Multi-Criteria DSS for Statewide Habitat Degradation Assessment by Hanna Ford

An assessment of the risks to the primary natural resource concerns plus the prioritization of agency programs to protect these resources was the goal of the Natural Resources Conservation Service (NRCS) Statewide Resource Assessment (SRA) project. The Arkansas NRCS office partnered with the Center for Advanced Spatial Technologies (CAST) at the University of Arkansas and many other natural resource partners to leverage the knowledge of our natural resource experts, the existing digital dataset and the modeling capabilities within ArcGIS to design and develop over twenty unique models to represent the relative risks to each NRCS targeted resource concerns within Arkansas. As part of the GIS technical support team for this project, we would like to share our experiences with this project and explain how the planning and model development process has sharpened the focus of resources from the NRCS staff and other natural resource decision makers within Arkansas.

Transcript

1. 2013 Arkansas GIS Users Forum Symposium September 12, 2013 | Rogers, Arkansas Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
2. Habitat Degradation Working Group NRCS Staff NRCS Conservation Partners Mike Sullivan, State Conservationist NRCS Luis Hernandez, MLRA Leader/State Soil Scientist NRCS Pam Cooper, Resource Inventory Specialist NRCS Reed Cripps, Assistant Conservationist for Easements NRCS Nancy Young, State Resource Conservationist NRCS Walt Delp, State Conservation Engineer NRCS Edgar Mersiovsky, Senior Soil Scientist NRCS Nelson Rolong, Assistant State Soil Scientist NRCS John Lee, State Agronomist NRCS Lane Johnson, Assistant State Conservation Engineer NRCS Charlotte Bowie, State Irrigation Engineer NRCS Wavey Austin, Environmental Engineer NRCS George Rheinhardt, Forester NRCS Ron Morrow, Grazing Land Specialist NRCS James Baker, Biologist NRCS Rich Joslin, Resource Conservationist NRCS Shawn Brewer, Hydraulic Engineer NRCS Terry Holland, USGS Arkansas Water Science Center Joe Krystofik, U.S. Fish and Wildlife Service Jason Milks, The Nature Conservancy - Arkansas Ethan Inlander, The Nature Conservancy - Arkansas Melissa Jenks, The Nature Conservancy - Arkansas Roger Mangham, The Nature Conservancy - Arkansas Matt Lindsey, The Nature Conservancy - Arkansas Bill Holiman, Department of Arkansas Heritage Chris Colclasure, Department of Arkansas Heritage Tom Foti, Department of Arkansas Heritage Dan Scheiman, Audubon Arkansas Dan Smith P.G., Arkansas Department of Health Ricky Chastain, Arkansas Game and Fish Commission Martin Blaney, Arkansas Game and Fish Commission David Long, Arkansas Game and Fish Commission Steve Filipek, Arkansas Game and Fish Commission Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
3. Background On February 1, 2011 the NRCS National Office issued directives to all states to create a consistent and defensible, science-based methodology for identifying areas “at-risk” to 31 natural resources concerns. The solution developed during Phase I & II of the NRCS State Resource Assessment project was successful in meeting the primary objectives of the NRCS State and National Offices. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
5. National Directives Summary Spreadsheet Potential At-Risk Acres for each NRC
6. NRCS Land Use Designations Land Use Designations are a “simplified” version of NLCD 2006 defined by the NRCS Headquarters in their Phase I Instructions and Documentation. Associated Ag Land Barren Land Crops Forest Pasture Range Urban Water
7. NRCS Land Use Designations Crosswalk
8. Current Working Model Structure • Coincidence/Overlay Model using raster algebra and a combination of binary and classified rasters; • Weighting determined by NRCS from conservation partner meetings. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
9. 9 Phase I – ArcGIS 10 ModelBuilder Model Structure (June 2011)
10. Binary - 1 x Weight Classified – Value (0-3) x Weight Part 1 – Input Layers Weighted Sum
11. NRCS Resource Assessment Phase I & II Model Results PHASE 1 – BY HUC8 PHASE 2 – BY HUC12 Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
