1. Multidecadal Oscillation of the Atlantic Meridional Overturning Circulation in Climate Models Bohua Huang Department of Atmospheric, Oceanic, and Earth Sciences George Mason University, Fairfax, Virginia Center for Ocean-Land-Atmosphere Studies, Maryland Collaborators: Shaoqing Zhang (GFDL) Zeng-Zhen Hu (CPC/NCEP) Jieshun Zhu (COLA)

2. Outline Motivation: AMOC and AMO Methodology: MSSA A quasi-30-year AMOC oscillation in CGCMs Evidence for an ocean internal mode Effect of coupling Summary

3. Atlanitc Meridional overturning circulation scheme (Schott 1902) (From Richardson 2008)

4. northward spreading of intermediate water in about 800 m (45S - 20N) southward spreading of deep water at 1500–3500 m (30N -> 55S) From Richardson (2008) Schematic Meridional Circulation along 30W (Merz 1925)

5. OBS at 26.5 o N (RAPID-MOCHA Array): 18.5±4.9 Sv (04/2004-10/2007, Johns et al., 2011)

6. Mean and Standard Deviation Gulf-Stream: 31.7 ±2.8Sv (Florida Strait) Ekman: 3.5 ±3.4Sv Upper Mid-Ocean: -16.6 ±3.2Sv (mooring densities) MOC: 18.5 ±4.9Sv (from http://www.noc.soton.ac.uk) RAPID-MOCHA observing system Joint UK/US Monitoring Array Observed Transport at 26.5°N (3.5 Years)

7. AMOC and Climate The Atlantic Ocean heat transport (OHT) is northward 60% of the OHT peak (1PW) at 20 o N is due to AMOC Surface air temperature change during years 20–30 after collapse of AMOC. Vellenga and Wood 2002 Water Hosing Experiment (HADCM3) Talley 2003;Boccaletti et al. 2005 What could happen in a world without AMOC (e.g., all glaciers melted)?

8. Long-term AMOC Variability A potential source of decadal predictability Forced change associated with global warming Multi-decadal oscillations due to internal dynamics Confidence in potential to predict decadal variability in AMOC reduced by model-dependence of both these features. From Schmittner et al. (2005) with additions (IPCC report).

9. Atlantic Multidecadal Oscillation (AMO/AMV)  Same sign in North Atlantic  Largest amplitude in the north  Major anomalies in the subtropical Atlantic  Basinwide horseshoe pattern  50-70 year time scales AMO Index based on Enfield et al. (2001) AMO and AMOC needs to be linked physically

10. AMO Global Effects OBSModel OBS Model EOF Patterns of Precipitation From Znang and Delworth 2006

11. From NOAA CPC Ocean Briefing, Oct 5, 2012 NAO and SST Anomaly in North Atlantic Fig. NA2. Monthly standardized NAO index (top) derived from monthly standardized 500-mb height anomalies obtained from the NCEP CDAS in 20ºN-90ºN (http://www.cpc.ncep.noaa.gov). Time-Latitude section of SST anomalies averaged between 80ºW and 20ºW (bottom). SST are derived from the NCEP OI SST analysis, and anomalies are departures from the 1981-2010 base period means. - High-latitude North Atlantic SSTA are closely related to NAO index (negative NAO leads to SST warming and positive NAO leads to SST cooling). -Negative NAO index persisted over five months, contributing to the strong warming in the high- latitude N. Atlantic.

12. NAO and SST Anomaly in North Atlantic Fig. NA2. Monthly standardized NAO index (top) derived from monthly standardized 500-mb height anomalies obtained from the NCEP CDAS in 20ºN-90ºN (http://www.cpc.ncep.noaa.gov). Time-Latitude section of SST anomalies averaged between 80ºW and 20ºW (bottom). SST are derived from the NCEP OI SST analysis, and anomalies are departures from the 1981-2010 base period means. - High-latitude North Atlantic SSTA is generally closely related to NAO index (negative NAO leads to SST warming and positive NAO leads to SST cooling). Negative NAO index has persisted for 7 months, contributing to persistent positive SSTA in high-latitude N. Atlantic, and also a warming in tropical N. Atlantic in Nov 2012. - In the past three hurricane seasons, positive SSTA in MDR was strong in 2010, and became weakening in subsequent two years. From NOAA CPC Ocean Briefing, Dec. 7, 2012

13. AMOC fluctuates faster in CGCMs (20-30yr, GFDL-CM2.1 Simulation) Delworth and Zeng (2012) AMOC Heat TransportTemperature Frankcombe et al. 2010

14. Quasi-30-yr oscillation is intermittent and model dependent Danabasolgu 2008 Danabasolgu et al. 2012 Leading AMOC mode, CCSM3

15. Guan and Nigam 2008 Nigam et al., 2012 Atlantic Annual SST and Fall Precipitation AMO indices 700hPa Z Moisture flux Precipitation Weaker subtropical anomaly Time series more wavy Is there a 30-yr oscillation on top of 70- yr?

16. England Temp Greenland Ice Core Frankcombe et al., 2010 Surface Temperature A 30-year signal in temperature? There is observational basis for 20-30yr variability in North Atlantic, faster than AMO

17. Questions How robust is the 20-30 year variability in climate models? What is its mechanism? What are its climate effects? What causes the model differences? Can we predict this multidecadal variability?

