Development of a Statewide Model of Culex tarsalis Habitat Suitability Using GIS

Abstract

My study combines the use of the maximum entropy approach to species habitat prediction with the need for a West Nile Virus (WNV) infection risk model for the state of Montana to produce a statewide null model of Culex tarsalis habitat to be used as the basis for a future infection risk model. My study assumes that C. tarsalis is the primary vector for WNV in Montana, and that MAXENT software can reliably generate a habitat suitability model for C. tarsalis throughout Montana using presence only data. I successfully generated a statistically sound, statewide model of C. tarsalis habitat using MAXENT software, C. tarsalis presence data, and readily available environmental datasets. Presence points were determined by verification in the lab via microscopic determination of the target species from mosquito samples collected over a four year period. Samples were acquired by me and my undergraduate colleagues, as well as by graduate students from MSU and a number of cooperators around the state, and the sample site locations were verified either through direct GPS measurement or a combination of GIS (Geographic Information System Software) and aerial photos. The final product is a map of the state showing potential C. tarsalis habitat in fourteen increments of increasing suitability (indicated by color), as well as a set of charts produced by the software demonstrating the statistical methods used in determining the various levels of suitability predicted by the model. In the process of creating the final model, a series of preliminary models were generated, and unnecessary environmental layers eliminated to ensure that the final model uses as few variables as possible to produce statistically significant results. Included along with the final model results, in appendices to the report are examples of statistical data and maps generated in preliminary model attempts that demonstrate the reasoning behind refinements made at various stages of the model’s evolution.