Maize, a staple food in most African countries, is prone to contamination by aflatoxins, toxic secondary metabolites mainly produced by Aspergillus flavus
and A. parasiticus
. Aflatoxins are known to cause liver cancer, and chronic exposure has been linked to other adverse health outcomes including growth faltering in children. To mitigate exposure in maize-dependent populations, there is need to identify the factors associated with aflatoxin contamination. This is difficult, however, because of high sampling cost and lack of affordable and accurate analytical methods. Publicly available, remotely-sensed data on vegetation, precipitation, and soil properties could be useful in predicting locations at risk for aflatoxin contamination in maize. This study investigates the utility of publicly available remotely-sensed data on rainfall, vegetation cover (indicated by normalized difference vegetation index or NDVI), and soil characteristics as potential predictors of aflatoxin contamination in Kenyan maize. Aflatoxin was analyzed in maize samples (n=2466) that were collected in 2009 and 2010 at 243 local hammer mills in eastern and western Kenya. Overall, 60% of maize samples had detectable aflatoxin. Global positioning system coordinates of each mill location were linked to remotely-sensed, spatially explicit indicators of average monthly NDVI, total monthly rainfall, and soil properties. Higher rainfall and vegetation cover during the maize pre-flowering period were significantly associated with higher prevalence of aflatoxin contamination. Conversely, higher rainfall and vegetation cover during the maize flowering and post-flowering periods (not including harvest) were associated with lower prevalence of aflatoxin contamination. Water stress throughout the growing season may cause increased plant susceptibility to fungal colonization and aflatoxin accumulation. Soil organic carbon content, pH, total exchangeable bases, salinity, texture, and soil type were significantly associated with aflatoxin. In conclusion, this study shows that remotely-sensed data can be regressed on available aflatoxin data highlighting important potential predictors that could reduce the cost of data collection and the cost of aflatoxin risk forecasting models.