1990). We investigated grizzly bear habitat use and selection at the 3rd order or ‘patch’ level (Johnson 1980) because management decisions are most often made at this scale (Nielsen et al. 2002). Our assumption was that habitats were equally available to individual animals during the temporal period of sampling (home range) (Boyce 1999).

Grizzly Bear Use and Selection of Habitat at the Population Level We used a univariate analysis to derive population level models of use and selection of habitat by season from our five datasets. We only considered observations where habitat availability was greater than 5% by home range in an attempt to minimize the effect of rare habitats and potential bias associated with non-zero replacement (Bingham and Brennan 2004) that has been used in compositional analysis (Aebischer 1993). We used a generalized linear model and the statistical software Stata (StataCorp 2009) to predict the mean and 95% confidence intervals of our dependent variables (use and selection) by interacting season with our categorical habitats for datasets 1-3. We also modelled habitat availability to aid in our interpretation. To meet the assumption of normality, we performed the logit transformation on the use and available proportions. We specified the robust cluster option (bear) to calculate variance using the HuberWhite sandwich estimator (White 1980) and weighted each observation by the inverse variance to account for precision (Thomas and Taylor 2006). Given the repeated measures of bears over multiple years (bear years hereafter BY), this appropriately assigned the unit of replication to the individual animal and avoided issues of pseudoreplication (Hurlbert 1984). We repeated this procedure for datasets 4-5, but used a three-way interaction between season, the redefined age class (1-20; 21-36; ≥37 years), and the binary classification of elevation (<1250m≤) or edge (≤70m<). For these population level models, we only considered home range observations when each age class of edge or elevation categories were both available. Because of the unit sum constraint and since not all habitats were available or used by bears (Aebischer 1993), we limited our inference to describing differences between the mean and variance of habitats used and selected by female grizzly bear. We accepted significant differences between habitats when confidence intervals did not overlap. Habitats were used at availability when confidence intervals did not overlap zero.

Results At the population level, we found no difference between the seasonal use and selection of operable or inoperable habitats, but female grizzly bear spent significantly more time (85%) in the operable land base (Table 5). Variation in use and selection was most pronounced during the spring and fall. In the fall, females used inoperable forest 11% more than in the spring/summer and corresponded to selection. Although confidence intervals overlapped compared to the spring, the fall significantly differed from the summer. Average selection in the spring/summer was lower (0.06) compared to the fall (0.30) with positive selection in 15 of 18 BY. Operable forest accounted for 89% of the available habitat on average.

138

Chapter 4-Bear Foods and Response to MPB management Compared to spring, females spent more time in leading stands of pine during the summer and fall. However, confidence intervals did not overlap with other stand types (Table 5). Although confidence intervals overlapped zero, in the spring females appeared to select for other forested stands (14 of 19 BY). On average, leading pine was the most used (59%) and available (61%) habitat. When female grizzly bears used pine leading stands, young clearcuts (1-20 years) were used less in the spring with a trend of increased use in the summer and fall, but confidence intervals overlapped (Table 6). This age class was used at availability in the spring and selected in the summer (15 of 22 BY) and fall (10 of 15 BY) (Table 6). With a 9% increase in use during the fall, selection more than doubled (0.28) compared with the summer (0.12) and was significantly higher than the spring season. Variation in the use/selection of this age class was most pronounced during the fall, but the lower end of the confidence limits were the same or higher than compared to summer. The use pattern of older harvested (21-35 years) stands by females was the opposite of young clearcuts. Use was highest in the spring, decreased slightly in summer and substantially more in the fall (Table 6). Grizzly bears used older harvested stands at availability in the spring/summer and avoided (7 of 10 BY) this age class in the fall (Table 6). Although the use of nonharvested (37-80 years) stands by females increased from spring to fall, use was significantly less than any other age class. Females avoided these stands in spring (9 of 10 BY) and used them at availability in the summer and fall. Grizzly bears used mature stands consistently in each season and more than any other age class (Table 6). Variation in use was high and confidence intervals overlapped harvested stands, but the lower limit was greater than what was predicted in harvested stands (Table 6). In the spring, females used mature stands in proportion to availability and avoided them in summer (16 of 23 BY) and fall (10 of 16 BY). On average, females spent 33% of their time in young clearcuts, 28% in older clearcuts, 8% in the 37-80 age class, and 47% in mature stands. On average, the availability of young and older harvested stands was comparable, while mature stands were the most available habitat in leading pine stands. Irrespective of age and season, female grizzly bears spent more time in clearcuts below 1228m (Table 7). Below 1228m, young clearcuts were used consistently by females and selected in the spring (7 of 12 BY), the summer (8 of 12 BY), and the fall (6 of 8 BY) (Table 7). Above this point, young clearcuts were used less in the spring with a trend of increased use in summer and fall, while use was proportional to availability in each season (Table 7). The use pattern of older clearcuts between seasons was consistent above and below 1228m (Table 7). Below 1228m, bears selected older clearcuts in the spring (5 of 7 BY) and use was proportional to availability in the summer/fall (Table 7). Above 1228m, the older clearcuts were used at availability in spring/summer and avoided in the fall (6 of 7 BY) (Table 7). The majority of variation in use by elevation can be attributed to non-harvested stands (â&#x2030;Ľ37 years). Below 1228m, females used non-harvested more in the spring with a trend of decreased use in the summer and fall (Table 7). This pattern reversed above 1228m with use being substantially higher in the fall compared to the spring (Table 7). In the spring and below 1228m, females used non-harvested stands at availability and avoided this age class in the summer (14 of 16 BY) and fall (9 of 10 BY) (Table 7). Again, the pattern flipped above

139

Chapter 4-Bear Foods and Response to MPB management 1228m as females avoided these stands in the spring (11 of 11 BY) and used them at availability in the summer and fall (Table 7). Non-harvested stands were the most available habitat (Table 7).

Discussion Inoperable areas distinguished by riparian buffer zones and steep slopes were used considerably less than operable habitats, but selected by female grizzly bears in the fall. Because they are often selected, riparian areas are often cited as important early and late season habitats (McLellan and Hovey 2001, Nielsen et al. 2002, Zager et al. 1983). Zager et al. (1983) found that grizzly bear selected creek bottoms in the spring and summer/fall, but compared to avalanche chutes (spring) and timbered shrub fields (summer/fall) use was substantially less. In our study area, increased use of inoperable may reflect annual variation in fruit production (McLellan and Hovey 2001, Zager et al. 1983). Bears may forage on Hedysarum spp. roots, forbs, and ungulates in riparian areas or the uplands whereas frugivory tends to be exclusive to the uplands (Hamer 1999, McLellan and Hovey 2001, Munro et al. 2006, Zager et al. 1983) and abundant in operable habitats depending on the year (I. Food Supply). Field observations in our study area confirm that grizzly bears may dig roots on steep dry south/west facing slopes and riparian areas, and tends to be less frequent given that it was rare within operable habitats (I. Food Supply, T. A. Larsen, unpublished data). Shrub species such as Shepherdia canadensis and Arctostaphylos uva-ursi may be associated with dry and/or steep terrain features (Hamer 1996, Hamer and Herrero 1987) and could be more common on steep slopes even though they were rare in operable habitats (I. Food Supply). Although riparian areas certainly are used and particularly in the fall, on an annual basis female grizzly bears meet the vast majority of their life history requirements within the operable land base even though these habitats were not selected. When using the operable land base, our results indicate that female grizzly bear spent more time in pine dominant habitats during the hyperphagic period (summer/fall) when food is most available (Nielsen et al. 2009). The use of other forested stands in the spring may be related to bears foraging on Equisetum spp. as this species tends to be more abundant at the microsite (30x30m) level (I. Food Supply) and in the region, can be an important dietary component (Larsen and Pigeon 2006). Hamer and Herrero (1987) found that grizzly bears foraged on Equisetum arvense in mature spruce forests but not exclusively. H. lanatum was also more abundant in other forest types (I. Food Supply), but females used leading pine more during the summer when bears tend to consume this food resource (Larsen and Pigeon 2006, Munro et al. 2006, McLellan and Hovey 1995). Use of leading pine during the fall was consistent with abundant Vaccinium spp. fruit (I. Food Supply). Pelchat and Ruff (1986) attributed the use of pine by black bears (Ursus americanus) to abundant Vaccinium spp. However, the similarity in which pine was used during hyperphagia and since harvested stands are primarily replanted to lodgepole pine, use may be more related to clearcuts than observed differences between stand types per se. Our broad comparison of stand types at the patch level may not be the appropriate scale that bears use/select food resources associated with canopy composition (Hamer and Herrero 1987, Mattson 1997, Nielsen et al. 2004a) and may require examining use/selection at the microsite scale (30x30m) (Johnson 1980). Regardless, bears

140

Chapter 4-Bear Foods and Response to MPB management spent more time in pine leading stands during the critical foraging period even these habitats were not selected. We found that when females used pine leading stands within the operable land base, seasonal patterns of use/selection differed by age class. The use of young (1-20 years) clearcuts increased with the progression of the growing season and peaked during the summer with the opposite pattern occurring in older (21-35 years) harvested stands. In the summer and fall, young clearcuts were selected and older clearcuts were avoided in the fall. These patterns reflected variation in the seasonal abundance plant based food resources (forbs and fruits) in the system. Forbs such as Equisetum spp. and H. lanatum increased considerably following timber harvest, while fruits of V. myrtilloides, V. caespitosum, and R. idaeus were abundant in the 1-20 age class. Greater variation in bear use/selection of young clearcuts during the fall may be related to annual pulses in Vaccinium spp. fruit availability (I. Food Supply). Because use/selection was equivalent or higher compared to the summer, bears may use alternative food resources in young clearcuts including forbs, ants, or ungulates (I. Food Supply, Nielsen et al. 2004c). Patterns were consistent with a study in west-central Alberta as grizzly bears selected younger clearcuts (<10 years) in the fall and older harvested (>10 years) during the spring. At the same time, bears also selected harvested stands in the fall that were older (36-46 years) than what was available to grizzly bear in our study area. Although responses were similar, certain foods (i.e. hedysarum spp. and S. Canadensis) associated with these age classes differed between study areas suggesting that regardless of the specific foods available, clearcuts are seasonally productive habitats for grizzly bears. The benefits of MPB harvesting associated with an increase in forbs and fruits may be a short term gain as fruit production in the system is predicted to decline. This decline is associated with a reduction in mature stands and a substantial increase in the 37-80 year age class (I. Food Supply). However, given that this age class was the least used and available habitat, our ability to assess grizzly bear response was limited. Overall, spring use was low and bears avoided this age class, but in the summer and fall use increased and was proportional to availability. At the same time, mature stands were the most available and used habitat by females. Unlike harvested habitats, mature stands were used consistently among the seasons, but females avoided this age class during the summer and fall. Because of this and since variation in use was high, harvesting these stands may be beneficial given the positive response to clearcutting. However, matures stands offer foraging opportunities related to fruiting shrub species (V. membranceum and V. vitis-idaea) that are predicted to decline. Furthermore, mature stands may be important during drought years since moisture deficits in clearcuts could influence plant phenology, forb distribution, and fruit abundance (I. Food Supply). In addition, mature stands may also be used for thermal and/or hiding cover near foraging sites or when resting (Munro et al. 2006). Grizzly bears not only used leading pine habitat seasonally according to age, but use differed by age associated changes in elevation. Irrespective of season, young clearcuts were favoured substantially by grizzly bears below 1228m and to a lesser extent above this point in summer and fall. On the other hand older harvested stands appear to benefit bears the most at lower elevations given that they are used less at higher elevations and particularly in the fall when bears avoided this

141

Chapter 4-Bear Foods and Response to MPB management age class. Grizzly bears appeared to shift from low to high elevations that can be attributed to phenological changes in forbs (Hamer and Herrero 1987). This is because females are better able to capitalize on plants when they are in an immature stage of development and higher in nutritive value (Hamer and Herrero 1987, McLellan and Hovey 1995). In our system, consistent use of lower elevation pine may be a foraging strategy whereby bears switch to alternative foods as they become available, while others may track the phenological stages of specific species with elevation (Hamer and Herrero 1987). Individuals may also use different fruiting species associated with elevation (Hammer and Herrero 1987, Raine and Kansas 1990). However, use of fruit may be unrelated to phenology as species tend to occur at distinct bands of elevation and are dependent on local environmental conditions that influence fruit production (age and canopy cover) (Hammer and Herrero 1987, Raine and Kansas 1990). Although not selected, the exceptionally high use of non-harvested stands above 1228m supports the contention that V. membranceum fruit is a potentially important food resource in the system (I. Food Supply). In addition, other fruiting species such as E. nigrum or Vaccinium scoparium or lupinus spp. may also be associated with these mature pine forests at higher elevation (Graber and White 1983, Raine and Kansas 1990, T. A. Larsen, unpublished data). Although plant phenology and the abundance of fruit relative to elevation appear to be key elements influencing bear use of pine habitat, reduced productivity (tree growth) with elevation (Wang et al. 2004) may explain lower use of harvested stands by females above 1228m. Population measures of use and selection provided complimentary information that aided in our interpretation of seasonal habitat associations and response of female grizzly bear to forest harvesting. As expected, the most available habitats such as mature pine were either used at availability or avoided, while those that were the least available tended to be used at availability or selected (Bayer et al. 2010). The dichotomy of resource use and selection is important and has received some attention in the literature (Long et al. 2009), yet measures of use and selection are rarely incorporated in the same study. Because selection alone may cloud important patterns of habitat use, we advocate that both measures should be integrated into habitat analysis. However, it should be cautioned that high use and selection may not be indicative of necessary habitat nor does avoidance signify unsuitable habitat (Pendleton 1998). These are assumptions that need to be tested (Pendleton 1998). Ultimately, there is the need to link grizzly bear response and the individual contributions of seasonal food resources to the health and reproductive performance of grizzly bear populations to better manage habitats relative to forest management planning.

Conclusions Lodgepole pine stands are associated with seasonally important fruit producing Vaccinium spp. The surge harvest and reduction of mature pine for MPB management led to a substantial increase in available forbs and fruits. However, the boost in habitat productivity and possible benefit to grizzly bear was short lived and merely offset the larger problem of a decline in fruit producing habitat at the landscape level. Because elevation spatially separated the two dominant fruit producing Vaccinium spp. in the system, not only is there an opportunity to maintain a seasonally critical bear food (V. membranceum) in mature pine, but also maximize fruit production given the predicted

142

Chapter 4-Bear Foods and Response to MPB management increase in V. myrtilloides associated with young clearcuts. If these same shrub species dominate the Upper Foothills across the eastern slopes of Alberta as evidence suggests, a provincial decline in fruit producing habitat may occur with the surge cut. Pine stands are associated with other fruiting producing shrub species that may be locally important to grizzly bear in other regions. Provincial inventories are likely necessary to fully assess the impacts of MPB management in Alberta to grizzly bear habitat supply. Grizzly bear spent most of their time in leading pine stands within the operable land base. When using these stands, females responded positively to harvesting, however, this was dependent on the associations between season, age class, and elevation. Bears responded positively to young clearcuts during the summer and fall. Conversely, older clearcuts tend to be favoured during the spring and summer. Although a positive response by females suggested that harvesting mature pine forest would benefit grizzly bears, caution is warranted since mature stands that were used the most were also the most available, and tended to be used at availability or avoided. However, when accounting for differences associated with elevation, it appeared that young and old clearcuts were most beneficial to grizzly bears below 1228m across seasons, while mature stands were used the most during the fall above this point. These results support our other finding that a negative change in food supply associated with harvesting mature stands may occur above 1228m. Although challenging given the need to harvest for MPB mitigation, management should focus on ways to maintain the seasonal habitat value of these higher elevation pine stands.

This report encompasses some of the major findings of a M.Sc. thesis. Additional analyses, results, and discussion on these topics can be found within the actual thesis dissertation.

Acknowledgements We gratefully acknowledge the Foothills Research Institute Grizzly Bear Program for financial and logistic support throughout the project. Additional funding support was provided by Alberta Conservation Association (ACA), Alberta Sports, Parks, Recreation, and Wildlife Foundation, Circumpolar/Boreal Alberta Research Grants, Forest Resource Improvement Association of Alberta, Yellowstone to Yukon, and Weyerhaeuser Grande Prairie LTD.

143

Chapter 4-Bear Foods and Response to MPB management

Table 5. Female grizzly bear population level use, selection, and availability of habitat by season, operability, and stand types.

Means and confidence intervals were estimated using individual bear as the cluster variable. Sample sizes are according to the number of bears during the spring (n=10), summer (n=12), and fall (n=9).

144

Chapter 4-Bear Foods and Response to MPB management

Table 6. Female grizzly bear population level use, selection, and availability of habitat by season and age class.

Means and confidence intervals were estimated using individual bear as the cluster variable. Sample sizes are according to the number of bears during the spring (n=10), summer (n=12), and fall (n=9).

145

Chapter 4-Bear Foods and Response to MPB management

Table 7. Female grizzly bear population level use, selection, and availability of habitat by season, elevation, and age class.

Means and confidence intervals were estimated using individual bear as the cluster variable. Sample sizes are according to the number of bears during the spring (n=8), summer (n=10), and fall (n=9).

146

Chapter 4-Bear Foods and Response to MPB management

Literature Cited Aebeisher, N. J., Robertson, P. A., Kenward, R. E., 1993. Compositional analysis of habitat use from animal radio-tracking data. Ecology. 74, 1313-1325. Alberta Environment. 2012. Government of Alberta: Environment and Water, Maps and Data Summaries. Available at http://environment.alberta.ca/forecasting/reports/index.html. Anon. 2003. Alberta Climate Prediction Model (ACPM) to provide climate predictions for any location in Alberta from its geographic coordinates. Unpublished report to Alberta Environment/Sustainable Resource Development, Edmonton, Alberta. Anonymous, 2006a. Appendix A - Grande Prairie Forest Management Area - Land Base Assignment. Weyerhaeuser Company Ltd., Grande Prairie, Alberta. Anonymous, 2006b. Appendix D - Mountain pine beetle detailed report â&#x20AC;&#x201C; Revised timber supply analysis 2004-2014. Weyerhaeuser Company Ltd., Grande Prairie, Alberta. Anonymous, 2007. Mountain pine beetle management strategy for Alberta. Sustainable Resource Development. Government of Alberta, Edmonton, Alberta. Anzinger, D. L. 2002. Big huckleberry (Vaccinium membranaceum Dougl.) Ecology and Forest Succession, Mt. Hood National Forest and Warm Springs Indian Reservation, Oregon. M.Sc. Thesis. Oregan State University, USA. Bainbridge, E. L., Strong, W. L., 2005. Pinus contorta understory vegetation dynamics following clearcutting in west-central Alberta, Canada. Forest Ecology and Management. 213, 133150. Beaudry, L., Martin, M., Paczkowski, J. 2001. Using silviculture to maintain and enhance grizzly bear habitat in six variants of the Prince George Forest Region. Habitat Branch, Ministry of Environment, Lands and Parks. Victoria, British Columbia. 1-58. Beasleigh, W. J., Yarranton, G. A., 1974. Ecological strategy and tactics of Equisetum sylvaticum during a postfire succession. Can. J. Bot. 52, 2299-2318. Beyer, H. L., 2004. Hawthâ&#x20AC;&#x2122;s Analysis Tools for ArcGIS. Available at http://www.spatialecology.com/htools. Beyer, H. L., Haydon, D. T., Morales, J. M., Frair, J. L., Hebblewhite, M., Mitchell, M., Matthiopoulos, J. 2010. The interpretation of habitat preference metrics under use-availability designs. Phil. Trans. R. Soc. 365, 2245-2254. Bingham, R. L., Brennan, L. A., 2004. Comparison of type I error rates for statistical analyses of resource selection. J. Wildlife Manage. 68, 206-212. Bokhorst, S., Bjerke, J. W., Bowles, F. W., Melillo, J., Callaghan, T. V., Phoenix, G. K., 2008. Impacts of extreme winter warming in the sub-Arctic: growing season responses of dwarf shrub heathland. Global Change Biology. 14, 2603-2612. Boyce, M. S., McDonald, L. L., 1999. Relating populations to habitats using resource selection functions. Tree. 14, 268-272.

147

Chapter 4-Bear Foods and Response to MPB management Burnham, K. P, Anderson, D. R., 2002. Model Selection and Multi-Model Inference. 2nd Ed., Corr. 3rd printing, 2002, XXVI, pp. 488. Chen, H. Y. H., Legare, S., Bergeron, Y. 2004. Variation of the understory composition and diversity along a gradient of productivity in Populus tremuloides stands of northern British Columbia, Canada. Can. J. Bot. 82, 1314-1323. Costello, C. M., Sage, Jr., R. W., 1994. Black bear habitat selection from food abundance under 3 forest management systems. Int. Conf. Bear Res. and Manage. 9, 375-387. Day, R. J. 1972. Stand structure, succession, and use of Southern Albertaâ&#x20AC;&#x2122;s Rocky Mountain Forest. Ecology. 53, 472-478. Duong, T., 2008. KS: Kernel smoothing. R package versison 1.5.9. Ellison, L. 1958. Production of herbaceous vegetation in openings and under canopies of western Aspen (Populus tremuloides). Ecology. 39, 337-345. Eriksson, O., Ehrlen, J., 1991. Phenological variation in fruit characteristics in vertebrate-dispersed plants. Oecologia. 86, 463-470. Evans, J. 2004. Compound Topographic Index. Available at http://arcscripts.esri.com/details.asp?dbid=11863 Farley, S. D., Robbins, C. T., 1995. Lactation, hibernation, and mass dynamics of American black bears and grizzly bears. Can. J. Zool. 73, 2216-2222. Felicetti, L. A., Robbins, C. T., Shipley, L. A., 2003. Dietary protein content alters energy expenditure and composition of the mass gain in grizzly bears (Ursus arctos horribilis). Physiological and Biochemical Zoology. 76, 256-261. Ferguson, S. H., McLoughlin, P. D., 2000. Effect of energy availability, seasonality, and geographic range on brown bear life history. Ecography. 23, 193-200. Friar, J. L., Fieberg, J., Hebblewhite, M. Cagnacci, F., DeCesare, N. J., Pedrotti, L. 2010. Resolving issues of imporecise and habitat-biased locations in ecological analysis using GPS telemetry data. Phil. Trans. R. Soc. 365, 2187-2200. Gitzen, R. A., Millspaugh, J. J., Kernohan, B. J., 2006. Bandwidth selection for fixed-kernel analysis of animal distributions. J. Wildlife Manage. 70, 1334-1344. Graber, D. M., White, M., 1983. Black bear food habits in Yosemite National Park. Conf. Bear. Res. Manage. 5, 1-10. Haeussler, S., Bedford, L., Boateng, J. O., MacKinnon, A. 1999. Plant community responses to mechanical site preparation in northern interior British Columbia. Can. J. For. Res. 29,1084-1100. Hall, I. V., Shay, J. M. 1981. The Biological Flora of Canada. 3. Vaccinium vitis-idaea L. var. minus Lodd. Supplementary Account. Canadian Field-Naturalist. 95, 434-464. Hamer, 1999. Forest fireâ&#x20AC;&#x2122;s influence on yellow hedysarum habitat and its use by grizzly bears in Banff National Park, Alberta. Can. J. Zool. 77, 1513-1520. Hamer, D., Herrero, S., 1987. Grizzly bear food and habitat in the front ranges of Banff National Park, Alberta. Int. Conf. Bear Res. and Manage. 7, 199-213.

148

Chapter 4-Bear Foods and Response to MPB management Harper, K. A., Macdonald, S. E., 2002. Structure and composition of edges next to regenerating clear-cuts in mixed-wood boreal forest. Journal of Vegetation Science. 13, 535-546. Hart, S. A., Chen, H. Y. H., 2006. Understory vegetation dynamics of North American boreal forests. Critical Reviews in Plant Sciences. 25, 381-397. Hendrix, S. 1984. Reactions of Heracleum lanatum to floral herbivory by Depressaria Pastinacella. Ecology. 65, 191-197. Hobson, K. A., McLellan, B. N., Woods, J. G. 2000. Using stable carbon (X13C) and nitrogen (x15N) isotopes to infer tropic relationships among black and grizzly bears in the upper Columbia River basin, British Columbia. Can. J. Zool. 78, 1332-1339. Howatt, S. 2005. Crop profile for wild blueberry in Canada. Pestcide Rick Reduction Program, Pest Management Centre, Agriculture and Agri-Food Canada. 1-39. Hurlbert, S. H., 1984. Pseudoreplication and the design of ecological field experiments. Ecological Monographs. 54, 187-211. Ihalainen, M., Pukkala, T. 2001. Modelling cowberry (Vaccinium vitis-idaea) and bilberry (Vaccinium myrtillus) yields from mineral soils and peatlands on the basis of visual field estimates. Silva Fennica. 35, 329-340. Ihalainen, M., Sala, K., Pukkala, T., 2003. Empirical prediction models for Vaccinium myrtillus and V. vitis-idaea berry yields in North Karelia, Finland. Silva Fennica. 37, 95-108 Jacobs, J., 1974. Quantitative measurement of food selection: a modification of the forage ratio and Ivlevâ&#x20AC;&#x2122;s electivity index. Oecologia. 14, 413-417. Johnson, D. H., 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology. 61, 65-71. Kardell, L., 1980. Occurrence and production of bilberry, lingonberry, and raspberry in swedensâ&#x20AC;&#x2122;s forests. Forest Ecol. Manage. 2, 285-298. Keating, K. A., Gogan, P. J. P., Vore, J. M., Irby, L. R., 2007. A simple solar radiation index for wildlife habitat studies. J. Wildlife Manage. 71, 1344-1348. Kerns, B. K., Alexander, S. J., Bailey, J. D. 2004. Huckleberry abundance, stand Conditions, and use in western Oregon: Evaluating the role of forest management. Economic Botany. 58, 668678. Kojima, S. 1996. Ecosystem types of Boreal forest in the North Klondike River Valley, Yukon Territory, Canada, and their productivity potential. Environmental Monitoring and Assessment. 39, 265-281. Krebs, C. J., Boonstra, R., Cowcill, K., Kenney, A. J., 2009. Climatic determinants of berry crops in the boreal forest of the southwestern Yukon. Botany. 87, 401-408. Larsen, T. A., Piegon, K. P., 2006. The diet of grizzly bears (Ursus arctos) in North-Central Alberta, Canada, part of a new food based model approach to mapping grizzly bear habitat. In. Stenhouse, G. S., Graham, K. (eds). Foothills Model Forest Grizzly Bear Research Program 2006 Annual Report. Hinton, Alberta. Lautenschlager, R. A. 1999. Environmental resource interactions affect red raspberry growth and its competition with white spruce. Can. J. For. Res. 29, 906-916.

149

Chapter 4-Bear Foods and Response to MPB management La Roi, G. H., Hnatiuk, R. J. 1980. The Pinus contorta forests of Banff and Jasper National Parks: a study in comparative synecology and syntaxonomy. Ecological Monographs. 50, 1-29. Long, J. S., Freese, J., 2005. Regression models for categorical outcomes using Stata. Second Edition. College Station, TX: Stata Press. Long, R. A., Muir, J. D., Rachlow, J. L., Kie, J. G. 2009. A comparison of two modeling approaches for evaluating wildlife-habitat relationships. J. Wildlife Manage. 73, 294-302. Lopez, L. M., Harper, K. A., Drapeau, P. 2006. Edge influence on forest structure in large forest remnants, cutblock separators, and riparian buffers in managed black spruce forests. Ecoscience. 13, 226-233. Mace, R. D., Waller, J. S., Manley, T. L., Ake, K., Wittinger, W. T., 1999. Landscape evaluation of grizzly bear habitat in western Montana. Conservation Biology. 13, 367-377. Manly, B. F. J., McDonald, L. L., Thomas, D. L., McDonald, T.L., Erickson, W. P., 2002. Resource selection by animals: statistical design and analysis for field studies, 2nd ed. Kluwer Academic Publishers, Dordrecht, the Netherlands. Martin, 1983. Factors influencing globe huckleberry fruit production in northwestern Montana. Int. Conf. Bear Res. and Manage. 5, 159-165. Martin, T. G., Wintle, B. A., Rhodes, J. R., Kuhnert, P. M., Field, S. A., Low-Choy, S. J., Tyre, A. J., Possingham, P., 2005. Zero tolerance ecology: improving ecological inference by modelling the source of zero observations. Ecology Letters. 8, 1235-1246. McLellan, B. N., 2011. Implications of a high-energy and low-protein diet on the body composition, fitness, and competitive abilities of black (Ursus americanus) and grizzly (Ursus arctos) bears. Can. J. Zool. 89, 546-558. McLellan, B. N., Hovey, F. W., 1995. The diet of grizzly bears in the Flathead River drainage of southeastern British Columbia. Can. J. Zool. 73, 704-712. McLellan, B. N., Hovey, F. W., 2001. Habitats selected by grizzly bears in a multiple use landscape. J. Wildlife Manage. 65, 92-99. Minore, D., Smart, A. W., 1978. Frost tolerance in seedlings of Vaccinium membranaceum, Vaccinium globulare, and Vaccinium deliciosum. Northwest Science. 52, 179-185. Minore, D. 1984. Vaccinium membranaceum berry production seven years after treatment to reduce overstory tree canopies. Northwest Science. 58, 208-212. Minore, D., Dubrasich, M. E. 1978. Big huckleberry abundance as related to environment and associated vegetation near Mount Adams, Washington. USDA Forest Service Research Note INT-240. Portland, OR: Pacific Northwest Forest and Range Experiment Station. 1-8. Minore, D., Smart, A. W., 1979. Huckleberry ecology and management research in the Pacific Northwest. USDA Forest Service General Technical Report PNW-93. Portland, OR: Pacific Northwest Forest and Range Experiment Station. Portland, Oregan. 1-50 Moola, F. M., Mallik, A. U. 1998. Morphological plasticity and regeneration strategies of velvet leaf blueberry (Vaccinium myrtilloides Michx.) following canopy disturbance in boreal mixedwood forests. Forest Ecology and Management. 111, 35-50.

150

Chapter 4-Bear Foods and Response to MPB management Munro, R. H. M., Nielsen, S. E., Price, M. H., Stenhouse, G. B., Boyce, M. S., 2006. Seasonal and diel patterns of grizzly bear diet and activity in west-central Alberta. Journal of Mammalogy. 87, 1112-1121. Natural Regions Committee 2006. Natural Regions and Subregions of Alberta. Compied by D. J. Downing and W. W. Pettapiece. Government of Albeta. Pub. No. T/852. Nielsen, S. E., Boyce, M. S., Stenhouse, G. B., 2004a. Grizzly bears and forestry I. Selection of clearcuts by grizzly bears in west-central Alberta, Canada. For. Eco. and Manage. 199, 5165. Nielsen, S. E., Boyce, M. S., Stenhouse, G. B., Munro, R. H. M., 2002. Modeling grizzly bear habitats in the yellowhead ecosystem of Alberta: taking autocorrelation seriously. Ursus. 13, 4556. Nielsen, S. E., Cranston, J., Stenhouse, G. B., 2009. Identification of priority areas for grizzly bear conservation and recovery in Alberta, Canada. Journal of Conservation Planning. 5, 38-60. Nielsen, S. E., Herrero, S., Boyce, M. S., Mace, R. D., Benn, B., Gibeau, M. L., Jevons, S., 2004b. Modelling the spatial distribution of human-caused grizzly bear mortalities in the Central Rockies ecosystem of Canada. Biological Conservation. 120, 101-113. Nielsen, S. E., Johnson, C. J., Heard, D. C., Boyce, M. S., 2005. Can models of presence-absence be used to scale abundance? Two case studies considering extremes in life history. Ecography. 28, 1-12. Nielsen, S. E., McDermid, G., Stenhouse, G. B., Boyce, M. S., 2010a. Dynamic wildlife habitat models: Seasonal food and mortality risk predict occupancy-abundance and habitat selection in grizzly bears. Biological Conservation. 143, 1623-1634. Nielsen, S. E., Munro, R. H. M., Bainbridge, E. L., Stenhouse, G. B., Boyce, M. S., 2004c. Grizzly bears and forestry II. Distribution of grizzly bear foods in clearcuts of west-central Alberta, Canada. Forest Ecology and Management. 199, 67-82. Nielsen, S. E., Rovang, S., Penzes, M. 2010b. Using line transects for rapid assessments of grizzly bear foods, in: Foothills Research Institute Grizzly Bear Program 2010 Annual Report. Stenhouse, G., Graham, K. (Eds.). Hinton, Alberta, 68-78. Nilsen, P. A. 2002. Scandinavian brown bear (Ursus arctos L.) foraging on temporary and spatially variable berry resources in the boreal forest. M.Sc. Thesis, 1-66. Noyce, K. V., Coy, P. L., 1990. Abundance and productivity of bear food species in different forest types of northcentral Minnesota. Int. Conf. Bear Res. and Manage. 8, 169-181. Partridge, S. T., Nolte, D. L., Ziegltrum, G. J., Robbins, C. T., Impacts of supplemental feeding on the nutritional ecology of black bears. J. Wildlife Manage. 65, 191-199. Pelchat, B. O., Ruff, R. L., 1986. Habitat and spatial relationships of black bears in boreal mixedwood forest of Alberta. Int. Conf. Bear Res. and Manage. 6, 81-92. Pendleton, G. W., Titus, K., DeGayer, E., Flatten, C. J., Lowell, R. E., 1998. Compositional analysis and GIS for study of habitat selection by goshawks in southeast Alaska. J. Agri. Bio. & Environ. Stat. 3, 280-295. Pritchard, G. T., Robbins, C. T., 1990. Digestive and metabolic efficiencies of grizzly and black bears. Can. J. Zool. 68, 1645-1651.

151

Chapter 4-Bear Foods and Response to MPB management Raine, R. M., Kansas, J. L., 1990. Black bear seasonal food habits and distribution by elevation in Banff National Park, Alberta. Int. Bear. Res. Mng. 8, 297-304. Redding, T. E., Hope, G. D., Fortin, M. –J., Schmidt, M. G., Bailey, W. G., 2003. Spatial patterns of soil temperature and moisture across subalpine forest-clearcut edges in the southern interior of British Columbia. Canadian Journal of Soil Science. 83, 121-130. Reynolds-Hogland, M. J., Mitchell, S., Powell, R. A. 2006. Spatio-temporal availability of soft mast in clearcuts in the Southern Appalachians. Forest Ecology and Management 237, 103-114. Reynolds-Hogland, M. J., Pacifici, L. B., Mitchell, M. S., 2007. Linking resources with demography to understand resource limitation for bears. Journal of Applied Ecology. 44, 1166-1175. Robbins, C. T., Fortin, J. K., Rode, K. D., Farley, S. D., Shipley, L. A., Felicetti, L. A., 2007. Optimizing protein intake as a foraging strategy to maximize mass gain in an omnivore. Oikos. 116, 1675-1682. Robinson, B. G., Hebblewhite, M., Merrill, E. H., 2010. Are migrant and resident elk (Cervus elaphus) exposed to similar forage and predcation risk on their sympatric winter range? Oecologia. 164, 265-275. Rogers, L. 1976. Effects of mast and berry crop failures on survival, growth, and reproductive success of black bears. Transactions of the North American Wildlife and Natural Resource Conference. 41, 431-438. Rover, 2008. Grizzly bears and forestry I: Road vegetation and placement as an attractant to grizzly bears. Forest Ecology and Management. 256, 1253-1261. Safranyik, L., Carroll, A. L., Regniere, J., Langor, D. W., Riel, W. G., Shore, T. L., Peter, B., Cooke, B. J., Nealis, V. G., Taylor, S. W., 2010. Potential for range expansion of mountain pine bettle into the boreal forest of North America. Can. Entomol. 142, 415-442. Selas, V. 2000. Seed production of a masting dwarf shrub, Vaccinium myrtillus, in relation to previous reproduction and weather. Can. J. Bot. 78, 423-429. StataCorp. 2009. Stats Statistical Software: Release 11. College Station, TX: StataCorp LP. Strong, W. L. 2002. Lodgepole pine/Labrador tea type communities of western Canada. Can. J. Bot. 80, 151-165. Suring, L. H., Goldstein, M. I., Howell, S. M., Nations, C. S. 2009. Resonse of the cover of berryproducing species to ecological factors on the Kenai Peninsula, Alaska, USA. Can. J. For. Res. 38, 1244-1259. Thomas, D. L., Taylor, E. J., 1990. Study designs and tests for comparing resource use and availability. J. Wildlife Manage. 54, 322-330. Thomas, D. L., Taylor, E. J., 2006. Study designs and tests for comparing resource use and availability II. J. Wildlife Manage. 70, 324-336. Timoshok, E. E. 2000. The Ecology of Bilberry (Vaccinium myrtillus L.) and Cowberry (Vaccinium vitis-idaea L.) in Western Siberia. Russian Journal of Ecology. 31, 11-16. Trusler, S. T., Johnson, L. M. 2008. “Berry Patch” As a kind of place-the Ethnoecology of Black Hucklerberry in Nortwestern Canada. Hum Ecol. 36, 553-568.

152

Chapter 4-Bear Foods and Response to MPB management Visscher, D. R., Merrill, E. H. 2009. Temporal dynamics of forage sucession for elk at two scales: implications of forest management. Forest Ecology and Management. 257, 96-106. Vuong, Q. H., 1989. Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica. 57, 307-333. Wang, G. G., Huang, S., Monserud, R. A., Klos, R. J. 2004. The Forestry Chronicle. 80, 678-686. Welch, C. A., Keay, J., Kendall, K. C., Robbins, C. T., 1997. Constraints on Frugivory by Bears. Ecology. 78, 1105-1119. White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometric 48, 817-838. Whitney, G. 1982. The productivity and carbohydrate economy of a developing stand of Rubus idaeus. Can. J. Bot. 60, 2697-2703. Worton, B. J., 1983. Kernel methods for estimating the utilization distribution in home range studies. Ecology. 70, 164-168. Wulder, M. A., Ortlep, S. M., White, J. C., Nelson, T., Coops, N. C., 2010. A provincial and regional assessment of the mountain pine beetle epidemic in British Columbia: 1999-2008. Journal of Environmental Informatics. 15, 1-13.

Young, D. R., 1985. Microclimatic effects on water relations, leaf temperatures, and the distribution of Heracleum lanatum at high elevations. Am. J. Bot. 72, 357-364.

