Mountain river discharge integrates rainfall, seasonal snowmelt, and glacier melt, yet most hydrological models conflate climatic sources with transport pathways, limiting diagnosis of how climate-driven shifts propagate through subsurface storage. We present a source-explicit extension to the spatially distributed SPHY model that preserves water-source identity by tracking rainfall-, snowmelt-, and glacier-derived recharge and partitioning groundwater discharge in proportion to evolving source-specific groundwater storage. We apply this framework to the glacierized Langshisha Basin (central Himalaya, Nepal) and force the model with bias-corrected ISIMIP3b climate projections under SSP1-2.6, SSP3-7.0, and SSP5-8.5 for 2025–2100. Across scenarios, annual discharge changes modestly by late century (−10 to −24%), but its composition reorganizes almost completely. In the early period (2025–2034), glacierized terrain supplies nearly 60% of total streamflow when direct glacier melt (~41%) and glacier-derived groundwater (~18%) are considered together. By 2090–2099, direct glacier melt declines to 13–16% and glacier-derived groundwater to 6–7%, revealing a large “hidden” loss of cryospheric contribution that is obscured when groundwater is treated as a single mixed baseflow component. Rainfall becomes the dominant streamflow source, increasing from ~31–32% to 65–74% depending on scenario, while the total magnitude of groundwater discharge remains comparatively stable but transitions from glacier-supported to rainfall-dominated recharge. These results demonstrate that stable or only modestly declining annual discharge can mask fundamental loss of glacier-derived buffering and increased sensitivity to monsoon timing and variability.