On the reconstruction of past levels of atmospheric carbon dioxide

Right now, with a concentration of 30% above background, the carbon dioxide content of the atmosphere is higher than anything experienced in the last 800,000 years. According to the "best-guess" reconstructions of Berner (1997, 1998), we have to go back tens of millions of years to reach values greater than 50% over present, and to the mid-Cretaceous to go beyond a factor of two. However, the error bars are very large indeed. According to new reconstructions based on carbon isotopes in unsaturated alkenones from deepsea carbonates, atmospheric CO(2) was in fact lower in the Miocene than today (Ennyu et al., 1998; Pagani et al., 1998), rather than higher, as is generally thought. If this is accepted, Neogene cooling must be derived from orographic and other geographic changes affecting albedo and redistribution of heat, rather than a drawing down of CO(2) from intensified weathering or burial of organic matter. The best documented past values of CO(2) are those derived from Antarctic ice, and the best-documented period of rapid change in these values is the time of deglaciation. The glacial to postglacial rise in temperature that may be ascribed to carbon dioxide is between 1 and 2 degrees C. A strong increase of alkalinity occurs within the ocean during deglaciation, which is counterintuitive. This increase (which produces the degiacial preservation spike) points to transfer of alkalinity from shallow to deep water in a stratified ocean (e.g., through coccolith blooms and shallow-water carbonate precipitation cum redeposition), and to nitrate and sulfate reduction at depth (from lack of ventilation). The carbon dioxide curve from ice cores can be extrapolated backward using regression on delta(18)O, with r(2) near 0.9, provided the equation has the form [CO(2) = f(x(2), x'(2))] where x is a transform of delta(18)O ranging between 0 and 1, and x' is its derivative. The ice core data, and perhaps the regression-based extrapolations for the last several 100 ky, provide a check on the quality of reconstructions using carbon dioxide proxies. The fact that such proxies need to be corrected for other parameters (e.g., temperature) introduces considerable uncertainty.