International Journal of Environment and Climate Change

This study investigates the influence of monsoonal wet periods on the yield of major Kharif crops in Gujarat, India.  Monsoon precipitation plays a crucial role in agriculture, as it replenishes water supplies and affects areas that rely on canals, tanks and groundwater for irrigation. During periods of minimal rainfall, the monsoon's impact on reservoir and groundwater levels can be significant, which ultimately affects agricultural output and the Indian economy. The study analyzed rainfall and crop yield data for the major kharif(paddy, pearl millet, groundnut, cotton, castor, maize, sesame, and pigeon pea)crops in 19 districts of Gujarat, which had more than 30 years of data. To achieve this, five different machine learning techniques were employed: Stepwise Multiple Linear Regression (SMLR), Artificial Neural Network (ANN), XGBoost Regression, Random Forest, and Support Vector Regression (SVR). These techniques were used to determine which monthly rainfall combinations had a significant impact on crop yield. The analysis was divided into a training phase, where 70% of the data was used, and a validation phase, which utilized the remaining 30% of the data. To evaluate the performance of the different techniques, error metrics such as Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE) were calculated. These metrics allowed for a comprehensive assessment of the accuracy and reliability of each technique in predicting crop yield based on monthly rainfall patterns. The results of the analysis indicated that XGBoost technique yielded the highest R² value, indicating a strong relationship between rainfall and crop yield and the lowest RMSE value, suggesting greater accuracy in predicting crop yield. Following XGBoost, the Random Forest technique performed well, followed by the Artificial Neural Network, Support Vector Regression and Stepwise Linear Regression techniques.