Turonian Subtropical Paleotemperatures from “Glassy” Foraminifera in East Africa and Their Implications for Modeling Greenhouse Climates

Specimens of “glassy” foraminifera from southern coastal Tanzania indicate sea surface temperatures up to 34°C for subtropical shelfal waters (assuming a seawater δ¹⁸O value of -1‰_SMOW) and a consistent planktic temperature gradient of 2-3°C through most of the Turonian. Using “glassy” preservation (preservation where there is no apparent test recrystallization at the submicron scale) as a criterion for evaluating δ¹⁸O results has greatly improved confidence in absolute paleotemperature estimates. However, data remain sparse for the Cretaceous. Further, existing data come largely from short intervals with unusual lithologic and faunal characteristics. These unusual characteristics raise the possibility that the results reflect temperatures or seawater compositions that were not typical for the time. Our new Tanzanian samples address these concerns. They contain a diverse assemblage of “glassy” planktic and benthic foraminifera preserved in otherwise unremarkable outer shelf claystones and silty claystones, and they span most of the Turonian (Whitinella archaeocretacea through Helvetoglobotruncana helvetica Zones). Additional samples from the Cenomanian, Coniacian, and Campanian are being processed.

Our calculated paleotemperatures are several degrees lower than some Cretaceous tropical estimates but are still ~6°C higher than summer seawater temperatures in comparable modern settings. Reproducing such temperatures in global climate models using Cretaceous boundary conditions requires atmospheric CO₂ concentrations of 2500 ppm or more (almost 10x pre-industrial CO₂ concentrations), a 10- to 20-fold increase in atmospheric methane combinations, or some combination of the two. These CO₂ levels are higher than most estimates for the Cretaceous atmosphere supporting the conclusion that sensitivity to greenhouse gas forcing, at least on long time scales, is too low in the current climate models.