Predicting Crop Yield Responses to Temperature and Precipitation Variability Using Statistical Models in Nellore District, Andhra Pradesh, India

Abstract

Statistical crop models are widely used to evaluate the impacts of climate variability on agricultural productivity. This study aims to evaluate the performance of statistical crop models in assessing the impacts of climate change—specifically changes in the mean and variability of temperature and precipitation—on maize yield in SPSR Nellore District, Andhra Pradesh. A perfect model framework using CropSyst was employed to simulate maize yields under baseline and synthetic climate scenarios. Model evaluation is conducted using statistical metrics such as the coefficient of determination (R²) and prediction accuracy. Results indicate that statistical models perform reliably when at least 10–20 observations per predictor variable are used. However, with sample sizes below 300, temporal disaggregation increases the risk of overfitting. Maize yield exhibits significant inter-annual fluctuations, ranging from 15 to 65 q/ha, with lower yields occurring during periods of rainfall deficit and higher yields associated with well-distributed precipitation. The study highlights the importance of adequate sample size and appropriate aggregation for reliable climate impact assessment. It further underscores the importance of improving climate data availability, strengthening adaptive agricultural practices, and enhancing irrigation and cropping strategies to build resilience. It is recommended that integrating statistical models with advanced machine learning techniques offers significant potential for enhancing predictive accuracy and supporting sustainable agricultural planning under changing climate conditions.