For our first lab in Applications in GIS we explored different methods of hotspot analysis through the lens of crime analysis in Washington DC and Chicago.
I enjoyed the amount of background available in this module; the examples and articles made it easier to tie in statistics-heavy concepts with the hotspot map imagery we see frequently for crime, school standards, and even Covid-19 impacts.
 |
| Grid-Overlay Hotspot Analysis |
 |
| Kernel Density Hotspot Analysis |
 |
| Local Moran's I Hotspot Analysis |
In creating the grid overlay analysis, we were given a pre-made 1/2 mile cell grid clipped to the City of Chicago boundary. This gave a basis for the analysis, and by using a spatial join to add a point shapefile of all 2017 homicides, students could create an additional field called Join_Count where the number of joins would be the number of homicides within that cell in 2017. I exported a new shapefile from that dataset containing only the cells with homicides greater than zero, and then followed the recommendation for a top 20% analysis. This was done by finding the top 20% most frequent homicide cells in 2017, and then exporting those cells into another shapefile. The shapefile was still a collection of squares, so I used the Dissolve tool to merge them together as a single polygon representing the top 20% homicide locations in 2017. This type of analysis may be most helpful for establishing high priority areas within a city grid.
The Kernel Density analysis relied primarily on the Kernel Density tool, for which students could input the homicides 2017 point shapefile, and receive a raster showing the different "kernels" of dense and frequent homicides in 2017. I manipulated the symbology on this raster -which comes with a wide gradient range- to have only two breaks in values. One break included everything from zero to three times the mean, the other break included everything from three times the mean to the maximum. From there I used the Reclassify tool to get the values into bins of either "1" or "2" and the Raster to Polygon tool to get the resulting raster into a polygon shapefile with areas as either of those bins as values. I exported only the "2" values into a separate shapefile, and this represented the areas of Chicago that had homicides in 2017 three times higher than the average.
The final type of hotspot analysis in this lab was a Local Moran's I analysis, where we were asked to use census data to determine the rate of homicides per household. The first step was to use spatial join and combine the 2017 homicide point shapefile with a shapefile containing census data and census tracts. Like the grid overlay, this created a Join_Count field where the number of joins was the number of homicides that had happened in that census tract in 2017. Using this Join_Count number and the number of total households form the census data, I created a new field that displayed the number of homicides per 1000 households. This was then the field that I input into the Cluster and Outlier Analysis (Anselin Local Moran's I) tool to return a shapefile, from which I exported the "High Highs" that showed census tracts with the most frequent and most clustered homicides in 2017. Like the grid overlay, these High-High tracts were still individual bordered shapes, so I combined them with the Dissolve tool to display the areas in 2017 Chicago that had frequent and clustered homicides.
Overall I greatly enjoyed this assignment, and it was a great reminder what can be accomplished by manipulating the fields in a shapefile.
No comments:
Post a Comment