A question of attribution

On the day the SPM was published, I wrote about the key IPCC claim that it was “extremely likely” (i.e. at least 95% certain) that human influence had caused most of the warming since 1950. This was an increase on the AR4 attribution statement of 90%, which seemed odd in the light of another six years with no warming, contrary to climate science predictions, and increasing number of papers from mainstream climate scientists acknowledging a discrepancy between models and observations. However, a direct comparison is not appropriate since the IPCC moved the goalposts from greenhouse gases to all human influences. At the time of that post the justification for the 95% claim was not available, but the SPM referred to Chapter 10 of the main report, “Detection and Attribution of Climate Change” in particular section 10.3.

A post at Realclimate tries to address the attribution question but does little more than re-state the IPCC claim, based on another IPCC figure and its error bars.

A post at the manicbeancounter blog notes that the uncertainty levels  for CO2 forcing has increased greatly between between AR4 and AR5, which seems inconsistent with the claim of increased certainty.

Apart from these two, I can’t find much detailed investigation of the 95% claim and how it arises from the main report.  Are there other blogs I missed?

Chapter 10 starts off with the statement that “More than half of the observed increase in global mean surface temperature (GMST) from 1951 to 2010 is very likely due to the observed anthropogenic increase in greenhouse gas concentrations.” This is almost exactly the same claim as made in AR4. So it seems that the inflation from 90% to 95% results from the switch from greenhouse gases to total anthropogenic contribution.

The second paragraph states the 95% claim in the SPM, with slightly different wording: “It is extremely likely that human activities caused more than half of the observed increase in global mean surface temperature from 1951–2010.”

Here is figure 10.5,  quoted at RC. (My screen-grabber has again garbled the colours, so apologies for any confusion)

It shows the observed warming trend at the top, followed by the trend attributable to greenhouse gases, the trend attributed to total anthropogenic factors, and the (cooling) trend due to man-made aerosols. Each of these has error bars associated with it. There is a very obvious problem with this figure, pointed out by Clive Best in his comment at RC:

“Can you explain why in figure 10.5 the error bar on ANT is so small? Naively I would expect this to be the sum of GHG and OA. This would then work out to be an error on ANT of sqrt(2*0.36) = 0.8C. This is also not explained in chapter 10.”

I had a look through the SOD version of Chapter 10, the version reviewed by scientists, to see if this strange figure was there – it isn’t.

The only way the figure makes any sense is if the IPCC has decided a priori that almost all the warming must be anthropogenic, because they can’t think of anything else, but are uncertain how to divide that warming into greenhouse and aerosol effects.

But even this explanation does not make sense, in view of this figure from Chapter 8, showing the total “effective radiative forcing” of GHGs, aerosols and the total anthropogenic.

In this case the error bars add up in a sensible way, as suggested by Clive – the error bar for total anthropogenic forcing is greater than that either GHGs or aerosols.

So the 95% claim seems to be based on a nonsensical figure that is contradicted by another graph elsewhere in the IPCC report.

[Culture spot: “A question of attribution” is a play by Alan Bennett about Anthony Blunt, custodian of the Queen’s pictures and Russian spy.  According to Wikipedia, Bennett described the piece as an “inquiry in which the circumstances are imaginary but the pictures are real.”  There is a great TV version with James Fox and Prunella Scales.]

The ocean heat graph (Part 2) – signal or noise?

In Part 1 I showed the ocean heat graph from the IPCC WG1 SPM, got rather distracted by the fact that I could not find the Supplementary Material that supposedly listed the datasets, showed the corresponding graphs from Chapter 3 and noted that they were described as ‘estimates’ there but not in the SPM.

The IPCC graph shows estimated total ocean heat content since 1950, in units of Joules, with the tick mark intervals on the axis showing an apparently enormous numbers such as 10^23 Joules.  What the IPCC does not tell the policymakers however, is what this corresponds to in degrees C.  In fact a very simple calculation shows that 10^23 Joules would correspond to a rise of only about 0.04C.

Lucia Liljegren produced a helpful version of the graph with additional axes showing the corresponding temperatures:

Her version is 0-2000m so is not directly comparable with the IPCC 0-700m one.

