Artificial Neural Networks for the Prediction of the Reference Evapotranspiration of the Peloponnese Peninsula, Greece

The aim of the study was to investigate utility artificial neural networks (ANNs) for estimation reference evapotranspiration (ETo) on Peloponnese Peninsula in Greece two representative months wintertime and summertime during 2016–2019 test if using fewer inputs could lead satisfactory predictions. Datasets from sixty-two meteorological stations were employed. available mean temperature (Tmean), sunshine (N), solar radiation (Rs), net (Rn), vapour pressure deficit (es-ea), wind speed (u2) altitude (Z). Nineteen Multi-layer Perceptron (MLP) Radial Basis Function (RBF) models tested compared against corresponding FAO-56 Penman Monteith (FAO PM) estimates a previous study, via statistical indices. MLP1 7-2 model with all variables as outperformed rest (RMSE = 0.290 mm d?1, R2 98%). results indicate that even ANNs simple architecture can be very good predictive ETo Peloponnese, based literature standards. determined Tmean, followed by u2, most influential factors ETo. Moreover, when one input used (Tmean, Rn), RBFs slightly MLPs < 0.385 ? 96%), which means sole-input ANN resulted predictions