Spatial Modelling of West Nile Virus Incidence in New Mexico: Utilizing Poisson Kriging and Mean Squared Error Analysis Over a 5-Year (2003 - 2022) Period

Abstract

West Nile Virus (WNV) poses a substantial public health threat in New Mexico, warranting a comprehensive investigation into its spatial distribution and incidence risk. This study, spanning from 2015 to 2019, employs Poisson kriging geostatistical techniques to model WNV patterns in Valencia County. The reported WNV cases over the 5-year period are aggregated to census tracts, and semivariogram modelling is applied to capture and integrate spatial autocorrelation in the prediction process. The resulting Poisson kriging interpolation generates continuous risk maps, providing a visual representation of WNV incidence patterns across the county. To assess and validate the model, mean squared error (MSE) analysis is conducted, systematically excluding each data point and predicting incidence at the omitted location. Low MSE values affirm the accuracy of incidence predictions. Notably, the results uncover significant clustering of high WNV risk in the western and southern regions of Valencia County, with core urban areas displaying lower incidence rates. A comparative analysis of observed and predicted incidence rates reveals an average prediction error of 1.81 cases. These findings furnish public health officials with a validated spatial decision-support tool, enabling targeted WNV surveillance and control initiatives in Valencia County. The employed methodology and analytic techniques underscore the efficacy of Poisson kriging for epidemiological mapping of environmentally driven diseases.