It’s no secret that the corresponding temperature increase is only a few hundredths of a degree. The website of the ARGO system for measuring ocean temperature mentions a temperature change of 0.06C since the 60s and the Levitus et al paper talks of  0.09C since 1955 in the abstract. But the IPCC SPM does not mention these small numbers, talking instead of a 0.11C/decade warming in the upper 75m of the ocean.

These very small temperature variations raise the question of whether there is any significant meaning in these graphs at all. How large are the error bars? How accurately can ocean temperature be measured? How accurately could it be measured in the 1950s? To within a hundredth of a degree? I think not.  As well as the temperature accuracy, there is the question of coverage. To be able to produce this graph accurately, you need to measure the temperature all over the ocean and at all depths, at all times back to the 1950s.

Here is an interesting graph from the Ishii and Kimoto 2009 paper, used as one of the IPCC sources

The lower curve, V6.7, surrounded by the grey band, is the version used in the AR5 report. The two upper curves, one solid and one dotted, V6.2 and V6.3, are from a 2006 paper by the same authors. Notice that the curves are quite different. The difference between them is of the same order as the variation in the graphs themselves! In the 2006 version, the heat content in the 1970s was about the same as it is now. So what is the difference? Well, the newer results incorporate what the authors refer to as ‘bias corrections’. An adjustment has been applied to the data, which just happens to work in such a way as to cool past temperatures, and produce a much more steadily increasing curve.

This all seems a bit familiar.

The ocean heat graph (Part 1)

This issue was discussed at Judith Curry’s blog, and Lucia’s Blackboard, referring to this post by Lubos Motl, following a post by Stefan Rahmstorf at RealClimate, all before the release of the AR5 report.

In the SPM, the upper ocean heat graph is Figure SPM.3c,

with the caption (on page 6)
(c) change in global mean upper ocean (0–700 m) heat content aligned to 2006−2010, and relative to the mean of all datasets for 1971 …  All time-series (coloured lines indicating different data sets) show annual values, and where assessed, uncertainties are indicated by coloured shading. See Technical Summary Supplementary Material for a listing of the datasets.
[My screen-grabber seems to have changed the colour scheme].

A first stumbling block is that I cannot find anything called “Technical Summary Supplementary Material” – there is a Technical Summary, but it does not have any supplementary material and does not list the datasets. Is this a mistake in the SPM, or is there some more material that has not been released yet? I have asked some IPCC authors; one says he doesn’t know.

In Chapter 3 there is the following version of the figure

which seems to show the same data, and indicates the sources. Here is the caption:
Figure 3.2: a) Observation-based estimates of annual global mean upper (0–700 m) ocean heat content in ZJ (1 ZJ =10^21 Joules) updated from (see legend): (Levitus et al., 2012), (Ishii and Kimoto, 2009), (Domingues et al., 2008), (Palmer et al., 2007), and (Smith and Murphy, 2007). Uncertainties are shaded, and plotted as published (at the one standard error level, except one standard deviation for Levitus, with no uncertainties provided for Smith). Estimates are shifted to align for 2006–2010, five years that are well measured by Argo, and then plotted relative to the resulting mean of all curves for 1971, the starting year for trend calculations.

Spot the difference in wording between the SPM and the main report.

Anyway here are the papers:

Levitus et al 2012,  World ocean heat content and thermosteric sea level change (0–2000 m), 1955–2010

Ishii & Kimoto 2009,  Reevaluation of historical ocean heat content variations with time-varying XBT and MBT depth bias corrections

Domingues et al 2008, Improved estimates of upper-ocean warming and multi-decadal sea-level rise

Palmer et al 2007, Isolating the signal of ocean global warming

Smith & Murphy 2007, An objective ocean temperature and salinity analysis using covariances from a global climate model

Poetry please

A post last week at the blog of climate scientist Ellie Highwood revealed remarkable literary talent not just from Ellie herself but also from climate sceptic Mango Chutney. Perhaps there is a hitherto unexploited poetic propensity on both sides of the climate debate.

So please post your poems below, and don’t be put off if you’re unable to reach the quality level of Ellie and Mango.

