Accounting for habitat-depth and seasonality effects on the UK′37 temperature proxy: calibration and insights into model-data comparison

Past ocean temperatures are widely used to constrain Earth System Models under climate states that differ substantially from the historical period, like the Last Glacial Maximum (LGM). However, the reliability of model–data comparisons critically depends on the robustness of the temperature proxies and on our ability to correctly interpret the signals they record. The alkenone-based U^K’₃₇ index is among the most widely applied proxies for reconstructing past sea surface temperatures (SST), but significant deviations persist at both low and high temperatures of the modern calibration. These biases highlight unresolved uncertainties in the environmental and biological controls on alkenone production, as most U^K’₃₇ calibrations implicitly assume that coccolithophores record surface temperature uniformly and continuously, neglecting ecological variability in depth habitat and seasonality.

Here, we present a revised calibration approach that explicitly incorporates the spatiotemporal ecology of alkenone-producing coccolithophores. We combine new global biomass estimates derived from a machine-learning reconstruction^(a) based on the CASCADE dataset^(b) with simple growth relationships^(c) to estimate coccolithophore net primary productivity. These depth- and season-resolved coccolithophores productivity estimates are then used to weight the global temperature field.

Our results indicate that accounting for the impact of coccolithophore ecology on the U^K’₃₇ signal leads to systematically warmer reconstructed temperatures in cold regions and cooler ones in warm regions relative to classical SST-based calibrations. This produces a slightly reduced calibration slope relative to previous field calibrations, although in good agreement with culture work, and a pronounced latitudinal structure in bias relative to mean annual SST. Applying the same framework to the analysis of the IPSL-CM5A2 present and LGM climate simulations (which explicitly represents the nano-phytoplankton group using the PISCES module), demonstrates that these biological biases vary across climate states and that using our revised calibration reduces model-data mismatches. Overall, this approach highlights the value of coupling biogeochemical information with climate simulations to improve proxies interpretation and strengthen the paleoclimate constraints imposed on climate models.

 

(a). de Vries, J., Poulton, A. J., Young, J. R., Monteiro, F. M., Sheward, R. M., Johnson, R., Hagino, K., Ziveri, P., and Wolf, L. J.: CASCADE: Dataset of extant coccolithophore size, carbon content and global distribution, Scientific Data, 11, 920, https://doi.org/10.1038/s41597-024-03724-z, 2024.

(b). de Vries, J., Monteiro, F. M., Poulton, A. J., Wiseman, N. A., and Wolf, L. J.: A diverse community constitutes global coccolithophore calcium carbonate stocks, in review, 2025.

(c). Hopkins, J. and Balch, W. M.: A New Approach to Estimating Coccolithophore Calcification Rates from Space, Journal of Geophysical Research, 123, 1447–1459, https://doi.org/10.1002/2017JG004235, 2018.