Ensemble learning-based estimation of reference evapotranspiration (ET_o)

Reference Evapotranspiration (ET_o) is crucial and influential in irrigation water management. Precise ET_o rate estimation is vital for successful agriculture water management. There are numerous techniques for ET_o rate simulation, but machine learning (ML) and deep learning (DL) approaches are currently popular. This study proposes an ensemble learning-based model for ET_o rate estimation. The proposed model leverages minimum meteorological parameters, i.e., minimum temperature (T_min), maximum temperature (T_max), relative humidity (RH), and mean wind speed (WS) as input features. The proposed model employs Random Forest Bagging and Gradient Boosting models as bagging and boosting ensemble techniques for the accurate ET_o rate estimation. The 10-fold cross-validation method is leveraged for the evaluation of the proposed model. The performance results of the proposed model are compared with the baseline model of ET_o estimation, i.e., the Food and Agriculture Organization Penman–Monteith (FAO-56 PM) and off-the-shelf deep learning models. The performance results indicate that Random Forest Bagging is significant as it yields Gradient Boosting and baseline models with 93.15% f-measure and reduces Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) by 17% and 10%, respectively.