GIS 5100 Module 1

We employed three distinct methods to identify homicide hotspots from 2017 data:

    Grid-Based Thematic Mapping: First, we created a fishnet grid and spatially joined it with the 2017 homicide data. We then filtered this grid to include only cells with at least one homicide and manually selected the top quintile of these cells. These selected features were then dissolved into a single set of polygons representing the hotspots.

Kernel Density Estimation: Next, we used the Kernel Density tool with the totalhomicides_2017 shapefile. We set the population field to none, specified homicides_kd as the output raster, a cell size of 100, and a search radius of 2630 square miles. We used the planar method for density calculation and incorporated the Boundary_chicago layer as a barrier. The resulting raster's symbology was edited to display two distinct values, which were then used to reclassify the raster. Finally, this reclassified raster was converted into a polygon file.

Local Moran's I Analysis: Lastly, we performed a Local Moran's I analysis using 2017 homicide data at the census tract level. We filtered the results to show only "high-high" values, indicating statistically significant clusters of high homicide rates, and dissolved these into a single set of polygons.

Our analysis suggests that kernel density estimation provides a more actionable outcome for predicting future crime hotspots than Moran's I, particularly concerning resource allocation. While Moran's I identified a greater total percentage of 2018 homicides within its 2017 hotspots, it did so by delineating a significantly larger area.

The challenge for law enforcement lies in the practical application of these predictions. Spreading a police presence over an extensive area, as suggested by Moran's I, significantly diminishes its effectiveness. Kernel density, however, demonstrates a superior density of homicides per square mile. This means that while it might capture a slightly lower overall percentage of future incidents, the areas it identifies are much more concentrated. This precision allows for a more efficient and impactful deployment of police resources, making it the preferred method for strategic crime prevention.

Comments

Post a Comment

Popular posts from this blog

GIS 5007 - About Me!

Image
 Hi - I joined the GIS program in the fall of 2024 to formalize my GIS training after more than a decade of learning to use GIS independently. I'm an Archaeologist, with my BA and MA in Anthropology, and have worked for the UWF Archaeology Institute as a research associate since 2015. In the past few years, I have become actively involved in cemetery preservation and research. Something that you can read a little bit more about on my story map . Ultimately I am hoping to add GISP to my credentials list. Resetting a marker in AME Zion Cemetery Exploring Oakland Cemetery in Atlanta with the Association for Gravestone Studies Conference.
Read more

GIS 5100 Module 5 Damage Assessment

Image
 For this week's exercise we were tasked with conducting a structural damage assessment of a few block area impacted by Hurricane Sandy. We were asked to create a map of the path of Hurricane Sandy We were then given a very large raster data set with a very small project area to assess structural damage. Knowing something about Hurricane Sandy and the amount of damage, I immediately ruled out the possibility that there were any structures with no damage or only affected or sustained minor damage. I then narrowed in on determining which structures were destroyed and which sustained major damage. To complete the analysis of damage within a distance of the shoreline I used the created shoreline to create buffers of 0-100 m, 100-200m, and 200-300m. I used a basic join to derive counts for the following result
Read more

GIS 5100 M4

Image
 For this week's exercise, we spent a considerable amount of time evaluating out-of-date elevation data, looking at the potential impact of storm surge data. This included comparing potential inundation models using LIDAR and USGS DEM-derived data. For the first exercise we were asked to compare Lidar data from New Jersey pre and post Hurricane Sandy. We were given several areas to asses and evaluate what has changed as well as looking at how the area has redeveloped based on a more recently derived buildings shapefile. We then turned our attention to the Tampa Bay Area where we looked at different methods of modeling storm surge and evaluated the predictive ability of each model. This included comparing counts of the types of buildings each building would see as flooded or not As well as comparing which models predicted a percentage of each type of building would be flooded to compare the accuracy of the different models Finally we created a map showing the agreement between the d...
Read more