153

Chapter 5-Bear Foods and Climate Change

CHAPTER 5: POTENTIAL IMPACTS OF CLIMATE CHANGE ON CRITICAL GRIZZLY BEAR FOOD RESOURCES IN ALBERTA David Laskin and Greg McDermid Foothills Facility for Remote Sensing and GIScience, University of Calgary, Department of Geography, 2500 University Dr. NW, Calgary, AB. T2N 1N4

Project Description Grizzly bears (Ursus arctos) inhabiting seasonal latitudes, such as Alberta, possess habitatuse patterns that are clearly driven by the timing and availability of high-nutrition foodplants (Munro et al. 2006; Hamer and Herrero 1987). As the season progresses, bears will move throughout their range in search of quality forage. As a result, grizzly bear habitat is not contained within a static boundary, but rather, is very dynamic in response to marked pulses of nutrition across the landscape (Nielsen et al. 2010). The study of the annual timing of recurring biological events is termed phenology, and provides a dynamic, integrative approach to ecological research (Post and Inouye 2008). A major driver of plant development is temperature (Slafer and Savin 1991), as plants accumulate exposure to heat they will predictably move from one life-cycle event to the next â&#x20AC;&#x201C; these events are called phenophases. Global climate-change threatens to affect regional seasonal temperatures, with a predicted increase in mean annual temperature across Alberta and western Canada (Mbogga et al. 2009; IPCC 2001). These changes will directly impact the timing and distribution of bear food-plants, and potentially impede germination success, fruit ripening, or expose them to early season frost risk (Bennie et al. 2009; Inouye 2008; Myking 1997). Such climatic influences on vegetation will have cascading effects throughout the ecosystem (Parmesan and Yohe 2003; Parmesan 2006), ultimately affecting the abundance and quality of critical habitat for grizzly bears within the province. Often, habitat models rely too heavily on simplistic environmental surrogates that remain static through time (Schlossberg and King 2009), therefore grizzly bear recovery efforts in Alberta would be greatly enhanced with an understanding of the timing and location of high quality food resources and how these may be affected by climate variation and change. This research project aims to determine the potential impact of climate-change on the timing and distribution of critical bear food-plant species, and the amount of available nutrition throughout grizzly bear habitat in the province. This will be accomplished through direct field observations of plant phenology using a network of time lapse-cameras, meteorological equipment, and site visits by personnel throughout five elevational transects in the Rocky Mountain foothills of Alberta. These ground data will be linked to 154

Chapter 5-Bear Foods and Climate Change daily overhead spectral and temperature data collected by the Moderate Resolution Imaging Spectroradiometer (MODIS), mounted on the Terra and Aqua satellite platforms. Climate impacts on plant physiology will be analyzed at the Alberta Innovates Technology Futures Bioresource Center in Vegreville (AITF). Here, the distinct phenological progression of food-plants will be monitored, along with the corresponding amount of nutrition, within climate-controlled growth chambers. The focus species include: Buffaloberry (Shepherdia canadensis), a fruiting plant; cow parsnip (Heracleum lanatum), a herbaceous food; and alpine sweet vetch (Hedysarum alpinum), which has an edible root. The laboratory results of principal phenophase timing and developmental temperature thresholds will be applied to the MODIS imagery to develop theoretical models, expressing the impacts of climate change on food-plant phenophase timing and distribution. The final objective will be to develop spatially explicit empirical models that provide wall-to-wall dynamic estimates of phenophase timing and available nutrition throughout the season, and across grizzly bear habitat in Alberta.

Progress To Date Summer 2010 – 40 observation sites were established along the five longitudinal transects that encompass the study area (5526158N to 6040873N). During this period, seeds and edaphic mycorrhizae of focus species were collected for germination in the growth chambers at the AITF facilities in Vegreville. These plants grown from seed are the individuals used for observation to determine the developmental temperature thresholds (biofix), and phenological response rate to climate scenario simulations. Seeds were collected from low, mid, and high elevations to control for any phenotypic adaptations to long-term meteorological trends at these specific elevations. Fall 2010 – Seed germination was preceded by threshing, and scarification at the AITF laboratory to improve germination success. Germination for Hedysarum was very high (90%), while Shepherdia was deemed too long-lived to reach maturity from seed within the duration of the research project. Rather, mature, equal-aged individuals were selected from the AITF nursery, and potted for use in the growth chambers. Heracleum was very difficult to germinate, and was met with poor success – this may be due to the bi-annual nature of the species, fertilization issues, or the multi-season chilling requirement necessary to break dormancy. As a result, it was decided to omit Heracleum from the growth chamber observations. Field equipment for phenological observations was procured during this period. This includes: 85 6MP PlantCam time-lapse camera units with stands; 120 Thermochron iButtons for hourly temperature logging; 4 Hygrochron iButtons for humidity monitoring at select sites (Kinky Lakes and Ya Ha Tinda). Summer 2011 – This was the first official phenology monitoring season in the study area. Field personnel deployed cameras and temperature loggers across all sites (Figure 1) and 155

Chapter 5-Bear Foods and Climate Change visited plots three times per month (April-November) to record phenological development of the focus species, check cameras/temp loggers and perform stand-level phenological observations along random 250m transects. At the Hinton, and Ya Ha Tinda transects, Shepherdia berry sugar content (% brix) was recorded in association with phenological development, there is a clear temporal lag in sugar % as elevation increases. Overall, with over 50GB of imagery collected, the preliminary results show clear trends in focus species development across elevation, with lower sites progressing up to two weeks earlier than high elevation sites. Two additional transects were selected for monitoring: The Bow Valley from Exshaw to Lake Louise and HWY 40 south of Highwood junction. These auxiliary transects were sampled and surveyed in the autumn of 2011 for observations beginning the following season. Winter 2011-12 â&#x20AC;&#x201C; Successfully germinated Hedysarum were matured in greenhouses over summer and placed into growth chambers for day shortening and chilling to initiate dormancy. Along with mature Shepherdia, these plants were kept below 0Â°C for a minimum of 60 days over winter to meet chilling requirements, and will begin their growing season under full observation within the growth chambers beginning March 2012. Analysis has been completed on a manuscript investigating the impact of climate, topographic, and growing degree days (GDD) as drivers of grizzly bear food-plant phenological timing. Preliminary results show a delay in growth and development of various forb and shrub species across an elevational gradient near Hinton. This manuscript will be completed in the Summer of 2012, and lays the foundation for the subsequent phenological research papers, of which the initial field data and image analysis has been performed to validate camera observations for preparations of growing degree day (GDD) distribution maps for each focus species. Spring 2012 â&#x20AC;&#x201C; Beginning in March, the phenological camera network and temperature loggers will be deployed at all of the field sites. The AITF experimental phenology observations of the focus species will also begin at this time in the growth chambers.

156

Chapter 5-Bear Foods and Climate Change

Figure 1. Grizzly bear food-plant phenology observation network

157

Chapter 5-Bear Foods and Climate Change

Deliverables Update This project has a duration of two years and will be completed by March 2013. The first year of this project was focused on the collection of field data for climate map (GDD) development and validation and the germination/maturation of climate-controlled growth chamber experiments (2011-12). Phenology model development and analysis will be a major focus of the second year along with growth chamber manipulations to examine plant response to simulated climate change scenarios. A final report of this project will contain the following results, models, and GIS applications: 1) May 2012 – Deliverables to the FRI will include developmental threshold temperatures (biofix) of the focus species derived from in situ field observations. The developmental threshold temperatures will be refined in the climate controlled growth chambers. This specific deliverable will be presented as the baseline timing of principal life phenophases (life cycle events such as fruiting) of Shepherdia, Hedysarum, and Heracleum at different elevations throughout the study area. June 2012 – Submission date for manuscript describing the broad phenological timing of bear-food plants growing near Hinton: Predictions of significant phenological events modeled using topographic and climatic covariates, 2008-2010 (Tracy McKay, Karen Graham, and Scott Nielsen, co-authors). 2) Sept 2012 – Theoretical models (maps) of the phenological timing and distribution of critical grizzly bear food plants using field observations, growth chamber observations, and MODIS land surface temperature data. This deliverable includes phenology maps for Hedysarum and Shepherdia. Maps for Heracleum may be compromised by germination issues affecting observation success. 3) December 2012 – Theoretical models of climate change impacts on critical grizzly bear food-plant distribution. This deliverable will also provide estimates for both increasing and decreasing mean annual temperatures, as well as long term climate forecasts for the region (approx. 20-30 year from now, using IPCC model data). These models to be validated using plant response to experimentally derived simulations of climate change within the growth chambers. 4) January 2013 – Identification of the nutritional energy values of Shepherdia (possibly Hedysarum) during each phenophase in conjunction with experimental and forecasted climatic conditions. 5) March 2013 – Spatially explicit empirical models of the phenology and spatiotemporal distribution of critical grizzly bear food-plants in Alberta, including climate impact estimates for both increasing and decreasing mean annual temperatures, as well as long term climate forecasts for the region. 6) March 2013 – Prepare plant energy models that will link observed phenology (NDVI) to landscape energy supply and availability. These models will identify the spatial and temporal distribution of plant energy available for grizzly bears currently, as well as for future forest and climatic scenarios. GIS applications will be developed to allow end users to vary forest management options to determine grizzly bear energy over the landscape. 158

Chapter 5-Bear Foods and Climate Change

It is the intent of the research team that all work associated with this project will be published in peer reviewed scientific journals. References Bennie, J., Kubin, E., Wiltshire, A., Huntley, B., and Baxter, R. (2009) Predicting spatial and temporal patterns of bud-burst and spring frost risk in north-west Europe: the implications of local adaptation to climate. Global Change Biology, 16, 1503-1514. Hamer, D. and Herrero, S. (1987) Grizzly bear food and habitat in the front ranges of Banff National Park, Alberta. International Conference on Bear Research and Management, 7, 199-213. Inouye, D.W. (2008) Effects of climate change on phenology, frost damage, and floral abundance of montane wildflowers, Ecology 89(2):353-362. IPCC (2001) (ed.) Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the International Panel on Climate Change. Cambridge, UK. Cambridge University Press. Mbogga, M.S., Hamann, A., and Wang, T. (2009) Historical and projected climate data for natural resource management in western Canada. Agricultural and Forest Meteorology, 149(5):881-890. Munro, R.H.M., Nielsen, S.E., Price, M.H., Stenhouse, G.B., and Boyce, M.S. (2006) Seasonal and diel patterns of grizzly bear diet and activity in west-central Alberta. Journal of Mammalogy, 87(6):1112-1121. Myking, T. (1997) Effects of constant and fluctuating temperature on time to budburst in Betula pubescens and its relation to bud respiration. Trees-Structure and Function, 12(2):107-112. Parmesan, C. (2006) Ecological and evolutionary responses to recent climate change. Annual Review of Ecology Evolution and Systematics, 37:637-669. Parmesan, C., and Yohe, G. (2003) A globally coherent fingerprint of climate change impacts across natural systems. Nature, 421(6918):37-42. Post, E.S., and Inouye, D.W. (2008) Phenology: response, driver, and integrator. Ecology, 89(2):319-320. Schlossberg, S., and King, D.I. (2009) Modeling Animal Habitats Based on Cover Types: A Critical Review. Environmental Management, 43(4):609-618. Slafer, G.A., and Savin, R. (1991) Developmental base temperature in different phenological phases of wheat (triticum-aestivum). Journal of Experimental Botany, 42(241):10771082.

159

Chapter 6-GIS Forecasting Tools

CHAPTER 6: INTEGRATED SUITE OF GIS FORECASTING TOOLS FOR GRIZZLY BEAR HABITAT PLANNING Jerome Cranston Arctos Ecological Services

Summary This report describes how a software application developed by the FRIGBP, called GBtools, can be used to convert static spatial habitat models into dynamic forecasting tools by regenerating these models to incorporate user-supplied development scenarios. The value of a spatially explicit habitat model lies in its ability to predict the value of some habitat attribute in areas where empirical data is lacking. They are limited in that they are static, and quickly become obsolete in a rapidly-changing landscape. A means of updating the habitat models is required, not only to maintain their currency, but to forecast the effect of anticipated landscape changes, and to compare quantitatively the effect of multiple scenarios on habitat quality. The habitat models thus become not only spatially predictive, but temporally predictive as well. Just as a reliable forecast of tomorrow’s weather is far more valuable than a report on yesterday’s weather, the ability to predict changes in habitat quality is of great benefit to land and resource managers. GBtools is a set of geoprocessing scripts, written in the Python language and integrated with Environmental Systems Research Institute’s (ESRI) ArcGIS Desktop GIS software (9.2 or higher), that regenerate the grizzly bear Habitat States model (a combination of the RSF and mortality risk models) from base landscape layers. GBtools was designed to provide land and resource managers with the means to answer two questions: 1) what is the current state of grizzly bear habitat in my area of interest, and 2) how can the effect of different disturbance scenarios on habitat condition be compared? In its latest version, GBtools can incorporate changes resulting from four types of anthropogenic disturbance: logging, road-building, oil and gas development, and mining/quarrying, and can forecast conditions up to 30 years post-disturbance by incrementing canopy conditions in regenerating cutblocks. The basic process used by GBtools can be applied to other models and other species. A similar application was developed for the Alberta Conservation Association (ACA) in 2006 to regenerate RSF models for elk and wolves; another was developed for Habitat 160

Chapter 6-GIS Forecasting Tools Stewardship Canada in 2011 to regenerate grizzly bear food models (Nielsen, 2003); and an application is currently under development to regenerate multiscale RSF models for caribou (Hebblewhite et al, in press).

Habitat Models Grizzly bears rely on their habitat to provide three essentials for life: resources, such as food and water, security, to prevent excessive rates of mortality; and mobility, to allow dispersal into new territory and maintain genetic diversity. The FRIGBP has developed highresolution (30m pixel size), raster-based spatial models that predict all three habitat attributes as a function of landscape variables, covering core and secondary grizzly bear conservation areas, plus all parks and protected areas to the west and south (87,000 km2)(Figure 1).

Figure 1: Habitat model extent

Resource availability is represented by the Resource Selection Function (RSF) model (Nielsen, 2002), which predicts relative probability of bear occurrence, classified into 10 ordinal bins. Bear occurrence, or occupancy, is used in this model as a surrogate for the abundance and distribution of various habitat resources including food, water, denning sites, and thermal cover. Models have been developed for three seasons (spring, summer and fall), based on seasonal variation in diet (Munro, 2003); and for all seven population units (Livingstone, Castle, Clearwater, Yellowhead, Grande Cache, Swan Hills, and 161

Chapter 6-GIS Forecasting Tools Chinchaga), based on differences in resource use between genetically-distinct regional population groups (Proctor, 2010). They are derived from at least 2 years of grizzly bear location data collected by GPS radio collars, combined with six basic landscape variables: landcover (eight broad vegetation classes (McDermid, 2005), canopy cover, conifer/deciduous mix, streams, regenerating forest mask, and Compound Topographic Index (CTI), a measure of soil wetness. Habitat security is represented by the mortality risk model (Nielsen 2004), which predicts the relative probability of human-caused mortality as a function of landscape variables, including terrain, proximity to roads and trails, and regional land-use. The model is based on multivariate logistic regression analysis of 297 anthropogenic grizzly bear mortalities within the Central Rockies Ecosystem (CRE), dating from 1971 to 2002. The model is derived from six base inputs: landcover, terrain ruggedness (TRI), roads and trails, White Zone (agricultural land), distance to streams, and protected areas. Habitat connectivity has been modeled using Graph Theory (Schwab, 2003), however, this model is not updateable due to practical computational limits, and so cannot be used for forecasting and is not included in GBtools.

Figure 2: Habitat States matrix

The Habitat States model (Nielsen 2006), combines the RSF model (representing occupancy, or habitat quality) with the mortality risk model (representing habitat security) to distinguish areas of population sink (good-quality but high-risk habitat) from good162

Chapter 6-GIS Forecasting Tools quality, low-risk areas that serve as a population source (Figure 2). The relative proportion of population sources and sinks within a defined area can serve as a baseline measure of overall habitat quality.

Processing Steps The GUI (Graphical User Interface) for GBtools is accessible through ArcToolbox (Figure 3). The dialog box for GBtools accepts eight user-supplied inputs, three of which are required and five are optional. The basic processing steps are: 1. clip out updateable inputs by an Area of Interest (AOI) 2. update base layers with new anthropogenic features (optional) 3. regenerate 3 seasonal RSF models 4. calculate a seasonal maximum RSF and reclassify into good, intermediate, and noncritical habitat 5. regenerate the mortality risk model from the updated base layers and reclassify into secure and vulnerable habitat 6. Combine the reclassified RSF and risk layers into a Habitat States layer, and calculate mean value across the AOI

163

Chapter 6-GIS Forecasting Tools

Figure 3: ArcToolbox script and dialog for GBtools

User Input: Area Of Interest (required) The first input the user is required to provide is a shapefile or feature class representing the userâ&#x20AC;&#x2122;s Area Of Interest (AOI). This polygon layer is clipped by a shapefile of the model extents, buffered by 1 km (to account for edge effects) and converted to a bounding grid. This grid is multiplied by the landcover, regeneration mask, and crown closure grids using Spatial Analyst MapAlgebra to extract an updateable portion of these layers. The regeneration mask is used to extract crown closure in both matrix and harvested (regenerating) forests. Roads and trails are also clipped by the buffered AOI. User Input: Population unit (required) The user is required to specify the population unit from a drop-down list. The options are: Grande Cache, Swan Hills, Yellowhead, Clearwater, and Livingstone-Castle. This parameter will determine which set of coefficients will be used for the RSF model. User input: New cutblocks (optional) If a polygon shapefile or feature class representing new cutblocks is entered, it is used to update the grids for regeneration mask, crown closure in regenerating forests (new values are set to zero by default, assuming clearcut harvest), and landcover (set to barren, or nonvegetated, within new cutblocks).

164

Chapter 6-GIS Forecasting Tools User Input: New pipelines (optional) If a line shapefile or feature class representing new pipeline corridors is added, those features are appended to the trails base layer, and also buffered by 30m to ensure the footprint is represented in the landcover layer. The landcover within the corridor is updated to class 3 (upland herb), on the assumption that most pipeline clearings are seeded with herbaceous plants but tree and shrub regeneration is suppressed. Crown closure within the right-of-way is set to zero. Distance to trails is a predictor for mortality risk. A trail is any linear feature accessible by motorized vehicle (ATV), such as a pipeline, powerline, or seismic cutline; a road, by contrast, is a trail that is accessible by a truck. User Input: New roads (optional) If a line shapefile or feature class representing new roads is added, those features are appended to the roads base layer, and also buffered by 30m to ensure the footprint is represented in the landcover layer. The landcover within the corridor is updated to class 7 (barren), and crown closure within the right-of-way is set to zero. User input: Deletions (optional) Deletions are anthropogenic disturbances that effectively remove land from available habitat. These may be mines, quarries, oil and gas well pads, or urban development. If a polygon shapefile or feature class is entered here, RSF values are set to zero for these areas. User input: forecast age (optional) If an integer between 1 and 30 is entered as a forecast age, crown closure within regenerating forest is incremented as a function of regeneration age. The script assumes that attributes of the surrounding forest, such as landcover class, leading species, and crown closure, are constant over time. The premise is that changes to undisturbed, mature forests are relatively small compared to the changes that occur within regenerating forests; also, the intent of the script is to isolate changes in habitat that result from harvesting, as distinct from naturally-occurring successional changes in stand structure. If no parameter is entered, the crown closure increment is assumed to be zero. Otherwise, a forecast age (an integer between 1 and 30) can be entered and crown closure increment (CC) will be calculated as follows: CC = -3.0977 + 2.4784 * [age] â&#x20AC;&#x201C; 0.0218 * [age]2 This formula was derived using a weighted least squares regression analysis of crown closure values for 35,000 cutblocks harvested in grizzly bear range between 1955 and 2003 (Figure 4). Crown closure values for cutblocks were calculated using Spatial Analyst Zonal Statistics with pixel values derived from the Phase 7+ crown closure model. Figure 4 shows the relationship between mean crown closure values vs cutblock age (blue squares, 2003 reference year). The black line is the predicted value and the grey lines are confidence intervals on regression model predictions. 165

Chapter 6-GIS Forecasting Tools

Figure 4: Crown closure vs. cutblock age

Crown closure within regen (cc_rgn) is then calculated in MapAlgebra as existing crown closure plus the increment: cc_rgn = cc_rgn + (CC * ( MAX ( 0, ( 70 â&#x20AC;&#x201C; cc_rgn) / 70 )) If no age parameter is entered, CC = 0 and the increment is zero. Otherwise, CC is calculated as a function of forecast age, and crown closure within regen is incremented according to the existing pixel values, and the length of the forecast period. The curve in Figure 4 approaches 70% for older blocks; there may be higher crown closure values within older blocks (the curve above is based on a mean value among hundreds of blocks of the same age), but it is assumed that crown closure values higher than 70 will not be incremented over the forecast period. Pixel values < 70 will be incremented to varying degrees, ranging from 0 (for pixel values approaching 70) to CC (for pixel values approaching 0). The MAX function ensures that existing cc_rgn values over 70 do not cause the increment to become negative (ie, a decrease in cc values over time). Regenerating forests are also redefined according to forecast age. The grid representing age of regenerating forest is incremented by the forecast age using MapAlgebra, and the 166

Chapter 6-GIS Forecasting Tools Spatial Analyst Con (Conditional) tool is used to remove regen pixels with values greater than 50. Regen is considered to be harvested (or burned) forest less than 50 years old. Regenerating stands greater than 50 years old are considered to become part of the matrix forest (upland tree class). User input: Output filename (required) A default filename is provided, or the user may choose a different filename for different scenarios, to prevent the outputs from being overwritten. Secondary inputs Once the base layers have been updated with the chosen scenarios, the models are regenerated. From the basic inputs, a number of secondary variables are derived. Cost Distance to Roads: Terrain Ruggedness is used as a cost surface to derive the negative exponential decay cost distance to roads, a predictor for mortality risk. Landcover is reclassified to two types, forest and non-forest, and a contour is generated from this surface to represent forest edges, from which a Euclidean distance to forest edge is derived for each landcover class. The landcover is also reclassified to extract a separate surface for each class, as each landcover type has its own coefficient in the RSF model. Focal Statistics was used to derive the mean proportion of upland tree class within a 17-km radius as a regional-scale predictor to the mortality risk model. Because generating this variable is computationally expensive, it is considered a static layer and does not change significantly with the addition of a few new openings.

Regeneration of RSF models These updated input layers, as well as the static layers such as decay distance to stream and CTI, are combined with the coefficients specific to the selected population unit in a MapAlgebra expression to create the RSF models for each season. A scalar value is applied to each raw RSF to account for regional occupancy (range contraction). The raw RSF scores are rescaled to between 0 and 1 using the expression exp(rsf_raw)/(1 â&#x20AC;&#x201C; exp (rsf_raw)) and classified into 10 ordinal bins using class breakpoints specific to the population unit. Once all three seasonal models have been generated, the MapAlgebra MAX function is used to carry forward the maximum pixel values for each seasonal model into a single RSF model; this ensures that areas of seasonally critical habitat are retained.

167

Chapter 6-GIS Forecasting Tools

Regeneration of Risk model The mortality risk model is regenerated by combining the updated inputs (landcover, decay distance to roads and trails, distance to forest edge) with the static layers (proportion of White Zone and proportion of protected area within 10km radius; decay distance to stream; terrain ruggedness) in a MapAlgebra expression. The raw risk scores are rescaled to between 0 and 1 as with the RSF, and classified into 10 ordinal bins using a predefined set of breakpoints.

Regeneration of Habitat States model The RSF model and the mortality risk model each describe different aspects of grizzly bear habitat, but it is the spatial interaction of the two models that determines overall demographic “fitness.” To regenerate the Habitat States model, the RSF model (seasonal maximum) is reclassified into three types:  Non-critical habitat (RSF <= 4) value = 0  Secondary habitat (RSF >=5 and RSF <= 7) value = 1  Primary habitat (RSF >=8) value = 2 Similarly, the risk model is reclassified into two types:  Secure habitat (risk <=5) value = 1  Vulnerable habitat (risk >=6) value = -1 The two reclassified surfaces are multiplied together in MapAlgebra to create a final Habitat States model for the AOI (Figure 5), whose name is specified as the final parameter in the dialog. Statistics for the AOI, including the mean, are calculated using Zonal Statistics and written to a dbase (*.dbf) file with the same name as the output model.

Figure 5: Habitat States model

168

Chapter 6-GIS Forecasting Tools The mean value of the Habitat States model across the AOI can be used as a benchmark against which to measure the effect of various development scenarios. The conversion of population source to population sink could be offset by mitigation in other areas to ensure a zero net loss of critical habitat. Note that this statistic cannot yet be calibrated against actual population trend. It cannot be said, for example, that if the mean Habitat State value for a given populationâ&#x20AC;&#x2122;s range is negative, N will decline, or that if the mean Habitat State value is positive, the population will increase. This will require long-term monitoring of population trend and the concurrent tracking of habitat conditions. References McDermid, G.J. 2005. Remote Sensing for Large-Area, Multi-Jurisdictional Habitat Mapping. PhD Thesis. Department of Geography, University of Waterloo, Waterloo, Ontario, Canada. Munro, R.H.M., Nielsen, S.E., Price, M.H., Stenhouse, G.B., and Boyce, M.S. 2006. Seasonal and diet patterns of grizzly bear diet and activity in west-central AlbertA. Journal of Mammalogy 87: 1112-1121. Nielsen, S.E., Boyce, M.S., Stenhouse, G.B., and Munro, R.H.M. 2002. Modeling grizzly bear habitats in the Yellowhead ecosystem of Alberta: Taking autocorrelation seriously. Ursus 13: 45-56. Nielsen, S.E., Boyce, M.S., Stenhouse, G.B., and Munro, R.H.M. 2003. Development and testing of phenologically driven grizzly bear habitat models. Ecoscience 10: 1-10. Nielsen, S.E., Herrero, S., Boyce, M.S., Mace, R.D., Benn, B.,Gibeau, M.L., Jevons, S., 2004. Modelling the spatial distribution of human-caused grizzly bear mortalities in the Central Rockies Ecosystem of Canada. Biological Conservation 120, 101â&#x20AC;&#x201C;113. Nielsen, S.E., Stenhouse, G.B., and Boyce, M.S. 2006. A habitat-based framework for grizzly bear conservation in Alberta. Biological Conservation 130: 217-229. Proctor, M., McLellan, B., Boulanger, J., Apps, C., Stenhouse, G., Paetkau, D., and Mowat, G. 2010. Ecological investigations of grizzly bears in Canada using DNA from hair, 19952005: a review of methods and progress. Ursus 21: 169-188. Schwab, B.L. 2003. Graph Theoretic Methods for Examining Landscape Connectivity and Spatial Movement Patterns: Application to the FMF Grizzly Bear Research. MSc Thesis. Department of Geography, University of Calgary, Calgary AB.

169

Chapter 7-Density and Population Trend

CHAPTER 7: ESTIMATION OF DENSITY AND POPULATION TREND FOR THE YELLOWHEAD 2011 DNA GRID Sarah Rovang1, John Boulanger2, Gordon Stenhouse3, and Scott Nielsen4 1

Department of Renewable Resources, University of Alberta, 751 General Services Building, Edmonton, AB T6G 2H1, rovang@ualberta.ca 2 Integrated Ecological Research, 924 Innes, Nelson, BC, V1L 5T2, 250-352-2605, boulange@ecological.bc.ca, www.ecological.bc.ca 3 Foothills Research Institute, Box 6330, Hinton, AB, gstenhouse@foothillsri.ca 4 Department of Renewable Resources, University of Alberta, 751 General Services Building, Edmonton, AB T6G 2H1, scottn@ualberta.ca

Introduction In 2010, Alberta listed grizzly bears (Ursus arctos) as a threatened species following six years of DNA-based mark-recapture population inventories that occurred across the province from 2004 to 2008 (ASRD & ACA 2010). While single point population estimates are known (ASRD & ACA 2010), no monitoring program is in place to: (a) identify trends, (b) determine whether or when populations reach established recovery targets, (c) evaluate the effectiveness of management actions, such as limitations on road densities (i.e., core and secondary habitats), and (d) determine population estimates necessary for setting harvest quotas. Such information, however, will be crucial to directing recovery actions and evaluating the success of those efforts. Although the hairsnag DNA-based mark-recapture approach is well-accepted and quite precise relative to other approaches for this species, such methods are unsuitable for their long-term use in population monitoring simply due to their high cost (i.e., $2.4M for the first population estimate). Standardized tools, techniques and sampling protocols that are cost effective, yet statistically robust for monitoring population size and trends are therefore needed. Two alternative approaches that may lower costs and be better suited for long-term population monitoring programs are to use a network of permanent sample plots in DNA-based markrecapture hair sampling designs, or to use multiple sources of DNA. This research evaluates tradeoffs between monitoring costs and detectability of grizzly bears using permanent DNA-based mark-recapture methods and investigates the efficacy of using multiple data sources, including rub trees and fecal DNA collections. This report details the estimation of density for the 2011 DNA grid within the Yellowhead population unit using DNA from hair collected using hair-snag and rub tree methods.

170

Chapter 7-Density and Population Trend

Methods Study Area A 1,500 km2 study area south of Hinton was delineated using 30 separate hexagon-shaped sampling cells each 50-km2 in size (Figure 1) with the purpose of spreading sampling effort across a gradient from upper to lower Foothills. Grizzly bear densities within the larger population unit were estimated at 4.75 bears per 1000km2 in 2004 (Boulanger et al. 2005b). The majority of the study area occurs in core grizzly bear conservation areas where management objectives are to keep road densities below 0.6-km/km2. The remainder of the study area occurs in secondary conservation areas/watersheds where the management objective is to maintain road density below 1.2 km/km2. The majority of the 2011 study area overlaps the 2004 DNA sampling grid (Figure 1).

2011 Survey Efforts Six hair-snag survey sessions for 60 permanent sample sites (360 visits) were monitored during the summer of 2011 (Figure 2). Following methods outlined by Woods et al. (1999) and Mowat & Strobeck (2000), hair traps consisted of approximately 30-m length of 4-prong barbed wire encircling 4â&#x20AC;&#x201C;6 trees at a height of 50 cm. A total of 2.5 L of scent lure, a 2:0.5 mix of aged cattle blood and canola oil (no physical reward), was poured on forest debris piled in the center of the corral at least 2m from the wire. Sites were visited on a two-week sample rotation (5 work days per week) from May 18 to August 25 for a total of six sampling sessions. At each visit, bait was refreshed and all hair was collected from each barb, dried at room temperature and stored in paper envelopes. Also during this time, a segment of GPS collared bears were monitored on the grid by the Foothills Research Institute. At each site a crew of two individuals searched a minimum 25 meter radius to identify the â&#x20AC;&#x153;bestâ&#x20AC;? or most used animal trail. Once found, this trail was searched for 250 meters in each direction (for a total of 500 meters) for scat and rub trees. Due to the unlikelihood of encountering scat or rub trees within 500 meters of a hair-snag site, at least 1 kilometer of additional trail independent of hair snag sites was searched in each hexagon. Seismic lines, pipelines, and quad trails were the preferred trail type due to the increased likelihood of animal movement on such trails. In the event where no natural rub tree was found, an artificial rub tree was created by stapling a small section of barbwire to the tree in a zigzag fashion and then spraying that barbwire with WD-40 (a fish oil), which served as a mild attractant. Because intersections of two game trails receive more traffic, artificial rub trees were often created at these locations. In total, four natural rub trees and twentyeight artificial rub trees were used in this study.

Multiple Data Source Analysis A multiple data source mark-recapture analysis approach (Boulanger et al. 2008) was used to estimate the average number and density of grizzly bears on the sampling grid. For this approach, the first session was represented by the list of radio collared bears known to be on the sampling grid, the second session was the list of detection of bears using rub trees throughout the entire time of sampling, and the third through eighth session was the x-matrix of bears detected in each of the hair snag sessions. The multiple data source approach has been shown to be more robust to heterogeneity variation in detection rates and display enhanced precision compared to single data source approaches (Boulanger et al. 2008, Kendall et al. 2008, Kendall et al. 2009, Gervasi et al. 2010). Preliminary analysis on the hair snag only dataset were performed to test whether a behavioural response may have been created by using only fixed sites for hair snag sampling. For

171

Chapter 7-Density and Population Trend this exercise, we contrasted the relative support of models that assumed that initial detection (p) and redetection probabilities (c) were different (suggesting a behavioral response was occurring) with models that assumed initial and redetection probabilities were equal (suggesting no behavioral response was occurring) for sex-specific and pooled sex detection and redetection rates. The multi-data source approach was also used to estimate density from the portion of the 2004 Yellowhead grizzly bear DNA inventory (Boulanger et al. 2005b) grid that overlapped the 2011 grid. The first session included collared bears known to be on the grid area and the second through fifth sessions were detections at hair snag sites, which included both fixed and moved sites as part of an experiment on optimal sampling strategies (Boulanger et al. 2006). We considered both fixed and moved site data for the analysis but also conducted a sensitivity analysis of the effect of inclusion of fixed and moved sites by comparing estimates to a moved-site only data set.

Estimation of Density Population density was estimated using the Huggins density estimation module in program MARK and using program DENSITY. The density estimation module in program MARK (White and Burnham 1999) uses information from DNA bears and bears that were collared to estimate density. The general estimator of population size using mark-recapture methods is the count of individuals detected (Mt+1) divided by the detection probability of individuals across all sampling occasions (p*) (eqn. 1) (Huggins 1991). Another way this can be expressed is the summation of individuals each divided by p* (eqn. 2). (̂ (eqn. 1) (̂

⁄ )

∑

)

(eqn. 2) This method assumes a closed population and therefore can lead to a positive bias in estimates if this assumption is violated. The MARK module estimates residency using logistic regression with the response variable being the ratio of locations that radioed bears were on the sampling grid compared to the total number of locations obtained for the duration of sampling efforts. Estimates of the average number of bears on the grid area at one time are estimated by substituting the 1 in the above equation with an estimate of residency (symbolized as ̃) (eqn. 3). Estimates of density were then obtained by dividing Nave by the area of the sampling grid (1,500 km2). (Nave) ( ̂

∑

)

(eqn. 3) Residency of radioed or DNA bears often is a function of the distance of the bears from the sampling grid edge (Boulanger and McLellan 2001) where bears that occur near the edge are more likely to be off the grid than bears in the middle of the study area. Therefore the distance from edge of the mean location of a radioed bears was used as a covariate in the logistic analysis to predict bear residency. An assumption of this method is that the radio collared bear residency is similar and is a good sample of the DNA bear residency. By using distance from edge as a covariate, the assumption of similar distributions of DNA and radio collared bears was relaxed, making this

172

Chapter 7-Density and Population Trend assumption more reasonable. Support for different models was evaluated using information theoretic model selection methods (Burnham and Anderson 1998).

Estimation of trend (2004-2011) Superpopulation size and trend from 2004 to 2011 were estimated using the Robust design (Pollock and Otto 1983) Pradel model (Pradel 1996) in program MARK (White and Burnham 1999). Because sampling was conducted in June and July in 2004 and from June to August in 2011, trends between the two years may be potentially influenced by the longer sampling duration in 2011. As a result, only the data from the first four hair-snag and rub tree sessions for 2011 were used. Population size and detection probability (p*) were estimated for each year using the Huggins closed population size model (Huggins 1991). Change in population size (), apparent survival () and rates of additions between years (f) were estimated using the Pradel model. Population rate of change is equivalent to the population size for a given sampling period divided by the population size in the previous sampling period (=Nt+1/Nt). Given this, estimates of  will be 1 with a stable population, less than 1 if the population is declining and greater than 1 if the population is increasing. The same general models and multiple data source approach used for the density analysis were also used to model variation in detection rates for the robust design analysis. Because sampling was conducted in 2004 and 2011, the sampling interval was not yearly but instead spanned seven yearly sampling intervals. To allow a standardized yearly metric for demography we scaled the Pradel model estimates to a yearly interval but also considered trend over the seven year interval between sampling events.

Results In total, we collected 664 hair samples at the 60 hair-snag sites, 29 rub tree hair samples at 32 sites, and 107 scat samples. Genotyping of hair samples was completed for 342 hair-snag samples and all 29 rub tree hair samples; however, poor genotyping success rates at Wildlife Genetics International (WGI)(Nelson, BC) prevented the analysis of scat samples. From the genotyped hair samples, we identified 21 unique grizzly bears (12 females and 9 males) within a 1,500-km2 area south of Hinton, Alberta (Figure 3). Of these 21 individuals, three were recaptures from the 2004 DNA survey, six were recaptures from other studies since 2004, and 12 were new individuals. Two bears (both male) were detected using rub trees, one of which was detected only by rub trees whereas the other was also detected using hair-snags and rub trees. Density for the 2011 grid was estimated at 11.7 bears ± 2.8 per 1000-km2 as compared to a 2004 estimate of 7.0 bears ± 1.3 per 1000-km2 for the same overlapping area (Boulanger et al. 2012) (Table 1). Based on these estimates, density increased in 2011 by a factor of 1.67 compared to 2004. However, because the extended sampling period in 2011 may influence the trend between the two years, a trend analysis was conducted that only used data collected in June and July. For this multiple data source analysis (Boulanger et al. 2012), the most supported model assumed sex-specific hair-snag detection rates and residency as well as sex-specific curves for residency and detection rates as a function of distance from the grid edge (Table 2, model 1). Models that assumed session-specific hair snag detection rates (models 6 and 7), different detection rates for bears that were previously collared (model 4), and no relationship between distance from edge and detection rate (model 2) were less supported. As a result, a negative behavioral response in re-

173

Chapter 7-Density and Population Trend capture rates is not evident. Model averaged estimate of session-specific detection rates was 0.15 for hair snags (Boulanger et al. 2012) (Table 3), therefore seasonal variations in capture probabilities were also unobserved. This suggests that hair sampling at a single site can be done during any part of the summer and without major concerns on recapture bias. Using the multiple data source mark-recapture analysis (Boulanger et al. 2012), superpopulation size for our study area was estimated at 25.1 bears (SE = 5.0) (11.7 ± 2.8 per 1000-km2) (Table 4). This compares with a superpopulation estimate of 16.1 bears (SE = 4.1) (7.0 ± 1.3 per 1000-km2) for the overlapping portion of the 2004 grid, which suggests an annual population growth rate of 5.8% (λyearly = 1.058, SE = 0.065); however, it should be noted that confidence intervals of λ overlap 1.0 (0.930 – 1.185). Furthermore, this study represents only one small area of the Yellowhead population (Figure 1).