12. Phase II Model Inputs BINARY LAYERS CLASSIFIED LAYERS • Conservation Delivery Network Priority Areas • Variable Riparian Buffer • NRCS Working Lands for Wildlife by HUC12 • Various Forest Bird and Wetland Prioritizations • Audubon’s Important Bird Areas • Forest Fragmentation (Patch Size) – Acres • Oil and Gas Wells – Density by HUC12 • AWAP Data – Terrestrial Scores by HUC12 • AWAP Data – Aquatic Combined Scores by HUC12 • Combination of… - 303(d) Impaired waterways and waterbodies HUC12– all impairments & - Ecologically Sensitive Waters HUC12 (ESW) & - Extraordinary Resource Waters HUC12 (ERW) • Juxtaposition to Public Lands & Easements to 4-miles/WRP Expansion & Connection Opportunities Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
13. Variable Riparian Buffer (wt. 1) Capturing stream banks and potential riparian areas
14. Juxtaposition to Public Lands and/or WRP Easement Boundary *see next slide for explanation of this layer Value 0 - No Priority 1 - Within 2-4 miles of Public Land and/or WRP Easement 2 - Within 1-2 miles of Public Land and/or WRP Easement 3 - Within 1 mile of Public Land and/or WRP Easement
15. Mean of Model Output Values by HUC12 Using the original model output values we calculated the MEAN of values by 12 digit HUC. MEAN BY HUC12 0 0.003 - 7.93 7.94 - 10.81 10.82 - 12.44 12.45 - 14.05 14.06 - 15.95 15.96 - 19.39 19.40 - 24.93
16. Variety of Model Output Values Using the original model output values we calculated the variety of values by 12 digit HUC. Lower variety numbers mean that there is little maybe only one or two model output values in that HUC12; the highest HUC12s have 11 different model outputs. VARIETY by 12 Digit HUC 1 2 3 4 5 6 7 8 9 10 11 HUC8 Boundaries
17. Variety (uniqueness) of Model Output Values Looking at just the 8-digit HUC Lower Ouachita by variety of model output values. Associated table shows the values for the selected HUC12. There are seven model values within this HUC (0, 12, 17, 19, 22, 24, 25). VARIETY by HUC12 2 3 4 5 6 7 8 9 HUC8 Boundaries
18. A Framework for Identifying Areas Potentially At-Risk for Habitat Degradation in the State of Arkansas using Available GIS Coverages M.A. Geography Thesis – In progress Hanna L Ford University of Arkansas Department of Geosciences Center for Advanced Spatial Technologies Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
19. Thesis Committee Dr. Fred Limp, Chair University of Arkansas, Department of Geosciences Center for Advanced Spatial Technologies Dr. Jackson Cothren University of Arkansas, Department of Geosciences Center for Advanced Spatial Technologies Dr. David Krementz University of Arkansas, Department of Biological Sciences Arkansas Cooperative Fish and Wildlife Research Unit Brian Culpepper University of Arkansas Center for Advanced Spatial Technologies Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
20. Objective To expand upon the existing working model developed for NRCS identifying areas “at-risk” for Inadequate Habitat for Fish and Wildlife by applying a Spatial Decision Support System framework using Simple Additive Weighting (SAW) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) techniques. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
21. Proposed Expert’s Questionnaire • Using expert input via questionnaire to establish: - Input layers (Criteria) - Maximize/Minimization of input layers - Weight (Importance) of input layers (sum of criterion importance =1) Target for questionnaire is 20- 25 experts from a variety of backgrounds (i.e. terrestrial, aquatic, habitat, etc…) • The questionnaire, results and validation of the expert’s responses will be structured and calculated following methodologies described by Kučas1 and Jakimavičius & Burinskiene2 1. Andrius Kučas. 2010. Location prioritization by means of multicriteria spatial decision‐support systems: A case study of forest fragmentation‐based ranking of forest administrative areas. Journal of Environmental Engineering and Landscape Management. Vol. 18, Iss. 4, 2010. 2. Marius Jakimavičius and Marija Burinskiene. 2007. Automobile transport system analysis and ranking in Lithuanian administrative regions. Transport Vol. 22, Iss. 3, 2007. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
22. Expert’s Questionnaire Online questionnaire - linked to published online maps for data review, providing experts opportunity to explore potential datasets for inclusion in the model. - explanations of the data layer - what do the attributes mean - how it was created - methodologies - from what organization the data was received Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
23. Simple Additive Weighting Prep • Standardize value ranges for criteria using a score range procedure: - Higher value of score, the more attractive the criterion, - Lower value of score, the less attractive the criterion, - Scale 0-1; Score Range Procedure is not proportional. • Benefit Criteria (Maximize) • Cost Criteria (Minimize) 3. Jacek Malczewski. 1999. GIS and Multicriteria Decision Analysis. Department of Geography, University of Western Ontario. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
