Evaporation modelling using different machine learning techniques

Accurate prediction of pan evaporation (Ep) is critical for water resource management. This article investigates the capabilities of three different soft computing methods at estimating monthly Ep at six stations in the Yangtze River Basin using climatic factors, including the air temperature (Ta), solar radiation (Rg), air pressure (Pa) and wind speed (Ws) for the period of 1961–2000. The first part of the study focused on testing and comparing model accuracy levels at each station using local input combinations. The results indicate that the fuzzy genetic (FG) model generally produces better results than adaptive neuro-fuzzy inference systems with grid partition (ANFIS-GP) and M5 model tree (M5Tree) specifications in terms of the root mean square error, mean absolute error and coefficient of determination values. The performance of the above models was also examined using cross-station applications (estimating Ep without local input or output data) in the second part of the study. The third part focused on estimating Ep using generalized FG, ANFIS-GP and M5Tree models. Collectively, the results demonstrate that the FG model can be successfully used to estimate Ep without any local inputs and outputs and that a single generalized FG model can also be used at six different locations.