Sunday, March 07, 2010
This is the angriest I have been for a while on the issue of supposed Catastrophic Anthropgenic Global Warming. The Met Office recently issued a new "review" of papers produced since the IPCC 2007 report. The usual alarmist suspects in our - increasingly surreal - mainstream media jumped on it immediately with spin completely disproportionate with the actual contents of the report. The reporting in the Independent contains outright lies, with the Guardian and BBC including a substantial dose of exaggeration. The Times offers the most neutral summary I've seen, though is let down badly by the headline - which, of course, is what will stick.
To give the BBC credit it looks like their reporter actually spoke to the lead author, Peter Stott. What is clear from all of the mainstream reports I have seen however is that none of the reporters have actually read the report themselves. I'm becoming sick to the back teeth of this "churnalism". And I find myself asking why it is interested individuals like us bloggers who take the time to actually carry out the "investigative" part of "investigative journalism" on the part of these lazy hacks.
Both the Independent and the Guardian, describe this as a "powerful riposte" and a "fightback" respectively against sceptics, despite the BBC reporting that "Dr Stott denies that the study has been published as part of a fight back by the climate research community." The Guardian, bless 'em, talk about the "latest battle" in the "climate change fightback". Don't let your bias show there chaps, will you? When did scientific discussion become like a pay-per-view wrestlemania match? Oh - that's right - when the likes of the Guardian infantilised the whole thing.
Pay per view
On that note - some vital statistics on this "review". The first one being that the paper is not publicly accessible. Probably one of the reasons our cast of lazy churnalists haven't read it. It costs $35 to gain 24 hours access pay-per-view for the article. Seriously. This is despite the fact that the Met Office is a massively expensive publicly funded institution. And despite the fact that Stott is quoted as saying: "I just hope people look at the evidence of how the climate is changing in such a systematic way. I hope they make up their minds on the scientific evidence." Sure Peter - "evidence" that Joe Public has to pay $35 for 24 hours access to. The only way to get it "free" is to access it via a University network, but even then that is only the case if your institution is subscribed to this journal. And it means the paper will not get the bredth of critical attention it really needs. Most people will rely on the mainstream's "analysis".
Second, and more importantly, is the neutrality of this "review". It covers 110 papers published since the IPCC-2007 report (it is *not*, as the Independent claims, an "update" on the IPCC report). A number of questions have been raised regarding how the 110 papers were selected (by people posting in the comments section to the "news" articles of course - not the journalists, silly!). Well - here's Guardian/BBC style neutrality for you: Out of the 110 citations, 24 are papers by Stott (first, or second author) and a further 22 are by other authors of the "review". So with 46 citations, 41% of the papers in this supposedly independent "review" are self citations.
Third, the reporting on this "review" give the impression of an extensive, comprehensive piece. It is not. The entire paper is a 20 page puff piece in 12-point font and lots of diagrams, and out of those 20 pages, 4.5 are the list of references.
And - a final statistic for you - the word "robust", or "robustly" is used eight times in the text.
Recycling old crap. (But if its behind a paywall, who notices?)
So, what is the primary basis of the paper? The paper claims to be able to "detect" climate change and "attribute" it appropriately. This is where the "human fingerprint" idea comes in. The paper's authors claim they can show a clear human influence.
So how do they go about this. First they have to engage in "detection" of a problem. How do they do this?
Wait for it -
They use the HadCRUT temperature dataset, complete with Michael Mann's hockey stick as their primary example of how to detect a human fingerprint.
Someone kill me please.
In case you're new to this controversy - here is John Daly's original critique of Mann's Hockey Stick, and here is Bishop Hill's analysis. Note that *both* of these critiques were prior to Climategate. The Register's Interview with Bishop Hill makes for slightly easier reading.Chris Booker links it all together - post climate-gate - here.
If after reading the above, you can't see why any subsequent research, or *any* assertion for that matter, based on the HadCRUT data set and, in particular, Michael Mann's hockey stick contribution, is automatically worthless then I can't help you and you best stop reading - go off and enjoy whatever catastrophic toss-fantasy the alarmists have in store for you.
Annoyed? It gets better. Stott et al. place the HadCRUT data over a set of 'control' data. Where does the 'control' data (which shows no significant variation before 1850 and after 2000) come from? It is: "A 'control' simulation of a couple ocean-atmosphere climate model over many centuries, with no changes in the external drivers of climate such as increases in greenhouse gas concentrations or in solar output, does not exhibit the sustained rise in temperatures seen in the observational data."
- Of course it doesn't you dimwit!
