2024, Vol. 7, Special Issue 8

Remote sensing technology has proven crucial in examining the relationship between paddy yield and various vegetation indices. This study, conducted in Petlurivaripalem, Andhra Pradesh, utilized satellite imagery from 2014 to 2023 to extract indices such as NDVI, GNDVI, SAVI, MSAVI, LAI, and LSWI. Accurate yield prediction is vital for India's economy, and this study evaluates the performance of several predictive models for paddy yield forecasting using these indices. The models assessed include traditional parametric approaches (ARIMAX, MLR), machine learning techniques (ANN, SVR, RFR), and advanced ensemble methods like XGBoost. The results indicate that XGBoost consistently outperforms other models, delivering the lowest error metrics across all vegetation indices. Specifically, XGBoost achieved the best results with the GNDVI index, recording an RMSE of 50.85, MAE of 42.1, sMAPE of 12.1, MASE of 1.086, and QL of 20.05. These lower error metrics highlight XGBoost's superior accuracy compared to traditional and machine learning models. This study underscores the importance of remote sensing technology in capturing crop development patterns and forecasting paddy yield with precision, providing valuable insights for agricultural planning and decision-making.