Applicability of a quantile-quantile (Q-Q) bias-correction method for climate dynamical downscaling at Beijing station

A statistical bias correction based on quantile-quantile (Q-Q) adjustment is applied to daily temperature and precipitation at Beijing simulated by the variational resolution model LMDZ4. After bias correction, the annual cycle, the average and frequency of temperature are all closer to observation, while the deviation of annual mean temperature decreases from -1.2℃ to -0.4℃. The bias correction can remove most of the spurious drizzles generated by the LMDZ4 model. The bias of rainy days decreases to 3.7% from 61.5%. The Q-Q adjustment shows a good performance of correction on precipitation intensity and frequency, and the deviation of annual mean precipitation decreases to 0.07 mm/d from 0.28 mm/d. After correction, precipitation peaks in July, which is consistent with observation, and the false extreme precipitation in the winter is removed. The Q-Q adjustment is separately operated for different seasons for both temperature and precipitation. The corrective effect for daily temperature is superior in the winter and summer, compared to that in the spring and autumn. Significant improvements are obtained for extremely high and low air temperatures. This statistical bias-correction method not only effectively eliminates drifts on the simulated climatological mean, but also increases the capability of reproducing extreme climate values. It is a relatively satisfactory correction scheme. Meanwhile, there still exist some uncertainties in Q-Q adjustment, and the corrective effect is influenced by observational data and model performance.