To address the challenges of runoff simulation caused by the sparse observation stations in the upper reaches of the Jinsha River and the complex hydrological processes in the alpine mountainous areas, a VIC model simulation scheme based on fusion of ERA5-Land reanalysis data and observed precipitation was proposed. Based on the simulation results, the snow coverage rate of remote sensing monitoring and model output was compared and analyzed, and the variation characteristics of snowmelt runoff in the basin and the influence of temperature on snowmelt runoff were discussed. The results show that the VIC model constructed only by observed precipitation is difficult to simulate the real runoff process of the basin, while the certainty coefficient of hourly runoff simulation is increased to 0.80 and 0.74 respectively in the calibration period and validation period by using the fusion data, and the relative error is significantly reduced to less than 20 %. The output snow coverage rate of the model is in good agreement with remote sensing monitoring. The interannual variation of snowmelt runoff in the basin is stable, and the difference between the years is large, showing bimodal characteristics. There is a large correlation between snowmelt runoff and temperature in spring.