A nonlinear model for resolving the temperature bias of branched glycerol dialkyl glycerol tetraether (brGDGT) temperature proxies

Abstract

Branched glycerol dialkyl glycerol tetraethers (brGDGTs) are a class of ubiquitous, bacteria-derived lipid biomarkers in terrestrial and aquatic environments. Many studies have demonstrated the potential for brGDGTs as paelotemperature proxies, but the observed seasonal bias in brGDGT-inferred temperatures still remains poorly understood, particularly for Arctic or seasonally dry climates. Here we introduce a new physical framework for understanding variations in the methylation index of branched tetraethers (MBT) by explicitly modeling the production and preservation of brGDGTs for a bacterial population with an exponential rate dependence on temperature. The dependence of MBT ratios on temperature is predicted to be nonlinear, and thus has a different form from other empirically defined MBT models. We apply the model to understanding how the methylation index depends on not only the mean annual air temperature (MAAT) but also how it depends on the amplitude of the seasonal temperature cycle and the phase and amplitude of the soil moisture content. We performed mesocosm growth experiments using natural lake waters that confirm that the model correctly accounts for lower brGDGT production rates during cold seasonal temperatures. Comparing the new model predictions with global compilations of MBT’5Me and MBT’ data, we determine a new calibration of best fitting set of model coefficients that more accurately account for the expected physical constraints on MBT-type data. Our new results account naturally for the expected saturation of MBT’5Me and MBT’ at low and high temperatures, a bias of up to 5–7 °C in MAAT for high seasonal temperature fluctuations, and an additional bias of up to 1–3 °C in MAAT for strongly seasonal soil moisture fluctuations. Taken together, the 3 effects explain differences of up to 15 °C in inferred MAAT compared with traditional empirical models for MBT’5Me and MBT’ and results in improved mean squared errors for both. At MAAT above 18 °C, MBT’ temperature errors are lower than those of MBT’5Me, suggesting MBT’ may still be a useful proxy in some situations despite its added complexity. While we apply the framework only to global MBT’5Me and MBT’ proxy data, the physical framework can be adapted to other types of paleoproxies and may therefore be more widely applicable. Preliminary application of our model to the Pliocene North Sea Hank Core yields ∼1 °C cooler temperatures than a previous calibration and application to 130 kyr-long Chinese Loess Plateau records yields ∼4 °C warmer glacial and glacial stadial temperatures compared to previous calibration.