Abstract:
Correctly estimation of evaporation is vital in agricultural production, management of water resources, irrigation management, and water budget. In recent studies, several meteorological data have been commonly used to estimate evaporation. In this study, we used monthly meteorological data ob-served from the year 1996 to the year 2017, to deter-mine the monthly pan evaporation of Lake Sapanca, located in the Marmara Region of Turkey. Correlation between monthly pan evaporation and other meteorological data was investigated by parametric Pearson test, by two kinds of non-parametric tests, Kendall's tau-b, and Spearman's rho test. Meteorological data such as sunshine hours (H-ss), average air temperatures (T-avg), minimum air temperatures (T-min) and maximum air temperatures (T-max) proved to have a strong correlation with monthly pan evaporation (EPm). For this cause, they were used to im-prove the modeling. In this study, three intelligent algorithmic approaches, Radial Bases Function (RBF), Self-Organizing Feature Map Network (SOFMN), and Recurrent Network (RN) were utilized to estimate monthly pan evaporation (EPm) in Lake Sapanca. On the other hand, a model was established to predict monthly pan evaporation using the Meyer's Formula, which is an empirical method. In the Meyer method, meteorological data such as monthly relative humidity (RH), and monthly wind speed (S-w), each consisting of 264 data, were used. The best combinations of models were calibrated using three kinds of performance criteria; Mean Square Error (MSE), Mean Absolute Error (MAE), and determination coefficient (R-2). The best results obtained by RBF, SOFMN, RN models, and Meyer formula were compared with observed data. The results have indicated that the performance of model RN with four input variables is superior to the other models in estimating the monthly pan evaporation of Lake Sapanca.
