Predicting severity of asian soybean rust epidemics with empirical rainfall models.

ABSTRACT Although Asian soybean rust occurs in a broad range of environmental conditions, the most explosive and severe epidemics have been reported in seasons with warm temperature and abundant moisture. Associations between weather and epidemics have been reported previously, but attempts to identify the major factors and model these relationships with field data have been limited to specific locations. Using data from 2002-03 to 2004-05 from 34 field experiments at 21 locations in Brazil that represented all major soybean production areas, we attempted to identify weather variables using a 1-month time window following disease detection to develop simple models to predict final disease severity. Four linear models were identified, and these models explained 85 to 93% of variation in disease severity. Temperature variables had lower correlation with disease severity compared with rainfall, and had minimal predictive value for final disease severity. A curvilinear relationship was observed between 1 month of accumulated rainfall and final disease severity, and a quadratic response model using this variable had the lowest prediction error. Linear response models using only rainfall or number of rainy days in the 1-month period tended to overestimate disease for severity <30%. The study highlights the importance of rainfall in influencing soybean rust epidemics in Brazil, as well as its potential use to provide quantitative risk assessments and seasonal forecasts for soybean rust, especially for regions where temperature is not a limiting factor for disease development.