Discussion Due to limitations in either the field or the laboratory, we currently do not suggest using rub tree hair sampling or fecal DNA for monitoring bear populations in the Hinton area. Because this area is largely characterized by forested habitat and low densities of trails, locating rub trees and scat samples is difficult and inefficient. Further work is needed to map locations of rub trees before further considering this technique in this area. Enhancements in laboratory techniques for genotyping fecal samples are needed before fecal DNA sampling can be used for population monitoring. Fecal DNA sampling may provide a useful tool, however for monitoring range expansion and contraction since it is successful in genotyping species (black versus grizzly bear) and may also be accurate for determining gender. In contrast, the results of this study suggest that a DNA-based hair sampling method based on a network of permanent sample plots is a reliable sampling strategy for population trend studies for bears. To be successful, a mark-recapture survey should strive to maximize capture probability (i.e. detection) while minimizing capture heterogeneity, which generally requires high capture probabilities (Lukacs and Burnham 2005b). Maximizing capture probability is especially crucial when sampling a low density population and even more so when using permanent DNA hair-snag methods where sites are not moved. As a result, the utility of this method is highly dependent on survey design. Remaining analyses for this project are therefore to investigate optimal sample design and environmental influences on detectability and capture probability.

Literature Cited Alberta Sustainable Resource Development (ASRD) and Alberta Conservation Association (ACA). 2010. Status of the grizzly bear (Ursus arctos) in Alberta: Update 2010. Alberta Sustainable Resource Development. Wildlife Status Report No. 37 (Update 2010). Edmonton, AB. 44 pp. Boulanger, J., G. Stenhouse, M. Proctor, S. Himmer, D. Paetkau, and J. Cranston. 2005b. Population inventory and density estimates for the Alberta 3B and 4B Grizzly Bear Management Area. Alberta Sustainable Resource Development, Hinton, Alberta Boulanger, J., M. Proctor, S. Himmer, G. Stenhouse, D. Paetkau, and J. Cranston. 2006. An empirical test of DNA mark-recapture sampling strategies for grizzly bears. Ursus 17:149-158.

174

Chapter 7-Density and Population Trend Boulanger, J., K. C. Kendall, J. B. Stetz, D. A. Roon, L. P. Waits, and D. Paetkau. 2008. Multiple data sources improve DNA-based mark-recapture population estimates of grizzly bears. Ecological Applications 18:577-589. Boulanger, J., S. Rovang, G.B. Stenhouse, and S.E. Nielsen. 2012. Estimation of density and population trend for the Yellowhead 2011 DNA grid. Report for ACA project. February 15, 2012. 14 pg. Burnham, K. P., and D. R. Anderson. 1998. Model selection and inference: A practical information theoretic approach. Springer, New York, New York, USA. Gervasi, V., P. Ciucci, F. Davoli, J. Boulanger, L. Boitani, and E. Randi. 2010. Addressing challenges in non invasive capture recapture based estimates of small populations: a pilot study on the Apennine brown bear. Conservation Genetics Online. Huggins, R. M. 1991. Some practical aspects of a conditional likelihood approach to capture experiments. Biometrics 47:725-732. Kendall, K. C., J. B. Stetz, J. Boulanger, A. C. Macleoud, D. Paetkau, and G. C. White. 2009. Demography and genetic structure of a recovering grizzly bear population. Journal of Wildlife Management 73:3-17. Kendall, K. C., J. B. Stetz, D. A. Roon, L. P. Waits, J. Boulanger, and D. Paetkau. 2008. Grizzly Bear Density in Glacier National Park, Montana. Journal of Wildlife Management 72:1693-1705 Lukacs, P.M., and K.P. Burnham. 2005b. Review of capture-recapture methods applicable to noninvasive genetic sampling. Molecular Ecology 14:3909-3919. Mowat, G., and C. Strobeck. 2000. Estimating population size of grizzly bear using hair capture, DNA profiling, and mark-recapture analysis. The Journal of Wildlife Management 64:183-193. Nielsen, S.E., J. Cranston, and G.B. Stenhouse. 2009. Identification of priority areas for grizzly bear conservation and recovery in Alberta, Canada. Journal of Conservation Planning 5: 38-60. Pollock, K. H., and M. H. Otto. 1983. Robust estimation of population size in closed animal populations from capture-recapture experiments. Biometrics 39:1035-1049. Pradel, R. 1996. Utilization of mark-recapture for the study of recruitment and population growth rate. Biometrics 52:703-709. White, G. C., and K. P. Burnham. 1999. Program MARK: Survival estimation from populations of marked animals. Bird Study Supplement 46:120-138. Woods, J.G., D. Paetkau, D. Lewis, B.N. McLellan, M. Proctor, and C. Strobeck. 1999. Genetic tagging of free-ranging black and brown bears. Wildlife Society Bulletin 27:616-627.

175

Chapter 7-Density and Population Trend

Table 1. Estimates of average population size and density for the 2004 and 2011 DNA surveys Statistic 2011 Average N Density 2004 Average N Density

Estimate

17.51 11.68

4.18 2.79

10.96 7.31

27.92 18.61

23.9%

10.49 7.00

2.00 1.33

6.57 4.38

14.42 9.61

19.1%

Table 2. Model selection of the Huggins density model for the 2011 grid No

HS detection

Residency

AICc

Î&#x201D;AICC

Deviance

sex+d+sex*d sex sex+d sex+collar constant sex+t7+t8 sex+session sex+d+sex*d sex constant constant

sex+d+sex*d sex+d+sex*d sex+d+sex*d sex+d+sex*d sex+d+sex*d sex+d+sex*d sex+d+sex*d sex+dcf+sex*dcf d+sex d+sex constant

5073.3 5075.0 5077.0 5077.1 5077.4 5078.1 5084.1 5128.5 5186.1 5188.5 6534.7

0.00 1.72 3.70 3.76 4.08 4.78 10.82 55.13 112.81 115.18 1461.35

0.51 0.22 0.08 0.08 0.07 0.05 0.00 0.00 0.00 0.00 0.00

10 8 9 9 7 10 13 9 7 6 5

5053.0 5058.8 5058.8 5058.8 5063.2 5057.8 5057.6 5110.2 5172.0 5176.4 6524.6

2 3 4 5 6 7 8 9 10 11

Table 3. Model averaged estimates of session-specific detection rates from the 2011 grid Statistic Session Estimate SE LCI UCI Radio p* 0.171 0.073 0.070 0.361 RT p* (all sessions) 0.057 0.041 0.013 0.212 RT p (session) 0.015 0.943 0.014 1.000 HS 1 0.150 0.051 0.075 0.277 2 0.150 0.051 0.075 0.277 3 0.150 0.051 0.075 0.277 4 0.150 0.050 0.075 0.277 5 0.152 0.054 0.073 0.290 6 0.147 0.051 0.072 0.278 Average residency 0.349 0.008 0.333 0.366

176

Chapter 7-Density and Population Trend Table 4. Model-averaged estimates of demography and superpopulation size from the Pradel model analysis . Statistics Estimate SE CI CV Demographics Apparent survival Additions (f) Trend 位 (yearly) Overall change 2004 to 11

0.731 0.327 1.058 1.479

0.101 0.110 0.065

0.499 0.154 0.930 0.604

0.881 0.565 1.185 3.275

Superpopulation 2004 2011

16.10 25.09

4.07 5.01

8.12 15.27

24.08 34.90

25.3% 20.0%

Figure 1. Yellowhead grizzly bear population unit illustrating terrain, protected areas (Jasper National Park has dotted pattern, while provincial parks are stippled), grizzly bear conservation areas (core or secondary) and the location of 2004 and 2011 grids and bear 177 detections.

Chapter 7-Density and Population Trend

Figure 2. DNA hair-snag sampling (monitoring) locations established during 2011 in the Hinton, Cadomin, and Robb areas. Spatial sampling intensity was two sites per 50-km2 hexagon. Background map represents the multi-seasonal habitat selection model (Nielsen et al. 2009).

178

Chapter 7-Density and Population Trend

Figure 3. Grizzly bear monitoring sites and detections of grizzly bears during the 2011 hair-snag DNA survey. Background map represents the multi-seasonal habitat selection model (Nielsen et al. 2009).

179

Chapter 8-Future of Alberta’s Forests

CHAPTER 8: FUTURE OF ALBERTA’S FORESTS FOR GRIZZLY BEARS: CLIMATE AND LANDSCAPE SCENARIOS AND SIMULATIONS PI: Nicholas Coops (UBC), Team: Mike Wulder , Wiebe Nijland2, Thomas Hilker2,3, Chris Bater2,4, Trisalyn Nelson5 1

Canadian Forest Service, 2University of British Columbia, 3NASA, 4Alberta Sustainable Resource Development, 5University of Victoria

EXECUTIVE SUMMARY The accurate and timely mapping of anthropogenic and natural disturbance patterns is critical to improve our understanding of habitat and bear movements. In addition an understanding of the climate of the region, now and into the future, as well as the underlying landscape cover and pattern, is critical to provide an overall management strategy for keystone species. In this annual report we continue our derivation of high-spatial (30 m) and -temporal (weekly or bi-weekly) resolution geo-spatial predictions of disturbance generated by the Spatial Temporal Adaptive Algorithm for mapping Reflectance Change (STAARCH) by extending our predictions from 2008 – 2011 providing over a decade of forest disturbance information. Research also continued into the monitoring of habitat at fine spatial scales. We again deployed a network of cameras over the 2010 and 2011 growing seasons across a range of sites in western Alberta, Canada, with the purpose of linking fine and broader scale images together, as well as combining plant phenology to nutritional data collected by our research partners. Lastly we began the development of a set of landscape simulation scenario tools developed within the landscape forest modeling framework LANDIS-II. To do so we have produced a number of base layers, and undertaken a suite of fire landscape simulations based on current landscape land cover and decadal fire hotspot data collected by satellite and aerial photography.

WORK PACKAGE 1: UPDATE LANDSCAPE CHANGE AND CONDITION PRODUCTS. 2008 TO 2010 Habitat structure and the distribution of wildlife species, such as grizzly bears, can be strongly influenced by the spatial and temporal distribution of vegetation structure, 180

Chapter 8-Future of Alberta’s Forests disturbance events and by vegetation phenology. Landsat imagery, with a spatial resolution of 30m, has been shown to allow the monitoring of land cover and ecosystem disturbance over large areas, however the minimum 16-day revisit cycle of the platform limits the ability to accurately determine dates of disturbance events, especially in regions where cloud contamination is a frequent occurrence. Fusion of high spatial resolution Landsat imagery with high temporal resolution imagery such as Moderate Resolution Imaging Spectroradiometer (MODIS) data can allow the generation of synthetic, Landsat-like, observations with high spatial (30m) and temporal (potentially daily) resolution. The Spatial Temporal Adaptive Algorithm for mapping Reflectance Change (STAARCH) (Hilker et al. 2009) was developed as an extended version of STARFM to allow the detection of disturbance events at spatial scales smaller than that of a MODIS pixel, through the generation of a spatial change mask derived from Landsat and an image sequence recording the temporal evolution of disturbance events (based on MODIS). STAARCH has been successfully applied for the core grizzly bear study area for the period of 2001 to 2008. The results of the STAARCH algorithm provided information on large scale spatial and temporal patterns in disturbance regimes and changes in vegetation structure, with implications for grizzly bear habitat modeling and management. In this annual report the STAARCH algorithm was run to extend the temporal coverage of the disturbance information to include the period of 2008 to 2011. As a result, the complete STAARCH dataset now spans a 10 year period from 2001 to 2011.

2008-2011 STAARCH data and processing The STAARCH runs undertaken from 2008 – 2011 were based on summer Landsat-5 scenes of 2008 and 2011 and 8-day MODIS composites (MODIS product MOD09-A1) excluding winter (November – February) MODIS scenes resulting in a total of 92 composites. The grizzly bear study area is covered by 16 Landsat scenes however as a number of scenes had extensive cloud cover in either 2008 or 2011 a number of scenes were excluded from the final analysis. The considerable overlap between the Landsat scenes in the area ensured almost 100% coverages with the 9 remaining scenes. Acquisition dates of the 2008 and 2011 Landsat-5 scenes were all between 25 July and 29 September (Table 1) . Table 1: Acquisition dates of the 9 Landsat-5 scenes used for the STAARCH disturbance detection

Path 41 42 42 43 43 43 45 45 47

Row 26 25 26 22 23 24 22 23 22

Landsat-5 2011 29 Sep 04 Sep 04 Sep 27 Sep 26 Sep 26 Aug 09 Sep 09 Sep 07 Sep 181

Landsat-5 2008 20 Sep 25 Jul 25 Jul 17 Aug 17 Aug 17 Aug 16 Sep 16 Sep 14 Sep

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests

The results of the STAARCH algorithm were converted to vector format with polygons attributed with the date of disturbance. As described in Gaulton et al. (2011) areas of disturbance smaller than 1 ha were removed from the results due to unreliable change detection often caused by slight differences in spatial registration of the two landsat scenes. In areas where information was available from multiple scenes, the best quality scene was chosen based on the lowest cloud contaminations. Furthermore the area of detection was limited to the grizzly bear research area excluding agricultural areas (based on the agriculture land cover layer).

Results The total area of disturbances detected between 2008 and 2011 was 199,527 ha, which is 1.52% of the total grizzly bear research area. The overall rate of disturbance was 0.51% of the total area per year compared to 0.4% in the period 2001-2008 (Gaulton et al, 2011).

Figure 1. Temporal distribution of disturbance events.

Figure 1 shows the temporal distribution of disturbance, as predicted by STAARCH, for the entire study area. Disturbances are most frequently predicted in the first few dates of the growing season (representing both disturbances during those periods as well as those occurring outside of the sequence, such as winter harvesting) as well as late in the season, during late August and September. A quick look of the STAARCH predicted layer is shown in Figure 2. This Figure shows the coverage of the Landsat scenes as well as the spatial distribution of the changed areas. Digital data layers of the STAARCH outputs were provided to FRI in January 2012.

182

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests

Figure 2. Overview of the STAARCH disturbance events for 2008-2011.

183

Chapter 8-Future of Alberta’s Forests

WORK PACKAGE 2: CLIMATE AND DEVELOPMENT OF LANDSCAPE LAYERS Deviation of Data: Mean monthly climate spatial surfaces are generated using ClimateWNA, which downscales precipitation and temperature data generated at 2 - 4 km by PRISM (Wang et al. 2006) Parameter-elevation Regressions on Independent Slopes Model (Daly et al. 2002) to 1 km. A 90m-Digital Elevation Model (DEM), obtained from the Shuttle Radar Topography Mission (SRTM), was resampled to 1 km to provide the required elevation data at the same resolution as the climatic data. The downscaling is achieved through a combination of bi-linear interpolation and elevation adjustment. We will compute monthly coverages of precipitation and minimum and maximum temperature from 1950 to 2010 (Figure 3). To simulate conditions under a future projected climate, we will utilize the Special Report on Emission Scenarios (SRES) climate scenarios developed by the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report, AR4 (Nakicenovik and Swart 2000; IPCC 2007). Three climate scenarios were produced: “a business as usual” scenario (A2), a scenario (B1) that assumes current emissions rates will remain steady until around 2040, and then slowly drop to about half of the current rate by the end of the century and a third in between scenario (A1B). Monthly climate layers were produced for three standard 30year periods, the 2020`s (2011-2040), 2050`s (2041 – 2070) and the 2080`s (2071 – 2100) (Figure 3). The provided datasets also include derivation of extended climate layers including the number of Frost Days, Vapour Pressure Deficit (VPD) and short wave incoming radiation (rad). To generate these layers, mean monthly climate spatial surfaces are generated using ClimateWNA, which downscales precipitation and temperature data generated at 2 - 4 km by PRISM (Wang et al 2006) (Parameter-elevation Regressions on Independent Slopes Model) (Daly et al. 2002) to 1 km. Mean monthly atmospheric VPD for daylight periods was calculated by assuming that the water vapor concentrations present throughout the day would be equivalent to that held at the mean minimum temperature (Kimball et al 1997). The maximum VPD was calculated each month as the difference between the saturated vapor pressure at the mean maximum and minimum temperatures. Mean daytime VPD was calculated as half of the maximum value (Waring 2000). The number of days per month with subfreezing temperatures (<-2 C) was estimated from empirical equations with mean minimum temperature (Coops et al 1998), such that: Number of Frost days = 11.62 – (Tmin * 1.57) Monthly estimates of total incoming short-wave radiation were calculated using a modeling approach detailed by Coops et al (2000) where the potential radiation reaching 184

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests any spot is first calculated and then reduced, based on the clarity (transmissivity) of the atmosphere (Goldberg et al 1979; Bristow and Campbell 1984; Hungerford et al 1989). Changes in the atmospheric transmissivity are mirrored in temperature extremes. With the digital elevation model, we adjusted for differences in slope, aspect, and elevation as well as for variations in the fraction of diffuse and direct solar beam radiation (Garnier and Ohmura 1968; Buffo et al 1972; Swift 1976; Hungerford et al 1989). The modeling approach, when compared with direct measurements, predicted both the direct and diffuse components of mean monthly incoming radiation with 93 - 99% accuracy on flat surfaces, and on sloping terrain accounted for >87% of the observed variation with a mean error < 2 MJ m-2 day-1 (Coops et al 2000).

185

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests

(a)

186

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests

(b)

Figure 3 Examples of generated climate data over Alberta. Annual Precipitation(a) in 2020, 2080 and Maximum July temperature (b) in 2020 and 2080 using 3 climate scenarios (A1B, A2 and B1).

187

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests

WORK PACKAGE 3: PHENOLOGY CAMERAS The phenological camera work for 2011 was focused on relating the detection of phenological events with the availability and quality of food for grizzly bears at the plot and forest stand scale. In previous work we have demonstrated the role that digital remote sensing cameras can play in the monitoring of the vegetation phenological cycle, across a range of sites in Alberta, Canada within grizzly bear habitats. Grizzly bears have diverse diets that change during the course of the year, and individual bears may travel large distances to locate high quality food sources (Rogers, 1987). The diet is known to comprise seasonally abundant, nutrient rich food (Hamer et al., 1991; Craighead et al., 1995; Munro et al., 2006). Despite the significant relationships between vegetation phenology and its remote observation, from a nutritional perspective changes in the vegetation phenological cycle do not directly correspond to the availability of high quality forage. For example, the nutritional content of Hedysarum alpinum (alpine sweetvetch) varies significantly throughout the growing season with its highest nutritional concentration occurring prior to, and during the initial green-up phase. Once the above ground vegetation has reached its phenological peak corresponding to the mid-point of the growing season, the nutritional value of the roots has decreased significantly. The main objectives of the study were therefore twofold. First, changes in vegetation phenology of key individual plant species, critical to the grizzly bear diet, were examined using very high spatial resolution digital camera data. Changes in vegetation phenology for a number of individual plants were examined over a full growing season, and changes in spectral greenness compared to both phenophase observations of the above ground vegetation components of the species, as well as to the nutritional content of the above and below ground components of the species. Once the link between observed changes in greenness and nutritional content is demonstrated, the second objective of the study is to demonstrate if these individual species based images can be scaled up using digital camera observations acquired at the forest stand scale.

Methods Three sites were located near the town of Robb, Alberta (53.2 oN; 117.0oW), to observe the range of phenological changes and growing season attributes across known grizzly bear habitat. At each site a pair of camera systems (Bater et al., 2011) was installed with two different fields; one focused on the individual plant scale, and the second at the forest stand scale. Six digital camera systems were installed at the 3 sites. At each of the three plots, one camera was mounted three metres above the ground on a tall and dominant tree and pointed north. This provided the forest stand scale imagery. A second camera was mounted close to the first camera but with a reduced field of view to approximately 5 x 5 m allowing a small number of individual plants to be imaged. This image set is referred to as the plant scale.

188

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests Field validation, phenophase codes and root nutrition data All camera sites were visited bi-weekly between April and October to record the phenophase codes of the vegetation. Species within the scenes which were sampled included Hedysarum alpinum (alpine sweetvetch), S. Canadensis (buffalo berry), Lathyrus ochroleucus (cream pea), Vaccinium vitis-idaea (lingonberry), Arctostaphylos uva-ursi (Bearberry), and Dryas octopetala (mountain avens). In addition to the phenophase observations H. alpinum plants just outside of the observable images were also sampled for nutritional content. In total, 66 samples were collected at the camera sites and analysed for crude protein content at the University of Alberta (Coogan, 2011). To allow comparisons between the camera information, phenology and the root protein data to be developed, the root samples were grouped into five phenological stages (pre-leaf, leaf, flower, seed, dormant).

189

Chapter 8-Future of Alberta’s Forests

Figure 4 Example images of a) RGB image, b) 2G-RB index, c) Greenup date, d) End of season

Image Analysis: In order to extract the phenological response from the image sequences, the images were ingested and the 5 images per day averaged into a single daily image to decrease the effects of differential exposure and shading in the images and then averaged using a 4 x 4 filter to reduce the data volume and remove the effects of vegetation movement by plant growth or wind. In order to extract a single spectral index from the blue, green and red spectral channels indicative for vegetation activity we calculated the 2G-RB index (Richardson et al., 2007). All days where snow was present in the scene were removed from further analysis and a smoothing spline with rigidity of 2/3 was fitted to the 2G-RBi data for each pixel (Richardson et al., 2009). Key dates from the fitted spline were then extracted including date of greenup, end of season, and flowering of the H.alpinum. To do so, greenup was defined as the first date the greenness was higher than that pixel’s robust halfmax (90th percentile – 10th percentile / 2); end of season was defined as the last date the greenness was higher than that pixel’s robust halfmax, and flowering of H.alpinum was detected as a local minimum in the greenness curve between the greenup date and the end of season. This minimum is caused by the pink flowers of alpine sweetvetch that have a negative 2G-RBi.

190

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests

Figure 5: Dates of mayor phenological events as observed in the field and as derived from the camera spectral response (R2 = 0.89, p < 0.001, n=16).

Results Approximately 850 images were obtained from each camera covering the period from midApril to mid-October. Figure 4(a) provides an example of a typical field of view of the plant scale imagery, taken at the time of flowering of Hedysarum and clearly shows the leaf structure and flowers of plants in the scene. The 2G-RBi image is shown in Figure 4(b) with white areas indicative of very green vegetation and darker areas showing non-vegetated scene components. The derived start and end of growing season images are shown in Figure 4(c) and 4(d) respectively. Relationships between the camera derived dates of phenological events and the field observed dates were strong (r2 = 0.89, p < 0.01, N=16)(Figure 5), except for a few evergreen shrubs (V.vitis-idea, A. uva-ursi) which have too little seasonality in greenness to be accurately detected. The relationship between the field measured phenophase and the nutritional status of H.alpinum is shown in Figure 6, and clearly demonstrates the reduction in protein content through the growing season with pre-leaf having the highest nutritional value and the flower and seed phenophases having the lowest. Also apparent is the high nutritional load of the below ground component when the species is dormant and the variability of root protein across the 3 sites.

191

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests

Figure 6 Average crude protein content of Hedysarum alpinum roots sampled in three locations in west-central Alberta, Canada in different phenological stages. Error bars are 1 standard error above and below the mean

Landscape forage quality Based on the connection between phenological development and root protein content of H.alpinum we are able to infer the nutritional value available in the landscape based on our time-series of stand scale photographs. The clear spectral signature of H.alpinum flowers allows for the detection of this speciesâ&#x20AC;&#x2122; spatial distribution within the landscape. The temporal patterns of major phenological events in the cells containing H.alpinum are shown in Figure 7. Greenup starts halfway through June with the peak of growth two weeks later, flowering lasts approximately three weeks, between mid-July and the first part of August. The decline of greenness starts around mid-August with all aboveground plant parts gone by mid-September. In nutritional terms this would mean that high protein roots are available before the start of July and again after the end of August. This pattern corresponds well with root digging patterns as observed in grizzly bear feces (Munro et al., 2006) which show most root consumption in late April to June, and again starting from late (early in mountain sites) August to October.

Conclusions While broad scale patterns in phenology can provide an overall assessment of available food resources and inter-annual changes in forest productivity, a key limitation of working at this broader scale was the inability to detect subtle or species specific phenological events critical for food modelling. In this research we used two sets of cameras at the plot and forest stand scale. This setup allows for the precise timing of initial leaf unfolding or the development of fruiting bodies to be observed and then linked to the nutritional value of the below ground resource. Placing cameras in close proximity to plants offers the advantage of reduced frequency of field visits for collection of phenological phase data, 192

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests which we demonstrate can be interpreted directly from the images. This study confirms that ground-based cameras can be employed to simultaneously monitor species-specific overstory and understory phenology, and that these images can then be linked to produce overall landscape assessments of vegetation nutritional and energy loads.

Figure 7 Histograms of Greenup, Flowering, End of Season for H.alpinum from the Cardinal Divide- Wide camera.

Forthcoming Publication: Nijland, W., Coops, N.C., Coogan, S., Bater, C.W., Wulder, M.A., Nielsen, S., McDermid, G., Stenhouse, G. (2012) Characterizing vegetation phenology and nutritional value for wildlife management using digital time-lapse photography. Ecological Indicators (in prep).

193

Chapter 8-Future of Alberta’s Forests

WORK PACKAGE 4: LANDIS SIMULATION SUMMARY As part of 2011/12 research we have also commenced the development of spatial layers to capture past and future landscape patterns due to fire and harvesting disturbance regimes. To do so we are using the LANDIS-II modelling framework which is a landscape scale model that works across a range of spatial and temporal scales. Our approach thus far to using the LANDIS-II has focused on the fire prediction and impact module. In order to parameterize the fire module within LANDIS we have utilised data from the August 20, 2008 release of the Canadian National Fire Database (NFDB). Locations of fire were selected within increasing buffers centred over the study area and for each buffer distance fire events will be separated into three decades: 1977-1986; 1987-1996; and 1997-2006. The maximum, minimum, and mean fire size was then calculated for each decade and the average decadal fire density determined by dividing the number of points within each decadal category by the total area. Table 2 and Figure 8 show the location, frequency and size of the fires in the region from the fire database. Table 2 Summary of fire prediction variables based on 100 and 200km radii averages around Hinton. SIZE (ha) 100km Decade interval 77-86 87-96 97-2006 1987-2006 (mean)

Min 0 0 0 0

Max 1186 4598 96130 50364

Mean 4.4 7.05 60 33

N 1776 2064 2491 2277

Average Annual fire density (fires/km2*year) 0.002 0.0024 0.00305 0.002725

The pattern of disturbances generated by STAARCH from 2000 – 2008 were combined with decadal Landsat coverage since 1975 to produce an overall stand age map which will provide detailed age information for stands less than 10 years old, decadal age estimates for stands between 10 and 40 and broad 30 – 50 year classes for stands older than 40 years. This base layer of stand age was provided to LANDIS-II to assess the impact of changing fire size and frequency on the landscape.

194

Chapter 8-Future of Alberta’s Forests

200 km radius 100 km radius

Figure 8: Sampling of fire around Hinton, Alberta.

Figure 9 provides an overview of the distribution of age classes for a 100 year simulation using fire frequency and size data summarised for each decade from Table 2. The figure shows the impact differing fire frequencies have on the overall stand age structure in the region. Periods with higher fire frequency and larger fires such as 1997 – 2006, and the average of 1987 – 2006, show a much younger stand age distribution that regions with lower frequencies (1977 – 1986). Table 3 provides data on the number of cells in the simulations that were burnt by age class as shown in Table 2. Maps of stand age based on the fire parameters are shown in Figure 10 and provide an initial example of how outputs from the LANDIS-II can be used to better understand past

195

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests and future changes in the landscape. Work in this area is continuing as part of year 2 of this climate / landscape simulation program.

Figure 9. Age class histogram (ha) showing distribution of age classes based on 4 scenarios summarized in Table 2 Table 3 Summary of area burned based on variables provided in Table 2

Stand Age 1 10 20 30 40 50 60 70 80 90 100 110 180

Area 1977-1986 2401 5320 5015 5125 392663 1500206 627888 3551 3557 2821 803 344151 330534

1987-1996 8666 16650 16388 15568 390799 1463805 617328 10723 10288 7840 2425 337044 317898

1997-2006 134763 218166 210310 200466 381545 906309 460545 126269 106555 78219 23485 212503 135927

196

1987-2006 270238 260265 183641 168201 266295 576144 318200 248463 177627 521609 8969 86955 100926

Chapter 8-Future of Albertaâ&#x20AC;&#x2122;s Forests

Figure 10: Initial simulations of stand age based on a range of forest fire scenarios described in Table 2.

197

Chapter 8-Future of Alberta’s Forests

References Bater, CW, NC Coops, MA Wulder, T Hilker, SE Nielsen, G McDermid, and GB Stenhouse. 2011. “Using digital time-lapse cameras to monitor species-specific understorey and overstorey phenology in support of wildlife habitat assessment.” Environmental Monitoring and Assessment 180 (1-4) (September): 1-13. Coogan S. 2011 “Getting to the root of the matter: grizzly bears and alpine sweetevtch in west-central Alberta, Canada.” University of Alberta, graduate research thesis. Edmonton, Alberta. Craighead, J.J., J.S. Sumner, and J.A. Mitchell. 1995. The grizzly bears of Yellowstone. Island Press Washington, DC. Hilker, T., Wulder, M.A., Coops, N.C., Seitz, N., White, J.C., Gao, F., Masek, J., & Stenhouse, G.B. (2009). Generation of dense time series synthetic Landsat data through data blending with MODIS using the spatial and temporal adaptive reflectance fusion model (STARFM). Remote Sensing of Environment 113: 1988-1999 Hilker, T., Wulder, M.A., Coops, N.C., Linke, J., McDermid, G., Masek, J., Gao, F. and White, J.C., 2009. A new data fusion model for high spatial- and temporal-resolution mapping of forest disturbance based on Landsat and MODIS. Remote Sensing of Environment, 113, 1613-1627. Gaulton, Rachel, T Hilker, M A Wulder, N C Coops, and Gordon Stenhouse. 2011. “Characterizing stand-replacing disturbance in western Alberta grizzly bear habitat, using a satellite-derived high temporal and spatial resolution change sequence.” Forest Ecology and Management 261 (4): 865-877. doi:10.1016/j.foreco.2010.12.020. Bristow, K L, Campbell, GS. 1984 On the relationship between incoming solar radiation and daily maximum and minimum temperature. Agricultural and Forest Meteorology 31:156-166. Buffo, J, Fritschen, LJ, Murphy. JL. 1972. Direct solar radiation on various slopes from 0 to 60 degrees north latitude. Pacific Northwest Forest and Range Experiment Station Forest Station. U.S. Department of Agriculture, Portland, Oregon, 74 pages. Coops, NC, Waring, RH, Landsberg, JJ.1998. Assessing forest productivity in Australia and New Zealand using a physiologically-based model driven with averaged monthly weather data and satellite derived estimates of canopy photosynthetic capacity. Forest Ecology and Management 104: 113-127. Coops, NC, Waring, RH, Moncrieff, JB. 2000. Estimating mean monthly incident solar radiation on horizontal and inclined slopes from mean monthly temperature extremes. International Journal of Biometeorology, 44, 204-211. Gaulton, R., Hilker, T., Wulder, M.A., Coops, N.C., Stenhouse, G. (2011) Characterising Stand Replacing Disturbance In Western Alberta Grizzly Bear Habitat, Using A Satellite-Derived High Temporal And Spatial Resolution Change Sequence. Forest Ecology and Management 261(4): 865-877 Garnier, BJ, Ohmura, A. 1968. A method of calculating the direct shortwave radiation income on slopes. Journal of Applied Meteorology 7: 796-800. 198

Chapter 8-Future of Alberta’s Forests Goldberg, B, Klein, WH, McCartney, RD. 1979. A comparison of some simple models used to predict solar irradiance on a horizontal surface. Solar Energy 23: 81-83. Hamer, D., and S. Herrero. 1987. “Grizzly bear food and habitat in the front ranges of Banff National Park, Alberta.” Bears: Their Biology and Management 7: 199–213. Kimball, JS, Running, SW, and Nemani, R. 1997. An improved method for estimating surface humidity from daily minimum temperature. Agricultural and Forest Meteorology 85: 87-98. Hungerford, RD, Nemani, RR, Running, SW, Coughlan, JC (1989) MT-CLIM: A mountain microclimate simulation models. USDA Forest Service Research Paper INT-414. Ogden, Utah. IPCC. Summary for Policymakers. In: Solomon SD, Qin M, Manning Z, Chen M, Marquis KB, Averyt M, Tignor H, Miller L, editors. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I; 2007. Munro, R.H.M., SE Nielsen, M. H. Price, Gordon B Stenhouse, and MS Boyce. 2006. “Seasonal and Diel Patterns of Grizzly Bear Diet and Activity in West-Central Alberta.” Journal of Mammalogy 87 (6) Nakićenović N, Swart R. Special Report on Emissions Scenarios. A Special Report of Working Group III of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge; 2000. Richardson, Andrew D, B.H. Braswell, D.Y. Hollinger, J.P. Jenkins, and S.V. Ollinger. 2009. “Near-surface remote sensing of spatial and temporal variation in canopy phenology.” Ecological Applications 19 (6): 1417-1428. Richardson, Andrew D, J P Jenkins, B H Braswell, D Y Hollinger, S.V. Ollinger, and M.L. Smith. 2007. “Use of digital webcam images to track spring green-up in a deciduous broadleaf forest.” Oecologia 152 (2): 323–334. Rogers, L N. 1987. “Effects of food supply and kinship on social behavior, movements, and population growth of black bears in northeastern Minnesota.” Wildlife Monographs 51 (2): 1-72. nutritional status of H.alpinum is shown in Figure 3, and clearly demonstrates the reduction in protein content through the growing season with preleaf having the highest nutritional value and the flower Swift, LW, Jr. 1976. Algorithm for solar radiation on mountain slopes. Water Resources Research 12: 108-112. Wang T, Hamann, A., Spittlehouse, D.L. & Aitken, S.N. 2006. Development of scale-free climate data for western Canada for use in resource management. International Journal of Climatology 26: 383-397. Waring, RH. 2000. A process model analysis of environmental limitations on growth of Sitka spruce plantations in Great Britain. Forestry 73: 65-79.

199

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB

CHAPTER 9: DESIGN OF DNA-BASED OCCUPANCY SAMPLING AND POPULATION ESTIMATION FOR GRIZZLY BEARS IN NORTHERN ALBERTA John Boulanger1, Gordon Stenhouse2, Jerome Cranston3 1

Integrated Ecological Research, 924 Innes, Nelson, BC, V1L 5T2, 250-352-2605, boulange@ecological.bc.ca 2 Foothills Research Institute, Hinton, Alberta, T7V 1X7 3 Arctos Ecological Services, Hinton, Alberta

Introduction This document summarizes a proposed design for DNA sampling of the Northern Alberta grizzly bear management area (GBMA) to determine relative occupancy of grizzly bears. Previous live capture efforts have suggested that there are very low densities of grizzly bears in this area (Stenhouse, unpublished). Given this, the primary objective of sampling efforts will be to estimate the extent of grizzly bear occupancy (MacKenzie et al. 2002) in this area. A secondary objective will be a baseline population estimate for the sampling area. Occupancy models use the sequences of detections and non-detections and sites to estimate the probability of detecting a species, and from that the actual number of sites that were occupied using a mark-recapture methodology that is very similar to estimation of population size (MacKenzie et al. 2002, Boulanger et al. 2008b). From this, the true proportion of occupied sites is estimated and termed the probability of occupancy (Ψ) (MacKenzie et al. 2006). Occupancy estimation involves repeated visits or samplings of a “site” so that the probability of detection of bears and the occupancy probability can be estimated. Factors such as site placement and lower efficiency of hair snag sites may influence the probability of detection. With repeat visits to each site, a series of detection and non-detections is obtained accompanied by the understanding that at some sites bears might have been present but not detected. This type of data collection then allows the estimation of occupancy probability to account for “missed” detections. We propose to design an occupancy sampling grid that uses DNA hair snag methods to obtain hair samples from grizzly bears. Sampling grids will be placed to ensure that they 200

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB are representative of habitats in the area, and sample areas where grizzly bears are most likely to occur based on previously developed food models and risk models (Nielsen et al. 2010). A secondary objective of sampling will be a baseline population estimate for the GBMA. We consider this a secondary objective since it is possible that the number of bears detected in the area may be too low for a precise population estimate. However, we did consider this objective when designing the sampling so that an estimate would be possible if a suitable number of bears were detected.

Methods We used an incremental approach to design the study where we first defined probable levels of both hair snag site detection and occupancy probabilities. Once this was done, we estimated the sampling intensity needed to provide precise occupancy estimates and also considered hybrid sampling designs to allow population estimates. Assessment of previous inventory projects to estimate detection probabilities of hair snag sites Unlike closed mark-recapture population estimation models the hair snag site, rather than the grizzly bear, is the sample unit for occupancy models and therefore detection probabilities for population estimates may not directly correspond to occupancy models. We used fixed site areas of the 2004 Yellowhead (Boulanger et al. 2006) and 2008 Grande Cache grid (Alberta Grizzly Bear Inventory Team 2009) to estimate detection probabilities across a range of grizzly bear densities. Table 1: Habitat variables used to describe habitat-relative occupancy relationships and factors influencing detection probabilities for DNA hair-snag locations in Alberta. Variables were summarized at the site and landscape (10K buffer) scale.

Habitat variable CC RISK DEM CTI TRI Dstream100

Description Crown closure (McDermid 2005) Risk of mortality (Nielsen et al. 2004) Elevation (metres) from a DEM Compound topographic index (a soil moisture index; Gessler et al. 1995) Terrain ruggedness index (Riley et al. 1999) Negative exponential decay (100m parameter) distance to stream

One of the main objectives for occupancy studies is placement of sites to maximize detection probabilities of grizzly bears. To assist in site placement we explored covariates that might affect detection probabilities of grizzly bears (Table 1). The Chinchaga study area is very different from the Yellowhead and Grande Cache study areas and therefore factors affecting occupancy may differ in these areas. However, we assumed that there 201

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB would be some similarity between the areas in terms of factors that affect smaller scale site placement of hair snags. To facilitate this further, we considered site scale detection probability covariates (defined by a 300 meter buffer around the site areas) under the assumption that similar smaller scale factors will influence detection probabilities in the different study areas. We built sequential models for both occupancy and detection probabilities using the covariates in Table 1. Models support was evaluated using the sample size adjusted Akaike Information Criterion (AICc). The model with the lowest AIC score was considered to be most supported by the data. The difference between the most supported models and other models was indexed by the Delta AICc value. Any models with Delta AIC values less than 2 were also considered. The estimates of detection and occupancy from each model were then averaged using the proportional support of each model as estimated by the AICc weight (wi) (Burnham and Anderson 1998). All analyses were done using program MARK (White and Burnham 1999). The analysis of live-capture efforts to define a range of likely occupancy probabilities for the study area We analyzed grizzly bear live capture data from the Chinchaga and Clear Hills area to assess approximate detection rates and occupancy rates. We used either occupancy models or logistic regression (Hosmer and Lemeshow 2000) for this analysis. Estimation of sampling effort needed to obtain occupancy estimates of adequate precision. We used likely levels of occupancy and detection probabilities of hair snags to estimate needed sampling effort (number of HS sites and sessions) to estimate occupancy with a desired level of precision using the formulas of MacKenzie and Royle (2005) Further consideration of sampling designs to obtain approximate estimates of population size A secondary objective of the project will be estimation of a baseline population size for the study area. Due to likely low densities of bears it may not be possible to obtain an estimate with similar levels of precision as the previous DNA inventories conducted in other provincial Bear Management Areas. However, a baseline estimate would still help assess how the NW Alberta area contributes to the overall population estimate for Alberta. We used habitat models based on seasonal food availability and relative risk (Nielsen et al. 2010) to further optimize sampling so that areas predicted to have higher grizzly bears received more sampling effort. This approach was used successfully to reduce sampling effort for the Grand Cache grid while still obtaining precise population estimates (Alberta Grizzly Bear Inventory Team 2009).

