The source of uncertainty in future warming and wetting projections over Northwest China

Northwest China (NWC) has experienced “warming and wetting” in recent decades. Reliable projections of temperature and precipitation are crucial for adaptation planning, yet CMIP6 models exhibit substantial spread. Using 26 CMIP6 models, we decompose projection uncertainty into internal variability, model differences, and scenario choice. During 1960-2020, the multi-model mean overestimates warming and underestimates wetting in NWC. For temperature, internal variability dominates near term projections (~54%), whereas scenario uncertainty prevails by the end of the century (~65%), with model uncertainty remaining substantial throughout (~27–49%). For precipitation, internal variability is overwhelmingly dominant (>80%) from near- to mid-term projections. Selecting high-skill models reduces projected anomalies and associated uncertainty for both temperature and precipitation. For the near-term future, temperature and precipitation anomalies decrease from 2.08 ± 0.92 °C and 6.58 ± 19.41 mm to 1.69 ± 0.77 °C and 5.87 ± 18.15 mm, respectively, after model selection. Across future periods, high-skill model selection reduces projected temperature uncertainty by ~16.3–20.9% (mainly from model uncertainty), whereas precipitation uncertainty is reduced only by ~3.6–6.5% (arising primarily from internal variability and model uncertainty). Overall, model selection cannot effectively reduce the substantial uncertainty associated with internal variability, particularly for near-term precipitation changes. Internal variability remains the primary source of near-term uncertainty, while scenario uncertainty dominates mid- to long-term temperature projections. These results indicate that isolating and representing key sources of internal variability will be critical over the next 10–30 years to improve NWC climate projections, thereby informing model development and adaptation decisions.​