The built in assumption is that there are no significant variances in the climate prior to 1850, so they can use a computer simulation to fill in for actual data! (Andthe few sources of real data we have for earlier periods are extremely problematic for the alarmists) One of the key parts of the scandal surrounding the HadCRUT data, and the "hockey stick", was its use of "proxies" such as tree-ring data, for previous temperature records. Only, the tree-ring data diverged from observed records in this century, so Michael Mann fraudulently "fudged" the data to make it fit the "hockey team" bias.
There is definitely a Mann-made "fingerprint" here. His dirty paws are all over the data. This presentation of "research" is one of the most shockingly disingenuous things I've seen since "climategate" itself: i) Stott et al. Assume no significant variation in climate outside their chosen window (roughly 1850-2000) so think it is OK to replace what little data we do have for the prior period with a "model" of no variation and ii) during the chosen window, humans began expelling CO2 into the atmosphere from industrial activities. The HadCRUT data (and other land based temperature data sets) showed a modest increase in temperatures in this period. However - not only are all the land based data sets now suspect (especially HadCRUT), but Mann forced his figures to fit with the increases in human CO2 contributions.
Stott goes onto establish statistical significance at the 5% level for warming in the 50 year period 1959-2008. Of course this will seem significant - the bald assertion is that the correlation between CO2 output and temperature can be clearly observed when compared to the "control simulation", because no such rises are present in the control.
This isn't science. It is complete rubbish.
A litany of caveats and qualifications
The paper goes on to cover various environmental phenomena and the supposed "detectable" human fingerprint. The process for obtaining this fingerprint is rarely made clear with the other environmental indicators the "review" subsequently covers - the gold standard for the initial "detection" phase is apparently the above completely rubbish HadCRUT analysis.
What is apparent from beginning to end though is that it is chock-full of caveats and qualifications. Many of them are very telling, because Stott et al. have effectively taken on a number of the points sceptics have made over the years. The MSM reporting missed this because none of them read the damn paper!
I'll go through a sample below:
"decade long trends with little warming or cooling are to be expected under a sustained long-term warming trend".
Uh huh. Might this mean that sceptics have been correct all this time that there are other factors than CO2, (including other Anthropogenic influences and natural influences) that can significantly affect the climate to an extent comparable with CO2?
"Further refinement of our understanding of the causes of decadal variability would benefit from tracking the changes of energy within the climate system and better understanding of the role of natural and human-induced external drivers of climate, including, for example, the effects of changing solar activity." [Emphasis mine]
^ Stott et al. Would be called "deniers" in some circles for saying such a thing.
Now, the "standard approach" Stott gives for identifying a "fingerprint" "is to use a climate model to determine the expected response to a particular forcing....Once the fingerprints have been derived, an analysis is carried out to determine if there is a significant manifestation of these fingerprints in the observations" So in other words, just as with the "great" example Stott gives above of "detection analysis" (HadCrut), it all rests on assumption first, observation second. This seems like more of that fuzzy science where an assumption about how things should work trumps actual observational data from which you *then* are supposed to form a theory.
Stott says later "The fingerprints...are estimated from the average of a finite number of simulations [Ed: hah!] with identical forcings but different initial conditions (typically 3 or 4 for most analyses), and are contaminated by internal variability (which reduces as more ensemble members are averaged)". So the "fingerprints" are derived from assumption-laden simulations. He at this point again invokes the HadCrut "detection" above as the basis for a sound way to estimate internal variability. So one assumption laden data-set of questionable integrity is used as the base line for the further-assumption laden simulation of "fingerprints". Is this crap squared?
More caveats and qualifications:
"The assumption of linearity is found to hold for some combinations of forced changes, particularly the direct effects of sulfate aerosols and greenhouses, although there is evidence that additivity does not hold so well for some other combinations, including greenhouse gases in combination with the indirect effects of earosols and greenhouse gases with solar forcing;"
"...the response to non-greenhouse gas anthropogenic factors is under-estimated"
So other factors *are* significant?
"it is appropriate to conclude that significant observed changes are attributable to human influence."
Take note here. "Human influence". Not just CO2.
Aerosols are a significant factor:
"it is very likely that aerosol cooling is suppressing a major portion of current greenhouse warming".
In true assumption-laden form though, the conclusion naturally is "As a result, additional warming is implied if aerosol pollution is removed from the atmosphere in future." (In other words - one of the classic alarmist hedges - our observed data does not provide a linear relationship between CO2 and temperature, so we'll hide the energy in another factor).
Models at sub-global scales.
Stott et al go on to analyse climate data at more local levels - "going down to the scales of climate model grid boxes or of order 500km". Grid boxes. Oh goody.
