Skip to main content

University of Guelph HomeHome

Search Type
Menu Search

Secondary menu

  • CSAHS
  • Awards
  • Academics
  • Alumni & Friends
  • Research

CSAHS

Navigation
Geography - Applied Geomatics (GEOG*4480*w18)
  • GEOG*4480
  • Geography
  • Login

You are here

Home

Objective 2: Develop a GIS-based multi-criteria evaluation model for suitable habitat prediction

A multi-criteria evaluation (MCE) is developed and applied to determine areas of most suitable habitat for wolves within Algonquin Provincial Park. A method of protecting threatened forest species, such as the Algonquin wolf, is to preserve their habitats from disturbances that alter them, such as wildfires. Multi-criteria decision making is the process by which information from various criteria is combined to create a single evaluation index (Chen et al. 2010). MCE methods provide the technical tools to standardize, weight, and integrate habitat needs of species, which facilitates the development of habitat suitability models (Store and Jokimaki 2003). The MCE approach has been demonstrated to be an ideal method in cases involving site selection based on various criteria and constraints. For example, a MCE approach was successfully used in a study that selected the best landfill sites based on a variety of factors (Gorsevski et al. 2012), which is a very similar set-up to the one outlined in this study.

MCE analyses are incredibly useful to make spatially-based decisions. There are two commonly used methods to conduct a MCE: the creation of Boolean overlays which are then combined using logical operators, as well as weighted linear combinations which use continuous, standardized criteria that are eventually combined via weighted averaging (Jiang and Eastman 2000). This study employs the latter, using criteria and constraints to create the GIS model. Given the scope of the analysis presented in this study, it is beneficial to use a nested approach to first create a MCE model that defines suitable habitat based on the weighted environmental variables. Following which, raster calculator is used to multiply a hazardous forest types criterion layer by the MCE model results, creating a final output with adjusted suitability scores in order to isolate areas of suitable habitat that are at low risk of wildfire disturbance (Equation 3). This approach will provide the opportunity to compare the proportion of total suitable habitat available with viable habitat at low risk of loss due to wildfire.

 

Factor Standardization 

To begin, each criterion included in the MCE model is reclassified to create individual criteria raster maps. This process ensures that they are weighted appropriately and standardized prior to overlay. Before weighting, base layers are reclassified on a scale from 0 – 100, with higher values indicating greater habitat suitability (Table 2). This procedure also promotes uniformity of classification between continuous and discrete variables. Finally, the produced layers are standardized using a basic linear transformation. As discussed above, there are several criteria that describe suitable habitat. The additional hazardous forest type criteria layer is to be combined with the MCE model using raster calculator in order to delineate suitable habitat at low risk of wildfire.

Table 2. Classification scheme for factors.

Factor

Classification

Standardization Scale

Criteria

   

Proximity to Water

Euclidian Distance

100 - 0

Distance from Roads

Euclidian Distance

0 - 100

Land Cover Type

Coniferous Trees = 100

Wetland (Marsh, Swamp, Fen, Bog) = 75

Deciduous and Mixed Trees = 50

Other Trees (Sparse, Upland, Plantations) = 25

Other = 1

Turbid and Clear Open Water = No Data

100 - 0

Slope

Degree

100 - 0

Constraints

   

Water

Other = 1

Water = 0

N/A

 

Weighting of Criteria 

Following reclassification and standardization, the weight of each criterion is determined using a pairwise comparison. This is achieved using a nine-point continuous rating scale (Table 3). As discussed by Drobne and Lisec (2009), a pairwise comparison deals with the relative importance of two criteria at a time and is repeated for all pairs of factors. The most important factors for wolves are proximity to water, distance from roads and land cover type, followed by slope (Table 4) (Benson et al. 2015). 

Table 3. 9-Point rating scale.

1/9

1/7

1/5

1/3

1

3

5

7

9

Extremely

Very strongly

Strongly

Moderately

Equally

Moderately

Strongly

Very strongly

Extremely

Less important

     

Equal

     

More important

 

Table 4. Pairwise comparison

Factor

Water

Roads

Land Cover

Slope

Water

1

1

1

3

Roads

1

1

1

3

Land Cover

1

1

1

3

Slope

1/3

1/3

1/3

1

Sum

3.33

3.33

3.33

10.00

 

Table 5. Factor weights

Factor

Water

Roads

Land Cover

Slope

Total Weights

Water

0.3000

0.3000

0.3000

0.3000

0.3000

Roads

0.3000

0.3000

0.3000

0.3000

0.3000

Land Cover

0.3000

0.3000

0.3000

0.3000

0.3000

Slope

0.1000

0.1000

0.1000

0.1000

0.1000

Sum

1.00

1.00

1.00

1.00

1.00

 

Finally, once weights are determined for each identified criterion (Table 5), a MCE algorithm is developed (Equation 2) to model the most suitable habitat for wolf den selection based off the General MCE Equation (Equation 1). This completes the MCE model of habitat suitability and produces a supplementary map showing suitable wolf habitat within Algonquin park without imposed fire risk.  

Equation 1. Generic MCE equation. 

SUITABILITY = ∑WeightLayerCriteriaLayer

 

Equation 2. Habitat suitability MCE equation. 

SUITABILITY = ConstraintWater * [(0.3000)*(CriteriaWater) + (0.3000)*(CriteriaRoads) + (0.3000)*(CriteriaLand Cover) + (0.1000)*(CriteriaSlope)]

 

Fire Risk Development

Completing the final model is achieved by combining the standardized wildfire risk layer with the habitat suitability MCE output using raster calculator (Equation 3). The wildfire risk layer was developed from the Hazardous Forest Type Layer (Figure 12 in Appendix I) based on the standardization scheme presented in Table 6. The outputted raster layer adjusts the initial suitability scores based on the relative wildfire risk at the same location. For example, low risk areas with a value of 1 would not alter the scores in areas of satisfactory habitat as they would remain unchanged. However, the initial suitability score in areas of extreme risk would be multiplied by 0, effectively classifying them as unacceptable. This final model is utilized to observe adequate wolf habitat that is at low risk for disturbance. The completed model using the nested approach highlights shifts in distribution and abundance of viable habitat when wildfire disturbance is considered as a significant influence.

Table 6. Standardized hazardous forest type layer values
Factor Classification Standardization Scale
Hazardous Forest type

Low = 1

Moderate = 0.75

Pine (Needs Evaluation) = 0.5

High = 0.25

Extreme = 0

Water = No Data

1 - 0

 

Equation 3. Suitable habitat and wildfire risk layer equation.

SUITABILITY = MCESuitable Habitat * CriteriaHazardous Forest

 

Research Approach
Powered by CCS