May 16, 2005

Determining Climate Sensitivity from Volcanoes: Observations vs. Models

Recently there have been several papers published that have attempted to use the evolution of the earth’s temperature after big volcanic eruptions as a determinate of the earth’s climate sensitivity—that is, how much the average temperature changes with a change in climate forcing (i.e. a change of energy input). Having a good understanding of the climate sensitivity is key to having a good understanding of future climate change.

Oftentimes, the sensitivity is reported as the temperature change resulting from an energy change that is equivalent to the one assumed for a doubling of the atmospheric carbon dioxide levels (from pre-industrial values). In its 2001 Third Assessment Report (TAR) the Intergovernmental Panel on Climate Change (IPCC) settled on a value of 3.7 watts per meter squared (W/m2) (see out last article for more information on energy units) for the energy change associated with a doubling of CO2. That’s the easy part. Figuring out how much the earth’s average temperature will change as a result has proven to be much more difficult.

The best the IPCC can do it to suggest that the temperature change for a doubling of CO2 is somewhere in the range of 1.5 to 4.5ºC. It seems that neither climate models, nor long-term climate observations, are yet able to provide enough information to narrow that range. For instance, the range of sensitivities from 15 mixed-layer climate models in the IPCC TAR was 2.0 to 5.1ºC for a CO2 doubling.

Trying to use long-term observations of the earth’s temperature history isn’t much better because we are unsure of how much the total energy change has been to date. According to the IPCC’s TAR, the total amount of positive climate forcing (energy increase) that has occurred since pre-industrial times has been about 3.3 W/m2 (made up a combination of increases in carbon dioxide, methane, nitrous oxide, halocarbons, tropospheric ozone, solar output, and black carbon soot, among other factors). Since the total temperature rise during the past 150 years or so has been about 0.8ºC, that would imply a climate sensitivity (for an equivalent CO2 doubling) of about 0.90ºC ((3.7/3.3)*0.8))—a value below the low end of the IPCC range. But, there have been other changes to the earth’s atmosphere that have led to negative climate forcing (or energy reduction). These are primarily associated with the cooling effects of atmospheric aerosols. The problem is that no one has a good idea about how large these negative forcings have been. The IPCC estimates vary wildly from somewhere around -0.9 to -2.9 W/m2. Combining this with the positive forcing estimate yields a range for the total change in forcing from pre-industrial times to present of somewhere between 0.4 to 2.4 W/m2, which, when combined with the observed temperature change, produces a climate sensitivity for an equivalent doubling of CO2 of 1.2 to 7.4ºC—not particularly insightful.

[As an aside, three recent papers appearing the in same issue of Science magazine, combine to indicate that the negative forcing from atmospheric aerosols has greatly diminished over the past two decades as air pollution controls have become effective. This cleaning of the air has reportedly led to a positive forcing increase of a whopping 5.6W/m2 since 1985! Since the global temperature change since 1985 has been about 0.36ºC, this suggests a sensitivity for a CO2 doubling of a miniscule 0.24ºC. These results, however seem highly suspect (for one thing, the worst-case IPCC negative forcing increase over the past 150 years is only 2.9W/m2—so how can a decrease in air pollution during the last 20 years lead to a forcing increase of nearly double that amount?). See here for more details about the Science articles.]

So, instead of trying to calculate the climate sensitivity using uncertain changes in forcings over longer time scales, some researchers have tried to study the temperature effects of shorter-term, better quantified forcing events—such as large volcanic eruptions.

Recent large eruptions, such as El Chichón in 1982 and Mt. Pinatubo in 1991, were fairly well measured in terms of how much material was injected into which parts of the atmosphere and how long it stayed there. From this information, it is possible to determine the magnitude and temporal evolution of the negative forcing that each eruption had on the climate system (primarily from ash and dust in the stratosphere reflecting back incoming solar radiation). Now all we need is the temperature response.

However, the global average temperature responds to other things besides volcanoes (e.g. El Niños, greenhouse changes, seasonal cycles, etc.) and this natural variability obscures some of the volcanic response details—details which prove to be vital to the determination of the sensitivity.

One way around all of these pesky, uncontrollable complicating factors present in the real world, is to step into the highly controllable world of climate models. With a climate model, you can set up the experiment in any way that you would like it (producing any result that you would like along the way, but we digress). This is precisely the approach taken by Tom Wigley and co-authors (Caspar Ammann, Ben Santer, and Sara Raper) in a recent paper appearing in the Journal of Geophysical Research. Wigley et al. used a collection of climate models to simulate the effect of historical volcanic eruptions of the global temperatures. By combining the results from many different models, they were able to remove most of the “natural” variability and thus better isolate the temperature change in response to just the volcanoes. To remove even more of the non-volcanic noise, Wigley demonstrated that the results produced by the full climate models could be closely replicated by a simpler climate model that could be set only to respond to the volcanic signal and nothing else. In this way, a pure global temperature response could be produced. Another advantage of the simple climate model is that it is quick to run (unlike full-scale models) and basic parameters, such as the climate sensitivity, can be set for each run. By running the model with a range of climate sensitivities while using the same set of volcanic forcings, a variety of temperature responses can be calculated.

