Climate warming has been reorganizing hydrological dynamics in high-mountain rivers by altering the relative contributions of snowmelt and glacier melt. However, inadequate representation of snow ablation processes can bias runoff-component partitioning and mislead the inferred roles of snowmelt and glacier melt in water availability and flooding. Here, we enhance the representation of snowpack dynamics in a cryosphere-hydrology model (SPHY-S) to quantify how changing snowmelt and glacier-melt contributions reshape flood regimes in the upper Yulongkashi River during historical (1980-2020) and future (2021-2050) periods. Results show that SPHY-S robustly reproduces observed snow-cover dynamics across monthly, seasonal, and interannual timescales, thereby improving the reliability of runoff-component estimates. By 2050, the relative contribution of snowmelt is projected to decrease by 2%-7% whereas that of glacier melt increases by 1%-6%, depending on the climate scenario. Under the high-emission scenario, glacier-melt runoff (increasing from 0.7 to 1.5 billion m³ yr⁻¹) will surpass snowmelt runoff (increasing from 0.9 to 1.2 billion m³ yr⁻¹) to become the dominant contributor. Annual streamflow is projected to increase by 37%-79% and flood frequency is projected to increase by 84%-105% by 2050, depending on the climate scenario, with floods shifting from snowmelt-driven to glacier-melt-driven under the high-emission scenario. The growing contribution of glacier melt to floods creates more sustained high flows, posing challenges for downstream reservoir operation and disaster mitigation. Our findings highlight that improved representation of snow processes is essential for reliable runoff partitioning and diagnosis of shifts in flood regimes, thereby supporting proactive water management strategies in climate-sensitive mountain rivers.