Correcting a systematic bias in an ocean drilling project site 882 alkenone sea surface temperature record

Abstract

Reconstructions of sea surface temperature (SST) in the geologic record are fundamental to our understanding of Earth's climate history and the evaluation of Earth's climate sensitivity to greenhouse gas forcing. SSTs are reconstructed with a variety of methods, including alkenone biomarker lipids produced by certain coccolithophore algae. One such alkenone SST reconstruction from the subpolar northwest Pacific Ocean Drilling Program (ODP) Site 882 (50.21° N, 167.35° E, 3244 m water depth) has played a large role in shaping the paleoclimate science community's view of global climate warmth during the Late Pliocene (3.6–2.6 million years ago) and the subsequent cooling that characterized the intensification of Northern Hemisphere Glaciation (Haug, 1995; Haug et al., 2005; Martínez-Garcia et al., 2010). First, we have found that the values reported in the PANGAEA archive for this ODP Site 882 alkenone dataset were inaccurately reported as U K' 37 values when they are instead U K 37 ( https://doi.org/10.1594/PANGAEA.315092 , Haug and Sarnthein, 2005). This error in the archived data table resulted in the incorporation of inaccurate absolute SST estimates by several studies that applied U K' 37 calibrations to this ODP Site 882 dataset (e.g., Brennan et al., 2022; Clark et al., 2024, 2025; Tierney et al., 2019, 2025b). Second, using other published data from ODP Site 882 (Studer et al., 2012) and nearby Site 883 (51.11° N, 167.46° E, 2384 m water depth; Herbert et al., 2016; Novak et al., 2024), we show that the original Haug (1995) alkenone SST record at ODP Site 882 systematically reports an amplified range of absolute SST values compared to the more recently generated data. This observation is consistent with the known concentration-dependent biases of the gas chromatography chemical ionization mass spectrometry (GC-CI-MS) analytical method used by the original ODP Site 882 study (Chaler et al., 2000, 2003; Haug, 1995; Hefter, 2008; Rosell-Mele et al., 1995). These concentration-dependent analytical biases complicate applying a uniform correction to the entire Haug (1995) dataset. However, we are able to leverage the published datasets to propose a correction and quantification of uncertainty for a subset of the Haug (1995) dataset measured at similar on-column analyte abundance. For these samples, we find an average analytical uncertainty equivalent to ±2.05 °C , which is greater than and in addition to the typical ±1.4 °C 1 σ prediction uncertainty of the U K' 37 sea surface temperature proxy. We then discuss the implications of the corrected dataset for our understanding of late Neogene and Quaternary climate in the Kuroshio Extension region.