Using Adaptive Neuro-Fuzzy Interface System (ANFIS) to estimate daily evaporation from pan

Evaporation process is considered to be one of the main components in hydrological cycle. Therefore, its accurate estimation plays a vital role in studies of irrigation management and river flow forecasting. In order to estimate evaporation from pan, 5 metrological parameters such as temperature, relative humidity, wind speed, vapor pressure deficit, and extraterrestrial radiation were daily collected during 1993-2007 at Shiraz Synoptic Station. In the first step, from 31 different combinations of input parameters, the best subset including relative humidity, vapor pressure deficit, and extraterrestrial radiation were detected by using the Gamma test (GT). Then, by applying M test, 1100 patterns were selected for an established training model and the rest of 3839 patterns were used for verification of the model. Afterward, based on the GT results, the evaporation model was established by applying adaptive neuro-fuzzy inference system (ANFIS). In addition, the results of this method compared with Marciano empirical equation, and calibrated Stephens–Stewart (S-S) equation. The results come from the evaluation of these models in the test period showed that the ANFIS model and S–S equation with determination coefficients of 0.9023 and 0.8604 on the line of (1:1) and RMSE with 1.342 and 1.614 mm day-1, respectively are more accuracy than other models.