Literature

Geographers Hwang and Smith (2010) discuss how spatial profiling is central to alleviating food insecurity as we analyse the factors impacting access to food both within and between communities (p.47-48). Extensive research on food deserts has been performed in the United States and researchers have even used participatory GIS to map food networks (Eckert and Shetty, 2011, p. 1218). Using participatory GIS can be beneficial as it combines qualitative and quantitative data to represent access to food (Hwang and Smith, 2012, p. 49). Visualizing and analysing accessibility has been quantified through the number of grocery stores, buffers, cost surface and distance analyses (Eckert and Shetty, 2011, p.1219; Burns and Inglis, 2007, p. 879)

In Eckert and Shetty (2011), they call upon McEntee and Agyeman to suggest that we move away from using network analyses in favour of distance and proximity analyses (p. 1219). One reason for this claim is that standard measures of accessibility allow researchers to compare different areas with each other (Eckert and Shetty, 2011, p.1219). Distance analyses have previously been conducted with a range of intervals from 500m, to 800m and even 10 miles to quantify food accessibility (Eckert and Shetty, 2011, p. 1219). Kibetu (2020), Eckert and Shetty (2011) performed distance analyses in their studies to measure physical assessibility.

Location Quotient (LQ) Calculation

The LQ is a statistic that was calculated using a python notebook and census data. The code created a LQ field and computed the value simultaneously. LQ was used to determine where the serviced population is located in Vancouver. The LQ value was used to determine subpopulation hotspots. A higher LQ value indicated that the population in a DA is higher in proportion to the city. A LQ value of 2 indicates that the DA has a population twice as large as the average for the city.

The LQ formula used was: (# of __ subpopulation in DA/ total # of __ subpopulation in Vancouver) / (Total Population in DA/ Total Population in Vancouver). The code block below is an example of what was used to calculate the LQ for Youth in Vancouver. The code was edited to calculate the LQ for each subpopulation: women, youth, seniors, low income and Indigenous.

censusUnit = "VancouverDA" 
arcpy.AddField_management(censusUnit, "LQ_Youth", "Float"
arcpy.CalculateField_management(censusUnit, "LQ_Youth",'(!LQ_Data.COL6! / 143405 ) / ( !totalpop.COL4! / 633138 )', "PYTHON3")

The LQ field for each subpopulation was symbolized using a reverted red-blue colour scheme. Looking at the histograms determined the appropriate classification method which for most, was Natural Breaks. Low Income was classified using Standard Deviation. The resulting LQ map was overlaid with food programs that serviced the particular population to see where they are located in relation to the target population.

Proximity Analysis

Using the “select by attribute” tool, I was able to query for DA hotspots for each subpopulation from the LQ maps. I queried using the minimum value of the highest class and exported the queried features to a new layer. For low income, I used the top 5% of the DAs as it was classified using standard deviation. To prepare for the distance calculation, I converted the DA polygons to a single point at the centre of where the DA was using the “feature to point” tool. The point would represent the average distance the DA would travel as using the polygon would show the shortest distance the boundary of the DA. The “near” tool was used to calculate the distance between the centre of each DA hotspot to the location of food assistance programs. The DA hotspot point was the input and the near feature was the location of the food program that services the subpopulation. The analysis created a new field in the hotspot layer with the distance in meters. I joined the hotspot point layer back to the LQ map and symbolized the distance using a green-red classification on top of the Vancouver DA file.

Scatterplots of LQ versus Distance 

The scatterplots are useful for visualizing the relationship between where there is a high population and its relative distance to the food program. Distance in meters was plotted as the independent variable and the LQ values are the dependent variable. An ideal scatterplot would indicate a negative relationship, where high LQ values correlate to shorter distances as that would indicate that high concentrations of populations are close to the food assistance program.  We can also determine the distribution of food programs in relation to the population, so the approximate distance that most food programs are away from the population.

Buffer Analysis

Another method to describe access was through the use of buffers. I used the “multiple ring buffer” to create a 500m, 1000m and 1500m buffer zone around the DA point. With the buffer, I am able to calculate how many hotspots are within 1500m radius of a food assistance zone and a percentage. This is different from the distance analysis as the proximity analysis only returns the shortest distance to any food program.