The article itself presents a Holocene temperature reconstruction that is very much at odds both with Marcott et al 2013 and Mann et al 2008. And, only a few weeks after IPCC expressed great confidence in the non-worldwideness of the Medieval Warm Period, Rosenthal et al 2013 argued that the Little Ice Age, Medieval Warm Period and Holocene Optimum were all global events.
Although (or perhaps because) the article apparently contradicts heroes of the revolution, Rosenthal et al 2013 included a single sentence of genuflection to CAGW:
The modern rate of Pacific OHC change is, however, the highest in the past 10,000 years (Fig. 4 and table S3).
In the Columbia and Rutgers press releases accompanying the article, this claim was ratcheted up into the much more grandiose assertion that modern warming is “15 times faster” than in previous warming cycles over the past 10,000 years (though the term “15 times faster” is not actually made in the peer reviewed article):
In a reconstruction of Pacific Ocean temperatures in the last 10,000 years, researchers have found that its middle depths have warmed 15 times faster in the last 60 years than they did during apparent natural warming cycles in the previous 10,000.
Rather than quoting the article itself, Michael Mann, an academic activist at Penn State University, repeated the claim from the press release in an article at Huffington Post entitled “Pacific Ocean Warming at Fastest Rate in 10,000 Years”.
However, both the claim in the press release and the somewhat weaker claim in the article appear to be unsupported by the actual data.
In the past, we’ve often observed that great care needs to be taken in the interpretation of spaghetti graphs containing both instrumental and proxy data. In the present case, none of the graphics in Rosenthal et al 2013 contain both instrumental and proxy data, a deficiency that I’ve attempted to remedy in today’s post.
First, here is Rosenthal’s figure (3B) showing their reconstruction of IWT (Intermediate Water Temperature) for the past two millennia, which shows a sharp decline in temperature from the medieval period to the Little Ice Age (which, as at Baffin Island also in the news lately, was the coldest period in the past 10,000 years), with a recovery in the 20th century, but to levels lower than those of the medieval and earlier periods.
On the far right, I’ve plotted Pacific ocean heat content, converted to deg C anomaly (red), together with its trend line. The two solid yellow lines show trend lines for 1100-1700 AD and 1600-1950 AD, two of the three periods considered in Rosenthal Table S4. It is true that the rate of change over the past 55 years is somewhat higher than the trend over 1600-1950, but it is not “15 times higher”. While I don’t think that one can safely reify the fluctuations in Rosenthal’s IWT reconstructions, on the other hand, these fluctuations appear to me to preclude any strong conclusions that the relatively modest increase is unprecedented.
Figure 1. Annotation of Rosenthal Figure 3B. Original caption: “Compiled IWT anomalies based on Indonesian records spanning the ~500- to 900-m water depth (for individual records, see fig. S7). The shaded band represents +-1 SD. Red- OHC Pacific 0-700m heat content converted to temperature using the 0-700m Pacific mass shown in the Rosenthal SI. The values are consistent with 0-700m temperature anomaly values at NOAA http://www.nodc.noaa.gov/OC5/3M_HEAT_CONTENT/index3.html.
In my annotation of their Figure 3B shown above, I’ve shown two trend lines, each of which more or less corresponds to the trends reported on lines 2 and 3 of Table S3: a trend of -0.15 deg C/century from 1100-1700 and a trend of 0.09 deg C/century from 1600-1950.
Rosenthal’s Holocene Figure
Rosenthal’s Figure 2 shows their temperature reconstruction over the past 10000 years (the version in their Figure 2C is shown below.) I’ve overlaid my digitization of their two-millennium reconstruction shown above (cyan). (Unfortunately, Rosenthal and coauthors didn’t archive anything: neither reconstruction nor underlying data, though, in response to my email, Rosenthal has undertaken to do so.) Once again, I’ve shown the temperature anomaly (as I’ve calculated it from instrumental ocean heat content) in red. At a Holocene scale, it is a very small squiggle – which I’ve accordingly highlighted by a circle around the squiggle). On the right is a blow-up of the graph in the left panel for the past 500 years.
Figure 2. From Rosenthal et al 2013 Figure 2C. Red- temperature anomaly converted from NOAA Pacific Ocean 0-700m ocean heat content. Cyan – Rosenthal Figure 3B reconstruction (my digitization). Orange trend line shows third comparison from Rosenthal SI, taken from first row in Table S3.
As with the two-millennium figure, there are many fluctuations within the reconstruction with much greater rates of change over a century than calculated from the modern instrumental OHC record.
Although spin by Mann and others has focused on the modern portion, Rosenthal and coauthors observe that their reconstruction differs on important points from that of Marcott et al 2013,
To the extent that our reconstruction reflects high-latitude climate conditions in both hemispheres, it differs considerably from the recent surface compilations, which suggest ~2°C MWP to- LIA cooling in the 30°N to 90°N zone, whereas the 30°S to 90°S zone warmed by ~0.6°C during the same interval (24- Marcott et al 2013). In contrast, our composite IWT records of water masses linked to NH and SH water masses imply similar patterns of MWP-to- LIA cooling at the source regions.
They showed the Marcott et al 2013 reconstruction in their Figure 2, apparently unaware that its 20th century uptick is (at best) an artifact. It is regrettable that Marcott and coauthors have not issued a corrigendum conceding that the uptick cannot be relied on.
