There has been considerable recent discussion of the fact that observations have been running cooler than models – see, for example, Lucia’s discussion of IPCC AR5 SOD Figure 9.8 (see here). However, Michael Mann at AGU took an entirely different line. Mann asserted that observations were running as hot or hotter than models. Mann’s assertion was taken even further by Naomi Oreskes, who asserted that climate models were under-estimating relative to observations. Oreskes squarely placed the blame for the supposed underestimates on climate skeptics.
In today’s post, I’ll look closely at the illustration in Mann’s AGU presentation, an illustration that gave an entirely different impression than the figure in the IPCC draft report. The reason for the difference can be traced to what I’ve termed here as “Mike’s AGU Trick”.
The IPCC AR5 SOD Graphic
An excerpt from IPCC AR5 SOD Figure 9.8 is shown below, clearly showing that the multimodel ensemble (red) is running noticeably hotter than observations (black). In my opinion, the difference is not merely “noticeable” but “statistically significant”, but that’s a story for a different day.
Figure 1. Excerpt from IPCC AR5 SOD Figure 9.8, comparing model ensemble to observations.
Mann’s AGU Presentation
However, Mann at AGU asserted that observations were running as hot or hotter than models. Mann’s model comparandum was Hansen’s Scenario B, which is widely regarded as the most reasonable scenario to use to interpret Hansen’s “forecast” – see past CA points on this issue.
I took the photo with a new phone, with which I was then unfamiliar and unfortunately can only provide a muddy zoom on the graphic. Despite the muddiness, you can see that observations (red) appear to cohere with Hansen’s 1988 forecast (blue). In the loop below, I’ve overplotted data for models and observations to show more clearly what was shown to the AGU audience. (There was a bit of detective work in figuring this out – see below.) Click on the figure below for a loop illustrating the components of the zoomed figure). (Note: see below for Mann’s use of his AGU Trick to hide the divergence in a presentation a few months earlier at Rutgers).
Mann’s AGU slide obviously has a completely different rhetorical impression than the IPCC graphic. Whereas the discrepancy between observations and models was immediately noticeable in the IPCC graphic, Mann’s AGU graphic showed no such discrepancy. There were two reasons for the difference, the combination of which I’ll call “Mike’s AGU Trick” and will analyse below.
Hansen 1988 versus Observations
First, I’ll show that Mike’s AGU Trick does not result from using Hansen’s Scenario B, as opposed to the IPCC multimodel ensemble. The next graphic compares Hansen’s original graphic (with Scenario B highlighted in blue) against observed GISS global land-and-ocean temperature (red). During the past decade, as with the IPCC multimodel ensemble, a noticeable divergence between observations and Scenario B has developed, reaching over 0.5 degrees C in 2012.
Figure 4. Annotated version of Hansen 1988 showing Scenario B (blue) and observed GISS land-and-ocean (red.) All data centered on 1958-1977. In the preparation of this graphic, I noticed something interesting about the centering of Hansen’s scenarios – an issue that occasioned considerable controversy in blog commentary: Hansen’s Scenarios are almost certainly centered on 1958-1977. The averages of both Scenario B and Scenario C are 0 to five significant digits with this centering. All data was accordingly centered on 1958-1977 for the present comparison. For GISS observations, the difference between 1958-1977 and 1951-1980 centering is -.0175 deg C – the difference is not material to comparisons, but there’s no reason not to do it as precisely as possible.
Mann and Kump, 2008
Analysis of Mann’s AGU graphic is fortuitously assisted by the use of a similar graphic in Mann and Kump, Dire Predictions, as illustrated in the loop below. Mann and Kump, published in July 2008, compared Hansen’s Scenario B to observations to 2005, though data was then available up to 2007. In addition, although GISS model results are obviously for Land-and-Ocean, Mann and Kump used Land-Only data (which runs hotter) [click on figure for loop]
Figure 5. Comparison of Hansen 1988 to observations in Mann and Kump, Dire Predictions – see https://www.e-education.psu.edu/meteo469/node/141
Mike’s AGU Trick
There were two components to Mann’s AGU trick. First, as in Mann and Kump, Mann compared model projections for land-and-ocean to observations for land-only. In addition, like Santer et al 2008, Mann failed to incorporate up-to-date data for his comparison. The staleness of Mann’s temperature data in his AGU presentation was really quite remarkable: the temperature data in Mann’s presentation (December 2012) ended in 2005! Obviously, in the past (notably MBH98 and MBH99), Mann used the most recent (even monthly data) when it was to his advantage. So the failure to use up-to-date data in his AGU presentation is really quite conspicuous.
