they induce would be similar to those shown for solar radiation reduction (Figure 5), since doubled CO2 and a 2 percent solar constant increase yield similar patterns of model response (Hansen et al., 1984).
The different forcings and natural variability discussed above have in many cases different geographical expressions (see, e.g., Figures 2, 3, and 5, and Table 1). For example, the forcing that is of radiative importance globally, such as reduced insolation or continual volcanic-aerosol injections, has a larger impact at inland locations away from the moderating influence of the ocean. In contrast, changes in ocean heat transport and NADW production have maximum influence in coastal locations. The former type of forcing would therefore imply maximum Little Ice Age cooling in central Eurasia, decreasing toward the Atlantic coast, whereas the opposite would prevail if NADW-formation changes were acting alone.
It might therefore appear to be possible to distinguish between them through reference to paleoclimatic data for the last several centuries. A joint NSF/NOAA project, Analysis of Rapid and Recent Climate Change (ARRCC), has begun reconstructing the geographical climate distribution for several time periods during the last few hundred years, with this goal in mind. Worldwide distributions of temperature and precipitation changes will be needed for this purpose. Whether the data are sufficiently comprehensive and accurate remains to be seen.
A complementary approach involves reconstructing the potential climate forcing for this time period, whether it be solar irradiance, volcanic aerosols, trace-gases, or ocean circulation. As noted above, there are many empirical estimates of solar irradiance variations, but they need to be quantified on the basis of physical principles. The actual volcanic-aerosol properties are similarly difficult to constrain (Bradley and Jones, 1992b). It may be possible to deduce ocean-circulation changes through biological/ geochemical studies in high-resolution sediment and coral cores, but that is not yet certain.
With both the climate response and climate forcing somewhat ambiguous for these time periods, in this paper we have had to rely on current climate models to suggest possible links. The major problem here is that we cannot be sure that the models have the proper climate sensitivity, either globally or locally. Analysis of climate forcing on the time scale of glacial/interglacial epochs suggests that we probably know the global sensitivity to within a factor of 2 (Lorius et al., 1990). The shorter-term response is hard to calibrate, and its latitudinal expression is likewise somewhat uncertain, beyond the relative certainty that there is a greater thermal response at high latitudes, especially in winter.
Given the above uncertainties, it does not appear as though we can rule out any of the proposed mechanisms for forcing climate fluctuations on the decade-to-century time scale, although there is so far little proof that any of them has a magnitude sufficient to strongly influence global temperatures. The radiative forcing due to the observed trace-gas increase is the climate perturbation that can be quantified best; our ability to quantify its historical effect is limited by uncertainty in the climate response time, and the potential offsetting influence of increased tropospheric aerosols. From the perspective of recent observations in the current century, it is perhaps more likely that a sustained insolation reduction of 0.25 percent could have occurred (associated with the Maunder Minimum) than that volcanic eruptions on the order of El Chichon could have arisen every five years. However, we do not know how typical this last century is, nor do we know what portion of the estimated larger insolation variations might have occurred at visible wavelengths capable of penetrating to the troposphere. While the unforced variations that arose in the GISS GCM may not ultimately prove to be physically realistic, this does not imply that unforced fluctuations are not possible, with perhaps some feedback involving ocean temperatures that operate on a longer time scale. The results also indicate that global cooling on the order of 0.5°C to 1°C is not sufficient to guarantee cooler temperatures everywhere, since advective changes, perhaps caused by the climate perturbation, can dominate the local response. Therefore, the possibility of global-scale forcings should not be dismissed if local regions of warming are uncovered.
Finally, it is certainly possible to have several of these effects working together. Reductions in high-latitude temperatures associated with volcanic aerosols or diminished insolation, if they are sufficiently large, might have initiated changes in high-latitude deep-water formation, especially in the North Atlantic; however, such variations could simply arise naturally, in conjunction with random atmospheric forcing. The lower concentrations of CO2 and methane found prior to the current anthropogenically influenced levels must have affected the climate system, perhaps preconditioning high-latitude regions, or at least acting in concert with other cooling mechanisms. The search for "the cause" of decade-to-century-scale variability may ultimately uncover a system that responds to a variety of forcings and exhibits unforced variability. The climate record of the past several centuries may be difficult to decipher if we explore it with a strictly linear cause/effect approach.
Modeling studies done in conjunction with the ARRCC project have been supported by the National Science Foun-