Abstarct: Actual evapotranspiration (AET) is one of the most important components of water balance in arid and semi-arid regions. Measuring this component is complex and time-consuming and generally leads to point values. Although the use of satellite imagine in the form of remote sensing-baxsed algorithms has overcome this problem, these algorithms themselves still have some limitations. In the present work, the performance of adaptive neuro-fuzzy inference system (ANFIS) artificial intelligence model in estimating AET in Neishaboor watershed (northeast of Iran) during 2001-2010 was evaluated. Meteorological data and indicators extracted from MODIS satellite images were used as input to this model under different scenarios. First, plant index images were received from the MODIS sensor and their data were extracted. Having the values of meteorological parameters and plant indices, first 8 input scenarios with a combination of meteorological parameters and plant indices were developed for the ANFIS model. Then 70% of the data were used for model training and 30% of the data were used for model testing and the performance of the model was evaluated under 8 scenarios. The results showed that the soil wetness deficit index (SWDI) had an important role in reducing the error rate of AET estimation, so that its addition to other remote sensing indices led to a decrease of 18.4% and 12.5% RMSE (root mean square error) in the training and testing sets, respectively. According to the results, the scenario including all inputs outperformed in AET estimation by assigning values of 13.1, 10.14 and 0.9 for RMSE, MAPE and NSE, respectively, in the training set and values of 11.93, 9.02 and 0.91 in the testing set. The performance of the proposed method in estimating AET can be evaluated in other watersheds using different satellite images and other artificial intelligence models.