Observations of the Meridional Overturning Circulation and
Implications for Past and Future Change
Carl Wunsch, MIT
NCAR July 2008
Oceanic change has become a public, “over-heated” subject….
Sea change: why global warming could leave Britain
feeling the cold
· No new ice age yet, but Gulf Stream is weakening
· Atlantic current came to halt for 10 days in 2004
James Randerson, science correspondent
Friday October 27, 2006
The Guardian, London a normally respectable newspaper
Scientists have uncovered more evidence for a dramatic
weakening in the vast ocean current that gives Britain its
relatively balmy climate by dragging warm water northwards from
the tropics. The slowdown, which climate modellers have
predicted will follow global warming, has been confirmed by the
most detailed study yet of ocean flow in the Atlantic.
Most alarmingly, the data reveal that a part of the current, which
is usually 60 times more powerful than the Amazon river, came to
a temporary halt during November 2004.
The nightmare scenario of a shutdown in the meridional ocean current which drives the Gulf stream
was dramatically portrayed in The Day After Tomorrow. The climate disaster film had Europe and
North America plunged into a new ice age practically overnight.
Lloyd Keigwin, a scientist at the Woods Hole
Although no scientist thinks the switch-off could happen that quickly, they do agree that even a
weakening of the current over a few decades would have profound consequences.
Oceanographic Institution, in Massachusetts, in
Warm water brought to Europe's shores raises the temperature by as much as 10C in some places
and without it the continent would be much colder and drier.
the US, described the temporary shutdown as
Researchers are not sure yet what to make of the 10-day hiatus. "We'd never seen anything like that
before and we don't understand it. We didn't know it could happen," said Harry Bryden, at the
"the most abrupt change in the whole [climate]
National Oceanography Centre, in Southampton, who presented the findings to a conference in
Birmingham on rapid climate change.
Is it the first sign that the current is stuttering to a halt? "I want to know more before I say that,"
record". He added: "It only lasted 10 days. But Professor Bryden said.
Lloyd Keigwin, a scientist at the Woods Hole Oceanographic Institution, in
Massachusetts, in the US, described 60 days, when most abrupt
suppose it lasted 30 or the temporary shutdown as "thedo you ring
change in the whole [climate] record".
the prime minister suppose it lasted 30 or 60 days, when
up He added: "It only lasted 10 days. But and say let's start do you ring
up the prime minister and say let's start stockpiling fuel? How can we rule out a longer
How canyear?" rule out that the flow rate of
stockpiling fuel?climate researchers lastwewith data suggesting a longer
Prof Bryden's group stunned
one next year?"he predicted, it wouldtonnestoofawaterdrop in the UK in1957next1998. If the
the Atlantic circulation had dropped by about 6m
current remained that weak, lead 1C
a second from
complete shutdown would lead to a 4C-6C cooling over 20 years.
The study prompted the UK's Natural Environment Research Council to set up an array of 16
submerged stations spread across the Atlantic, from Florida to north Africa, to measure flow rate and
other variables at different depths. Data from these stations confirmed the slowdown in 1998 was not
a "freak observation"- although the current does seem to have picked up slightly since.
Temperature Proxies: The M. Bender, GISP2 core
Greenland ice core
D-O events Bolling-Allerod
Last Glacial Maximum
Not temperature, but clearly related to it.
The Dominant features:
D-O events appear to represent major
warmings occurring sometimes in less than 10
(2) Apparent Holocene stability
(3) How could such things happen?
years! Glacial period instability (rapid fluctuations,
the `Dansgaard-Oeschger,’ D-O, events)
The favored explanation (see Al Gore, The Day After
Tomorrow, The Guardian (London), The Independent
(London), The Economist, Nature, Science, ….) is that the
is supposed to shut down occasionally---when fresh water
floods the shuts off in the North Atlantic.
ocean circulationNorth Atlantic---preventing sinking.
Is this credible?
A GREAT GRAPHIC! BUT THE NOTION THAT OCEANIC FLOW
IS A ONE-DIMENSIONAL RIBBON IS A LOVELY FANTASY.
Nature, December 2005 Claims a 30% reduction in
overturning circulation, 50%
reduction in lower NADW
transport since 1957. Huge
Data used are 5
Some basic, background issues.
The expression “thermohaline circulation” should be suppressed. on
the basis that it has become essentially meaningless of currents that arises
Example: "The thermohaline circulation (THC) is a global pattern through
from gradients in density, and hence hydrostatic pressure, between different regions in
the world's oceans." E. Hawkins and R. Sutton, 2007, Clim. Dyn., 29, 745-762.
