Several assumptions were made in the creation of this MCE model, most notably involving the vegetation data, land use data, and the weightings each parameter was assigned. In addition to these assumptions, we cannot rule out possible sources of error that may be associated with the assemblage of our model, and the data used.
VEGETATION DATA
As described in detail in the Methodology page under Parameters, data missing from the vegetation data set provided by DMTI Spatial Inc. were assumed to be areas of no vegetation. These areas were therefore assigned a value of 1, corresponding to high risk of landslides. In order to remain conservative in our landslide risk evaluation, areas that were designated as “cloud shadow” were also assigned a high value, as the data was inconclusive.
LAND USE DATA
The data set obtained from HectaresBC for Land Use: Baseline Thematic Mapping was collected in 2006, with an unknown date of modification. We must assume for the purposes of our analysis that this is the current region of urban, agriculture, residential agriculture mixtures, and recreational activity areas in the West Vancouver- Sunshine Coast- Sea to Sky district. In reality, areas such as West Vancouver, Squamish and Whistler have since greatly expanded the areas residential and recreational activities. Therefore, there may landslide susceptible regions with direct human implications that are not identified in this study.
WEIGHTINGS
The multi-criteria evaluation was created by weighting various factors of landslide susceptibility. Factor weights were taken from a multi-criteria analysis evaluated by Shahabi & Hashim (2015) to determine landslide susceptibility using GIS and remote sensing data in Malaysia. The study focuses on tropical rain forests in Malaysia, an ecosystem different than the temperate rain forests of B.C. It is assumed that the factors contributing to landslide occurrence are similar for all regions, and can be applied to this study as well. Vegetation types would vary between regions, however the presence and absence of plant cover would have the same implications on soil cohesion. Heavy precipitation occurs in both regions and therefore should not greatly affect the weighting distribution. In addition, slope and proximity to fault lines should not vary based on these two locations.
Due to the limited collection of data sets available and limited scope of our analysis, five of ten variables analysed by Shahabi and Hashim (2015) were used to assess landslide susceptibility. The variables analysed in our project include slope, precipitation, vegetation cover, bedrock geology and proximity to fault lines. Variables included in determining landslide susceptibility in Shahabi & Hashim (2015) that were excluded for the purpose of our study included soil, land cover, aspect, distance to road, and distance to drainage. This may lead to a greater dependence of each variable in determining landslide susceptibility in our analysis in comparison to physical landslide occurrence in the region. However, although the weightings themselves may differ from the 2015 Shahabi & Hashim study, the order in which parameter was ranked remained comparable.
POSSIBLE SOURCES OF ERROR
Numerous data sets used in the analysis are of different geographic coordinate systems (Table 1). Each data layer was transformed and projected onto NAD83 to align the data for the purposes of analysis. This approach is sufficient to map and display data, however it may cause slight differences in results. Edits and analysis are generally applied to the geographic coordinate system, and therefore may cause slight variance in landslide susceptibility distribution. However, our area of study is quite small and therefore would show less variance between coordinate systems.
Vegetation and bedrock geology data were converted from polygon to raster format. This may have caused errors during analysis as polygons were redefined into raster cells. Areas on the perimeters of polygons may be generalized into a different category depending on the percentage cover of raster cell.