Through these various model runs, Wigley and colleagues found that not only does the climate sensitivity affect the magnitude of the maximum temperature response (cooling) but also the speed of the recovery—basically, the lower the climate sensitivity, the less the total cooling and the quicker the return to normal conditions. Based upon these model results, and a comparison with a value that Wigely and Santer had determined from previous papers for the maximum cooling observed for the past three large volcanic eruptions, Wigley et al. determined the climate sensitivity for an equivalent doubling of atmospheric CO2 to be in the range of about 1.54 to 3.03ºC, although the 95% confidence range extended from 0.30 to 7.73. In other words, values not too different from the IPCC range of 1.5 to 4.5. All that work for apparently so little gain!

Actually, maybe not.

Wigley suggests that earlier work which attempted to relate the observed volcanic signal with the observed temperature response using real world observations, citing specifically some work the we had done a few years ago (Michaels and Knappenberger, 2000), was too simplistic in that we assumed that the temperature response was virtually contemporaneous with the forcing. Wigley might have a point here, for in actuality there is probably some degree of thermal inertia such that the temperature recovery is slower than the forcing recovery. And, as we mentioned earlier, Wigley has determined that the temperature recovery time tells you something about the climate sensitivity. In Wigley’s opinion, we had used too quick a recovery time, and thus our climate sensitivity was too low, and our conclusions were “suspect.”

The problem, as we see it, however, is that Wigley’s conclusions are primarily drawn from the results he obtained from climate models and not real world observations (we relied on the latter for our analysis). So, accepting that there may be some degree of thermal inertia in the response to volcanic forcing, we attempted to determine what amount of delay between the forcing and the temperature was required to best fit the observed temperature history of the lower atmosphere during the period 1979-2000 (spanning two volcanic eruptions and replicating the analysis in Michaels and Knappenberger, 2000). To do this, we substituted the observed volcanic forcing signal with one that had a prescribable exponential decay (to match the form suggested by Wigley). We then altered the decay time and via multiple regression determined how well the prescribed volcanic signal, along with a measure of ENSO (e.g. El Niños/La Niñas), matched the observed temperature history. Figure 1 shows our results as we lengthened the exponential decay times from 12 to 45 months. It is clear that the temperature observations are best replicated when the decay time is around 20 months. Wigley and colleagues, using their climate models, find that the best decay time for both the El Chichón and Pinatubo eruptions are 37 months—nearly twice the value found by using observations rather than models.

Time Decay

Figure 1. Percent variance explained (the higher the number the better the match) of the observed temperatures in the lower atmosphere from 1979 to 2000 by a combination of a measure of ENSO and volcanic eruptions. The decay time of the temperature response to the volcanic forcing was varied from 12 to 45 months.

Remember that Wigley et al. found that the shorter the recovery time, the lower the climate sensitivity. Therefore, our results with observations suggest that the climate sensitivity as determined using climate models is overestimated and in fact, lies closer to the low end of the IPCC range than to the high end.

Similar results were recently reported by other researchers who have set out to use real world observations of the climate response to the eruption of Pinatubo to determine the climate sensitivity. David Douglass and Robert Knox from the University of Rochester, reported in Geophysical Research Letters that, based upon their calculations and their fit of the observations, that the decay time for the temperature response to Pinatubo was several times shorter than the 37 months calculated by Wigley et al. using climate models, and that the resulting climate sensitivity to a doubling of CO2 was only about 0.6ºC—far beneath climate model estimates.

From all of this, it is seem apparent that real world observations of the climate response to recent volcanic eruptions suggest that the earth’s climate sensitivity is quite a bit less than the sensitivity as determined by climate models. Observations suggest that the sensitivity lies near or below the low end of the IPCC range of 1.5 to 4.5ºC for a doubling of CO2.

The lower the climate sensitivity, the lower the potential future temperature rise, and the lower the potential impact. This is precisely the same conclusions that one arrives at when relating observed temperature changes to other observed quantities such as carbon dioxide emissions or carbon dioxide concentration (see here for example).

As time goes on, and research continues, it is becoming clearer and clearer that the high end of the range of temperature increases projected by the IPCC is simply scientifically unsupportable. All the while, the evidence mounts suggesting that the low end is the more probable outcome.


Douglass, D.H., and R. S. Knox, 2005. Climate forcing by volcanic eruption of Mount Pinatubo. Geophysical Research Letters, 32, doi:10.1029/2004GL022119.

Michaels, P.J., and P.C. Knappenberger, 2000. Natural signals in the MSU lower tropospheric temperature record. Geophysical Research Letters, 27, 2905-2908.

Santer, B.D., et al., 2001. Accounting for the effects of volcanoes and ENSSO in comparisons of modeled and observed temperature trends. Journal of Geophysical Research, 106, 28,033-28,059.

Wigley, T.M.L., 20000. ENSO, volcanoes and record breaking temperatures. Geophysical Research Letters, 27, 4101-4104.

Wigley, T.M.L., et al., 2005. Effect of climate sensitivity on the response to volcanic forcing. Journal of Geophysical Research, 110, doi:10.1029/2004JD005557.

No Comments

No comments yet.

RSS feed for comments on this post. TrackBack URI

Sorry, the comment form is closed at this time.

Powered by WordPress