Estimating Daily Rice Crop Evapotranspiration in Limited Climatic Data and Utilizing the Soft Computing Algorithms MLP, RBF, GRNN, and GMDH

Evapotranspiration represents the water requirement of plants during their growing season, and its accurate measurement at farm is essential for agricultural planners managers. Field measurements evapotranspiration have always been associated with many difficulties that led researchers to seek a way remotely measure this component in horticultural areas. This study aims investigate an indirect approach daily rice crop (ETc) by machine learning (ML) techniques least available climatic variables. For purpose, meteorological variables were obtained from three ground stations cultivation regions northern Iran 2003–2016. The ETc rates calculated seven variables, FAO-56 Penman-Monteith equation, regional calibrated coefficient considered as reference data. MLs, including Multilayer Perceptron (MLP), Radial Basis Function (RBF), Generalized Regression Neural Network (GRNN), Group Method Data Handling (GMDH), utilized modeling. Different input combinations applied, based on use minimum effective input. Results showed models most performances combination four variables: sunshine duration, maximum temperature, relative humidity, wind speed. Investigating accuracy growth phases estimation error belonged initial stage, which increased mid-season late-season stages. A comparison GMDH model performed better against other competitors. model, both Nash-Sutcliffe (NS) R2 greater than 0.98, Root Mean Square Error (RMSE) ranged between 0.214 0.234 mm per day all stations. current promising results modeling only common can be reliably applied variable over farms. studied will research value crops similar/different conditions.