Correction And Projection For Surface Air Temperature And Precipitation In The Early 21st Century Based On CMIP5 Models Over China

This study focuses on investigating and correcting the deviation of surface air temperature and precipitation projected by multi-model ensemble over China using the outputs of 24 climate models in CMIP5 under the RCPs scenarios in comparison with the latest grid temperatures and precipitations data from NOAA. The distributions, differences between RCPs scenarios and interannual differences of deviations have been analyzed. After corrected by historical bias and multiple linear regression, the temperature and precipitation of the next 10-20 years are projected by correction results. Main conclusions are as follows:(1) The multi-model ensemble has a capacity of projecting the distributions of annual temperature and precipitation and the seasonal variation. The model-ensemble temperature are underestimated in most parts of China, while the deviation is larger in the western region and it is largest in Tibet Plateau and its north regions, even more to 6℃. In summer, the uncertainty is lower than other seasons. For precipitation, the model-ensemble precipitation is overestimated in northern and western parts of China and it is over 80%, while underestimated in coastal zone about-20%. In winter-half year, the phenomenon of overestimating precipitation is obviously enhanced, while it is underestimated in the eastern monsoon zone in summer-half year.(2) The consistency of temperature among 24 models is higher than precipitation. The differences of projected temperature among models are more obvious in the east. And there are obvious differences in projected precipitation among models in Northwest China and eastern coastal zone. Interannual variability of precipitation deviations is higher in northwest, while the interannual differences of temperature deviations are only higher in the northeastern.(3) With the increase of emission intensity, the impact of the emission intensity is more evident to the western regions, while the overestimates in precipitation in the northwest and the underestimates in temperature would be intensified.(4) After the terrain correction, the deviation of model temperature could reduce about 20%, the pattern of the complex terrain region is effectively correction. The deviation of temperature and precipitation would decrease about 72% and 84% by deducting the climate drift. The models temperature is still underestimated and the deviations are about ±1℃ in most parts of China. For projection precipitation, it turns to underestimation and the percentage is about ±20%. The deviations are only significantly overestimated in west regions of northwest.(5) By using multiple linear regression method, the deviation of projection temperature is decreased within ±0.5℃. And the phenomenon of underestimation is weak significantly and the overestimated regions are increased significantly. For corrected precipitation, the deviations are decreased to ±20% in most regions and are only obvious in the west of northwest and the Tibetan plateau.(6) Multiple linear regression method could more effectively improve the credibility of future climate projection. There are significant increasing of temperature in the early century(2016-2035) than the early 10 years, and the increasing is more obvious in most of the northwest than other areas. For precipitation, annual mean precipitation in future changes about ±100mm in many parts of China, and it only changes obviously in the southern region.Various contrast and analysis of deviations show that, the different deviations seemingly result from the defects of the models themselves, such as physical process, cumulus convection parameterization, topography treatment and spatial resolution. These deviations imply that it would be of severe uncertainty if directly using CMIP5-model outputs. It is necessary to correct the uncertainty so as to reduce potential risk for users. Moreover, the results of different correction methods also indicate that, besides the climate drift of deviation, the non-stationarity of deviation still has great influence on model forecast, and the areas where the deviations are lower are vulnerable to the impact of the error rate. Therefore, except the climate drift of models, the temporal evolution characteristics of deviations must be considered in the more detailed error correction methods.