Error & uncertainty
Resolution is of course a source of uncertainty, specifically the resolution of the DEM which the slope, elevation and aspect layers were based on. It also contributed to the creation of the viewshed layer, distance from roads, distance from transmission lines, distance from coast and distance from water features layers. Due to its widespread use, it is of specific importance in this discussion. The resolution of the DEM was 30 arc seconds (approximately one kilometer).The resolution of all other layers used during both stages of analysis is also important to consider. Our analysis is only as good as the data used, so we did our best to find data that was of a good standard. This was ensured by reading metadata and comparing some things (such as land use) to google maps satellite imagery.
Another source of error and uncertainty are the weights used. The weights in an MCE come down to a normative decision, although we attempted, to the best of our abilities, to base our decisions off our literature review (Figure 2). However, whilst we based our decisions off the literature, none of the studies used were looking at the context of British Columbia, meaning we had to assess the relative importance of each factor in the context of its respective study and in the context of our analysis. Additionally, each of the studies relied on a different set of factors which fit their unique context and data availability, as did our analysis. This meant that once again we had to make judgements based on the relative weights used within studies and between studies. However, despite this, using the AHP model tool available at 123ahp.com we had a consistency ratio of 0.0582, which reflects our efforts to think through the relative weights of factors thoroughly (Figure 1). Furthermore, we completed a sensitivity analysis with respect to the weights to be certain they would not jeopardize our results entirely (Figure 9).
Some error and uncertainty will have also been introduced during the process of normalizing our data. However, we attempted to suit the normalization method to the data as best as possible so as to minimize error and uncertainty.
Furthermore, some areas along the coastline were not included in the output from the MCE analysis due to not being included in all data sets (Figure 4). However, these were likely not to be the most suitable areas anyway given the surrounding area.
In the second part of our analysis error and uncertainty will have been introduced based on all our decisions with respect to buffer distances. Although, once again, we based our decisions in the literature in an attempt to minimize normative or subjective decisions.
In deciding where to focus for our second portion of the analysis, the threshold of 20% was selected more or less randomly. After looking at multiple thresholds, 20% was used because it clearly highlighted one large region of BC as containing the most area of suitable locations, which is what we were looking for (Figure 5). Again, this decision introduced some error or uncertainty as we could have used a different threshold. Furthermore, even using the 20% threshold, there were still other areas that were of high suitability outside our selected study region, however, we decided to focus on the north-east due to the large amounts of suitable area in the region. Therefore, this study is by no means entirely conclusive. Other analyses may have made different decisions along the way, and came up with other suitable regions.
Sensitivity Analysis
Having completed our analysis, we concluded that the stage where the most uncertainty or error may have been introduced was in the assigning of weights. As such, we performed a sensitivity analysis to make sure that the weights chosen did not jeopardize our results entirely. For our sensitivity analysis we reran the Weighted Sum tool with all the layers weighted equally. Then we compared the two maps at both a provincial wide scale and within the study area (Figure 9). We concluded that the weights we had chosen did not affect our findings, as the north-east area of BC was, respectively, more suitable when factors were weighted equally in comparison with the rest of the province.
Figure 9.
Discussion of initial provincial wide MCE results
The general trends for our suitability output layer were largely what is to be expected for the province of British Columbia (Figure 4). Both the Coastal and Rocky Mountains having relatively low suitability is to be expected due to orthographic effects and the difficulty of the terrain. The general trend of decreased suitability moving north (excluding the eastern side of the Rocky Mountains), is also to be expected due to the latitude. Despite the high latitudes of the north-eastern (east of the Rocky Mountains) section of the province, it contained much of the most suitable area. Although we did not expect it to be unsuitable, we were surprised at how suitable the region was. Especially after taking the subset of the most suitable 20 percent of areas (Figure 5). This speaks to the importance of this type of analysis, as the results are not always what is expected. This also speaks to the success Alberta could have with solar power.
Discussion of local results
Whilst all the constraints included had some effect on the available suitable areas in the study region, they did so to differing degrees (Figure 7). The land use (including urban, suburban, and agricultural land) associated with Fort St John, Dawson Creek, and the surrounding area eliminated a major portion of the more southern cluster of suitable land within the study area. This reflects the importance of considering the best use of different land as areas may be suitable for many forms of development. The northern cluster of suitable land was constricted mostly by the range of the Boreal Caribou. Although we applied a large 2km buffer, as indicated by the literature, it is worth noting that compared to the other constraints (which have more definitive borders such as land use), it is much harder to determine species ranges. This applies to the ranges of the Boreal Caribou, the Southern Mountain Caribou, and the Northern Myotis, as well as the water features layer as its buffer distance is largely based on habitat considerations. Furthermore, when embarking on any development projects, the shifting ranges of animals due to climate change should be considered as well as the ranges of other animals who may become endangered.
Whilst Indigenous reserves were used as a constraint, we acknowledge that British Columbia is the traditional, ancestral and unceded territory of many Indigenous nations, and as such construction of a solar farm should not be undertaken without the consent of the nation whose land it is to be on. Furthermore, one of the many benefits of solar power developments is that they can very easily be built by or in partnership with First Nations. The Tŝilhqot’in Nation finished their first solar farm in 2019 bringing not only clean power to their community, but jobs and new economic opportunities (Ecosmart, 2014).
The 5029 square kilometer region of the subset of the 20% most suitable land that was left after eliminating the areas of the constraints is still a very large area (Figure 8). For reference, what would be the world’s largest solar farm in the world has been proposed in Australia, and if it gets built it will cover 120 square kilometers (Dockrill, 2020) making it far and away the largest solar farm by area. Furthermore, most solar farms are much smaller. Therefore, whilst our goal was to find suitable locations for a solar farm in BC, further analysis would need to be done to determine the exact location within the area resulting from our analysis. Although our subset suitability raster indicates which parts of the remaining area have higher or lower suitability, so it could still be used even within a smaller scale analysis.
It is also worth noting that having looked at only the most suitable 20 percent of land means that in actuality there is much more area that could be used, if any of the problems associated with the factors used in our MCE can be resolved. Furthermore, we did not look into other areas of the province that did fall into the most suitable 20 percent, so other regions may be able to build solar farms as well (Figure 5).
Final conclusions
Given that hydro power cannot be relied on to match future energy consumption in BC, especially taking into account the provincial government’s plans and goals, this analysis is pertinent as the potential of solar power in the province has not been realised. Having located over 5029 square kilometers of highly suitable land, BC clearly has the potential to increase the contribution solar power makes towards meeting the province’s energy demands.