A Drift-Free Decadal Climate Prediction System for the Community Earth System Model

Abstract

Performance of a newly developed decadal climate prediction system is examined using the low-resolution CommunityEarthSystemModel(CESM).Toidentifykeysourcesofpredictabilityanddeterminetheroleof upper and deeper ocean data assimilation, we first conduct a series of perfect model experiments. These experiments reveal the importance of upper ocean temperature and salinity assimilation in reducing sea surfacetemperaturebiases.However,toreducebiasesintheseasurfaceheight,dataassimilationbelow300m in the ocean is necessary, in particular for high-latitude regions. The perfect model experiments clearly emphasize the key role of combined three-dimensional ocean temperature and salinity assimilation in reproducing mean state and model trajectories. Applying this knowledge to the realistic decadal climate predictionsystem,we conductedan ensemble of oceanassimilationsimulations withthe fullycoupledCESM covering the period 1960–2014. In this system, we assimilate three-dimensional ocean temperature and salinity data into the ocean component of CESM. Instead of assimilating direct observations, we assimilate temperature and salinity anomalies obtained from the ECMWF Ocean Reanalysis version 4 (ORA-S4). Anomalies are calculated relative to the sum of the ORA-S4 climatology and an estimate of the externally forcedsignal.Asaresultofapplyingthebalancedoceanconditionstothemodel,ourhindcastsshowonlyvery littledriftandinitializationshocks.Thisnewpredictionsystemexhibitsmultiyearpredictiveskillsfordecadal climate variations of the Atlantic meridional overturning circulation (AMOC) and North Pacific decadal variability. c. 2019 American Meteorological Society.