We begin with the possibility of dynamically induced climate changes. Inherent variability in the atmosphere, possibly associated with non-linear dynamics, is thought to be capable of generating intraseasonal variability. Hansen and Sutera (1986) utilized a planetary-wave activity index to resolve two modes of Northern Hemisphere winter 500 mb height distributions, one with low and one with high amplitudes. The differences between the height fields are centered over the Aleutians, north central Canada, and western Europe. If these two modes represent interannual tendencies that might be selectively activated, they could conceivably set in motion climate feedbacks (snow, sea ice, ocean mixed-layer depth) with climate implications and longer time scales. It is not known whether the modes are forced or simply the result of non-linear dynamics. This type of "natural" variability must be considered speculative when applied to the climate domain. Additional decadal variability may be found in meridional modes of the coupled ocean-atmosphere system that propagate from pole to pole over 3- to 20-year time periods (see, e.g., Mehta, 1992). Their relationship to climate variability is also unproven.

Hansen et al. (1988) ran a GCM with a mixed-layer ocean for 100 years (Table 1). The model produced an oscillation of 0.5°C (peak to peak) over that interval without any external forcing (Figure 1, top). As emphasized by Barnett et al. (1992), an inversion and slight time shift of


Top: Global-mean (unforced) surface air temperature variations in a 100-year control run with an atmospheric GCM and mixed-layer ocean model. (From Hansen et al., 1988; reprinted with permission of the American Geophysical Union.) Bottom: An inversion and time shift of the results shown in the top figure, compared with observed temperature changes over the past 100 + years. (From Barnett et al., 1992; reprinted with permission of the American Geophysical Union.)

this record results in matches to a number of the features in the observed temperature record from 1850 to 1950 (Figure 1, bottom). The implication is that the system is capable of producing such temperature oscillations without the need for causative factors (i.e., forcing).

As analyzed by Barnett et al. (1992), the primary contributions to the temperature oscillation occurred in the tropics (Figure 2); over 90 percent of the variability was located within 30° latitude of the equator. The oscillation was largely the result of a cloud cover/sea surface temperature correlation: Low clouds increase as the ocean cools, further reducing the net heating of the surface. At least part of this correlation was due to the sub-grid-scale temperature parameterization used in the GCM. In contrast to these results, the observed variability over the last century has been dominated by high-latitude contributions (Hansen et al., 1983a). Thus there is good reason to doubt that the variability found in the GISS GCM 100-year run is a convincing indicator of the natural variability in the climate system. The inclusion of additional processes, such as ocean dynamics, variations in cloud optical thickness, and the heat exchange with the ocean below the mixed layer would improve the model's realism (Barnett et al., 1992).

Manabe et al. (1991) produced similar-looking temperature global variations in a 100-year simulation with a fully coupled atmosphere-ocean GCM. The greatest persistence was found in the polar oceans of both hemispheres, due to their weaker thermal damping (reduced surface and radiative heat fluxes) and deep mixing. The relevance of this run to real-world oscillations is limited by the continuously specified input of heat and salinity to the ocean; while this input does not vary from year to year, its very existence will perturb the model's own natural variability, and the necessity for incorporation of these fluxes raises doubts about the model's accuracy.

In yet another unforced 100-year simulation with a coarse-resolution model and a mixed-layer ocean, Houghton et al. (1991) found decadal and longer-time-scale variability in sea surface temperatures; in fact, the sea surface temperature at low and middle latitudes had a distinctly red spectrum. The percentage of variance in the low-frequency range (4 to 100 years) showed a small increase with latitude, with high-latitude effects associated with decadal-scale sea-ice variability (as has been noted by Mysak et al. (1990) in the Greenland and Labrador seas). The simulated global surface air temperature did not display significant multidecadal-scale variability, mainly because little such variability arose in the tropical middle-latitude domain (in contrast to the GISS model results). Cloudiness was not allowed to change in these experiments, which it could in the GISS model.

In summary, the unforced decade-to-century-scale oscillations of global temperature that characterize the natural climate system may well be on the order of several tenths

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