One of the most important and complex processes in the watersheds is the identification and prediction of surface water changes. The main processes associated with surface water include precipitation, percolation, evapotranspiration and runoff. In this research, the semi-distributed model, SWAT, was used to simulate ground water and surface water in Semnan catchment in a monthly scale. A sensitivity analysis was perfomed to evaluate and demonstrate the influence of the model parameters on the four major components of water budget including surface runoff, lateral flow, groundwater and evapotranspiration. River discharge data from 2004 to 2014 were used for the calibration and those of 2014 to 2016 were applied for the validation. The results of sensitivity analysis showed that the most sensitive parameters were: SoL_K(Saturated hydraulic conductivity), CH_K2 (Effective hydraulic conductivity in main channel), RCHRG_DP(Deep aquifer percolation fraction and CN2 (Moisture condition II curve number). The simulation accuracy using Nash-Sutcliffe and coefficient of determination for Shahrmirzad, Darjazin, and Haji Abad hydrometric stations was about 0.60 to 0.80 and 0.80 to 0.90 for the calibration and validation period, respectively, showing a good performance in the simulation of river flow. According to the water balance results, about 87.6% of the total inflow into the watershed was actual evapotranspiration, 3% was surface run off, 3% was percolation, and the rest was related to the soil moisture storage.