Research On Decadal Prediction With BCC Climate System Model

The use of climate system model to carry out decadal climate prediction has become a hot topic of climate change research worldwide in recent years.The decadal prediction is influenced not only by external forces,but also by the initial conditions,in which the model initialization method plays a very important role.This study establishes an ocean data assimilation system to generate a new assimilated ocean dataset.We use this dataset to initialize the BCC_CSM1.1 climate system model developed by the National Climate Center and perform a new set of decadal prediction experiments(EnOI_HadInit)annually from 1960 to 2005.Results of EnOI_HadInit are compared with the SODAInit,which are initialized by SODA reanalysis data.The major achievements of this study are as following:(1)Establishment of global ocean data assimilation system with ensemble optimal interpolation(EnOI)method.The assimilated data and model are more coordinated,providing reliable and reasonable initial value for decadal experiments.Based on the BCC ocean model MOM4_L40,the EnOI assimilation method has been used to establish the ocean data assimilation system.Sixty years of assimilation is carried out since 1949.The assimilated ocean data is able to reproduce the global warming trend of Sea Surface Temperature(SST).The correlation coefficient between the assimilated and the observed SST anomalous time series is 0.92.The Root Mean Square Error(RMSE)of assimilation experiment is about 30% lower than the control experiment.(2)Improvement in decadal prediction of global SST with EnOI assimilated initial condition.The EnOI_HadInit decadal prediction experiment using the assimilated data as the initial value can significantly reduce the systematic bias.Compared with SODAInit,the hindcast skill of SST in the tropical western Pacific,the South Indian Ocean and the northeastern part of the extratropical North Atlantic is significantly improved.RMSE is reduced by 20% to 30% in most region,and the ACC skill is improved obviously.EnOI_HadInit significantly improves the ability to predict the AMO compared with SODAInit,which is considered as a benchmark for the decadal prediction.The correlation coefficient between the hindcast by EnOI_HadInit and the observation is 0.52 in the past 40 years.EnOI_HadInit decadal prediction also reproduces the warm phase of PDO in the 1977-1998.The ACC of the hindcast of the PDO leading by 6-9 year and the observation is 0.41.The hindcast skill of Southern Indian Ocean Dipole(SIOD)by EnOI_HadInit is improved significantly as well.(3)Improvement in the decadal prediction of global and East Asian climate with EnOI assimilated initial condition.The effects of oceanic initialization can increase the prediction skill of the global climate through ocean-atmosphere interaction.EnOI_HadInit has higher hindcast skill in the global average surface air temperature,as well as surface air temperature in Europe and the western America compared with SODAInit.EnOI_HadInit also shows high skill in the hindcast of surface air temperature in the central eastern China.(4)Prediction for climate in the next 10-20 years with EnOI assimilated initial condition.The warming amplitude predicted by EnOI_HadInit is smaller than the RCP4.5 projection.The global mean surface air temperature continues to increase in the next 10-20 years predicted by both EnOI_HadInit and RCP4.5.The global average annual surface air temperature will increase by 0.9? in 2035 compared with climatology of 1961-1990.The North Atlantic surface temperature will decrease during 2016-2025 and increase during 2026-2035.The North Pacific and the Indian Ocean will continue to warm in the next 20 years.Based on the above analysis,by using ocean assimilated data in the initialization of BCC_CSM1.1 model developed by this study,the skill of the decadal prediction has improved in many aspects.