Machine learning-based forewarning models for rice pests and diseases using climatic parameters in Madurai district, Tamil Nadu

Rice is a staple crop extensively cultivated in tropical regions like the Madurai district of Tamil Nadu, where agriculture plays a central role in livelihoods and food security. However, the crop is increasingly vulnerable to climatic variability, which significantly influences pest and disease outbreaks. Given the rising unpredictability of weather patterns, there is a pressing need for early warning systems that can forecast pest and disease incidences to support timely and effective management. This study aimed to develop forewarning models for major rice pests (stem borer, leaf folder, brown planthopper) and diseases (false smut, sheath rot, brown spot) using weather parameters recorded during 2022-2024. Pearson correlation analysis revealed strong associations between weather variables and biotic stress incidence, with brown spot showing the highest sensitivity exhibiting a significant negative correlation with maximum temperature (r = -0.779**) and a strong positive correlation with relative humidity (r = 0.844**). Machine learning (ML) models including Random Forest, Support Vector Machine (SVM), Linear Regression and XGBoost were evaluated for prediction accuracy. Among these, Random Forest and SVM provided superior performance in terms of R² and RMSE metrics across multiple targets. The study demonstrates the potential of integrating climatic data with predictive modelling tools to enable timely and localized interventions, forming a scientific basis for sustainable rice pest and disease management strategies in the region.