Some comments on the Bryden et al paper (Nature, 1st Dec. 2005) on Atlantic Circulation changes

links:   (look here for short presentation)

            original Bryden et al. paper 

date: 04 Dec 2005 
last update 10 Jan 2006:
file: bryden_discuss01.html
compiled by Francis Massen

Stuart Staniford: (from

Thanks for responding. If we take these axioms:

a) The wind driven gyre fills the entire basin (between the rapid western boundary current and the slower southward returning currents in the rest of the Atlantic) , and extends vertically down to the top of the NADW.

b) The global conveyor says there's a net transfer of warm shallow water northwards.

There is no way that these facts can be reconciled without requiring an imbalance between the mass transport in the two arms of the gyre. There's nowhere else for northbound shallow water to sneak through. The more northerly transfer of shallow water one would like, the more imbalance one would expect to see. So if the conveyor is slowing down, one would expect that to show up as a decreasing imbalance between the western and eastern halves of the gyre.

The question that remains is whether the required reduction in imbalance would be achieved by 1) a slowing of the Gulf stream, 2) a speeding up of the Eastern half of the gyre, or 3) some combination of the two. Bryden's data suggest 2), but why should this be so?

(For background, I have no training in oceanography, but do have a rusty PhD in Physics).

It appears to me that the surface thermohaline currents in the conveyor must be mediated by sea-level inhomogeneities (long-lived, average ones). A given chunk of water moves because of the net forces on it, and so a net northward transfer of water not driven by friction must require that the average isobar surfaces slope downwards in the shallow ocean as you go north. In general, the regions of expanding warming upwelling water in the Indian Ocean, North Pacific, or wherever they are, must create slight bulges in the surface, and the regions of shrinking, cooling, sinking water in the Arctic must create slight depressions in the sea surface (again, I mean in a very low pass sense - obviously storms, tides, etc, create all kinds of short-terms signals obscuring this). I can't find any confirmation of this in a quick search, but I don't see how else it could be. (Surely someone must study sea level inhomogeneties and make maps of them? I can't find it).

Now in the Gulf stream, this isobar gradient would be expected to reinforce the frictional wind forces, while in the eastern atlantic, it would oppose the frictional wind forces (thus leading to the necessary imbalance between the arms of the gyre I noted above). However, the portion of the Atlantic that is the southern gyre is much much larger than the portion involved in the boundary current. We would expect isobar deviations to more or less radiate out symmetrically from areas of higher sea level. Thus this isobar gradient is able to have a far larger impact on the total transport through the broad eastern side of the gyre than it is on the very narrow western side of the gyre. Thus changes in the THC mainly show up as changes in the Eastern side of the gyre.

I realize this is only a cartoon of the complex physics that must go on. However, if basically correct it would explain what Bryden et al have observed.

Martin Visbeck (expert in this particular area of the science)

I am a rare visitor of your site (, link added by f.Massen) but like the forum that allows interested folks to ask questions. The paper by Bryden reports on a particular calculation done to infer the volume transport of the Atlantic Ocean along 25N at five instances in time (now granted this is expensive so we cannot get to do many of these types of transects). The method he applied is only briefly described, but seems to be using a fixed zero reference level. This way of getting at transports was done in the pre 1990s area of oceanography and has been shown by many subsequent researchers to be not optimal. More modern approaches use inverse methods to calculate the reference velocity for each station pair under a set of physical contraints (Such as the Ganachaud and Wunsch reference). This sounds a little complicated for the non experts, but needs to be done since no DIRECT measurements of flow were taken.

An even more up to date way to get transports in the ocean is to use ocean models constraint by observations (so called data assimilation or ocean state estimation in several forms). Both of these methods have demonstrated to be more accurate and can yield significantly different results for two reasons:

1) The assumed level of zero motion trend out to be not optimal or non existant.

2) The particular week, when the measurements were taken had a transport that was higher or lower than the say annual average. (aliasing)

Bryden reports an error bar of 5Sv which is based on an invese model result, and could be even higher for his method. His method does not allow to calculate errors easily. So I would not be to confident in that number for his method.

The sampling error can be large as results from realistic ocean model consistently show. To alleviate this problem the same group has one of the largest arrays of moorings in the water to get daily measurements of this transport (RAPID/MOCA array funded jointly by the UK and USA). They have not analyzed the first year of data yet, but in my lab we have looked at results from a similar set of moorings at 15N (Uwe Send's work) and find rather significant variability on weekly to monthly time scales (but no trend over the 4 years of data). This togther with the model findings should raise some flags (as also Quadfasel did) that the results might not stand the test of time.

Secondly, someone else in the comments and Quadfasel in the N&V piece mentioned the direct flow measurments in the Faroe - Shettland channel, one of the two main pathways for dense water out of the Nordic Seas, as supporting evidence. That record showed a declining transport by up to 15% until 1999, when the Science paper was written. Bogi Hansen has recently shown in several seminars that the updated time series until 2003 sends a different message: The flow is back up to the 1995 level. i.e. no change over the last decade... (unfortunately these updates usually does not make a Science or Nature paper in particular if the bottom line of the argument has shifted...)

Similar signals of interannual variability without a discernable trend come from direct flow measurements in the western boundary current at the exit of the Labrador Sea (Fritz Schott and colaborators) (north of Brydan's section) and Uwe Send's record at 15N to the south. In my assessment that is an issue that needs to be looked into in a larger synthesis paper, and not just a single point analysis.

None of the data assimilation models show a strongly declining trend in the mass flux at 25N in fact many of them show a weak increase overall...

And Gavin you are right ... the way the calculation is done mass needs to balance (since flow is only inferred via thermal wind and a reference level...) So the compensation is part of the method.

My personal take is that this remains an open issue in oceanography and climate research. We simply have not all the tools and observing systems in place to give these numbers with confidence. One thing is clear, though, that only the combined model-data synthesis will give us robust and believable answers. This paper unfortunatly does not attempt to do that.


Pat Michaels:  read comment in TCS here


Co2science: read here comment on a related paper


Die Welt: (4th Dec.): read here


Carl Wunsch:   Gulf Stream safe if wind blows and Earth turns 

Sir - Your News story "Gulf Stream probed for early warnings of system failure" (Nature 427, 769 (2004)) discusses what the climate in the south of England would be like "without the Gulf Stream." Sadly, this phrase has been seen far too often, usually in newspapers concerned with the unlikely possibility of a new iceage in Britain triggered by the loss of the Gulf Stream.

European readers should be reassured that the Gulf Stream's existence is a consequence of the large-scale wind system over the North Atlantic Ocean, and of the nature of fluid motion on a rotating planet. The only way to produce an ocean circulation without a Gulf Stream is either to turn off the wind system, or to stop the Earth's rotation, or both.

Real questions exist about conceivable changes in the ocean circulation and its climate consequences. However, such discussions are not helped by hyperbole and alarmism. The occurrence of a climate state without the Gulf Stream anytime soon - within tens of millions of years - has a probability of little more than zero.

Carl Wunsch
Earth, Atmospheric and Planetary Sciences,
Massachusetts Institute of Technology" (Nature 428, 601, April 8, 2004)