• Soil and Vegetation layers
  

  There may be human errors associated with our soil and vegetation layers because it was a subjective part of our analysis. By ranking the drainage efficiency of the vegetation and soil types from 1-7, we discounted the magnitude of importance between each soil or vegetation type within the soil and vegetation layers. If more research into how each soil type or vegetation cover compared to one another could be done, this would improve the accuracy of our results. Also, we could not take all the confounding factors into account to predict soil and vegetation effects on flood risk. For example, although Dunelands are very permeable to water, a lack of vegetation could increase risk of floods.

• Drainage density and precipitation layers
  Drainage density only includes rivers that are visible at 1:1,000,000 scale, which means that there are a few exclusions of small streams and rivers that can potentially affect the results of our drainage density calculations. The scale of 1:2,000,000 streams data was used to attempt the analysis but due to the large volume of the data, the ArcGIS software was unable to load fast enough. In addition, we assumed that the precipitation for each subbasin to be the sum of the annual precipitation rate for each of the raster cell in the subbasin. This was done to account for the flow of water from the mountains to the accumulation areas down in the flat zones. However, by calculating the total annual precipitation and drainage basin density for each subbasin, uncertainty was added to our results because this decreased the spatial resolution of our analysis.
  

• Accumulation area
  

  For the accumulation layer, we assumed that the top 97% of the accumulation area to be prone to flooding. This number was determined by examining the map and trying different reclassification values. 97% was found to be the most reasonable number as it gave most of the areas of accumulation in the low-lying regions of our study area. At the same time, this posed an uncertainty in our analysis. Further research needed to be done in order to have a more reasonable percentage during reclassification. In addition, the uncertainty also occurs when the operation of buffers surrounding the accumulation areas. We have assumed a distance of 5km for the buffer but it is totally arbitrary and further investigation needs to be done in order to produce a more valid result. The resolution is also a source of error for the analysis of this layer. Since the the accumulation area is such a small portion of the study area, a smaller resolution (rather than a cell size of around 1044m) should be used to produce a more accurate result. Finally, the conversion of this layer back and forth between raster and vector data (for the buffer operation) might have also posed some errors in our analysis.

• Wetland
  

  As shown on the map (see the "Method" section), wetlands usually occur where the rivers are. However, with respect to the drainage density criteria, it should be the subbasin that has the higest abundance of rivers and streams that are more likely to flood. Therefore, we suspect that the inclusion of the wetland into the MCE analysis contradicts the fact that floods tend to occur near rivers and more research needs to be done to present a more reliable result.

• AHP

• Results 

In order to quantify the differences when the priority of flood-risk factors are changed, we have preformed a sensitivity analysis on the effect of changing the weight delegations of our risk factors. Our approach to this analysis was based on correlational factors that are likely to affect flooding. Perhaps future flood analyses could minimize errors by applying hydrological models that are more process based rather than rely on correlational factors.

The four maps for the sensitivity analysis are present below (Figure 11):

Figure 11.  Sensitivity analysis results
 a.  wtd_accu   b. wtd_soil    c. equalwtd    d. wtd_precip


Figure 12. MCE result map

Table  6.   Sensitivity analysis summary table