November 17, 2010

Kracked Up Over Krakatoa: Models Have It All Wrong

Filed under: Climate Models

It was all the rage a few years back to claim that long ago volcanic eruptions—for instance Krakatoa in 1883—were still acting to mask a large fraction of the oceanic warming that should have occurred because of anthropogenic carbon dioxide emissions. The epitome of this argument was published in Nature magazine, by an all-star cast of scientists ever-eager to suggest that it is all our fault and then some. The authors included Tom Wigley, Ben Santer, Karl Taylor, Krishna AchutaRao, Jonathan Gregory, and lead author Peter Gleckler.

The accompanying Editors’ Summary of the 2006 Nature article by Gleckler et al. provides the gist:

The 1883 eruption of the volcano Krakatoa in Indonesia has echoed down the centuries in art and in legend. Now an analysis of a suite of 12 climate models shows that Krakatoa also made its presence felt well into the twentieth century in the form of reduced ocean warming and sea-level rise. The changes lasted much longer than was previously suspected and were sufficient to offset much of the ocean warming and sea-level rise caused by more recent human activities.

The IPCC incorporated this finding into their Fourth Assessment Report (AR4) to show that models better match the observed history of the increase in oceanic heat content and sea level rise from thermal expansion when modern (since 1880) volcanic eruptions were included along with anthropogenic forcings. The implication was two-fold; 1) the climate models were now able to closely match reality (so they should be considered reliable), and 2) the cooling from volcanoes was offsetting a large fraction of the influence of anthropogenic global warming (i.e. our influence was even worse than we thought).

Now, a new study comes along, performed by one of the et al.’s of the Gleckler study, that basically shows that the conclusions of that original paper were quite likely incorrect, because the climate models examined had been equilibrated to an improper set of “background” conditions—conditions unnaturally free of any and all volcanic eruptions.

Had the climate models been equilibrated to more realistic conditions—after all, big climatologically important volcanic eruption are a fairly common part of the earth’s natural environment and not just a phenomenon of the pat 120 years—Krakatoa and subsequent volcanoes would not have induced a large, long-term warming-offsetting cooling tendency. And in that case, the apparent match between climate models and reality would fall apart as models would be warming their oceans far more rapidly than observations show the real oceans are (i.e., the models don’t work so well after all).

Jonathan Gregory writing in Geophysical Research Letters put it this way:

This artefact [the model response to improper equilibration] could be misleading in comparisons and attribution of observed and simulated changes in ocean heat content.

No kidding.

Let’s review what Gleckler, Santer, Wigley, et al. claimed in Nature in 2006.

In model simulations, Krakatoa has long-lasting effects, offsetting a large fraction of the changes in ocean heat content and thermal expansion caused by twentieth-century anthropogenic influences. These results are robust to current uncertainties in climate models and in the historical forcings applied to them.

Just how large a “fraction of the changes in ocean warming and sea-level rise caused by twentieth-century anthropogenic influences” was offset by the string of volcanic eruptions during the past 130 years is illustrated in the Figure below taken from the Gleckler et al. paper. The top panel of Figure 1 shows how much models expected the oceanic heat content to have increased considering observed changes in greenhouse gases, aerosols, and solar output (green shading), and the lower panel depicts how much ocean heat content was projected to rise when the additional influence of volcanic eruptions since 1880 was included in the model runs (blue shaded area). This history of the eruptions is shown (in purple) along the top of the figure.

Figure 1. Model evolution of the oceanic heat content since 1880 in runs which include the major natural and anthropogenic forcings except volcanic eruptions (green shading) and those which include volcanic eruptions (blue shading) (from Gleckler et al., 2006).

The difference between the evolution of ocean heat content between the volcanic and non-volcanic model runs is very large. Gleckler et al. described it like this

“[T]he distinction between the simulations with and without [volcanoes] in Fig. 1 is striking.”

Just how “striking”? Well, for one thing, oceanic heat content ties directly into the volume of the oceans through the thermal expansion of water. Warmer oceans generally take up more volume which lead to higher sea levels. Here is how Gleckler et al. put it:

Global mean thermal expansion is highly correlated with changes in heat content, and so comparisons of thermal expansion between the [volcanic] and non-[volcanic] simulations look much like Fig. 1a. Increases in thermal expansion at the end of the twentieth century (relative to 1882, the year before Krakatoa) are appreciably less for simulations with [volcanoes] (average: 1.7 cm; with 1 s.d.: 1.8 cm) than for the simulations without [volcanoes] (average: 6.3 cm; with 1 s.d.: 2.2 cm).

In other words, Gleckler et al. found that Krakatoa and subsequent volcanic eruptions reduced the rise in global sea levels caused by thermal expansion by nearly 75%.

Another impact of cooler oceans is that they help slow the rate of atmospheric warming. So if volcanic eruptions slow the projected rate of ocean warming from increasing greenhouse gases down somewhat, then the rate of warming in the overlying surface air will also be slowed somewhat. The Gleckler team found that, according to climate models, the series of volcanic eruptions since Krakatoa reduced the heating of the oceans by 75% from where it would have otherwise been from rising greenhouse gases. The authors don’t report how much this would have slowed the rise in global air temperatures, but no doubt it would have had a sizable impact.

Gleckler et al. ultimately concluded:

Inclusion of volcanic forcing from Krakatoa (and, by implication, from even earlier eruptions) is important for a reliable simulation of historical increases in ocean heat content and sea-level change due to thermal expansion.

Well, now it appears that parenthetical phrase “(and, by implication, from even earlier eruptions)” has come back to bite them in the posterior. In a major way.

See, the models used by Gleckler et al. to arrive at their conclusion, in fact, did not include eruptions earlier than Krakatoa.

