Data used
The data we used for our analysis falls into two categories. The data that was used to represent the fuzzy factors that went into the MCE and the constraints that were used in the second part of our analysis.
In terms of MCE factors, slope, elevation and aspect were determined based on simple analyses of a digital elevation model (DEM). The DEM was also used in combination with a population centers shapefile from the Canadian Government in order to create the viewshed layer. Distance from transmission lines and distance from roads were both determined from individual shapefiles using a distance accumulation analysis. For these, the DEM was used as an “input surface raster”, so that the distance would be representative of the topography. The distance from water bodies was determined from a water features shapefile, and similarly the distance from coastline was determined from a shapefile of the coast. For both of these, a distance accumulation analysis was once again performed but the DEM was used as an “input horizontal raster” so that output was representative of the distance away from water with respect to water vapour. Using a two-dimensional distance accumulation made more sense here as the interior of BC should be represented as further from the coast than the tops of the Coastal Mountains. Finally, the potential photovoltaic output layer was downloaded as a raster layer.
After determining the general region to be focused on for the second portion of the analysis, a land use shapefile was downloaded as well as a layer of Indigenous reserves. The water features layer was also used again as a Boolean constraint. Finally, a geodatabase of endangered habitat was downloaded, and the habitat’s that coincided with the study area were selected and clipped to the study area.
Analysis overview
This analysis was performed in two stages. A provincial wide MCE was conducted first in order to determine general suitable areas for a solar farm. Subsequently, a localized Boolean overlay analysis was completed to eliminate areas where a solar farm was not able to be constructed.
MCE factors
As outlined in the Data used section, the MCE factors were (in order of importance) potential photovoltaic output, distance from transmission lines, viewshed, aspect, slope, elevation, distance from roads, distance from the coast, and distance from water bodies. These factors were determined based on a literature review (Figure 1). In order to assign weights to these factors, we used the AHP model tool available at 123ahp.com, and based our decisions off the literature (Figure 2) and our knowledge about the context of British Columbia. The output from the AHP model had a consistency ratio of 0.0582. The weights are also outlined in Figure 2.
Figure 1.
Figure 2.
After we created raster layers for all the MCE factors (explained in the Data used section), we then normalized them, so that they could be compared to each other using the weights from Figure 2. To normalise the rasters, we used the Fuzzy Membership Spatial Analyst tool. A linear membership type was used for all the normalizations except for aspect and viewshed. We selected a linear membership type because we did not believe that the statistical shape of the data should influence the weighting of individual areas. For aspect, a Gaussian membership type was used due to the nature of the data being ideal at the middle of it’s range (South) and then getting progressively more inadequate in either direction. A spread of 0.0001 was used in order to create a distribution of data (Figure 3) that we felt reflected the aspect values that were appropriate for a solar farm. For Viewshed, an MS Small membership type was used, because the linear membership type discounted the viewshed of all but the largest urban areas more than we felt comfortable with.
Figure 3.
Once all the data layers were prepared, the MCE was completed using the Weighted Sum tool. Weighted Sum multiplies each raster layer by its respective weight and then sums them together. After completing the MCE, a select by attribute command was used to select only the top 20 percent of suitable areas. This subset of the suitability raster was then used to determine where to focus the second portion of the analysis.
Constraints
Given the general region selected for the second portion of the analysis, the most important forms of land use that required consideration were agricultural land and urban and suburban development. After being clipped to the study area, a select by attribute method was used on the land use shapefile to select polygons within any of the classes: bare land strata, building strata, common ownership, crown subdivision, parks, and subdivision. Although some of the names imply they may be suitable land use types for a solar farm, the data was explored in depth and it was determined that even types such as “bare land strata” existed within developed areas and were therefore unsuitable for a solar farm. Through the selection of all these classes, agricultural land and urban and suburban development areas were all included. As the land use layer was a land parcel map, every property was a unique polygon and none of them were touching. This was solved however when the layer was buffered 500 meters to match the suggestions found in the literature (Noorollahi et al., 2016), and then dissolved into one layer (as individual property information was not necessary).
The shapefile of Indigenous reserve land was also clipped to the study area and buffered 500 meters as well before being used as a constraint.
Due to the environmental impact of developing alongside streams, rivers, or lakes, the water features shapefile was also used as a constraint. The layer was first clipped to the north-east BC study area before being buffered by 500 feet to create a riparian zone (Hawes & Smith, 2005).
The three endangered habitat layers that coincided with the north-east BC study area were: boreal population caribou, southern mountain population caribou and northern myotis bat habitat. To align with the literature (Noorollahi et al., 2016), these were all buffered 2 kilometers before being used as constraints.
The land use, riparian zones, Indigenous reserves, and critical habitat layers were then overlaid on the north-east portion of the suitability subset raster as Boolean constraints. The constraints were also merged into one layer, which was erased from the study area. The resulting layer, which represented all areas without constraints, was then used to clip the suitability subset raster. This final output represented all of the areas that were in the top 20 percent of suitability within the north-east study area that were not affected by the constraints. In order to calculate the total area left, the raster was converted from floating type to an integer raster using the Raster Calculator and Int tools. It was then converted to a polygon (without altering the geometry or shape whatsoever) where the whole thing was dissolved together and a field was added and the area calculated using Calculate Geometry.