"The first systematic investigation of conteinental scales to use the optimal detection regression approach described above was by Stott". By the author of the very paper we are reading? Fantastic. And this wonderful new "optimal detection regression approach" is applying statistical multivariate regression on the basis of the assumption-laden "fingerprint" models that use the assumption laden HadCRUT "detection" model as a control for natural variability. For the non-statistically au fait, "regression" is a statistical technique that allows you to predict trends over time, varying some variables and holding others constant to see what the results might be. The entire practice, while an interesting exercise, is heavily dependent on initial assumptions. In other words, it doesn't produce what I regard as "data" - it just shows you were your assumptions might take you with "real" data (assuming any real inputs are provided - it looks in many cases in the Stott et al. cited research that this itself is a tenuous assumption). A good example of this is given later: "When signals were regressed individually against the observations, an anthropogenic signal was detected in each of 14 regions except for 1, central North America, although the results were more uncertain when anthropogenic and natural signals were considered together." Read: regression + assumptions + real observed data = some results contrary to assumed expectations when extrapolated into projected trends.
Stott goes on (about his own research) "This study found a detectable change over the 20th century in decadal mean temperatures over each of the six populated continental areas...and furthermore found that these changes could only be reproduced with the inclusion of anthropogenic greenhouse gas emissions."
Really? No mention of Urban Heat Island effect in the data sets then? You know - that effect that it's possible for a 12 year old kid to detect, using an excel spreadsheet?
More uncertainty when we start looking at regional variations:
"Relevant model simulations considering the different forcing factors separately are often not available, so the attribution to different forcings is limited to a consistency analysis rather than a full attribution analysis in which all plausible forcing factors are considered. For example, increases in irrigation in California have been important for regional temperature trends, while land cover change can be important for regional temperature changes."
Another interesting point - a lot of the regression simulations only go up to 2000. With real data available for the last decade also, why did they not continue the "simulation" and match it against real data? Is it perhaps because they don't match?
"attribution at regional scales is limited at present by the relatively lower signal-to-noise ratios, the difficulties of separately attributing the effects of the wider range of possible forcing factors at these scales, and lmitations of models in capturing some characteristics of regional climate variability."
Right. Some question begging is occurring here. It is very similar to the "climate is not weather" conundrum. At what point do the separate "weather" data points morph into "climate". Similarly, the "global" data is made up of "local" data. Stott et al. seem to carry out regular hedges on the regional data, where it doesn't fit their preconceptions (even sometimes the "simulated" and "regressed" data FFS!). The fact is that as soon as they look at more local areas, they are forced to apply more (local) factors and thus end up with more complex models less likely to support their position. The 'global' models are more simplistic and so therefore more likely to be amenable to smoothing.
Another strange statement here:
"Human influence is estimated to have more than doubled the likelihood of positive warming trends in every region considered except central North America."
How interesting. Especially when you consider the 1934 problem.
On the hydrological (water) cycle
Yet more hedges, caveats and qualifications....
An anticipated consequence of these flux and transport changes is that wet regions should become wetter and dry regions drier. Many of these anticipated changes, reasoned from physical principles, have been observed and confirmed by climate model simulations."
Sorry, wait. How does a "simulation" confirm anything? And how on earth do Stott et al. compromise this bald statement with this research cited in National Geographic: "Scientists are now seeing signals that the Sahara desert and surrounding regions are greening due to increasing rainfall.If sustained, these rains could revitalize drought-ravaged regions, reclaiming them for farming communities.This desert-shrinking trend is supported by climate models, which predict a return to conditions that turned the Sahara into a lush savanna some 12,000 years ago." Uhhhh......
"Mitchell et al. theorized that the latent heat of condensation in the troposphere is balanced by radiative cooling."
Carefull Peter, that sounds a bit "denialist" to me....
"Liepert and Previdi show that 20 years may not be sufficient to determine whether models and observations agree on the rainfall response to global warming."
"Generally, however, detection and attribution of regional precipitation changes remains difficult because of low signal-to-noise ratios and poor observational coverage. To date there have been no detection and attribution studies of precipitation over oceans because the available satellite data sets... are short and not considered to be sufficiently reliable for this purpose."
This is one of the clearest expressions of doubt in the whole paper. So no alarmist fingerprint to be found in regional precipitation then. Which means global precipitation too, on the basis of the available evidence.
Runoff and Drought:
"Monitoring and understanding changes in runoff and drought is more difficult thatn for temperature and precipitation because soil moisture is poorly observed, and soil moisture and runoff changes are difficult to constrain from the residual difference between precipitation and evaporation, both of which are also relatively poorly observed.
Many factors can cause soil moisture and runoff changes, including changes in climate, land use, stream management, water withdrawal, and water use efficiency by plants in high CO2 environments."
Despite these expressed caveats, Stott then goes on to cite the Palmer Drought Severity Index, without further explanation or comment, claiming that increases in the index "has been attributed to anthropogenic influence".
