Improving mosquito population predictions in the Greater Toronto Area using remote sensing imagery

Abstract

Listen

Translate article

West Nile Virus (WNV) and St. Louis Encephalitis (SLE) are two of the most common mosquito-borne diseases in North America. WNV and SLE have sporadic spatial and temporal outbreaks, making their outbreaks difficult to predict. However, recent studies have found that mosquito abundance is correlated with WNV and SLE transmission, providing researchers with a starting point for the development of mosquito-borne disease forecasting systems. Mosquito populations are controlled by a variety of environmental variables, including humidity, temperature, vegetation, and available water habitat for breeding. Current mosquito population forecasting models heavily weigh precipitation and temperature inputs, as they are traditionally seen as the best estimates of available breeding space in a region of interest. Although rainfall data are easy to acquire, precipitation data may not actually be the best estimates of mosquito habitat, as water does not flow evenly across landscapes. Furthermore, precipitation data generally come at a spatial resolution of 800 m to 2,500 m, and while this resolution can help predict mosquito abundances on large spatial scales, it inhibits the estimation of mosquito populations in urban areas with granular landscape heterogeneity. To overcome these limitations, this research explores the use of multispectral imagery for predicting mosquito populations, specifically in the Greater Toronto Area. Multispectral imagery is an attractive data source for predicting mosquito abundance due to its consistent collection and comparatively high spatial resolution (e.g., 30 m for Landsat). We derive a monthly time series of standard spectral indices from multispectral imagery over the Greater Toronto Area from 2004 to 2011. We then explore how spectral indices perform as a predictor for combined Cx. restuans and Cx. pipiens mosquito populations, with the ultimate aim of using multispectral imagery to forecast mosquito-borne diseases in highly urbanized areas.