Extreme events and catastrophes

In recent  years the media has been full of hyped-up stories about increasing extreme events, such as storms, floods and droughts, and keen to attribute them to climate change, along with claims of imminent catastrophe.  To what extent is this supported by the IPCC, and has there been a climb-down in IPCC claims about this?  Roger Pielke Jr suggested (see point 5 of his post) that there has been a ‘walking back’ of claims regarding extreme events, and says ‘kudos to the IPCC’ for this.  In a follow-up post he gives a number of direct quotes from AR5 chapter 2, all of which use phrases like “limited evidence”, “no significant trend”, “low confidence”, but he laments that this is unlikely to slay the storm-caused-by-global-warming zombie.

There are two relevant tables in the AR5 report.  Regarding the chance of some catastrophic event (such as the collapse of the Gulf Stream) there is Table 12.4 in Chapter 12 of the main report.   At  Bishop Hill, Katabasis has already discussed Table 12.4, pointing out that almost all of the catastrophes listed in Table 12.4 are either described as “very unlikely” or assessed with “low confidence”.  Judith Curry has also discussed this, “Did the AR5 take the ‘dangerous’ out of AGW?”

Possible increases in events such as storms and floods are described in Table SPM.1, page 23 in the SPM.   It is interesting to compare Table SPM.1 with the corresponding one from AR4. Fortunately we don’t have to compare two tables side by side because this comparison is included in the table. The text in black is the AR5 assessment, the text below in blue is from the 2012 SREX report, and the last line in red (orange) is from the 2007 AR4 (irritatingly, although Table SPM.1 is on page 23, its caption is on page 4!)

The table shows that for tropical cyclones (which includes hurricanes) there has been a reduction in confidence (for example from ‘likely’ to ‘more likely than not’ in some regions). And similarly for droughts, where the phrase ‘low confidence’ appears three times.  On the other hand, risk of extreme high sea levels in the future seems to have gone up, from ‘likely’ to ‘very likely’. Similarly, likelihood of human contribution to heat waves has gone up.

So overall, is it a climbdown or a ratcheting up of  extreme event risk?

The decadal mean temperature graph

Perhaps the most ridiculous graph in the IPCC AR5  SPM is this one showing “decadal mean” temperature.

A few points about this graph:

• As far as I know, such a graph has not been used in any peer reviewed publication
• The graph was not in the draft version of the SPM subjected to expert review
• No such graph has been used in any of the previous IPCC reports.
• I’m not aware of any such graph being used in any other field of science – any examples?

So why is the graph so bad? It’s hard to see why it is necessary to point this out.  Firstly, it takes a graph with about 160 data points on it, and reduces this to just 16, effectively throwing away most of the data.  Secondly, the appearance of the graph depends very much on how you choose to do the 10-year cuts. They seem to have chosen either 0-9 or 1-10 bins (it’s not clear which) so that the last two or three years aren’t included at all. But if we chose 5-4 bins, the picture would probably look quite different (has anybody done this?).  The introduction of this graph into AR5, with no such graph in the previous reports, leaves the IPCC open to accusations of trying to “hide the decline” in warming this decade,  though of course the levelling off is clear in the graph above, so the graph seems quite pointless.

In the draft version of the SPM reviewed by scientists, this graph was not there, perhaps because the authors were aware that it might be criticised.  This illustrates the point about the authors having carte blanche to insert whatever they like into the final version after the review.   The decadally averaged graph was there in the main section of the report, in chapter  2, Fig 2.20 (In the final draft version, it is Fig 2.19).  In my review comments, I was very critical of this graph (“Fig 2.20 – I am surprised to see this absurd figure still in the SOD. No such figure appears in the cited paper Morice et al, or in any other published paper I am aware of , or in previous IPCC reports. Such a figure would be widely and rightly ridiculed as an attempt disguise the recent slow-down of warming.”)

The IPCC responded to my criticism by putting the graph in the SPM.

I have not seen much comment on this graph.  But Reiner Grundmann tweeted “Summary for policymakers dodges issue of ‘pause’ in global warming. New fig.1 makes problem invisible” and “So SPM replaced the ‘dodgy sandwich’ graph with an ‘elevator’ graph of decadal temp rise. Good PR, but is it sustainable?” on the day the SPM was published.

This type of graph seems to have originated in a Met Office press release from 2009, although the IPCC version seems to be based on this one from 2012 from the EEA.

Update, October 2: Thanks to John Kennedy for pointing out that the decadal mean graph appeared in the 200+ page report, “State of the climate in 2009” (Fig 2.3).