Application of innovative data-driven techniques for modeling reference evapotranspiration

Modeling reference evapotranspiration (ETO) as close as possible to actual measurement is critical in agricultural water management. It aids in irrigation planning and improvement, which is necessary for crop sustenance. Evaluating ETO with precision improves water conservation, food production, and drought management. This study attempts to develop a model for ETO using data-driven methodologies and compare the findings with the standard FAO-56 Penman-Monteith (PM) method. This research paper focuses on machine learning, namely Artificial Neural Networks (ANN), Multiple Linear Regression (MLR), and Random Forest (RF), and presents a comparative review of their practical applications in addressing hydrological challenges. So, the ANN, RF, and MLR models were trained first with the training data andthen checked for their effectiveness. R2, RMSE, MAE, MAPE, MSE, and NSE metrics were used to assess the effectiveness of the models. The ANN model had the best accuracy for the estimation of evaporation, with an R2 equal to 0.999 and the lowest RMSE and MAE amongthe models. Besides, RF performed well but was less precise than ANN, while the MLR model revealed the lowest performance. These findings have important implications for the management of agricultural water and triangulate the fact that ANN models can potentially improve ETO estimation across a wide range of agro-climatic zones. In the future, studies onhybrid models and the inclusion of other parameters will be undertaken to further enhance theaccuracy of the prediction.