Choice of scale for integrating land use in malaria risk monitoring

There were nearly 37,000 reported cases of malaria in Peru in 2009 alone. With over 30% of the population identified as being at "high risk" for exposure, detailed risk mapping, along with early detection and warning systems, are in critical need. While there is evidence that the increased formation of puddles arising from deforestation increases the breeding of the rainforest's primary malaria vector, Anopheles darlingi, neither the spatial structure of land uses/land cover changes (LUCC) nor the area of influence of LUCC on mosquito density has been systematically addressed. The radius of influence that LUCC - particularly areas of deforested land and other regions likely to see increases in stagnant water formation - has on mosquito presence is of particular importance, both for the design of warning systems and to inform future malaria transmission studies. Here, we present the results of satellite-based analysis of land use patterns and mosquito density along the Iquitos-Nauta road in the Peruvian Amazon. Comparing supervised classifications of Landsat images of the Iquitos region from 1996 and 2001 , land cover features around each of 832 mosquito sites were tabulated by percent at six different radii: 250m, 500m, 1000m, 2000m, 3000m, and 5000m. These results were then used as inputs in a mosquito prediction model that determined the most pertinent spatial scale necessary to predict both adult and larvae Anopheles mosquitoes (darlingi, benerocchi, oswaldoi, mattogrossenis, and rangeli). The application of this study is to provide a systematic means of determining which areas are at the highest risk of malaria infection in order to inform design of warning systems and future studies of land use and malaria in the Amazonian frontier.