中国6个CMIP5模式对全球降水年际-年代际变率模拟的定量评估

Quantitative Evaluation on the Interannual and Interdecadal Precipitation Variability Simulated by Six CMIP5 Models of China

本文利用美国全球降水气候中心(GPCC)的降水资料和中国参加国际第五阶段耦合模式比较计划(CMIP5)的6个气候模式[BCC_CSM1.1、BCC_CSM1.1(m)、BNU-ESM、FGOALS-s2、FGOALS-g2和FIO-ESM]的历史模拟试验的降水数据,采用可以表征降水变率相对和绝对量级的方法,定量评估了6个模式对降水年际-年代际变率的模拟能力。研究表明,观测降水的年际变率一般占总方差的65%~80%,年代际变率占总方差的10%~35%。在CMIP5历史试验中,6个模式平均的降水年际分量方差对总方差的贡献(超过70%)较观测偏强,模拟降水年代际分量的方差对总方差的贡献较小(约为10%~20%)。模式总体低估了全球平均总降水、年际降水和年代际降水的变率,但是高估了年际降水对总降水的贡献、低估了年代际降水对总降水的贡献。与观测相比,6个模式对东亚和澳大利亚地区的年代际降水的模拟都比较好,模拟与观测年代际降水方差的比值为1左右。在非洲、南美洲和海洋性大陆,BCC_CSM1.1模式模拟的降水年代际变率最接近观测;在欧亚和北美,BNU-ESM模式模拟的降水年代际变率与观测最接近。在欧亚大陆上,BCC_CSM1.1模式模拟的降水年际分量与年代际分量的方差比最接近观测;在非洲和美洲,FGOALS-s2模式模拟的降水年际分量与年代际分量的方差比最接近观测。本文的研究结果有助于理解中国当前气候模式对降水年际-年代际变率的模拟能力,以及未来改进模式。

By using the precipitation data of the US Global Precipitation Climate Center（GPCC）and the outputs of six Chinese climate models[BCC_CSM1.1,BCC_CSM1.1（m）,BNU-ESM model,FGOALSs2,FGOALS-g2 and FIO-ESM]participating in the historical simulations of Coupled Model Intercomparison Project Phase 5（CMIP5）as well as the methods featuring the relative and absolute magnitudes of precipitation variability,this paper evaluated the capabilities of the six models quantitatively to capture the interannual and interdecadal variations of global precipitation.The research results showed that the interannual variance of observational precipitation generally accounts for 65%-80% of the total variance,and the interdecadal variance accounts for 10%-35% of the total variance.In the CMIP5 simulations,the interannual variance of the ensemble mean precipitation of the six models is generally stronger than observation,accounting for over 70% of the total variance,while the variance of the interdecadal component contributes less to the total variance（about 10%-20%）.Compared with observation,the variability of the globally averaged total,interannual and interdecadal precipitations were underestimated,the contribution of interannual precipitation to total precipitation was overestimated,and the contribution of interdecadal precipitation to total precipitation was underestimated by the models.The interdecadal precipitation in East Asia and Australia was well simulated by the six models,and the ratio between simulated and observed interdecadal precipitation variance is about 1.For Africa,South America and Maritime Continent areas,the simulated interdecadal precipitation variability of BCC_CSM1.1 model is the closest to observation,and for Eurasia and North America,the simulated interdecadal precipitation variability of BNU-ESM model is also close to observation.In Eurasia,the ratio between interannual and interdecadal precipitation variances simulated by BCC_CSM1.1 model is the closest to observation,so did the FGOALS-s2 model in Africa and America.The results of this study would help to understand the current simulation ability of the six Chinese climate models and to improve the models in the future.