Predicting monthly evaporation using linear and nonlinear time series models (Case study: Ekbatan Dam station)

Prediction of evaporation as a key component of the hydrological cycle is one of the most important issues in water resources management and meteorology studies. In this study, the performance of ARIMA, SARIMA, gene expression programming, multiple linear regression, Monte Carlo and Thomas Fairing models in prediction of monthly evaporation values of Ekbatan Dam station, west of Iran in a 47 years period (1971-2017) were evaluated. For calibration of these models, 40 years data (1971-2010), and for validation, data from 2011-2017 (7-year) were used. The statistical metrics of the correlation coefficient, root mean square error, standard error, the Akaike information criterion, and NSE were selected for evaluation and comparison of models. The results showed that the SARIMA model has more accurate performance in predicting monthly evaporation. The GEP model, ARIMA, and MLR are ranked second to fourth. However, since the GEP model is easier to use than the SARIMA model and requires fewer variables than the SARIMA model, it shows promise to generate faster results, therefore, the GEP models can be the preferred option compared to others.