Multi-model Super Ensemble Prediction Of The East Asia Monsoon Interannual And Interdecadal Variability Based On The CMIP5Dataset

By utilizing the historical hindcast and future climate prediction datasets of the Coupled Model Intercomparison Project Phase5(CMIP5), prepared for the upcoming Intergovernmental Panel on Climate Change-Fifth Assessment Report (IPCC-AR5), we analyzed and evaluated the climate change in historical scenario for the20th century. Based on the multimodel ensemble forecasts, multimodel superensemble prediction on the summer monsoon in East Asia during the period from2010to2049in the21st century in the RCP8.5scenario has been conducted.First of all, the optimal model number and optimal length of the training period of multimodel superensemble prediction experiments have been conducted. The results show that the root-mean-square errors (RMSE) of the hindcast decreases and the anomaly correlation coefficient (ACC) will be improved as the model number increases. However, as the model number reaches a certain number, the decrease of the RMSEs and the improvement of ACC will be no longer evident.12models will be chosen for the multimodel superensemble prediction. In addition, optimal length of the training period could be decided for a particular length of the forecast period. In addition, as the length of training period increases, both of the RMSEs and ACC will be improved. Nevertheless, as the length of the training period exceeds a certain length the, the improvement of the RMSE and ACC will not be significant. In this study,20-40yr will be chosen as the optimal length of the training period for multimodel ensemble prediction.The hindcast data of sea level pressure and the zonal wind at850hPa and200hPa during the period from1990to2009have been analyzed. The results show that the RMSE of multimodel superensemble is less than that of any individual model and multimodel ensemble mean (EMN), while its ACC is higher than that of any individual model and EMN. Furthermore, the multimodel superensemble may reasonably hindcast the climatological mean of the sea level pressure and zonal wind at850hPa and200hPa, and the interannual variations of the summer mean of the sea level pressure and zonal wind at850hPa and200hPa as well. As a whole, the multimodel superensemble has a better hindcast than any individual model and the multimodel ensemble mean.Finally, the multimodel superensemble prediction of climate for the period from2010to2049has been conducted. Under the RCP8.5scenario, the East Asian summer monsoon would become stronger during the period from2010to2049with reference to the period from1980-2009. In addition, there are2-3yrs, quais-4yrs and6-8yrs period in the interannual variations of the East Asian summer monsoons during the period from2010to2049. The intensity of the East Asian summer monsoon would be relatively stronger during the2020s, and the increase trend would begin in2020.