Abstract: Based on observational data from 1962 to 2020 and the multi-model decadal hindcast and historical simulation experiments from the Decadal Climate Prediction Project (DCPP) under the Coupled Model Intercomparison Project phase 6 (CMIP6), this paper primarily evaluates the interdecadal predictive skill of the significant summer warming and wetting trends in Northwestern China over the past 60 years and its main sources. The interdecadal predictive skill of the multi-model ensemble for warming in Northwestern China, as indicated by the correlation coefficient with observations, is above 0.9. The high predictive skill is mainly attributed to the external forcing component, contributing between 80% and 99%. However, due to the combined influence of external forcing and initialize, there is significant variation in the predictive skill for wetting trends. The contribution of initialize to the predictive skill varies widely among models, ranging from 19% to 94%. This suggests that to achieve accurate short-term projections of future temperature and precipitation changes in Northwestern China, it is crucial to consider not only the impact of external forcing but also the influence of initialize on precipitation.