Spatiotemporal prediction of reference evapotranspiration in Araban region, Turkey: A machine learning based approaches

Abstract

Accurate prediction of Reference Evapotranspiration (ET₀) is crucial for climate change mitigation, water resources management and agricultural activities. Therefore, this study aimed at investigating the applicability of a recently developed Machine Learning (ML) model called Gaussian Process Regression (GPR), for the prediction of ET0 in Araban station, Gaziantep region Turkey. Artificial Neural Network (ANN) was also developed for comparison. Several meteorological variables including temperatures T_min, T_max and T_mean (minimum, maximum and mean), surface pressure (PS), wind speed (U₂) and relative humidity (RH) from 1990 – 2021 were used as the inputs. Determination coefficient (R²), root mean square error (RMSE) and mean absolute deviation (MAD) were used as performance evaluation criteria. The obtained results revealed that GPR led to better performance with MAD = 0.0174, RMSE = 0.0236 and R2 = 0.9940 in the validation step. The general results demonstrated that GPR could be employed successfully to accurately predict ET0 in Araban station and thus, could be useful to decision makers and designers of water resources structures.