24. Simple Additive Weighting (SAW) 4. Ghafoori and Syyed Ali. 2007. MCDM Multi-Criteria Decision Making. Presentation of unknown origin, searching for details. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
25. • Not Normalized HUC12 Watersheds (Alternatives) Criteria (Input Layers) Avg Forest Patch Size Oil/Gas Well Dens Avg Dist to Public Land % Imper Surface % Forested Rip Buff AWAP Score 0802030102 Bull Creek-Cyprus Bayou 150 acres 2% of HUC 12 area 15 miles 30% of HUC 12 area 10% of Riparian areas – Forested Avg. 1102 ~1,565 HUC12s In Arkansas; some only partial -- -- -- -- -- -- Importance (Weight from expert questionnaire) 0.225 0.115 0.220 0.180 0.190 0.070 Function MAX MIN MIN MIN MAX MAX Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment Simple Additive Weighting (SAW)
26. • Normalized using Score Range procedure HUC12 Watersheds (Alternatives) Criteria (Input Layers) Avg Forest Patch Size Oil/Gas Well Dens Avg Dist to Public Land % Imper Surface % Forested Rip Buff AWAP Score 0802030102 Bull Creek-Cyprus Bayou 0.60 0.47 0.21 0.48 0.45 0.22 ~1,565 HUC12s In Arkansas; some only partial -- -- -- -- -- -- Importance (Weight from expert questionnaire) 0.225 0.115 0.220 0.180 0.190 0.070 Function MAX MIN MIN MIN MAX MAX Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment Simple Additive Weighting (SAW)
27. Example results 0802030102 Bull Creek-Cyprus Bayou = 0.42255 0.60 x 0.225 + 0.47 x 0.115 + 0.21 x 0.220 + 0.48 x 0.180 + 0.45 x 0.190 + 0.22 x 0.070 The best alternative (HUC12) would be the one closest to 1. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment Simple Additive Weighting (SAW)
28. TOPSIS Technique for Order Preference by Similarity to Ideal Solution - Identifies the alternative (HUC12) closest to the “Ideal Alternative” and farthest from the “Negative Alternative.” 4. Ghafoori and Syyed Ali. 2007. MCDM Multi-Criteria Decision Making. Presentation of unknown origin, searching for details. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
29. TOPSIS Example - Normalized Decision Matrix; transformation of attributes to dimensionless attributes allowing for cross-criteria comparison - Calculate for each entry; then use to divide each column: = / 2 =1 1/2 1. Andrius Kučas. 2010. Location prioritization by means of multicriteria spatial decision‐support systems: A case study of forest fragmentation‐based ranking of forest administrative areas. Journal of Environmental Engineering and Landscape Management. Vol. 18, Iss. 4, 2010. 2. Marius Jakimavičius and Marija Burinskiene. 2007. Automobile transport system analysis and ranking in Lithuanian administrative regions. Transport Vol. 22, Iss. 3, 2007. 4. Ghafoori and Syyed Ali. 2007. MCDM Multi-Criteria Decision Making. Presentation of unknown origin, searching for details. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
30. HUC12 Watersheds (Alternatives) Criteria (Input Layers) Avg Forest Patch Size Oil/Gas Well Dens Avg Dist to Public Land % Imper Surface % Forested Rip Buff AWAP Score 111401060604 Bull Creek 0.075 0.928 0.761 0.748 0.388 0.697 111101050301 Bull Creek-Poteau River 0.047 0.037 0.507 0.374 0.892 0.578 110100120205 Bullpen Creek- Strawberry River 0.996 0.005 0.406 0.548 0.233 0.424 2 1/2 Denominator Value 2007.866 0.021541 19.72308 0.4011234 0.2578759 1581.73 Importance (Weight from expert questionnaire) 0.225 0.115 0.220 0.180 0.190 0.070 Function MAX MIN MIN MIN MAX MAX Normalized Decision Matrix (Scaled for Cross-Criteria Comparisons) Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
31. TOPSIS • Example - Weighted Normalized Decision Matrix; multiply each column by it’s expert determined weight = Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment 1. Andrius Kučas. 2010. Location prioritization by means of multicriteria spatial decision‐support systems: A case study of forest fragmentation‐based ranking of forest administrative areas. Journal of Environmental Engineering and Landscape Management. Vol. 18, Iss. 4, 2010. 2. Marius Jakimavičius and Marija Burinskiene. 2007. Automobile transport system analysis and ranking in Lithuanian administrative regions. Transport Vol. 22, Iss. 3, 2007. 4. Ghafoori and Syyed Ali. 2007. MCDM Multi-Criteria Decision Making. Presentation of unknown origin, searching for details.
