To address the difficulty in runoff forecasting in alpine areas, a runoff forecasting method coupling the iRainSnowHydro model and weighted Markov chain was proposed. The iRainSnowHydro model considered the regulation and storage of snowmelt water by constructing a snow reservoir, precisely depicting the process of snowmelt water outflows from wet snow. The method also introduced a weighted Markov chain to capture the transmission and transfer rules of simulation errors, achieving real-time error correction. The forecasting method was applied to the Wunonglong Basin in the upper reaches of the Lancang River. The results show that the determination coefficient and Nash efficiency coefficient in the calibration and validation periods were both over 0.8, and the absolute value of the percentage bias was less than 10%, indicating that the iRainSnowHydro model has a good simulation effect, especially during the snowmelt period (March to May). The weighted Markov chain effectively corrects the systematic underestimation of summer runoff by the model, reducing the average relative error from 15.68% to 7.52% and significantly improving the forecasting accuracy.