Some Dyspepsia on Global Warming. by Andrew R. Solow.  October 2001.

Global warming has firmly established itself as a preeminent public policy issue, as industrialized nations consider costly reductions in carbon dioxide emissions, and a reluctant Bush administration faces serious political damage for refusing to go along. All the more surprising, then, that some of the basic assumptions underlying measured global warming have not received more careful thought. Three in particular bear closer scrutiny: the significance and meaning of global warming detection; how to respond given the uncertainty surrounding the measurement of warming; and finally, the explicit and implicit goals of its mitigation and the tradeoffs involved.

Usually, the detection question is posed quite simply: Has the mean annual surface temperature of the earth increased over the past century or so? That seems obvious enough, but the measurement is not straightforward. Within any particular year, the mean annual surface temperature is defined as the average of the temperatures at every point on the earth's surface at every time during that year. If temperature had been measured at every point on the earth's surface at every time during the past century, then obtaining the average would be a simple matter of arithmetic and the detection question could be answered unambiguously. Of course, temperatures have not been measured at every point at every time during any of the past 100 years. Instead, climatologists have used what temperature data are available to construct estimates of mean annual surface temperature. Although these estimates--which involve somewhat different data and averaging methods--differ in some details, all of them exhibit an irregular warming of approximately 0.5 degrees Celsius over the past century.

Given the consistency of the estimates, the only real question is whether they are representative of the earth as a whole. To be sure, such sampling gives rise to potentially significant errors. For example, the stations recording temperatures have expanded in geographical coverage over time. If the growth in this coverage had occurred predominantly in the tropics, where temperatures are higher, this by itself could produce an apparent warming. But climatologists are keenly aware of such problems and, while none would claim that the estimates are perfect, there is no real doubt that on average the Earth's surface has warmed by a few tenths of a degree Celsius over the past century. Although objections to this conclusion have been raised (based on the apparent disagreement between terrestrial and satellite temperature measurements), these point mainly to features of the vertical structure of temperature in the atmosphere. In fact, the satellite record is too short for this disagreement to cast significant doubt on the reality of the warming of the past century.

The more serious problem with measured detection is that it is silent on the cause of the warming. If the only possible cause of the warming were human activity, its detection would obviously point to an anthropogenic warming. But there are other possibilities. It is known from indirect measurements--for example, tree rings and isotope measurements in sediments and ice cores--that global temperature has varied substantially over the century time-scale with no aid from humans. Even numerical models of climate change exhibit this type of persistent natural variability. As a result, it is not possible simply to attribute the warming of the past century to human activity. By the same token, it would not be possible to dismiss the existence of anthropogenic effects should temperatures stabilize or even fall during the next few decades.

The recognition that natural climate variability can be large and persistent has led to a more refined version of the detection question: namely, is the warming of the past century consistent with natural variability alone or is some or all of it due to human activity? This is a problem of separating signal from noise. Doing so requires an understanding of the character of natural climate variability (the noise) and the potential pattern of anthropogenic climate change (the signal).

This is no simple task. A purely empirical analysis of natural variability on the century time-scale would require an observational record of at least a few thousand years, precluding the use of direct measurements. Tree rings, isotope measurements, and other so-called proxies can be used to reconstruct historical temperatures, but interpreting proxy records, which can be influenced by factors other than temperature, can be difficult. Numerical models can be used to simulate natural variability, but the results are only as good as the model. Given the limited observational record, evaluating such models is difficult.

Characterizing the potential pattern--the so-called footprint--of anthropogenic climate change is also difficult. One of the more obvious problems is that while carbon dioxide and other greenhouse gas emissions can cause warming, burning fossil fuels and biomass can emit aerosols that also can exert a cooling effect. The effects of aerosols on temperature are complicated and poorly understood, making it very difficult to incorporate them into a model of the footprint.

Recognizing such difficulties, the best available science nonetheless suggests that the warming over the past century cannot be attributed solely to natural variability. This represents a significant scientific achievement. But the science alone fails to address a critical issue: What is the relevance of this finding to public policy? The fact that climate is affected by human activity may not be sufficient, by itself, to engender a public policy response. Rather, it is the potential damage posed by climate change that would motivate a response. Clearly, the level and seriousness of these effects would correspond to the extent to which society should respond. That, in turn, would depend on more than just the existence of climate change, but on the details of the change.

