Hot Spots, Clusters and Outliers

Analysis Procedures

For the Fort Worth Fire Department analysis, the Cluster and Outlier Analysis tool (Anselin local Moran’s I) in ArcMap was run two times utilizing two different conceptualization types. In both runs of the tool, the fire department’s incident data for January 2015 was used. This data set has a “FEE” field which provides the values of the department’s ten-point call priority ranking scale. The output of the tool was visualized over US 2010 Census data where census blocks were symbolized in graduated colors according to median household income.

For the Dallas County Economic Development Office, similar US Census data for only Dallas County was used in ArcMap’s Hot Spot Analysis (Getis-Ord Gi*) tool.

For the first analysis for the Fort Worth Fire Department, I added data for all incidents from January of 2015 into ArcMap as the feature class in the Cluster and Outlier Analysis tool. The input field was set to the call priority ranking scale and the Inverse Distance conceptualization was used with zero permutations. Other variables were left at defaults and I ran the model. The output of the tool was a new feature class that plotted four cluster outlier types (CO Types); those with high positive values (HH), low positive values (LL) and negative value outliers (LH and HL). I adjusted the symbology of the layer to highlight the CO types. I then added census data for 2010 and adjusted the symbology of the layer to display the median household income with graduated colors. This resulted in very little distinction in income where clustering was occurring. To fix this, I examined the median household income field in the attribute table and determined that clustering was occurring in areas of low income. To differentiate between various levels of income, I adjusted the 5 colors of the graduated color scheme to represent income levels in increments of $10,000 starting at zero.

Application and Reflection

Problem description:

In a previous project, I discussed using Ripley’s K or spatial autocorrelation to discover if there was clustering between Venus flytrap populations and habitat conditions. If clustering exists, which I suspect it does, I would be interested to use the Cluster and Outlier Analysis tool to determine where specific clusters of plants are located.

Data needed:

This data has been collected by the North Carolina Natural Heritage Program. The NCHP has organized researchers and volunteers to locate and document plant populations including habitat conditions. This data is available from the program's website. These results along with wetland and soil data collected by the USGS would be used to visualize the data.

Analysis procedures:

By doing this it would be possible to find areas with high positive values where optimal conditions exist for transplanting specimens. Areas with low positive values would identify areas where conservation and habitat remediation efforts could be focused. Outliers would be of value in showing the locations for further research opportunities on understanding the species. Where habitat conditions are good, but plant populations are low, researchers might find a limiting factor to the plant’s success such as a polluted source of runoff. Although probably few in number, outliers where flytrap populations exist but shouldn’t according to our understanding of the plant’s optimal conditions would be worthy of additional study to better understand the species.