Accurate temperature records for the deep geological past are a vital component of paleoclimate research. Distributional changes of branched glycerol dialkyl glycerol tetraether (brGDGT) lipids in geological archives including paleosoils are a promising indicators to infer past continental air temperatures. However, the 'orphan' status of the brGDGTs, the potential effect of temperature-independent parameters on their relative distribution, and the uneven geographical distribution of the soils used for calibration contribute to the high uncertainty of brGDGT-based transfer functions (root mean squared error, RMSE: ± 5 °C). Here, we expand the soil dataset from the previous calibration(s) with new and published soil data. We use Bayesian statistics to calibrate the relationship of the 5-methyl brGDGTs (MBT'5Me) and mean annual air temperature (MAAT). The addition of soils from warm (>28 °C) environments from India substantially increases the upper limit of the Bayesian calibration (BayMBT) from 25 to 29 °C, aiding in the generation of temperature records for past greenhouse climates, such as the Eocene. The BayMBT model also effectively minimizes the structured MAAT residuals prevalent in previous calibrations, therefore giving the opportunity to explore confounding factors on the calibration. We formulate a set of alternative calibration models to test the effect of specific environmental parameters and show that soils at mid-latitudes with temperature seasonalities >20 °C are not well described by the BayMBT model. We find that the MBT'5Me index is best correlated to the average temperature of all months >0 °C, called the BayMBT0 model. This finding supports the hypothesis that brGDGT production ceases or slows down in the winter months. However, a persistent feature of the BayMBT model and previous calibrations is the significant scatter at mid-latitudes, which is speculatively linked with a possible increase in diversity of microbial brGDGT-producing communities in these locations.
|Date made available||2019|