Spatial Pattern Analysis
Mapping Spatial and Geographical Clustering of Emergency Medical Services
The Fort Worth Fire Department wants to know if there is spatial and geographical clustering of emergency medical services (EMS) calls for them to consider placing their resources near places where they are needed most.
I used ESRI Guide GIS Tutorial 2, 4th edition to complete six different exercises utilizing six different spatial statistical analysis tools provided in ArcMap version 10.6.1.
Exercise 8-1: I. First, I ran a definition query using the “Incident” layer restricting the data to a few specific incident types. Then I used the Average Nearest Neighbor Tool to test the hypothesis if the calls for service are geographically clustered or are randomly distributed. The tool calculated the Average Nearest Neighbor Index, Observed Mean Distance, Expected Mean Distance, Z-score, and Confidence level.
Exercise 8-2: I used the Calculate Distance Band from Neighbor Count from the General G statistic tool (Getis-Ord General G) in ArcMap to test the hypothesis if the Priority Ranking values for service “incident” calls from January are geographically clustered.
Exercise 8-3: I used the Multidistance spatial cluster tool (Riply’s K function) to determine clustering of January 15 calls. I computed the difference between observed values and HiConfEnv and displayed the values as graph – Y=DIiffK, and X=Expected K.
Exercise 8-4: Spatial join the 300 ft grid layer with the Patron location. Created a definition query to allow count values greater than 0 and used “Spatial Autocorrelation (Moran’s I)” tool to look for clustering of service calls. I calculated Z-score and confidence level for 6 distances ranging from 2800 ft to 3800 ft in 200 ft increments.
Exercise 9.1: Performed Cluster and Outlier analysis to determine hot spots and cold spots and their relationship to the median household income. Reran the analysis using the fixed distance band of 900 feet.
Click to enlarge
Click to enlarge
Application and Reflection
I really enjoyed working on these exercises since the topic has a widespread application in epidemiology such as an outbreak of infectious diseases or cancer clusters in a defined geographic area, e.g., county.
Problem: Lyme disease results from ticks. It is spreading in NC, especially in Mecklenburg County, but is underreported. The Mecklenburg County Health Department (MCHD) wants to know if there are clusters of Lyme disease cases present in the county.
Data needed: MCHD has a list of all cases (along with the addresses) of Lyme disease that has occurred in the last five years. The Mecklenburg County Shapefile will be downloaded from the Open Mapping website of the county.
Analysis: The county shapefile and the list of cases will be imported in ArcMap. The Cluster and Outlier Analysis tool (Anselin local Moran’s I) will be used to identify clustering of Lyme disease cases in the county.