18. Characteristics of phenomenon multiple time scales Intermittent oscillation Method (SSA) separate periodic and non-periodic signals sensitive to time scales identify oscillations (even if intermittent) objectively no rigid frequency constraint allow spatial-temporal propagation Singular Spectrum Analysis (SSA)

19. An extension of EOF analysis to time-lag Most easily explained in 1-D case Given a time series, x(t), an M-lagged vector can be built as covariance matrix has M eigenvalues and eigenvectors: is the time EOFs of The The kth principal component is

20. The kth reconstructed component (RC) is In particular, an oscillatory mode is characterized by a pair of degenerate EOFs, i.e., This procedure can be easily extended into multivariate (or multichannel) SSA, i.e., MSSA vary coherently and 90 o out of phase. is the filtered oscillatory mode and Singular Spectrum Analysis (SSA)

21. Anything in common among models in the “fast” modes? Multi-channel Singular Spectrum Analysis (MSSA) Separating “Fast” and “Slow” Signals MSSA Variance: 10.8%+10.4%MSSA Variance: 9.8%+ 9.2%

22. Similar spatial structures center at 40-50 o N 1 st EOF patterns of the “fast” AMOC modes

23. Time Scale of the “Fast” AMOC Modes A Quasi-30-year Mode?

24. Construction of Phase Composite The 1 st principal component of the RC field and its derivative, both normalized, can be written as A cycle can be characterized by eight chosen phase intervals, The average of the original data for any variable A over all occurrences in phase m is called the phase composite A m (m=1,…,8)

25. Both models show similar AMOC phase evolution

26. Weaker modes also follow a similar pattern

27. Both models show similar AMOC phase evolution

28. North Atlantic SST is increased following stronger AMOC

29. North Atlantic HCA forces SSTA. Both are induced by AMOC heat transport

31. Potential Temperature (θ) and Density (σ θ ) at 300m, 40 o -50 o N σθσθ

32. East Observed multidecadal variability in North Atlantic sea level height Frankcombe and Dijkstra, 2009 West Subsurface Temperature 200-300 meters, T 0-400 meters, HC 10 o N-60 o N Frankcombe et al., 2010

33. An Internal Ocean Mode?  Warm anomaly in north-central induces a negative zonal overturning  Upwelling (downwelling) causes warm (cold) anomaly in the west (east)  Negative zonal ΔT induces a negative MOC  Upwelling in north-central reverses zonal overturning Uncoupled Ocean Model Simulations Westward propagation of thermal anomalies Sustained oscillation in idealized thermal boundary condition Thermal overturning: Salinity is secondary Te Raa and Dijkstra (2002), see also, Huck et al. (1999)

34. Phase A Phase B Phase A Phase B Geostrophic Self-Advection Sévellec and Fedorov 2012, J Clim., in press

35. Effects of Ocean-Atmosphere Interaction Surface heat flux Subpolar gyre circulation Atmospheric response Surface evaporation

36. Surface latent heat flux anomalies damp SSTA

37. North Atlantic SST is increased following stronger AMOC Surface latent heat flux anomalies damp SSTA

38. Mean Heat Content (HC, 0-500 m) Surface Current Surface Drifter WOA09 OBS GFDL-CM3.0 CCSM3.0 HCA/SSTA along Gulf Stream Extension and North Atlantic Current A “deformation” of subpolar gyre

39. The AMOC oscillation is associated with NAO (“delayed in CCSM3”)

40. SSTA forces divergence/convergence near the surface (SSTA is “delayed in CCSM3”)

41. Air temperature is warm after a strong AMOC

42. Strong AMOC induces sea ice melting

43. SST SSS σtσt Surface salinity is dominant, due to evaporation

44. Temperature Salinity Potential Density

45. SST SSS σtσt

47. Summary A quasi-30-year oscillation appears in climate models, centered at 40 o -50 o N. GFDL-CM3 and CCSM3 show vigorous oscillation with similar lifecycles. Associated with a strong AMOC, eastern subpolar gyre is warmed up while its south cools down. The gyre is “deformed”. HCA forces SSTA. Warm SSTA in northern North Atlantic weakens NAO, reduces sea ice, and expands warm air to North America. There is a westward propagation of subsurface temperature around 40 o -50 o N. Evaporation damps SSTA but enhances SSSA, making salinity dominant near surface.

48. Further Questions Any feedback from the NAO? What roles evaporation play? AMOC active or passive? What roles the AMOC-heat transport play? How strongly mean state affects oscillation? Is there any tropical-extratropical feedback? Is there an optimal perturbation? How predictable is the oscillation? How to initialize our prediction in real world? ……

49. Is there any potential feedback from NAO to AMOC? Can the extratropical processes affect the tropics? Some CFSv1 results

50. 99.5% 74.2%23.0% 80.1%18.6% Leading MSSA Patterns CFSv2 300-yr simulation Mode 1(22.3%) Mode 4 (5.0%), 5 (4.8%) Mode 2 (7.6%), 3 (6.4%) EEMD Century EEMD Matidecadal EEMD Decadal

51. Is there a Subtropical Connection? Huang et al. Clim. Dyn., 2012 Multidecadal Mode

59. Summary II The multidecadal AMOC oscillation may be affected by ocean-atmosphere feedback within the Atlantic sector. Low-frequency NAO fluctuation plays a more active role. Southward expansion of the SLP anomalies affects the subtropical gyre circulation Delayed heat transport of the subtropical shallow cell affects the oscillation.