Methods

The area designated for sampling includes East Side of Vancouver bordered by Boundary Road on the east and Main Street on the west. The study area denoted by "E" in the middle is about 115 square km and appears less green on satellite imagery due to its lower green cover compared to Vancouver West Side.

Study Area

GIS layers used in this study were public .shp files available from the City of Vancouver VanMap (http://data.vancouver.ca/datacatalogue/index.htm ).
Detailed descriptions of the zoning districts were obtained from the City of Vancouver (http://vancouver.ca/home-property-development/map-of-zoning-districts.aspx).
Census of Canada 2006 was the source of socioeconomic data. GIS data detailing family income per dissemination areas for the study region was obtained from UBC Department of Geography (GVRDda06.shp).

Data source

Geoprocessing was performed using ArcGIS 10.1. Vancouver boundary and zoning layers were overlayed. Polygons from east and west of the Main street were clipped. New layers were then created from each. The polygon centroids for different zones were calculated and displayed as a layer.

The zoning categories were converted to represent more general categories. The categories decided upon for the purpose of this study include C: Commercial, CD: Comprehensive Development, I: Industrial, RS: Single Family Dwelling District, RT:Two Family Dwelling District, RM: Multiple Family Dwelling District

Zoning categories with very few occurrences such as HA: Historical sites were ignored.

Geoprocessing

In each zoning polygon, the polygon centroid was taken as the centre of the core plot. Per each polygon, 3 random points were generated with minimum allowed distance of 37 m from the core plot as the subplot centres.

Randomization template:

import arcpy

arcpy.CreateRandomPoints_management("OUT_PATH","OUTPUT NAME","CONSTRAINING FEATURE CLASS","CONSTRAINING EXTENT", "NUMBER OF POINTS", "MINIMUM ALLOWED DISTANCE", "MULTIPOINT", "MULTIPOINT SIZE")

Following (17), 400 square m plots were adopted as the sampling units. To achieve this, circular buffers with the radius of 11.28 m were created around the plot centres.

Each zone was considered as a stratum. Subselecction of the plots to form the sampling frame was done based on proportional allocation (21). In doing so, weights were calculated in respect to the number of polygons in strata:

Wg= Ng/N where Wg is the weight for each stratum, Ng is the number of polygons in each stratum, and N is the total number of polygons across strata.

As a general rule, 200 plots in a stratified random sample in a city will yield a standard error of about 10% for an estimate for the entire city (i.e., tree cover in the city) (19). Given the total number of samples n, minimum sample per stratum is obtained as ng = Wg*n.

Quantifying the vegetation cover (response value) in each plot, one can use sample mean as the estimator of the population mean (i.e. vegetation cover, or inversely the remainder or potential area available for planting trees).

Establishment of the plots

Sampling frame

Inclusion of additional variables

Canada Census 2006 data represented in GVRDda06.shp was joined with East Side and West Side Vancouver layers. A subset of residential plots (n= 30) from each side was randomly selected. The three residential categories RS, RT, RM were equally represented in the sample, and objects with the same DAUID were avoided. As the population shows unequal variances for family income per dissemination area between East Side and West side, two sample t-test with the assumption of unequal variances was used to compare the average family income between the plots of East Side and West Side.