Modeling Low-Frequency Climate Variability: The Greenhouse Effect and the Interpretation of Paleoclimatic Data



Detection of the decade-to-century time-scale effects of any external forcing agent on global-mean temperature requires a knowledge of internally generated natural variability on these time scales. Methods for estimating the magnitude of this variability are described. While the observed twentieth-century warming agrees well with model estimates based on greenhouse-gas and aerosol forcing alone, it is shown that changes of similar magnitude could have occurred solely through internally generated variability. Similarly, observed changes in global-mean temperature over the past 10,000 years (in particular, the numerous Little Ice Age events) could have been internally generated. At the very least, if these events were externally forced (e.g., by solar irradiance changes) then internal variability would probably have modified the response markedly from a "pure" response to external forcing, making the interpretation of the paleoclimatic record extremely difficult.


Understanding the past is often said to be the key to predicting the future. The reverse is equally true: Attempts to predict the future can help us to understand the past. These maxims will be illustrated below.

An important but nebulous question raised in the context of the greenhouse effect is, "Have we detected it?" This issue concerns not just anthropogenic greenhouse-gas-induced climate change but past climate changes as well. The primary question in both cases is the identification of an externally forced change (a signal) in the presence of natural variability (background noise). Of course, detection requires more than just identifying a climate change that is consistent with some hypothesis—it requires also the demonstration of a cause-effect relationship and the attribution of some part of the observed changes to the particular cause. The present paper deals mainly with identification, rather than attribution.

We begin by comparing predicted and observed changes in global-mean temperature over the instrumental period, highlighting the role of low-frequency natural variability. We then review methods that may help to elucidate the character of this variability. Finally, we broaden the detection issue to consider the detection of external forcing effects


University Corporation for Atmospheric Research, Boulder, Colorado


Climatic Research Unit, University of East Anglia, Norwich, United Kingdom

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement