The Return of `Moby Dick'


John L. Daly

Critical comment on -

Levitus, S. et al., "Anthropogenic Warming of Earth's Climate System"
Science, vol.292, p.267-270, 13 April 2001 [8]

Barnett, T. et al., "Detection of Anthropogenic Climate Change in the World's Oceans"
Science, vol.292, p.270-274, 13 April 2001 [2]


These two papers were published in the same issue of the journal `Science', on consecutive pages, both of which attracted a barrage of media publicity, most of whom like `New Scientist' [6] seemed to know the contents of the papers well before their actual release.   `New Scientist' declared "The strongest evidence to date for global warming undermines a discrepancy often cited by sceptics.",   a comment typical of other major media outlets. In fact, the papers did nothing of the kind.

Those PR considerations aside, did these papers tell us anything new about the state of the Earth's climate, or of human influence?  In other words, were they good science, even if released for headline grabbing?   

Both papers were somewhat complementary in that both covered much the same ground (thus the reason for `Science' publishing them on consecutive pages in their journal).  Good publicity they may have been, but good science they were not.

To briefly summarise their findings, the Levitus paper presented a calculation of their estimate of  increased heat content of the world's oceans,  atmosphere, and of the cryosphere (ice covered regions) during the latter half of the 20th century, their conclusion being that  the oceans had gathered nearly 10 times as much heat as the rest of the atmosphere or cryosphere combined.  This calculation was based on separate estimates of increased heat content of the oceans, the atmosphere, continental glaciers, Antarctic sea ice extent, mountain glaciers, northern hemisphere sea ice extent, and perennial Arctic sea ice volume.  Each variable was estimated separately and then combined into one global estimate of heat content increase.  Having made the grand total, the authors concluded, "the observed increase in ocean heat content may largely be due to the increase of anthropogenic gases in Earth's atmosphere."  `May'...,  but then maybe not.

By contrast, the Barnett paper concentrated entirely on the world's oceans, comparing ocean temperature data from the 1950s to the 1990s and comparing the trends with those produced by a new generation model called the `Parallel Climate Model'.  This was described as `state of the art' and regarded by the authors as superior to all previous models.  When the historical data for surface and deep ocean temperature was matched to the model, they found that both the model and the observations agreed closely when the model was forced with the greenhouse gas increases we know have occurred in the last 45 years. Since the two correlated, the authors concluded that the greenhouse gases caused the ocean temperature increase in both the model and the observed data. 


Where Levitus et al. went wrong

It is a rule in science that you must always check and verify your assumptions. The Levitus paper was dependent on some very rubbery assumptions to arrive at its conclusions.

Firstly, they accepted as a prior assumption that the oceans had in fact warmed when a lack of reliable global data for such a claim makes such an assumption premature.  Furthermore, the data they cite dates from 1955, a known cold period worldwide, with the consequent distortion which comes from beginning any data series from a known anomalous starting point.  Had the ocean temperature data been taken from the 1930s, a known warm period worldwide, a much different picture would have emerged.  That's rubbery assumption no.1

Secondly, their estimate of decrease in the mass of continental glaciers included a critical assumption that sea levels had risen +1.8 mm per year during 1955-1996.  This assumption is highly suspect, especially given that the Topex-Poseidon satellite which has been measuring sea level from space for the last 8 years, and which included the big El Niņo of 1997-98 (which itself raises sea level for a time) only amounts to +0.8 mm per year during the 1990s.  Other evidence suggests that sea levels since 1955 have changed hardly at all, particularly around the Australian coast which has recorded a net sea level rise over the same period, facing three oceans, of only +0.3 mm/yr [1].   If Adelaide is excluded from that series (due to a measured land subsidence there), the sea level rise around Australia has only been +0.16 mm/yr, or less than one tenth of that assumed by Levitus et al.  Rubbery assumption no.2.

Thirdly, their estimate for `Decrease in Antarctic sea ice extent' was based entirely on one paper which appeared in Nature in 1997 [5], which analysed a whaling records database from the 1930s to 1970s, concluding that sea ice extent had retreated over 300 kilometres between the late 1950s and the early 1970s (a period during which whaling had been suspended with no-one around to actually witness this real or imagined event).  Having examined the same database of whaling records myself, it was incredible that such absurd conclusions could be drawn from it, let alone pass peer review, or get published in `Nature'.  A detailed critique of that paper is available here [4] and is recommended reading before proceeding further.  Levitus et al. even admitted that this assumption about whaling records and Antarctic sea ice extent was questionable, but went ahead and included it in the overall calculation anyway.  Rubbery assumption no. 3.