Results Estimation of HS detection rate Three hundred forty eight sites from the Yellowhead and Grand Cache grids were used for the occupancy analysis. Of these, 192 were from the Yellowhead 2004 grid (176 fixed 7x7 202

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB km cell sites and 16 transect sites) and 156 were from the Grande Cache 2008 grid (125 7x7 km cell fixed sites and 31 10x10 km occupancy sites). Model selection results suggested that the sampling grid (Grande Cache or Yellowhead), terrain ruggedness (TRI evaluated at the site (300m) scale), and whether the site was helicopter access affected detection probability (Table 2). Occupancy was influenced by terrain ruggedness, compound topographic index, risk factor, and elevation. Models that also considered the effect of proximity to stream and canopy closure on detection probability were also supported. Table 2: Model selection for occupancy model of fixed site data from the Yellowhead (2004) and Grand Cache (2008) DNA inventory projects. No detection probability occupancy AICc Î&#x201D;AICc wi K 1 Grid+ TRI(site) + heli TRI+CTI+risk+DEM 952.1 0.00 0.15 10 2 Grid+ TRI(site) + heli TRI+TRI2+CTI+CTI2+risk+DEM 952.8 0.71 0.10 12 3 Grid+ TRI(site) + heli TRI+CTI+risk+DEM+cell 952.9 0.86 0.10 11 4 Grid+ CC+ TRI(site) + heli TRI+CTI+risk+DEM 953.1 1.00 0.09 11 5 Grid+ stream+ TRI(site) + heli TRI+CTI+risk+DEM 953.1 1.03 0.09 11 6 Grid+ TRI(site) + heli TRI+TRI2+CTI+CTI2+risk+DEM 953.3 1.23 0.08 11 7 Grid+ TRI(site) +stream+ heli TRI+TRI2+CTI+CTI2+risk+DEM 953.5 1.43 0.07 13 8 Trend+Grid+TRI(site) + heli TRI +CTI+risk+DEM 954.0 1.98 0.06 12

As noted earlier, the primary objective of the analysis was exploration of factors affecting detection probabilities. Evaluation of model parameter estimates and odds ratios suggested that sites with higher terrain ruggedness and helicopter access had higher detection rates (Table 3). Odds ratios were based on standardized coefficients and therefore a unit of change in odds ratio represented change of one standard deviation for each of the covariates. In addition, the detection rates for the Grande Cache grid were 4.98 times higher than the Yellowhead grid. In addition, models of lesser support suggested that areas of higher canopy closure had lower detection rates (odds ratio=0.87, CI=0.67-1.12) and detection probability decreased with increasing distances from stream areas (odds ratio=0.90 CI=0.75-1.09). Odds ratios also suggested that occupancy increased in areas of higher risk and elevation on the Grande Cache grid while decreasing in areas of higher terrain ruggedness and compound topographic index (Table 3). Note that terrain ruggedness was evaluated at the larger 10K scale for occupancy whereas detection rates were at the site scale. So this suggests that sites in areas of higher terrain ruggedness are more likely to detect bears even though bears may be less likely to occupy areas of larger scale terrain ruggedness. One example of this might be placement of sites in avalanche chutes or valley bottoms (of high local terrain ruggedness).

203

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB Table 3: Parameter estimates from the most supported occupancy model along with odds ratios of slope estimate parameters

Label

Estimate SE

LCI

UCI

Odds ratio

Detection probabilities

intercept TRI(site) Helicopter GC vs YH

-2.33 0.60 0.74 1.61

0.37 0.11 0.38 0.39

-3.06 0.38 -0.01 0.84

-1.61 0.81 1.50 2.37

1.81 2.10 4.98

1.46 0.99 2.32

2.25 4.47 10.68

0.43 -1.11 -1.73 1.42 1.14 0.46

0.58 0.72 0.70 0.49 0.48 0.66

-0.71 -2.52 -3.09 0.46 0.21 -0.84

1.57 0.29 -0.37 2.38 2.08 1.76

0.33 0.18 4.14 3.14 1.58

0.08 0.05 1.58 1.23 0.43

1.34 0.69 10.79 8.00 5.80

Occupancy

intercept TRI (10k scale) CTI (10k scale) Risk Elevation GC vs YH

Model averaged estimates of occupancy and detection rates revealed slightly higher levels of occupancy for the Grande Cache study area and much higher detection probabilities (Table 4). It should be noted that the fixed sites for the Grande Cache area were in Jasper Park where site placement was primarily in valley areas whereas fixed sites for the Yellowhead grid were interspersed among foothill areas. Therefore, it is likely that optimal site placement and selection was easier for the Jasper area compared to Yellowhead grid. Table 4: Model averaged estimates of occupancy and detection probability (from models in Table 2)

Paramter Estimate Occupancy rates

LCI

UCI

2004 2008

0.62 0.72

0.14 0.08

0.34 0.55

0.83 0.84

0.10 0.36

0.03 0.04

0.06 0.29

0.17 0.43

Detection rates 2004 2008

Estimation of occupancy from snare data Twenty three live capture trapping sites were established in the Chinchaga and Clear Hills areas in 2006 and 2007. Sites were trapped on average for 17.6 days (min=2 max=29 days) for a total of 410 trap nights. Four grizzly bears were captured in 4 different sites for the entire trapping effort. We initially considered the use of occupancy models for this data set but the number of detections were too low for models to converge. We therefore used logistic regression to estimate the na誰ve occupancy (the probability of capture X true occupancy). From this we estimated na誰ve occupancy at 0.009756 (CI=0.0036-0.0256). 204

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB This is the per-session probability of detecting a grizzly bear at a snare site. The main challenge in interpreting this estimate is that the detection or per-session capture probability of the snare sites is unknown and therefore true occupancy cannot be determined. To explore potential values we considered a range of true occupancy values as a function of a range of per-session detection rates. More exactly, we solved the equation na誰ve occupancy=true occupancy * detection rate for true occupancy using the na誰ve occupancy value of 0.0097) (True occupancy 率 =0.009756/detection rate). From Figure 1 it can be seen that true occupancy values up to 0.4 are possible if the per-session detection probabilities of the snares are as low as 0.05. 0.45 0.4 True Occupancy

0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0

0.1

0.2

0.3

0.4

0.5

Live capture probability (per trap night)

Figure 1: Levels of occupancy assuming as a function of nightly capture probabilities of live capture sites assuming a na誰ve occupancy rate of 0.00975 of snare sets.

Designs needed to obtain adequate occupancy estimates Using the formulas of MacKenzie and Royle (2005), we assessed the number of sites and sampling sessions needed to obtain occupancy estimates with a coefficient of variation of less than 20% (Figure 2). From this, it can be seen that if detection rates are low then at least 4 sessions are needed to obtain adequate precision of estimates with a reasonable number of sites. MacKenzie and Royle (2005) comment that if occupancy is low (and detection rate is reasonable (>0.2), then it is optimal to sample more sites. If detection rate is low (<0.2), but occupancy is reasonable (>0.2) then it makes more sense to sample for more sessions.

205

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB

6000 Sites visited

5000 4000

3000 2000 1000 0 0

6 Sessions

p(0.2) Ψ(0.2) p(0.36) Ψ(0.4)

p(0.36) Ψ(0.2) p(0.1) Ψ(0.36)

Figure 2: The estimated number of sites and sessions needed to obtain occupancy estimates with a coefficient of variation of 20% as a function of detection probabilities (p) and occupancy (Ψ).

Given these limitations, we suggest that a 4 session design will be needed if the number of sites is kept to a moderate (200-300) number. Assuming a 200 site design we further considered levels of detection rate and occupancy needed to obtain estimates with adequate levels of precision for a 2 session and 4 session sampling design (Figure 3). For the 2 session design, it was difficult to obtain occupancy estimates with adequate precision if detection probability was 0.1 or lower. If detection probability was 0.2 or higher then occupancy needed to be at least 0.3 to obtain estimates of adequate precision. For the 4 session design, it was possible to get estimates of adequate precision with detection rates of 0.1 if occupancy was 0.3 or higher. This analysis identifies thresholds needed to obtain estimates of occupancy with adequate precision. Namely, if 2 sessions are conducted then detection rate must be above 0.2 given that occupancy may be as low as 0.3. If 4 sessions are conducted, then detection rates can be as low as 0.1 if occupancy is near 0.3. As discussed earlier, occupancy is basically the percentage of sites that grizzly bears encountered (while accounting for detection probability). So, in essence, these designs apply to the assumption that 20-30% of the sites will be in habitat that is traversed by grizzly bears. We suspect that detection probabilities may be between 0.1 and 0.2 given likely challenges in optimal site placement in the Chinchaga. Given this, we recommend that a 4 session design is most optimal to ensure occupancy estimates with adequate precision.

206

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB 2 sampling sessions: 0.6

CV occupancy

0.5 0.4 0.3 0.2 0.1 0 0

0.2

0.4

0.6

0.8

Occupancy 0.1

0.2

0.3

0.4

4 sampling sessions 0.4 0.35

CV occupancy

0.3

0.25 0.2 0.15 0.1 0.05 0 0

0.2

0.4

0.6

0.8

Occupancy 0.1

0.2

0.3

0.4

Figure 3: Estimated coefficient of variation of occupancy estimates as a function of occupancy and associated detection probabilities (lines) for a study with 2 and 4 sampling sessions and 200 sampling sites.

Grid cell size needed for occupancy estimation The total area of the Chinchaga study area is 19,824 km2 (Figure 4). Using this area, we approximated the number of cells with 1 site that would be needed to sample the entire 207

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB study area (Table 5). From this, a 10x10 grid cells size was chosen to be the optimal sampling intensity for occupancy estimation given that it resulted in the number of sites being close to 200. Unlike estimation of population size, the site is the sample unit for occupancy estimation and therefore the closer spacing of sites to redetect bears is not as essential as sampling an adequate number of sites needed to obtain precise occupancy estimates. Table 5: The approximate number of cells needed to sample the Chinchaga study area 19,824 km2 Cell size Number of cells 7x7 (49 km2) 405 8x8 (64 km2) 310 9x9 (81 km2) 245 10x10 (100 km2) 198

Figure 4 illustrates a potential grid setup with 10x10 km cells. To further refine the boundaries of the grid we used predictions of habitat value based on seasonal food and mortality risk models for grizzly bears (Nielsen et al. 2010). This grid contains 209 10x10 cells.

208

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB

Figure 4: Food model based predictions of bear habitat in the Chinchaga (June 1-15).

Sampling requirements to allow estimation of population size A second objective of sampling is to provide a baseline population estimate for the Chinchaga study area. For this objective, the occupancy based site density is less optimal and ideally site densities should be similar to other DNA inventory projects that were conducted in Alberta (1 site per 49 km2). However, to sample the entire area at this site density is not necessarily cost effective given the low habitat value and likely low grizzly bear density for much of the Chinchaga. We propose that 2 sites be placed in 10x10 km cells that traverse the higher food value areas and the area between the 2 patches. This sampling intensity will approximate that of previous DNA inventories (i.e. 1 site per 50 km2) which would allow a meta-analysis approach (Boulanger et al. 2002, Alberta Grizzly Bear Inventory Team 2007;2009) that combines data from previous studies to assist in the modelling of detection probabilities for estimation of population size and density. When 2 sites are placed in a cell they would

209

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB be spaced at least 5 kilometers apart to minimize issues with non-independence of sample sites for estimation of occupancy.

Figure 5: Proposed sampling grid with cells that would receive 2 sites outlined in blue. This amounts to 98 sites. Other cells would receive 1 site.

This stratified approach would provide higher recapture rates for bears that likely inhabit the higher habitat value area as well as sample any movement between these 2 areas. For modeling of occupancy the cells with 2 sites would have higher detection rates than cells with one site. To account for this issue, occupancy analyses would be stratified by the 2 cell types to account for difference in detection probabilities. For population estimates, bear detection rates would be modeled based upon the relative proportions of detections in the higher and lower site density areas therefore accounting for heterogeneity in detection probabilities that could be caused by sampling at 2 different spatial intensities. In addition, spatially explicit mark-recapture methods (Efford 2004, Efford et al. 2004, Efford et al. 2007) would also be used to model the effect of differential site densities on detection rates and obtain estimates of population density. 210

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB Optimized study area size design One alternative design that was considered sampled a proportion of the overall study area to further reduce field costs. The strategy behind this design is to intensively sample the core area of highest and medium habitat value to allow a solid estimate of occupancy and baseline population size. Areas to the east would receive low intensity sampling with the primary objective of estimating occupancy for this area (Figure 6). The core area is composed of 61 10x10 km cells that would receive 2 sites each and 35 10x10 km cells in medium quality habitat that would receive 1 site per cell. The occupancy area to the east would receive 1 site per 20X20 km cell (23 cells). Overall, this would result in 180 sites in total given that 61 cells would receive 2 sites. The analysis strategy for this design would be to estimate population size for the core area alone given that site coverage (1 site per 20x20 km cell) would not be sufficient for population estimation in the eastern area. Variation in detection rates caused by differential site densities would be modeled using individual covariates that pertained to the proportion of detection in areas of different site densities. As with the full sampling design, spatially explicit mark-recapture methods (Efford 2004, Efford et al. 2004, Efford et al. 2007) would also be used to model the effect of differential site densities on detection rates and obtain estimates of population density. Occupancy would be modeled for the entire sampled grid with an emphasis on determining covariates that influence occupancy. Population size would then be extrapolated to the eastern area using the baseline densities estimated from the core area. This basic method was also used to estimate population sizes for the Swan Hills management area (Boulanger et al. 2009). Sampling the lower quality habitat to the east of the core area will ensure that this area is well represented in occupancy models. A key covariate will be distance from the core higher quality habitat areas for it would be expected that occupancy rates would decline at further distances from these areas.

211

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB

Figure 6 Proposed optimized sampling grid with core area cells that would receive 2 sites outlined in blue (61 cells), cells that would receive 1 site outlined in green (35 cells), and 20x20 km cells that would receive 1 site (23 cells).

Figure 7 displays threshold occupancy and detection probabilities needed for this design (assuming 4 sampling sessions). For this design, occupancy would have to be 0.2 if detection probabilities were 0.2, or occupancy would have to be 0.35 if detection probabilities were 0.1. In summary, this design is most viable if it can be assumed that the majority of bears inhabit the higher habitat value area and occupancy rates are reasonably high in this core area.

212

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB

0.45 0.4

CV occupancy

0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0

0.2

0.4

0.6

0.8

Occupancy 0.1

0.2

0.3

0.4

Figure 7: Estimated coefficient of variation of occupancy estimates as a function of occupancy and associated detection probabilities (lines) for the reduced area study design with 4 sampling sessions and 180 sampling sites.

Budgets for proposed designs Figure 8 below provides an assessment of road or quad access to the cells in the Chinchaga study area. From this we estimate that roads occur within 174 of the 209 cells in the study area. Using random site placement, we estimated that 90 of the 209 sites are within 1 km of a road (blue), 129 of the 209 sites are within 2 km of a road. Therefore, we estimate that approximately 140 sites should be accessible using ATV or foot access for the full study area and 80 sites should be accessible by road for the reduced study area. Eventually, the actual sites for the study area will be selected based on habitat features which would allow a more refined estimate of site access. The estimated cost for the full area design is $450,250. The estimated cost for the optimized design is $312,950. The main difference in cost between these 2 designs is reduced helicopter time with the optimized design. The details of each budget are given in Appendix 1.

213

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB

Figure 8: The Chinchaga study area with optimized sampling areas and road access areas denoted. Each road is buffered by 1 kilometer.

Discussion We emphasize that it is challenging to sample a large and relatively remote area that is likely to have low densities of grizzly bears. The designs we propose are optimized to provide an estimate of occupancy of grizzly bears in the Chinchaga study area as well as provide information on the primary factors influencing occupancy in the study area. As discussed in the next section, the actual optimal design will depend on management priorities for this grizzly bear management area. We cannot guarantee that any of the proposed designs will provide precise population estimates if very few bears (<25 grizzly bears) are detected on the study grid. It will be essential that sites are placed in optimal grizzly bear habitat and be optimized for smallerscale habitat features. Our analyses suggest that areas of higher terrain ruggedness, near streams, increases detection probabilities of grizzly bears. Use of remote sensing technology for a-priori site selection will help assist in determining the general areas for sites. In addition, collection of covariates such as weather, small-scale site characteristics as well as remote sensing based larger scale factors will be essential for the modeling of 214

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB occupancy and detection probabilities. The exploratory models used for the occupancy analysis (Table 2) had 10-12 parameters and therefore there needs to be at least 120 sites sampled to support these models. The current design (with approximately 270 sites) will therefore provide adequate sample sizes for occupancy analysis. We also suggest that alternative data sources such as rub trees and scat sampling be utilized if possible during field sampling. Genotyping success of scat samples is still low in Alberta field trials, however additional work on this topic continues within the Foothills Research Institute Grizzly Bear Program. Rub tree sampling has been successively applied to Banff National Park and the Northern Divide Ecosystem in Montana (Boulanger et al. 2008a, Kendall et al. 2009). However, it is uncertain how well rub tree sampling can be applied to the Chinchaga given the large difference in forest types as well as likely lower densities of grizzly bears in the area. If rub tree sampling can be applied it will increase both estimates of occupancy and population size by allowing the addition of an extra sampling session (Boulanger et al. 2008a).

Recommendations The relative costs and risks associated with each design are outlined in Table 6. The actual design that should be used depends on the relative risk that is acceptable with the given objectives of the study. If an estimate of population size is of most importance than we suggest that the full area design is best. The full area design samples the majority of the study area uniformly and therefore has the least risk for obtaining an occupancy estimate and baseline population estimate. The sampling intensity for the highest habitat value area and area connecting these 2 patches is similar to that used in previous DNA surveys. Therefore, if a reasonable population size of bears is in this area then it should be possible to obtain a population estimate. While sampling is more intense in the core area it also samples the areas to the east with reasonable intensity which will allow the development of more detailed occupancy models for this area. Therefore, estimates of occupancy and population size will be reasonably robust even if bear distribution is different than that predicted by food models. The optimized design assumes that the majority of the bears in the area inhabit areas of higher habitat value, and that a model-based method can be used to estimate densities in the area of lesser intensity (20x20 km grid cells). If bear distribution does not follow that predicted by the food models (for example more bears occur to the east) then it is likely that both occupancy and population estimates will be less precise than the full 4 session design. Therefore, this design has more inherent risk associated with it than with the full design. Finally, we emphasize that obtaining a precise population estimate for this GBMA may be problematic if the actual population size of bears in the sampling area is small (<25 bears). The designs that were considered for this study therefore are optimized for occupancy 215

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB estimation with the added secondary objective of a baseline population estimate if sample size of detected grizzly bears are suitable. Table 6: The benefits and risks with various proposed designs for the NW Alberta Chinchaga study area. See Appendix 1 for budget details.

Design Full area-4 sessions

Details  209 10X10 km cells  98 cells with 2 sites  120 cells with 1 site  316 sites total  4 sampling sessions $ Cost= $ 405,250.00

Benefits and risks  Basic 4 session design is compatible with previous DNA studies in Alberta to allow meta-analysis to estimate N  Samples the highest value habitat and moderate realized habitat (Figure 4)  Precise occupancy estimate if detection probabilities ≥ 0.1 and occupancy is ≥ 0.3 or detection probabilities ≥ 0.2 and occupancy ≥ 0.18 (Figure 3)  Allows a robust population estimate for the entire study area if the number of bears in the area ≥ 25.

Optimized area - 4 sessions



67 10X10 km cells with 2 sites each  35 10x10 km cells with 1 site  23 20x20 km cells with 1 site  182 sites total  4 sampling sessions $ Cost= 312,950.00



 



Basic 4 session design is compatible with previous DNA studies in Alberta to allow meta-analysis to estimate N in the core areas. Samples the highest value habitat assuming lower densities/occupancy in lesser value habitats (Figure 7). Uses model based method to extrapolate density to eastern areas. Precise occupancy estimate if detection probabilities ≥0.1 and occupancy is ≥0.35 or detection probabilities ≥0.2 and occupancy ≥0.2 (Figure 8) Allows baseline population estimate for core area if the number of bears in the core area ≥ 25. Model based methods to extrapolate to areas to the east of the core area (Figure 6)

216

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB Appendix 1: Budget details Item

Estimated costs Full area Optimized area

Quantities

Comments

Contracts/ Professional Fees J. Boulanger

5,500

10 days

n/a

10 days

25,000

100days

6,750

15 days

20,000

3,500

60,750

76,800

60,000

4,700

3,200

81,500

63,200

18,000

15,000

3,000

15,000

12,000

GS Travel

2,000

General Travel

3,000

41,000

35,000

G. Stenhouse

Project Field Leader Jerome Cranston Wildlife Genetics International FRI Research crew costs Total

5,500 n/a

Project statistician, study design, analyses Assist Project coordinator, site selection, study design Field leader, site selection, crew management, communications GIS Support Genetic lab costs

20 days

Logistic assistance

Wages & Benefits Part Time Staff Payroll Costs Total

4 teams = 8 people $200/day

Travel Meals, Food Accommodation Training Accommodation Field

Total Helicopter

100,000

50,000

Fuel

25,000

14,000

Total

125,000

64,000

Charter

100hours

20 days @5 hours/day = 100 hrs 1.30/litre

Vehicles ATV Costs

9,000

Gas & Oil

12,000

Insurance

4,000

20,000

Truck Lease Costs

217

2K x2.5 months x 4 vehicles

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB Item

Estimated costs Full area Optimized area

Repairs & Maintenance

4,000

49,000

Bait/ Blood

4,000

3,000

Sea Can

2,000

GPS

2,000

Field Supplies

5,000

4,000

13,000

11,000

Office Supplies

2,000

Administrative Support

5,000

Total

7,000

Radio, Sat & Cell Phone

6,000

First Aid & Quad

2,000

Contingency

20,000

15,000

Total

28,000

23,000

Total

Quantities

Comments

Equipment

Total

Extra fish or blood Bait storage container & move

Office & Administration Maps & printing

Utilities, training, and contingency

Total Expenses

$405,250

$312,950.

Literature cited AlbertaGrizzlyBearInventoryTeam. 2007. Grizzly bear population and density estimates for the 2006 Alberta Unit 5 Management Area inventory. Alberta Sustainable Resource Development, Fish and Wildlife Division. _____. 2009. Grizzly bear population size and density estimates for the 2008 DNA Inventory of the Grande Cache Bear Management Area (BMA 2). Alberta Sustainable Resource Development, Fish and Wildlife Division. Boulanger, J., J. Cranston, S. E. Nielsen, and G. Stenhouse. 2009. Estimation of grizzly bear population size for the Swan Hills management unit using DNA sampling and habitat-relative occupancy models. Alberta Sustainable Resource Development. Boulanger, J., K. C. Kendall, J. B. Stetz, D. A. Roon, L. P. Waits, and D. Paetkau. 2008a. Multiple data sources improve DNA-based mark-recapture population estimates of grizzly bears. Ecological Applications 18:577-589.

218

Chapter 9-Occupancy Sampling/Population Estimate for Northern AB Boulanger, J., M. Proctor, S. Himmer, G. Stenhouse, D. Paetkau, and J. Cranston. 2006. An empirical test of DNA mark-recapture sampling strategies for grizzly bears. Ursus 17:149-158. Boulanger, J., G. C. White, B. N. McLellan, J. G. Woods, M. F. Proctor, and S. Himmer. 2002. A meta-analysis of grizzly bear DNA mark-recapture projects in British Columbia. Ursus 13:137-152. Boulanger, J., G. C. White, M. Proctor, G. Stenhouse, G. MacHutchon, and S. Himmer. 2008b. Use of occupancy models to estimate the influence of past live captures on detection probabilities of grizzly bears using DNA hair snagging methods. Jounal of Wildlife Management 72:589-595 Burnham, K. P., and D. R. Anderson. 1998. Model selection and inference: A practical information theoretic approach. Springer, New York, New York, USA. Efford, M. 2004. Density estimation in live-trapping studies. Oikos 106:598-610. Efford, M., D. L. Borchers, and A. E. Byrom. 2007. Density estimation by spatially explicit captureâ&#x20AC;&#x201C;recapture: likelihood-based methods. Ecological and environmental statistics. In review. Efford, M., D. K. Dawson, and C. S. Robbins. 2004. DENSITY: software for analysing capturerecapture data from passive detector arrays. Animal Biodiversity and Conservation 27:217-228. Hosmer, D. W., and S. Lemeshow. 2000. Applied Logistic Regression Analysis: Second Edition. Wiley and Sons, New York. Kendall, K. C., J. B. Stetz, J. Boulanger, A. C. Macleoud, D. Paetkau, and G. C. White. 2009. Demography and genetic structure of a recovering grizzly bear population. Journal of Wildlife Management 73:3-17. MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248-2255. MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. L. Bailey, and J. Hines. 2006. Occupancy estimation and modelling: Inferring patterns and dynamics of species occurences. Academic Press, New York. MacKenzie, D. I., and J. A. Royle. 2005. Designing occupancy studies: general advice and allocating survey effort. Jounal of Applied Ecology 42:1105-1114. Nielsen, S. E., S. Herrero, M. S. Boyce, R. D. Mace, B. Benn, M. L. Gibeau, and S. Jevons. 2004. Modeling the spatial distribution of human-caused grizzly bear mortalities in the Central Rockies ecosystem of Canada. Biological Conservation 120:101-113. Nielsen, S. E., G. McDermid, J., G. Stenhouse, and M. S. Boyce. 2010. Dynamic wildlife habitat models: Seasonal foods and mortality risk predict occupancy-abundance and habitat selection in grizzly bears. Biological Conservation 143:1623-1634. White, G. C., and K. P. Burnham. 1999. Program MARK: Survival estimation from populations of marked animals. Bird Study Supplement 46:120-138.

219

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden

CHAPTER 10: HABITAT SELECTION FOR REGENERATING CLEAR-CUT FORESTS BY BROWN BEARS: A COMPARATIVE ANALYSIS BETWEEN ALBERTA AND SWEDEN Scott E. Nielsen1, Andreas Zedrosser2, Jerome Cranston3, Jonas Kindberg4, Gordon B. Stenhouse5, and Jon E. Swenson6 1

Department of Renewable Resources, University of Alberta, Edmonton, Canada T6G 2H1

Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, Post Box 5003, NO-1432 Ås, Norway and Department of Integrative Biology and Biodiversity Research, Institute for Wildlife Biology and Game Management, University of Natural Resources and Applied Life Sciences, Vienna, Gregor Mendel Str. 33, A-1180 Vienna, Austria

Arctos Consulting, Hinton, Alberta

Wildlife, Fish, and Environmental Studies, Swedish University of Agricultural Sciences, SE-901 83 Umeå, Sweden

G.B. Stenhouse, Foothills Research Institute, P. O. Box 6330, Hinton, Alberta T7V 1X6, Canada

J.E. Swenson, Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, Post Box 5003, NO-1432 Ås, Norway and Department of Integrative Biology and Biodiversity Research, Institute for Wildlife Biology and Norwegian Institute for Nature Management, NO-7485 Trondheim, Norway.

Executive Summary Brown bears are generalist omnivores that frequently use disturbed habitats – including young regenerating clear-cut forests – to meet their resource (food) demands. Little is known, however, about how specific forest management practices, including silvicultural treatments and harvest design (e.g., size and shape of clear-cuts), affect bear habitat, how this varies among sex and offspring status classes and populations, and how availability of these or alternate habitats affects their habitat use (i.e., functional responses in habitat selection). To address these knowledge gaps, we examined seasonal habitat selection of regenerating clear-cut forests ≤ 60 years of age by brown bears for two long-term studies occurring in different hemispheres (Alberta and Sweden), but within a shared Holarctic boreal forest ecosystem. Using ~1.9 million GPS telemetry locations for 157 individuals across 12 different years (1999-2010 in Alberta; 2005-2010 in Sweden), we developed seasonal (monthly) habitat selection models between May and October for each individual bear and evaluated secondly how selection (response variable) varied among bears (sexoffspring classes), seasons, habitat availability, and study areas using generalized linear models. We found that brown bears selected regenerating clear-cut patches relative to all 220

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden other habitats available (mostly mature forests) with peak selection occurring most often between 10 to 25 years post-harvest in Alberta and 20 to 40 years post-harvest in Sweden. Bears in Sweden selected clear-cuts consistently across all seasons, while bears in Alberta reduced their selection for clear-cuts in late summer and fall when younger-aged clear-cuts were selected. This seasonal reduction in habitat selection for clear-cuts in Alberta may be due to negative effects associated with mechanical site preparation (scarification) that reduce populations of long-lived fruiting shrubs, while providing a short-term pulse in short-lived fruiting resources, such as raspberry, in younger clear-cuts. In both study areas, dominant bears selected larger clear-cut patches over that of smaller clearcuts suggesting that security (hiding cover) was less important than the size of the resource patch which presumably would increase foraging efficiency. Recent trends in Alberta towards harvesting larger clear-cuts may therefore benefit dominant individuals assuming human access is managed. Females with cubs of the year, however, selected for smaller clear-cuts presumably for greater security and avoidance of dominant males. We found evidence for functional responses in habitat selection where availability of habitats being selected affected selection for those habitats. Future work should consider the effects of silvicultural treatments, including silvicultural thinning which is common in Sweden, but rare in Alberta, on food resource abundance and habitat use by brown bears.

Introduction Brown bears are generalist omnivores that frequently use or even depend on disturbed habitats to meet their resource (food) demands. Dominant disturbance regimes common to brown bear habitat include both natural sources, such as fire (Zager et al., 1983), avalanches (McLellan and Hovey, 2001) and insect epidemics (McLellan, 1989), and anthropogenic sources, such as roadside verges (Roever et al., 2008a) and forest clear-cut harvesting (Nielsen et al., 2004a). In some cases the dependence upon these disturbances may lead to attractive sink dynamics (Nielsen et al., 2006; 2008) whereby anthropogenically-derived disturbances, such as roadsides or clear-cuts, attract bears to these areas due to the presence and concentration of food resources (Nielsen et al., 2004b; Roever et al., 2008b), although use of these areas also increase mortality risk (Nielsen et al., 2004c). How best to manage anthropogenic disturbances in order to minimize their negative effect (i.e., reduced survival associated with human-bear conflict), while also potentially enhancing their positive effect (i.e., habitat carrying capacity) is a major management challenge (Nielsen et al., 2008). Since many brown bear populations occur in human-managed forested environments, modifications to forest management likely offers the greatest opportunity for managing brown bear habitat and populations. Although some studies have found that brown bears avoid clear-cut harvests (Zager et al., 1983; McLellan and Hovey, 2001), many others have found habitat use of clear-cuts to be either equal to availability (neutral) or selected (Mace et al., 1996; Wielgus and Vernier, 2003; Nielsen et al., 2004a; Stewart et al., 2012). Nielsen et al. (2004a) suggested that differences observed among studies may simply be due to landscape effects as it relates to availability of natural disturbances (e.g., fires, avalanches) and natural openings. Where openings are limited, such as in fire-suppressed forested ecosystems, bears readily use 221

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden clear-cut harvests as an alternate habitat. This suggests the presence of a functional response in habitat selection (i.e., Mystrud and Ims, 1998) where selection for clear-cuts should be inversely related to the amount of natural disturbances or openings. As well as understanding landscape-level compositional effects on use of clear-cuts by grizzly bears, more information is needed on how specific forest management practices, including silvicultural treatments and harvest design (e.g., size and shape of clear-cuts), affect bear habitat. Although numerous studies have examined general habitat selection for clear-cuts relative to other habitats, few have examined the details regarding how the characteristics (age, size, site history, etc.) of individual clear-cuts themselves affect their use by bears (see however, Nielsen et al., 2004a; Stewart et al., 2012). If clear-cut size and shape affect the use of those habitats by brown bears, then simple changes can be made to forest harvest designs to benefit bears. Likewise, understanding the specific ages or conditions when brown bears select clear-cuts would help to better design harvest rotations, silvicultural treatments to extend or limit attractiveness to bears, and timing for managing human access. Given these knowledge gaps and management needs, we tested 3 primary hypotheses: (1) regenerating forest hypothesis; (2) landscape structure hypothesis; and (3) functional response (landscape composition) hypothesis. The regenerating forest hypothesis (H1) predicted that brown bears would select regenerating clear-cuts over mature forests due to dietary advantages associated with disturbed habitats. The landscape structure hypothesis (H2) predicted that irregular-shaped and smaller-sized clear-cuts would be selected (on a per unit area basis) over more regular-shaped and larger clear-cuts when clear-cuts were young since there would be little hiding (security) cover. Conversely, larger clear-cut patches should be selected over smaller clear-cuts for older clear-cuts since food resource abundance should be more concentrated due to the larger size of the resource patch and hiding cover sufficient for security purposes. And finally the functional response (landscape composition) hypothesis (H3) predicted that the strength of selection for clearcuts would be locally dependent on the availability of clear-cuts or other open-vegetated (alternate) habitats. We assessed these hypotheses using a comparative analytical approach between two study areas â&#x20AC;&#x201C; Alberta and Sweden â&#x20AC;&#x201C; where brown bears have been studied within an actively managed forest landscape. This comparative approach emphasizes natural differences in forest management practices between regions, while also evaluating whether consistent patterns of habitat selection emerge despite study area differences. If so, this would suggest common behavioral mechanisms of habitat selection in brown bears for recent disturbances in forested landscapes, while also helping to identify the habitat value of actively managed forests and different harvest designs (e.g., clear-cut size, shape, etc.) for brown bears.

