Development of a Disease Forecasting Model for Rhizoctonia Aerial Blight (AG1-1A) of Soybean in the Central Agroclimatic Zone of India

Rhizoctonia aerial blight (RAB), incited by Rhizoctonia solani, poses a significant risk to soybean production, reducing yield and grain quality. No single or multivariate predictive models employing subset or stepwise regression analysis currently exist for RAB for central zone of India’s principal soybean growing regions using susceptible soybean genotypes.

Weekly RAB incidence data from six susceptible genotypes, including JS 95–60, were correlated with weekly weather variables using Pearson’s analysis. Significant variables were used to develop single and multivariate regression models. Models with superior performance metrics were validated using cross-validation and evaluated for predictive accuracy through case-control classification and ROC analysis.

Disease was initiated during pod formation stage and gradually progressed up to harvest. Disease development was favoured by conducive microclimatic conditions, including dense canopy, warm temperatures (28-30°C), high relative humidity (>85%) and well-distributed rainfall. Significant variation in area under the disease progress curve (AUDPC) values across genotypes, years and their interaction was observed in the current study. Two single-variable models and one multivariable model were developed using stepwise and subset regression. These models identified minimum temperature, rainfall and the number of rainy days as key factors influencing the development of RAB disease. Multivariable regression model outperformed single-variable models with higher accuracy (85.90%), area under the curve (AUC=0.860) and better performance indicators based on cross-validation and ROC analysis. These models show promise for quantitative risk assessment of RAB epidemics up to 15 days in advance.