Using predicted Anopheles gambiae larval habitat locations to explain heterogeneity in adult An. gambiae spatial distribution

Declines in morbidity and mortality from malaria have been seen both within most endemic countries and worldwide over the last decade, but malaria continues to affect around 200 million people per year and kill about 1 million people per year. The spatial distribution of malaria has been documented in numerous ecological and epidemiological settings, showing that malaria prevalence is often heterogeneous across a landscape, differing among households within a community. The locations of people’s houses relative to the locations of larval Anopheles habitats have been shown to influence the spatial distribution of malaria infections in certain landscapes. In lowland, western Kenya, Anopheles gambiae larval habitats are numerous and widespread, and the relationship between larval habitat distribution and malaria transmission is less clear. Accurately predicting the locations of larval habitats will allow us to test this relationship across a large area. We used geographic information about a 10 by 10 km landscape in Asembo, Kenya to model the locations of An. gambiae larval habitats. Thirty-­‐one sample quadrats (500 by 500 m) were exhaustively surveyed, and all potential larval habitats were georeferenced with GPS. Rasters of four geographic variables were calculated at a 20 m resolution for the study area: elevation, distance to the nearest stream, land use categories, and a topographic wetness index. These variables were tested for their ability to predict habitat locations using logistic regression modeling. Habitats were more likely in areas of low elevation, especially where the topographic wetness index values predicted pooling water. They were also more likely closer to streams and in agricultural land. The resulting predicted An. gambiae larval habitat locations were then used to explain heterogeneity in adult An. gambiae spatial distribution.