Competing resource selection modeling predicts risk for preventing and mitigating impacts to flying birds from industrial wind energy development Tricia Miller, Robert Brooks, Michael Lanzone, David Brandes, Jeff Cooper, Kieran O’Malley, Adam Duerr, Todd Katzner
There are direct effects on wildlife from wind energy development, but the costs are not equal among facilities, seasons or species. •Site to site variation •Within site variation •Seasonal and other temporal variation Photo: Alameda County
Indirect effects of development are difficult to see and measure, but may be increasingly important as more facilities come online. • Habitat loss • Habitat fragmentation • Displacement
Photo: John Terry
Small population of Golden Eagles migrates in a narrow band through the central Appalachian Mountains United States
Katzner et al. 2012
Canada
Wind power development is located in the migratory path of Golden Eagles
2012 - 41 existing or planned facilities with over 1191 turbines
Objective: Develop a tool to reduce direct and indirect effects of wind power development on Golden Eagles
To reach our objective: 1. Created a spatial model of habitat selection by low flying eagles
M. Lanzone
2. Created spatial model predicting “habitat� selection for wind turbines
Photo: John Terry
3. Combined these to create a risk map
Photo: John Terry, http://vawind.org/
Study area consists of three topographically distinct regions Northern Plateau
Allegheny Mountains
Ridge & Valley
Generalized estimating equation to model resource selection for eagles and wind turbines Elevation
“Good� Winds
Northness Eastness Updraft Potential
Data sources: USGS, Nat. Renewable Energy Labs, Ecological Land Units (TNC) Updraft: Brandes & Ombalski 2004
Side Slopes Steep Slopes Summit
Habitat quality for wind power
Overlap = Risk
Habitat quality for eagles
Miller et al. 2014 Conservation Biology
Ranked risk by comparing habitat quality
Risk
Low Value
Eagle Habitat Quality
Poor
Low Risk Poor Mod – Fair – Ext Risk Excellent Moderate Risk Fair High Risk Good Extreme Risk Excellent
Wind Power Habitat Quality Poor Fair Excellent Poor Fair Excellent Fair Excellent Fair Excellent
Eagles used more high risk habitat in the Ridge & Valley NP
AM
RV
Every eagle in the Ridge & Valley encountered highest risk areas at least once during migration NP
AM
RV
Most turbines in the Allegheny Mountains are low risk NP
AM RV
Can apply the model to wind facilities and individual turbines
Identify turbine risk
But there are suitable areas for wind power where there is NOT overlap.
Risk Low Value Low Risk Mod - Extreme Moderate Risk High Risk Extreme Risk
We can use our models for conservation by making recommendations for siting turbines in low risk areas
We identified regional variation in risk Guide regional level management decisions •Restrict development •Increase pre-/postconstruction monitoring •Additional requirements to reduce risk
We identified project level risk Use to guide management decisions •Restrict development •Require mitigation for existing projects •Use with turbine level information prioritize monitoring •Make site and/or turbine level adjustments
Applied framework to understand seasonal variation in risk
We will apply this framework to other regions and species to mitigate risk from wind turbines
Funding: Penn State Ecology Penn State Earth and Environmental Institute Fellowship State Wildlife Grant Lindbergh Foundation Hydro Quebec Quebec MNRF Grant PA Wild Resources Conservation Fund Virginia Department of Game & Inland Fisheries Grant
Field assistants: Jeanette Parker, Dave Kramer, Eric Frank, Frank Nicoletti, B. Baillargeon, & Philipe BeauprĂŠ
Thank you!
Regional variation in risk – Ridge & Valley has NP
AM
RV
Eagles are present in the eastern US during fall, spring & winter
There may be greater risk during fall and winter when eagles fly at lower altitudes n = 48 (79,315)
n = 19 (25,778)
n = 43 (28,883)
Calculated logistic GEEs for eagles and wind developments in each region using 75% of data Left out 25% of data for model validation Repeated measures : Individual eagles Wind turbine facilities Backward step-wise selection, p<0.05 Calculated spatial models: resource selection functions for eagles – used vs. available w * = exp( β 1 x 1 + ... + β n x n )
resource selection probability functions for turbines – used vs. unused ∧
w =
exp( β 0 + β 1 x 1 + ... + β n x n ) 1 + exp( β 0 + β 1 x 1 + ... + β n x n )
Manly 2001
Reclassified continuous spatial models into four habitat quality bins Eagle Models Turbine Models
Validated eagle models postreclassification by comparing observed to expected per bin Assessed: R2>
•slope different from 1 •slope different from 0 •intercept different from 0 •goodness-of-fit
0.98
Random model slope = 0 Ideal model, slope = 1, intercept = 0 Johnson et al. 2006
Validated wind models pre-reclassification using area-under-the-curve & Kappa
AUC = 0.908±0.016 K = 0.572±0.051
AUC = 0.960±0.008 K = 0.761±0.035
AUC = 0.977±0.008 K = 0.794±0.041
Most projects have at least one high risk turbine NP
AM
RV