There is a circulation of heat (temperature) and of salt. They are very
different (they have entirely different boundary conditions). "...the
Or, Bulletin of the American Meteorological Society, June 2006, P. 803,The
expression has measured in the Florida Straits, was near the long-term
thermohaline circulation, asbeen used in at least several mutually
mean.“ (What does this mean?)
(1) The circulation of heat and salt (which are different).
(2) The circulation driven by heat and salt fluxes at the sea surface.
(3) The circulation driven by density anomalies
(4) The circulation driven by pressure anomalies
(5) The abyssal circulation.
(6) The abyssal circulation driven by abyssal density anomalies
Note, at least, that the ocean circulation is almost indistinguishable from
geostrophic balance. Does flow drive pressure gradients or vice-versa?
Contrary to hundreds of papers about box models, “conveyor belts”,
zonally integrated models, etc., the ocean circulation is three-dimensional
and highly time-dependent. What does one see?
The ocean is a “noisy” place and so it’s easy to be fooled
From Estimating the Circulation and Climate of the Ocean (ECCO) consortium
(MIT, JPL, GFDL, U. of Hamburg) results.
A very dynamic, basically turbulent system, whose sampling is a major
Red segment denotes
western end of
Dudley Chelton, 2008.
Gridded altimetric data alone. 1 cm produces about 7Sv transport at mid-latitudes
Macdonald and Wunsch, 1996, Nature. Some people seem to
be able to sit at their
Oversimplified versions of the desks and determine
how this was different
in the past and how it
will be in the future. An
Lumpkin & Speer, JPO, 2008
The complicated pathways are an essential ingredient in understanding
how the system works, how it might change in the future, and how it might
have been in the past.
How fast can the ocean gross baroclinic structure change?
Recall Veronis and Stommel (1956):
This is the initial signal velocity, not the adjustment time.
The energetics of the system largely determine how fast things
Ferrari & Wunsch, 2009,
Ann. Rev. Fluid Mechs.
Rates of exchange of energy between ocean and atmosphere, and
between the components of oceanic energy involved in the general
circulation are all O(1012W). Changing the abyssal N. Atlantic
temperature by 1 degree C leads to a PE change of order 1022J for a
time scale of 1010s or about 300 years---unless energy transfer rates are
greatly changed in the process.
T. Stocker, Science, 1998
(referring to Dansgaard-Oeschger events)
Numerous modeling studies (13) have shown that changes in the meridional heat
transport in the Atlantic Ocean, caused by sudden changes of the Atlantic's
thermohaline circulation, are resulting in antiphase behavior of north and south. A
sudden increase of the northward meridional heat flux draws more heat from the south
and leads to a warming in the north that is synchronous with a cooling in the south
(14). It has been shown that the stability of this circulation is limited (5) and that
changes in the surface salinity can trigger major reorganizations of this circulation.
More importantly, simulations demonstrate that the amplitudes and rapidity of events
compare well with the paleoclimatic record (15).
Total ocean+atmosphere heat transport from (mainly)
observations. Total from ERBE. Ocean from hydrography
(Ganachaud), atmosphere as residual.
heat radiated to heat in from sun heat radiated to
Wunsch, 2005, J.
At high northern latitudes, the poleward transport of heat is
dominated by the atmosphere. Suppose one interrupted the
North Atlantic part. Can the North Atlantic Ocean tail wag the
atmospheric dog? Recall, too, Bjerknes compensation.
What is observed to be going on?
Among the Data Types
Argo T/P, Jason
How to synthesize? Estimation/optimal control problem:
Use a model (MITgcm) and its adjoint:
An independent estimate, in which we
attempt to use all the data, no matter
what type it is, from 1992 onward.
How to put those together to create an
understanding of what the three-
dimensional ocean is doing over days to
About 1 billion data constraints used.
Some withheld data (TOGA/TAO, drifter
The ECCO-GODAE setup, v2
• 1 degree horizontal
• covering 80N to 80S
• 23 vertical levels
• GM/Redi eddy
• KPP vertical mixing
• covers 1992 to 2006
• forcing: 6-hourly NCEP
ECCO-GODAE estimates are from ordinary least-squares solutions obtained
by “adjoining” the model to a model-data misfit function using
an ancient mathematical trick: Lagrange multipliers:
misfit to Initial conditions
misfit to the observations
adjustable parameters (controls)
vectors of Lagrange multipliers, AKA, the adjoint or dual solution
and seek the stationary point.
In control engineering, called the Pontryagin Minimum Principle,
in meteorology 4DVAR, in oceanography the adjoint method, ….
After adjustment, the model is run forward in time, in ordinary free mode,
Solved by iteration relying upon knowledge of the partial derivatives
using the adjusted parameters.
of J with respect to x(t), u(t), using automatic/algorithmic
differentiation (AD) software tools. Will skip all that here.
