NORTH ATLANTIC DECADAL-TO- MULTIDECADAL VARIABILITY - MECHANISMS AND PREDICTABILITY Noel Keenlyside Geophysical Institute, University of Bergen Jin Ba, Jennifer Mecking, and Nour-Eddine Omrani NTU International Science Conference, Sep. 2012
Multidecadal temperature fluctuations: Internal versus externally driven?
Pronounced 70-80 year variation in observed North Atlantic Sea Surface Temperature Climatic impacts of socioeconomic importance (R Zhang) 1. What causes Atlantic Multidecadal variability Observations Models 2. Can it be predicted? 3. Do we underestimate the role of ocean-atmosphere coupling?
North Atlantic: wintertime ocean deep convection and a thermohaline circulation Meridional overturning circulation (MOC) Marshall & Schott, 1999 Pickart et al. 2003
Warm AMV and negative winter NAO Composite analysis of 10yr running mean AMV-index, 1870-2009 SST (JFM, C) SLP (JFM, Pa)
Normalised Multi-decadal fluctuations in Atlantic Sector variability: ocean-stratosphere Multi decadal variability explains 48% of AMV index and 14% of NAO index
Observational evidence that NAO drives AMV NAO leading by several years Latif and Keenlyside, 2011
Bjerknes conjecture: Ocean drives mid-latitude SST and turbulent heat fluxes on multi-decadal timescales AMV SST and turbulent heat flux indices Can these changes drive multi-decadal shifts in the winter atmosphere? Gulev et al. submitted
What is the potential to predict the decadal Observed shifts in characteristics the global monsoons? of AMV North Atlantic supports winter-time deep convection and thermohaline overturning circulation Winter NAO and AMV tend to vary in anti-phase on multi-decadal timescales NAO drives winter convection, and may drive AMOC changes Subpolar North Atlantic SST controlled by ocean dynamics NAM in the stroposphere covaries with NAO
Simulated Atlantic multi-decadal variability Some robust features among climate models Delworth et al. 1993 mechanism, and role of salinity Stochastic null hypothesis Other aspects Coupled feedbacks involving salinity External forcing
Models simulate similar decadal variability independent of external forces Kiel Climate Model, preindustrial control simulation Atlantic multi-decadal variability index Regression pattern 0.3-0.3 0 Model year 1000-1 2.5 Ba et al. submitted
Models simulate a range of internal variability Spectra of AMOC 30N CMIP3 Pre-industrial Runs BCM IPSL ECEARTH MPIOM MPI MIUB CCSM3 GFDL MIROC CSIRO INMCM4 100 50 25 100 50 25
AMOC fluctuations tend to drive subpolar gyre SST variations on decadal timescales Tropical North Atlantic Subpolar North Atlantic year year
Relation between NAO and AMOC variability not clear in models NAO spectra Cross-correlation NAO and AMOC Systematic error in winter convection sites may explain some of these differences
Delworth et al. 1993 mechanisms for AMOC decadal variability Found in a number of models Anomalous strong SPG drives salty water to convection sites Salinity anomalies enhance oceanic convection Convection enhances AMOC Stronger AMOC reverses SPG anomaly Excited by NAO like variability through heat flux anomalies
Illustration from KCM: SPG and convection in Irminger Sea Ba et al. submitted
Illustration from KCM: Salinity variations key to mechanism Control simulation Simulation with salinity restored to climatology over the subpolar North Atlantic Ba et al. submitted
Stochastic null hypothesis following Hasselman (1976) Delworth & Greatbatch (2000) Latif and Keenlyside (2011)
Period (years) Ocean driven by stochastic atmospheric variability: No indication of oscillatory ocean only mode Wavelet spectra: ocean model simulation driven by stochastic NAO forcing Time (years) Courtesy: Jenny Mecking
AMOC variations best described by an auto regressive process of order 7
Coupled feedbacks involving ITCZ Vellinga and Wu (2004): Menary et al. 2011 Salinity variations may also come from Arctic (Jungclaus et al. 2006)
AMV may also be driven by external forcing But controversy exists See also Ottera et al. 2010 Booth et al. 2012