In a soon-to-be-published paper in Geophysical Research Letters (not quite as glamorous an address as Nature magazine), Jonathan Gregory reports that had “even earlier eruptions” been included, virtually all of the findings from the Nature paper would have been different, and the major conclusions would not have held.

Gregory was concerned that in most climate models which introduced volcanic eruptions in the late 1880s, the response time was so long that each new eruption seemed to add to the cooling effect of the previous eruptions, somewhat ad infinitum. Gregory reasoned that this could not go on forever. In fact, he pondered about this in a paper he published in the Journal of Climate back in 2006:

The time series of [climate model simulated sea level response to episodic volcanic eruptions] suggest the idea that sea level would continue to fall in steps indefinitely when the system is subjected to episodic negative forcing. This could not really be the case, of course. If the sequence of volcanoes is continued, a statistically steady state must be reached eventually.

Good point. And you would have thought that this would have been somewhat obvious to Gleckler’s team as well (which included Gregory).

If the impact of Krakatoa persisted in the models until the end of the 20th century, more than 100 years after the eruption, then didn’t Gleckler et al. think that the climate at the time of the Krakatoa eruption in 1883 was still being impacted by the monstrous eruption of Tambora in 1815? And what about the one before that? And before that? Figure 2 shows the history of climatologically-significant volcanic eruptions for the past 1000 years. Notice that big volcanic eruptions happen fairly frequently.

Figure 2. Climatologically important volcanic eruptions (top) of the past 1,000 years (source: IPCC AR4).

Gregory, at least, was thinking about this. Again from his 2006 paper:

Given the long time scale of recovery from volcanic eruptions [as indicated by climate models], it seems likely that the signal of Tambora must have persisted beyond the end of the nineteenth century.

Gregory finally pinpointed why the models were behaving they way they were—the models are being equilibrated to the wrong set of natural conditions.

Before climate models runs are subjected to anthropogenic forcing factors, like increasing greenhouse gas concentrations and/or changing aerosol levels, they are first allowed to reach equilibrium to the initial background (i.e. “natural”) conditions. Once equilibrium is established and the models seem to be able simulate a somewhat stable climate (although, to this day, the climate in most models exhibits a slow, long-term drift and this drift must be subtracted out before anthropogenic trends can be quantified), then anthropogenic effects (or any other changes) are applied. But, for virtually all climate models, the background “natural” conditions do not include volcanic eruptions.

The net result of this is that when the first volcano erupts into this virgin, volcano-free environment, it has a huge and long-lasting impact as the modeled climate has to incorporate this new perturbation throughout the entire system. The impact of subsequent volcanoes are less than the first, but still produce nearly additive results as the modeled climate is still adjusting to this new form of (negative) climate forcing.

In his new paper, Gregory details the above-described model behavior, and goes on to suggest that if climate models were to be equilibrated to conditions which include periodically occurring volcanic eruptions, the impact of new eruptions is much shorter-lived and doesn’t lead to a long-term trend in oceanic heat content (or, thus, sea level rise).

The general idea is illustrated in Figure 3.

Figure 3. Climate model evolution of oceanic temperature changes from various methods of including volcanic eruptions. The black and pink lines show the impact of successive eruptions, the first occurring in a volcano-free environment. The blue line shows the impacts of a constantly applied negative forcing equal to the time average of the forcing from volcanoes. The red line is the temperature anomaly (from the blue line) induced by the volcanic eruptions. (source: Gregory, 2010).

The black and pink lines are the modeled impact on oceanic heat content of a succession of volcanic eruptions in a climate that has been volcano-free. Notice the continued downward step with each new eruption. The blue line is the modeled climate response to the average forcing from volcanoes over the study period. It is smoother than the black and pink lines because there are no episodic events, but it is downward at the same overall slope as the black and pink lines, indicating that the climate responds to an average initially-applied forcing over the long-term the same way that it does to the combination of episodic forcings. Finally, the red line—which Gregory proposes best captures the impact of volcanic eruptions—is the difference between the pink line and the blue line. This is basically an anomaly response to long-term conditions. The influence of individual eruptions is felt, and the timing of the eruptions can, and does, lead to trends lasting from years to decades. But in the long-term (like on the century time scale) volcanoes don’t induce any trend in ocean heat content, sea level, or, for that matter global surface air temperatures.

The story told by the red line in Figure 3 is vastly different from the ones told by the black and pink lines. And so too is the story told by Gregory from the one originally told by Gleckler et al. in Nature magazine back in 2006.

According to Gleckler et al. (and incorporated by the IPCC) models with a large overall cooling influence from volcanoes since 1880 matched observations better than models without such a cooling influence. The non-volcanic models warm the oceans and raise sea levels too quickly in response to anthropogenic CO2 emissions. Volcanoes were embraced as a good excuse for the reasons why.

Now, it turns out, that that excuse no longer holds.

This is major blow to the climate models, for it means that they are improperly handling the exchange of heat through the earth’s atmosphere/ocean system, with the ultimate result being that they are over-responsive to rising levels of greenhouse gases.

Getting these things fixed will mostly likely lead to much more modest projections of future temperature rise and accompanying impacts.

It seems as if it is time for the modelers to head back to the drawing board.


Gleckler, P. J., T. M. L. Wigley, B. D. Santer, J. M. Gregory, K. AchutaRao, and K. E. Taylor, 2006. Krakatoa’s signature persists in the ocean. Nature, 439, 675, doi:10.1038/439675a.

Gregory, J. M., J. A. Lowe, and S. F. B. Tett, 2006. Simulated global-mean sea-level changes over the last half-millennium. Journal of Climate, 19, 4576–4591.

Gregory, J. M., 2010. The long-term effect of volcanic forcing on ocean heat content. Geophysical Research Letters, in press.

No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

Powered by WordPress