General circulation models have been widely applied to the reconstruction of climate in the geologic past. The output of these models has usually been validated by comparison with paleoclimate indicators, such as coals, tillites, or evaporite deposits. However, these model/data comparisons are usually far from rigorous. Most comparisons have relied on subjective visual interpretations (i.e., comparing by ?eye?) for several reasons. First of all, model performance can vary spatially on both global and regional scales, so that climate predictions may be highly generalized. Second, paleoclimate indicators could be formed by processes not considered within general circulation models. For example, GCM?s do not include regional drainage, an important aspect of soil formation, and thus would not be able to accurately predict the spatial distribution of key soil types. Third, although the boundary conditions of climate change at multiple scales, each climate model run represents only a single climate state. In contrast, paleoclimate indicators are time averaged and thus can incorporate multiple states, including extreme conditions. Finally, the definition of what constitute a ?match? between model results and indicators is also problematic. Most paleoclimate indicators are defined in general terms, such as hot and seasonally wet, but such definitions are difficult to assess against the more precise climate models. Model/data comparisons have thus generally been qualitative rather than quantitative. New approaches are clearly needed to improve climate model/data comparisons.

We are addressing this issue using the distribution of evaporites as a test case. Evaporites are generally defined as forming in warm and dry regions that are seasonally wet. In reality, a wide variety of conditions can lead to evaporite formation. We are producing a quantitative calibration of the occurrence of recent evaporites against a climate model for modern boundary conditions. This calibration is then used to produce a predictive relationship between climate variables and evaporite formation. Models of ancient climates can then be used to predict evaporite occurrence in the geologic past. The predicted distributions are then compared with known occurrences to validate the approach.