Although IPCC recently reported much confidence in the non-globalness of the MWP, Rosenthal et al took the opposite view:
The inferred similarity in temperature anomalies at both hemispheres is consistent with recent evidence from Antarctica (30), thereby supporting the idea that the HTM (Holocene Thermal Maximum), MWP, and LIA were global events.
Given all the publicity about the supposed unprecedentedness of the recent increase in Rosenthal et al 2013, the data in the article offers little support for the assertion that the modern rate of increase is 15 times greater than any previous increase – or, indeed, for the weaker proposition that the modern increase is unprecedented. Moreover, it stands against claims that modern temperatures are themselves unprecedented, not only within the Holocene, but within the last two millennia.
Rosenthal’s Boxplot and Table S3
Rosenthal’s assertion that recent OHC trends are unprecedented are derived from their Figure 4B, shown below. This figure is shown as a box plot – a standard technique for showing distributions. However, it is, so to speak, a fake box plot, as I’ll discuss below.
First and most obviously, the length of the periods being compared are totally different. The modern instrumental period is 55 years, while the first boxplot is a decline over 5500 years, within which proxy data has substantial fluctuations.
The figure “15 times” appears to be related somehow to the comparison between the instrumnental increase (fourth box) and the rate of increase in the post-LIA reconstruction (the other two trends being negative.) The instrumental box median OHC rate of change is 15E22 joules/century (compare to Table S3 fifth column) while the LIA recovery box (third box) “median” rate of change is 1.5E22 joules/century (also compare to Table S3). (I’ll discuss these questionable calculations below.) Their ratio is 10, not 15. Perhaps Phil Jones did the calculation for them. Or maybe something else.
Rosenthal et al 2013 Figure 4B. Reconstructed rates of OHC change during the main transition periods. Reconstructed anomalies and rates are compared with modern observations for the 2000 to 2010 and 1955 to 2010 CE periods, respectively (5). The middle line at each box represents an average estimate for 50% of the Pacific volume between 0 and 700 m, whereas the top and bottom quartiles of the box represent 62.5 and 37.5% of the total volume in this depth interval, respectively. The bottom whiskers represent 25% of the volume; the top whisker denotes 75%. The modern value is based on the entire Pacific volume for 0 to 700 m.
Table S4 is shown below. Its 2nd column shows temperature deltas. The 1600-1950 change (third row) is shown as 0.25 deg C, while the 1955-2010 instrumental (from OHC) is shown as 0.11 deg C. The next column purports to convert temperature change into ocean heat content change.
In the reconstructions, the temperature change is converted to Pacific Ocean Heat Content change by multiplying the estimated temperature change by 50% of the mass of the Pacific Ocean (they use a mass of Mz=1.12E20 Kg and a specific heat of 4 J/deg C-g.) Watch this carefully: this calculation assumes a temperature change of zero on the rest of the ocean – an assumption that radically contradicts their observations that they are observing global phenomena. And while one can understand the reluctance to extrapolate to the entire ocean, this inability means that the relevant instrumental comparison ought to be to half the Pacific Ocean as well. Watching this pea also shows the absurdity of the “quartiles” in the boxplot. These “quartiles” are nothing more than applying zero to 37.5% and 62.5% of the ocean respectively.
While this basic assumption of Table S3 and Figure 4 makes no sense, there are other problems as well. Look closely at Table S3. The delta-temperature is less than half (0.11 versus 0.25), but the delta OHC is 50% greater (8.4 versus 5.6). One can “get” the delta-OHC for the reconstruction periods by multiplying the deg C by the Pacific Ocean mass (Mz=1.12E20 Kg and a specific heat of 4 J/deg C-g) and multiplying by 50%. If the same procedure were applied to the instrumental period, one would get a delta-OHC of 2.6E22 joules (as compared to 8.4). Applied to the rate of change, the implication is that the rate of change in the instrumental period would be 2.8 times the average rate of change in the post-LIA period (1600-1950), rather than 15 times. Even if one accepted the Rosenthal et al assumption that one could apply zero to half the reconstruction ocean, there seems to be something wrong with the arithmetic to 8.4E22 joules. There also seems to be an error with the value of 0.032 deg C/century (fifth column fourth row) as this is inconsistent with other columns.
While the calculations in this table seem peculiar – ranging from apparent arithmetic errors to questionable extrapolation from temperature change to ocean heat content, one should not presume that a corrected version of this table necessarily makes sense: the entire enterprise of attempting to compare changes in ocean heat content based on proxy data on Intermediate Water seems both ill-conceived and forced.
Update Mar 2, 2014. Last year, I requested the underlying data for this article from Rosenthal and from Sciencemag. Rosenthal promised on several occasions to provide the data, but repeatedly failed to live up to his promise. In January 2014, Rosenthal archived the data in a revised SI at Sciencemag – without the courtesy of notify me that he had done so.
In my blog article on Rosenthal et al 2013, I had pointed out that Table S3 appeared to contain numerous errors.
The revised SI states that Table S3 contains errors. Rosenthal didn’t acknowledge or thank me for pointing out these errors.
The re-stated table S3 is shown below. The largest change is in their delta-T per century, where the reported value has been increased from 0.032 to 0.24. This was presumably some sort of mechanical error in the previous article. Other changes include shortening the modern Levitus period from 55 years to 45 years – I’m not sure why this was done. Changing the delta-T in the holocene (first line) from 8.4 to 6 – I still don’t see how they get this number. As before, their comparison of historical to modern depends on the assumption that the heat content in one half of the historical ocean changes without change in the other half: this doubles the modern comparison, thereby exaggerating the effect.