In present study, a lumped conceptual hydrological model, NAM (MIKE11), is calibrated while optimizing the runoff simulations on the basis of minimization of percentage water balance (% WBL) and root mean square error (RMSE) using measured stream flow data of eight years from 1991 to 1998 for Yerli catchment (area = 15,701 km2) of upper Tapi basin, Maharashtra in Western India. The sensitivity of runoff volume and peak-runoff has been undertaken with reference to nine NAM parameters using the data of calibration period. The runoff volume and peak-runoff have been found to be highly sensitive with reference to maximum water content in root zone storage (Lmax) and overland flow coefficient (CQOF) respectively. On the other hand, runoff volume is found to be moderately sensitive with maximum water content in surface storage (Umax). The calibrated model has been validated for independent stream flow data of Yerli gauging site for years 2001–2004, and Gopalkheda gauging site for years 1991–1998 and 2001–2004. The model performance has been assessed using statistical performance indices, and compared the same with their yardsticks suggested in published literature. The simulated results demonstrated that calibrated model is able to simulate hydrographs satisfactorily for Yerli (NSE = 0.86–0.88, r = 0.93–0.96, EI = 1.05–1.12) as well as Gopalkheda subcatchments (NSE = 0.76–0.92 and r = 0.88–0.96, EI = 0.89–0.91) at monthly time scale. The model also performs reasonably well in simulating the annual hydrographs at daily time scale. The calibrated model may be useful in prediction of water yield and flooding conditions in the Purna catchment.