In preparation for a climate change impact assessment study for East Asia, we analyzed an RCM dataset from a 22-year regional climate hindcast. East Asia is among the regions that are most vulnerable to climate change. Socio-economical developments in the region has been adapting to frequent droughts and floods , as well as availability of water resources, as this region has been suffering from natural disasters and lack of water resources historically. Hence, shifts in regional water cycle due to the climate change induced by anthropogenic emissions of greenhouse gases will inevitably affect human sectors in this region. As this region is experiencing unprecedented growths in human population and economic development in recent years, the impacts of climate change on regional water cycle will be a crucial factor in planning for sustainable development. As projections of future climate and the climate change induced by anthropogenic greenhouse gases are usually generated by dynamical models, uncertainties originating from model errors will remain an important concern. As it is impossible to avoid model errors, close examination of model results and the characteristics of model errors is an important step in projecting future climates.

In this study, we examined the results of a regional climate hindcast generated for the 22-year period 1979-2000 in which the MAS model was driven by the ?-level NCEP R2. The RCM was successful in reproducing many important regional climate characteristics such as precipitation and temperature, upper-level fields, and ENSO-related rainfall variability. The agreement between the hindcast and available observations, however, are often only qualitative. The RCM results showed notable biases locally, suggesting that the regional climate change signals from RCM simulations may need bias corrections before they are applied to assessing the impacts of climate change. Further examination of the hindcast dataset showed that some scaled variables, especially the coefficient of variation, agree more closely with observations than the raw RCM variables. Based on these findings, climate change signals based on percentage-changes may be more useful for impact assessment than those based on simple differences.