222

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden

Material and methods Study areas Alberta: The Alberta study area covered 52,537 km2 and was located in the eastern foothills of the Canadian Rocky Mountains in west-central Alberta (~53.5oN, -117.5oW; Figure 1). Elevations vary from 574 m to 3490 m in rolling low mountainous terrain (just east of the Rocky Mountain ranges). Summer and winter temperatures average 11.5 and 6.0 ºC, respectively, with a normal annual precipitation of 538 mm (Beckingham et al., 1996). Snow cover lasts from late October until early May with the growing season averaging 160 to 185 days. The area is managed intensively with a number of resource extraction activities occurring including forestry, oil and gas, and open-pit coal mining (White et al., 2011). The forest harvest footprint covers 12.2% of the study area (Table 1) with road densities averaging 0.54 km/km2. Linear exploration disturbances (seismic lines) are also common throughout the area and are frequently kept open due to human recreational activity. Closed conifer forests are dominated by lodgepole pine (Pinus contorta), and to a lesser extent by spruce (Picea glauca, P. mariana, P. engelmannii). Deciduous stands of trembling aspen (Populus tremuloides) and/or balsam poplar (P. balsamifera) are more uncommon in the southern (Yellowhead) part of the study area (Nielsen et al., 2004a), but more common or even dominant in parts of the northern (Grande Cache) study area. Prior to 1950, periodic stand-replacing fires were the primary disturbance, averaging 20% of the landscape burned per 20-year period yielding a 100-year fire cycle (Andison, 1998). Since the 1950s, however, there has been a reduction in fires which is associated with the initiation of industrial forestry and fire suppression (Andison, 1998; Nielsen et al., 2008). Although some stands in remote regions are in advanced stages of succession due to fire suppression, large areas have been or continue to be harvested providing the only major mechanism of forest disturbance. The majority of the study area occurs on crown land (federally owned), but is managed by the province of Alberta for timber under long-term (40-year) forest management agreements with forest companies. Traditional forest management in the area centers on a two-pass harvest design with small clear-cuts (<40 ha) designed in a checkerboard pattern within a larger management block (Nielsen et al., 2008). Adjacent stands are subsequently harvested after at least 15 years (reforestation green-up period) has passed (Smith et al., 2003). Mean cutblock size in the study area in Alberta was 25.4 ha. Timber extraction in the Alberta study area is still within its first rotation. Sweden: The study area in Sweden covered 15,933 km2 and was located in rolling low mountainous terrain in the Dalarna and Gävleborg counties of south-central Sweden (~61oN, 14oE; Figure 1). Elevations range from 200 m in the east to 1000 m in the west with timberline at 750 m. Very little of the landscape is above timberline. Summer and winter temperatures averaged 15 and -7.0 ºC, respectively, and precipitation averaged 500–1,000 mm annually. Snow cover lasts from late October until April with a growing season of approximately 180 days. The area is covered by an intensively managed boreal forest 223

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden interspersed with natural bogs and lakes. Roads are common within the Sweden study area with an average density of 0.99 km/km2. Forests are dominated by Scotch pine (Pinus sylvestris) and Norway spruce (Picea abies), although deciduous trees such as mountain birch (Betula pubsecens), silver birch (Betula pendula), European aspen (Populus tremula), and gray alder (Alnus incana) are also common (Zedrosser et al., 2006). Commercial exploitation of timber in the Swedish study area started in the late 1800s (Linder and Östlund, 1998). Prior to exploitation the forest was dominated by old forest stands >200 years of age. Forest commons and the State Forests were established in the late 1800s, when delineation (separation of private and public land) and enclosure (division of land among the different landowners in the villages) were carried out. Most of the unclaimed forest land was allotted as forest commons to landowners in existing villages, but some parts were designated State Forests under the administration of the National Forest Service (Linder and Östlund, 1998). This resulted in an owner structure that was characterized by the possession of several spatially disconnected and irregular-shaped and often small stands. Subsequent intensive forest management and fire protection have resulted in a forest landscape dominated by relatively young and dense stands (Linder and Östlund 1998). Since the late 1800s, both the number of large trees and the volume of snags have been reduced by about 90%, and the area of old stands has diminished to < 1% (Linder and Östlund, 1998). Until the 1940s, natural regeneration without any soil treatment was the most common regeneration technique, and commercial and precommercial thinning of stands, with the intention of creating evenly-aged stands was introduced. From the 1950s, large-scale clearcutting was the dominant felling system, which in most cases was followed by planting of tree-seedlings. Since the 1990s, silvicultural practices have been directed towards the promotion of more natural conditions so the size of clearcuts has decreased, and natural regeneration has again increased (Linder and Östlund, 1998). Due to the historic development of landownership and exploitation, cutblocks in Sweden are usually irregular in shape and unevenly distributed along forestry roads. Dead-end roads are commonly built to gain access to new cutblocks. The mean cutblock size in the Swedish study area was 40.7 ha (~60% larger than those in Alberta) with timber extraction in its second and third rotation. Study area population characteristics Based on >25 years of research in Sweden and >16 years of research in Alberta, brown bears between the two areas have comparable life-history characteristics (Table 2) (Zedrosser et al., 2012). Both are arctic-interior populations, with similar ecological conditions, life-history traits (Ferguson and McLoughlin, 2000) (Zedrosser et al., 2012), and diets (Dahle et al., 1998; Munro et al., 2006). Protein consumption is highest in spring and early summer for both populations, consisting mainly of ungulate neonates and insects (mainly ants (Formica spp., Camponotus spp.) in Sweden (Dahle et al., 1998) and Canada (Munro et al., 2006). Insect consumption is highest in both study areas in July prior to fruit ripening. Late summer and fall diets are dominated by graminoids, herbs, and berries. The primary source of berries in Canada is from russet buffaloberry/soopolallie (Shepherdia canadensis), mountain huckleberry (Vaccinium membranaceum) and to a lesser extent dry 224

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden hillside blueberry (V. myrtilloides) and lingonberry (V. vitis-idaea) (Munro et al., 2006). In Sweden berries are mainly from dwarf Ericaceae shrubs, especially blueberry (V. myrtillus), crowberry (Epetrum spp.), and lingonberry (Dahle et al., 1998). Neither study population has access to spawning salmon (Oncorhynchus spp., Salmo spp.). In both areas, brown bears face resource extraction, agriculture, urbanization, and recreation (Nielsen et al., 2004; Nellemann et al., 2007). In Sweden, however, brown bears have increased from ~130 animals in the 1930s to >3,200 animals in 2008 (Kindberg et al., 2009), and have been hunted since the 1940s (Swenson et al., 1994). In contrast, they are designated as threatened in Alberta, with about 691 bears in the province (Alberta Sustainable Resource Development and Alberta Conservation Association, 2010), within an area ~50% larger than that of Sweden (661,848 km2 vs. 449,964 km2). Regulated hunting occurred until 2002, became increasingly restricted and was stopped in 2006 (Alberta Sustainable Resource Development and Alberta Conservation Association, 2010). Selection of bears for analysis We considered three sex and offspring dependent classes of bears for this study: males, females, and females with dependent cubs of the year. We did not distinguish solitary females from females with yearlings or 2-year-olds since previous research has demonstrated that habitat selection in adult females differs only when having cubs of the year (Steyaert et al., 2012). Because cub loss is common in Sweden and occurs mostly during the months May and June (i.e. the mating season; Swenson et al., 1997; Zedrosser et al., 2009), females that lost their cubs were assigned to the â&#x20AC;&#x2DC;femaleâ&#x20AC;&#x2122; sex-offspring class for the month following cub loss. Animal location (habitat use) and home range definitions All bears were captured as a part of long-term research projects. GPS-location data were collected from bears captured from 2005-2010 in Sweden, and from 1999-2010 in Alberta using GPS radiocollars. Capture and handling are described in Cattet et al. (2003) for Alberta and Arnemo et al. (2011) for Sweden. In Sweden bears were fitted with GPS Plus, Vectronic Aerospace GmbH radiocollars with a relocation frequency scheduled at 30 minutes, although some bears were programmed with relocation frequencies at 1 or 10 minutes. In Alberta, bears from the Yellowhead and Grande Cache populations units were fitted with either a Televilt GPS-Simplex or an ATS (Advanced Telemetry Systems) GPS radiocollars with a relocation frequency of every 1 or 4 hours. Daily behaviour patterns of bears between the study areas were similar with diurnal patterns typically having the highest activity during dusk and dawn hours and a resting behaviour during midday and midnight hours (Munro et al., 2006; Moe et al., 2007). In this study we did not distinguish between diurnal behaviours in analyses of habitat selection, but rather evaluated general seasonal changes in habitat selection at a monthly temporal resolution during the active period (i.e., May through October). Across both studies, we gathered animal use information from GPS telemetry on 157 unique bears totaling 1,916,188 relocation events (Alberta: 152,514 relocations from 72 225

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden bears; Sweden: 1,763,674 relocations from 85 bears). Because we were interested in testing whether sex-offspring status affected habitat selection of clear-cuts, animal locations were labeled as either male, female with cubs of the year, or females without cubs of the year (i.e., solitary or with yearlings). Based on these sex-offspring classes, sample sizes for each study area consisted of 38 male and 34 female bears in Alberta with 8 female bears monitored while having cubs of the year. In Sweden sample sizes for sexoffspring classes consisted of 35 male and 50 female bears with 26 female bears monitored while having cubs of the year. In many instances, females with cubs of the year were also categorized as females in different years (or months if cubs were lost in their first year). Home ranges were defined for each bear using Minimum Convex Polygons (MCPs) using the ArcView 3.2 extension Animal Movement (Hooge and Eichenlaub, 1997). For some peripheral study bears we clipped MCP home ranges to jurisdictional boundaries, such as the border with British Columbia for the Alberta population. Home ranges were used to define availability of habitats for each animal using random sampling. A sample intensity of 5 locations per square kilometer with a minimum distance between random locations of 50 m was used for each animal based on MCP polygons. Characterizing habitat classes for 3rd order habitat selection We defined our study area into 6 broad habitat categories including 4 seral forest stages and 2 non-forested habitat types. These 6 categories were: (1) young regeneration (seedling to sapling stages defined as 0–9 years of age), (2) advanced regeneration (poles, defined as 10–25 years age), (3) young maturing forest (26–60 years of age), (4) mature forest (>60 years of age), (5) non-treed vegetated habitats (e.g., meadows, wetlands, shrubland), and (6) non-habitat (e.g., barren, snow, ice, or water). Young regenerating forests (seedling to sampling stage) were characterized by having an open canopy structure resulting in the dominance of a herbaceous and shrub groundcover. Advanced regenerating forests (regenerating pole stage) were transitional habitats where tree canopy cover begins to develop, as well as maturation of dwarf fruiting shrubs. Young maturing forests (26-60 years post-disturbance) were characterized by vertical rather than horizontal stand growth with dominance by trees being established thus resulting in reductions in food resource availability and habitat selection by bears (Nielsen et al., 2004a; 2004b). And finally, mature forests are considered fully-developed forests. To match animal location data among years, annual habitat maps were defined for each study area (1999 to 2010 in Alberta; 2005 to 2010 for Sweden) using base maps (1998 for Alberta and 2000 for Sweden), Alberta Vegetation Inventory (AVI) data, and remote sensing imagery (e.g., Landsat and SPOT5). Characterizing age, size, and shape of clear-cut patches for 4th order selection For each clear-cut, we collected information on year of harvest to calculate age of harvest at time of animal use, the size of harvest in hectares, and the shape of clear-cuts. Age was based on the year of harvest from GIS databases. In contrast, size and shape of clear-cuts were based on landscape definitions of ‘patches’ defined as forest stands having similar aged and distinct habitat boundaries (edges), not necessarily individual clear-cut 226

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden boundaries (Figure 2a). For instance, two young stands harvested in the same cutblock and year, but separated only by a road or two young maturing forest stands adjacent to one another that were harvested a few years apart were considered a single ‘patch’ since no real difference in forest structure existed between harvest polygons. To define boundaries (edges) of these patches, we considered stand boundaries with significantly different stages of growth (age) or forested and non-forested boundaries. Because the structure and habitat characteristics of a regenerating forest change more rapidly in early stages of succession than in later stages, the importance of age differences between stands was itself a function of stand age. Specifically, a boundary between stands was considered to form an ‘edge’ (E1) if the difference in ages between the two stands was equal to or greater than half the sum of the two ages or, E1 if . (eq. 1) Otherwise it was considered part of the same patch (E0). This emphasized differences in adjacent early seral stands harvested in different years (e.g., 2 versus 5 year old stands), but not adjacent mature stands harvested in different years having relatively similar stand structure (Figure 2a). Like that of the first analysis for 3rd order habitat selection, annual landscapes defining even-aged patches were produced for each study area and year. Once patches were defined, patch size (hectares) was calculated. Using defined patches, a shape index was calculated for each patch to indicate how similar in shape the harvest patch was to a circle. More specifically, we used the shape index from Hunter (1990), which is defined as, (eq. 2). ( ) √

Based on this index, a circle would have a value of 1, while increasingly complex polygons would be increasingly larger than one. Figure 2b illustrates shape indices for defined clearcut patches for a cutblock in west-central Alberta. Statistical methods 3rd order habitat selection (habitat types): We estimated general habitat selection for clear-cuts ≤60 years of age using habitat selection ratios (wi, Manly et al. 2002) for each bear and month combination. Selection ratios were summarized with a one-sample t-test against a null value of 1.0 (proportion use = proportion available) to identify general selection patterns by season and study area. Only those selection ratios for a bear-month combination having at least 50 monthly telemetry observations were considered. As well, only bears having a reasonable amount of clear-cut habitats available from which to select from were considered. Here we set that threshold at 5% of their home range. To more fully understand sex-offspring, study area, and seasonal effects, we also estimated monthly habitat selection coefficients across the active period (May through October) for 6 defined ‘habitat’ types, including 3 clear-cut age classes, using resource selection functions (RSFs) with logistic regression models used to estimate selection coefficients (Manly et al., 2002; Johnson et al., 2006). Given that our predictors of habitat selection were based on categorical variables (habitat types), we used binary (0 or 1) ‘dummy’ coding to identify 227

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden each habitat type with mature forests (>60 year old stands) used as the reference category (control). Selection of different clear-cut seral stages were therefore compared (and reported) against selection with mature forests. To account for potential differences among animals (sex, reproductive class, etc.), a twostage RSF analysis (Manly et al., 2002; Nielsen et al., 2002; 2009) was used where a RSF model was fit for each individual bear and month combination (minimum sample size of use locations equal to 50) and subsequently combined into sex-offspring groups by month classes (second-stage) using summary statistics (mean and standard errors) with sample weights (analytical weights) of habitat selection coefficients. We used the two-stage analytical approach to acknowledge the final unit of replication was the animal, not the telemetry location, while also allowing a random intercept and multiple random slopes to be fit for each factor for each bear (Nielsen et al., 2009). Although the two-stage model is seemingly similar to a mixed effect RSF model (e.g., Gillies et al., 2006), mixed effect models cannot easily estimate numerous random slopes and are further limited in their ability to evaluate secondary hypotheses related to animal-level differences in selection, such as differences between sex-offspring groups, study area effects, and functional responses in habitat selection due to changes in habitat availability. Sample weights for each observation (selection coefficient by bear-month) were estimated as inverse variance weights (Hedges, 1983) using the estimates of standard errors from the selection coefficient (Nielsen et al., 2009), as in meta-analyses (Ellis, 2010). This accounted for differences in precision of selection coefficients among bears. All analyses were performed in STATA 12.1 (StataCorp 2011). Final group-level estimates of selection for the three regenerating forest classes were graphed by month to illustrate general similarities and differences among seasons, sexoffspring classes, and study site combinations. The regenerating forest hypothesis (H1) and the functional response hypothesis (H3) were evaluated for 3rd order habitat selection. Methods for evaluating these hypotheses are discussed in more detail below (‘Evaluating support for major hypotheses’). 4th order habitat selection (within clear-cut patches): Fourth order (within clear-cut patch) habitat selection was evaluated for each bear and month combination using only the GPS telemetry (30,588 observations in Alberta and 860,371 observations in Sweden) and random locations occurring within clear-cuts ≤ 60 years of age. For each location, the age of the regenerating forest, the size in hectares (log [area + 1]) of the patch, and the shape of clear-cut patch were identified using GIS. As in 3rd order habitat selection methods, RSF models were estimated for each bear and month combination. However, due to high collinearity between clear-cut patch shape and patch area (size), patch shape was removed from all models. We fit a quadratic term for stand age since previous work has shown that bears typically select for intermediate-aged regenerating forests (see Nielsen et al., 2004a). And finally, we fit an interaction term between age and size (log[area + 1]) of clear-cut patch to test for ‘security’ or hiding cover effects. By evaluating these factors, we could test two competing hypotheses that have forest management and planning 228

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden applications; specifically, the ‘habitat security’ hypothesis versus a ‘food resource concentration’ (optimal foraging) hypothesis. In the former hypothesis, we expected that the size of the clear-cut patch would be less important as the clear-cut aged since forest cover would reduce visibility of animals proving security (cover). The alternate hypothesis, on the other hand, would suggest that food resource concentration would be the most important factor and thus if a certain age of clear-cut resulted in the greatest concentration of resources then it would be most ‘profitable’ to forage in larger patches. As in the above two-stage RSF methods for 3rd order analyses, summary statistics were estimated for each sex-offspring group and month combination using sample weights to estimate the ‘second-stage’ of the RSF models. Final coefficients were tabulated by group (study site and sex-offspring class) and month and tested for significance (from 0) based on a simple intercept model using a generalized linear model (Gaussian family and identity link) with sample weights. Specific hypotheses related to landscape structure (H2) and functional responses (H3) were evaluated for 4th order habitat selection. Methods for evaluating these hypotheses are discussed in more detail in the following section. Evaluating support for major hypotheses We tested for support of our competing hypotheses and for general functional responses, sex-offspring and study area (Alberta versus Sweden) effects in habitat selection coefficients for each month and scale of analysis (i.e., 3rd or 4th order habitat selection) using generalized linear models (Gaussian family and identity link) with sample weights (inverse variance) with Akaike Information Criteria (AICc) used to rank support of models (Burnham and Anderson, 2002). In total, 10 a priori models (hypotheses) were tested, including a null (intercept) model and different combinations of sex, sex-offspring, study site, and functional response variables (Table 3). Functional responses for 3rd order habitat selection were based on availability of seral forest classes (0-9 years, 10-25 years, and 26-60 years) and amount (%) of natural openings, while 4th order functional responses in habitat selection were based on average age of clear-cut within each animals home range since we assumed that responses to age should be dependent on the range of ages available, particularly for brown bears in Alberta where many home ranges of bears only encompassed recent or young clear-cuts.

Results 3rd order habitat selection (habitat types) Amount of clear-cut habitats (≤60 years old) within brown bear home ranges was noticeably different between study areas. Alberta’s bears ranged from having a minimum of 1.6% of their home range in clear-cuts to a maximum of 42.8% of their home range, while Sweden ranged from a minimum of 19.8% to a maximum of 61.5%. Habitat selection ratios were estimated for 851 unique bear and month combinations whose home ranges contained at least 5% clear-cut habitat and a minimum of 50 use locations per month per bear. Regardless of study area, sex-offspring class, season (month), or age of regenerating clear-cuts (≤60 years old), brown bears selected (i.e., selection ratios, w0-60) clear-cut 229

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden habitats (w0-60 = 1.29, SE = 0.03, t = 11.25, df =850, p < 0.001) relative to the remaining matrix of habitats thus supporting our regenerating forest hypothesis (H1). Evaluations of selection ratios by study area, however, revealed that overall selection ratios were noticeably higher for bears in Sweden (w0-60 = 1.36, SE = 0.020, t = 17.89, df = 528, p < 0.001) than in Alberta (w0-60 = 1.18, SE = 0.06, t = 3.08, df = 321, p = 0.001). Habitat selection of clear-cuts by habitat types, including the 3 regenerating forest habitats, demonstrated that selection varied by age of regenerating clear-cut, month, sexoffspring class, and study area (Figure 3A). As a general rule, recent clear-cuts (0–9 years) were selected more by brown bears in Alberta than in Sweden. Peak selection for recent clear-cuts in Alberta occurred in August with the lowest period of selection in May. In contrast, bears in Sweden generally used recent clear-cuts according to their availability (use=available), although increases in selection were notable during September and October for all sex-offspring classes and earlier in the year (July-August) for male bears. Model comparisons of our 10 a priori hypotheses demonstrated that differences in habitat selection among animals for recent (0–9 years) clear-cuts were explained most (AICc weights) between July and September by sex and study area effects (1b), by functional responses and sex-offspring classes in June and October (4a), and by a null (0 or constant model) during May (Figure 4A). Selection of young regenerating clear-cuts (10–25 years) was consistent for all sex-offspring classes and study sites, especially during June, July and August when bears were about two times more likely to use young regenerating clear-cuts than mature forests (Figure 3B). Selection for young regenerating clear-cuts diminished noticeably, however, for bears in Alberta during September and October, especially for females with cubs of the year, while bears in Sweden continued to select young regenerating clear-cuts throughout the late summer and fall period. Model comparisons of our 10 a priori hypotheses demonstrated that differences in habitat selection among animals for young regenerating (10–25 years) clear-cuts were explained best (AICc weights) by models containing study area effects. During the second half of the active season (August – October), a functional response in habitat availability and study area interaction (3b) explained habitat selection of young regenerating clear-cuts best, while a sex and study area interaction (1b) described habitat selection best during May and July, and finally a functional response-study area interaction and sex-offspring class and study area interaction (4c) was most supported during June (Figure 4B). Study area differences were generally due to lower levels of selection for young regenerating clear-cuts in Alberta relative to Sweden (especially late in the active season), while the sex-offspring study area interaction was most notable for the differences in selection of young regenerating clear-cuts during September-October by female bears with cubs of the year relative to female bears without cubs of the year (Figure 3B). Patterns of selection for old regenerating clear-cuts (26–60 years) by brown bears in Alberta and Sweden were the most dissimilar of the age-classes tested. In Sweden bears consistently selected older clear-cuts regardless of month, while selection in Alberta was 230

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden evident only during the first half of the active season (Mayâ&#x20AC;&#x201C;July) with a major decline in selection of older clear-cuts for bears in Alberta during August, September, and October (Figure 3C). These changes were reflected in the candidate models tested for individuallevel differences in habitat selection. During May, September, and October, differences in habitat selection were explained best by the global model (4c) defining functional responses by study area and sex-offspring study area interactions, while habitat selection of old clear-cuts in June and August were explained best by a functional response study area interaction (Figure 4C). And finally, in July habitat selection of old clear-cuts were explained best by sex-study area interactions. 4th order habitat selection (within clear-cut patches) Overall, brown bears selected clear-cut patches that were intermediate in age (10-30 years) and depending on sex-offspring class and study area either large or small in size (Table 4; Figure 5). Specifically, male bears in both study areas selected large clear-cuts over that of small clear-cuts on a per unit area basis with peak habitat selection for clearcut ages of 10 and 25 years in Alberta (Figure 5A) and 20 to 30 years in Sweden (Figure 5D). Female bears (without COY) in Sweden selected habitats similar to male bears in Sweden (i.e., large clear-cuts 20 to 30 years old; Figure 5E). In contrast, females without COY in Alberta selected for smaller clear-cut patches, but at similar ages (10-25 years) to that of male bears (Figure 5B). And finally, female bears with COY displayed segregation of habitats with the other sex-offspring classes. In Alberta, females with COY selected for very young clear-cuts (< 5 years old) regardless of size (Figure 5C), although sample sizes were very small and influenced strongly by selection for young clear-cuts in May (Table 4). Females with COY in Sweden on the other hand, selected for small clear-cut patches opposite to that of males and females without COY, although age (20 to 30 years) of clearcut selected was still similar to the other sex-offspring classes (Figure 5F). Differences in habitat selection among animals within clear-cuts of different ages were explained most (AICc weights averaged for Age + Age2) by functional responses (average clear-cut age) during May and September (i.e., model 3a), by a functional response by study area interaction in August and October (3b), and by study site differences with a functional response, sex-offspring interactions (4b) during June and July (Figure 6A). Differences in habitat selection among animals based on size of clear-cuts were supported most by a null model (0) during October, by a sex difference (1a) in May and September, by a functional response and study site interaction (3b) during August, and by study site differences with a functional response, sex-offspring interactions (4b) during June and July (Figure 6B). And finally, age and size of clear-cut interactions were supported most for sexoffspring study area interactions (model 2b) during May, by functional response and study area interactions (3b) during June, by study site differences with a functional response, sexoffspring interactions (4b) during June and July, and by a null model (0) during September and October (Figure 6C).

231

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden

Discussion Regenerating forest hypothesis In both Alberta and Sweden, brown bears selected for clear-cut patches over the surrounding matrix habitats. This supports our regenerating forest hypothesis (H1) where generalist omnivores, such as brown bears, respond positively to forest disturbances, particularly where forests dominate the landscape like that of the Holarctic boreal forests and where natural fire regimes are suppressed. Our results further support previous studies showing neutral (used according to availability) to positive associations in selection of clear-cuts including, Mace et al. (1996) in the Swan Mountains of Montana, Wielgus and Vernier (2003) in the Selkirk Mountains of British Columbia, Nielsen et al. (2004a) in the Alberta foothills, and Swenson et al. (1999) in central Sweden. In contrast to these supporting studies, Zager et al. (1983) and McLellan and Hovey (2001) showed negative associations with clear-cuts, although in both of these cases human traffic on roads associated with clear-cuts may have negatively affected habitat use of clear-cuts as well as the presence of alternate, productive regenerating forest habitats that were associated with major historic fires. Research on food resource availability and abundance in clearcuts supports a pulse in food resources exceeding that of mature forests prior to canopy closure (Martin, 1980; Swenson et al., 1999; Nielsen et al., 2004b), although silvicultural site preparation techniques that damage the roots of long-lived fruiting shrubs negatively affects late season food resources (Martin, 1980; Nielsen et al., 2004b). The presence of alternate resource patches, such as large historic fires, suggests that a functional response in habitat selection (Mysterud and Ims, 1998) for clear-cuts (i.e., the functional response hypothesis, H3) is an important consideration in understanding why some populations appear to favor clear-cut forests more than others. Landscape characteristics hypothesis Although numerous studies have examined general 3rd order habitat selection for clearcuts, few have examined the details of the characteristics (e.g., age, size, site history) of individual clear-cuts being selected (see however, Nielsen et al., 2004a). When examining habitat selection among general structural classes or ages of clear-cuts, recent clear-cuts (0-9 years) were selected more by brown bears in Alberta than in Sweden with peak selection of recent clear-cuts in Alberta occurring in August when grizzly bears typically forage on fruit (Munro et al., 2006). In Sweden, use of recent clear-cuts was largely based on availability (neutral) except late in the year (September-October) when selection for recent clear-cuts increases. Lower use of clear-cut patches in Sweden may reflect the openness of clear-cuts where hiding cover (security) may be more important. Young regenerating clear-cuts (10-25 years) were strongly selected by both brown bear populations, although selection was attenuated late in the year for brown bears in Alberta, particularly for females with cubs of the year. This age class corresponds to the period of maximum fruit and ant resource abundance just prior to canopy closure (Swenson et al., 1999; Nielsen et al., 2004b). Older (26-60 years) clear-cuts were neutral to moderately

232

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden selected, although an even stronger decline in selection was notable for the late season period in Alberta. The age of the clear-cut selected may, however, depend on the size of the clear-cut patch which we refer to as our landscape structure hypothesis (H2). Specifically, trade-offs may occur between hiding cover (security) needs in recent clear-cuts, especially for vulnerable sex-offspring classes, and for foraging efficiencies gained by larger resource patches (i.e., larger clear-cut patches). Our results support this hypothesis. When brown bears selected more for younger clear-cut patches, those patches tended to be smaller in size, while selection for older clear-cut patches was associated with larger patch sizes. Interactions with sex-offspring classes were noticeable, however, with male brown bears in both study areas consistently selecting for the larger, intermediate-aged clear-cut patches, while female bears in Alberta and females with cubs of the year in Sweden both selected for smaller clear-cut sizes. In the case of females with cubs of the year in Alberta, patch size and age were both inversely related to habitat selection. These patterns could be due to sexually-segregated habitat selection associated with intra-specific competition (Nielsen, 2005), sexually-selected infanticide mechanisms promoting sexually-segregated habitat selection (Swenson et al., 1997), or both, although sample sizes of females with cubs of the year in Alberta were small. The fact, however, that females with cubs of the year in Sweden still selected small clear-cuts over large clear-cut even after a period of years when hiding/security cover would be present suggests that these animals were segregating themselves from other bears (and not humans) by using smaller resource patches. In contrast to patterns of selection for size of clear-cut patch, selection for age of clear-cut patch appeared to be more consistent across sex-offspring classes and study areas, although seasonal differences were evident. This suggests that bears were selecting clearcuts at maximum or peak availability in food resources, which in this study averaged between 10 to 25 years post-harvest in Alberta and 20 to 40 years post-harvest in Sweden. This is consistent with prior research on food resource availability and habitat selection of brown bears (Martin, 1980; Nielsen et al., 2004a; 2004b). This is also the time period when regenerating trees begin to dominate the site reducing light availability. Seasonal differences, however, were present with the strongest effects evident during July when both populations were actively using clear-cut patches, especially for myrmecophagy activities (Swenson et al., 1999; Munro et al., 2006). Functional response hypothesis Our functional response in habitat selection (Mysterud and Ims, 1998) hypothesis was supported for most variables and seasons (months) by second-stage analyses of habitat selection among individual bears and comparisons with other a priori candidate models explaining differences in habitat selection. This suggests that the landscape context to which an animal is exposed is critical to understanding observed selection patterns. Further work is needed to explore these relationships in full detail and to make specific predictions for management purposes. Regardless, it is apparent from prior research that when alternate natural disturbances, such as regenerating forests from wildfire, are present, 233

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden brown bears will generally select for those habitats over that of similar-aged clear-cuts (Zager et al., 1983; McLellan and Hovey, 2001). The benefits of forest harvesting are therefore greatest in fire-suppressed landscapes dominated by mature forests.

Conclusions Although increases in human activity associated with expanding forest management in Alberta is a major conservation concern (Nielsen et al., 2006; 2008), it is evident that forest harvests provide an important mechanism for regenerating young seral forests in Holarctic boreal landscapes where natural upland openings are rare and major natural disturbances, such as fire, are suppressed (Johnson et al., 2001). Intermediate-aged (10 to 40 years postharvest) clear-cuts were selected for most by brown bears with a positive relationship between selection and patch size evident for the dominant sex-offspring classes. This suggests that patch size of the resource is more important than security, although regenerating trees likely provide sufficient cover by that age of stand. When younger regenerating stands were selected, such as by females with cubs of the year in Alberta, size of clear-cut was inversely related to habitat selection suggesting that security (hiding cover) was more important at that age of stand when tree cover is low and for that sexoffspring class. Although past trends in Alberta have been towards smaller clear-cut sizes (Stewart et al., 2012), natural disturbance-based harvest designs (Hunter 1990) with larger clear-cuts are now being implemented. Our results suggest that this would positively affect brown bear habitat selection for dominant bears, but potentially negatively affect female bears with cubs of the year as they tended to select for smaller clear-cuts. Future work should address whether in-block retention of trees as patches or diffuse standing trees would mitigate the negative responses to clear-cut size observed by females with cubs of the year. Future work should also consider the effects of silvicultural treatments on food resource abundance and habitat use by brown bears, particularly for long-lived fruiting shrubs. Prior work by Nielsen et al. (2004a; 2004b) suggests that certain mechanical site preparation techniques can negatively affect food resource abundance and habitat selection by brown bears, although use of prescribed fire appears to increase abundance of long-lived fruiting shrubs (Martin, 1980; Nielsen, personal obs.) More studies are needed to evaluate whether these responses are consistent among populations and what practices beside prescribed fire can mitigate these effects. Finally, silvicultural thinning which is commonly used in Scandinavia, but rare in Alberta, should be experimented with in Alberta for wildlife enhancement purposes shortly after stand closure.

Acknowledgements Data collection in Alberta was supported financially by the program partners of the Foothills Research Institute Grizzly Bear Program. Data collection in Sweden was supported financially by the Swedish Environmental Protection Agency, the Norwegian Directorate for Nature Management, and the Research Council of Norway. Funding support for this international research collaboration was supported financially by Alberta InnovatesBiosolutions. Funding for support of SEN was provided by NSERC. We thank the many field staff who assisted in the capture of bears in both studies. 234

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden

Literature cited Andison DW, 1998. Temporal patterns of age-class distributions on foothills landscapes in Alberta. Ecography 21:543-550. Arnemo JM, Evans A, Fahlman Å, 2011. Biomedical protocols for free-ranging brown bears, wolves, wolverines and lynx. Trondheim, Norway: Norwegian Directorate for Nature Management; 14. Association ASRDaAC, 2010. Status of the Grizzly Bear (Ursus arctos) in Alberta: Update 2010. Edmonton, Canada: Alberta Sustainable Resource Development; 44. Beckingham JD, Corns IGW, Archibald JH, 1996. Field guide to ecosites of west-central Alberta. . (Natural Resources of Canada CFS, Northwest Region, ed). Edmonton, Alberta, Canada: Northern Forest Centre. Burnham KP, Anderson D, 2002. Model selection and multi-model inference: A practical information-theoretic approach. 2nd edition. Springer, New York, 488 pg. Cattet MRL, Caulkett NA, Stenhouse GB, 2003. Anesthesia of grizzly bears using xylazinezolazepam-tiletamine or zolazepam-tiletamine. Ursus 14:88-93. Dahle B, Sorensen OJ, Wedul EH, Swenson JE, Sandegren F, 1998. The diet of brown bears Ursus arctos in central Scandinavia: Effect of access to free-ranging domestic sheep Ovis aries. Wildlife Biology 4:147-158. Ellis, PD, 2010. The Essential Guide to Effect Sizes: An Introduction to Statistical Power, Meta-Analysis and the Interpretation of Research Results. Cambridge University Press, Cambridge, United Kingdom, 173 pg. Ferguson SH, McLoughlin PD, 2000. Effect of energy availability, seasonality, and geographic range on brown bear life history. Ecography 23:193-200. Gillies CS, Hebblewhite M, Nielsen SE, Krawchuk MA, Aldridge CL, Frair JL, Saher DJ, Stevens CE, Jerde CL, 2006. Application of random effects to the study of resource selection by animals. Journal of Animal Ecology 75, 887-898. Hedges LV, 1983. A random effects model for effect sizes. Psychological Bulletin 93: 388395. Hooge PN, Eichenlaub B, 1997. Animal movement extension to arcview. ver. 1.1. Alaska Biological Science Center, U.S. Geological Survey, Anchorage, AK, USA. Hunter, M.L. Jr. 1990, Wildlife, forests, and forestry: principles of managing forests for biological diversity. Prentice Hall, Inc., Englewood Cliffs, N.J. Johnson CJ, Nielsen SE, McDonald TL, Merrill E, Boyce MS, 2006. Resource selection functions based on use-availability data: theoretical motivation and evaluation methods. Journal of Wildlife Management 70: 347-357. Johnson EA, Miyanishi K, Bridge SRJ, 2001. Wildfire regime in the boreal forest and the idea of suppression and fuel buildup. Conservation Biology, 15(6):1554-1557. Kindberg J, Swenson JE, Ericsson G, 2009. Björnstammens storlek i Sverige 2008 - länsvisa uppskattningar och trender. Umeå, Sweden: Scandinavian Brown Bear Research Project. Linder P, Östlund L, 1998. Structural changes in three mid-boreal Swedish forest landscapes, 1885-1996. Biological Conservation 85:9-19. 235

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden Mace RD, Waller JS, Manley TL, Lyon LJ, Zuuring H, 1996. Relationships among grizzly bears, roads and habitat in the Swan Mountains, Montana. Journal of Applied Ecology 33:1395–1404. Mascarùa Lòpez LE, Hareper KA, Drapeau P, 2006. Edge influence on forest structure in large forest remnants, cutblock spearators, and riparian buffers in managed black spruce forests. Ecoscience 13:226-233. Manly BFJ, McDonald LL, Thomas DL, McDonald TL, Erickson WP, 2002. Resource Selection by Animals: Statistical Design and Analysis for Field Studies, 2nd ed. Kluwer Academic Publishers, Dordrecht, the Netherlands. Martin P, 1980. Factors influencing globe huckleberry fruit production in Northwestern Montana. International Conference on Bear Research and Management, 159-165. McDermid GJ, 2005. Remote Sensing for Large-Area, Multi-Jurisdictional Habitat Mapping. PhD Thesis. Department of Geography, University of Waterloo, Waterloo, Ontario, Canada. McLellan BN, 1989. Dynamics of a grizzly bear population during a period of industrial resource extraction. I. Density and age–sex composition. Canadian Journal of Zoology 67(8):1856-1860. McLellan BM, Hovey FW, 2001. Habitats selected by grizzly bears in a multiple use landscape. Journal of Wildlife Management 65:92–99. Moe TF, Kindberg J, Jansson I, Swenson JE, 2007. Importance of diel behaviour when studying habitat selection: examples from female Scandinavian brown bears (Ursus arctos). Canadian Journal of Zoology, 85: 518-525. Munro RHM, Nielsen SE, Price MH, Stenhouse GB, Boyce MS, 2006. Seasonal and diel patterns of grizzly bear diet and activity in west-central Alberta. Journal of Mammalogy 87:1112-1121. Mysterud A, Ims RA, 1998. Functional responses in habitat use: Availability influences relative use in trade-off situations. Ecology 79:1435-1441. Nellemann C, Stoen OG, Kindberg J, Swenson JE, Vistnes I, Ericsson G, Katajisto J, Kaltenborn BP, Martin J, Ordiz A, 2007. Terrain use by an expanding brown bear population in relation to age, recreational resorts and human settlements. Biological Conservation 138:157-165. Nielsen SE, 2005. Habitat ecology, conservation, and projected population viability of grizzly bears (Ursus arctos L.) in west-central Alberta, Canada. PhD Dissertation, University of Alberta, Edmonton, Alberta, Canada. Nielsen SE, Boyce MS, Stenhouse GB, Munro RHM, 2002. Modeling grizzly bear habitats in the Yellowhead Ecosystem of Alberta: Taking autocorrelation seriously. Ursus 13: 45-56. Nielsen SE, Boyce MS, Stenhouse GB, 2004a. Grizzly bears and forestry: I. Selection of clearcuts by grizzly bears in west-central Alberta, Canada. Forest Ecology and Management 199:51-65. Nielsen SE, Munro RHM., Bainbridge E, Boyce MS, Stenhouse GB, 2004b. Grizzly bears and forestry II: distribution of grizzly bear foods in clearcuts of west-central Alberta, Canada. Forest Ecology and Management 199: 67-82. 236

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden Nielsen SE, Herrero S, Boyce MS, Benn B, Mace RD, Gibeau ML, Jevons S, 2004c. Modelling the spatial distribution of human-caused grizzly bear mortalities in the Central Rockies Ecosystem of Canada. Biological Conservation 120:101-113. Nielsen SE, Stenhouse GB, Boyce MS, 2006. A habitat-based framework for grizzly bear conservation in Alberta. Biological Conservation 130:217-229. Nielsen SE, Boyce MS, Beyer H, Huettmann F, Stenhouse GS, 2008. Can natural disturbance-based forestry rescue a declining population of grizzly bears? Biological Conservation 141: 2193-2207. Roever CL, Boyce MS, Stenhouse GB, 2008a. Grizzlybears and forestry: II: Grizzly bear habitat selection and conflicts with road placement. Forest Ecology and Management 256:1262-1269. Roever CL, Boyce MS, Stenhouse GB, 2008b. Grizzly bears and forestry I: Road vegetation and placement as an attractant to grizzly bears. Forest Ecology and Management 256: 1253-1261. StataCorp, 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP. Stewart BP, Nelson TA, Wulder MA, Nielsen SE, Stenhouse G, 2012. Impact of disturbance characteristics and age on grizzly bear habitat selection. Applied Geography 34:614625. Steyaert, SMJG., Kindberg J, Swenson JE, Zedrosser A, 2012. Making the best of a bad situation: resource selection by a nonsocial carnivore in a risky landscape. Journal of Animal Ecology, submitted. Swenson JE, Sandegren F, Bjarvall A, Soderberg A, Wabakken P, Franzen R, 1994. Size, trend, distribution and conservation of the brown bear Ursus arctos population in Sweden. Biological Conservation 70:9-17. Swenson JE, Sandegren F, Soderberg A, Bjarvall A, Franzen R, Wabakken P, 1997. Infanticide caused by hunting of male bears. Nature 386:450-451. Swenson JE, Jansson A, Riig R, Sandegren F, 1999. Bears and ants: myrmecophagy by brown bears in central Scandinavia. Canadian Journal of Zoology 77: 551-561. White JC, Wulder MA, Gómez C, Stenhouse G, 2011. A history of habitat dynamics: Characterizing 35 years of stand replacing disturbance. Canadian Journal of Remote Sensing 37: 234-251. Wielgus, RB, Vernier, PR, 2003. Grizzly bear selection of managed and unmanaged forests in the Selkirk Mountains. Canadian Journal of Forest Research 33:822–829. Zager P, Jonkel C, Habeck J, 1983. Logging and wildfire influence on grizzly bear habitat in Northwestern Montana. Proceedings of the Sixth International Conference on Bear Research and Management, 124–132. Zedrosser A, Dahle B, Stoen OG, Swenson JE, 2009. The effects of primiparity on reproductive performance in the brown bear. Oecologia 160:847-854. Zedrosser A, Dahle B, Swenson JE, 2006. Population density and food conditions determine adult female body size in brown bears. Journal of Mammalogy 87:510-518.

237

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden Table 1. Area (km2) and percent composition of land cover classes for the Alberta, Canada and Sweden study areas.

Land cover class

Alberta Area (km2) Percent

Sweden(2005) Area (km2) Percent

Clear-cut (0-60 yrs.) Recenta

1,398

3.7

701

4.8

Young regeneratingb

1,970

5.2

1,176

8.0

Old regeneratingc

1,242

3.3

3,591

24.6

4,610

12.2

5,468

37.4

Mature forestd

25,877

68.1

6,651

45.5

Vegetated non-forest land

3,943

10.4

1,853

12.7

Barren land

3,591

9.4

642

4.4

Sub-total

0 to 9 years post-harvest; 10 to 25 years post-harvest; 26 to 60 years post-harvest; >60 years since post-harvest

238

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden Table 2. Brown bear life-history parameters in Canada (Alberta) and Sweden (see also Zedrosser et al., 2012). The number of individuals a parameter estimate is based on is given in parentheses.

Life history parameter

Alberta

Sweden

Age (years) at primiparity

5.8 (13)

5.0 (59)

Inter-litter interval (years between successful litters)

2.5 (10)

2.3 (124)

Mean date of den entry: Adult male Adult female without dependent offspring

November 22 (15) November 9 (41)

October 27 (33) October 25 (43)

Mean date of den exit: Adult male Adult female without dependent offspring Adult female with cubs of the year

April 4 (13) April 11 (24) April 17 (11)

April 4 (33) April 13 (13) May 7 (21)

~May 15 – July 31

~May 15 – July 7

899 (22)

833–1,055 (36)

273 (39)

217–280 (52)

~14

~29

Timing of breeding season Median home range size (km2): Adult male Adult female (without dependent offspring) Mean population density (per 1000 km2)

239

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden Table 3. Description of secondary hypotheses tested for 3rd and 4th order habitat selection of clear-cut patches in Alberta and Sweden by brown bears.

