Runoff component quantification and future streamflow projection in a large mountainous basin based on a multidata-constrained cryospheric-hydrological model

Abstract. The Yarlung Tsangpo River (YTR) is one of the several major rivers originating from the Tibetan Plateau (TP) and plays a pivotal role in providing invaluable fresh water to its downstream countries. Large uncertainties existed in the studies related to streamflow variations in this basin, and the investigation is difficult due to the widely distributed snowpack, glaciers and permafrost and their complex effects on hydrological processes. In this study, we conducted a systematic analysis on the streamflow variations and runoff components in the YTR basin, using a physically-based hydrological model validated by streamflow and multiple datasets related to cryospheric processes. Main findings include (1) The contributions of snowmelt and glacier melt runoff to streamflow were limited, both for about 5~6 % for the whole basin, which might be overestimated by previous studies. (2) Under the climate change, the annual runoff would increase evidently in the future. The relative change of annual streamflow could exceed 90 mm (~38 %) at the outlet station in the far future compared to the historical period under the high emission scenario, while the amount and contributions of meltwater runoff would both decrease. (3) Adopting more observational data to calibrate the hydrological model played a critical role in reducing the uncertainty of hydrological simulation. The biases of snow and glacier simulation for data unconstrained led to a marked overestimation of contributions of snowmelt and glacier melt runoff to streamflow and further brought about an underestimation of the increasing trends of annual runoff by approximately 5~10 % in future projection. These results provide a relatively reliable reference of the streamflow change and the runoff components in both historical and future periods in the YTR basin, and have the potential to serve as a “reference value” in this region because we used more datasets to constrain the model uncertainty compared to previous studies.