Determining the best empirical method for estimating water surface vaporation based on sensitivity analysis in paddy rice field in Amol and comparing them with Artificial Neutral Network

Evaporation takes place in from water surfaces such as lakes and also from soil surface. In this study, several methods were applied to estimate evaporation of water surfaces including Mayer, Marciano, Shahtin, Hefner, Ivanof, Tichomirof, USBR and neutral network. The estimated values were compared with measured pan evaporation values. Sensitivity of these methods to input parameters of wind speed, temperature and relative humidity were evaluated using variation rates of ±10%, ±20% and ±30%. Analyzing results by statistical indices such as R2, MBE, RMSE and 1:1 line showed that estimated water surface evaporation using Ivanof method offers the best correlation with pan evaporation values followed by, Ivanof, Meyer and Tichomirof methods. In situations with probability of systematic and human faults in measuring input parameters, Ivanof, Marciano and USBR methods are more acceptable rather than other methods, because of lower sensitivity to input parameters. Obtained results showed a high efficiency of Multilayer perceptrons model for estimating water surfaces evaporation. The values of R2, RMSE and MBE in the best models) neutral network (are equal to 0.95, 0.72 and 0.22, respectively. Accurate estimation of evaporation from surface water provides a vital tool to better manage essential decisions for decreasing harmful effects of drought crisis and protect water resources.