Decadal climate predictions with the CMCC-CM coupled OAGCM initialized with ocean analyses

1. Istituto Nazionale di Geofisica e Vulcanologia EGU General Assembly 2011,Vienna, April 2011 – CL3.1. Decadal climate predictions with the CMCC-CM coupled OAGCM initialized with ocean analyses A. Bellucci1, S. Gualdi1,2, E. Scoccimarro2, A. Navarra1,2, S.Masina1,2 and A. Storto1 CMCC - Centro Euro-Mediterraneo per i Cambiamenti Climatici (Euro-Mediterranean Centre for Climate Change), Bologna, Italy INGV - Istituto Nazionale di Geofisica e Vulcanologia (National Institute for Geophysics and Volcanology), Bologna, Italy Contact: alessio.bellucci@cmcc.it ABSTRACT In this work the effects of realistic oceanic initial conditions on a set of decadal climate predictions performed with a state-of-the-art coupled ocean-atmosphere general circulation model (OAGCM), under the framework of the COMBINE (Comprehensive Modelling of the Earth System for Better Climate Prediction and Projection) EU Project, are investigated. The decadal predictions are performed in both retrospective (hindcast) and forecast mode. Specifically, the full set of prediction experiments consists of 3-members ensembles of 30-years simulations, start- ing at 5-years intervals from 1960 to 2005, using CMIP5 historical radiative forcing conditions (including green- house gases, aerosols and solar irradiance variability) for the 1960-2005 period, followed by RCP4.5 scenario settings for the 2005-2035 period. The ocean initial state is provided by ocean syntheses differing by assimilation methodologies and assimilated data, but obtained with the same ocean model. The use of alternative ocean anal- yses yields the required perturbation of the full three-dimensional ocean state aimed at generating the ensemble members spread. A full-value initialization technique is adopted. The predictive skill of the system is analysed at both global and regional scale as well as the processes underlying the enhanced predictability exhibited over specific regions (most notably, in the North Atlantic) The CMCC-CM Coupled Model Model Drift Atlantic Meridional Overturning Circulation ATMOSPHERE ECHAM5 Res. = T159L31 (~80 Km) Fig.4 Time series of the maximum Atlantic MOC (AMOC, Sv; 1 Sv=106 m3s-1) at 26oN from CMCC ocean analyses (used to initialize decadal predictions) using OI and 3DVAR data assimilation schemes. The AMOC diagnosed from 2 different ODA schemes is affected by large uncertainties. Note that OI and 3DVAR assimilate different dataset as well: in situ temperature and salinity vertical profiles are assimilated by both OI and 3DVAR, while after 1992 SLA anomalies are assimilated by 3DVAR only (not OI). This is reflected by the larger discrepancies between OI and 3DVAR arising from the 90s onward: assimilating altimetry data determines a stronger (and more realistic: see fig. 5) AMOC. This results in large perturbations on the initial state of the MOC for different ensemble members. [Fig.4b] 1997/98 ENSO Fig.1 Fig.2 T,S & SLA (92 onward) COUPLER OASIS 2.3 Coupling frequency = 1.5h NO FLUX ADJUSTMENT OCEAN OPA 8.2 + SEA ICE LIM [Madec et al., 1998, Fichefet 1997] Horiz. Res. = ORCA2 (0.5o to 2o) Vertical Res. = 31 Levels T,S Fig.4 OBS Hindcast The CMCC-CM Coupled Model Fig. 4b:Hindcasts initialized on1995 with OI (black) and 3DVAR (red). Experiment Setup • Full-value initialization introduces a strong drift from realistic initial conditions towards the model own attractor. • Fig.1 shows global mean SST (GMSST) for the full set of decadal predictions (grey) and HadISST data (black). • In Fig.2 the reconstructed GMSST obtained by ensemble averaging model predictions over 30-yrs (red), 10 yrs (green) and 6 yrs (black) are shown. It is evident that retaining only the time slice which is closer to the initial conditions leads to a better agreement with the observations. On the other hand, including the long-term tail of decadal predictions has