GIS4048 - Module 1 - Crime Analysis

-

For this lab, we explored 3 different techniques to map and analyze crime hotspots in Chicago using 2017 homicide data. The main objective was to assess how the three different spatial analysis methods: Grid Overlay, Kernel Density Estimation, and Local Moran’s I clustering can be used to highlight areas of high crime concentration and could even be an accurate measure to predict where homicides occurred in 2018. 

Each technique approached the concept of a "hotspot" differently. The Grid Overlay method simply counted how many homicides occurred within evenly sized ½-mile grid cells across the city. After ranking the grid cells by homicide count, I selected the top 20% with the most homicides and dissolved them into a single hotspot zone. This technique provided a very targeted result, covering a small area with the highest density of crimes per square mile as compared to the other hotspot maps.

Grid Overlay

Next, we practiced using Kernel Density Estimation (KDE) to create a smooth and continuous surface showing homicide intensity. This method doesn’t rely on boundaries but calculates density based on proximity between incidents. I used a cell size of 100 feet and a search radius of 2,630 feet, then classified only the highest density areas, being above three times the average as hotspots. The KDE technique provided a well balanced view of both concentrated and moderately intense areas (and in my opinion the most visually appealing). 

Kernel Density Estimation (KDE)

Lastly, we used Local Moran’s I technique, which is a statistical clustering method that detects significant spatial clusters of high (or low) values. In this case, I calculated the homicide rate per 1,000 housing units by census tract and applied Local Moran’s I to find “high-high” clusters. These are tracts with high homicide rates surrounded by others with similarly high rates. These clusters were then dissolved into one hotspot area so they could be compared. 

Local Moran’s I

After creating all three hotspot maps, we calculated their effectiveness by seeing how many 2018 homicides occurred within the 2017 hotspots. Local Moran’s I captured the most 2018 homicides (45%) but also had the largest footprint (35.38 mi²) and the lowest crime density. The Grid Overlay method was the smallest (15.45 mi²) and had the highest density, but only predicted 27% of 2018 homicides. KDE seemed to be the best middle ground, predicting 43% of 2018 incidents within 25.75 mi². This trade off between area and predictive power is important since, if we designated the entire city of Chicago as a hotspot, we’d “predict” 100% of future crimes, which wouldn’t be helpful for targeting resources like law enforcement. More advanced analysis would be needed to statistically compare the effectiveness of each method beyond just raw percentages and densities. I think this lab was a great introduction however to crime analysis techniques and has made me really interested in GIS forensics.

Comments

Post a Comment

Popular posts from this blog

About Me

-
Hi everyone! My name is Dalton Inman, I'm a full time marine bio undergraduate at UWF, and just switched from the minor to the certificate in GIS and will be graduating this December! I really enjoyed the introductory classes and wanted to dive deeper into understanding and using GIS for marine conservation and rehabilitation. My aim in taking these GIS Courses for my certificate is to add to my current GIS skill set. I noticed some LiDAR labs later in this course which makes me excited since I've been interested in learning about it. I am currently an intern at the Escambia County Marine Resource Division where I help organize data in Excel and create maps from Boat Data and deployed artificial reef condition data. When I'm not spending time outdoors flying drones or snorkeling in the ocean, I typically spend it inside with my now fiancé!! and my three cats: Kiwi, Pumpkin, and Mango. I hope to learn a bunch from this course and from everyone I meet in GIS!  If you want to ...
Read more

GIS4048 - Module 6 - Suitability Analysis (Part 2)

-
Image
For the second part of this module, we were tasked to identify a least cost wildlife corridor between two sections of the Coronado National Forest. The goal was to identify potential movement routes for black bears by combining environmental criteria into a suitability model, converting that model into a travel cost surface, and then using corridor analysis to locate low cost connection areas between the parks.  I did this by creating three suitability rasters on a scale of 1 - 10 using the reclassify tool. These were the distance to roads, elevation, and land cover, with values assigned according to their relative importance for black bear habitat. I then used the Weighted Overlay tool to combine these layers, assigning weights of land cover 60%, elevation 20%, and distance to roads 20% creating a suitability map. I then inverted the suitability raster (10 – suitability) to create a cost surface where highly suitable areas became low cost travel zones. I calculated the cost d...
Read more

GIS4930 - Module 2.2 - Interpolation

-
Image
For this lab, we used three different interpolation methods: Thiessen , Inverse Distance Weighting (IDW) , and Spline , to map Biochemical Oxygen Demand (BOD) across Tampa Bay, and compared how each method handles spatial variation and data distribution. The easiest and most simple method was Thiessen interpolation , which assigns each location the value of its nearest sample point, which means the interpolated surface matches the exact values at the sample sites. This wasn't the best type of interpolation though as it creates abrupt transitions between the polygons which made it not as good at modeling the continuous environmental data like BOD. Spline interpolation fits a curved surface that must pass directly through all of the input points. This created a smoothed, and honestly, very visually appealing interpolation, but seemed to be overestimating or underestimating values in areas with low data. This is what was happening in certain splots of the study area which is wh...
Read more