Model ID

Model structure

Habitat selection hypothesis

Null (intercept)

Constant

Sex

Sexually-segregated

Sex × Site

Sexually-segregated and site interaction

Sex-offspring

Sex-offspring segregated

Sex-offspring × Site

Sex-offspring segregated and site interaction

Functional responses

Varies as a function of habitat availability

Functional responses (FR) × Site

Varies as a function of habitat availability and site interaction

FR + Sex-offspring

Varies as a function of habitat availability and sexoffspring status

FR + Sex-offspring + Site

Varies as a function of habitat availability, sexoffspring status and site

FR × Site + Sex-offspring × Site

Varies as a function of habitat availability and site interaction, as well as sex-offspring status and site interaction

240

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden Table 4. Average (sample size of bears in parentheses) seasonal (May – October) 4th order habitat selection of clearcut patch characteristics (e.g., age and size of patch) used by brown bears in Alberta and Sweden. Average coefficients reported by sex-offspring class and study site (Ab = Alberta; Sw = Sweden) with significant (p<0.05) differences from zero indicated by a superscript § symbol. Variable and group May June July August Sept. Oct. May-Oct β Age Ab-Male

0.165§ (8)

0.060 (10)

0.058 (13)

-0.167 (8)

0.252 (7)

-0.087 (5)

0.047

Sw-Male

0.036§ (34)

0.081§ (34)

0.081§ (31)

0.055§ (27)

0.058§ (21)

0.061§ (17)

0.062§

Ab-Female

0.097 (12)

0.021 (11)

0.152§ (15)

-0.003 (14)

0.132 (5)

0.080 (6)

0.080§

Sw-Female

0.038§ (46)

0.073§ (48)

0.108§ (46)

0.061§ (45)

0.048§ (40)

0.059§ (37)

0.065§

Ab-FemaleCOY

-0.256 (2)

-0.129 (3)

-0.020 (3)

-0.183 (3)

-0.037 (2)

-0.008 (2)

-0.106

Sw-FemaleCOY

0.008 (19)

0.056§ (19)

0.067 (15)

0.071§ (8)

0.035 (8)

0.058 (6)

0.049§

Ab-Male

-0.035§ (8)

-0.006 (10)

-0.044§ (13)

-0.028 (8)

-0.029 (7)

0.005 (5)

-0.023§

Sw-Male

-0.004§ (34)

-0.010§ (34)

-0.011§ (31)

-0.008§ (27)

-0.010§ (21)

-0.011§ (17)

-0.009§

Ab-Female

-0.012 (12)

-0.001 (11)

-0.040§ (15)

-0.036§ (14)

-0.054§ (5)

0.000 (6)

-0.024§

Sw-Female

-0.005§ (46)

-0.010§ (48)

-0.015§ (46)

-0.008§ (45)

-0.007§ (40)

-0.009§ (37)

-0.009§

Ab-FemaleCOY

-0.013 (2)

-0.015 (3)

-0.040 (3)

0.025 (3)

-0.011 (2)

-0.017 (2)

-0.012

Sw-FemaleCOY

-0.002 (19)

-0.006 (19)

-0.008§ (15)

-0.011§ (8)

-0.006 (8)

-0.009 (6)

-0.007§

Ab-Male

-0.368§ (8)

0.124 (10)

-0.099 (13)

-0.077 (8)

0.613 (7)

0.058 (5)

0.042

Sw-Male

0.078§ (34)

0.195§ (34)

0.254§ (31)

0.173§ (27)

0.186§ (21)

0.101 (17)

0.165§

Ab-Female

0.177 (12)

0.033 (11)

-0.130 (15)

-0.404§ (14)

-0.375§ (5)

-0.812§ (6)

-0.252

Sw-Female

0.001 (46)

0.174§ (48)

0.157§ (46)

0.168§ (45)

0.093§ (40)

0.083 (37)

0.112§

Ab-FemaleCOY

-2.060§ (2)

-0.263 (3)

-0.838§ (3)

-0.022 (3)

-0.025 (2)

0.332 (2)

-0.479

Sw-FemaleCOY

-0.079 (19)

0.066 (19)

0.070 (15)

0.102 (8)

0.110 (8)

-0.069 (6)

0.033

0.004 (8)

-0.001 (10)

0.018§ (13)

0.036§ (8)

-0.017 (7)

0.018 (5)

0.010

β Age2

β Size

β Age x Size Ab-Male

241

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden

Sw-Male

-0.002 (34)

-0.004§ (34)

-0.004§ (31)

-0.005§ (27)

-0.003§ (21)

-0.003 (17)

-0.003§

Ab-Female Sw-Female

-0.009 (12) -0.001 (46)

0.001 (11) -0.003§ (48)

0.003 (15) -0.002§ (46)

0.024§ (14) -0.003§ (45)

0.011 (5) -0.002§ (40)

-0.004 (6) -0.003§ (37)

0.004 -0.002§

Ab-FemaleCOY

0.081 (2)

0.032 (3)

0.044§ (3)

0.016 (3)

-0.002 (2)

-0.020 (2)

0.025

Sw-FemaleCOY

0.001 (19)

-0.004§ (19)

-0.002 (15)

0.000 (8)

-0.001 (8)

-0.006 (6)

-0.002

Age is reported as coefficients to the power of 1x10

242

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden

Figure 1. Locations of two brown bear study areas in west-central Alberta, Canada and central Sweden.

243

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden a.

Figure 2. Example of patch edge delineation (a.) and shape index (b.) for clear-cuts in west-central Alberta, Canada.

244

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden

Figure 3. Patterns of habitat selection by sex-offspring class and study area for 3 different post-harvest age classes (a. 0-9 years; b. 10-25 years; c. 26-60 years). Selection coefficients (ď&#x201A;ąSE) are based on averaged (inverse variance weights) animal-specific RSFs.

245

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden 1.0

0.8

1b 1b

0.6

w0-9 yrs

0.4

4a 0

0.2 0.0

May 1.0

0.8

June

July

August

September

0 1a 1b 2a 2b 3a 3b 4a 4b 4c

1b 3b

0.6

w10-25 yrs

October

0.4

0.2 0.0

May 1.0

w26-60 yrs

0.8

June

July

August

September 3b

4c 3b

October 4c

0.6 0.4 0.2 0.0 May

June

July

August

September

0 1a 1b 2a 2b 3a 3b 4a 4b 4c

October

Figure 4. Akaike weights (wi) from AICc scores ranking support among 10 a priori models describing 3rd order habitat selection for clear-cuts by brown bears in Alberta and Sweden based on general age classes (A. 0–9 years; B. 10–25 years; C. 26–60 years). See Table 3 for description of hypotheses.

246

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden 1000

A. Alberta - Male

D. Sweden - Male

1000

100

Size

100

10 0.1 0.2 0.3 0.4 0.5 0.6

0.55 0.60 0.65 0.70 0.75

1 0 1000

B. Alberta - Female

Age 1000

Age E. Sweden - Female

100

Size

100

10 0.1 0.2 0.3 0.4 0.5 0.6

0.55 0.60 0.65 0.70 0.75

1 0 1000

Age C. Alberta - Female with COY

1000

100

Age F. Sweden - Female with COY

Size

100

10 0.0 0.1 0.2 0.3 0.4

0.55 0.60 0.65 0.70

1 0

Age

Figure 5. Multi-seasonal (average) predicted habitat selection responses to age and size of clear-cuts by study area (Alberta vs. Sweden) and sex-offspring class.

247

Chapter 10-Comparing Regenerating Clear-Cut Forests in Alberta and Sweden 1.0

wAge

0.8

A. = 3a

0.6 3a

0.4

0.2 0.0 May 1.0

June

July

August

September

October

0.8

wSize

0.6 4b

0.4 1a

0.2 0.0 May 1.0

July

August

September

4b 3b

0.6

0 0

0.4 2b

0.2 0.0 May

June

July

August

September

0 1a 1b 2a 2b 3a 3b 4a 4b 4c

October

0.8

wAge x Size

June

0 1a 1b 2a 2b 3a 3b 4a 4b 4c

October

Figure 6. Akaike weights (wi) from AICc scores ranking support among 10 a priori models describing 4th order habitat selection within clear-cuts by brown bears in Alberta and Sweden based on age (A.), size (B.), and age x size interactions (C.). See Table 3 for descriptions of hypotheses.

248

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress

CHAPTER 11: ADVANCEMENTS IN THE DEVELOPMENT OF HAIR CORTISOL CONCENTRATION AS AN INDICATOR OF LONG-TERM STRESS IN BROWN BEARS Marc Cattet1, Gordon B. Stenhouse2, Andreas Zedrosser3, Jon E. Swenson4, Bryan J. Macbeth5, David M. Janz6, Mathieu Dumond7, Michael Proctor8, Harry Reynolds9, Luvsamjamba Amgalan10, Odbayar Tumendemberel11, and Derek Craighead12 1

Canadian Cooperative Wildlife Health Centre, Western College of Veterinary Medicine, University of Saskatchewan, 52 Campus Drive, Saskatoon, Saskatchewan S7N 5B4, Canada. 2 Foothills Research Institute, P. O. Box 6330, Hinton, Alberta T7V 1X6, Canada and Alberta Sustainable Resource Development, 1176 Switzer Drive, Hinton, Alberta T7V 1V3, Canada. 3 Department of Integrative Biology and Biodiversity Research, Institute for Wildlife Biology and Game Management, University of Natural Resources and Applied Life Sciences, Vienna, Gregor Mendel Str. 33, A-1180 Vienna, Austria. 4 Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, Post Box 5003, NO-1432 Ă&#x2026;s, Norway and Norwegian Institute for Nature Management, NO-7485 Trondheim, Norway. 5 Department of Veterinary Biomedical Sciences, Western College of Veterinary Medicine, University of Saskatchewan, 52 Campus Drive, Saskatoon, Saskatchewan S7N 5B4, Canada. 6 Department of Veterinary Biomedical Sciences, Western College of Veterinary Medicine, University of Saskatchewan, 52 Campus Drive, Saskatoon, Saskatchewan S7N 5B4, Canada and Toxicology Centre, University of Saskatchewan, 44 Campus Drive, Saskatoon, Saskatchewan S7N 5B3, Canada. 7 Nunavut Wildlife Division, Department of Environment, Government of Nunavut, Box 377, Kugluktuk, Nunavut X0B 0E0, Canada. 8 Birchdale Ecological, P. O. Box 606, Kaslo, British Columbia V0G 1M0, Canada. 9 Reynolds Alaska Wildlife Institute, PO Box 80843, Fairbanks, Alaska 99708, USA. 10 Mongolian Academy of Sciences, Jukov Avenue, Ulaanbaatar 51, Mongolia. 11 Laboratory of Genetics, Institute of Biology, Mongolian Academy of Sciences, Jukov Avenue-77 Ulaanbaatar-210351 Mongolia. 12 Craighead Beringia South, PO Box 147, Kelly, Wyoming 83011, USA.

249

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress

Executive Summary Measurement of the concentration of cortisol in hair has recently emerged as a promising indicator of long-term stress across a suite of mammals that includes humans, domestic and laboratory animals, and captive and free-ranging wildlife. Despite a spate of publications on the application of HCC, few report on its validation which includes the identification of variables that could compromise its soundness as an indicator of long-term stress. In this study, we extend previous developmental research to investigate what factors influence HCC at the level of individual brown bears3 (Ursus arctos), with the aim of eventually employing HCC at the population level as a wildlife conservation tool. We compared HCC values from 785 brown bear hair samples collected by various methods (hair snag, live-capture, hunter-kill) over 16 years and representing four geographic areas (Alberta, Sweden, Nunavut, and Mongolia) across the species’ Holarctic distribution, and found a four-fold difference between the highest (Alberta live-capture: 2.90 pg/mg) and lowest median HCC (Mongolia hair snag: 0.77 pg/mg). HCC was influenced by sex and age, as well as physical attributes (body mass and body length), but not in a consistent manner across populations. Through a detailed analysis of data from live-captured bears in Alberta, we also identified body condition, season of hair collection, and several capture-related factors that influence HCC, including the presence of radio-telemetry devices (radio-collars and/or radio ear-tags. In general, we found no evidence to discount HCC as a reliable indicator of long-term stress in individual bears, and suggest the next step is to widen the application of HCC to ascertain its effectiveness at the population level. We provide two examples of how this might be done using HCC values from five bear management units in Alberta. The first applied HCC as a direct measure of long-term stress and underscores the need to confirm associations between HCC and fitness parameters to firmly establish the value of HCC for monitoring of long-term stress in populations. The second applies HCC as an indirect marker of body condition status to show there may be some value in using HCC to monitor body condition trends, especially in food-limited environments.

Introduction Long-term stress is increasingly recognized as a primary determinant of the health of individual animals. This trend is evident across several fields of knowledge, including human medicine (Juster et al. 2010, Kudielka and Wüst 2010), animal science (Korte et al. 2009, Riondato et al. 2008), laboratory animal medicine (Konkle et al. 2010, Laber et al. 2008), veterinary medicine (Accorsi et al. 2008, Smith et al. 2010), and wildlife conservation (Busch and Hayward 2009, Johnstone et al. 2012). Whereas the physiological stress response when elicited over seconds or minutes is generally beneficial to an individual, this same response when sustained or repeated frequently over days or weeks becomes detrimental to health and fitness (Romero 2004, Wingfield et al. 3

Brown bear and grizzly bear refer to the same species, Ursus arctos, with “brown bear” being the more general term applied to the species across its Holarctic distribution. To avoid confusion within this report, we use brown bear in reference to all populations.

250

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress 1998). A prolonged state of elevated stress, widely accepted as a measurable increase in hypothalamus-pituitary-adrenal (HPA-) axis activation, can suppress immune function, growth, tissue maintenance, and reproduction; promote protein loss, deposition of fat, and hypertension; disrupt second messenger systems and signal transduction cascades; and initiate neuronal cell death (reviews in Busch and Hayward 2009, McEwen 2008). Populations rather than individuals provide the context for interest in long-term stress in the field of wildlife conservation (Busch and Hayward 2009, Ellis et al. 2012). For example, how does stress influence population regulation and spatial patterns of animal distribution, particularly in the face of anthropogenic environmental degradation and habitat fragmentation? Can monitoring population stress provide early warning capability and an opportunity to mitigate stressors before negative consequences begin to manifest? Despite these significant questions, the importance of understanding how stress operates at the individual level cannot be understated. The impact of stressors is manifested on individuals at a physiological level before negative consequences are expressed and observed at a population level (Ellis et al. 2012). If long-term stress is localized and affecting few animals or wide-spread and affecting many, if it is sex- or age-biased, or if it affects reproduction and/or survival, will all influence if and how it manifests at the population level. Furthermore, the likelihood of detecting long-term stress is much greater at the level of individuals than populations. So, increased attention toward sampling or measuring of stress in individual animals has potential to provide early warning of population-level effects and an opportunity to take preventative action. Although quantifiable, physiological measurements to estimate HPA-axis activation in vertebrates are numerous, many are sensitive to short-term (acute) stress or both short- and long-term (chronic) stress (Johnstone et al. 2012). Few if any are sensitive only to long-term stress, but a measure that has emerged as a promising long-term stress index is the glucocorticoid concentration of keratinous tissue, such as hair (fur) and feathers (Bortolotti et al. 2008, Gow et al. 2010, Sheriff et al. 2011). In contrast to other biological media (blood, saliva, urine, feces), keratinous tissues are relatively stable media known to incorporate blood-borne hormones and xenobiotics during periods of growth (Bortolotti et al. 2010, Pragst and Balikova 2006). The quantification of cortisol in hair (or corticosterone in feathers) represents an integration of HPA activity over a period of weeks to months (Bortolotti et al. 2008, Davenport et al. 2006, Russell et al. 2012), thus making it a particularly appropriate measurement for evaluating long-term stress in wildlife populations (Koren et al. 2002, Macbeth et al. 2010, Sheriff et al. 2011). Further, hair can be collected non-intrusively without capture or restraint (Macbeth et al. 2010), it can be transported and stored at ambient temperature, and under optimal conditions, cortisol and other molecules incorporated into hair may remain detectable for years to centuries (Kintz 2004, Webb 2010). We recently adapted a technique for the measurement of cortisol in rhesus macaque (Macaca mulatta) hair (Davenport et al. 2006) and refined it for application to brown bear (Ursus arctos) hair (Macbeth et al. 2010). Beyond working out the procedures for hair preparation and cortisol analysis, and the accuracy and precision of the measurement, Macbeth et al. (2010) also sought to determine how a range of factors, including hair colour, exposure to inclement weather for 18 days,

251

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress and laboratory storage for more than one year affect the hair cortisol concentration. The intent behind this work was to establish the robustness of the measurement and, by extension, build its utility as a wildlife conservation tool. In this report, we build on the foundation study of Macbeth et al. (2010) to investigate what factors influence the hair cortisol concentration (HCC) at the level of individual bears, again with the intent of developing HCC as a wildlife conservation tool. Our approach is both opportunistic and stepwise. It is opportunistic in the sense that we make use of HCC data collected by several different projects, all focused on brown bears but involving different populations, different research objectives, and different methods of hair collection. It is stepwise in the sense that we start with a broad comparison of HCC levels across projects and then ask whether differences between projects correctly reflect differences in long-term stress or are there confounding factors at play. Controlling for appropriate covariates, such as sex, life-history stage, and season is essential to the interpretation of HCC data (Busch et al. 2009). What follows is essentially a series of questions with each subsequent step in our analysis determined by the outcome of the previous step. In the end, we demonstrate two ways in which HCC can be applied as a monitoring tool for populations.

Materials and Methods Sources of brown bear hair We obtained 785 brown bear hair samples from six different projects representing four geographic areas across the species’ Holarctic distribution (Table 1). Although hair collection methods varied by project, hair samples were handled similarly in that they were air-dried and stored in a paper envelope under low light at room temperature (~20C) until the time of laboratory analysis. Details of each project follow with the circumstances around hair collection and availability of supplementary data summarized in Table 1. 1. Alberta Grizzly Bear DNA Inventory for Population and Density Estimates (Alberta, 2004-2008): As part of grizzly bear management and recovery in Alberta (Canada), the provincial government conducted DNA inventories from 2004-2008 to estimate population size and density for five bear management units (Alberta SRD 2010). The total area covered approximately 132,000 km2, mostly in and adjacent to the eastern slopes of the Rocky Mountains, from northern Montana (49°N, 113°W) to the city of Grande Prairie (55°N, 118°W). Hair was collected by barbwire snag during June and July of each year with study design details provided in a series of technical reports (Boulanger et al. 2005a, 2005b, 2007, 2008). Absolute locations were determined for all samples by Global Positioning System (GPS) receiver and suitable samples were genotyped to confirm species, sex, and individual identity (Paetkau 2003). 2. Eastern Slopes Grizzly Bear Project (Alberta, 1994-2002): The Eastern Slopes Grizzly Bear Project was conducted over a 9-year period with the goal of contributing toward a scientific understanding of grizzly bear biology, ecology, and demography in an area of west-central Alberta and east-central British Columbia known as the Central Canadian Rocky Mountain Ecosystem (50-52N, 114-117W) (Herrero 2005). Hair samples were collected from live-

252

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress

captured bears with details on capture and handling procedures provided in Garshelis et al. (2005). In addition to hair samples, we also were provided with supplementary data that included the identification, sex, age, location, body mass, morphometric measurements, and capture history for each bear sampled. Foothills Research Institute Grizzly Bear Program (Alberta, 1999-present): The Foothills Research Institute Grizzly Bear Program, now in its 13th year, was initiated in 1999 to provide knowledge and planning tools to land and resource managers to ensure the long-term conservation of grizzly bears in Alberta (Stenhouse and Graham 2011). The study area encompasses the entire distributional range of grizzly bears within the province (49-58N, 113120W) with yearly research effort typically targeted toward one or two bear management units. Hair samples were collected from live-captured bears with details on capture and handling procedures provided in Cattet et al. (2008). In addition to hair samples, we also have an extensive sum of data for each bear that includes its identification, sex, age, location, body mass, morphometric measurements, and capture history. Scandinavina Brown Bear Research Project (Sweden, 1984-present): The Scandinavian Brown Bear Research Project (SBBRP) was initiated in 1984 with the primary goals to (i) document the basic ecology of the Scandinavian brown bear, (ii) provide management authorities with data and interpretations of the results to help meet bear management objectives, and (iii) provide information about brown bears to the general public. The Swedish study area (Zedrosser et al. 2006) covers approximately 21,000 km2 of intensively managed boreal forest in a rolling landscape in south-central Sweden (61°N, 14°E), and mountainous national parks and adjacent forested land in the north (67°N, 18°E). Under SBBRP coordination, hair samples were collected by hunters from harvested brown bears in Sweden from August to October, 2008. By law, successful brown bear hunters are required to present bear carcasses to an officially appointed inspector on the day of the harvest and to provide information about harvest methods, sex of harvested bears, body mass, and harvest location to the Swedish Hunters Association and the National Veterinary Institute of Sweden (Bischoff et al. 2008). The information and samples are archived by the National Veterinary Institute of Sweden. Western Kitikmeot Grizzly Bear Project (Nunavut, 2004-present): The major goal of the Western Kitikmeot Grizzly Bear Project is to assess the status of grizzly bear populations in the Western Kitikmeot region of Nunavut (Canada) using genetic information extracted from hair trapped by barbwire snag on an annual basis since 2005 (Dumond et al. 2011). The study area covers approximately 40,000 km2 in western Nunavut (66.7-69.0N, 112.7-113.5W). We received hair samples collected from May to August 2008, as well as absolute locations and confirmation of species, sex, and individual identity for each sample (Paetkau 2003). Mongolian Gobi Bear Research Project (Mongolia, 2006-present): The primary objectives of the Mongolian Gobi Bear Research Project are to: (i) determine population size and assess potential limiting factors, (ii) determine habitat use and movement patterns between and within the three oasis complexes, (iii) determine genetic status, (iv) train and enhance capacity of Mongolian bear specialists in techniques for bear capture, handling, and data analysis, and (v) provide training on data collection and monitoring for rangers of the Great Gobi Strictly

253

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Protected Area (Tumendemberel et al. 2011). The study area is the Great Gobi Strictly Protected Area of southwestern Mongolia (42-44N, 95-98E). A DNA-based population inventory was conducted in 2008 and 2009 to estimate the population, ascertain sex ratio, document inter-oases movements of individual bears, and explore genetic variability of Gobi bears (Proctor et al. 2010). We received hair samples trapped by barbwire snag during this effort, as well as absolute locations and confirmation of species, sex, and individual identity for each sample (Paetkau 2003). Table 1. Source, sampling situation, and supplementary data for brown bear hair samples provided to this study for the determination of hair cortisol concentration as a marker of longterm stress.

Source

Year(s)

Sample (N)

Sampling situation

Alberta

20042008

350

barbwire snag

Alberta

19942010

192

Supplementary data Individual identification, sex, location

live-capture

Individual identification, sex, age, location, body mass, morphometric measurements, capture history Individual identification, sex, age, location, body mass, morphometric measurements

Sweden

2008

149

hunter-kill

Nunavut

2008

barbwire snag

Individual identification, sex, location

Mongolia

20082009

barbwire snag

Individual identification, sex, location

Laboratory analysis of hair cortisol concentration We used only guard hairs for the determination of hair cortisol concentration as recommended by Macbeth et al. (2010). Surface contamination was removed by washing hairs with methanol as described in detail elsewhere (Macbeth et al. 2010). Following decontamination, hair was dried and then ground to a fine powder using a ball mill. Ground hair samples were immersed in 0.5 ml of high resolution gas chromatography grade methanol, gently swirled (10 s), and placed on a slowlyspinning rotator to extract for 24 hours. The hair extract was dried under nitrogen gas (38°C), concentrated, reconstituted in phosphate buffer (0.2 ml), and analyzed on a commercially available enzyme linked immunoassay kit (Oxford EA-65 Cortisol EIA kit, Oxford Biomedical, Lansing, MI, USA). Any remaining guard hairs, as well as undercoat hair, was stored at room temperature in the dark and surplus reconstituted hair extract was stored at -80°C.

254

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress

Statistical analysis We analyzed data using IBM-SPSS version 19.0.0 (IBM Corporation, Armonk, New York, United States) with the significance level () set at ≥ 0.95 and power (1 - ) set at ≥ 0.80 for statistical tests. Before performing tests though, all variables were assessed for normality and the need to remove influential outliers and/or apply natural log-transformations. Consequently, sample size varied to some degree between tests. To compare mean hair cortisol concentration between sources, we performed a univariate analysis of variance (ANOVA) with source as the factor, including the separation of Alberta bears into two sources based on the circumstances around hair collection, live-capture or barbwire snag. For any bears sampled more than once within the same cycle of hair growth (August to May), we used only the HCC value recorded for the first sampling. HCC values for bears sampled more than once, but in different years (different hair cycles), were treated as individual observations. We used Tukey’s btest for a post-hoc comparison of means. Mongolia data was excluded from this and subsequent analyses due to small sample size (n = 10). We also calculated median values for the HCC given that its distribution is likely to always be positively skewed (skewed to the right) with the minimum possible value set at 0.16 pg/mg cortisol, which is half the limit of detection of the assay (Macbeth et al. 2010). Median values were compared between sources using an independent samples median test with identification of homogeneous subsets based on all pair-wise comparisons. We compared mean and median HCC values between female and male brown bears from each source (except Mongolia) using a t-test for comparison of means and an independent samples median test, respectively. The availability of supplementary data for live-captured bears from Alberta and hunter-killed bears from Sweden allowed us to compare between sources the influence of a common set of potential predictor variables (sex, age, body mass, chest girth, and contour length) on the HCC. For this, we used generalized linear models in SPSS with the Wald 2-statistic to describe model effects and the adjusted-R2 (also called pseudo-R2) value to quantify the proportion of variation in HCC explained by the complete model. We also repeated the same procedures for live-captured bears from Alberta, but expanded the range of potential predictor variables to include: (i) (ii) (iii) (iv) (v)

Body condition status – categorized as ‘normal’ or ‘poor’ Season of capture – categorized as ‘hypophagia’ (den emergence to June 14), ‘early hyperphagia’ (June 15 to August 7), or ‘late hyperhagia’ (August 8 to den entry) Capture method – categorized as ‘by culvert trap’ or ‘by other methods’, i.e., by leg-hold snare or by remote drug delivery from helicopter Radio-telemetry – categorized as ‘not wearing a radio-collar and/or radio-ear tag prior to capture’ or ‘wearing a radio-collar and/or radio-ear tag prior to capture’ Capture number – number of times captured previously + 1

255

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress The categorization of body condition was based on the body condition index (BCI) value for each bear as described by Cattet et al. (2002). We assigned bears as ‘normal’ or ‘poor’ by using the 25th percentile BCI value by sex and capture season as the cut-off value. Female brown bears (n = 162) in poor condition had BCI values of ≤ -0.66 during hypophagia, ≤ +0.02 during early hyperphagia, and ≤ +1.28 during late hyperphagia. Male brown bears (n = 147) in poor condition had BCI values of ≤ +0.04 during hypophagia, ≤ -0.26 during early hyperphagia, and ≤ +0.93 during late hyperphagia. We also performed several smaller analyses that were directed at three specific questions arising from the analyses described above. First, to determine if physical injury was greater than expected in bears wearing radio-telemetry devices (radio-collar and/or radio-ear tag), we used a Pearson 2test to compare the frequency of physical injury observed for bears wearing radio-telemetry devices versus that for bears without these devices. Second, to determine if there was any evidence to show that sudden death can cause an immediate rise in HCC (acute effect), we compared mean and median HCC values using a t-test for two independent samples and an independent samples median test, respectively, between a select group of “normal-health” Alberta live-capture bears and twelve Alberta bears that had died suddenly as a consequence of research (n = 4), management (n = 6), or hit-by-vehicle (n = 2). The normal-health group comprised twenty bears that met the following criteria: (i) HCC values were measured, (ii) no physical injuries were recorded, (iii) no complete blood count abnormalities were identified, (iv) no serum biochemistry abnormalities were identified, (v) and the bear had not been captured previously. Third, to determine if capture effects alone could explain differences in median HCC values between bears that were sampled following live-capture and bears that were sampled by barbwire snag, we divided bears from Alberta that were sampled by barbwire snag into three groups as follows: (i) bears that had never been captured (n = 299), (ii) bears that were captured previously, but during a different cycle of hair growth (n = 41), and (iii) bears that were captured previously, but during the same cycle of hair growth (n = 10). We then compared the three groups using an independent samples median test followed by pair-wise comparisons for homogeneous subsets. The final facet of our analysis was focused on determining ways in which the HCC could be applied as a conservation tool to monitor long-term stress at the population level. For this analysis, we used the HCC values for bears from Alberta that were sampled by barbwire snag across five bear management units (n = 350; Fig. 1). In the first approach, we considered using the HCC as a direct measure of long-term stress, calculated median HCC values for each bear management area (population), and then compared between populations using an independent samples median test followed by pair-wise comparisons for homogeneous subsets. In the second approach, we considered using the HCC as an indirect measure of body condition. For this, we selected a subset of Alberta live-captured bears that had not been captured previously and were not captured by culvert trap when sampled (n = 76). We divided these bears into two groups,

256

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress bears with poor body condition status (n = 24) and bears with normal body condition status (n = 52) with the categorization of ‘normal’ and ‘poor’ as described above. We then overlapped the distributions of HCC values for the two groups and selected a threshold HCC value that would provide high sensitivity (≥ 90%) that the observed HCC ≥ the threshold HCC if a bear is in poor body condition. We gave less attention to specificity as a test performance measure on the assumption that, from a conservation recovery standpoint, it would be more important to correctly identify bears in poor body condition than to correctly identify bears in normal body condition. Following this, we used the threshold HCC to estimate the proportion of bears sampled by barbwire snag that were in normal versus poor body condition within each of the five bear management units. Finally, we performed a nonparametric multiple comparison of proportions to see if we could detect population-level differences in body condition.

Figure 1. Alberta grizzly bear management areas sampled by barbwire hair snag from 2004-2008 for DNA inventories to estimate population size and density.

Results 257

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress We found significant differences (univariate ANOVA – F = 140.9, p ≤ 0.001) in mean hair cortisol concentrations (HCC) between sources (Fig. 2a). However, for some pair-wise comparisons, it appears that these differences may have been attributable to differences in the situation of hair collection rather than the source of hair. For example, the mean HCC for Alberta bears sampled during live-capture (mean  SE: 4.74  0.248 pg/mg) was almost four-fold greater than that for Alberta bears sampled by barbwire hair snag (1.25  0.199 pg/mg). For other comparisons, the situation of hair collection was similar suggesting that differences in mean HCC related more to the source of hair. For example, the mean HCC for Nunavut bears (3.71  0.375 pg/mg) sampled by barbwire snag was three-fold greater than that for Alberta bears sampled in the same manner. The mean HCC for Sweden bears (2.16  0.116 pg/mg) was intermediate to other sources, but the situation of hair collection also was unique for this source. The mean HCC for Mongolia was the lowest (0.93±0.20 pg/mg) among sources, but the sample size for this source was quite small (n = 10).

Figure 2a. Comparison of mean hair cortisol concentration between different sources of brown bear hair and different methods of hair collection. Sample sizes are provided in parentheses along the abscissa. Letters ‘a’-‘d’ are used to indicate significant differences (p ≤ 0.05) between means where a > b > c > d based on Tukey’s b-test for a post-hoc comparison of means.

Although the median HCC was less than the mean for each source (Alberta-hair snag: 0.86 pg/mg, Alberta-live capture: 2.90 pg/mg, Sweden: 1.97 pg/mg, Nunavut: 2.23 pg/mg, Mongolia: 0.77

258

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress pg/mg), the pattern of differences between sources and/or situations of hair collection was essentially the same (Fig. 2b; independent samples median test – 2 = 124.1, p ≤ 0.001). Mean and median HCC were similar between female and male brown bears within each source except Nunavut (Tables 2a and b) where values for males were approximately 50% greater than values for females.

Figure 2b. Comparison of median hair cortisol concentration between different sources of brown bear hair and different methods of hair collection. Sample sizes are provided in parentheses along the abscissa. Letters ‘a’ and ‘b’ are used to indicate significant differences (p ≤ 0.05) between medians where a > b based on all pair-wise comparisons.

259

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Table 2a. Comparison of mean hair cortisol concentrations between female and male brown bears from several sources.

Source Alberta (hair snag)

Hair cortisol concentration (pg/mg) Females [mean ± SE, Males [mean ± SE, N] N] 1.270.133 1.180.138 184 156

Statistical significanceA 0.430

Alberta (livecapture)

5.100.668 91

4.420.437 101

0.845

Sweden

2.200.158 60

2.130.163 89

0.629

Nunavut

2.900.430 48

4.780.969 36

0.020

Mongolia

0.73 1

0.960.217 9

Statistical significance (p ≤ 0.05) based on t-test for comparison of means using ln-transformed data. ‘na’ indicates not applicable due to small sample size.

Table 2b. Comparison of median hair cortisol concentrations between female and male brown bears from several sources.

Source Alberta (hair snag)

Hair cortisol concentration (pg/mg) Females [median, N] Males [median, N] 0.87 0.81 184 156

Statistical significanceA 0.650

Alberta (live-capture)

2.89 91

3.00 101

0.885

Sweden

2.07 60

1.96 89

0.462

Nunavut

1.97 48

2.94 36

0.008

Mongolia

0.73 1

0.81 9

Statistical significance (p ≤ 0.05) based on independent samples median test. ‘na’ indicates not applicable due to small sample size.

260

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress The availability of supplementary data from live-captured bears in Alberta and hunter-killed bears in Sweden (Table 1), and the fact that the same types of measurements were recorded at both sources, allowed us to begin exploring what factors influence HCC and if these factors are similar between sources. Using the same combination of biological and physical attributes, we identified some differences between sources in explanatory variables (Table 3). For example, HCC increased with age in Sweden bears but showed no association with age in Alberta bears. HCC did increase with body (contour) length in Alberta bears, but the overall model for Alberta lacked in explanatory power accounting for only 4% of the variation in HCC. In contrast, the Sweden model which was composed of the same explanatory variables accounted for approximately 28% of the variation in HCC. Also notable with the Sweden model was that a weak but statistically significant difference between female and male bears was uncovered when controlling for age and physical attributes.

Table 3. Model coefficients describing variation in the hair cortisol concentrations of brown bears sampled in Sweden and Alberta based on analyses using the same potential explanatory variables.

Alberta (N = 93, adjusted R2 = 0.04) b SE p

Sweden (N = 101, adjusted R2 = 0.28) b SE p

-0.095 0.183 0.604 0 (reference category)

-0.237 0.115 0.039 0 (reference category)

Age

-0.040

0.062

0.518

0.139

0.041

<0.001

Age2

0.003

0.002

0.269

-0.004

0.002

0.023

Mass

-0.965

0.452

0.033

0.269

0.337

0.424

Chest girth

0.337

0.627

0.591

-1.557

0.609

0.010

Contour length

4.328

1.298

<0.001

-1.754

0.978

0.073

-17.978

5.658

<0.001

14.907

3.448

<0.001

Variable Sex: female male

Constant

To further evaluate the HCC of live-captured bears from Alberta, we expanded the suite of potential explanatory variables to include body condition, season of capture, and several capture-related variables (Table 4). Bears in normal body condition tended to have a lower HCC (adjusted mean  SE: 4.56  1.160 pg/mg, n = 72) than bears in poor body condition (6.42  0.883 pg/mg, n = 21; Wald 2 = 5.11, p = 0.024). This difference also remained consistent over the three capture seasons (Fig. 3a) although HCC tended to be lower during late hyperphagia than during the preceding seasons (Wald 2 = 6.26, p = 0.044). Bears captured by culvert trap had markedly higher HCC (7.77 

261

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress 1.196 pg/mg, n = 23) than measured in bears captured by leg-hold snare or by remote drug delivery from a helicopter (3.21  0.619 pg/mg, n = 70; Wald 2 = 41.52, p ≤ 0.001; Fig. 3b). Bears that were captured while wearing a radio-telemetry device (radio-collar and/or radio-ear tag) from a previous capture had higher HCC (6.38  0.730 pg/mg, n = 32) than bears without these devices (4.60  1.085 pg/mg, n = 61; Wald 2 = 9.06, p = 0.003; Fig. 3c). The frequency of physical injuries also was higher in bears wearing radio-telemetry devices (43 of 145 [30%] vs. 17 of 159 [11%]) (Pearson 2 = 17.22, p ≤ 0.001). Although injuries were not recorded in any standardized manner, a review of our field records indicated that injuries to bears wearing radio-telemetry devices were associated with ear-tags (both radio and non-radio types) in 19 cases, radio-collars in 14 cases, previous snareinduced injuries in 5 cases, and injuries of unknown cause in 5 cases. Injuries to bears not wearing radio-telemetry devices were associated with ear-tags in 5 cases and unknown cause in 12 cases. The number of times that a bear was captured was inversely associated with the HCC (Wald 2 = 13.39, p ≤ 0.001). This expanded model for Alberta bears accounted for approximately 34% of the variation in HCC, a marked improvement over the model presented in Table 3. We were able to consider the question of whether or not sudden death can cause an immediate, significant increase in HCC (acute effect) by comparing HCC between twenty “normal-health” Alberta live-capture bears and twelve Alberta bears that died suddenly. We found no significant differences in mean HCC between the two groups (normal-health: 4.38  0.901 pg/mg, sudden death: 3.49  0.863 pg/mg; t = 0.429, p = 0.671]. A comparison between median HCC yielded similar results (normal-health: 2.62 pg/mg, sudden-death: 2.53 pg/mg; independent samples median test – p = 1.000). These findings do not give credence to the possibility that the higher HCC in Sweden bears relative to Alberta hair-snag bears and Mongolia bears could be explained by the fact that hair samples were collected from hunter-killed bears in Sweden. To determine if long-term capture effects on HCC were significant enough to account for the difference in HCC between Alberta bears sampled during live-capture and Alberta bears sampled by hair snag, we compared HCC between three sub-groups of Alberta bears sampled by hair snag. Bears that were sampled twice within the same hair growth cycle, first by live-capture and then by hair snag, had higher mean HCC (3.78  2.229 pg/mg, n = 10) than did Alberta bears sampled first by live-capture and then by hair snag, but in different hair growth cycles (1.05  0.293 pg/mg, n = 41), and Alberta bears sampled by hair snag only (1.25  0.101 pg/mg, n = 299) (univariate ANOVA – F = 4.59, p = 0.010, observed power = 0.775). Although the power of this analysis was < 0.80, the outcome was similar when comparing median HCC (same hair cycle: 2.02 pg/mg, n = 10; different hair cycle: 0.51 pg/mg, n = 41; hair snag only: 0.90 pg/mg, n = 299; independent samples KruskalWallis test, p = 0.010).

