We can see from the figures that the ALR areas identified as most prone to climate change and agricultural runoff are the mountainous areas (Figures 7-9), where precipitation trends are highest. However, they should not be taken into account since we are only interested in the low, flat areas suitable and reserved from agriculture. Near the southern US-Canada border under Hope, there is a region of ALR with the Highest Risk value from all 3 of our MCE analyses (Figure 7-9). We do not know if it is currently being farmed but would advise checking in to see its status. If it is farmed, we suggest a relocation of the farmed area, or a limit of fertilizer and pesticide application. Because the Sensitivity Analysis (Figure 7) also gives Highest Risk values for that area, we know that all of the 5 criteria play a role in determining the level of runoff risk.

There are also relatively high risk values for ALR land near Agassiz, Laidlaw and north of Abbotsford in Mission (Figures 7-9), therefore respective policymakers and farm authorities should consider altering their management of fertilizer and pesticide use. The higher risk of these areas is likely due to the higher precipitation trends from our preliminary calculations (Table 1). There is unusually high risk values around the Sumas Canal area, and while there is little documentation and studies done on the agricultural practices for farms in this regions, we can assume that this divergence can be explained by the regions’ proximity to the numerous water bodies in this region (Figure 3). Anticipated trend in precipitation for this location is minimal, therefore climate change is not the main factor in determining the high risk of agricultural runoff in this region, but rather explained by the region’s topography and proximity to lakes.

For our Weighted MCE analysis (Figure 8), we chose to give precipitation the highest weight since it is the only factor that is changing (Province of British Columbia, 2015) and is our main focus of the analysis. The changes in other factors (slope, TWI, proximity to river and lakes) are negligible since they are generally in a steady-state and the changes only become significant on a long time scale (millions of years). With the onset of climate change and changes in weather patterns, our primary concern should be precipitation patterns, as the other factors evaluated in the MCE are not changing much throughout time and should not pose any additional effect on agricultural runoff, if precipitation was constant. But it’s not, and therefore we must evaluate other attributes to the farmland such as topography of the land and location near water bodies in assessing the additive risks.

Comparing the two MCE analysis, we can see that areas with higher risk in general increased in the Weighted Sum scheme (Figure 7) than in the Sensitivity Analysis scheme (Figure 8). These changes must be due to the influence of factors whose weight is higher in the Weighted Sum analysis, such as precipitation and TWI. That means that one of these factors for sure will have a big impact on agricultural runoff.  Subjectively, the factors in our MCE could be weighted differently, however for our analysis, we are only concerned with one scheme which we think is the most appropriate, given our rationale and available literature. Combining the Weighted Sum MCE and the Sensitivity Analysis in the Combined MCE map, we can observe a more realistic projection of agricultural runoff risk in the ALR of the Lower Fraser Valley region as uncertainties with the weights are reduced (Figure 9).

As environmental data tend to be hard to collect and analyze, there are potential sources of error associated with our data:

• Precipitation data is subjected to collection technology errors and analysis errors. We are not sure how the data was collected, but every instrument has uncertainty associated with it. Also, there may have been a switch in measuring instruments since new technologies are developed and replaces older ones, as well as the addition of more weather stations and measuring instruments throughout the 40 year period of our study. There were also some missing precipitation data which we chose to ignore, since we do not think it will have a big effect on the long-term trend. Another important assumption is that future precipitation trend will be a continuation of historical trend.
• There could be interpolation errors due to missing data from northern weather stations and not incorporating data from outside the study area. We have very limited number of weather stations which have a continuous 40 year data in precipitation, so we can only chose the ones that has relatively complete data, which limited our choice of weather stations.
• We treated all ALR areas as farmed areas, but in reality, not all ALR areas are farmed, or farmed for the same purposes. However, they cannot be used for industrial or residential development either, so we think it’s a fair assumption to use ALR as proxies for farmlands.
• Not understanding the physical characteristics of the lakes or streams, and assuming that lakes are more affected by agricultural runoff than streams (Motew et al., 2017).
• Uncertainty with the AHP decision rule and the MCE analysis itself.