Using Geographic Information Systems in Predicting Reference Communities for Landscape Scale Restoration by ESRA OZDENEROL, PhD University of Memphis Department of Earth Sciences
A nonprofit, community-based organization that exists to help communities restore, manage and learn about their natural environment through volunteer involvement.
Oak Savanna
The Vision …
BIG RIVERS PARTNERSHIP PROJECT AREA
Collaborators: Cynthia Lane, Ph.D. Greg Noe, Ph.D. Bart Richardson
Methods: 1. Land Cover Classification data 2. Environmental data 3. Data categorized 4. Statistical Analyses 5. Predictive Model 6. Filters
1. LAND COVER CLASSIFICATION DATA MLCCS Hierarchical Classification System: Cultural or Natural/Semi-natural Five level system beginning with vegetation type or dominant cover type % impervious estimated for cultural cover types Modifiers for adding information for specific polygons
2. ENVIRONMENTAL DATA Data obtained for each HQN and Restorable site (polygon): Soil Texture and Drainage Slope and Aspect Shade
Soil Drainage and Texture USDA-NRCS, Official Soil Series description Soil characteristics commonly affecting the establishment and persistence of perennial native vegetation Predominant drainage and texture in upper horizon Soil
Soils Drainage: Drainage class: 1. = Excessively drained, Somewhat Excessively drained 2. = Well drained, Moderately well drained 3. = Somewhat Poorly Drained, Poorly drained, Very poorly drained Drainage class diversity
Soil Texture: S = Sand L = Loam O = Organic Texture class diversity
U.S.G.S. 30 meter digital elevation model Converted to grid format using ArcView Spatial Analyst Slope - mean and standard deviation for each site Slope and Aspect:
Aspect: Mean aspect & angular dispersion (aspect variability) Mean aspect converted to sine and cosine using circular statistics
Shade layer generated using DEM and ArcView Spatial Analyst Hottest day and time of day modeled Shade:
High Quality Native Community Disturbed Unsuitable Unknown Restorable 3. SITES CATEGORIZED
Disturbed = Disturbed = soils classified as “urban lands”, “udorthents”, and “gravel pits”; >75% impervious cover Unknown = Unknown = no soils data or aspect Unsuitable = Unsuitable = wetlands; >90% impervious
High Quality Native Community# sites Oak Forest (mesic, dry) 228 Maple Basswood Forest 44 Aspen Forest (temporarily flooded) 14 Floodplain Forest (silver maple) 212 Lowland Hardwood Forest 48 White Pine Hardwood Forest 2 Oak Woodland Brushland 120 Mesic Prairie 8 Dry Prairie (barrens, bedrock bluff, sand gravel) 58 Wet Meadow (shrub) 11 Dry Oak Savanna (sand gravel) 12 Mesic Oak Savanna 23
Restorable Cover Type # polys #ha Sparse trees + turf/grassland Agricultural crops Turf/grassland Deciduous trees Boxelder/Green ash forest Mixed woodland, disturbed Mixed coniferous & deciduous Coniferous trees
3. STATISTICAL ANALYSES 1. Tested relationship between High Quality Native Communities and environmental characteristics 2. Applied results of analysis to Restorable polygons to predict target community
3. STATISTICAL ANALYSES Factor analysis Linear discriminant function analysis
High Quality Native Community Oak Forest (mesic, dry) Maple Basswood Forest Aspen Forest (temporarily flooded) Floodplain Forest (silver maple) Lowland Hardwood Forest White Pine Hardwood Forest Oak Woodland Brushland Mesic Prairie Dry Prairie (barrens, bedrock bluff, sand gravel) Wet Meadow (shrub) Dry Oak Savanna (sand gravel) Mesic Oak Savanna 21 full, 12 aggregated
RESULTS – Full Analysis All environmental variables significantly different (Wilks’ Lamda, P<.00001) 94.8% of variation explained 6 discriminant functions statistically significant 9 communities reliably predicted >50%
RESULTS – Aggregated Analysis All environmental variables significantly different (Wilks’ Lamda, P<.00001), except shade 98.3% of variation explained 6 discriminant functions statistically significant 6 communities reliably predicted >50%
Predicted Native Communities
Undifferentiable communities: Oak forest Maple basswood forest Oak woodland brushland Mesic prairie Aspen forest Lowland hardwood forest
3. FILTERS Cost, Ease of restoration Rare native community Landscape – Patch size & Connectivity
Cost Filter: Ease of Conversion Patch size (polygon size) % impervious surface Access (slope)
Conversion matrix:
Patch Size & Connectivity Filter: Straight line allocation Existing native communities used as targets Undifferentiable sites converted to nearest native community
Prioritize restoration sites: Target rare community for restoration Communities reliably predicted from full analysis
SUMMARY: 9 full, 6 aggregated communities reliably predicted Target refined using cost and landscape filters Method can be used to prioritize sites based on project goals
Acknowledgements: Legislative Commission on Minnesota Resources Mississippi National River and Recreation Area Minnesota Department of Natural Resources, Conservation Partners Grant