Fig. 1 Observation (a) and low-resolution MME (b) and high-resolution MME (c) of mean annual precipitation, and the relative deviation (d, e) between simulated results and observation in 1961-2014 (The dots indicate that 78% or more of models agree on the sign of biases)

Fig. 2 The spatial distribution of model biases in mean annual precipitation in 1961-2014. (a) Low-resolution models, (b) high-resolution models (The number corresponds to Table 1)

Fig. 3 Box-and-whisker plots for multi-year average monthly precipitation distribution in China simulated by low-resolution models (purple) and high-resolution models (black) during 1961-2014 (The upper and lower boundaries of the box represent the 75% and 25% quantiles respectively)

Fig. 4 Taylor diagram for low-resolution models (a), high-resolution models (b) simulated seasonal and annual precipitation in China during 1961-2014 compared with the observation (The different colors represent the season, and the different numbers represent the model)

Fig. 5 Low-resolution models (b) and high-resolution models (c) simulated EOF1 of interannual variation of summer precipitation spatial distribution in eastern China during 1961-2014 and observation (a). (b0 and c0 are MME. Dot area in panels indicates that they are all statistically significant at the 95% confidence level. The green box indicates that the spatial positive correlation with observation is statistically significant at the 95% confidence level)

Fig. 6 As in Fig. 5 except for EOF2

Fig. 7 As in Fig. 5 except for the EOF1 of interdecadal variation

Fig. 8 As in Fig. 7 except for EOF2