Weather-based rice yield prediction in Kerala using ANN, SMLR and normal regression

Rice is the staple food crop in Kerala which accounts for nearly all of the state's food grain production and is mainly cultivated under rainfed conditions during the Kharif season. The critical influence of weather on rice productivity necessitates accurate and timely yield forecasts to aid agricultural planning. This study aimed to develop district-level rice yield prediction models for Kerala by analyzing the effects of essential weather variables: temperature, rainfall, relative humidity, and solar radiation on yield outcomes. Three statistical methods were employed: normal regression, Artificial Neural Networks (ANN) and Stepwise Multiple Linear Regression (SMLR). Among these models ANN demonstrated the highest predictive accuracy, with superior R²; values across all districts, indicating its robustness in modeling yield variability based on weather parameters. The ANN model has the ability to capture complex, nonlinear relationships among weather variables underscores its reliability as a tool for rice yield forecasting in all districts of Kerala. This enhanced forecasting potential holds substantial value for proactive agricultural planning and decision-making, allowing stakeholders to better manage resources and mitigate climate risks to ensure stable rice production in the state.