Unfortunately, the best available science does not help us address many of these details. The uncertainty surrounding what is arguably the best understood measure of climate change--the equilibrium response of mean global surface temperature to a given increase in the atmospheric concentrations of greenhouse gases--is very large: a range of between 2 and 5 degrees Celsius. The uncertainty over the rate of warming, its regional distribution, and its effects on precipitation, sea level, hurricanes, and other climate processes is even greater.

Clearly then, deciding how to respond to climate change is more than a scientific problem. The response must somehow incorporate the uncertainties that the science yields, and the problem of global warming becomes one of decision-making under uncertainty.

While it is a truism that uncertainty should not be used as an excuse for inaction, the implied alternative is to act as if there were no uncertainty, or to respond as if the worst possible outcome is a certainty. This is just as clearly false. Instead, decision-making should take the uncertainty into account. We know how to do this. It involves taking immediate low-cost or cost-free measures, or measures that have additional benefit and make sense for other reasons. There is no shortage of these. But it also involves playing for time, gathering information that reduces uncertainty, and promoting future flexibility by keeping existing options open and developing new ones. Such responses are not necessarily signs of moral failure: they are part of a sensible response in the face of uncertainty.

One aspect of the climate change problem that seems to pose a particular problem for decision-making under uncertainty is the possibility, however remote, of a truly catastrophic change. Two commonly mentioned possibilities are the collapse of the West Antarctic ice sheet, which could raise sea level by several meters, and the shutdown of the thermohaline circulation in the ocean, which could lead to rapid cooling over parts of North America and Europe. While these possibilities are not utterly without scientific basis, fixating on them seems to reflect a psychological attitude somewhat akin to a fear of flying. It is almost always possible to concoct a plausible, catastrophic scenario to accompany any decision whose consequences are uncertain. For example, a mandated large-scale reduction in the use of fossil fuels could--conceivably--plunge the world into an economic depression, exacerbate regional tensions and set off a nuclear war. The point is not to assess the probability of apocalyptic scenarios, but to understand that allowing such scenarios to dominate or determine policy responses may not be sound.

A final word about uncertainty. Emissions reductions are commonly advocated as a form of insurance against uncertain future climate changes. Insurance has a nice ring to it. Prudent people buy insurance. But curtailing greenhouse gas emissions, however beneficial that may be, is not insurance. Insuring an automobile does not mean leaving it in the garage. It means making an investment so that, should an accident occur, resources are available to cover the damages. Automobile insurance works by pooling risk: lots of people buy it, but only a few collect on it. Yet because climate change would affect everyone, such risk-pooling is not straightforward.

In principle, it is possible to self-insure against the risk of climate change. That would entail setting aside resources to cover potential future damages, akin to the idea of saving for a rainy day. As applied to global warming, society could, for example, make current investments to improve future adaptability to climate change. Such clearly relevant questions are seldom, however, discussed with any real seriousness in the global warming debate.

Also implicit in all such discussions is the assumption that the reason to respond to global warming is to avoid human suffering. Under this assumption, deciding how to respond to global warming would require comparing the returns to social investment in this area to returns in other areas.

Such comparisons--once they are examined directly--raise critical questions. There is no shortage of human suffering that is completely unrelated to global warming. Half the world's population lacks such basic necessities as food, water, and medical care, let alone education and human rights. There is nothing new or uncertain about this. There is absolutely no doubt that the return to a dollar invested in these areas is many times--perhaps orders of magnitude--greater than the return to a dollar invested in stabilizing the composition of the atmosphere. Such comparisons highlight the inherent contradictions between evincing concern over human suffering and advocating a costly response to global warming. At the very least, these should necessitate taking a clearer look at the choices and tradeoffs involved in addressing the global warming problem. Ignoring such questions does not solve them.

Andrew R. Solow is an associate scientist at the Woods Hole Oceanographic Institution.