A pretty funny couple of statements in the context of recent events (and alarmist claims regarding precipitation):
"In a warmer world, less winter precipitation falls as snow"
"There has been a reduction in the ratio of precipitation falling as snow in the Western US that cannot be explained by climate models including only the natural effects of solar and volcanic forcings and which has been attributed to anthropogenic forcings."
Right Peter. Whatever you say.
Arctic / Antarctic:
Stott refers to "dramatic sea ice retreat" and - guess what? - "Human influence on Arctic sea ice is detectable in an optimal detection analysis." Really? Where is the mention of the recent recovery?
This is hilarious. All the Antarctic gets is:
"In contrast, Antarctic sea ice has not significantly decreased."
Whoops! No further comment is given whatsoever as to why this might be the case.
"anthropogenic circulation changes are poorly characterized"
OK - so we don't have a clue on any anthropogenic influences here then?
Well, Stott then cites himself (again): "were able to detect an anthropogenic response independently of the natural response and with an amplitutde consistent between model and observation." - so Peter, given your first comment, do we trust this research or not?
"It has been estimated that over 80% of the excess heat built up in the climate system by anthropogenic forcing has accumulated in the global oceans."
This is a hedging assertion used by alarmists to explain where the energy goes when the observed pattern does not exhibit a proportional temperature increase with CO2.
So how much do Stott et al think they know about this?
"..the subsurface ocean has been sparsely observed in many regions, and sampling errors remain an issue when comparing observed and modeled timeseries of ocaen properties, with the choice of infilling method being potentially important in poorly sampled regions."
No shit sherlock. In other words, not much, and too many gaps in the data to enable them to fill the gaps with bullshit without anyone noticing.
Apparently we can attribute "...the short term cooling episodes to volcanic eruptions and the multidecadal warming to anthropogenic forcing." Right - so nature *does* have an in built cooling mechanism (which will presumably continue), and if Stott et al are thereby going to attribute the warming primarily to anthropogenic forcing (N.B. Note *again* this does not state CO2), then without it we would all be significantly colder?
And then we have the issue of salinity - this has caused a lot of confusion in people commenting on how the journalists have presented this. Interesting because Stott only gives a paragraph to the whole topic - here is what he says:
"It has been suggested that freshening at high latitudes is consitent with observed increases in precipitation at high latitudes although climate model studies suggest that Atlantic freshening could be associated with changes in northward advection associated with variability of the meridional overturning circulation.
An optimal detection analysis [Ed: LOL] of Atlantic salinity changes by Stott et al. [Ed: LOL again] detected a human influence on the observed increases in salinity at low latitudes but found that high-latitutde changes, including a recent reversal of the freshening observed previously, are consistent with internal variability."
So, despite the way the salinity issue in the "review" has been reported in the press, there is nothing significant here and a good deal of it is attributed to natural variation.
This is possibly the best section. The alarmists are continually banging on about extreme weather. But even Stott et al. couldn't fit their simulated data to the pattern.
He writes it off for hurricane activity for example:
"...the importance of the anthropogenic increase in sea surface temperature in the cyclogensis region for past and future changes in hurrican activity is still poorly understood."
And his conclusion is pretty clear:
"In conclusion, while there has been progress since AR4, there are still many gaps in our understanding of changes in extremes and in our ability to attribute observed changes to particular causes."
Sorry alarmists - no evidence here for your extreme events fantasies.
Is a little bizarre given the actual content of the paper - and appears to have been what has grabbed the headline writers attention:
"The wealth of attribution studies reviewed [Ed: including all 46 of our self-citations] in this article shows that there is an increasingly remote possibility that climate change id dominated by natural rather than anthropogenic factors."
Notice what Stott has done here. Taken as a whole, his review has shown, in fact that both natural, and non-CO2 anthropogenic factors have a significant role to play in the climate. In many ways, he has actually presented quite a sceptical position in some areas in that he has shown (his "optimal detection analyses" notwithstanding) that climate is being driven by a range of factors, of which human contributed CO2 is only one. Remember the thrust of the paper is anthropogenic influences NOT CO2.
My own conclusion is that, not only does the "review" completely fail to support the claims being made for it by the quisling useless media hacks, where it does make a case at all, it is one for much more doubt over climate issues, and for the inclusion of a great many factors, including non-CO2 anthropogenic factors. Notice how the BBC wrote their title: "Climate change human link evidence 'stronger'" - many people will immediately read that as - 'case for human CO2 forced catastrophic climate change stronger'. Which isn't what the review shows - and that's assuming you trust the methods used by Stott et al. Which I don't at all. The combination of extensive self-citation, with an extremely dubious statistical modelling method (the "optimal detection analysis") that often seems to skip real data simply looks like utter utter crap to me.
If we had a media deserving of the name we'd all know this.
Posted by Katabasis at 1:10 PM