Developing a GIS-based Multi-Criteria Evaluation Model

An MCE is the technique chosen for this analysis as it considers many criteria and their importance relative to one another. This model involves standardizing each variable so that they can be compared against one another (Bonnycastle et al, 2014). The now normalized criteria rasters will be on a range of 0 to 100, with 0 being the least suitable and 100 being the most suitable. For the criteria in this study, the higher the standardized value, the most suitable the land is for moose habitat. The equation (1) that is used to standardize the criteria is as follows:

Equation (1): Cell Standardized Score = Range [(Cell Value – Min Cell Value) / (Max Cell Value – Min Cell Value)]

Next, to determine the weightings for each criterion, a pairwise comparison method is used to maintain a level of consistency between the most important and least important variables (Table 1) (Kumar et al., 2014). The pairwise comparison method involves a decision matrix, in which relationships between each of the criteria factors are determined (Kumar et al., 2014). In this case, the criteria that are most critical to moose habitat are previous fire locations, forest cover and wetlands, ecosite 6 and all other ecosites, in this order.

Table 1. The scale used to calculate the relative weights of each criteria and constraints. Weightings on the right are considered more important while weightings on the left are considered less important.

The justification for the criteria weighting is important to this study. The previous fire extents layer is weighted the most important because it correlates to high food availability for moose populations (Pastor et al., 1998). Forest cover and wetlands are both considered moderately important because forest cover is necessary for shelter, however shelter is only important if water is available nearby (Ardea Biological Consulting, 2004). This will be accounted for using the euclidean distance tool. The fourth highest weighting is Ecosite 6, this is because the Eastern Lowlands contain vegetation that is consistent with moose's feeding habitats, such as shrubs and tree buds within swamps and marshy areas (Rominger & Oldemeyer, 1989). The last criteria to consider is all other ecosites, this is considered the least important because moose are the least likely to inhabit these regions due to food availability and forest cover.

For constraints, a Boolean overlay analysis is used to determine unsuitable sites. This approach uses a mathematical tool to evaluate the values of inputs in order to determine the output values based on Boolean logic (ArcGIS, 2017). In other words, the land-use constraints such as roads, non-forest cover and parks are identified as uninhabitable areas and subsequently given a value of 0.  Since the goal of this study is to identify the six most suitable locations for moose habitat, the Boolean overlay analysis is partially used to reclassify the constraint layers. The roads, parks and non-forest cover constraints are given a value of 0, as moose will not be able to reside in residential, agricultural, industrial or urban areas and they must remain off all roads and park areas (Christie & Nason, 2004). These societal factors are considered to be uninhabitable, making these areas unsuitable for moose habitat.

The pairwise comparison is then run through a consistency matrix to determine appropriate ratios, only then is this MCE formula inputted into the raster calculator. The equation (2) that is used is as follows:

Equation (2): Suitability = [Σ (Weight*Criteria)] * (Constraints)                                                             

Equation 2 calculates the suitability scores and creates a raster that determines the most suitable locations for new moose habitat post fire. The most suitable locations will have the highest values, with the maximum cell value being 100, unsuitable locations for moose habitat will have the lowest values, with a minimum cell value being 0. Using an MCE approach, the most suitable sites using weighted factors for moose habitat are identified with the highest scores for each MCE being compared to determine optimal moose habitat sites. The six polygons pertaining to the most suitable moose habitats are determined using focal statistics, where clusters of sites are dissolved into one larger suitable habitat area of 10 km². After analyzing the outcomes, a map is produced to represent the six most suitable locations for new moose habitat in the Big Tracadie River Wildlife Management Zone of New Brunswick, Canada. A conceptual visualization of the MCE model is shown below (Figure 2).

Figure 2. Conceptual visualization of the GIS-based MCE model that will be implemented to determine the most suitable moose habitat sites.