Nowadays, accurate estimation of evaporation as one of the important elements of hydrological cycle can play an important role in sustainable development and optimal water resources management of the countries facing water crisis. Up to now, empirical methods and formulas on estimation of non-linear and complex process of daily pan evaporation have been developed that is of uncertainty. These methods and formulas do not have high accuracy and also access to their input parameters is difficult or their measurement requires high cost and time. In this study, performances of two non-linear models of NN-ARX and ANFIS have been evaluated to estimate daily pan evaporation under arid and hot climate conditions including dry and warm climate (Iranshahr), dry and coastal warm (Chahbahar), and semi-arid and warm temperate (Saravan). For this purpose, the best combination of model inputs was selected by using Genetic Algorithm embedded in Gama Test software for each of Synoptic stations located in these regions for the 5years period(2005-2010), then daily pan evaporation was estimated by using NN-ARX and ANFIS models. By employing the statistical criteria including R2، RMSE and MAE, performances of ANFIS model with three Gaussian membership functions and NN-ARX model were evaluated for each of the selective Synoptic stations. The obtained results indicate the accuracy of ANFIS model is higher than the one of NN-ARX model in estimating daily pan evaporation in different climatic conditions.