tions of a whole host of parameters describing climate have been collected for centuries. These observations and other indirect but equally valid ones definitely show that climate varies on a short time scale and can change significantly on longer time scales. These facts have been especially well summarized elsewhere in this volume and in two recent reports, Understanding Climatic Change: A Program for Action (U.S. Committee for the Global Atmospheric Research Program, 1975) and The Physical Basis for Climate and Climate Modelling (WMO/ICSU, 1975).
These documents recommend many new monitoring observations from satellite platforms. Space technology has made truly global observations possible for the first time. Because spacecraft carry the same instruments over different parts of the earth, regional differences and variations should be more easily detected. Despite these important advantages there are also significant limitations. Obviously, spacecraft orbit well above the earth’s atmosphere, and certain common climatic parameters such as temperature can only be inferred from the electromagnetic radiation emanating from the atmosphere below. One can hardly expect to achieve the same intrinsic accuracy that an in situ thermometer would achieve. On the other hand, other quantities such as the extent of sea ice and possibly even the thickness of sea ice can, at least in principle, be far better determined from space than is economically feasible using observers on the earth.
With these new possibilities for climate monitoring, will we be able to measure man’s influence on climate? Perhaps, but most likely not. The difficulty arises from the fact that changes in climate that are significant in their effects on man may be scarcely detectable on a global scale. The situation is not quite so difficult on the regional scale. Regional changes are often larger in magnitude but compensated for by changes in the opposite direction in other areas. Thus small global changes that are difficult to measure may be manifested in regional shifts that we can observe quantitatively.
Even though our ability to monitor the climate of the earth is very much better now than ever before, and even though this new capability will make it possible to obtain a better set of observations of many key climatic parameters, it does not appear possible at present to measure how man is changing the global climate. On the other hand, it may be possible to measure regional climate changes because these changes are larger. We are already certain that we can observe changes in local climate because that has already been done without space platforms. We probably can obtain certain local observations even better with them. On the local scale, we may even be able to separate the changes caused by man from those caused by nature.
The most productive approach to devise a climate observing system is through a combination of modeling and monitoring. One of the most successful accomplishments of the Global Atmospheric Research Program (GARP) so far is the clear specification of what is required to describe the initial state of the atmosphere so its short-term transient behavior (i.e., the weather) can be predicted. Numerical atmospheric simulation schemes, often called numerical models, require that the state of the atmosphere be specified at some initial time for several levels over several thousand grid points spaced over the earth. The specification of the atmospheric state can be real (from measurements) or fictitious (from guesses) or both, but the data set cannot be empty or only partly filled. A model that simulates atmospheric behavior is an especially powerful tool not only because it can predict the future weather, but, equally important, it specifies what observations are the key ones. Moreover, sensitivity tests with the model can be used to learn how good, how often, and how closely space d these observations must be. These modeling tools have had a great influence on our meteorological satellite program.
What can be measured from space must not only be novel, it must also be useful. Sensitivity tests indicate just how useful the observations will be. For example, our ability to observe the atmosphere’s initial state is now considered promising enough—on the basis of model experiments—to warrant a large international cooperative program—the First GARP Global Experiment (FGGE)—which will be conducted in 1978–1979.
The status of climate modeling in the late 1970’s is not so fully developed as weather modeling was in the late 1960’s when GARP was first proposed. The problem is that we do not have our theory of climate in as good order as our theory of weather was at that time. Stated simply, the short-term future behavior of our atmosphere depends on its present physical arrangement and almost unchanging boundary conditions. But the present state of the atmosphere is unimportant for climate because the atmosphere soon “forgets” its present state. Its statistical behavior depends more on the boundary conditions, and the transient behavior of the atmosphere can slowly change the boundary conditions. Such feedback mechanisms can be positive or negative.
In summary, models give us the basis for determining what observations are required for the study of climate and how well they must be obtained. These same questions have been looked into in detail at several study conferences. What follows has been extracted from those conference reports at which the author was a participant. These have been summarized and added to in an attempt to present the latest consensus on the observing requirements. In doing this, the author acknowledges the contributions of the other participants in these meetings (listed in the referenced reports) and assumes full responsibility for any change in emphasis, deliberate or inadvertent, that such a synopsis may incur.
If one takes the time to read all the recent documents on the requirements for a global climate observing system, one comes away with two strong impressions. First, these documents do not read like novels. Secondly, there is a considerable difference between what the modelers want and what has been proposed as feasible by those familiar with how the observations might be obtained. Part of this confusion stems from the lack of a satisfactory theory of climate and part from our ignorance of the best approach