Aflatoxin contamination in maize (Zea mays), primarily caused by Aspergillus flavus, is strongly influenced by meteorological conditions and remains a major food safety concern. Predictive models have been developed to support contamination risk assessment and management strategies, typically relying on meteorological data from local weather stations. While these inputs offer high accuracy, their limited spatial coverage and lack of forecasting capability reduce their application as early warning systems and in-season decision support. This study evaluates the integration of ERA5-Land reanalysis and seasonal climate forecasts into AFLA-maize, a mechanistic model for predicting aflatoxin B1 contamination, to extend its spatial and temporal applicability. Using historical data from the Emili…
Aflatoxin contamination in maize (Zea mays), primarily caused by Aspergillus flavus, is strongly influenced by meteorological conditions and remains a major food safety concern. Predictive models have been developed to support contamination risk assessment and management strategies, typically relying on meteorological data from local weather stations. While these inputs offer high accuracy, their limited spatial coverage and lack of forecasting capability reduce their application as early warning systems and in-season decision support. This study evaluates the integration of ERA5-Land reanalysis and seasonal climate forecasts into AFLA-maize, a mechanistic model for predicting aflatoxin B1 contamination, to extend its spatial and temporal applicability. Using historical data from the Emilia-Romagna, Italy (2008–2025), we tested bias-adjusted ERA5-Land inputs and developed a hybrid forecasting approach combining reanalysis data for the early-season with adjusted forecasts for the later crop stages. Results show that simulations driven by ERA5-Land reproduced contamination probability in 91% of years compared to station-based results, supporting its use in regions with low quality observations or missing station networks. Hybrid integrations of bias-adjusted forecasts provided contamination risk assessments 3–8 weeks ahead of harvest, showing the trade-off between lead time and accuracy. Forecast initialized later in the season (August) achieved much higher accuracy (up to 91%) but offer less time for action, whereas earlier initializations (June, July) bear larger uncertainty (accuracy of 88%–93%) but extended the decision windows. Furthermore, the availability of ensemble forecasts allows to quantify the uncertainty, providing probability ranges across members that support risk communication and early warning outputs. In summary, the presented approach extends AFLA-maize into a scalable and transferable tool for anticipatory risk management, supporting climate-resilient agriculture and food safety through publicly available meteorological data.