Our subject of discussion in this chapter is global climate prediction and the uncertainties of such predictions. What do we mean by a prediction of climate? E. Lorenz, the father of the chaos theory (Gleick, 1988), once clarified the important difference between forecasts of climate anomalies, such as the one caused by the El Nino phenomenon, and forecasts of the state of climate caused by changes in the solar forcing or by changes in the composition of the atmosphere. The first of these phenomena can in principle be predicted per se with useful skill, while in the second case only changes in the statistical structure of climate can be predicted. We will not be able to say whether a particular summer or winter will be warmer or colder than normal but only say, for example, that the number of summers with a temperature above a certain value will be more common than what it was previously. In this chapter I will use the expression climate prediction only in the context of the ability to simulate or predict the overall statistics of climate.

Even if the second kind of prediction is less precise, it is nevertheless very important, since knowledge of the average condition of climate, including its statistical structure in space and time, is of importance to the society and to the environment. Only a modest change in the average temperature or precipitation may imply changes in the statistical distribution of extremes. Precautions for flooding in most cities and municipalities, for example, have been designed to withstand extreme events, which on average occur only once a century. A warming on the order of one degree may lead to changes in the moisture content of the atmosphere by some 6%, with the consequences that extreme rainfall may be more common.

A fundamental question that first needs to be addressed is whether climate is at all unique or transitive, that is, for a given set of external forcing it follows that there exists a unique set of cli mate statistics, or whether several possible sets of climate statistics are possible for a given set of forcings? Such an intransitive climate would then be a priori unpredictable, since infinitesimal changes in the initial data or in the forcing may change the climate in ways similar to the chaotic processes that limit the length of useful weather forecasts (Lorenz, 1968). We believe, based on numerical modeling studies, that this is not the case with the atmosphere when forced from prescribed boundary conditions, but there is no indication that it will not be the case when we incorporate the full feedback with the oceans and the land surfaces. In fact, we have several indications of the nonuniqueness of the earth's climate, one of them related to the thermohaline circulation of the ocean. Such a mechanism, indicated by Stommel (1961), arises from the influence of the ocean salinity on the vertical heat exchange with the deep ocean, whereas salinity does not influence the interaction with the atmosphere (Bryan, 1986; Maier-Reimer and Mikolajew-icz, 1989). There are other similar examples from the interaction between the atmosphere and the land vegetation which can change the regional climate at least as significantly (Claussen, 1998; Brovkin etal, 1998).

These considerations and the additional fact that climate is the integral of weather over long periods of time, the weather itself being unpredictable, mean that the predictability of climate is a fundamental issue. The uncertainties of global climate prediction are a broad subject and I will here restrict my presentation to a time scale of a few hundred years and thereby concentrate on the time from early industrialization to the middle of the next century or so.

In Section 2,1 will discuss the observational evidence of climate change. As a suitable point of reference in my review, I will use the recent attempts by Mann et al, (1998, 1999) to reconstruct the surface temperature of the Northern Hemisphere for the last millennium. In Section 3, I will address the physical rationale underpinning climate change modeling, and finally in Sections 4 and 5,


Copyright 200! bv Academic Press. Ali rights of reproduction in any form reserved.

I will be concerned with the modeling aspects and present and analyze some general results of numerical experiments.

0 0

Post a comment