Automatic detection of surface-water bodies from Sentinel-1 images for effective mosquito larvae control

Authors:

Abstract
Surface-water bodies are important breeding habitats of mosquitoes, thus, surface-water body maps together with other parameters (meteorological and phenological) are often used as predictors of mosquito larvae presence.1 Regular monitoring of surface-water bodies in near real-time (NRT) is necessary for the timely prevention of outbreaks, such as the Rift Valley Fever2 or the West Nile Virus (WNV). The latter accounts for high numbers of human cases and fatalities, and for large outbreaks in Greece (during the periods 2010–2014 and 2017–2019, a total of more than 1.200 human cases were identified) and other European countries. A way of limiting the number of infections is through the effective mosquito population control, where exhaustive larviciding is the most commonly used practice worldwide. The mapping of surface-water areas can be achieved through observation and recording in the field. However, such a process is time-consuming and costly in human resources and budget,3,4 especially when a wide area needs to be covered. The use of unmanned aircraft systems (UAS) for the acquisition of aerial photographs in the areas of interest is considered a widespread practice, but the resources and equipment required (UAS, suitable sensors, human resources for the operation of the UAS, and the analysis of aerial photographs) are prohibitive for large-scale water maps.5 While several studies have used satellite data (optical and/or radar) to identify surface-water areas at different spatial and temporal scales, the requirements for an operational surface-water monitoring system for mosquito larvae prediction (accuracy, ease of integration, automation, frequency of update, and robustness under cloudy conditions) are not all met.