Well, this Nature paper seems to be flavour of the week in the science news. It's an analysis of some measurements which seem to indicate a dramatic reduction in the thermohaline circulation which helps to keep western Europe mild in winter. Of course, such a scenario was (over-)dramatised in the film "The Day After Tomorrow" last year. So, are we about to enter a new cold spell after all? Personally, I doubt it, but maybe I'm not quite as sure as I was :-)
I should make it quite clear that I'm not going to be presumptious enough to criticise what they've done - the work looks fine, and they are up-front about their estimated uncertainties in their result. But the press does seem to have gone overboard on what is an interesting, but perhaps slightly anomalous result. Firstly, I would never want it to be thought that I prefer models to measurements - quite the reverse - but such a strong effect seems a bit unlikely, given what is understood about the dynamics. The 30% reduction in overturning circulation that they infer is a realistic estimate of the response to a century or more of substantial future warming, but no model suggests anything of this nature under recent historic conditions. In fact, according to these new figures, all of the slow-down happened prior to 1998, and half of it prior to 1992. Since 1998, things have not changed significantly. Assuming a drift speed of 3cms-1 (feel free to correct this calculation), 10 years represents 10,000km of travel, plenty of time to cover the distance from tropics to UK latitudes. But there hasn't been any significant cooling observed yet in the northern Atlantic, let alone the UK climate.
I also have to say that I'm increasingly jaded about the type of stuff that gets published in Nature. Sure, this is easily (and indeed correctly :-)) dismissed as the bitter and twisted response of someone who's never likely to get published there. But their deliberate policy to select only the most "exciting" results, which are then picked up and amplified by the press, pretty much guarantees that the outliers are given a prominence that substantially overstates their true significance (I'm wondering if I should even draw attention to this again, or if it would be more polite to let it die a natural death). When all the world's climate models predicted a steady modest decline in THC which would only partially offset the future warming, it only rated an unheralded paper in GRL. I don't see this new work in itself as being enough to justify the suggestion that the UK Govt should now start planning for a colder, rather than warmer, future. And talk of a "smoking gun" also seems a bit premature. The uncertainty in the analysis is "uncomfortably close" (their words) to the observed effect, and that's assuming they haven't underestimated their errors, which is always a possibility. But it's certainly interesting, and I look forward to future analyses.
Update
Oh ok, time for a minor nit-pick that I thought of while cycling home. They infer a change of about 8-9Sv which they admit is "uncomfortably close" to the error estimate of 6Sv on each value. But (unless I've misread something) the 6Sv refers to the uncertainty of each value individually, so the error on the difference between two values is 6*sqrt(2) = 8.5Sv. That's a bit more than uncomfortably close to the estimated change, in my book. Set against that, both the recent values are low and both the first two values are high, so I'm not claiming to have completely wiped out their result with a spot of mathematical sleight-of-hand.
Update 2
I see RC also give a somewhat guarded (and more scientifically detailed) assessment here.
I should make it quite clear that I'm not going to be presumptious enough to criticise what they've done - the work looks fine, and they are up-front about their estimated uncertainties in their result. But the press does seem to have gone overboard on what is an interesting, but perhaps slightly anomalous result. Firstly, I would never want it to be thought that I prefer models to measurements - quite the reverse - but such a strong effect seems a bit unlikely, given what is understood about the dynamics. The 30% reduction in overturning circulation that they infer is a realistic estimate of the response to a century or more of substantial future warming, but no model suggests anything of this nature under recent historic conditions. In fact, according to these new figures, all of the slow-down happened prior to 1998, and half of it prior to 1992. Since 1998, things have not changed significantly. Assuming a drift speed of 3cms-1 (feel free to correct this calculation), 10 years represents 10,000km of travel, plenty of time to cover the distance from tropics to UK latitudes. But there hasn't been any significant cooling observed yet in the northern Atlantic, let alone the UK climate.
I also have to say that I'm increasingly jaded about the type of stuff that gets published in Nature. Sure, this is easily (and indeed correctly :-)) dismissed as the bitter and twisted response of someone who's never likely to get published there. But their deliberate policy to select only the most "exciting" results, which are then picked up and amplified by the press, pretty much guarantees that the outliers are given a prominence that substantially overstates their true significance (I'm wondering if I should even draw attention to this again, or if it would be more polite to let it die a natural death). When all the world's climate models predicted a steady modest decline in THC which would only partially offset the future warming, it only rated an unheralded paper in GRL. I don't see this new work in itself as being enough to justify the suggestion that the UK Govt should now start planning for a colder, rather than warmer, future. And talk of a "smoking gun" also seems a bit premature. The uncertainty in the analysis is "uncomfortably close" (their words) to the observed effect, and that's assuming they haven't underestimated their errors, which is always a possibility. But it's certainly interesting, and I look forward to future analyses.
Update
Oh ok, time for a minor nit-pick that I thought of while cycling home. They infer a change of about 8-9Sv which they admit is "uncomfortably close" to the error estimate of 6Sv on each value. But (unless I've misread something) the 6Sv refers to the uncertainty of each value individually, so the error on the difference between two values is 6*sqrt(2) = 8.5Sv. That's a bit more than uncomfortably close to the estimated change, in my book. Set against that, both the recent values are low and both the first two values are high, so I'm not claiming to have completely wiped out their result with a spot of mathematical sleight-of-hand.
Update 2
I see RC also give a somewhat guarded (and more scientifically detailed) assessment here.
1 comment:
John Fleck says -
James, your comment about Nature's power of amplification is right on the money. And Nature's press management system encourages this amplification - sending out an embargoed press kit a week ahead of time, building a crescendo of behind-the-scenes buzz about a story until the embargo lifts and all we "in the know people" (scientists and reporters who share in this collusion) breathlessly rush to share it with others. By preventing scientists from publicly discussing results beforehand, Nature in fact actively inhibits the sort of free-wheeling pre-publication discussion that ought to be the lifeblood of this sort of frontier science.
PNAS and Science do the same, along with a number of the big medical journals.
I admit to full guilt in playing this game, but not without great discomfort.
Post a Comment