Projected runoff responses to climate and vegetation changes on the Tibetan Plateau

Global warming has intensified precipitation and vegetation greening on the Tibetan Plateau (TP) over the past four decades, with important implications for regional hydrological processes. However, systematic biases in climate forcing data remain a major source of uncertainty in runoff projections. Consequently, how future climate change will alter vegetation dynamics and subsequently affect regional water resources on the TP remains largely unclear. In this study, we developed a bias-correction framework that integrates detrended quantile mapping (DQM) and U-Net deep learning model to improve meteorological forcing data from the Coupled Model Intercomparison Project Phase 6 (CMIP6) models on the TP. The corrected climate data were then used to force land surface models that simulate runoff and vegetation dynamics, enabling projections of further runoff changes under three shared socioeconomic pathways (SSP1–2.6, SSP2–4.5, SSP5–8.5). Ridge regression and partial least squares structural equation modeling (PLS-SEM) were further applied to quantify the relative contributions of temperature, precipitation, and vegetation to runoff variability. The proposed correction framework substantially improves the forcing data, reducing the root mean square errors (RMSE) of temperature and precipitation by 27.08% and 22.07%, respectively, and increasing the spatial correlation of temperature by 0.22. In the Inner basins, runoff decreases under the SSP1–2.6 and SSP2–4.5 scenarios, while under SSP5–8.5 it initially decreases and then increases. In contrast, the remaining basins exhibit a persistent increase in annual runoff under all three scenarios, with larger increases under higher-emission scenarios. Precipitation remains the dominant control on runoff in the southeastern and northwestern TP under three scenarios, despite continued warming and vegetation greening on the TP. In semi-arid regions, vegetation exerts stronger hydrological influence under the low-emission scenario (SSP1–2.6), whereas temperature becomes the primary driver under medium- and high-emission scenarios (SSP2–4.5 and SSP5–8.5). Increasing radiative forcing enhances runoff sensitivity to precipitation while reducing its sensitivity to vegetation. The results reveal complex climate-vegetation-hydrology interactions on the TP under different emission pathways and provide new insights into future water resource management and ecosystem conservation on the TP.