The sustainability of water resources in mountainous areas has a significant contribution to the stabilization and persistence of the ecological and agriculture systems in arid and semi-arid areas. However, the insufficient understanding of hydrological processes in ungauged mountainous catchments (UMCs) is not able to scientifically support the sustainable management of water resources. The conventional parameter transferability method (transplanting the parameters of the donor catchment model with similar distances or attributes to the target catchment model) still has great potential for improving the accuracy of the hydrological simulation in UMC. In this study, 46 river catchments, with discharge survey stations and multi-type catchment characteristics in Xinjiang, are separated into the target catchments and donor catchments to promote an improved model parameter transferability method (IMPTM). This method synthetically processes the SWAT model parameters based on the distance approximation principle (DAP) and the attribute similarity principle (ASP). The performance of this method is tested in a random gauged catchment and compared with other traditional methods (DAP and ASP). The daily runoff simulation results in the target catchment have relatively low accuracy by both the DAP method ( N S = 0.27, R 2 = 0.55) and ASP method ( N S = 0.36, R 2 = 0.65), which implies the conventional approach is not capable of processing the parameters in the target regions. However, the simulation result by IMPTM is a significant improvement ( N S = 0.69, R 2 = 0.85). Moreover, the IMPTM can accurately catch the flow peak, appearance time, and recession curve. The current study provides a compatible method to overcome the difficulties of hydrological simulation in UMCs in the world and can benefit hydrological forecasting and water resource estimation in mountainous areas.