262

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Table 4. Model coefficients describing variation in the hair cortisol concentrations of brown bears sampled in Alberta based on analysis using the full suite of potential explanatory variables.

Alberta (N = 93, adjusted R2 = 0.34) b SE p

Variable Sex: female male

-0.010 0

0.157 0.950 (reference category)

Age

-0.042

0.055

0.444

Age2

0.003

0.002

0.124

Mass

-0.838

0.368

0.023

Chest girth

0.728

0.360

0.043

Contour length

3.081

1.259

0.014

Body condition status: normal poor

-0.379 0

0.168 0.024 (reference category)

Season of capture: hypophagia (den emergence – Jun 14) early hyperphagia (Jun 15 – Aug 7) late hyperphagia (Aug 8 – den entry)

0.562 0.450 0

0.227 0.013 0.257 0.080 (reference category)

Capture method: culvert trap leg-hold snare or helicopter capture

0.877 0

0.136 <0.001 (reference category)

Radio-telemetry: no radio-collar and/or radio-ear tag wearing radio-collar and/or radio-ear tag

-0.401 0

0.133 0.003 (reference category)

Capture number

-0.198

0.054

<0.001

Constant

-17.978

5.658

<0.001

263

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress

Figure 3. The effects of a) body condition, b) capture method, and c) radiotelemetry device on the predicted hair cortisol concentration for an adult brown bear during different capture seasons. Predicted mean values and 95% confidence intervals (hatched lines) were estimated from the model presented in Table 4 and adjusted for an 8-year old bear.

264

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress When applying HCC as a direct measure of long-term stress across five bear management units (BMU) in Alberta (Fig. 1), the highest median HCC occurred in the Livingstone, Castle, and Grande Cache BMUs (Table 5). The lowest median HCC occurred in the Yellowhead BMU, while the median HCC for the Clearwater BMU was intermediate in value (Independent samples median test – 2 = 23.02, p ≤ 0.001; homogeneous subsets ‘a’, ’ab’, and ‘b’ based on all pair-wise comparisons where a ≥ ab ≥ b and a > b). Although not shown in Table 5, we found similar results when comparing mean HCC between BMUs ([Livingstone – 1.60  0.266 pg/mg, n = 75; Castle – 2.04  0.770 pg/mg, n = 26; Grande Cache – 1.05  0.067 pg/mg, n = 144] ≥ [Clearwater – 1.25  0.227 pg/mg, n = 43] ≥ [Yellowhead – 0.75  0.132 pg/mg, n = 52]; univariate ANOVA – F = 4.22, p = 0.002; Tukey’s b-test for post hoc comparison of means – p ≤ 0.05).

Table 5. Median hair cortisol concentration (HCC) and predicted body condition status of Alberta grizzly bear management units (BMU) based on the HCC of hair samples collected from 340 grizzly bears during hair-snag genetic inventories conducted in the province from 2004-2008.

BMU (Year) Grande Cache (2008)

Median HCCA (pg/mg) [minimum maximum] 0.89a [0.16 - 4.31]

Predicted Body Condition StatusB (count, percent) Normal (HCC < 2.05 Poor (HCC ≥ 2.05 pg/mg) pg/mg) 127 (88.2%)

17 (11.8%)

Yellowhead (2004)

0.35b [0.16 - 4.85]

47 (90.4%)

5 (9.6%)

Clearwater (2005)

0.86ab [0.16 - 6.57]

34 (79.1%)

9 (20.9%)

Livingstone (2006)

1.07a [0.16 - 14.94]

60 (80.0%)

15 (20.0%)

Castle (2007)

1.01a [0.16 - 15.62]

22 (84.6%)

4 (15.4%)

Independent samples median test – 2 = 23.02, p ≤ 0.001; homogeneous subsets ‘a’, ’ab’, and ‘b’ based on all pair-wise comparisons. B Multiple comparison of proportions by BMU – Pearson 2 = 5.05, p = 0.282. A

To apply HCC as an indirect measure of body condition we calculated sensitivity and specificity values as test performance measures across a range of HCC values from 1.63-10.00 pg/mg (Fig. 4). We obtained a sensitivity of 91.7% (0.917) and a specificity of 44.2% (0.442) when selecting HCC ≥2.05 pg/mg as the threshold value. At this level of performance, 22 of 24 bears in poor body condition were correctly “diagnosed” as poor body condition (i.e., high sensitivity), whereas 29 of 52 bears in normal body condition were incorrectly diagnosed as poor body condition (i.e., low specificity). When we used the threshold HCC to divide Alberta bears sampled by hair snag (n = 340)

265

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress based on estimated body condition status, we did not detect significant differences between BMUs in their proportions of bears in poor body condition (Table 5; multiple comparison of proportions by BMU â&#x20AC;&#x201C; Pearson ď Ł2 = 5.05, p = 0.282).

Figure 4. Performance curves for the application of hair cortisol concentration (HCC) as an indicator of body condition status in Alberta brown bears. The selection of an HCC threshold value of 2.05 pg/mg provides high sensitivity (91.7%) to detect bears in poor body condition, but low specificity (44.2%) to detect bears in normal body condition.

Discussion In this study, we explored what factors influenced hair cortisol concentration (HCC) at the level of individual bears with the intent of ascertaining if HCC is a reliable indicator of long-term stress with potential application as a wildlife conservation tool. HCC differed significantly between brown bears from different areas of their distributional range that included several populations in Canada, Sweden, and Mongolia. By examining HCC in conjunction with supplementary data collected from the same animals, we were able to identify a range of biological, physical, and capture-related attributes that appear to influence HCC over the long-term, i.e., weeks to months. We also were unable to identify any measurable difference in HCC between normal-health bears sampled during live capture and bears that were sampled following sudden death which suggests that death does not cause an immediate (acute) increase in HCC. Taken together, these findings support the

266

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress measure of HCC as an indicator of long-term stress that is unlikely to be influenced by acute stressors that occur on an infrequent basis. At a population-level, we demonstrated how HCC could be used remotely (e.g., sampling by barbwire hair snag) as a direct measure of long-term stress to compare between populations or monitor populations over time. We also demonstrated how HCC could be used in much the same manner as a medical diagnostic test to remotely detect bears in poor body condition with high sensitivity, but low specificity. These final steps point toward ways in which the HCC could be applied as a wildlife conservation tool.

Life history effects on HCC The most plausible explanation for the finding that HCC was influenced by gender in Nunavut and to a lesser extent Sweden, and by age in Sweden is that hypothalamus-pituitary-adrenal (HPA-) axis activity can vary by life-history stage in brown bears. The recognition that life-history stages and the activity of the HPA-axis are intimately linked is well established across a wide variety of mammals (reviews in Boonstra 2005, Reeder and Kramer 2005). Typically, females have higher baseline glucocorticoids (with cortisol being the major glucocorticoid in most mammals) and a more robust stress response than males; an observation that has been attributed to the differential effects of sex steroids on females and males (Handa et al. 1994, McCormick et al. 2002). However, relative to males and non-pregnant females, stress-induced glucocorticoids are markedly decreased during pregnancy, as well as during lactation, as a consequence of increased protein-binding (mostly corticosteroid-binding globulin) of glucocorticoids (Chow et al. 2011, Lightman et al. 2001, Stern et al. 1973). This likely has the effect of buffering females from elevated glucocorticoids or large fluctuations in glucorticoid levels, both of which can adversely affect developmental processes and the subsequent health and survival of offspring (Wadhwa et al. 2001, Weinstock 2005). Because the cortisol extracted from hair is in an active form (free, not protein-bound), the higher HCC detected in male bears from Nunavut and Sweden suggests a few possibilities. One is that males may face more stressful environmental conditions that females. Sexual segregation is known to occur in some brown bear populations (Dahle and Swenson 2003, Rode et al. 2006) and, under this condition, so presumably would exposure to environmental stressors. Another possibility is that the proportions of pregnant and/or lactating bears represented in the hair samples received from Nunavut and Sweden were greater than in those received from sources in Alberta. In this case, female bears would be predicted to have lower HCC on average than males due to increased protein-binding of glucocorticoids. Confirmation of either or both explanations would require additional data lacking from this study. Superimposed on sex differences in HPA activity, aging is believed to be associated with reduced control of the HPA axis with the nature of this change varying between and within species (Reeder and Kramer 2005). For example, baseline glucocorticoid levels increase with age in rats (Dellu et al. 1996), but remain unchanged in aging non-human primates (Goncharova and Lapin 2002). However, one age-related change that appears to be consistent for many mammals is a gradual impairment of the negative feedback system for the HPA axis which results in a slower return to baseline glucocorticoid levels following exposure to a stressor (Sapolsky et al. 1983). This may explain the greater HCC that we observed in older brown bears from Sweden. The reason why we

267

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress didn’t also detect an age effect in live-captured bears from Alberta is not certain. The age statistics and ranges were quite similar between sources (mean  SE, minimum-maximum: Sweden – 6.3  0.54 yrs, 1-25 yrs, n = 101; Alberta – 7.5  0.50 yrs, 2-22 yrs, n = 93), so it is unlikely that differences in age distribution could provide an explanation. Perhaps any similar age-related increase in the HCC of live-captured bears from Alberta was simply obscured by the apparent greater influence of body length (see Tables 3 and 4). Although the results are not presented in this report, we did find a weakly significant (partial correlation analysis – p = 0.049), positive association between age and HCC when controlling for contour length.

Body condition effects on HCC The finding that HCC was lower in Alberta bears with normal body condition than those with poor body condition is consistent with findings in other studies where associations between body condition and glucocorticoid levels in other biological media, including blood (Romero and Wikelski 2001, Wingfield and Kitaysky 2002) and feces (Cabezas et al. 2007, Gladbach et al. 2011), have been examined. In some cases, energy deficiency resulting from insufficient food availability and limited physiological energy stores is the primary stressor causing a prolonged stress response and loss of body condition (du Dot et al. 2009, Ortiz et al. 2001). The association between increasing stress and diminishing body condition is typically tight in this situation. In other cases, adverse environmental conditions aside limited food availability initiate a prolonged stress response which in turn may draw on physiological energy stores to meet the requirements of essential metabolic processes (McNamara and Buchanan 2005). In this situation, the association between stress and body condition may be weak or absent. The association between HCC and body condition (as reflected by the body condition index [BCI]) in Alberta bears was weak (Pearson r = -0.21, p = 0.014, n = 137) which suggests that for many bears it was not primary energy deficiency but other stressors affecting HCC levels. Although we selected the 25th percentile BCI value by sex and capture season as the threshold between normal and poor body condition status, this decision was somewhat arbitrary in the sense that we do not know at what BCI level is health likely to be compromised. Nonetheless, when we estimate percent body fat (BF) from BCI values using a prediction curve developed by Baldwin and Bender (2010; see Fig. 2 on p. 49), the equivalent BF values for females are 3% during hypophagia (den emergence to June 14), 7% during early hyperphagia (June 15 to August 7), and 21% during late hyperphagia (August 8 to den entry). Recently, Robbins et al. (2012) determined in a study of six captive female brown bears that no female with a BF ≤20% immediately prior to den entry produced cubs even though breeding occurred. This suggests that our selection of the 25th percentile BCI is at least close to the body condition level, especially prior to den entry, where reproduction may be compromised for females. The equivalent BF values for males are 8% during hypophagia, 5% during early hyperphagia, and 17% during late hyperphagia, but we do not know if adverse health consequences are likely to occur at these levels. Seasonal variation in baseline and stress-induced HPA-axis activity appears to be the norm for many animals (Romero 2002, Vera et al. 2011). The pattern that we observed in Alberta brown bears

268

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress does fit that described for some mammals, as well as reptiles, amphibians, and birds, in which glucocorticoid levels are highest during the breeding season and lower at other times of the year (Romero 2002). The highest HCC were measured during hypophagia and early hyperphagia which also is concurrent with the breeding season for brown bears in Alberta (Stenhouse et al. 2005). However, seasonal differences in HCC for Alberta bears also could be explained by seasonal differences in body condition. The season of late hyperphagia is when bears are attaining their greatest amount of body fat in preparation for entering winter dens. Given the inverse association between HCC and body condition, HCC should be lowest at this time. Regardless of cause, controlling for seasonal variation should be an important consideration when monitoring HCC in a population over time or comparing HCC among populations.

Capture effects on HCC The combined effects of capture-related factors accounted for most of the variation in HCC (28 of 34%) explained by the model shown in Table 4, and was enough to explain the difference in mean HCC between Alberta bears sampled following live-capture (4.74  0.248 pg/mg) and Alberta bears sampled by barbwire snag (1.25  0.199 pg/mg). Capture method had the most influence with bears captured by culvert trap having on average an HCC over two-fold higher than HCC values measured in bears captured by leg-hold snare or by remote drug delivery from helicopter. Macbeth et al. (2010) previously reported this culvert trap-effect and demonstrated that contamination of hair with urine and feces while a bear is held in a trap can alter the permeability of the hair and through inward diffusion increase the concentration of cortisol within the hair shaft. This, however, may not provide the full explanation because not all bears are exposed to urine and feces in culvert traps and because, in cases where bears have been contaminated, field personnel preferentially seek non-soiled hair to sample (Macbeth et al. 2010). Further to this, we have recently found that serum corticosteroid binding globulin (CBG) levels in many of the bears for which HCC was measured were lower in those captured by culvert trap (4.13  0.403 g/ml) than those captured by other methods (5.85  0.185 g/ml) (Brian Chow, unpublished results). This finding also could explain a higher HCC in culvert trapped-bears because with lower CBG presumably there would be more circulating free cortisol to diffuse into growing hair. However, CBG levels are believed to be relatively insensitive to short-term stressors associated with capture and handling (Chow et al. 2010, Tinnikov 1999), thus the basis for the association between high HCC and capture by culvert trap remains uncertain. The presence of a radio-telemetry device, whether it is a radio-collar and/or a radio-ear tag, deployed at a previous capture was associated with a significantly higher HCC (6.38  0.730 pg/mg, n = 32) than measured in bears without these devices (4.60  1.085 pg/mg, n = 61). While some of this difference may have been attributable to physical injury as a direct result of wearing these devices, e.g., neck lacerations and chronic-active inflammation and infection around the ear tag insertion hole, injuries were not observed in the majority of bears (70%) wearing radio-telemetry devices which suggests that the presence of these devices alone was a significant stressor for some animals. A critical assumption of radio-telemetry studies has long been that the radio-transmitters themselves do not influence an animal’s behavior, physiology, or survival (Coté et al. 1998, Pollock

269

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress et al. 1989, Barron et al. 2010). However, several recent studies have challenged this assumption by documenting negative effects of radio-telemetry or attachment devices in a variety of animals, including reptiles (Knapp and Abarca 2009, Lentini et al. 2011), fishes (Jepsen et al. 2008, Stakėnas et al. 2009), birds (Barron et al. 2010, Saraux et al. 2011), and both terrestrial and marine mammals (Swenson et al. 1999, Léchenne et al. 2011, Walker et al. 2011). Although we are not aware of previous reports on the potential for negative effects of radio-telemetry devices on the behavior and physiology of brown bears, or other species of Ursidae for that matter, the results from this study show that the increase in HCC associated with wearing these devices was similar to the increase in HCC associated with being in poor body condition. The implications of this finding are potentially far-reaching and suggest the possibility that, in the worse case, research findings and management or policy decisions derived from telemetry-based studies could be unsuspectingly biased. We strongly recommend further study to address several key questions that should include: (1) Do telemetry devices have an overall effect on the fitness (reproduction and survival) of bears? (2) Which features of bear behavior, physiology, and ecology are affected? (3) What characteristics of bears influence radio-transmitter effects? (4) What characteristics of telemetry devices influence their effects? The long-term goal in addressing these questions is to be able to knowledgeably balance the benefits of using radio-telemetry against the potential costs to bears and the reliability of the data obtained (Barron et al. 2010). The number of captures that a bear experienced also influenced the HCC, but in a manner that was contrary to what might be predicted. In general, bears that were captured several times had lower HCC than bears that had not been captured previously. However, the statistical significance of this apparent capture effect appears to have been highly influenced by a single bear that had been captured eight times and had a low HCC (2.10 pg/mg) relative to other live-captured bears in Alberta. When we excluded the data for this individual from the analysis, the association between HCC and number of captures was no longer significant (Pearson r = 0.03, p = 0.736, n = 92). Further, comparing mean HCC between bears captured once (4.23  0.514 pg/mg, n = 59) and bears captured 2-5 times (4.56  0.598 pg/mg, n = 33) did not yield a significant difference (t-test for independent samples – p = 0.360). Thus, with this correction, we found no evidence to suggest number of captures influenced HCC. We found no evidence to indicate the sudden death of a bear could cause an immediate increase in HCC. Although current opinion holds widely that the HCC should not increase measurably following an acute stress response, Sharpley et al. reported both diurnal variation in HCC (Sharpley et al. 2010) and transient localized changes in HCC in response to local pain (Sharpley et al. 2009). They attributed these transitory fluctuations to cortisol production by skin and hair follicles via their own neuroendocrine system known as the “peripheral HPA axis”, a distinct entity from the central HPA axis which involves cortisol production and secretion from the adrenal glands (Arck et al. 2006, Ito et al. 2005, Paus et al. 2006). Cortisol levels in blood often increase rapidly to very high levels during stressful agonal periods immediately prior to death (Erkut et al. 2004), but cortisol levels in blood collected following death may be similar to normal values obtained during life (Finlayson 1965, Palmiere and Mangin 2012). However, despite the possibility of agonal changes in blood

270

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress cortisol or peripheral HPA axis activity in brown bears, we were unable to detect any differences in HCC between twenty “normal-health” Alberta live-capture bears and twelve Alberta bears that died suddenly. Therefore, we have no reason to conclude that the HCC values measured in hunter-killed bears from Sweden were confounded by a fatal gunshot or the preceding hunt.

HCC as a conservation tool Thus far, we’ve considered factors that influence HCC levels in an individual bear. This knowledge is essential to determining if and how HCC can be applied as a marker of long-term stress in brown bears, or more broadly in wild mammals. Our findings support using HCC in this manner, but also give insight into what other factors must be considered in its application. The next step in widening the application of HCC as a marker of long-term stress is to ascertain its effectiveness at the population level because it is only at this level that its utility as a wildlife conservation tool can be made certain. We suggest two approaches, one that requires more knowledge to be fully effective, and the other that can be used now but has clear limitations. In order for HCC to serve as a useful tool for conservationists, it should fluctuate with environmental variables and correlate with fitness parameters (Breuner et al. 2008, Busch and Hayward 2009). From concurrent research investigating linkages between HCC and environmental attributes in Alberta brown bears, we do know that the combination of inter-annual variability in local weather and regional climate, and local anthropogenic features, explain approximately 30% of the variation in HCC (Stenhouse et al., manuscript in preparation). Presumably, most of this explanatory power is additive to the 34% already explained by the model in Table 4. What we are lacking, however, is information on the association between HCC and the two measures most relevant to fitness, survival and reproductive success. With this information, which is the focus of our current studies, we will be able to use the HCC as a direct measure of long-term stress at the population level. More specifically, we should be able to identify threshold HCC values above which survival is likely to be reduced or reproductive success is likely to be compromised. At this point, however, we’re limited to making comparisons among groups of bears (as in Figures 2a and 2b or Table 5), but unable to fully explain the ramifications of significant differences. Although survival and reproductive rates are not available for all bear management units (BMU) in Alberta, it is noteworthy that the median HCC values by BMU shown in Table 5 parallels the density of bears within each BMU adjusted for the area of lands protected from land-use activity. The Yellowhead BMU had the lowest median HCC at 0.35 pg/mg and lowest density at 6 bears per 1000 km2 of protected land. Conversely, the Livingtone and Castle BMUs had the highest median HCC at 1.07 pg/mg and 1.01 pg.mg, and the highest densities at 32 and 100 bears per 1000 km2, respectively. For some species, high population densities have been correlated with increased stress levels with this association attributed to increase antagonistic interactions and competition (Raouff et al. 2006, Sheriff et al. 2012). Hair cortisol concentration may also be applied at the population level as an indirect marker of poor body condition in much the same way that diagnostic tests are used in the health professions to detect disease, e.g., serum high density lipoprotein (HDL) cholesterol as a marker for coronary

271

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress heart disease. However, as an indirect marker of body condition, the usefulness of HCC will be determined by its strength of association with body condition. In this study, the association in Alberta grizzly bears was weak. So, we took a conservative approach in selecting a threshold HCC value (â&#x2030;Ľ 2.05 pg/mg) by ensuring it was highly sensitive (i.e., correctly identify 91.7% of the bears in poor body condition in a population), but not very specific (i.e., correctly identify 44.2% of the bears in normal body condition in a population). In a situation where energy deficiency is a primary stressor, for example in food-limited populations such as the Nunavut brown bears sampled in this study (Gau et al. 2002), the association between HCC and body condition should be considerably stronger offering the possibility of establishing a threshold HCC that is both highly sensitive and highly specific.

Conclusions With this study, we extended the work of Macbeth et al. (2010) to determine what factors influence HCC at the level of the individual bear. We found significant differences in HCC levels between brown bears representing different geographic areas across the speciesâ&#x20AC;&#x2122; distributional range, as well as between different bear management units within one geographic area. Through a series of analyses using available supplementary data, we determined that HCC can be influenced by sex and age, as well as physical attributes (body mass and body length), but not necessarily in a consistent manner across populations. Through a detailed analysis of brown bears in Alberta, we also identified body condition, season of hair collection, and several capture-related factors that influence HCC. Most troubling in this regard was the finding that the presence of radio-telemetry devices on bears was associated with significantly higher HCC than measured in bears without these devices. Because of the potentially far-reaching implications of this finding, further study is urgently needed to better understand the overall effects of telemetry devices on bears and other wildlife, and to knowledgeably balance the benefits of using radio-telemetry against potential wildlife health and welfare costs and the reliability of the data obtained. Overall, we found no evidence to discount HCC as a reliable indicator of long-term stress. Thus, we suggest that the differences in HCC levels between brown bears representing different geographic areas as shown in Figures 2a and 2b truly represents differences in long-term stress levels. The differences in mean HCC between Alberta bears sampled following live-capture and Alberta bears sampled by barbwire snag were largely explained by long-term effects of capture and handling. We propose that the high HCC levels in Nunavut bears may reflect long-term nutritional stress in a food-limited environment. Although we lack plausible explanations for the relatively high HCC in Sweden bears and low HCC in Mongolia bears, these findings could certainly serve as a basis for future research hypotheses. We also believe that the differences in HCC levels between Alberta BMUs reflect real differences in long-term stress, but again a full appreciation of causative factors is lacking. Although differences in population density may play a role, survival and reproductive rates by BMU are what is required to confirm the effectiveness of HCC as a wildlife conservation tool. If associations between HCC and fitness parameters are strong, then HCC should have considerable value for monitoring of long-

272

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress term stress in populations, particularly in conjunction with genetic sampling using DNA from bear hair collected with barbed wire hair traps. There also may be some value in using HCC to monitor body condition trends, especially in food-limited environments, but further research is required to substantiate this application.

Literature Cited Accorsi, P. A., E. Carloni, P. Valsecchi, R. Viggiani, M. Gamberoni, C. Tamanini, E. Seren. 2008. Cortisol determination in hair and faeces from domestic cats and dogs. General and Comparative Endocrinology 155: 398–402. Alberta Sustainable Resource Development (SRD). 2010. Grizzly bears. Government of Alberta Sustainable Resource Development website (see http://www.srd.alberta.ca/FishWildlife/WildlifeManagement/BearManagement/GrizzlyBea rs/Default.aspx - last accessed on 29 April 2012) Arck, P. C., Slominski, A., Theoharides, T. C., Peters, E. M. J., Paus, R. 2006. Neuroimmunology of stress: skin takes center stage. Journal of Investigative Dermatology 126: 1697–1704. Baldwin, R. A., Bender, L. C. 2010. Development of equations predictive of size and condition for black bears in Rocky Mountain National Park, Colorado. The American Midland Naturalist, 164: 44–51. Barron, D. G., Brawn, J. D., Weatherhead, P. J. 2010. Meta-analysis of transmitter effects on avian behaviour and ecology. Methods in Ecology and Evolution 1: 180–187. Bechshøft, T. Ø., C. Sonne, R. Dietz, E. W. Born, M. A. Novak, E. Henchey, J. S. Meyer. 2011. Cortisol levels in hair of East Greenland polar bears. Science of the Total Environment 409: 831– 834. Bischof, R., Fujita, R., Zedrosser, A., Soderberg, A., Swenson, J. E. 2008. Hunting patterns, ban on baiting, and harvest demographics of brown bears in Sweden. Journal of Wildlife Management 72: 79–88. Boonstra, R. 2005. Equipped for life: The adaptive role of the stress axis in male mammals. Journal of Mammalogy 86: 236–247. Bortolotti, G. R., T. A. Marchant, J. Blas, T. German. 2008. Corticosterone in feathers is a long-term, integrated measure of avian stress physiology. Functional Ecology 22: 494–500. Boulanger, J., G. Stenhouse, M. Proctor, S. Himmer, D. Paetkau, J. Cranston. 2005a. 2004 Population inventory and density estimates for the Alberta 3B and 4B grizzly bear management area. Report prepared for Alberta Sustainable Resource Development, Fish and Wildlife Division, May 2005 (with updates November 2005). Edmonton. 28 pages. Boulanger, J., G. Stenhouse, G. MacHutchon, M. Proctor, S. Himmer, D. Paetkau, J. Cranston. 2005b. Grizzly bear population and density estimates for the 2005 Alberta (proposed) unit 4 management area inventory. Report prepared for Alberta Sustainable Resource Development, Fish and Wildlife Division, December 2005. Edmonton. 31 pages. Boulanger, J., G. Stenhouse, G. MacHutchon, M. Proctor, D. Paetkau, J. Cranston. 2007. Grizzly bear population and density estimates for the 2006 Alberta unit 5 management area inventory.

273

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Report prepared for Alberta Sustainable Resource Development, Fish and Wildlife Division, May 2007. Alberta Sustainable Resource Development. Edmonton. 37 pages. Boulanger, J., G. MacHutchon, G. Stenhouse, M. Proctor, J. Cranston, D. Paetkau. 2008. Grizzly bear population and density estimates for Alberta Bear Management Unit 6 and British Columbia Management Units 4-1, 4-2, and 4-23 (2007). Report prepared for the Alberta Sustainable Resource Development, Fish and Wildlife Division, British Columbia Ministry of Forests and Range, British Columbia Ministry of Environment, and Parks Canada. 46 pages. Breuner, C. W., Patterson, S. H., Hahn, T. P. 2008. In search of relationships between the acute adrenocortical response and fitness. General and Comparative Endocrinology 157: 288– 295. Busch, D. S., L. S. Hayward. 2009. Stress in a conservation context: A discussion of glucocorticoid actions and how levels change with conservation-relevant variables. Biological Conservation 142: 2844–2853. Cabezas, S., Blas, J., Marchant, T. A., Moreno, S. 2007. Physiological stress levels predict survival probabilities in wild rabbits. Hormones and Behavior 51: 313–320. Cattet, M., Boulanger, J., Stenhouse, G., Powell, R., Reynolds, M. 2008. Long-term effects of capture and handling in ursids: implications for wildlife welfare and research. Journal of Mammalogy 89: 973–990. Cattet, M. R. L., Caulkett, N. A., Obbard, M. E., Stenhouse, G. B. 2002. A body condition index for ursids. Canadian Journal of Zoology 80: 1156–1161. Chow, B. A., Hamilton, J., Alsop, D., Cattet, M. R. L., Stenhouse, G., Vijayan, M. M. 2010. Grizzly bear corticosteroid binding globulin: Cloning and serum protein expression. General and Comparative Endocrinology 167: 317–325. Chow, B. A., Hamilton, J., Cattet, M. R. L., Stenhouse, G., Obbard, M. E., Vijayan, M. M. 2011. Serum corticosteroid binding globulin expression is modulated by fasting in polar bears (Ursus maritimus). Comparative Biochemistry and Physiology, Part A 158: 111–115. Coté, S. D., Festa-Bianchet, M., Fournier, F. 1998. Life-history effects of chemical immobilization and radiocollars on mountain goats. The Journal of Wildlife Management 62: 745–752. Dahle, B., Swenson, J. E. 2003. Seasonal range size in relation to reproductive strategies in brown bears Ursus arctos. Journal of Animal Ecology 72: 660–667. Davenport, M. D., S. Tiefenbacher, C. K. Lutz, M. A. Novak, J. S. Meyer. 2006. Analysis of endogenous cortisol concentrations in the hair of rhesus macaques. General and Comparative Endocrinology 147: 255–261. Dellu, F., Mayo, W., Vallée, M., Maccari, S., Piazza, P. V., Le Moal, M., Simon, H. 1996. Behavioral reactivity to novelty during youth as a predictive factor of stress-induced corticosterone secretion in the elderly—a life-span study in rats. Psychoneuroendocrinology 21: 441–453. du Dot, T. J., Rosen, D. A. S., Richmond, J. P., Kitaysky, A. S., Zinn, S. A., Trites, A. W. 2009. Changes in glucocorticoids, IGF-I and thyroid hormones as indicators of nutritional stress and subsequent refeeding in Steller sea lions (Eumetopias jubatus). Comparative Biochemistry and Physiology, Part A 152: 524–534.

274

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Dumond, M., Boulanger, J., Paetkau, D. 2011. Barren ground grizzly bear remote monitoring in the Canadian Arctic. Poster presentation number 112 at the 20th International Conference on Bear Research & Management, July 17-23, 2011 Ottawa, Ontario, Canada. Ellis, R. D., T. J. McWhorter, M. Maron. 2012. Integrating landscape ecology and conservation physiology. Landscape Ecology 27: 1–12. Erkut, Z. A., Klooker, T., Endert, E., Huitinga, I., Swaab, D. F. 2004. Stress of dying is not suppressed by high-dose morphine or by dementia. Neuropsychopharmacology 29: 152–157. Finlayson, N. B. 1965. Blood cortisol in infants and adults: a postmortem study. Journal of Pediatrics 67: 284–292. Gau, R. J., Case, R., Penner, D. F., McLoughlin, P. D. 2002. Feeding patterns of barren-ground grizzly bears in the central Canadian arctic. Arctic 55: 339–344. Garshelis, D. L., M. L. Gibeau, S. Herrero. 2005. Grizzly bear demographics in and around Banff National Park and Kananaskis Country, Alberta. The Journal of Wildlife Management 69: 277–297. Gladbach, A., Joachim Gladbach, D., Koch, M., Kuchar, A., Möstl, E., Quillfeldt, P. 2011. Can faecal glucocorticoid metabolites be used to monitor body condition in wild Upland geese Chloephaga picta leucoptera? Behavioral Ecology and Sociobiology 65: 1491–1498. Goncharova, N. D., Lapin, B. A. 2002. Effects of aging on hypothalamic–pituitary–adrenal system function in non-human primates. Mechanisms of Ageing and Development 123: 1191– 1201. Gow, R., S. Thomson, M. Rieder, S. Van Uum, G. Koren. 2010. An assessment of cortisol analysis in hair and its clinical applications. Forensic Science International 196: 32–37. Handa, R. J., Burgess, L. H., Kerr, E. R., O’Keefe, J. O. 1994. Gonadal steroid hormone receptors and sex differences in the hypothalamo-pituitary-adrenal axis. Hormones and Behavior 28: 464– 476. Herrero, S. (editor). 2005. Biology, demography, ecology and management of grizzly bears in and around Banff National Park and Kananaskis Country: The final report of the Eastern Slopes Grizzly Bear Project. Faculty of Environmental Design, University of Calgary, Calgary, Alberta. 248 pages. (Available at http://www.canadianrockies.net/grizzly/final_report.html - last accessed on 29 April 2012). Ito, N., Ito, T., Kromminga, A., Betterman, A., Takigawa, M., Keesd, F., Straub, R. H., Paus, R. 2005. Human hair follicles display a functional equivalent of the hypothalamic-pituitary-adrenal axis and synthesize cortisol. FASEB Journal 19: 1332–1334. Jepsen, N., Mikkelsen, J. S., Koed, A. 2008. Effects of tag and suture type on survival and growth of brown trout with surgically implanted telemetry tags in the wild. Journal of Fish Biology 72: 594–602. Johnstone, C. P., R. D. Reina, A. Lill. 2012. Interpreting indices of physiological stress in free-living vertebrates. Journal of Comparative Physiology B: Biochemical, Systemic, and Environmental Physiology (DOI: 10.1007/s00360-012-0656-9). Juster R-P., B. S. McEwen, S. J. Lupien. 2010. Allostatic load biomarkers of chronic stress and impact on health and cognition. Neuroscience and Biobehavioral Reviews 35: 2–16.

275

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Kintz, P. 2004. Value of hair in post-mortem toxicology. Forensic Science International 142: 127– 134. Knapp, C. R., Abarca, J. G. 2009. Effects of radio transmitter burdening on locomotor ability and survival of iguana hatchlings. Herpetologica 65: 363–372. Konkle, A. T. M., A. C. Kentner, S. L. Baker, A. Stewart, C. Bielajew. 2010. Environmentalenrichment-related variations in behavioral, biochemical, and physiologic responses of Sprague–Dawley and Long Evans rats. Journal of the American Association for Laboratory Animal Science 49: 427–436. Koren, L., O. Mokady, T. Karaskov, J. Klein, G. Koren, E. Geffen. 2002. A novel method using hair for determining hormonal levels in wildlife. Animal Behaviour 63: 403–406. Korte, S. M., J. Prins, C. H. Vinkers, B. Olivier. 2009. On the origin of allostasis and stress-induced pathology in farm animals: Celebrating Darwin’s legacy. The Veterinary Journal 182: 378– 383. Kudielka, B. M., S. Wüst. 2010. Human models in acute and chronic stress: Assessing determinants of individual hypothalamus–pituitary–adrenal axis activity and reactivity. Stress 13(1): 1–14. Laber, K., L. M. Veatch, M. F. Lopez, J. K. Mulligan, D. M. R. Lathers. 2008. Effects of housing density on weight gain, immune function, behavior, and plasma corticosterone concentrations in BALB/c and C57BL/6 mice. Journal of the American Association for Laboratory Animal Science 47: 16–23. Léchenne, M. S., Arnemo, J. M., Bröjer, C., Andrén, H., Ågren, E. O. 2011. Mortalities due to constipation and dystocia caused by intraperitoneal radio-transmitters in Eurasian lynx (Lynx lynx). European Journal of Wildlife Research 58: 503–506. Lentini, A. M., Crawshaw, G. J., Licht, L. E., McLelland, D. J. 2011. Pathologic and hematologic responses to surgically implanted transmitters in eastern Massasauga rattlesnakes (Sistrurus catenatus catenatus). Journal of Wildlife Diseases 47: 107–125. Lightman, S. L., Windle, R. J., Wood, S. A., Kershaw, Y. M., Shanks, N., Ingram, C. D. 2001. Peripartum plasticity within the hypothalamo–pituitary–adrenal axis. Progress in Brain Research 133: 111–129. Macbeth, B. J., M. Cattet, G. B. Stenhouse, M. L. Gibeau, D. M. Janz. 2010. Hair cortisol concentration as a non-invasive measure of long-term stress in free-ranging grizzly bears (Ursus arctos): considerations with implications for other wildlife. Canadian Journal of Zoology 88: 935–949. McCormick, C. M., Linkroum, W., Sallinen, B. J., Miller, N. W. 2002. Peripheral and central sex steroids have differential effects on the HPA axis of male and female rats. Stress 5: 235– 247. McEwen, B. S. 2008. Central effects of stress hormones in health and disease: understanding the protective and damaging effects of stress and stress mediators. European Journal of Pharmacology 583: 174–185. McNamara, J. M., Buchanan, K. L. 2005. Stress, resource allocation, and mortality. Behavioral Ecology 16: 1008–1017.

276

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Ortiz, R. M., Wade, C. E., Ortiz, C. L. 2001. Effects of prolonged fasting on plasma cortisol and TH in postweaned northern elephant seal pups. American Journal of Physiology - Regulatory, Integrative and Comparative Physiology 280: R790–R795. Paetkau, D. 2003. Genetical error in DNA-based inventories: insight from reference data and recent projects. Molecular Ecology 12: 1375–1387. Palmiere, C., Mangin, P. 2012. Postmortem chemistry update part II. International Journal of Legal Medicine 126: 199–215. Paus, R., Theoharides, T. C., Arck, P. 2006. Neuroimmunoendocrine circuitry of the ‘brain-skin connection’. Trends in Immunology 27: 32–39. Pollock, K. H., Winterstein, S. R., Bunck, C. M., Curtis, P. D. 1989. Survival analysis in telemetry studies: the staggered entry design. Journal of Wildlife Management 53: 7–15. Pragst, F., M. A. Balikova. 2006. State of the art in hair analysis for detection of drug and alcohol abuse. Clinica Chimica Acta 370: 17–49. Proctor, M. F., McLellan, B. N., Boulanger, J., Apps, C., Mowat, G., Paetkau, D., Stenhouse, G. 2010. Ecological investigations of Canadian grizzly bears using hair-snagged DNA, 1995-2005. Ursus. 21: 169–188. Raouf, S. A., Smith, L. C., Brown, M. B., Wingfield, J. C., Brown, C. R. 2006. Glucocorticoid hormone levels increase with group size and parasite load in cliff swallows. Animal Behaviour 71: 39– 48. Reeder, D. M., Kramer, K. M. 2005. Stress in free-ranging mammals: Integrating physiology, ecology, and natural history. Journal of Mammalogy 86: 225–235. Riondato, F., A. D’Angelo, B. Miniscalco, C. Bellino, R. Guglielmino. 2008. Effects of road transportation on lymphocyte subsets in calves. The Veterinary Journal 175: 364–368. Robbins, C. T., Ben-David, M., Fortin, J. K., Nelson, O. L. 2012. Maternal condition determines birth date and growth of newborn bear cubs. Journal of Mammalogy 93: 540–546. Rode, K. D., Farley, S. D., Robbins, C. T. 2006. Sexual dimorphism, reproductive strategy, and human activities determine resource use by brown bears. Ecology 87: 2636–2646. Romero, L. M. 2002. Seasonal changes in plasma glucocorticoid concentrations in free-living vertebrates. General and Comparative Endocrinology 128: 1–24. Romero, L. M. 2004. Physiological stress in ecology: lessons from biomedical research. Trends in Ecology and Evolution 19: 249–255. Romero, L. M., Wikelski, M. 2001. Corticosterone levels predict survival probabilities of Galápagos marine iguanas during El Niño events. Proceedings of the National Academy of Sciences of the United States of America 98: 7366–7370. Russell, E., Koren, G., Rieder, M., Van Uum, S. 2012. Hair cortisol as a biological marker of chronic stress: Current status, future directions and unanswered questions. Psychoneuroendocrinology 37: 589–601. Sapolsky, R. M., Krey, L., McEwen, B. S. 1983. The adrenocortical stress-response in the aged male rat: impairment of recovery from stress. Experimental Gerontology 18: 55–64.

