Uncertainty in Earth System Components and Implications for Climate Change Risk

Uncertainty is inevitable in a complex process such as climate change, where human activities are influencing the entire Earth system in ways that will continue for centuries. The fact of this uncertainty is used both as justification for waiting to take action and as a reason for urgent measures to limit the human influences. Assessments of uncertainties in the human-climate interaction, and the associated societal risks, are needed to clarify this debate and to inform decision-making and policy formulation. To properly achieve this end requires careful attempts to describe uncertainty in quantitative terms, and to assess the extent to which mitigation policies may be able to reduce the risk of extreme climatic change and other dangerous levels of interference in the Earth system. Even with best efforts, however, the existence of deep structural uncertainty means that existing models may underestimate the chance of extreme changes, or fail to capture processes that could lead to abrupt changes, and the Program seeks to investigate these processes as well.

A central contribution of this work has been to empirically estimate the likelihood of given amounts of climate change, conditional on specific assumptions about future climate policy. As such, the work recognizes that uncertainty in such projections derives from both uncertainty in our understanding of the physical and biological Earth system, and in the growth of the economy, resource availability, and changing sets of available technologies. The key study describing this work is Uncertainty analysis of climate change and policy response, which describes the likelihood of exceeding significant levels of global and latitudinal temperature and sea level rise. This provides the foundation for the MIT "Greenhouse Gamble" -- roulette wheels depicting the probabilities of different levels of climate change. These images have proved invaluable in communicating to many audiences the risks of climate change and how policy measures can reduce but not eliminate those risks.

The forward projections of likelihood are, however, only as good as the inputs to the exercise. In deterministic modeling the phrase "garbage-in, garbage-out" is a modeler's reminder of the importance of care in establishing estimate of input values. Modeling uncertainty does not make the input values any less critical -- if anything it makes the job more difficult. An accurate characterization of not only a median value is needed, but also the entire range of plausable inputs and the distribution of their likelihoods, known as the probability density function.

A key element of the effort on physical and biological feedbacks to climate change is a flexible modeling system that can represent a range of feedback responses to find that set (and their joint probability) that is most consistent with observations of atmospheric and ocean temperatures, to detect and quantitatively attribute climate change to various anthropogenic and natural causes. The method is generally known as "optimal fingerprint" as it uses not just a global average temperature but the particular vertical temperature gradient. Pioneering work in the Program included Quantifying uncertainties in climate system properties with the use of recent climate observations. That early work has been updated and generally suggests more rapid atmospheric warming than is generally predicted by other groups, because it finds that the ocean is taking up heat more slowly than is projected by many climate models.

Incorporating uncertainty about human activities in projections of climate change is also important, and the Program has pushed for the need for clearer quantitative statements of uncertainty in major climate assessments (Uncertainty and climate change assessments). Given advances in understanding the Earth system, and more recent evidence on the growth of emissions from human activities, and on technology options, a major update of our empirical estimates of future warming is underway. Unfortunately, our preliminary indicators are troubling as they show much more warming than we previously thought was likely. So we are uncertain. Should policy-makers wait until uncertainty is resolved? Or does the very fact that there are risks mean that we should bring the energy system to a halt until low-carbon solutions can be brought on line? One can find both of those recommendations coming from different groups with obvious axes to grind.

The decision process we must confront, in the terminology of statistics, is a classic example of "Type I" and "Type II" error: treating a problem you thought you had, only to find out you didn't, and taking some risks with the treatment; or delaying treatment on the bet that you didn't have the problem, and then when it is clear that the problem is serious, the treatment is much more difficult or much less effective. At its basic level this is a problem of calculating probabilities and expected values of different actions. But can you just take the expected value or does inertia in the system bias the solution to doing either more or less? This question is investigated in The curious role of "learning" in climate policy: Should we wait for more data?. What about the response of the social system: Will we get used to a changing climate and thereby underestimate the risks? Once we see that we have a problem will we act in time.

Unfortunately, much of the risk literature is based on risks that are relatively easy to quantify, or affect a limited number of people at any one time, and so risk-pooling (i.e., through insurance of other similar mechanisms) is a viable option. Climate risks are poorly understood and threaten the entire planet. It's not what we know, but what we don't know, that is of most concern. How to best make decisions under those circumstances is unclear. The Program seeks to investigate where strong non-linear feedbacks may exist. This includes work on the ocean thermohaline circulation and the risk of its collapse and feedbacks. Investigation of potentially significant changes in Arctic areas is advancing under a relatively new research grant that aims to quantify climate feedbacks from abrupt changes in high-latitude ecosystems. And continuing work is focused on the possible methane and CO2 release from Arctic tundra systems as they warm.