The GIS model that was applied to each subset of data takes the general form of a multiple criteria analysis (MCE) or index model. The main approach involved converting the vector data to raster format and classifying the data sets either as binary data, ordinal data, or continuous data. This data fed the sub-models which compose the final MCE. The output of the MCE was then interpreted and effectively classified to create a suitability map that describes the level of protection and the possibility of allowing development across the areas governed under the Greenbelt Act.
Figure 1. Concept diagram of MCE process illustrating workflows.
The environmental model is composed of 3 separate sub-models comprised of ecological, hydrological and agricultural factors. Each sub-model was divided into groups of features that represent the same type of data and were processed in a similar manner.
All protected areas files (including Federal Protected Area, NGO Nature Reserve, and Provincial Park Regulated) were vector files. These files were converted to raster and reclassified as either a 0 or 1, representative of protection status. This ensured that all protected areas remain protected (0). Provincially tracked species was used as a proxy to track biodiversity at a 1000m cell size by first converting the vector to a raster format. The data was standardized and classified on a 1-10 scale representative of the number of species present in each cell. The Wetland and Wooded Area files were all converted from vector to raster. Their significance was then individually standardized on a 1-10 scale based on the relative area of each wetland, wooded area or wilderness area. (see Figure 2 below).
Figure 2. Natural Areas Sub-Model
Table 1. Pairwise comparison for natural areas to be able to achieve a single output by combining elements via raster calculator for further analysis. (*PTS: Provincially Tracked Species)
All hydrological features were given a 30m buffer and converted to raster. They were given the value 0 within buffer areas and 1 outside of buffers. A series of buffers was created at progressively increasing distances (100m, 200m, 300m, 400m, and 500m). The 100m buffer was based on the same scale as the 30m buffer (0 or 1) however, the other consecutive buffers were based on a scale of 0 for buffer areas and 2 for areas outside of the buffer. Raster calculator was then used to perform a simple addition of the buffer areas and generate a final layer that classified land based on the proximity to water features on a scale of 0, inside the 30m buffer, to 10, outside the 500m buffer. (see Figure 3 below).
Figure 3. Hydrological Sub-Model
The agricultural viability was assessed based on the Canada Land Inventory (CLI) index. This data came in vector format with each polygon represented on a scale of 0-7 with the addition of W for water and O for organics. Since 0 also represents organics, the O feature class was reclassified to 0. Since water also has no capability for agriculture, it was reclassified as 7 (Government of Canada, 2013). There are also a minimal number of polygons with no data which were classified as 7 to reflect no added weight for protection. The data was then reclassified on a standardized 0-10 scale using the raster calculator. (see Figure 4 below).
Figure 4. Agriculture Sub-Model
The environmental model incorporated the ecological, hydrological and agricultural data by combining the data in the raster calculator. Weights were assigned by pairwise comparison and multiplied through to provide an output with the same 0-10 scale. A reclassification of each layer was necessary before combining the data to ensure areas of no data had values (of 0) and could be added. (see Figure 5 below).
Figure 5. Environmental Model
Table 2. Pairwise comparison for environmental factors to be able to achieve a single output by combining elements via raster calculator for further analysis.
The policy data layers were ranked based on the level of protection and development allowable within each layer, as they correspond to the Greenbelt Plan, Niagara Escarpment Plan and the Oak Ridges Moraine Conservation Plan as well as supporting policy governing features such as urban river valleys. The ranking system allowed for an unambiguous assignment of weighting without involving a pairwise comparison. Instead, all the policy layers were ranked on a scale using the policy guidelines to assign weighting subjectively. The data layers were then converted to raster and reclassified based on their established ranking (0 being most protected and 10 being least protected). Indian Reserve and Greenbelt Specialty Crop were all classified as 0. (see Figures 6, 7 and 8 below).
Figure 6. Greenbelt Plan Sub-Model
Figure 7. Niagara Escarpment Plan Sub-Model
Figure 8. Oak Ridges Moraine Sub-Model
Since the areas of overlap in policy were ranked using the more specific policy, the lesser policy was given values of 0 in these areas. The mosaic tool compiled the policy data taking the maximum value, eliminating the complexity of using pairwise comparison and raster calculator. The output was based on the same standardized scale used to rank the policy layers (0-10 scale). A zero represented the highest level of protection requirements, and 10 represented the best available land for development. (see Figure 9 below).
Figure 9. Policy Model
An MCE model was chosen as the primary analysis method as it allows GIS software to manage a complex model with conflicting geospatial data (Carver, 1991). As many inputs are supporting an array of different land uses and features, an MCE forms the body of the decision-making process (Carver, 1991). Standardization of data allows qualitative and quantitative data from different sources to be compared (Carver, 1991). The MCE was broken down into sub-models which preprocessed the data into more meaningful and manageable categories. This prevented complications and allowed for a more straightforward determination of weighting by comparing smaller groups of related factors (Ross, 1988). Both binary and index models were applied at the sub-model scale as a conventional method of determining land surface suitability. The binary models were used for all constraint layers to classify land as absolutely protected or not (Eastman, 1999). Index models were used to evaluate the ordinal data and form a ranked output (York University, n.d.). The sub-models were then combined into the Policy and Environment Models which were further combined in a simplified MCE to produce a final output (Ross, 1988). (see Figure 10 below).
Figure 10. Final Model
The final model was then overlaid across municipal boundaries and zonal stats were produced. Using the mean values, an Incremental Spatial Autocorrelation was performed and the peak z-score distance was used to complete a Hot Spot Analysis (Getis-Ord Gi*).