Raster Based Predictive Ecosystem Mapping As A Tool For Habitat Modeling In Southeastern British Columbia

Paper number 807

Habitat Modeling With GIS Moderated Session

Esri World Users Conference

July 12, 2001, San Diego California

Maureen V. Ketcheson M.Sc. R.P.Bio

Tom Dool BES

A Predictive Ecosystem Mapping (PEM) model for mapping the Biogeoclimatic Site Series Classification for Southeastern British Columbia (Braumandl and Curran 1992) was developed using a raster-based approach (Ketcheson, Dool and Wilson 2001a, 2001b). The model is composed of spatial data input layers, in this case as many as 19 separate layers with 52 separate variables, which are registered to a common base. The input layers were developed in a grid format with a 25 m. pixel size. The input layers were derived from topographic mapping, terrain mapping, Landsat 7 imagery, forest inventories, and geological mapping. The individual layers developed were as follows:

· Biogeoclimatic subzone variant

· Slope class

· Aspect class

· Solar radiation shading class

· Landscape shape class

· Toe slopes

· Satellite imagery classification

· Wetlands

· Alpine Forest

· Forest Inventory Type Group

· Forest Height

· Drainage

· Materials texture

· Stream Density Class

· Gullies and avalanche paths

· Neural network classification probability

· Field Plot Site Series and UTM Coordinates

A knowledge table was developed which initially, subjectively, assigns a value to each input layer variable in relationship to the site series classification using an informal Bayesian approach. The PEM model derives a cumulative score for site series based on the input layer variables and the knowledge table values. The site series scores are noted and the top scoring site series is assigned to the pixel, while the second place site series is noted and the "point spread" is documented. The model is run and plot "goodness of fit" of the PEM result against the field plot data is determined. Through several iterations the fit of the model to the field plot data is improved until either a satisfactory level of fit is achieved, or the financial resources for the mapping are exhausted.

Mapping for a wide variety of Biogeoclimatic subzones and variants in dry, moist and wet macroclimates was completed over two years. An independent field data set was used to assess the accuracy of the mapping. Accuracy varied with the site series being mapped. In general mid-slope units were the best mapped, and flatter units were more poorly mapped. The model needs more work in subdued landscapes. It is presently not effective at differentiating rich sites from poor sites in local situations.

A structural stage model was developed for the project within the dry climatic region. It was based on forest cover information and can only be considered as accurate as the inventory input information. At this point the structural stage model is not reliable.

The site series classification can be applied to a wildlife habitat capability rating (RIC 1999) via a cross-walk table. The result is a wildlife habitat map that indicates the value of each pixel for an activity and season of use for each species of interest. Two kinds of habitat maps can be generated. A capability map, that indicates the potential of the landscape to support the animal seasonally, regardless of the present day structural stage of the stand and a suitability map that indicates the present day status to the animal based on stand structural characteristics.

We extended this model to a large area where Mountain Goat habitat mapping was needed (Sinclair, Dool and Ketcheson, 2001) and were able to identify high value habitats where potential conflicts with ski area developments and heli-hiking were of concern. We also used the site series model to rate each unit for its capability to support certain Grizzly Bear food plants (Vaccinium spp. and Shepherida canadensis), and riparian habitats (Ketcheson, Sinclair, Dool and Burns 2001.

We used a similar PEM approach to mapping US Habitat Types (Johnson and O’Niel 2001) across the 49^th parallel (Ketcheson, Mack and Littlewood, 2001) with good success. This is a cooperative project between the BC Ministry of Forests and the Northwest Power Planning Council and the Northwest Habitat Institute of the United States. This extends their classification of 15 applicable habitat types into the Columbia Basin within British Columbia. These are landscape level habitat types that are used to assist resource manager’s decisions as they relate to species distribution and wildlife-habitat issues within the Columbia Basin international watershed.

The potential for this method of mapping ecosystems and wildlife habitats, inexpensively (approximately $0.35/ha cdn for site series level and $0.0064/ha for US Habitat Type level) over large areas of landscape are great very exciting. More work is needed on the accuracy of the model in wetter sites of differing trophic status and within subdued terrain.

References Cited

Braumandl T.F. and M. Curran. 1992. A Field Guide to Site Identification and Interpretation in the Nelson Forest Region. Land Management Handbook NUMBER 20. BC Ministry of Forests.

Breiman, L., J. H. Friedman, R. A. Olshen, and C. J. Stone. 1984. Classification and regression trees. Wadsworth & Brooks, Monterey, CA.

Johnson, D.H. and T.A.O’Neil 2001. Wildlife Habitat Relationships in Washington and Oregon, Oregon State University Press, Corvallis, Oregon.

Ketcheson, M.V. and T. Dool. 2001a. Arrow TSA Site Index Adjustment Project, Predictive Ecosystem Mapping, Year Two. An unpublished Report to the Arrow IFPA Committee, Nelson, BC.

Ketcheson, M.V.,T. Dool, and S. Wilson. 2001b. Canal Flats Operating Area Predictive Ecosystem Mapping (PEM). (TEMBEC Industries Inc., Cranbrook, BC). Unpublished Report and Maps for TEMBEC Industries Inc., Cranbrook, BC.

Ketcheson, M.V., D. Mack and C. Littlewood. 2001. Classification and Mapping of US Wildlife Habitat Types In The Columbia River Basin of British Columbia. An unpublished report and maps to Bryan Nyberg, Manager, Wildlife Policy and Adaptive Management, Forest Practices Branch, BC Ministry of Forests, Victoria, BC.

Ketcheson, M.V., B. Sinclair, T. Dool and G. Burns. 2001. Flathead Special Resource Management Zone, Predictive Ecosystem Mapping (PEM) and Grizzly Bear Habitat Capability Assessment Project.

Resources Inventory Committee (RIC) 1999. British Columbia Wildlife Habitat Rating Standards, Version 2.0. Ministry of Environment, Lands and Parks, Resource Inventory Branch. Victoria, BC.

Sinclair, B., T. Dool and M.V. Ketcheson 2001. Mountain Goat Habitat Capability Assessment Mapping Model for the Golden TSA, Unpublished report to Rob. Neil, BC Ministry of Environment, Cranbrook, BC.

Statistica. 2000. Statistica neural networks. Statsoft, Inc. Tulsa, OK.

Maureen V. Ketcheson M. Sc. R.P.Bio.

Tom Dool BES

JMJ Holdings Inc.

Suite 208 – 507 Baker Street

Nelson, BC Canada V1L 4J2

(250) 354-4913 fac (250) 354-1162