The climate of the next few centuries is unlikely to be similar to any of the recent geological past, both in terms of forcing and response [ Crowley 1990; 1993; Overpeck et al., 1991; 1992; T. Webb et al., 1993]. For this reason, predictive models, rather than geological analogs, will have to be used to assess future climatic change. Unfortunately, the instrumental record of climatic variability does not include enough change to be of use in testing how well different models simulate the response to large changes in forcing. Furthermore, most predictive models are tuned to the present-day climate. The paleoclimatic record of change provides the only opportunity to determine how well state-of-the-art models simulate climate change and climatic conditions unlike the present day.
The last five years have seen impressive growth in the development of
paleoenvironmental frameworks for testing and improving predictive
models. Almost all of this effort has focused on the equilibrium responses
of models, and most studies have centered on understanding the global
sensitivity to a known radiative forcing, and/or the regional climatic
response to a known forcing. The current consensus is that more
paleoclimatic data need to be generated and quantified better before
significant paleoclimate-based reductions can be made in the uncertainty
associated with the current greenhouse global warming projection of 1.5
to 4.5 C [ Hansen and Lacis, 1990; Lorius et al., 1990;
Hoffert and
Covey, 1992; Houghton et al., 1992; Crowley, 1993]. However, these
paleoclimatic studies support the likelihood that climate system feedbacks
(e.g., related to water vapor, clouds, snow and ice cover) will amplify the
direct 4 Wm
radiative forcing due to an anthropogenic trace-gas
doubling, and will yield a global temperature increase of at least 1.5
C by the end of the next century [ Lorius et al., 1990;
Hoffert and Covey,
1992]. Without factoring in additional warming due to continued trace-gas
emissions, the Earth may be warmer by the end of the next century, than
at any time in the recent geological past.
The global array of paleoenvironmental data necessary for a complete analysis of how well predictive models simulate the regional response to altered climatic forcing is still years away, but many recent studies illustrate the power of this approach. Most comparisons between climate simulations and paleoclimatic data have focused on the last 20,000 years due to the abundance of well-dated (i.e., by radiocarbon) data for this period, and because the change in the Earth's boundary conditions are well known for this period relative to earlier ones. These late-Quaternary studies have laid the foundation for the Paleoclimate Modelling Intercomparison Project, and have demonstrated that the current generation of global general circulation models can simulate many subcontinental aspects of how the climate system responded to the complex array of insolation, sea-surface temperature, trace-gas, and ice changes since the last glacial period [ COHMAP, 1988; Prentice et al., 1991; Wright et al. 1993 ; Hostetler et al., 1994]. Comparisons between model simulations and empirical reconstructions of pre-Quaternary climate are complicated by greater uncertainty in terms of known climatic forcing (e.g., atmospheric trace gas concentrations) and reconstructed climate, but they still provide a large range of climate change scenarios useful for understanding climate sensitivity and model response [ Rind and Chandler, 1991; Rind, 1992; Sloan and Barron, 1992; Crowley, 1993; Barron et al., 1993]. Assessments of possible future change will be most useful if they are based on models that have a demonstrated ability to simulate past environmental change [ Overpeck et al., 1991].