277

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Saraux, C., Le Bohec, C., Durant, J. M., Viblanc, V. A., Gauthier-Clerc, M., Beaune, D., Park, Y.-H., Yoccoz, N. G., Stenseth, N. C., Le Maho, Y. 2011. Reliability of flipper-banded penguins as indicators of climate change. Nature 469: 203–208. Sharpley, C. F., Kauter, K. G., McFarlane, J. R. 2009. An initial exploration of in vivo hair cortisol responses to a brief pain stressor: Latency, localization and independence effects. Physiological Research 58: 757–761. Sharpley, C. F., Kauter, K. G., McFarlane, J. R. 2010. Diurnal variation in peripheral (hair) vs central (saliva) HPA axis cortisol concentrations. Clinical Medicine Insights: Endocrinology and Diabetes 3: 9–16. Sheriff, M. J., Dantzer, B., Delehanty, B., Palme, R., Boonstra, R. 2011. Measuring stress in wildlife: techniques for quantifying glucocorticoids. Oecologia 166: 869–887. Sheriff, M. J., Wheeler, H., Donker, S. A., Krebs, C. J., Palme, R., Hik, D. S., Boonstra, R. 2012. Mountain-top and valley-bottom experiences: the stress axis as an integrator of environmental variability in arctic ground squirrel populations. Journal of Zoology 287: 65– 75. Smith, F., J. E. Clark, B. L. Overman, C. C. Tozel, J. H. Huang, J. E. F. Rivier, A. T. Blisklager, A. J. Moeser. 2010. Early weaning stress impairs development of mucosal barrier function in the porcine intestine. American Journal of Physiology Gastrointestinal and Liver Physiology 298: G352–G363. Stakėnas, S., Copp, G. H., Scott, D. M. 2009. Tagging effects on three non-native fish species in England (Lepomis gibbosus, Pseudorasbora parva, Sander lucioperca) and of native Salmo trutta. Ecology of Freshwater Fish 18: 167–176. Stenhouse, G. Boulanger, J., Lee, J., Graham, K., Duval, J., Cranston, J. 2005. Grizzly bear associations along the eastern slopes of Alberta. Ursus 16: 31–40. Stenhouse, G. B., Graham, K. (editors). 2011. Foothills Research Institute Grizzly Bear Program 2010 annual report. Foothills Research Institute, Hinton, Alberta. 180 pages. Stern, J. M., Goldman, L., Levine, S. 1973. Pituitary–adrenal responsiveness during lactation in rats. Neuroendocrinology 12: 179–191. Swenson, J. E., Wallin, K., Ericsson, G., Cederlund, G., Sandegren, F. 1999. Effects of ear-tagging with radiotransmitters on survival of moose calves. The Journal of Wildlife Management 63: 354–358. Tinnikov, A. A. 1999. Responses of serum corticosterone and corticosteroid-binding globulin to acute and prolonged stress in the rat. Endocrine 11: 145–150. Tumendemberel, O., Proctor, M., Reynolds, H., Amgalan, L., Tserenbataa, T., Batmunkh, M., Craighead, D., Yanjin, N., Paetkau, D. 2011. Gobi Bear population survey 2009. Final report May 2011, Beringia South Publications, Kelly, Wyoming, USA. 24 pages. Vera, F., Antenucci, C. D., Zenuto, R. R. 2011. Cortisol and corticosterone exhibit different seasonal variation and responses to acute stress and captivity in tuco-tucos (Ctenomys talarum). General and Comparative Endocrinology 170: 550–557. Walker, K. A., Trites, A. W., Haulena, M., Weary, D. M. 2011. A review of the effects of different marking and tagging techniques on marine mammals. Wildlife Research 39: 15–30.

278

Chapter 11-Hair Cortisol as an Indicator of Long-term Stress Webb E., S. Thomson, A. Nelson, C. White, G. Koren, M. Rieder, S. Van Uum. 2010. Assessing individual systemic stress through cortisol analysis of archaeological hair. Journal of Archaeological Science 37: 807–812. Weinstock, M. 2005. The potential influence of maternal stress hormones on development and mental health of the offspring. Brain, Behavior, and Immunity 19: 296–308. Wingfield, J. C., Kitaysky, A. S. 2002. Endocrine responses to unpredictable environmental events: Stress or anti-stress hormones? Integrative and Comparative Biology 42: 600–609. Wingfield, J. C., D. L. Maney, C. W. Breuner, J. D. Jacobs, S. Lynn, M. Ramenofsky, R. D. Richardson. 1998. Ecological bases of hormone–behavior interactions: the ‘‘emergency life history stage”. American Zoologist 38: 191–206. Zedrosser, A., Dahle, B., Swenson, J.E., 2006. Population density and food conditions determine adult female body size in brown bears. Journal of Mammalogy 87: 510–518.

279

APPENDIX 1:LIST OF PROGRAM PARTNERS (1999 – 2010) Ainsworth Lumber Co. Ltd. Alberta Advanced Education and Technology (formerly Innovation and Science) Alberta Conservation Association Alberta Environment Alberta Fish & Game Association AB Innovates (ARC) Alberta Newsprint Company Alberta Sustainable Resource Development Alberta Tourism, Parks and Recreation Anadarko Canada Corporation Anderson Exploration Ltd. AVID Canada B P Canada Energy Company BC Oil and Gas Commission Buchanan Lumber – Tolko OSB Canada Centre for Remote Sensing Canadian Association of Petroleum Producers (CAPP) Petroleum Technology AllianceCanada (PTAC) Environmental Research Advisory Council (ERAC) Fund Alberta Upstream Petroleum Research Fund Canadian Natural Resources Ltd. Canfor Corporation Center for Wildlife Conservation ConocoPhillips Canada (formerly Burlington Resources Canada Ltd.) (formerly Canadian Hunter Exploration Ltd.) Conservation Biology Institute Daishowa Marubeni International Ltd. Devon Canada Corp Enbridge Inc. Encana Corporation ENFORM

Foothills Research Institute (formerly FoothillsModelForest) Forest Resources Improvement Association of Alberta (FRIAA) G&A Petroleum Services GeoAnalytic Inc. Government of Canada CanadianForest Service, Natural Resources Canada Canadian Wildlife Service Environment Canada – HSP Human Resources and Skills Development Canada Natural Sciences and Engineering Research Council of Canada (NSERC) Parks Canada BanffNational Park JasperNational Park Grande Cache Coal Corporation Hinton Fish and Game Association Hinton Training Centre Husky Energy Inc. Komex International Ltd. Lehigh Inland Cement Limited Luscar Ltd. Gregg River Resources Ltd. Manning Diversified Forest Products Ltd. Manning Forestry Research Fund Millar Western Forest Products Ltd. Mountain Equipment Co-op Nature Conservancy NatureServe Canada Nexen Inc. Northrock Resources Ltd. Peregrine Helicopters

Persta Petro Canada Ltd. Peyto Energy Trust Precision Drilling Corporation Rocky Mountain Elk Foundation - Canada Shell Canada Limited Sherritt – Coal Valley Resources Slave LakeDivision – Alberta Plywood SprayLake Sawmills Ltd. Suncor Energy Inc. Sundance Forest Industries Ltd. Talisman Energy Inc. TECH – CardinalRiver Operations (formerly Elk Valley Coal) Telemetry Solutions TransCanada Pipelines Ltd. University of Alberta University of Calgary University of Lethbridge University of Saskatchewan WesternCollege of Veterinary Medicine University of Washington University of Waterloo Veritas DGC Inc. West Fraser Mills Ltd. Alberta Plywood Blue Ridge Lumber Inc. Hinton Wood Products SlaveLake Pulp SundreForest Products Weyerhaeuser Company Limited WilfredLaurierUniversity World Wildlife Fund Canada Yellowstone to Yukon

280

APPENDIX 1 – List of 2011 Program Partners (1999-2011)

APPENDIX 2: LIST OF PROGRAM PARTNERS (2011) Alberta Conservation Association Alberta Ecotrust Foundation Alberta Newsprint Company Alberta Upstream Petroleum Research Fund Canadian Cooperative Wildlife Health Center Coalspur Mines Ltd. ConocoPhillips Canada Devon Canada Corp. Daishowa-Marubeni International Ltd. Encana Corp. Environment Canada Foothills Research Institute Grande Cache Coal Corp. Government of Alberta Alberta Tourism, Parks and Recreation Alberta Sustainable Resource Development Alberta Employment and Immigration – Summer Temporary Employment Program AB Innovates - Biosolutions Government of Canada Natural Resources Canada Human Resources and Skills Development Canada – Canada Summer Jobs Parks Canada Husky Energy Inc. Nature Conservancy Shell Canada Ltd. Sherritt Corp. – Coal Valley Resources Inc. Suncor Energy Inc. Sundance Forest Industries Ltd. Talisman Energy Inc. TransCanada Pipelines Ltd. University of Alberta University of British Columbia University of Calgary University of Saskatchewan University of Victoria University of Waterloo Weyerhaeuser Company Ltd. Yellowstone to Yukon

281

APPENDIX 1 â&#x20AC;&#x201C; List of 2011 Program Partners (1999-2011)

APPENDIX 3: LIST OF REPORTS, THESES AND PUBLISHED PAPERS Alberta Grizzly Bear Inventory Team. 2007. Grizzly bear population and density estimates for the 2006 Alberta Unit 5 Management Area inventory. Report prepared for Alberta Sustainable Resource Development, Fish and Wildlife Division. 37 pp. Alberta Grizzly Bear Inventory Team. 2009. Grizzly bear population and density estimates for the 2008 DNA inventory of the Grande Cache Bear Management Area (BMA 2). Report prepared for Alberta Sustainable Resource Development, Fish and Wildlife Division. 39 pp. Bater, C. W., N. C. Coops, M. A. Wulder, T. Hiker, S. E. Nielsen, G. McDermid, and G. B. Stenhouse. 2011. Using digital time-lapse cameras to monitor species-specific understorey and overstorey phenology in support of wildlife habitat assessment. Environmental Monitoring and Assessment. 180:1-13. Bater, C. W., N. C. Coops, M. A. Wulder, S. E. Nielsen, G. McDermid, and G. B. Stenhouse. 2011. Design and installation of a camera network across an elevation gradient for habitat assessment. Instrumentation Science and Technology 39:231-247. Berland, A., Nelson T., G. Stenhouse, K. Graham, and J. Cranston. 2008. The impact of landscape disturbance on grizzly bear habitat use in the Foothills Model Forest, Alberta Canada. Forest Ecology and Management 256::1875-1883. Boulanger, J., M. Proctor, S. Himmer, G. Stenhouse, D. Paetkau, and J. Cranston. 2006. An empirical test of DNA mark-recapture sampling strategies for grizzly bears. Ursus 17:149-158. Boulanger, J., G. Stenhouse, G. MacHutchon, M. Proctor, S. Himmer, D. Paetkau, and J. Cranston. 2005. Grizzly bear population density estimates for the 2005 Alberta (proposed) Unit 4 Management Area inventory. Report prepared for Alberta Sustainable Resource Development, Fish and Wildlife Division. 31 pp. Boulanger, J., G. Stenhouse, and R. Munro. 2004. Sources of heterogeneity bias when DNA mark-recapture sampling methods are applied to grizzly bear (Ursus arctos) populations. Journal of Mammalogy 85:618-624. Boulanger, J., G. Stenhouse, M. Proctor, S. Himmer, D. Paetkau, and J. Cranston. 2005. 2004 population inventory and density estimates for the Alberta 3B and 4 B grizzly bear management area. Report prepared for Alberta Sustainable Resources Development, Fish and Wildlife Division. 28 pp. Boulanger, J., G. C. White, M. Proctor, G. Stenhouse, G. Machutchon, and S. Himmer. 2008. Use of occupancy models to estimate the influence of previous live captures on DNA-based detection probabilities of grizzly bears. Journal of Wildlife Management 72:589-595. Boyce, M. S., S. E. Nielsen, and G. B. Stenhouse. 2009. Grizzly Bears Benefit from Forestry Except for the Roads. BC Forest Professional January-February:19. Boyce, M. S., J. Pitt, J. M. Northup, A. T. Morehouse, K. H. Knopff, B. Cristescu , and G. B. Stenhouse. 2010. Temporal autocorrelation functions for movement rates from global 282

APPENDIX 1 â&#x20AC;&#x201C; List of 2011 Program Partners (1999-2011) positioning system radiotelemetry data. Philosophical Transactions of The Royal Society B: Biological Sciences 365:2213-2219. Carra, B. L. 2010. Spatial and spatial-temporal analysis of grizzly bear movement patterns as related to underlying landscapes across multiple scales. PhD Thesis, Department of Geography, Wilfrid Laurier University, Waterloo Ontario, Canada. Cattet, M., J. Boulanger, G. Stenhouse, R. A. Powell, and M. J. Reynolds-Hogland. 2008. An evaluation of long-term capture effects in Ursids: Implications for wildlife welfare and research. Journal of Mammology 89::973-990. Cattet, M., G. Stenhouse, and T. Bollinger. 2008. Exertional myopathy in a grizzly bear (Ursus arctos) captured by leghold snare. Journal of Wildlife Diseases 44:973-978. Cattet, M. R., A. Bourque, B. T. Elkin, K. D. Powley, D. B. Dahlstrom, and N. A. Caulkett. 2006. Evaluation of the potential for injury with remote drug-delivery systems. Wildlife Society Bulletin 34:741-749. Cattet, M. R. L., N. A. Caulkett, M. E. Obbard, and G. B. Stenhouse. 2002. A body-condition index for ursids. Canadian Journal of Zoology 80:1156-1161. Cattet, M. R. L., N. A. Caulkett, and G. B. Stenhouse. 2003. Anesthesia of grizzly bears using xylazine-zolazepam-tiletamine or zolazepam-tiletamine. Ursus 14 :88-93. Cattet, M. R. L., D. Christison, N. A. Caulkett, and G. B. Stenhouse. 2003. Physiologic response of grizzly bears to different methods of capture. Journal of Wildlife Diseases 39:649-654. Chow, B. A., J. Hamilton, M. R. L. Cattet, G. Stenhouse, M. E. Obbard, and M. M. Vijayan. 2011. Serum corticosteroid binding globulin expression is modulated by fasting in polar bears (Ursus maritimus). Comparative Biochemistry and Physiology, Part A 158:111-115. Chow, B. A., J. W. Hamilton, D. Alsop, M. Cattet, G. Stenhouse, and M. M. Vijayan. 2010. Grizzly bear corticosteroid binding globulin: cloning and serum protein expression. General and Comparative Endocrinology 167:317-325. Collingwood, A. 2008. Satellite image classification and spatial analysis of agricultural areas for land cover mapping of grizzly bear habitat. MSc. Thesis. Department of Geography, University of Saskatchewan, Saskatoon, Saskatchewan. Collingwood, A., S. E. Franklin, X. Guo, and G. Stenhouse. 2009. A medium-resolution remote sensing classification of agricultural areas in Alberta grizzly bear habitat. Canadian Journal of Remote Sensing 35:23-26. Coogan, S. 2012. Getting to the root of the matter: grizzly bears and alpine sweetvetch in westcentral Alberta, Canada. MSc. Thesis, University of Alberta, Edmonton, Alberta, Canada. 98 pp. Coops, N. C., T. Hilker, C. W. Bater, M. A. Wulder, S. E. Nielsen, G. McDermid, and G. Stenhouse. 2012. Linking ground-based to satellite-derived phenological metrics in support of habitat assessment. Remote Sensing Letters 3:191-200. Cristescu, B., G. B. Stenhouse, M. Symbaluk, and M. S. Boyce. 2011. Land-use planning following resource extraction - lessons from grizzly bears at reclaimed and active open pit mines. 283

APPENDIX 1 â&#x20AC;&#x201C; List of 2011 Program Partners (1999-2011) Mine Closure 2011 2:207-218. Curry, B. L. 2008. A comparison of relative and absolute change detection for measuring forest disturbance. MGIS Thesis, Department of Geography, University of Calgary. Frair, J. L., S. E. Nielsen, E. H. Merrill , S. R. Lele, R. S. Boyce, R. H. M. Munro, G. B. Stenhouse, and H. L. Beyer. 2004. Removing GPS collar bias in habitat selection studies. Journal of Applied Ecology 41:201-212. Franklin, S. E. 2009. Remote Sensing for Biodiversity and Wildlife Management: Synthesis and Applications. McGraw-Hill Professional, New York346. Franklin, S. E., Y. He, A. D. Pape, X. Guo, and G. J. McDermid. 2011. Landsat-comparable land cover maps using ASTER and SPOT images: a case study for large-area mapping programmes. International Journal of Remote Sensing 32:2185-2205. Franklin, S. E., P. K. Montgomery, and G. B. Stenhouse. 2005. Interpretation of land cover changes using aerial photography and satellite imagery in the Foothills Model Forest of Alberta. Canadian Journal Remote Sensing 31:304-313. Franklin, S. E., D. R. Peddle, J. A. Dechka, and G. B. Stenhouse. 2002. Evidential reasoning with Landsat TM, DEM and GIS data for landcover classification in support of grizzly bear habitat mapping. International Journal of Remote Sensing 23:4633-4652. Franklin, S. E., G. B. Stenhouse, M. J. Hansen, C. C. Popplewell, J. A. Dechka, and D. R. Peddle. 2001. An integrated decision tree approach (IDTA) to mapping landcover using satellite remote sensing in support of grizzly bear habitat analysis in the Alberta Yellowhead Ecosystem. Canadian Journal of Remote Sensing 27:579-592. Franklin, S. E. M. B. L. M. A. W. G. B. S. 2002. Change detection and landscape structure mapping using remote sensing. The Forestry Chronicle 78:618-625. Franklin, S. E. M. J. H. G. B. S. 2002. Quantifying landscape structure with vegetation inventory maps and remote sensing. The Forestry Chronicle 78(6):866-875. Gau, R. J., R. Mulders, L. M. Ciarniello, D. C. Heard, C. B. Chetkiewicz, M. Boyce , R. Munro, G. Stenhouse, B. Chruszcz, M. L. Gibeau, B. Milakovic, and K. Parker . 2004. Uncontrolled field performance of Televilt GPS-Simplex collars on grizzly bears in western and northern Canada. Wildlife Society Bulletin 32:693-701. Gaulton, R., T. Hilker, M. A. Wulder, N. C. Coops, and G. Stenhouse. 2011. Characterizing standreplacing disturbance in western Alberta grizzly bear habitat, using a satellite-derived high temporal and spatial resolution change sequence. Forest Ecology and Management 261:865-877. Graham, K., J. Boulanger, J. Duval, and G. Stenhouse. 2010. Spatial and temporal use of roads by grizzly bears in west-central Alberta. Ursus 21:43-56. Grizzly Bear Inventory Team. 2008. Grizzly bear population and density estimates for Alberta Bear Management Unit 6 and British Columbia Units 4-1, 4-2, and 4-3 (2007). Report prepared for the Alberta Sustainable Resource Development, Fish and Wildlife Division, British Columbia Ministry of Forests and Range, British Columbia Ministry of Environment, and Parks Canada. 46 pp. 284

APPENDIX 1 â&#x20AC;&#x201C; List of 2011 Program Partners (1999-2011) Hamilton, J. W. 2007. Evaluation of indicators of stress in populations of polar bears (Ursus maritimus) and grizzly bears (Ursus arctos). M.Sc. Thesis. Department of Biology, University of Waterloo, Waterloo, Ontario. He, Y., S. E. Franklin, X. Guo, and G. B. Stenhouse. 2009. Narrow-Linear and small-area forest disturbance detection and mapping from high spatial resolution imagery. Journal of Applied Remote Sensing 3:1-20. He, Y., S. E. Franklin, X. Guo, and G. B. Stenhouse. 2010. Object-oriented classification of multiresolution images for the extraction of narrow linear forest disturbance. Remote Sensing Letters 2:147-155. Hilker, T., N. Coops, R. Gaulton, M. Wulder, J. Cranston, and G. Stenhouse. 2011. Biweekly disturbance capture and attribution: case study in western Alberta grizzly bear habitat. Journal of Applied Remote Sensing 5, 053568 (2011), DOI:10.1117/1.3664342. Hilker, T., M. A. Wulder, N. C. Coops, N. Seitz, J. C. White, F. Gao, J. Masek, and G. B. Stenhouse. 2009. Generation of dense time series synthetic Landsat data through data blending with MODIS using the spatial and temporal adaptive reflectance fusion model (STARFM). Remote Sensing of Environment 113:1988-1999. Hilker, T. W. M. A., N. C. Coops, J. Linke , G. McDermid, J. Masek, F. Gao, and J. C. White. 2009. A new data fusion model for high spatial- and temporal- resolution mapping of forest disturbance based on Landsat and MODIS. Remote Sensing of Environment 13::16131627. Hird, J. and G. J. McDermid. 2009. Noise reduction of NDVI time series: An empirical comparison of selected techniques. Remote Sensing of Environment 113: 248-258. Hird, J. N. 2008. Noise reduction for time series of the Normalized Difference Vegetation Index (NDVI): An empirical comparison of selected techniques. MSc Thesis, Department of Geography, University of Calgary. Huettmann, F., S. E. Franklin, and G. B. Stenhouse. 2005. Predictive spatial modelling of landscape change in the Foothills Model Forest. The Forestry Chronicle 81:525-537. Hunter, A. 2007. Sensory Based Animal Tracking. PhD Thesis. Department of Geomatics Engineering, University of Calgary, Calgary, Alberta, Canada. Lindsjo, H. J. A. 2009. Development and application of a health function score system for grizzly bears (Ursus arctos) in western Alberta. MSc. Thesis. Department of Veterinary Pathology, University of Saskatchewan, Saskatoon, Saskatchewan. Linke, J. 2010. Development and application of a framework for flexible and reliable landscape monitoring: changes in the Alberta foothills and the impact on grizzly bear habitat. PhD Thesis, Department of Geography, University of Calgary, Calgary, Alberta, Canada. 163 pp. Linke, J. 2003. Using Landsat TM and IRS imagery to Detect Seismic Cutlines: Assessing their Effects on Landscape Structure and on Grizzly Bear (Ursus arctos) Landscape Use in Alberta. MSc. Thesis. Department of Geography, University of Calgary, Calgary, Alberta, Canada. 285

APPENDIX 1 â&#x20AC;&#x201C; List of 2011 Program Partners (1999-2011) Linke, J., M. Betts, M. B. Lavigne, and S. E. Franklin. 2007. Introduction: Structure, function and change of forest landscapes. Pages 1-30 in M. A. Wulder and S. E. Franklin, eds. Understanding Forest Disturbance and Spatial Pattern: Remote Sensing and GIS Approaches. CRC Press, Boca Raton. Linke, J., S. E. Franklin, M. Hall-Beyer, and G. B. Stenhouse. 2008. Effects of cutline density and land-cover heterogeneity on landscape metrics in western Alberta. Canadian Journal of Remote Sensing 34:390-404. Linke, J., S. E. Franklin, F. Huettmann, and G. B. Stenhouse. 2005. Seismic cutlines, changing landscape metrics and grizzly bear landscape use in Alberta. Landscape Ecology 20:811826. Linke, J. and G. J. McDermid. 2011. A conceptual model for multi-temporal landscape monitoring in an object-based environment. Journal of Selected Topics in Earth Observations and Remote Sensing 4:265-271. Linke, J., G. J. McDermid, D. N. Laskin, A. J. McLane, A. D. Pape, J. Cranston, M. Hall-Beyer, and S.E. Franklin. 2009. A disturbance-inventory framework for flexible and reliable landscape monitoring. Photogrammetric Engineering and Remote Sensing 75:981-995. Linke, J., G. J. McDermid, A. D. Pape, A. J. McLane, D. N. Laskin, M. Hall-Beyer, and S. E. Franklin. 2009. The influence of patch delineation mismatches on multitemporal landscape pattern analysis. Landscape Ecology 24: 157-170. Macbeth, B. J., M. R. L. Cattet, G. B. Stenhouse, M. L. Gibeau, and D. M. Janz. 2010. Hair cortisol concentration as a noninvasive measure of long-term stress in free-ranging grizzly bears (Ursus arctos): considerations with implications for other wildlife. Canadian Journal of Zoology 88:935-943. Mace, R., D. W. Carney, T. Chilton-Radandt, S. A. Courville, M. A. Haroldson, R. B. Harris, J. Jonkel, B. McLellan, M. Madel, T. L. Manley, C. C. Schwartz, C. Servheen, G. Stenhouse, J. S. Wallyer, and E. Wenum. 2011. Grizzly bear population vital rates and trend in the Northern Continental Divide Ecosystem, Montana. Journal of Wildlife Management 9999:1-10. McDermid, G. J. 2005. Remote Sensing for Large-Area, Multi-Jurisdictional Habitat Mapping. PhD Thesis. Department of Geography, University of Waterloo, Waterloo, Ontario, Canada. McDermid, G. J., F. E. Frankling, and E. F. LeDrew. 2005. Remote sensing for large-area habitat mapping. Progress in Physical Geography 29:449-474. McDermid, G. J., R. J. Hall, G. A. Sanchez-Azofeifa, S. E. Franklin, G. B. Stenhouse, T. Kobliuk, and E. F. LeDrew. 2009. Remote sensing and forest inventory for wildlife habitat assessment. Forest Ecology and Management 257:2262-2269. McDermid, G. J., J. Linke, A. D. Pape, D. N. Laskin, A. J. McLane, and S. E. Franklin. 2008. Objectbased approaches to change analysis and thematic map update: challenges and limitations. Canadian Journal of Remote Sensing 34:462-466. McDermid, G. J., N.C. Coops, M.A. Wulder, S.E. Franklin, and N. Seitz. 2009. Critical Remote 286

APPENDIX 1 – List of 2011 Program Partners (1999-2011) Sensing Data Contributions to Spatial Wildlife Ecological Knowledge . Pages 193-221 in F. Huettmann and S. Cushman, eds. Spatial Complexity, Informatics, and Wildlife Conservation. Springer, Tokyo. McLane, A. J. 2007. Processing discrete-return profiling LiDAR data to estimate canopy closure for the purpose of grizzly bear management in northern Alberta. MGIS Thesis, Department of Geography, University of Calgary. McLane, A. J., G. J. McDermid, and M. A. Wulder. 2009. Processing discrete-return profiling LiDAR data to estimate canopy closure for large-area forest mapping and management. Canadian Journal of Remote Sensing 35:217-229. Mowat, G., D. C. Heard, D. R. Seip, K. G. Poole, G. Stenhouse, and D. W. Paetkau. 2005. Grizzly Ursus arctos and black bear U. americanus densities in the interior mountains of North America. Wildlife Biology 11: 31-48. Munro, R. H. M., S. E. Nielsen, M. H. Price, G. B. Stenhouse, and M. S. Boyce. 2006. Seasonal and diel patterns of grizzly bear diet and activity in west-central Alberta. Journal of Mammalogy 87:1112-1121. Nielsen, S. E. 2004. Habitat ecology, Conservation, and Projected Population Viability of Grizzly Bears (Ursus arctos L.) in West-Central Alberta, Canada. PhD Thesis. Department of Biological Sciences, University of Alberta, Edmonton, Alberta, Canada. Nielsen, S. E., M. S. Boyce, and G. B. Stenhouse. 2004. Grizzly bears and forestry I. Selection of clearcuts by grizzly bears in west-central Alberta, Canada. Forest Ecology and Management 199:51–65. Nielsen, S. E., M. S. Boyce, G. B. Stenhouse, and R. H. M. Munro. 2003. Development and testing of phenologically driven grizzly bear habitat models. Ecoscience 10:1-10. Nielsen, S. E., M. S. Boyce, G. B. Stenhouse, and R. H. M. Munro. 2002. Modeling grizzly bear habitats in the Yellowhead ecosystem of Alberta: Taking autocorrelation seriously. Ursus 13:45-56. Nielsen, S. E., J. Cranston, and G. B. Stenhouse. 2009. Identification of priority areas for grizzly bear conservation and recovery in Alberta, Canada. Journal of Conservation Planning 5:38-60. Nielsen, S. E., G. McDermid, G. B. Stenhouse, and M. S. Boyce. 2010. Dynamic wildlife habitat models: Seasonal foods and mortality risk predict occupancy-abundance and habitat selection in grizzly bears. Biological Conservation 143:1623-1634. Nielsen, S. E., R. H. M. Munro, E. L. Bainbridge, G. B. Stenhouse, and M. S. Boyce. 2004. Grizzly bears and forestry II. Distribution of grizzly bear foods in the clearcuts of west-central Alberta, Canada. Forest Ecology and Management 199:67–82. Nielsen, S. E., G. B. Stenhouse, H. L. Beyer, F. Huettmann, and M. S. Boyce. 2008. Can natural disturbance-based forestry rescue a declining population of grizzly bears? Biological Conservation 141:2193-2207. Nielsen, S. E., G. B. Stenhouse, and M. S. Boyce. 2006. A habitat-based framework for grizzly bear conservation in Alberta. Biological Conservation 130:217-229. 287

APPENDIX 1 â&#x20AC;&#x201C; List of 2011 Program Partners (1999-2011) Northrup, J. M. 2010. Grizzly bears, roads and human-bear conflicts in southwestern Alberta. MSc. Thesis. Department of Biological Sciences, University of Alberta, Edmonton, Alberta, Canada. Northrup, J. M. and G. B. B. M. S. Stenhouse. 2012. Agricultural lands as ecological traps for grizzly bears. Animal Conservation doi: 10.1111/j.1469-1795.2012.00525.x. Pape, A. 2006. Multiple Spatial Resolution Image Change Detection for Environmental Management Applications. MSc. Thesis. Department of Geography, University of Saskatchewan, Saskatoon, Saskatchewan, Canada. Pape, A. D. and S. E. Franklin. 2008. MODIS-based change detection for grizzly bear habitat mapping in Alberta. Photogrammetric Engineering and Remote Sensing 74:973-985. Pereverzoff, J. L. 2003. Development of a Rapid Assessment Technique to Identify Human Disturbance Features in a Forested Landscape. MGIS Thesis, Department of Geography, University of Calgary, Calgary, Alberta, Canada. Popplewell, C. 2001. Habitat Structure and Fragmentation of Grizzly Bear Management Units and Home Ranges in the Alberta Yellowhead Ecosystem. MSc. Thesis. Department of Biological Sciences, University of Calgary, Calgary, Alberta, Canada. Popplewell, C., S. E. Franklin, G. B. Stenhouse, and M. Hall-Beyer. 2003. Using landscape structure to classify grizzly bear density in Alberta Yellowhead Ecosystem bear management units. Ursus 14:27-34. Proctor, M., B. McLellan, J. Boulanger, C. Apps, G. Stenhouse, D. Paetkau, and G. Mowat. 2010. Ecological investigations of grizzly bears in Canada using DNA from hair, 1995-2005: a review of methods and progress. Ursus 21:169-188. Proctor, M. F., D. Paetkau, B. N. McLellan , G. B. Stenhouse, K. C. Kendall, R. D. Mace, W. F. Kasworm, C. Servheen, C. L. Lausen, M. L. Gibeau, W. L. Wakkinen, M. A. Haroldson, G. Mowat, C. D. Apps, L. M. Ciarniello, R. M. R. Barclay, M. S. Boyce, C. C. Schwartz, and C. Strobeck. 2012. Population fragmentation and inter-ecosystem movements of grizzly bears in western Canada and the northern United States. Wildlife Monographs 180:146. Ram, A. E. 2011. Changes in grizzly bear habitat due to human disturbance in the Rocky Mountain Foothills of Alberta from 1985-2005. MGIS Thesis, Department of Geography, University of Calgary, Calgary, Alberta. Ritson-Bennett, R. A. 2003. Assessing the Effects of a Heli-portable 3D Seismic Survey on Grizzly Bear (Ursus arctos horribilis) Distribution. MSc. Thesis, Department of Geography, University of Calgary, Calgary, Alberta, Canada. Roever, C. L. 2006. Grizzly Bear (Ursus arctos L.) Selection of Roaded Habitats in a Multi-use Landscape. MSc. Thesis, University of Alberta, Edmonton, Alberta, Canada. Roever, C. L., M. S. Boyce, and G. B. Stenhouse. 2010. Grizzly bear movements relative to roads: application of step selection functions. Ecography 33:1-10. Roever, C. L., M. S. Boyce, and G. B. Stenhouse. 2008. Grizzly bears and forestry I: Road vegetation and placement as an attractant to grizzly bears. Forest Ecology and 288

APPENDIX 1 â&#x20AC;&#x201C; List of 2011 Program Partners (1999-2011) Management 256:1253-1261. Roever, C. L., M. S. Boyce, and G. B. Stenhouse. 2008. Grizzly bears and forestry II: Grizzly bear habitat selection and conflicts with road placement. Forest Ecology and Management 256:1253-1261. Sahlen, E. 2010. Do grizzly bears use or avoid wellsites in west-central Alberta, Canada? MSc Thesis, Swedish University of Agricultural Sciences, Faculty of Foresty, Department of Wildlife, Fish and Environmental Studies. Sanii, S. 2008. Assessing the effect of point density and terrain complexity on the quality of LiDAR-derived DEMs in multiple resolutions. MGIS Thesis, Department of Geography, University of Calgary. Schwab, B. L. 2003. Graph Theoretic Methods for Examining Landscape Connectivity and Spatial Movement Patterns: Application to the FMF Grizzly Bear Research. MSc Thesis. Department of Geography, University of Calgary, Calgary AB. Schwab, C., B. Cristescu, M. S. Boyce, G. B. Stenhouse, and M. Ganzle. 2009. Bacterial populations and metabolites in the feces of free roaming and captive grizzly bears. Canadian Journal of Microbiology 55:1335-1346. Schwab C., Cristescu B., Northrup J.M., Stenhouse G.B., and Ganzle M. 2011. Diet and environment shape fecal bacterial microbiota composition and enteric pathogen load of grizzly bears. PLoS ONE 6:e27905. doi:10.1371/journal.pone.0027905. Smulders, M., T. A. Nelson, D. E. Jelinski , S. E. Nielsen, and G. B. Stenhouse. 2010. A spatially explicit method for evaluating accuracy of species distribution models. Diversity and Distributions 16:996-1008. Smulders, M. C. A. 2006. Spatial-Temporal Analysis of Grizzly Bear Habitat Use. MSc. Thesis. Department of Geography, University of Victoria, Victoria, British Columbia. Sobol, A. 2009. The impact of sampling intensity and failed GPS acquisition on animal home range estimation. MSc. Thesis, Department of Geography, University of Calgary, Alberta. Stenhouse, G., J. Boulanger, J. Lee, K. Graham, J. Duval, and J. Cranston. 2005. Grizzly bear associations along the eastern slopes of Alberta. Ursus 16 :31-40. Stenhouse, G. B., J. Boulanger, M. Cattet, S. Franklin, D. Janz, G. McDermid, S. Nielsen, and M. Vijayan. 2008. New Tools to Map, Understand, and Track Landscape Change and Animal Health for Effective Management and Conservation of Species at Risk. Final Report for Alberta Advanced Education and Technology and Alberta Sustainable Resource Development. 391 pp. Stewart, B. P., T. A. Nelson, M. A. Wulder , S. Nielsen, and G. Stenhouse. 2012. Impact of disturbance characteristics and age on grizzly bear habitat selection. Applied Geography In Press. Stewart, B. P., M. A. Wulder, G. J. McDermid, and T. Nelson. 2009. Disturbance capture and attribution through the integration of Landsat and IRS-1C imagery. Canadian Journal of Remote Sensing 35:523-533. 289

APPENDIX 1 â&#x20AC;&#x201C; List of 2011 Program Partners (1999-2011) Timmins, T. 2010. Modelling Spatial Dependence in Multivariate Regression Models of Grizzly Bear Health in Alberta, Canada. M.Sc. Thesis, Department of Geomatics Engineering, University of Calgary, Calgary Alberta. Wang, K., S. E. Franklin, X. Guo, and M. Cattet. 2010. Remote sensing of ecology, biodiversity and conservation: a review from the perspective of remote sensing specialists. Sensors 10:9647-9667. Wang, K., S. E. Franklin, X. Guo, A. Collingwood, G. B. Stenhouse, and S. Lowe. 2010. Comparison of Landsat multispectral and IRS panchromatic imagery for landscape pattern analysis of grizzly bear habitat in agricultural areas of western Alberta. Canadian Journal of Remote Sensing 36:36-47. Wang, K., S. E. Franklin, X. Guo, Y. He, and G. J. McDermid. 2009. Problems in remote sensing of landscapes and habitats. Progress in Physical Geography 33:747-768. Wasser, S. K., B. Davenport, E. R. Ramage, K. E. Hunt, M. Parker, C. Clarke, and G. Stenhouse. 2004. Scat detection dogs in wildlife research and management: application to grizzly and black bears in the Yellowhead Ecosystem, Alberta, Canada. Canadian Journal Zoology 82:475-492. White, J. C., M. A. Wulder, C. Gomez, and G. Stenhouse. 2011. A history of habitat dynamics: characterizing 35 years of stand replacing disturbance. Canadian Journal of Remote Sensing 37:234-251. Wulder, M. A. and S. E. Franklin. 2007. Conclusion: Understanding forest disturbance and spatial pattern, information needs, and new approaches. Pages 233-243 in M. A. Wulder and S. E. Franklin, eds. Understanding Forest Disturbance and Spatial Pattern: Remote Sensing and GIS Approaches. CRC Press, Boca Raton . Wulder, M. A., B. P. Stewart, M. E. Andrew, M. Smulders, T. Nelson, N. C. Coops, and G. B. Stenhouse. 2009. Remote sensing derived edge location, magnitude, and class transitions for ecological studies. Canadian Journal of Remote Sensing 35:509-522. Wunderle, A. L. 2006. Sensitivity of multi-resolution satellite sensor imagery to regenerating forest age and site preparation for wildlife habitat analysis. M.Sc. Thesis, Department of Geography, University of Saskatchewan. Wunderle, A. L., S. E. Franklin, and X. Guo. 2006. Regenerating boreal forest structure estimation using SPOT-5 pansharpened imagery. International Journal of Remote Sensing 28:4351-4364. Yalte, K. 2007. Relative change detection: an overlooked advantage. MGIS Thesis, Department of Geography, University of Calgary.

290