Natural Variability and Climate Sensitivity

One of the central tasks of climate science is to predict the sensitivity of climate to changes in carbon dioxide concentration. The answer determines in large measure how serious the consequences of global warming will be. One common measure of climate sensitivity is the amount by which global mean surface temperature would change once the system has settled into a new equilibrium following a doubling of the pre-industrial CO2 concentration. A vast array of thought has been brought to bear on this problem, beginning with Arrhenius’ simple energy balance calculation, continuing through Manabe’s one-dimensional radiative-convective models in the 1960’s, and culminating in today’s comprehensive atmosphere-ocean general circulation models. The current crop of models studied by the IPCC range from an equilibrium sensitivity of about 1.5°C at the low end to about 5°C at the high end. Differences in cloud feedbacks remain the principal source of uncertainty. There is no guarantee that the high end represents the worst case, or that the low end represents the most optimistic case. While there is at present no compelling reason to doubt the models’ handling of water vapor feedback, it is not out of the question that some unanticipated behavior of the hydrological cycle could make the warming somewhat milder — or on the other hand, much, much worse. Thus, the question naturally arises as to whether one can use information from past climates to check which models have the most correct climate sensitivity.

In this commentary, I will discuss the question "If somebody were to discover that climate variations in the past were stronger than previously thought, what would be the implications for estimates of climate sensitivity?" Pick your favorite time period – Little ice age, Medieval Warm Period, Last Glacial Maximum or Cretaceous – the issues are the same. In considering this question, it is important to keep in mind that the predictions summarized in the IPCC reports are not the result of some kind of statistical fit to past data. Thus, a revision in our picture of past climate variability does not translate in any direct way into a change in the IPCC forecasts. These forecasts are based on comprehensive simulations incorporating the best available representations of basic physical processes. Of course, data on past climates can be very useful in improving these representations. In addition, past data can be used to provide independent estimates of climate sensitivity, which provide a reality check on the models. Nonetheless, the path from data to change in forecast is a subtle one.

Climate doesn’t change all by itself. There’s always a reason, though it may be hard to ferret out. Often, the proximate cause of the climate change is some parameter of the climate system that can be set off from the general collective behavior of the system and considered as a "given," even if it is not external to the system strictly speaking. Such is the case for CO2 concentration. This is an example of a climate forcing. Other climate forcings, such as solar variability and volcanic activity, are more clearly external to the Earth’s climate system. In order to estimate sensitivity from past climate variations, one must identify and quantify the climate forcings. A large class of climate forcings can be translated into a common currency, known as radiative forcing. This is the amount by which the forcing mechanism would change the top-of-atmosphere energy budget, if the temperature were not allowed to change so as to restore equilibrium. Doubling CO2 produces a radiative forcing of about 4 Watts per square meter. The effects of other well-mixed greenhouse gases can be accurately translated into radiative forcings. Forcing caused by changes in the Sun’s brightness, by dust in the atmosphere, or by volcanic aerosols can also be translated into radiative forcing. The equivalence is not so precise in this case, since the geographic and temporal pattern of the forcing is not the same as that for greenhouse gases, but numerous simulations indicate that there is enough equivalence for the translation to be useful.

Page 1 of 6 | Next page