| This paper uses observational daily temperature and precipitation data over Yangtze-Huaihe river basin, ERA-40 reanalysis dataset and data from eight CMIP5 (Coupled Model Intercomparison Project Phase 5) climate models (BCC-CSM1.1(m), MPI-ESM-MR, IPSL-CM5A-MR, CMCC-CM, ACCESS 1.3, CESM1-CAM5, HadGEM2-ES, EC-EARTH) to construct statistical downscaling models to project the future changes of climate for the early(2016-2035), middle(2046-2065) and end(2081-2100) of the twenty-first century based on the combination of empirical orthogonal function (EOF) and canonical correlation analysis (CCA). The main findings can be summarized as follows:(1) According to different downscaling experiments, the optimal predictors for the downscaling of summer temperature and precipitation over Yangtze-Huaihe river basin are the combination of geopotential height at 500hPa and temperature at 850hPa, and the combination of geopotential height at 500hPa, temperature at 850hPa, and specific humidity at 500hPa, respectively.(2) The validation with independent sample (1991-2002) justifies that the statistical downscaling models have good abilities to simulate the temporal variation and spatial distribution of summer climate. For example, the stations with correlation coefficient between the downscaled temperature, precipitation and observed series exceeding 0.50 account for 100% and 61%, respectively. Most annual spatial correlations between the downscaled temperature, precipitation and observed climatological average state of precipitation are greater than 0.90 and 0.34, respectively.(3) The abilities of the coupled models to simulate the local climate can be significantly improved by statistical downscaling, and statistical downscaling multimodel ensemble (SDMME) performs better than others. After downscaling, the biases of area-averaged temperature are reduced to 0.11～0.76℃ with a decrease of 0.01～1.80℃, while the absolute values of the relative errors of precipitation are reduced to 1%～7%, associated with a decrease of 15%～41%. The bias of temperature and the absolute value of relative errors of precipitation simulated by SDMME is 0.41℃ and 1.3%, respectively, which are better than most individual models’ downscaling results. And the absolute values of temperature biases at most stations are reduced from above 1.0℃ to less than 0.5℃,while the absolute value of relative errors of precipitation are reduced from above 15% to less than 15%. Besides, the spatial correlation between the downscaled temperature, precipitation and observation are increased from below 0.30 and 0.40 to greater than 0.98 and 0.60, associated with RMSE reducing to 0.12～0.17℃ and 0.40～0.80 mm/d, respectively. The simulation of the standard deviation is also improved.(4) The projections of SDMME show that under the RCP4.5 scenario, relative to 1986-2005, the temperature in the early, middle and end of the twenty-first century is expected to increase 1.52℃,2.62℃ and 3.39℃, respectively. And the increasing trends of summer temperature in the eastern region are more obvious than the west. The projected domain-averaged relative changes of summer precipitation for the early, middle and end of the twenty-first century is -1.8%,6.1% and 9.9%, respectively, associated with more intense precipitation appears in western part of the region. The SNR (signal-to-ratio) of temperature at almost all the stations are greater than 3, which indicates high reliability in the projection. The stations with higher reliability are mainly located over the southern area of 30℃ and the early period has the highest reliability. But the reliability of the projected precipitation is much smaller than temperature that the SNR is mainly in the range of 0～1.5. And the reliability of the projected precipitation over the eastern area is higher than the west for the early of the twenty-first century, and stations with higher reliability mainly appear in the north of the region for the latter periods. |
PDF Full Text Request