32. HUC12 Watersheds (Alternatives) Criteria (Input Layers) Avg Forest Patch Size Oil/Gas Well Dens Avg Dist to Public Land % Imper Surface % Forested Rip Buff AWAP Score 111401060604 Bull Creek 0.017 0.107 0.167 0.135 0.073 0.048 111101050301 Bull Creek-Poteau River 0.011 0.004 0.111 0.067 0.169 0.040 110100120205 Bullpen Creek- Strawberry River 0.224 0.0005 0.101 0.099 0.044 0.030 2 1/2 Denominator Value 2007.866 0.021541 19.72308 0.4011234 0.2578759 1581.73 Importance (Weight from expert questionnaire) 0.225 0.115 0.220 0.180 0.190 0.070 Function MAX MIN MIN MIN MAX MAX Weighted Normalized Decision Matrix Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
33. TOPSIS • Example - Select Ideal Positive Variants + ; If the criterion is minimized, take the minimal value. and - Ideal Negative Variants − ; If the criterion is maximized, take the maximal value. 1. Andrius Kučas. 2010. Location prioritization by means of multicriteria spatial decision‐support systems: A case study of forest fragmentation‐based ranking of forest administrative areas. Journal of Environmental Engineering and Landscape Management. Vol. 18, Iss. 4, 2010. 2. Marius Jakimavičius and Marija Burinskiene. 2007. Automobile transport system analysis and ranking in Lithuanian administrative regions. Transport Vol. 22, Iss. 3, 2007. 4. Ghafoori and Syyed Ali. 2007. MCDM Multi-Criteria Decision Making. Presentation of unknown origin, searching for details. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
34. HUC12 Watersheds (Alternatives) Criteria (Input Layers) Avg Forest Patch Size Oil/Gas Well Dens Avg Dist to Public Land % Imper Surface % Forested Rip Buff AWAP Score 111401060604 Bull Creek 0.017 0.107 0.167 0.135 0.073 0.048 111101050301 Bull Creek-Poteau River 0.011 0.004 0.111 0.067 0.169 0.040 110100120205 Bullpen Creek- Strawberry River 0.224 0.0005 0.101 0.099 0.044 0.030 2 1/2 Denominator Value 2007.866 0.021541 19.72308 0.4011234 0.2578759 1581.73 Importance (Weight from expert questionnaire) 0.225 0.115 0.220 0.180 0.190 0.070 Function MAX MIN MIN MIN MAX MAX Ideal Positive Variants + Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
35. HUC12 Watersheds (Alternatives) Criteria (Input Layers) Avg Forest Patch Size Oil/Gas Well Dens Avg Dist to Public Land % Imper Surface % Forested Rip Buff AWAP Score 111401060604 Bull Creek 0.017 0.107 0.167 0.135 0.073 0.048 111101050301 Bull Creek-Poteau River 0.011 0.004 0.111 0.067 0.169 0.040 110100120205 Bullpen Creek- Strawberry River 0.224 0.0005 0.101 0.099 0.044 0.030 2 1/2 Denominator Value 2007.866 0.021541 19.72308 0.4011234 0.2578759 1581.73 Importance (Weight from expert questionnaire) 0.225 0.115 0.220 0.180 0.190 0.070 Function MAX MIN MIN MIN MAX MAX Ideal Negative Variants − Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
36. TOPSIS Example - Ideal Positive Variants + {0.224 0.0005 0.101 0.067 0.169 0.048} - Ideal Negative Variants − {0.011 0.107 0.167 0.135 0.044 0.030} 1. Andrius Kučas. 2010. Location prioritization by means of multicriteria spatial decision‐support systems: A case study of forest fragmentation‐based ranking of forest administrative areas. Journal of Environmental Engineering and Landscape Management. Vol. 18, Iss. 4, 2010. 2. Marius Jakimavičius and Marija Burinskiene. 2007. Automobile transport system analysis and ranking in Lithuanian administrative regions. Transport Vol. 22, Iss. 3, 2007. 4. Ghafoori and Syyed Ali. 2007. MCDM Multi-Criteria Decision Making. Presentation of unknown origin, searching for details. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
