Multi-Criteria Evaluation (MCE) analysis was used to identify the best potential areas for prescribed burns, using ESRI ArcGIS. MCE's are a useful tool to address management issues that must consider multiple and potentially conflicting criteria and values. They generally involve the following five steps:

 

1)     Determining Important Criteria

In identifying the best suitable sites for prescribed burn areas I considered sites in relation to three criteria which include proximity to: caribou habitat, known existing Whitebark Pine plots and water (major lakes and rivers).[1] Due to the overlapping nature of Whitebark Pine and caribou habitat, water was included as a natural fire break; it was important to take natural breaks into account when considering areas that are most suitable to burn in a continuous landscape. Furthermore, development such as Jasper town-site, roads and ski areas were not included in this first MCE layer because Whitebark Pine exists at fairly high elevations and is unlikely to overlap with areas of high development.

 

2)     Creating and Normalizing Layers to a Common Scale

First, caribou habitat was represented by a layer showing kernel density representation of year-round caribou distribution. Second, due to the lack of accurate spatial data on the distribution of Whitebark Pine throughout the park, two data sources were merged to ensure a more accurate representation of where Whitebark Pine currently exists. The first dataset used contains information on where Whitebark Pine is predicted to exist based on specific biophysical site attributes.[2] The second dataset used existing research and monitoring coordinate locations of Whitebark Pine within the park. Four plots were missing Universal Transverse Mercator coordinates and their coordinates were estimated using descriptions of their location and Google Earth.  Using two datasets ensured that the Whitebark Pine ELC areas identified as potential places to burn in later steps were closer to existing seed sources. Moreover, the proximity to existing seed sources is important as recent literature suggests that Clark’s Nutcracker are more likely to cache seeds in burned areas close to their home range (Unite States Department of Agriculture 2011). Third, and finally, the water layer (including only larger lakes and rivers) was created from a vector land cover feature class.

For each layer, a separate Euclidean distance representation was created at a 10m raster resolution. This surface covered the entire park and categorized areas as having high or low values depending on whether they were near or far from the criteria. For instance, the caribou Euclidean distance representation categorized areas within and close to caribou habitats with low values, and areas farther from the habitat with high values. I then normalized the layers to a common scale from 0 to 1 so they could be overlaid and compared using the Fuzzy Membership spatial analysis tool of linear type.

 

3)     Determining Weights For Each Layer

To determine weights for each layer I used an online demonstration from Make It Rational's Collaborative Decision Making tool (Make It Rational 2012). Using a weighting process wherein every criterion is compared to every other criterion –in order to determine how much weight each layer should be given– I weighted caribou habitat at 77.02%, proximity to existing plots at 16.18%, and proximity to water at 6.8%. As I was most interested in identifying potential burn areas outside of caribou habitat, I weighted the caribou layer most heavily.

 

4)     Conducting an Overlay Using MCE Algorithms

I conducted a linear weighted overlay using ArcMap’s Weighted Sum tool, with the output at 10m raster resolution. The final weighted MCE map ranks all ELC areas predicted to contain Whitebark Pine based on the criteria previously identified: proximity to caribou habitat, proximity to known existing Whitebark Pine plots, and proximity to water as a natural break. A hill-shaded digital elevation model (DEM) was used as a background to show generally where the plots were located in the park, and the estimated elevation. Each research and monitoring plot is symbolized based on the project type (blister rust monitoring site, cone caging site, PhD research site, or incidental field observation).

 

5)      Performing a Sensitivity Analysis

A sensitivity analysis is important in any MCE, as it shows the effect that different weighting schemes have on the results. In reality, determining weights for each layer can be a controversial step, as the weights assigned depend on what is valued most. In this case, I performed a sensitivity analysis using the same input layers with equal weights.

 

 

---

 

[1] For a full description of each layer and its source, see Appendix A: Metadata, located on the References page. 

 

[2] This is from the Ecological Land Classification which these areas will now be referred to as Whitebark Pine ELC areas.  Until very recently this was the most accurate distribution map for Whitebark Pine within the park.