Bias-corrections on regional climate simulations over the northeastern United States

The delta method is applied to correct the biases of 10-year regional climate simulations over northeastern United States from two popular regional climate models (RCMs): the Weather Research & Forecasting Model-version 3.4.1 (WRFv3.4.1) and the Regional Climate Model-version 4.3(RegCM4.3) driven by the observed boundary conditions for period January 1983 to December 1992. The raw RCM simulations are first compared with observations from the Climatic Research Unit (CRU) high-resolution data separately for two 5-year periods, and the comparison shows that although the RCMs can reasonably reproduce the spatial patterns of summer and winter temperature and precipitation over the northeastern U.S., there are still some obvious biases, especially for precipitation. Then, the observed and simulated monthly temperature and precipitation values for the first period are used to generate the change factors, which are applied to the simulations of second period. The cross validation shows that, after the bias-corrections, both the simulated mean and variance of the summer and winter temperature and precipitation fields for the second period are improved, although the performances differ from region to region due to the raw RCM simulations. These improvements also indicate that bias-correction may also be needed in the simulation of future regional climate simulations driven by the general circulation models (GCMs).