37. TOPSIS Example - Calculate the variant’s deviation from the Ideal Positive Variant + + = − + 2 =1 - Calculate the variant’s deviation from the Ideal Negative Variant − − = − − 2 =1 4. Ghafoori and Syyed Ali. 2007. MCDM Multi-Criteria Decision Making. Presentation of unknown origin, searching for details. 1. Andrius Kučas. 2010. Location prioritization by means of multicriteria spatial decision‐support systems: A case study of forest fragmentation‐based ranking of forest administrative areas. Journal of Environmental Engineering and Landscape Management. Vol. 18, Iss. 4, 2010. 2. Marius Jakimavičius and Marija Burinskiene. 2007. Automobile transport system analysis and ranking in Lithuanian administrative regions. Transport Vol. 22, Iss. 3, 2007. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
38. HUC12 Watersheds (Alternatives) Criteria (Input Layers) Avg Forest Patch Size Oil/Gas Well Dens Avg Dist to Public Land % Imper Surface % Forested Rip Buff AWAP Score 111401060604 Bull Creek 0.043 0.011 0.004 0.0046 0.0092 0 111101050301 Bull Creek-Poteau River 0.045 0.0000123 0.0001 0 0 0.000064 110100120205 Bullpen Creek- Strawberry River 0 0 0 0.0010 0.0156 0.000324 2 1/2 Denominator Value 2007.866 0.021541 19.72308 0.4011234 0.2578759 1581.73 Importance (Weight from expert questionnaire) 0.225 0.115 0.220 0.180 0.190 0.070 Function MAX MIN MIN MIN MAX MAX Deviation from Ideal Positive Variants + {0.224 0.0005 0.101 0.067 0.169 0.048} − + 2 calculated for each entry Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
39. Deviation from Ideal Negative Variants − {0.011 0.107 0.167 0.135 0.044 0.030} − − 2 calculated for each entry HUC12 Watersheds (Alternatives) Criteria (Input Layers) Avg Forest Patch Size Oil/Gas Well Dens Avg Dist to Public Land % Imper Surface % Forested Rip Buff AWAP Score 111401060604 Bull Creek 0.000036 0 0 0 0.000841 0.000324 111101050301 Bull Creek-Poteau River 0 0.010609 0.003136 0.004624 0.015625 0.0001 110100120205 Bullpen Creek- Strawberry River 0.045369 0.0113423 0.004356 0.001296 0 0 2 1/2 Denominator Value 2007.866 0.021541 19.72308 0.4011234 0.2578759 1581.73 Importance (Weight from expert questionnaire) 0.225 0.115 0.220 0.180 0.190 0.070 Function MAX MIN MIN MIN MAX MAX Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
40. TOPSIS Example - Calculate the proportional variant’s deviation from the Ideal Variant = −/ + + − 4. Ghafoori and Syyed Ali. 2007. MCDM Multi-Criteria Decision Making. Presentation of unknown origin, searching for details. 1. Andrius Kučas. 2010. Location prioritization by means of multicriteria spatial decision‐support systems: A case study of forest fragmentation‐based ranking of forest administrative areas. Journal of Environmental Engineering and Landscape Management. Vol. 18, Iss. 4, 2010. 2. Marius Jakimavičius and Marija Burinskiene. 2007. Automobile transport system analysis and ranking in Lithuanian administrative regions. Transport Vol. 22, Iss. 3, 2007. Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
41. TOPSIS- Variant’s Deviation Results Variable Alternative Options Bull Creek Bull Creek- Poteau River Bullpen Creek- Strawberry River + 0.268 0.213 0.130 − 0.035 0.185 0.250 0.115 0.465 0.658 Best alternative: Closest to Ideal Positive Variant AND Farthest from Ideal Negative Alternative; Best Candidate for Conservation Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
42. Sensitivity Analysis • Sensitivity Analysis - Methodically altering expert input to evaluate the changes on model results - Separate experts into broad groups and run models using subsets of expert input Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment
43. A Framework for Identifying Areas Potentially At-Risk for Habitat Degradation in the State of Arkansas using Available GIS Coverages Pam Cooper, NRCS Jackson Cothren, Center for Advanced Spatial Technologies Brian Culpepper, Center for Advanced Spatial Technologies Hanna Ford, Center for Advanced Spatial Technologies Multi-Criteria Decision Support for Statewide Habitat Degradation Assessment