A Spatio-Temporal investigation of risk factors for aflatoxin contamination of corn in southern Georgia, USA using geostatistical methods

Aflatoxin is a mycotoxin produced by the Aspergillus flavus fungi that can severely contaminate corn grain. The U.S. Food and Drug Administration (FDA) have set a limit of 20 ppb, total aflatoxin, for interstate commerce of food and feed as it can induce liver cancer in humans and animals. Contamination is exacerbated by high temperatures, drought conditions and light-textured soil which are all common in Georgia (GA). Lack of irrigation infrastructure can further amplify drought stress and aflatoxin contamination. Accurate aflatoxin assessment requires the collection of multiple corn samples, is expensive and conducted at harvest which does not allow for the use of in-season mitigation strategies to reduce the risk. Given the expense of measurement and the consequences of crop loss, an important goal for agricultural extension services is the prediction and identification of years and counties at higher risk of aflatoxin contamination. This would allow growers to deploy management tactics to reduce risk and to reduce unnecessary expense on aflatoxin testing. In this research, aflatoxin levels were analysed by Poisson kriging and used to validate a strategy for identifying high risk years and counties. It is based on mapping risk factors (Maximum June temperatures, June rainfall, % corn planted area and % soil drainage types) that are above key thresholds. The aflatoxin data used were county level, collected unevenly in space and time from 1977 to 2004 in 53 counties in southern GA. Averaging and typical geostatistical methods were unreliable for producing a temporal summary of the spatial patterns because aflatoxin data were highly skewed and approached a Poisson distribution, and averages for counties with fewer observations are less reliable. Poisson kriging down-weights the influence of these in variogram computation and the estimation process. Comparison tests confirmed significant differences in aflatoxin levels between counties and years that were identified as having different levels of risk using the risk factors approach. Sensitivity analysis for Poisson kriged aflatoxin risk showed that the more years of data are clearly better for this analysis, but fewer than 15 years of data were not advisable.