Regional climate models have become a common research tool for downscaling
global climate simulations. To further examine their usefulness for climate
studies and the impacts that different physical parameterizations have on
the simulations, an intercomparison experiment has been performed where thr
ee regional climate models are used to simulate an extreme flood event. Alt
hough the dynamical components of the models are almost identical, the phys
ical parameterizations used to represent clouds, radiative transfer, turbul
ence transport, and surface processes are very different. The models were u
sed to simulate the heavy precipitation during the 1991 summer which caused
severe flooding over the Yangtze River in China. This extreme event is sel
ected to highlight the differences among regional climate models. Results f
rom the intercomparison show that all models simulated the gross flood cond
itions reasonably well, although each model reproduced the observed rainban
d to varying degrees, and significant differences are found in the simulate
d energy and hydrological cycles, especially over cloudy areas. Through det
ailed analyses of model simulations at different spatial and temporal scale
s, several reasons are found to cause the departure of model simulations fr
om each other. These include the simulation of the amount and vertical dist
ribution of clouds, the treatment of cloud-radiative feedbacks, and the rep
resentation of land surface processes. This study suggests that aspects oth
er than surface temperature and precipitation of the regional climate simul
ations need to be more carefully evaluated. One specially important evaluat
ion criterion is the radiation balance which has serious implications for l
ong-term climate simulations.