Mechanisms for AMV based on models AMOC variations drive significant part of AMV AMOC changes driven by salinity modulated convection Stochastic atmospheric forcing excites oceanic variability Coupled feedbacks in the tropics could play a role External forcing may also contribute
To what extent is AMV predictable: IPCC AR5 near-term predictions
Skill of (7) CMIP5 initialized nearterm forecasts Indian Ocean and western Pacific North Atlantic Guemas et al. 2012
North Atlantic subpolar gyre very predictable Bias corrected Raw CCSM4 CCSM4 predictions predictions of SPG of SPG heat heat content content anomalies anomalies 10 year long; 10 members Budget analysis shows predictability comes from ocean circulation changes Yeager et al. 2012
AMV predictable up to 5 years in advance Kim et al. 2012
Uncertainties in initial conditions Keenlyside & Ba (2010)
Large uncertainty model projections of AMOC AMOC at 30N, CMIP3 models, 20C/A1B Schmittner et al. (2005)
Quantifying uncertainty in decadal AMOC change Following Hawkins and Sutton (2009)
Near-term predictions of AMV AMV appears potential predictable, particularly in the subpolar region Challenges Model uncertainty Ocean initial conditions uncertain
Stratosphere drives decadal shifts in winter NAO Changes forced by a strengthening of stratospheric polar votex SLP (DJF, hpa) Zonal average zonal wind (DJF, ms - 1 ) Scaife et al. (2005) Can AMV drive the stratospheric changes, and in turn the NAO?
Ocean-atmosphere coupling: Potential role of stratosphere Observed SST anomaly NCEP/NCAR 1000hPa GPH anomaly Observed 1951-1960 anomalies Lower stratosphere resolving Entire stratosphere resolving Simulated (ECHAM5)resp onse to Atlantic SST 1951-1960 1000hPa GPH Omrani et al., submitted
Simulated and observed weakening of early winter stratospheric polar vortex 20 hpa geopotential height (NDJ) High-top NCEP/NCAR
Normalised Investigating the stratosphere s role Warm (1951-1960) Reference (1961-1990) Experiments: High top: MAECHAM5 at T63L39, surface to 0.01hPa (80km) Parameterisation for gravity wave momentum flux Low top: standard ECHAM5 at T63L19, surface to 10hPa (30km) Enhanced horizontal diffusion in upper layers (sponge layer)
m/s Stratosphere resolving model has dynamic polar vortex High Top 80 Daily zonally averaged wind 60N, 10hPa Control Warm phase -60 Aug Jun Low-top Control Warm phase
Stratosphere resolving coupled model support warm AMV drives negative NAO ECHAM6/MPIOM T63L95 atmos. res, model top 0.01hPa (~80km) 1.5deg/L40 ocean res 500-year-long preindustrial coupled control integration Composite analysis of unfiltered winter AMV index Stand alone ECHAM6 simulations driven by coupled model SST
Coupled model results warm phase associate with negative NAO like SST (JFM) pattern 500 hpa GPH (Feb)! "#$%*%
Warm phase Atmospheric pattern largely driven by North Atlantic SST! "#$%*% 500 hpa GPH (Feb) Uncoupled Coupled
Warm phase Stratospheric polar vortex weakening largely driven by North 30 hpa GPH (Jan) Uncoupled Atlantic SST! "#$%) % Coupled
AMV contributes to low-frequency NAO variations stratosphere key Observations and model indicate warm AMV drives negative NAO, with stratosphere playing a key role: Upward propagating waves weaken the vortex in early winter Weakening of the westerlies propagates down and causes a negative NAO in late winter Robustness for cold case less clear Potenial implications for understanding multidecadal variability in the North Atlantic sector Omrani et al. 2012a,b in revision
Power Power Not 10clear if resolving stratosphere interaction 3 enhances simulated AMV E20 C02 Spectra of AMOC at 30N from 500 yr long high- and low top coupled models 10 3 High top ECHAM5 T63L47/MPIOM E20 C02 10 2 10 2 Low top ECHAM5 T63L31 /MPIOM 10 1 10 1 10 2 Year
Systematic error in winter convection sites Standard deviation of winter mixed layer depth
Cold phase results less clear! "#$%*% 500 hpa GPH (Feb) Uncoupled Coupled
Figure from Holger at 48N