Toward a Better Approach to Data-Model Comparison

Abstract:

The quantitative census of a typical faunal assemblage recovered from a deep-sea core sample is the result of an integration of many different factors including but not limited to environmental (e.g., temperature, salinity, dissolved oxygen, nutrient availability, etc.), temporal (e.g., cycle of annual productivity), and pre- and post-depositional processes (e.g., dissolution, preservation potential, bioturbation, etc.). Extraction of a single environmental variable (e.g., temperature) requires poorly supported assumptions that introduce additional error and results in loss of valuable information contained within the assemblage. Conversely, temperatures generated by a numerical climate model for a given locale do not have any associated error. Thus, comparison of faunal assemblage-based temperature estimates with simulated temperatures from GCMs is an oversimplification of a complex system.

It is critical that like data sets are compared. For example, proxies for temperature may estimate conditions at the surface or at various depths or during particular seasons. These estimates are not directly comparable, and pooling Mg/Ca, alkenone and assemblage-based estimates of ocean temperature is inherently wrong. A more appropriate data-model comparison would involve comparing the faunal assemblage found in a core sample to one estimated by a paleoecological model driven by a GCM. An ideal comparison would score a suite of environmental factors gleaned from a faunal assemblage with those simulated by a GCM. Quantification of these factors, however, is not practical in some cases.

A paradigm shift in marine paleoclimate reconstruction is overdue, and comparison of paleoenvironments to model simulations requires a revised approach. The PRISM4 reconstruction presents a holistic and nuanced interpretation of multidimensional oceanographic processes and responses that is lost when reduced to a single variable such as temperature. Beyond the global approach, we incorporate regional climate dynamics with emphasis on processes, integrating multiple environmental proxies wherever available, to better characterize the environment. A more appropriate methodology is needed to adequately compare these robust and intricate paleoenvironmental reconstructions to those simulated by GCMs.