Objective 3: To apply a MCE model that determines the most suitable moose habitats post fire.
Implementing a GIS-based Multi-Criteria Evaluation Model
To begin, all data is collected from the Department of Energy and Resource Development or the Canadian National Fire Database (table 2). Next, all layers are converted to the ‘NAD 1983 CSRS New Brunswick Stereographic’ coordinate system and cells are resampled to a cell size of 500. Both constraint and criteria layers are then clipped to Zone 8's 150 km buffer, converted into raster layers and re-sampled using equation (1) to ensure an equal comparison.
Table 2. Data source used for geospatial analysis of moose habitat upon the occurrence of a forest fire in New Brunswick, Canada.
The constraints for this assessment are the roads, non-forest cover and layers, meaning that any area that falls within the boundary of these layers is unsuitable as moose habitat. Also, since this assessment is being run individually for the 10 fires in zone 8, the fire that is being examined is also considered a constraint. All layers are converted from a shapefile format to a raster file using the “Polygon to Raster” tool. Then, to input this information into the MCE both layers are reclassified using the “Reclassify” tool, giving a value of “0” to any area containing the features in each layer and a value of “1” to all other suitable areas.
The criteria for this assessment are the forest cover, wetlands, ecosite 6, all other ecosites, as well as non-forest cover, roads and parks layers. Again, since this assessment is being run individually for each of the 10 fires, any fire that has occurred within a 5 to 20 year range of the fire being examined is also a criteria. To create the criteria layers for this assessment, all layers are first converted to raster format using the method described in the previous section. Next, the “Euclidean Distance” tool is used on the forest cover, wetlands, fire extents, non-forest cover, parks, and roads layers to calculate the distance from each cell to the nearest cell. This tool takes into consideration the distance from desirable and undesirable features (Su et al., 2012). For the constraint layers, they are then are standardized using the raster calculator to give higher preference to areas further away from the constraint layers. For the criteria, the layers are standardized using the raster calculator to giver higher preference to areas closer to the criteria layers (Su et al., 2012). To differentiate between the desirable and non-desirable areas, the layers are each put into the “Raster Calculator” tool using equation (3) for the desirable layers: forest cover, wetlands, ecosite 6, all other ecosites and past fire extents, and equation (4) for the non-desirable layers: non-forest cover, roads, and parks.
Equation (3): Desirable = 100 * [“CRITERIA” / (MAX-MIN)]
Equation (4): Non-Desirable = 100 * [1 – (“CRITERIA” / (MAX-MIN))]
The pairwise comparisons are then used to calculate the final weights for each criteria (Table 3).
Table 3. Pairwise comparison table (pairwise rankings) used for determining the weight of each criteria and constraint.
Table 3 (Continued). Pairwise comparison table (individual rankings) used for determining the weight of each criteria and constraint.
Once all criteria rasters are standardized and the weights calculated, the MCE formula, equation (2) is then used to create 10 suitability rasters, one for each fire, using all criteria and constraint layers. (Note: the pariwise rankings are completed again for the fires without post-fire areas. This comparison has all criteria and contraints except fire extents, subsequantly all other factors are slightly more important, ranging on a scale of 1/5 to 5).
For Fires with post-fire areas:
Suitability = [(0.368 * "OTHER FIRES") + (0.190 * "Forest") + (0.190 * "Wetland") + (0.096 * "Ecosite 6") + (0.039 * "Other Ecosites") + (0.039 * "Roads ED") + (0.039 * "Non-Forest ED") + (0.039 * "Parks ED")] * ("Non-Forest Con" * "Parks Con" * "Roads Con" * "FIRE #")
For Fires without post-fire areas:
Suitability = [(0.310 * "Forest") + (0.310 * "Wetland") + (0.148 * "Ecosite 6") + (0.058 * "Other Ecosites") + (0.058 * "Roads ED") + (0.058 * "Non-Forest ED") + (0.058 * "Parks ED")] * ("Non-Forest Con" * "Parks Con" * "Roads Con" * "FIRE #")
After creating all 10 individual MCEs, the “Cell Statistics” tool is used to average the results from all of the individual fires. Next, the “Clip (Data Management)” tool is used to clip the output to the study area, making sure to check the “Use Input Features for Clipping Geometry” box. This produces an MCE that extends only to the extent of the study area. From here the “Focal Statistics” tool is used to find all areas which meet the maximum suitability and 10 km2 area requirement for moose. To do so, the radius of a 10 km2 circle is found in meters and converted into cells. Since this conversion produces a decimal, it is not accepted by the tool, so the number is rounded from 3.5 down to 3. This still meets the moose habitat range equally with an average suitability of 7km2 (instead of the original 10km2). Next, the “Integer” tool is used to convert each cell value of the raster into an integer. Finally, the “Extract by Attribute” tool is used on the second highest value of the modified MCE to find the top sites. This produces 196 top sites which are then given a 1,500 m2 buffer, based off of the rounded cell value used in the “Focal Statistics” tool. All 196 buffered sites create six clusters which are selected and dissolved to create six distinct polygons. These six polygons represent the six most suitable regions for moose habitat relocation in the Big Tracadie River Wildlife Management Zone of New Brunswick, Canada.