Statistical Bias Correction Methods Of Regional Climate Model Projection

Credible estimates of climate variables are the basis and premise of the assessment of the climate change impacts. The elimination of various types from climate models is the important aspect of improving climate projection. Therefore, this paper aims to make a research on the different statistical correction methods based on Delta Change (DC) and Distribution-based Scailing (DBS) used to reduce the bias from PRECIS projection, such as P, Tm, Tmax and Tmin, in China under SRES A1B scenario.1962.12-1992.11is chosen as the control period and1992.12～2002.11is chosen as the validation period. Compare the correction results in terms of average spatial distribution, annual cycle, probability distribution, extreme indices and find that DBS can correct the average biases as well as DC, but it can correct annual cycle, probability distribution and extreme indices over most regions better than DC. There are some circumstances that DC acts better than DBS. In general, DBS is better than DC, since it takes not only mean value but also the standard deviation and extreme values into account when correcting the variables.1. DC and DBS do a good and comparative job in terms of simulated mean precipitation and temperature, especially decreasing the large biases significantly. The absolute biases decrease up to80%in China and the precipitation (temperature) R increase to0.95(1.0) from0.5(0.9). DBS(DC) can correct Tmax(Tmin) over most regions better. Meanwhile, DC overcorrects Tmax in tropical regions.2. DC and DBS do a good and comparative job in terms of simulated annual cycle of precipitation and temperature. The absolute biases of Tmin decrease the largest (42%,53%) and Tmax the least (18%,33%). The precipitation R increase to0.8from0.7and the temperature R is similar to the simulated R, which is close to1. Except DC does better at Tmax in HN region and XN region, DBS can do better at precipitation and temperature over most regions. While the two methods both overcorrect P in HN region and DC also overcorrects P and Tmax in HD region.3. DC and DBS can correct the simulated probability distribution of precipitation and temperature and DBS is better. The grid ratio that pass the K-S test in every season precipitation (temperature) is from29%(8%) up to49%(21%) and74%(36%). There are some circumstances that DC acts better than DBS, e.g. the PDF of Tmax in South HD region and North ZB region. There are some overcorrections, e.g. DC overcorrects precipitation and temperature in Shanghai and Zhejiang. The precipitation in summer and Temperature in winter and summer have been corrected the least, which indicates extreme values are important to PDF correction and are the key to improve the correcting technique.4. DC and DBS can correct the simulated CDD, R5D, ETR, FD and HTWI2much significantly, and the R95T as well. While they can not correct HTWI1.1or HTWI1.2significantly. The region-averaged R of FD and HTWI2can be larger than0.95. DBS can decrease the absolute biases of all the indices much more than DC and increase the region-averaged R of these indices, except FD and HTWI2, more than DC. DC can do a better job in some regions, e.g. FD in QZGY region and HTWI2in DB region. The overcorrections exist more in R95T and HTWI1and DC overcorrects more than DBS.As a result, in the impact assessment researches, it is necessary to choose the better statistical correction method to correct the climate projections as the input of assessment models, based on the different regions, climate variables and research goals.