4.4 Calibrating the clumped isotope-temperature relationship in cold (<30°C) carbonates
Our lab-grown A. islandica shells offer more control on formation temperature than naturally grown carbonates precipitated under variable temperatures. Ideally, the temperature of these natural samples is monitored so an average temperature can be calculated for the targeted growth period (e.g. Kele et al., 2015; de Winter et al., 2020; 2021b; Huyghe et al., 2021). However, formation temperatures are often indirectly estimated through other proxies (e.g. δ18Oc) and/or estimates of the living environment (e.g. water depth) of the carbonate producer, accumulating uncertainty (e.g. Peral et al., 2018; Piasecki et al., 2018; Meinicke et al., 2020). These caveats obscure the full uncertainty on the formation temperatures of natural carbonates as well as the effect of this unknown uncertainty on the calibrations. Considering the methods by which the “known” temperatures of natural carbonates are estimated in previous studies, it seems possible that the ~1.5°C temperature offset between Anderson et al. (2021) and Meinicke et al. (2020; 2021; see Fig. 3 ) and the 2.71 ± 2.03°C offset between Anderson et al. (2021) and our A. islandica data is partly caused by uncertainty on the formation temperatures of the calibration dataset. The similarity between Δ47 values of our data with Meinicke et al. (2021; ΔΔ47 0.004 ± 0.007‰) and the theoretical calcite temperature dependence by Guo et al. (2009; ΔΔ47 0.002 ± 0.007‰; Fig. 1 , Fig. 3 ;S8 ) shows that precise control on formation temperatures leads to tighter constraint on the clumped isotope temperature dependence of low-temperature samples. Precisely temperature-controlled carbonates thus better constrain the slope of the ∆47-\(\frac{1}{T^{2}}\) relationship for cold carbonates (improving calibration accuracy) while reducing the uncertainty on the calibration (improving calibration precision).
The ~1.5°C difference in reconstructed temperature between the calibrations in the low temperature range (<30°C) may seem trivial and requires the complete A. islandica dataset (N = 278; see Fig. 4 ) to resolve. However, in paleoclimate reconstructions (e.g. Petersen et al., 2016; de Winter et al., 2017; 2021a; Vickers et al., 2020b; Meckler et al., 2021; Agterhuis et al., 2021), this temperature offset may have significant consequences. A ~1.5°C cold bias in temperature reconstructions may lead to a significant underestimation of climate sensitivity to CO2 forcing, biasing the physical science basis for informing policymakers about future climate change (e.g. Dennis et al., 2013; Modestou et al., 2020; Westerhold et al., 2020; Tierney et al., 2020; IPCC, 2021). Accurate clumped isotope-based temperature reconstructions therefore require calibration datasets with precisely constrained formation temperatures tailored to the temperature range of the samples.