a deteriorating effect on the reconstructed GMSST due to the large drift towards the cold-biased model attractor (red curve). • The upward GMSST trend determined by the 20C radiative forcing is skillfully reproduced by the hindcasts • Note that vigorous ENSO events (e.g 1997/1998) are not captured by the hindcast, as the selected start dates do not coincide with ENSO years, and thus do not include the El Niño signal in the initial conditions. Fig.5 Time series of the maximum Atlantic MOC (AMOC, Sv; 1 Sv=106 m3s-1) at 26oN from the full set of hindcasts, retaining the initial 6 (black), 10 (green) and 30 yrs (red). Available estimates of the AMOC around 25oN (Bryden et al. 2005, [BLC05] and Cunningham et al. 2007 [C07]) are also shown. The effect of the model drift towards its own attractor (~12 Sv) is clearly visible in the 30-yrs reconstructed AMOC. When considering only the most skillful segment of the time series (5 to 10 yrs; black and green curves) the resulting AMOC drift is much reduced. Also, 6-yrs and 10-yrs AMOC display enhanced energy at decadal timescales, with respect to 30-yrs AMOC, and a closer agreement with the available observations (BLC05 and C07). 20C 3M North Atlantic Fig.3 Fig.3 Time series of an SST index defined as the area-averaged SST in the [0-60N; 50W-10W] region. The blue lines are derived from the initial 5-years time slice from each hindcast simulation initialized by three different ocean re-analysis.The red line is the ensemble mean and observations are in black. Anomalies for each time series are computed with respect to the corresponding long-term climatology. Part of the decadal SST fluctuations in the North Atlantic are skillfully reproduced by the hindcasts. Fig.5 Initialization & Perturbation • SUMMARY • A set of decadal predictions performed with the CMCC-CM coupled GCM initialized with ocean syntheses full-values and externally forced with CMIP5 radiative forcing (20C+RCP4.5) display predictability skill in the North Atlantic region. Skill in reproducing the long term upward trend of global mean SST reflects increasing radiative forcing from boundary conditions. • Full-value initialization introduces a strong drift. Retaining the initial 5-yrs time segment from decadal hindcasts yields a better agreement with observations. • Atlantic MOC is very sensitive to assimilation methodology and data used to produce the initial conditions. The large uncertainties associated with the AMOC in the ocean syntheses introduce a strong perturbation on the initial state of the model MOC. Perturbation of ocean initial conditions using different ocean analyses CMCC OI (Bellucci et al.2007): Assimilating in-situ T,S observations (ENSEMBLES EN3 dataset) using an Optimal Interpolator assimilation system. CMCC 3DVAR(Storto et al., 2011): Assimilating EN3 in-situ T,S data (as OI) + along-track sea-level anomaly observations using a three-dimensional variational data assimilation system. References Bellucci A, S. Masina, P. Di Pietro and A.Navarra 2007: Using temperature-salinity relations in a global ocean implementation of a multivariate data assimilation scheme, Mon. Wea. Rev.,135,3785-3807. Bryden, HL, Longworth, HR and Cunningham, SA 2005: Slowing of the Atlantic meridional overturning circulation at 25 degrees N,NATURE, 438,655-657 Cunningham, SA and Co-authors 2007: Temporal variability of the Atlantic Meridional Overturning circulation at 26.5N, SCIENCE, 317,935-938. Storto A., S. Dobricic, S. Masina and P. Di Pietro 2011 : Assimilating along-track altimetric observations through local hydrostatic adjustment in a global ocean variational assimilation system, Mon. Wea. Rev., in press. Acknowledgements This work was funded by the EC FP7 COMBINE (Comprehensive Modelling of the Earth System for Better Climate Prediction and Projection) Project, Grant Agreement number 226520. Valuable contributions from E. Manzini, P.Fogli and C.Cagnazzo are also acknowledged.