Lab 6 - Crime Analysis

This week's lab centered on how to run various types of hotspot analysis. Our focus for this use of GIS was specifically on determining crime hotspots - although I can see alternate uses in terms of generating hotspots of other sorts (statistically significant concentrations of certain types of archaeological sites or artifact types, for example).

Overlay of three different hotspot mapping techniques.

The above map shows the results of three separate hotspot analyses. The most concentrated area (shown in red) was derived from the Kernel Density analysis tool. Running this particular tool is superficially easy - but the results will vary depending on the variables used. For example, would a search radius of 0.25 miles be enough? Or up to 1 mile? I had run the tool 4 separate times, from 0.25 miles to 1.0 mile and finally settled on the results from the 0.5 mile search radius. How one displays the results can also alter the final view - the above results are based on all values that were 3 times above the mean - or in other words, the magnitude of the number of crimes per unit area (raster cell size = 100 feet, and the search radius was 0.5 miles).

The second most concentrated area is the Grid-based overlay, shown in bright yellow above. The results are based on values added to a uniform grid - in this case, 2007 burglaries per grid cell. The final results represent only the top 20% quintile, so only the most concentrated areas of burglaries are shown.

The final view also takes up the most map space - this is the results of the Local Moran's I analysis, shown in blue. While this analysis uses spatial autocorrelation in its final output it covers too much area - in my mind this might be a deterrent to decision makers who simply wish to know where to concentrate their resources.