The Nature article that has caused so much angst about the possibility we are entering a decade of cooling — “Advancing decadal-scale climate prediction in the North Atlantic sector” (subs req’d) — has, in fact, been widely misreported. I base this in part on direct communication with the lead author.
In fact, with the general caveat from the authors that the study as a whole should be viewed in a very preliminary fashion, and should not be used for year-by-year predictions, it is more accurate to say the Nature study is consistent with the following statements:
- The “coming decade” (2010 to 2020) is poised to be the warmest on record, globally.
- The coming decade is poised to see faster temperature rise than any decade since the authors’ calculations began in 1960.
- The fast warming would likely begin early in the next decade — similar to the 2007 prediction by the Hadley Center in Science (see “Climate Forecast: Hot — and then Very Hot”).
- The mean North American temperature for the decade from 2005 to 2015 is projected to be slightly warmer than the actual average temperature of the decade from 1993 to 2003.
Before explaining where the confusion came from — mostly a misunderstanding of how the Nature authors use the phrase “next decade” — let’s see how the media covered it:
The UK Telegraph says “Global warming may ‘stop’, scientists predict” — “… Researchers studying long-term changes in sea temperatures said they now expect a ‘lull’ for up to a decade.”
National Geographic News blares, “Cooler Climate May Hit N. America, Europe Next Decade.”
Revkin at the NYT wonders, “Can Climate Campaigns Withstand a Cooling Test?” and says the Nature study forecast “some Northern Hemisphere cooling in the coming decade.”
No surprise, global warming denier Sen. James Inhofe leaped on this with his own press release: “ ‘Global Warming Will Stop,’ New Peer-Reviewed Study Says: Global Warming Takes a Break for Nearly 20 Years?”
But none of these headlines accurately portray what the data presented in the paper actually says. Let’s look at the paper’s key figure, the one that looks at past and (forecast) future global temperatures, “Hindcast/forecast decadal variations in global mean temperature, as compared with observations and standard climate model projections” (click to enlarge)
Let me try to explain this complicated figure (I’ll reprint the authors’ own caption at the very end).
The first thing to know about the figure — indeed, one major source of confusion — is that “each point represents a ten-year centred mean.” That is, each point represents the average temperature of the decade starting 5 years before that point and ending 5 years after that point.
Second, the red line is the actual global temperature data from the UK’s Hadley Center for Climate Prediction and Research. Why does the red line stop in 1998 and not 2007? Again, it is a running 10-year mean, and the authors use data from a Hadley paper that ends around 2003 (I believe), so they can’t do a ten-year centered mean after 1998.
Third, the black line is one of the IPCC scenarios, A1B. It is a relatively high-CO2-growth model — but actual carbon emissions since 2000 have wildly outpaced it (see here).
Fourth, the solid green line is the “hindcast” of the authors — how well their model compares to actual data (and the A1B scenario). It is then extended (in dashes) through 2010 and finally to 2025, where it meets up with A1B, since their model only imposes decadal variability on the inexorable climb of human-caused global warming.
[Fifth, the short purple line is with radiative forcing (i.e greenhouse gas concentrations) frozen at 2000 levels, which, of course, didn’t happen.]
So you can clearly see that the green line rises and then plateaus, repeatedly, until it really starts to take off in the decade of the 2010s. Perhaps the source of much of the media’s confusion is that the authors describe their results in the final line of the abstract this way:
Our results suggest that global surface temperature may not increase over the next decade, as natural climate variations in the North Atlantic and tropical Pacific temporarily offset the projected anthropogenic warming.
But what they mean by that statement is not what a simple reading of that sentence would suggest: They do not mean that “the global surface temperature may not increase over the next ten years starting now.” What they mean is what the lead author, Dr. Noel Keenlyside, wrote me last night when I asked for a clarification:
Thus, based on our results we don’t expect an increase in the mean temperature of the next decade (2005–2015).
They are predicting no increase in average temperature of the “next decade” (2005 to 2015) over the previous decade, which, for them, is 2000 to 2010! And that’s in fact precisely what the figure shows — that the 10-year mean global temperature centered around 2010 is the roughly the same as the mean global temperature centered around 2005.
The authors have not predicted the next 10 years won’t see any warming. They have, however, offered an explanation for why temperatures have not risen very much in recent years, and, perhaps, why ocean temperatures have also not risen very much in the past few years (see here). Dr. Keenlyside continues:
However, as you correctly point out, our results show a pick up in global mean temperature for the following decade (2010–2020). Assuming a smooth transition in temperature, our results would indicate the warming picks up earlier than 2015.
Again, at that point, Dr. Keenlyside reiterates the disclaimer that this analysis can’t be used for year-by year predictions. Indeed, he notes that his main conclusion is not really quantitative, but qualitative:
Given the uncertainties that exist in such kinds of preliminary studies, I believe it is more useful to point out that climate on decadal timescales may be quite different from that expected only considering external radiative forcing (as in the IPCC). This is actually an obvious, but I believe mostly overlooked fact. Our results highlight this.
WHAT THIS STUDY TELLS US
Before coming to a final conclusion, I would add three points. First, as you can clearly see in the figure — the actual observed runnning average temperatures from the Hadley Center since 1995 have been between the IPCC scenario projection and Dr. Keenlyside’s forecast, which does suggest that his model may be underestimating warming. Indeed, the lack of agreement between the model’s “hindcast” and actual temperatures since 1995 should remind us again to view this only as a very preliminary analysis with predictive ability that is much more qualitative than quantitative.
Second, since carbon emissions since 2000 have been racing past all projections, and been accompanied by soaring CO2 concentrations, (see “NOAA: Atmospheric Carbon Dioxide, Methane Rise Sharply in 2007”), we would again naturally expect actual temperatures to be slightly higher than Dr. Keenlyside’s forecast (just as frozen concentrations yield results below his forecast).
Third, this general prediction — internal variability leading to slower than expected warming in recent years through 2010, followed by accelerated warming — is almost exactly the same prediction that the Hadley Center made last summer in Science (see here). They concluded:
… at least half of the years after 2009 predicted to exceed the warmest year currently on record.
… [2014 will] “be 0.30° ± 0.2°C warmer than the observed value for 2004.”
So I take both these admittedly preliminary short-term forecasts to suggest that warming is going to be a roller coaster ride, with much short-term variation, but we are probably going to get quite hot quite fast early in the 2010s.
One final caveat: After reading the first draft of this post (which I have since revised), Dr. Keenlyside writes me this morning, “All our figures are decadal means, and it is hard to say (due to high frequency internal variability) at which point [after 2010] a rapid increase will occur.” That is, his study does not necessarily predict the rapid warming will actually start, in say, 2011, though his results are not inconsistent with that possibility. He reiterates that his paper is not designed to make such detailed year-by-year predictions. Indeed, the paper was designed to show that any such predictions are complicated by decadal-scale climate factors.
Here is the authors’ own caption for the figure:
Model projections are twentieth century-RF [radiative forcing] followed by A1B scenario simulations (‘20C-RF/A1B’); ‘Stabilization’ forecasts assume greenhouse gas concentrations fixed at year 2000 levels. Each point represents a ten-year centred mean; vertical bars indicate ensemble spread; verification and forecast periods are indicated (dark shading begins 2008, indicating the start of the true forecast period). Three additional decadal means (joined by a dotted line) show the evolution of the initialized and un-initialized 2005 predictions extended till 2030. Correlation of both hindcasts and climate model projections with observations are given in brackets. Correlation of the twentieth century-RF simulation with observations is greater than that of the hindcasts, but only marginally at the 5% significance level. Observed global mean temperature anomalies are from HadCRU3