1. GIS RS Habitat Modeling Approaches to Identify Riparian Communities
on the Pine Ridge Reservation
* Charles Jason Tinant
Don Belile
Helene Gaddie
Devon Wilford
* Corresponding Author, Oglala Lakota College 490 Piya Wiconi Road – Kyle, South Dakota
(USA)

2. Overview Populus deltoides are an
White River Group
Medicine Root Creek
Populus deltoides are an
early successional species colonizing point bars;
Recruitment is correlated with floods.
Damming and river alteration effects depend on channel type:
For meandering wash load streams (Missouri River) become hardwood forests;
For braided gravel streams (Platte River) cottonwoods woodlands extent increases.
Arikaree Group
Recruit 3 – 4 Native American STEM undergraduates to work as research interns
Enhance environmental science courses at OLC by incorporating GRIPP research into 2 – 4 classes over the next four semesters
Disseminate GRIPP results to K-12 students and undecided undergraduate students on the Pine Ridge reservation through participation in 4 – 6 outreach activities in the next two years.
Measure the canopy cover and geomorphology at 60 sampling units
Evaluate the health of woodlands by enumerating stand composition and age class of woody riparian species for at least 500 subplots
Identify geomorphic parameters significant to native woody riparian species recruitment
Create a predictive map of woody riparian species on the Pine Ridge Reservation at an 80% level of accuracy
Porcupine Creek

3. Great Plains Riparian Protection Project (GRIPP) Research Objectives
1) Understand PRR woodlands distribution and demography;
3) Predict woodlands community type using GIS remote sensing techniques.

4. Methodology – Using RS to Identify Sites
Unsupervised classification of 2-m DOQ;
Pull out remotely sensed “tree” layer;
Buffer streams 50-m from center of stream;
Buffer roads 250-m from roads;
Intersect and use output to clip “tree” layer;
Draped 100-m grid and randomly selected points.
Figures are courtesy
of Jim Sanovia

5. Methodology - Fieldwork
Sampled 22 plots in 2007 and 26 plots in 2008;
Estimated canopy cover at 4 community levels;
Enumerated trees to species at 5 age classes.
- Measured stream morphology (2007 only)
13 cross-sections by Rosgen Method.
White River Group
Medicine Root Creek

6. Analytical Approaches
Distinguishes juniper from cottonwoods
Identifies invasive Russian olive
Cloud cover!!
Doesn’t distinguish cottonwoods from hardwoods
Remotely Sensed
Landsat – 7
Correctly Identifies Woodlands > 80%
Needs Additional Information (underlying geology)
Computationally complex process
Misclassified watersheds
Physiographic Regions
SSURGO + DEM
Correctly identifies woodlands > 70%
Scale of mapping overlooks features below about 1:50,000 scale
Geology
Provides a context for understanding the underlying abiotic and ecologic processes
Lacks spatial context
Multivariate Approaches (DA + Clustering)
Final Habitat Model
MaxEnt

7. Remotely Sensed Approach -Final Classified Landsat - 7 Image
Distinguishes juniper from cottonwoods
Identifies invasive Russian olive
Cloud cover!!
Doesn’t distinguish cottonwoods from hardwoods

8. Computationally simple process
Geology for Pine Ridge Reservation has a need for stratigraphic revision
Correctly Identifies Woodlands ~ 70%

10. Physiographic Regions Logic Model - ArcGIS
Pourpoint
shapefile
10-m DEM
Shannon
10-m DEM
Jackson
Mosaic DEM
Depressionless
DEM
Streamflow
Model
Flow Direction
Flow Accumulation
Set Null Functions
Strahler
Model
Watershed
Model
10-m DEM
Bennett
Apply Sink and Fill Functions
Project to UTM
Zone 13
Mosiac Rasters
Add pourpoints and Iterate
Hydrologic
Properties
shapefile
SSURGO
database
MUKEY
Flat file
database
Hydrologic
Properties
shapefile
Hydrologic
Properties
shapefile
Hydrologic
Properties
shapefile
SSURGO
shapefiles
Hydrologic
Properties
31 - Hydrologic
Properties
Rasters
Select Hydrologic Properties
Tie to MUKEY
Join database to SSURGO shapefile by MUKEY
Select Hydrologic Properties
Tie to MUKEY
Apply Zonal Statistics
(Mean, Std. Dev, Max, Min)
Mosaic DEM
Rasters
Terrain
Rasters
Strahler
Model
Spatial Analyst
(Slope, Curvature)

