I downloaded Landsat 8 OLI data from USGS Glovis and created a variety of images using different band combinations in ArcGIS Pro. The intent was to help identify vegetation in Vancouver, so band combinations that highlighted green space were of particular interest. To determine the overall greenness of Vancouver, I used the Landsat 8 OLI data to compute the normalized difference vegetation index (NDVI) for the city using ArcGIS Pro’s NDVI tool. To help corroborate the NDVI analysis, I looked at the following band combinations:

6-5-4 Vegetation Analysis

4-3-2 Natural Color

5-4-3 Color Infrared (vegetation)

I then computed some spatially constrained multivariate clustering analysis’ to identify clusters of distinct, spatially contiguous characteristics to help identify the relationship between EDI scores and green space. The first task was to determine the distribution of EDI scores and determine if there was a pattern in the data. This was achieved by computing two spatially constrained multivariate clustering analyses. This analysis did not include green space, as it was purely to see patterns in the distribution of EDI scores throughout Vancouver. The first clustering analysis looked at 5 categories; physical, social, language, communication, and total scores. The second analysis looked just at physical and social scores. Both analysis only used the point data for the EDI scores. I then compared these results to the NDVI and other band combination maps to visually see any patterns in the data. 

I then used the Summarize Within (Analysis) tool in ArcGIs Pro to determine the total amount of park space in each DA, and the mean values for individual EDI scores (physical, social, language, communication, and total scores). Once this was accomplished, I completed two more spatially constrained multivariate clustering analysis’, this time looking at mean EDI scores and total hectares of green space within each DA. The first clustering analysis looked at mean physical, social, language, communication, total scores and sum hectares. The second only looked at mean physical scores, mean social scores, and sum hectares within each DA. 

Next, I computed the geographically weighted regression for physical score, with the dependent variable being mean physical scores, and the explanatory variable being sum hectares. 

Lastly I conducted one more spatially constrained multivariate clustering analysis’ and two more geographically weight regression to try and get a better representation of how green space impacts children’s development by looking at overall greenness of a DA instead of just using total hectares of park space. Before I could do this though, I needed to get the mean NDVI values for each DA. This was accomplished by converting the NDVI raster layer into point data, and then using the summarize within tool to get the mean NDVI values for each individual DA. I then conducted the spatially constrained multivariate clustering analysis, looking at physical, social, language, communication, total scores and NDVI. Finally I completed two geographically weighted regressions, one looking at mean physical scores and mean NDVI and the other looking at mean social scores and mean NDVI. 

Data

Landsat 8 OLI data from USGS Glovis

Early Development Instrument (EDI) data for Vancouver

Park, street, and school information from City of Vancouver Open Data Portal

Vancouver Census Dissemination Areas

Vancouver Border