Predictive modelling of aflatoxin contamination to support maize chain management

The aim of this work was to design the potential support given by predictive models to maize management in a chain vision aimed at minimising aflatoxin contamination and human and animal exposure. There are some predictive models available but only AFLA-maize, which is a mechanistic model, is suitable for aflatoxin risk prediction worldwide. Weather data are the mandatory input for aflatoxin risk prediction and the output depends strictly on data sources, thus being influenced by both the time and distance scale of meteorological data. A user friendly summary interface of output data from predictive models is represented by risk maps in which the spatial gradient of the risk is highlighted. Actual (day by day throughout the maize growing season), historical (collected in the past) and future (predicted) data can be considered from single weather stations, a network of weather stations or a data base to support a single farm, a group of farms or an area, respectively. Past scenarios are the output generated by historical data, predictions related to actual data describe the risk situation of the current growing season and future data support the prediction of future scenarios. Model predictions cannot really support operational decisions throughout the maize growing season, but they are useful approaching crop ripening, when it is suggested that early harvest is associated with high risk and a switch from contaminated grain to non-food/ feed use can be decided. Scenarios generated by historical data can support preseason decisions with more careful management in high risk areas, while climate change scenarios are mainly destined for strategic actions deputed to institutions and policy makers. Model predictions destined for all the actors in the chain (farmers, extension services, stakeholders, politicians, institutions and researchers) can further support crop management, being also useful as communication and risk management tools.