Had Mann shown a comparison of Hansen’s Scenario B to up-to-date Land-and-Ocean observational data, the discrepancy would have been evident to the AGU audience, as shown in the loop below.
Pierrehumbert Condemns Data Truncation
In a 2007 realclimate article, Raymond Pierrehumbert condemned use of non-updated temperature data, when the effect of the failure to use up-to-date data was that the image gave an entirely different impression to the reader. In that situation, Pierrehumbert even called into question the ethics of the author.
there is no legitimate reason — in a paper published in 2007 — for truncating the temperature record at 1992 as they did. There is, however, a very good illegitimate reason, in that truncating the curve in this way helps to conceal the strength of the trend from the reader, and shortens the period in which the most glaring mismatch … occurs.
There does not appear to be any relevant difference between Mann’s AGU technique and the graphic so vehemently criticized by Pierrehumbert, though, to my knowledge, Pierrehumbert has not taken exception to Mann’s AGU Trick.
Mann’s AGU Trick appears to have wrongfooted his mini wingman, Naomi Oreskes.
Like Mann, Oreskes gave three AGU presentations (sessions PA13B – Countering Denial and Manufactured Doubt of 21st Century Science I; GC22B – Communicating Climate Science—Seeking the Best of Old and New Paradigms I; GC33F – Construing Uncertainty in Climate Science I). Oreskes’ starting point was that models had supposedly under-estimated relative to observations – a starting-point that seems oddly disconnected to the IPCC graphic shown above but, hey, Oreskes is an expert in manufactured disinformation. If Oreskes was not in fact wrongfooted by Mann, then one would like to know the provenance of her assertion that models were “underestimating” observed temperature increases.
Oreskes then purported to ponder on the institutional factors that supposedly caused such under-estimates by climate scientists. Oreskes had no doubt as to where the “blame” lay. Not with the scientists themselves. of course not. Oreskes placed the blame squarely on climate skeptics. According to Oreskes, their intimidation had led climate scientists to pull their punches and make forecasts that were too conservative.
[Note (March 2, 2013 4:22 pm.] While I was mostly offline, it turns out, as a reader below points out and as Pielke Jr has communicated to me by email, that Oreskes has published the argument outlined in her AGU presentation here. Pielke Jr has an excellent review of the Oreskes article here. Pielke Jr quoted the following from the Oreskes article – a quote which is very much consistent with the AGU presentation:
[O]ne possible reason why scientists may have underestimated the threat of anthropogenic warming is the fear that if they don’t, they will be accused by contrarians (as was Schneider) of being alarmist fear-mongers. That is to say, pressure from skeptics and contrarians and the risk of being accused of alarmism may have caused scientists to understate their results.
In other words – and this is the implication of Oreskes’ presentation not what she told the audience – were it not for the moderating influence of climate skeptics, the discrepancy between observations and the models that climate scientists would have presented in an Oreskes world, would have been even larger than the present discrepancy.
Surely this is an outcome that should cause Oreskes to thank skeptics, rather than condemn them. But, needless to say, no such thanks were forthcoming at the AGU of Oreskes, Gleick, Lewandowsky and Mann.
Update: As reader DGH observed in a comment below, Mann’s presentation at Rutgers also employed Mann’s AGU Trick to hide the divergence between Hansen Scenario B and observed temperature, not showing data after 2005. As noted above, not using up-to-date data in virtually identical circumstances was characterized by Pierrehumbert as “ugly” and “illegitimate”:
As reader ZT pointed out, Mann also used his AGU Trick to hide the divergence in his TEDx talk
As reader ^ pointed out, in 2006, Hansen did his own comparison of his 1988 scenarios to observations, a diagram that appears to be the iconographic predecessor of Mann’s AGU diagram, a diagram that clearly shows an important difference between Hansen’s ethics and Mann’s ethics. Hansen’s article included up-to-date temperature data. In contrast, Mann’s AGU diagram, over 6 years later, still used 2005 data. In addition, Hansen showed both Land-Only and Land-and-Ocean data sets, showing the greater divergence between models and Land-and-Ocean data (arguing, not entirely convincingly, for an intermediate.) In contrast, Mann only showed the most advantageous version.