Unraveling proxy noise and climate signal in paleotemperature records: a replicate study in the northwest Pacific

Geochemical proxies are used to reconstruct changes in sea surface temperature (SST) prior to the instrumental era. However, proxy records can be influenced by non-climatic factors, including instrumental errors and sediment heterogeneity, which introduce uncertainty and further complicate the interpretation. To decipher the climate signal and noise from proxy reconstructions, here we assess the replicability of two commonly used SST proxies, foraminiferal Mg/Ca and U^K’₃₇, by analyzing records from four nearby sediment cores in the northern Okinawa Trough collected from a radius of 10 km. The results show that all records of the same proxy type display consistent glacial-interglacial trends but differ in the degree of high-frequency variability across sites in Mg/Ca records. This variability cannot be reproduced among sites, thus may reflect uncertainties in instrumental analysis or sedimentary heterogeneity rather than actual SST changes. In addition, this variability contributes to differing inter-proxy deviations across sites, demonstrating that proxy uncertainty may influence comparisons between proxies. Despite this, averaging proxy records reveals a systematic offset between Mg/Ca and U^K’₃₇. One possibility is that the discrepancy arises from different seasonal productions, as supported by modern proxy observations and the closer alignment of U^K’₃₇ with model-derived annual mean temperatures compared to Mg/Ca. However, Mg/Ca is influenced by non-thermal factors which, if taken into account, can also resolve the aforementioned discrepancy between proxies. Overall, our findings indicate that the first-order glacial-interglacial patterns in paleotemperature records are reproducible among sites, but Mg/Ca records exhibit additional high-frequency variability that may reflect proxy noise. These results may be site-specific due to the core sites being located in the depocenter of a small basin, highlighting the need for replicate studies in diverse depositional settings. Data-model comparisons can refine interpretations of proxy records. We encourage models to account for sedimentation processes, enabling more precise quantification of the impact of sedimentary heterogeneity. This approach will improve the robustness of data-model comparisons and provide a more comprehensive framework for reconstructing past climate variability.