Anomalous Warmth? Context for Comments on, Critiques of, Study in Science

A March report published in Science magazine prompts widespread coverage and substantial online back-and-forths. But what’s it all mean for our understanding of past and future global temperatures?

An early March publication on past temperatures by Oregon State University earth scientist Shaun Marcott and colleagues in Science has generated significant media attention and lots of online commentary, criticism, and praise.

The study, published in the March 8, 2013, issue of Science, is entitled “A Reconstruction of Regional and Global Temperature for the Past 11,300 Years.”

“Recent heat spike unlike anything in 11,000 years” wrote Seth Borenstein of the Associated Press. The New Scientist reported that “Earth’s temperature is changing faster now than at any time since the last ice age.”

Analysis

A month later, after much online comment and criticism, a somewhat more nuanced picture has emerged: The rapid spike in temperatures in recent times in the figures by Marcott et al is, by their admission, “not statistically robust, cannot be considered representative of global temperature changes, and therefore is not the basis of any of our conclusions.”

This qualification does not mean that temperatures haven’t risen dramatically over the past century: There is ironclad evidence in the form of instrumental temperature records that they have. Rather, the estimate of past temperatures using the researchers’ proxy measurements cannot show variations in such a short timeframe. As a result, while unlikely, it is not possible, based on the results of the Marcott research, to rule out similar variations in the past. That said, the paper provides a novel and interesting look at temperatures over the past 11,000 years, and it makes a compelling case that continuing releases of CO2 and other greenhouse gases will push Earth’s global temperature well above what has been observed since the end of the last ice age.


View larger image

Marcott et al provided an estimate of temperature variations over the Holocene, a period from the end of the last ice age (around 11,000 years ago) to present. Their research uses a set of 73 climate proxies, including marine fossils, pollen records, and ice cores. These temperature proxies provide information on the temperature at the time they were created, but the data they provide is often imprecise and an imperfect measure of temperature given potential interference by other factors. In addition, they do not provide precise dates. For most of the record produced by Marcott and his colleagues, shown in the figure above, the proxy information allows only an estimate of the average temperature over a period of 400 years or so. It therefore fails to capture variability over shorter periods of time (though a rise equivalent to that over the past century could still show up in the record so long as it was effectively recorded by the proxies and not canceled out by an equal amount of cooling in the subsequent century).

The major problem identified in the Marcott paper involves the dramatic spike at the present day end of the record. It turns out that the presence (or absence) of the spike is strongly dependent on the methods used. Indeed, it may simply be an artifact of colder proxies tending not to extend up to the present day, with their drop-out from the record artificially warming the results. The application of another method by Marcott, called RegEM, in fact avoids this issue by infilling missing data points, and it does not show any equivalent large uptick in recent years. The chart above shows the Standard method, the RegEM method, and the instrumental record (via the National Climatic Data Center) over the past 2,000 years. There has also been some criticism of the date choices for recent proxy measurements, but this makes little difference when using a RegEM approach.

The only real difference in results between the Marcott standard and RegEM approaches occurs at the present day; the overall Holocene reconstructions are effectively identical. However, without the eye-catching spike at the end, the results appear much less dramatic, raising the question of how best to compare modern temperatures to proxy reconstructions. The original spike in the Marcott paper was, as they acknowledge, “not robust”. Similarly, simply grafting on the instrumental temperature record to the end of the proxy reconstruction risks comparing apples and oranges, as the proxy record reflects a long-term smoothed (e.g. low-resolution) estimate.


View larger image

To get around this issue, Marcott and his co-authors took a rather interesting approach in their paper. They created an estimate of the amount of decadal and century-scale variability in global temperatures using both instrumental records and existing higher-frequency proxy reconstructions (particularly that of Mann et al 2008). They then added this variability to their low-frequency Holocene reconstruction to get a better sense of how much multi-decadal variability might actually occur. While this approach is admittedly imperfect (it assumes that higher-frequency proxy reconstructions present an accurate picture of actual multi-decadal variability, and will not include any events distinct from those that were captured in proxies over the past two millennia), it provides a much better comparison for recent temperatures than the original reconstruction.


View larger image

Marcott et al presented the results of this analysis in Figure 3 of their paper. It’s probably the most important, and most interesting, of their three figures, but it has received far less media and public attention, most likely because of its complexity.

The figure, above, shows a frequency distribution — effectively a histogram — of estimated temperatures over the past 11,000 years and includes a number of different variations on methods used. The most important of these distributions is that shown in solid black. It represents their basic method with the inclusion of high frequency noise (Standard 5×5 + high frequency). The authors compare this to current temperatures (from 2000 to 2009) and to those at the beginning of the century. Interestingly, they find that “global mean temperature for the decade 2000–2009 has not yet exceeded the warmest temperatures of the early Holocene (5,000 to 10,000 yr B.P.),” though they note that they are warmer than 72 percent of the record.

On the other hand, temperatures at the beginning of the century were some of the coldest in the Holocene, colder than 95 percent of the record. They note that this dramatic change, from some of the coldest temperatures to temperatures decidedly on the warm side, was dramatic, though they do not perform any specific tests to determine how unlikely such rapid changes would be in the past. Marcott et al also point out that under any of the IPCC scenarios for future warming over the next century, temperatures will be well outside anything seen in the Holocene, as shown by the grey bands representing different future warming scenarios.

Raymond Pierrehumbert from the University of Chicago succinctly sums up the usefulness and limitations of Marcott’s paper:

I think this paper represents a very nice advance in the characterization of Holocene climate; ever since the various papers on cooling trends of the past 2,000 years pre-industrial started coming out, I had been yearning for something that covered back to the Altithermal [early Holocene]. This kind of paper will be very valuable in refining our understanding of how the climate responds to the precessional cycle, particularly once it becomes clearer whether the reconstructed temperatures are really annual averages, as opposed to biased toward summer …

It is conceivable that there are individual centuries in the Altithermal where the temperature rose as fast as today, and to the same extent or more, but these would not show up in a record smoothed to 100 year time resolution. I think this is very unlikely, but the paper doesn’t strictly rule out the possibility. This remark applies only to the warming of the past 100 years. Where we are going in the next century is so extreme it would show up even if smoothed down to the centennial resolution, I think.

While most admirers and critics of the Marcott paper have focused on the hockey-stick style graph, where the spike on the end was not robust, the real advances in Marcott paper lie in improving understanding of temperatures over the Holocene, and examining how future warming will compare to past temperatures.

Unless climate models are well off the mark and climate sensitivity is dramatically lower than now understood, it seems very likely that Earth’s warming will exceed any average temperatures experienced in the planet’s recent history.

We are on the early portions of the blade of a hockey stick, and how high that blade will go depends largely on steps taken, or not taken, by us humans.

Zeke Hausfather

Zeke Hausfather, a data scientist with extensive experience with clean technology interests in Silicon Valley, is currently a Senior Researcher with Berkeley Earth. He is a regular contributor to The Yale Forum (E-mail: zeke@yaleclimatemediaforum.org, Twitter: @hausfath).
Bookmark the permalink.

Comments are closed.