Science Briefs

Accounting for Climate's Backseat Drivers

The climate would be a much easier system to study if there was only one thing going on at a time. For instance, if the only external driver of climate change was the Sun, or if everyone agreed to halt greenhouse gas emissions when there was a big volcano going off, it would be simple to isolate the effects of these different factors.

Unfortunately, all of the different external drivers (often termed "forcings" in the climate literature) happen independently, and particularly so over the 20th Century. In fact, many of them happened at the same time. If we make an analogy to a bus where each passenger can drive, it seems as though everyone in the bus is trying to grab the steering wheel at the same time.

Photo of Canadian oil refinery.

Refineries servicing the oil sand fields in Alberta, Canada. How well can we separate the effect of greenhouse gas emissions produced by such human activity from other forcings in increasingly complex climate models? (Image credit: Kris Krüg via Flickr)

The question is, if the bus veers off course, who should we blame? Can we make any clear attributions to the individual factors? The answer appears to be a qualified yes.

In a recently published analysis, we looked at climate model simulations where only one thing was allowed to change at one time, and we compared them to simulations where everything was allowed to change all together.

The different external influences involved are varied. They can be natural in origin, like changes in the Sun's output or volcanic eruptions, or they can be human-caused, like greenhouse gas emissions, air pollution, or the "ozone hole". Because models can be used as numerical laboratories, we can use them to look at the effects of individual factors, something that is rarely possible in the real world, where the many different factors can be hard to isolate.

As climate models get more sophisticated, they explicitly incorporate more realistic physical and chemical processes. This means that processes that were fixed in simpler models are now calculated. This can complicate attribution. For instance, in a model that does not explicitly simulate atmospheric chemistry, the amount of ozone in the atmosphere can be specified, and it will make sense to ask what the impact of ozone changes are on temperatures and circulation. However, in a more sophisticated model the ozone will be calculated as a function of precursor emissions, the climate, and the atmospheric circulation. Thus, it is no longer possible to cleanly separate out the effect of just the ozone change, though there is an analogous suite of simulations that break up the drivers in a slightly different way.

We looked at simulations performed using three models. Two of them (the NCAR CCSM4 and the GISS-E2 R-NINT models) are complex coupled ocean-atmosphere models but do not have interactive chemistry or aerosols. The third (known as GISS-E2 R-TCADI) also included interactive chemistry. All of these simulations are available in the public domain as part of the Coupled Model Intercomparison Project Phase 5 (CMIP5 at LLNL; CMIP5 at GISS).

Figure with three panels

1900–2000 single-forcing and historical ensemble member temperature and precipitation signal-to-noise ratios for (a) CCSM4 (b) GISS-NINT (c) GISS-TCADI. The 99% significance levels for century-scale trends relative to internal variability are shown as gray boxes. The purple box represents the 99% confidence interval for the sum of single-forcing T and P signal-to-noise ratios. The sum of the mean single-forcing responses is shown as a white circle. (Figure modified from Fig. 1 in Marvel et al., 2015.)

We examined two different large scale measures of climate change: the global mean temperature and the global mean precipitation amount. Over the 20th Century simulations there was a significant global mean temperature increase in all models. In the first two models, we found that for both rainfall and temperature, the sum of the effects of the individual drivers over the century is very similar to the effects of all of the drivers added together (see figure above). That's good — it means that the different drivers do not noticeably interact amongst each other, and that, at least for these global quantities, we can sensibly talk about attribution of changes to individual drivers. For instance, the dominant factor for the temperature rise is the rise in greenhouse gases.

However, in the more sophisticated third model, the picture is a little more complicated. The different factors do not add up as cleanly. This means that attributions of the changes to the specific drivers need to also account for some interactions between the drivers. Looking more deeply into this, the main interaction (or non-linearity) seems to involve atmospheric ozone changes, which affect how rainfall and circulation patterns respond to other factors. This makes most of a difference in the 1980s and 1990s when the ozone hole was expanding due to the increase in ozone depleting substances.

So in summary, it is still a pretty good approximation to say that the whole is the sum of the parts, but as we get into more details, the non-linearities loom larger. That means in future we are going to need to be more careful in judging the importance of the different drivers in climate change. It may be that it really takes a number of drivers all interacting together to explain the climate bus’s erratic journey.

Reference

Marvel, K., G.A. Schmidt, D. Shindell, C. Bonfils, A.N. LeGrande, L. Nazarenko, and K. Tsigaridis, 2015: Do responses to different anthropogenic forcings add linearly in climate models? Environ. Res. Lett., 10, no. 10, 104010.

Contact

Please address all inquiries about this research to Dr. Kate Marvel.