Predictive Vegetation Modelling & Mapping for 6e10 D. Puric-Mladenovic Information Management & Spatial Analysis Unit Southern Science and Information March -08-2011
Predictive Vegetation Modelling • How was it developed? • How can it be used? • Limitations
How was it developed?
• Sampled – – – –
6e10 and St.L. NP Bruce NP and Peninsula Kawartha PP, Trent-Severn Waterway
• Sampling continues – Eco-districts: 6e15, 6e11, 6e12 – Niagara Escarpment Planning Area
6e10-GPE • 1080 field plots • Over 100 wetland plots • Photo-plots – pitch pine – open rock-barren sites
Field plot distribution in 6e10-GPE
http://www.forestry.ut oronto.ca/imsa/VSP
Plot based sampling
Site-groups mapped by modeling then classification Species
Classification Species-groups mapped by modeling then classification
Modeling
Vegetation Data
Classification
Site-groups mapped by classification then modeling
Site-group Modeling Species-group
Species-groups mapped by classification then modeling
6e10 Modeling Products • It is a digital atlas – ~ 180 vegetation and habitat maps
PVM – 6E10 SLNP How can it be used?
• One map / product or class is not enough • Support multiple uses and applications • Use field data today , but also in the future • Integrate field data with spatial technology (e.g., higher resolution remote sensing data) • Get more, if not for less but for the same amount of $.
Climate change Wildlife and SAR Goods and services
Invasive species
VEGETATION MAPS
Vegetation management
Plant species composition
Significant woodlands and wetlands
NHS
How this mapping differ from the traditional vegetation mapping? Traditional mapping
PVM mapping
+
How this differ from the traditional vegetation mapping?
PVM Mapping outcomes
Wetlands Classes
Vegetation
Forest
Habitat
Types/ classes
structure
characteristics
Trees and shrubs
ELC classes
Mature forest
Soil moisture
Ground veg.
Silvicultural Cl.
Tree abundance
Species
Species
assemblages
6e10-GPE area has over 180 vegetation and habitat maps
Rock-barren Canopy closure
Some habitat requirements for forest dwellings SAR
?
SAR
• Do we have sufficient spatial information to support SAR protection and recovery planning across landscapes? • Can we target specific areas for SAR sampling?
6e10 -SAR modeling • Typha marsh
– Evaluated wetlands for modeling – Field data for validation
Wetland classes Probability of typha marsh
6e10 -SAR modeling • Canopy closure • Forest maturity • Silvicultural classes
Five-lined Skink (Eumeces fasciatus)
Rock-barren
Sustaining What We Value • Vegetation is used as a surrogate for overall landscape biodiversity • Vegetation mapping enables wildlife and SAR habitat modeling and mapping • Vegetation mapping supports estimates of ecological goods and services • Vegetation mapping provides necessary input for defining and setting various functional targets.
NHS application - Representation
• Vegetation classes are used as conservation features • For each specific targets (in ha) is set
Emerald Ash Borer - Agrilus planipennis
Emerald Ash Borer - Agrilus planipennis
White Ash abundance 0-40%
Soil moisture
Habitat characteristics
Basal area and stand density
Basal area
Predicted biomass and carbon Predicted Biomass & carbon
• Biomass estimates – Productivity, hydrology, wildlife, • Standing carbon estimates • Carbon uptakes estimates* • Carbon credits – Pilot project in eastern Ontario – EOMF, NCC ttp://www.forestry.utoronto.ca/imsa/ BruceBiomass/index.html
Limitations
Modeling issues • Spectral information – Landsat (spring, summer, fall)
• Modeling: species first – Statistically not possible to separate large number of different vegetation classes – Modeling veg. class – yielded more than 65% misclassification error rate
• Classification continues to evolve and change
Sampling confusion • Classifying rock barren, savannah, woodland or forest • Fuzzy boundary (threshold) Open
Savannah
Woodlands
Forest
< 10%
10 to 34%
35 to 59%
60 % to 100%
Solution Collect base information
Not modeled and mapped • • • • •
Meadows Abandoned fields Disturbed vegetation Successional shrub and other communities ……
• Solution - sample this vegetation too – We do not need classification to start sampling
Vegetation mapping perception • • • •
Lines are more accurate pixel Interpretation is more accurate than modeling Pixels are “jaggy” Validation and errors – Not knowing vs. knowing
• Species assemblages ( veg. types) are stable • Looking for wrong vs. right • Scale of application
Running the Numbers: An American Self-Portrait (2006 - 2009)
Running the Numbers: An American Self-Portrait (2006 - 2009)
Running the Numbers: An American Self-Portrait (2006 - 2009)
Running the Numbers: An American Self-Portrait (2006 - 2009)
Running the Numbers: An American Self-Portrait (2006 - 2009)
Cans Seurat, 2007 60x92" Depicts 106,000 aluminum cans, the number used in the US every thirty seconds Running the Numbers: An American Self-Portrait (2006 - 2009)
Validation • Field validation ( and independent data set) • Every field visit is an opportunity to collect field data – For validation – But also to enhance the inventory • “A database of validation efforts over the years which may have helped refine the quality of the product beyond pure stats” (Greg Saunders, SLNP)
Climate change Wildlife and SAR Goods and services
Invasive species
VEGETATION MAPS
Vegetation management
Plant species composition
Significant woodlands and wetlands
NHS
Diverse products
Diverse applications NHS
SAR
Biodiversity
SOR
Invasive sp.
Forest mng.