Determining the spatial autocorrelation of dengue vector populations: influences of mosquito sampling method, covariables, and vector control.

Abstract

We investigated spatial autocorrelation of female Aedes aegypti L. mosquito abundance from BG-Sentinel trap and sticky ovitrap collections in Cairns, north Queensland, Australia. BG-Sentinel trap collections in 2010 show a significant spatial autocorrelation across the study site and over a smaller spatial extent, while sticky ovitrap collections only indicate a non-significant, weak spatial autocorrelation. The BG-Sentinel trap collections were suitable for spatial interpolation using ordinary kriging and cokriging techniques. The uses of Premise Condition Index and potential breeding container data have helped improve our prediction of vector abundance. Semiovariograms and prediction maps indicate that the spatial autocorrelation of mosquito abundance determined by BG-Sentinel traps extends farther compared to sticky ovitrap collections. Based on our data, fewer BG-Sentinel traps are required to represent vector abundance at a series of houses compared to sticky ovitraps. A lack of spatial structure was observed following vector control treatment in the area. This finding has implications for the design and costs of dengue vector surveillance programs.