Abstract: Snowmelt Runoff Model(SRM) is a concept and semi-distributed hydrological Model and its primary parameters can express the internal response mechanism of catchment runoff to snowmelt. In this paper, we proposed a progressive calibration approach for discretized parameters optimization of SRM which is applied to simulate the snowmelt-runoff process of the Manas River basin in Xinjiang, China. The snowmelt-runoff coefficient and rainfall-runoff coefficient are determined as the two sensitive parameters to be calibrated based on the model structure parsing, and it is analyzed that the time-varying characteristic of catchment snowmelt-runoff response expressed by the model parameters which are discretized to the multiple time scale of month, half month, ten days and five days. The different object functions and optimization methods are used to calibrate the year model parameters discretized ten days scale by simulating the snowmelt-runoff process during snowmelt period from 2001 to 2012. The simulation accuracy evaluation shows that the progressive calibration method is better than the discrete time segment calibration method. Finally, oriented three objectives, the model parameters are optimized with 2001-2008 as calibration period and 2009-2012 as validation period, the results of verification simulations show that the average of efficiency coefficient is 0.87, the proposed calibration method for SRM parameters optimization are proved to be effective and the optimized model can be used to predict snowmelt-runoff process