Simulation And Projection Of Climate Change Over Qinling Mountains Using Multi-model Based On Statistical Downscaling

Climate change and its impact assessment are hot issues in current global change research.The global climate model is a reliable tool for estimating future climate change,it has a good simulation capability on a large scale,but the global climate model does not well in simulate many physical processes on a regional scale.The statistical downscaling approach as a tie can be used to connect the global climate model data and regional climate change characteristics.Although the current global climate model can give a nearly credible result,there is still a great deal of uncertainty,the uncertain of regional climate change estimates results can be effectively reduced by the multi-modal integration method to some extent.The main data involved in this paper are from the daily mean temperature and precipitation datum of 10 meteorological stations in Qinling area,the NCEP reanalysis datum and the seven global climate models datum from 1961 to 2100.The statistical downscaling approach provides by ASD model is used to simulate and predict the mean temperature and precipitation in the study area.The main research works and conclusions are as follows:(1)Two kinds of statistical downscaling approaches of multiple linear regression and ridge regression were used to simulate the mean temperature and precipitation in the period of the calibration and the validation period.The results indicate that,in the site scale,under two statistical downscaling approaches the explained variancesof mean temperatureexceed 91%,root mean square errors less than 0.03,the explained varianceof precipitation were above 14%,the root mean square error less than 0.25,and theexplained variancesof mean temperature and precipitation of multiple linear regression statistical downscaling approach are more than ridge regression statistical downscaling,while root mean square errorsare smaller than the ridge regression statistical downscaling approach.From the study area as a whole,both statistical downscaling approaches performed well when simulating the change features of mean temperature and precipitation,during the calibration and validation periods,in most month the simulation capacity of rmultiple linear regressions are better than ridge regressions downscaling approaches.Compared the performances in calibration and validation periods of two statistical downscaling approaches over the Qinling mountains,we conclude that multiple linear regression is more suitable than ridge regressions in the Qinling mountains.(2)Based on the multiple linear regressions statistic downscaling approach,the reliability of the statistical relationship established under the two interpolation conditions and the simulation ability of the seven global climate models after the statistical reduction of the mean temperature and precipitation in the base period are evaluated.The results show:whether it is GCM ? NCEP interpolation order or NCEP ? GCM interpolation sequence,the established statistical relationship is stable and effective in the study of climate change statistics in Qinling area.After statistical downscaling 7 global climate models can simulatemean temperature very well.Both on amonthly scale and on a quarterly scale,in most months,after statistical downscaling most of the global climate models are slightly underestimate the mean temperatureand the minimum value of mean temperature,slightly overestimate the mean temperature standard deviation and the maximum value of mean temperature.After statistical downscaling 7 global climate models can simulateprecipitationona certain extent.While the simulation capacity of MPI-ESM-MR,IPSL-CM5A-LR,CNRM-CM5 and MPI-ESM-LR global climate models are better than GFDL-ESM2G,Can-ESM2 and BNU-ESM global climate modelsin the reference period.On the monthly scale,after statistical downscalingthe seven global climate models are overestimated the mean valuesof precipitation,underestimated the maximum number of consecutive dry days.After statistical downscalingmost of the global climate models are overestimated the standard deviation of precipitation and precipitation Probability.On aquarterly scale,after statisticaldownscaling 7 climate models are overestimated the mean value of precipitation,the standard deviation of precipitation and the probability of precipitation,and underestimate the maximum number of consecutive dry days.(3)Based on multiple linear regressions statistic downscaling approach,using the output datumof 7 global climate models under the RCP45 and RCP85 scenarios to ensemble predictionthe mean temperature,and using the output datumof 4 global climate models under the RCP45 and RCP85 scenarios to ensemble predictionthe precipitation.The results show that the mean temperature of the Qinling area ensemble prediction by the seven global climatic models,which is significantly increased in the three periods of the 21st century,and the larger radiative forcing the higherincrement.The growing rate ofin the 6-9 months is higher than that of other months.The growing rate in August is the most obvious.While the growing rate of mean temperature in January is the smallest.The seasonal variation of the mean temperature in the three periods of the 21st century is presented as a whole.The highest growing rates appear in the summer,followed by the autumn,the spring and the winter.Under the RCP45 scenario,in the three time periods the growing rate of mean temperatureare0.888?,1.631 ? and 1.855?,respectively;under RCP85 scenarios is more obvious,which is 0.992?,2.124? and 3.442?,respectively.Under two RCPs scenarios,the precipitation in the Qinling area,which is estimated by the four global climatic models,is not significantly reduced in the three periods of the 21st century.The precipitation fluctuation decreases with the increase of radiative forcing.The amplitudeof variationof precipitationin the 6-9 months is more obvious than that of other months.The highestamplitudeof variationof precipitation appears in the summer,summer,followed by the fall,the spring and the winter.In the annual scale,under RCP45 scenarios,three time periods,the amplitudeof variation of precipitation are-1.453mm,-1.095mm and-1.578mm,respectively;under RCP85 scenarios,three time periods,the amplitudeof variation of precipitation are-0.644mm,-0.562mm and 0.13 9mm,respectively.