Fourthly, their assumptions about northern hemisphere sea ice extent and volume are based on satellite imagery of Arctic sea ice which shows a one-off shrinkage in 1978, but has changed little since -  shown in this satellite graph of arctic sea ice extent [11]

This one-off transition in the late 1970s closely matches a change in the frequency of the Southern Oscillation where several decades of predominantly La Niņa conditions (an ocean cooling influence) gave way in the late 1970s to a dominance of El Niņo conditions (an ocean warming influence) as shown below -

The dominance of El Niņo since the late 1970s will inevitably warm the oceans anyway as El Niņo is itself a natural ocean warming phenomenon, particularly in the Pacific.  It should be no surprise to find that warming effect appear in the observational record.  In the last 3 years, La Niņa has reasserted itself, but too late for the period of study covered by the two papers.  As for sea ice thickness, this has seen some decrease since the 1950s, again consistent with the 1950s being anomalously cold (probably due to the dominance of La Niņa), thus giving another distorted view of trends when the start date is anomalous. However, the 40% decrease in ice thickness claimed by Levitus et al is well off the mark and itself based on some faulty assumptions relating to inter-annual variability of sea ice thickness [9]. A full discussion of Arctic sea ice extent and thickness history is presented here [3].  Rubbery assumption no.4.

And just how much warmer did the oceans become?  That's where Barnett et al. enter the story... 


Where Barnett et al. went wrong

The Barnett team also examined ocean temperature data since 1955, most of it for the North Pacific and North Atlantic where data sources were considered by the authors to be sufficient.  They presented a picture of overall warming for those two basins between 1950 and 1995 of up to +0.15°C at depths to 500 metres, +0.1°C to depths of 1,000 metres, and +0.05°C to 2,000 metres.  These results were hardly news, having been published a year earlier by Levitus [7].   It should be remembered that ocean temperatures away from the tropics change many degrees due to seasonal forcing alone, while the Decadal Pacific Oscillation (PDO) and the El Niņo-La Niņa Southern Oscillation (ENSO) also impose temperature changes of similar magnitude in the Pacific.  The North Atlantic Oscillation (NAO) does the same in the North Atlantic [10].  All these are natural phenomena unconnected with human activity.

Barnett's primary purpose in the paper was not to merely repeat the work of Levitus a year earlier in presenting a picture of trace warming in the oceans,  but rather to match that warming with the new Parallel Climate Model to see if the model could reproduce the same trend if forced with the same input of greenhouse gases which we know has happened over the last 4 decades.  Although the model was forced with these gases, it was not forced with any other known external input - nothing from the volcanoes, nothing from ENSO, PDO, or the NAO, and nothing from the Sun.  In other words, the model was only allowed to run with internal forcings (ocean currents, seasonal changes etc.) and the only external forcing included was the trace gases from human activity, mainly CO2 and also the assumed (but unproved) effect of direct sulfate emissions.

The Parallel Climate Model is not a single model per se.  It is actually a mega-model made up from separate models. The atmospheric component of PCM is the CCM3 model at NCAR in Boulder.  The University of Texas model provides the rivers component, the Parallel Ocean Program model provides the ocean component, finally a sea-ice model at the Naval Postgraduate School represents the cryosphere. 

Naturally, the model produced just the result expected, particularly since the model used the same questionable assumptions about the temperature effects of CO2 since its atmospheric component came from the NCAR model, but which has never been proved by observation.  The model's ocean temperature trend approximately matched the observed trend reported by Levitus in 2000, albeit with an unexplainable discontinuity around 1970-1980.

The results of the  model run matched against observations are shown left. 

Since the model  roughly tracked the temperature trend, the authors concluded that this model better represented climate change than any other model. Since the model produced that result through greenhouse gas forcing, they reasoned that the observed ocean warming must therefore have been produced in exactly the same way.

This is  closed logic.

The model was constrained so that the only external forcing it could react to was greenhouse gases and aerosols - the sensitivity and magnitude of that response defined by the NCAR CCM3 atmospheric model.  They did not include key forcings such as the Sun and El Niņo, nor did they include in the forced model runs for detection other man-made inputs such as indirect sulfates, aerosols, smoke, or ozone.  By excluding these other forcings, of course the PCM produced the expected result.  How could it do otherwise?

As an analogy, it's like pumping up a balloon with air and noting that the balloon expands.  The closed  logic comes if we then attribute all subsequent occasions of balloon expansion to air being pumped in, when we know full well that additional heat or a fall in external air pressure can have exactly the same effect.  In other words, the equations `air pumped in' equals `expansion', while `expansion' equals `air pumped in' excludes by definition any other external variable - an example of closed logic where the two allowed variables prove the existence of each other in a closed system.

The oceans are not a closed system. Ocean temperature also responds to all the other forcings, but which were excluded by the PCM model so that greenhouse gases could get an unimpeded `clear run'.  The authors got the result they expected by accepting a set of model conditions which favoured that outcome, and then proceeded to announce they had `proved' an anthropogenic fingerprint.  They had done nothing of the kind.

Had the model been forced with ENSO or  PDO or NAO or the volcanoes or the Sun (or all of them), but without the greenhouse gas forcing, much the same result would have been achieved because all these other forcings  have  also affected ocean temperature during the same 40 year period, mostly toward slight warming, exactly what has been observed.   The model validation experiment might have been more credible had all external forcings been incorporated, just as the real climate has to cope with all forcings, both internal and external.

In their quest to validate this new mega model, Barnett et al. dropped the rest of the greenhouse industry in `deep water'.  In spite of being limited to give a clear run to greenhouse gases only, the Parallel Climate Model still produced a warming slightly higher than the observed data.  Worse still, having announced the model as validated by the observations, they proceeded to attack the credibility of all other models as shown in the quote (left) from their paper.