Two major difficulties: the size of the problem, and the need
to understand errors in everything.
Errors and uncertainties
• The case for remaining vigilant:
– current instruments at the edge of technology, error estimates
• Recent corrections published for almost all observations:
– Argo (depth-errors, …, Lyman et al., 2007)
– XBT (fall rates, …, Gouretski & Koltermann, 2006)
– Altimetry (geocenter motion, …, e.g. Lavallee et al., 2006)
– SST (Thompson et al., 2008)
• Error/uncertainty estimates and updates remain crucial,
for new observations and model representation
• G. Forget and C. Wunsch, 2006: Global hydrographic variability and the data weights in oceanic
state estimates. J. Phys. Oceanogr., 2007.
– Ponte, R. M., C. Wunsch and D.Stammer, 2007: Spatial mapping of time-variable errors in Jason-1 and
TOPEX/POSEIDON sea surface height measurements. JAOT, 2007.
Thompson et al.,
Zonal sum, 15-year time-mean
meridional vol. transport/meter
upper 300m only
Zonal integrals are not particle
Zonal integral, vertical velocity, 15-year time mean.
Within error bars, there is consistency of the meridional heat
transport of the model and that from independent calculations using
only thermal wind balance and Ekman flows.
W. Broecker, Science, 2003:
The Younger Dryas
The prevailing view of this cold snap is that it was triggered by a catastrophic release of
fresh water stored in proglacial Lake Agassiz (6). This release was initiated when the
retreating margin of the Laurentide ice sheet opened a lower outlet, allowing much of the
lake's stored water to flood across the region now occupied by the northern Great Lakes
into the St. Lawrence valley and from there into the northern Atlantic (Fig. 1). On the
basis of reconstructions of the pre- and post-diversion shorelines of Lake Agassiz, it has
been estimated that 9500 km3 of water was released (6). If released over the course
of a single year, this flood would match today's net annual input of fresh water to the
Atlantic Ocean region north of 45°N. In most ocean models, an input of this magnitude
cripples formation of deep water in the northern Atlantic (i.e., it greatly weakens or even
shuts down the model's conveyor circulation).
fresh water anomalies created
at Greenland/Antarctic margins
Stammer, JGR, 2008
Consider the basic time scale of change.
Do we see anything inconsistent with the
elementary theoretical ideas?
Stammer, 2008, JGR
upper (0-1165m, 1165-4450m,
4450-bottom), intermediate, and
abyssal ocean mass flux
heat (enthalpy) fluxes
26N North Atlantic
3- monthly average mass flux
Expanded scale, variability month to month of midwater mass flux,
1165-4450m. Aliasing is a serious issue in the use of synoptic sections
to estimate time-average values.
26N in Pacific
3-monthly avg. zonal integrals
Antarctica to Australia
~44% of the variance.
First EOF, zonal volume monthly data
First EOF zonal volume,
annual mean data ~40% of
the low frequency variance
2nd EOF meridional volume flux based on monthly means. ~15% of
low frequency variance
First EOF, meridional enthalpy (heat)
transport. Ocean is taking up heat,
3.2W/m2 (in the NCEP/NCAR
reanalysis---ocean data inadequate to
force a smaller value).
First EOF, atm. forcing
Why is there an obsession with the North Atlantic? About 10% of the
area of the world ocean. The atmosphere is in contact with it for a
comparatively brief time. Signals of oceanic influence on the
atmosphere are marginal at best.
(Wikipedia article asserts that the MOC exists only in the Atlantic
Ocean. One wonders what the author knows of the rest of the
A Summary Statement
There is no observational evidence of any major change taking place in
the overturning circulation of the ocean over the past 15 years when
global ocean observations became available.
All of our theory suggests that in the perturbation regime, mid and high
latitude baroclinic shifts require multiple decades for adjustment.
The influence of the ocean on the atmosphere is difficult to detect on
decadal time scales. (Didn’t talk about this. Weak signals.)
It appears extremely unlikely that D-O events are ocean generated. How
could the ocean undergo major baroclinic adjustment in under 10 years?
What physics would operate? How would the energy pathways be
accomplished? (Sea level change timing is not consistent….)
The circulation is very noisy. Whatever true trends exist will almost surely
require many decades to detect. Five and 10 year observational programs
will tell us something about interannual variability but little else. Climate is
an intergenerational problem.
Coarse resolution models exhibiting violent respone to massive fresh
water injections need to be understood in terms of energy transfer rates
and numerical artifacts.
If you are going to forecast, bet on ocean persistence….