11. Physiographic Regions Logic Model - Erdas Imagine
Hydrologic
Properties
shapefile
Hydrologic
Properties
shapefile
Hydrologic
Properties
shapefile
Hydrologic
Properties
shapefile
PCA Stack
15 Layers
31 - Hydrologic
Properties
Rasters
Hydrologic Properties
Stack - 31 Layers
Import into Imagine
Layer Stack
PCA to reduce
dimensionality
Physiographic Regions Model
– Based on USGS Nomenclature (when possible)
Initial Classification
20 classes
Intermediate
Classification
classes
Sand Hills
Eolian Sands
Fertile Lands
Tablelands
Foothills
Escarpment
Badlands
Alluvial
River Breaks
Isomeans
Clustering
Recode Results
Overlay
Mask Mixed Classes
Geology
Shapefile
DOQ
DEM

13. Correctly Identifies Woodlands > 80%
Aa class needs additional information on bedrock geology
Computationally complex process
Misclassified watersheds

14. Unconfined Channels High Peak Flows
Multivariate Approach – Clustering Dendrogram
Unconfined Channels High Peak Flows
Confined Channels
Narrow Flood Plains
Foot slopes
Active Point Bars
Cottonwood
Willow
Woodlands
Russian Olive
Woodlands
Juniper
Woodlands
Boxelder
Green Ash
American Elm

15. Microhabitat Niches by Geologic Unit
White River Group and Pierre Shale – Plains cottonwoods and willows species: erodible sediments with sparse vegetation, unconfined flood plains, high peak flows, frequent channel migration
Arikaree Formation - Green Ash, Boxelder, American Elm: cohesive sediments, mixed-grass prairie uplands, confined flood plains, attenuated peak flows, stable channels

16. Maximum Entropy Model
Uses ascii rasters and sample locations in csv format as model inputs;
Used 30m ascii rasters in UTM14 prepared using ArcGIS Spatial Analyst;
Model calculates omission rate, sensitivity, marginal and correlated response curves, model variable contributions and a jackknife test of model variable importance;
The following slides are results from MaxEnt model runs analyzing 28 variables from SSURGO soils data;
SSURGO quality for Shannan, Jackson, and Bennett counties (last updated in 1960s) has an effect on the quality of the model results;
The final model will incorporate SSURGO data, geology data, gridded precipitation data, classified Landsat imagery, and NVDI data.

17. Cottonwood/Willow Prediction using SSURGO Soils Variables
Variable Percent Contribution
dem 29.3
ec 23
kw 17.3
grass 9.9
slope 7.4
gypsum 3.9
water 2.8
silt 1.8
albedo 1.3
sar 0.7
om 0.6
caco3 0.6
shrub 0.5
ksat 0.4
hardwood 0.3
conifer 0.1

18. Cottonwood/Willow Prediction using SSURGO Soils Variables

19. Conclusions
Cottonwoods and hardwoods species on the Pine Ridge reservation are end-members distributed along a disturbance gradient;
The disturbance gradient corresponds with geomorphic response to precipitation events, which can be predicted by bedrock geology;
Landscape level variables accurately predict riparian community type on the Pine Ridge Reservation;
MaxEnt software predicts riparian community occurrence at a finer level of spatial detail than other landscape or watershed level analyses.

20. Acknowledgements Funded by: National Geospatial Agency
NSF Tribal College and University Program (TCUP)
Project is supported by:
OLC Math and Science Department:
Hannan LaGarry, Al Eastman, Chris Lee, Kyle White, Elvin Returns, Michael DuBray, Dylan Brave, Michael Thompson, Beau White, Jeremy Phelps, Landon Lupe (SDSU), Jim Sanovia (SDSMT)
MaxEnt reference:
Maximum Entropy Modeling of Species Geographic Distributions – Phillips, Anderson, and Shapire, Ecological Modeling ,Vol 190, 2006.