Site Description
The Fraser Valley Regional District is located in the southeastern region of British Columbia, Canada, bordering the State of Washington to the South. It has a total area of 13,361.74 squared kilometres with 6 municipalities: Abbotsford, Chilliwack, Hope, Mission, Harrison Hot Springs, and Kent, and 7 electoral areas. The Agricultural Land Reserves constitute 71,865 hectares (BC Ministry of Agriculture, 2015) in the Fraser Valley Regional District, protecting the land from non-agricultural uses (Provincial Agricultural Land Commission, 2017).
Figure 1. Agricultural Land Reserves (bright green) in the Lower Fraser Valley from Google Earth (http://www.fvrd.ca/ and Google Earth).
FVRD Agricultural Sector & Climate
The Fraser Valley Regional District has multiple conditions favorable for agriculture: proximity to large markets, high quality soils for growing crops, access to water, and mild climate for a moderate growing season; with conditions like these, it’s no wonder the FVRD ranked as #1 on BC’s gross farm receipts in 2005 with $921,425,274 (FVRD, 2006). Taking this into consideration, we understand the economic benefits of agricultural practices in the FVRD, and the reliance of the Canadian economy on the productivity of the FVRD for feeding its growing populations and increasing demands. However, with climate change, we are anticipating changes temperature and precipitation patterns over the world, therefore we developed our project to identify ALR areas more prone to contributing agricultural runoff in the next 40 years, based on predicted precipitation trends in the region and the topographic and geographic characteristics of the ALR land. We hope that our project findings will be a framework for farmers and municipal authorities alike to become more aware of the risks of certain areas of ALR land in contributing agricultural runoff, given the precipitation intensities, topography and geography of the region. We also hope that our project raises public awareness on the severity of agricultural runoff on aquatic ecosystems, water quality, and human health, especially in a world with growing food demand and a changing climate.
For our analysis, we obtained and transformed data from the following sources:
DEM (GeoGratis Canada, 2017)
- Converted from tiff to raster
- Calculated TWI raster from DEM
- Created Slope raster from DEM
ALR Polygons (Data BC, 2017)
- Proxy for current and future farmlands in the FVRD
- MCE analysis on ALR areas prone to runoff
Streams & Lakes (Data BC, 2017)
- Used Euclidean Distance tool to calculate proximity to streams and lakes
Precipitation Data (Environment Canada, 2017)
- Gathered from 7 weather stations in the Lower Fraser Valley: Abbotsford, Mission, Chilliwack, Hope, Agassiz, and Sumas Canal
- Calculated the fall trend from 1967 to 2007 for a 40-year period (Sept-Nov)
- Interpolated trend across the study area using Spline tool
We performed a Multi-Criteria Evaluation analysis on the following parameters:
- Topographic Wetness Index / TWI (given a weight of 0.245)
It is a steady-state wetness index that do not change overtime, since topography changes very slowly. We first created a flow direction raster and slope raster from our projected DEM (from Geogratis Canada). Then, from the flow direction raster, we created a flow accumulation raster, and used the following function to calculate TWI:
Ln ((“FLOWACC”*900) / Tan(“SLOPE”))
Figure 2. Topographic Wetness Index raster prior to smoothing or standardization of the values.
Because the TWI raster is naturally sensitive to topographic abnormalities, it produces a rather “grainy” distribution (Figure 2), therefore we used the Focal Statistics tool to smooth out any discontinuities. Higher values for the TWI were deemed as more prone to contributing to agricultural runoff. We then normalized the TWI raster using Linear Fuzzy Membership
2. Slope (given a weight of 0.185)
First, we created a Slope raster from the DEM using the tool “Slope” in Arctoolbox. Then, we normalized it using the Spatial Analyst > Map Algebra tool by typing in the equation: “SLOPE” / 84.5 (maximum value of slope). The highest slopes would get values closer to 1 and visa versa. This makes sense because we want to identify the areas with higher slopes, where runoff speed will be greater due to gravitational pull (Borghi, 2015). Due to farmlands typically being situated on flat ground, we decided to place more importance on the Topographic Wetness Index rather than the slope as agricultural lands tend to be on low slope.
3. Precipitation Trend (given a weight of 0.322)
Precipitation data from 1967 to 2007, a 40 year period, at each of the 7 weather stations were gathered from Environment Canada and their trend was analyzed using Excel (2017). We only used data for September, October and November since those are the times where, after fertilizer application in the summer (because it’s the most suitable season for plant growth), nutrient runoff is going the be the greatest. The monthly precipitation was averaged over the 3 month period and a trend was extrapolated using Excel for cm/ month / year increase. We then interpolated the average precipitation trend to the FVRD using the “Spline Interpolation” tool and normalized the values using the Linear Fuzzy Membership tool. Since the weather stations we chose are spread out evenly across the study area, we think the interpolation would provide a good estimate of the actual precipitation trend across the region.
4. Proximity to water bodies and watercourses (given a weight of 0.141 and 0.107, respectively)
As runoff is the leading source of nonpoint source pollution to rivers (watercourses) and lakes (water bodies) (EPA, 2005), we identified the proximity of the ALR lands to the streams and lakes vectors, using the “Euclidean Distance” Tool which calculated the shortest path distance from any cell to a stream or a lake (Figure 3). The stream and lake data was retrieved from Data BC (2017). We normalized these two euclidean distance rasters using the Linear Fuzzy Membership. However, we assigned higher importance and a larger weight for lakes rather than streams due to the poorer circulation in lakes than rivers and potential sedimentation in lakes, leading to reduced sunlight penetration, depleted oxygen and harmful algal blooms (EPA, 2005).
Figure 3. Euclidean shortest path distance from every unit to a lake polygon. Lower values denote proximity to lakes. We can see that the western regions have greater numbers of lakes.
The different variable weights were determined using the Analytic Hierarchy Process (http://www.123ahp.com/Default.aspx) where we assigned the following weights: TWI = 0.245, Slope = 0.185, Proximity to Streams = 0.107, Proximity to Lakes = 0.141 and Precipitation = 0.323, which we later on changed to 0.322 to make the weights equal to 1.000 for the MCE analysis (Figure 4).
Figure 4. Flow diagram of the Analytic Hierarchy Process for the 5 criteria: TWI, Slope, Proximity to Streams, Proximity to Lakes, and Expected Precipitation Trend.
Figure 5. Evaluation of the importance of the five criterias considered in the MCE using the Analytical Hierarchy Process (https://bpmsg.com/academic/ahp.php).
A Sensitivity Analysis was also performed prior to the MCE using equally weighted variables and assigned a weight of 0.2 to all 5 variables. We then combined the Sensitivity Analysis and the MCE with different weights and graphed a Combined MCE Analysis. For all the MCE Analyses, we reclassified the values into the 5 classes: lowest risk, low risk, medium risk, high risk, and highest risk (Figure 6).
Figure 6. Classifying the break values for the Sensitivity Analysis, Weighted Sum, and the Combined MCE Analysis, respectively.