They state the PCM has a `relatively low sensitivity', hardly surprising given that the model was required to reproduce the observed ocean temperature.  Having achieved that with a low sensitivity, thereby involving a much smaller theorised human impact on climate, the validity of the model could only be affirmed by a direct attack on the other models in current use, particularly those with higher sensitivities and  yielding the largest warmings.  The worst-case  IPCC scenario of a 6°C warming by 2100 is clearly in that league, and Barnett et al. undermined the credibility of these IPCC scenarios in an attempt to prove the worth of the PCM model.



These papers, like many before them, were trumpeted by the media as cutting the legs from under the arguments put by greenhouse sceptics that the warming from greenhouse gases will only be very slight (measureable in only tenths of a degree, not whole degrees), and that most studies reporting real or imagined warming trends in the atmosphere, oceans, or cryosphere, reveal serious flaws in the statistical treatment of the data, and/or in the methods of collecting that data.  

These two papers are little different.  The Levitus paper makes so many bland and questionable assumptions that their calculation of global heat gain is highly questionable as serious science.  The Barnett paper merely repeats the earlier work of Levitus in reporting a slight warming in the ocean, a warming which can just as easily be attributable to PDO, ENSO, NAO, volcanoes, and/or the Sun.  The only way they could target greenhouse gases as the cause of this miniscule warming was to exclude these other variables from the Parallel Climate Model.  In doing so, they neither proved the accuracy of the ocean warming reported earlier by Levitus, nor the validity of the model.

But greenhouse sceptics were doubtless amused to see Barnett et al. proceed to attack all  the other models, claiming they were showing much too large a human impact due to excessive model sensitivity.  Having taken the small observed ocean warming as the baseline for the sensitivity settings of the Parallel Climate Model, there could be no other conclusion - the other models were wound up much too high and giving much higher predicted temperature rises than the oceans would allow.

If anything, the Barnett paper affirms what sceptics have been saying all along, that human impact will be very small and that the present scary predictions by the IPCC are scientifically unsupportable.  Unfortunately, the media rarely reads past the abstract of such papers.

Update 1 May 2001:  As a point of clarification, it has been pointed out in email correspondence that the PCM model was forced with sulfate aerosols in addition to the greenhouse gases already mentioned.  The Barnett paper is ambiguous on this point as it states in one part of the paper that the direct effect of such aerosols was included, but then two pages later said that `the forced runs used for detection did not include indirect sulfate, ozone, so-called black aerosols, and other anthropogenic factors, not did they include external forcing due to solar variability and volcanic activity.'  The ambiguity would therefore hang on the definition of `direct' and `indirect' anthropogenic aerosols.

Given that the assumed cooling effect of sulfate aerosols is still only a theory, the fact that the sulfate-free southern hemisphere has not warmed at all since satellite temperature measurement of the lower troposphere began in January 1979, while the sulfate-rich northern hemisphere has seen a slight warming (a decadal trend of +0.113°C), suggests that the theory lacks substance.  Had it been otherwise, the southern hemisphere would have warmed relative to the northern.

It was also a point of discussion as to whether El Nino, the North Atlantic Oscillation, and the Pacific Decadal Oscillation (none of which were included as forcings in the PCM model) were `external' or `internal' forcings to the climate system. Given that one of the guest authors on this website, Dr Theodor Landscheidt, has demonstrated that all three of these major climate oscillations are closely correlated with solar activity and motions, it is therefore considered for the purposes of this review that these oscillations are indeed `external'. For more information on this solar connection to the three major oscillatioins, see Landscheidt's papers on El NIņo, on the NAO, and on the PDO.



[1] Australian Mean Sea Level Survey, National Tidal Facility, Adelaide.

[2] Barnett, T. et al., "Detection of Anthropogenic Climate Change in the World's Oceans", Science, vol.292,
      p.270-274, 13 April 2001

[3] Daly, John L., "The Top of the World: Is the North Pole Turning to Water?, 2000,

[4] Daly, John L., "Moby Dick's Revenge". 1997,

[5] de la Mare, William K., 1997, "Abrupt mid-twentieth-century decline in Antarctic sea-ice extent from whaling records"
      Nature, vol.389, pp 87-90, 4 Sept 97

[6] Hecht, Jeff, "Warm Water", New Scientist, 15 Apr 2001,

[7] Levitus, S. et al., Science vol.287, p.222, 2000

[8] Levitus, S. et al., "Anthropogenic Warming of Earth's Climate System", Science, vol.292, p.267-270, 13 April 2001

[9] McLaren, A. et al., "Variability in Sea-Ice Thickness over the North Pole from 1977 to 1990", Nature, v.358,
       p.224, 16 July 1992

[10] NAO winter index chart -

[11] National Assessment Synthesis Team (NAST), "Climate Change Impacts on the United States: The Potential
        Consequences of Climate Variability and Change"
- Overview document, USGCRP, June 2000

Return to `Still Waiting For Greenhouse' main page, or use your "back key"

FastCounter by bCentral