Rainfall variability in China affects agriculture, infrastructure and water resource management. Assessing the fidelity of the representation of interannual and intraseasonal hydrological variability in numerical simulations is an important step towards improving the predictions of flooding and drought. We identify spatial patterns of coherent regional interannual and intraseasonal rainfall variability in China by applying Empirical Orthogonal Teleconnection (EOT) analysis to high-resolution precipitation data from observations. This analysis technique does not maximize the percentage of explained space-time variance, like Empirical Orthogonal Functions, but instead identifies regions of strong coherent variability, and produces timeseries that are tied to specific points in China. Through regression analysis, rainfall variability in each region is connected to local and large-scale atmospheric and coupled atmosphere-ocean processes.  

The analysis of interannual variability of total seasonal rainfall reproduces known relationships between the El Ni？o-Southern Oscillation and variability in eastern China in winter, along the Yangtze River in summer, and in southeast China during spring. New discoveries are made for spring, autumn and winter, when rainfall variability in China is connected to midlatitude synoptic waves. The causes of subseasonal precipitation variability in China are investigated for extended winter (November-April) and summer (May-October). The three dominant patterns of intraseasonal precipitation variability in winter together explain 43% of the total space-time variance and have their origins in midlatitude disturbances that appear ten days in advance. In summer, all patterns have a strong relationship with the Boreal Summer Intraseasonal Oscillation. 

This comprehensive catalog of the mechanisms of observed rainfall variability in China is used as a basis for assessing numerical simulations performed with the Met Office Unified Model. We apply the same diagnostics to six climate simulations at various horizontal resolutions (from ~200km to ~40km) and with and without atmosphere-ocean coupling, to test how well the patterns of precipitation variability, their associated atmospheric circulation anomalies, and observed global teleconnections are captured in the simulations.