An assessment of biomarker-based multivariate classification methods versus the PIP25 index for paleo Arctic sea ice reconstruction

• Published in: Organic Geochemistry
• Main Authors: Koseoglu, Denizcan, Belt, Simon T., Husum, Katrine, Knies, Jochen
• Format: Article in Journal/Newspaper
• Language: English
• Published: Elsevier 2018
• Subjects: VDP::Mathematics and natural science: 400::Geosciences: 450::Hydrogeology: 467; VDP::Matematikk og Naturvitenskap: 400::Geofag: 450::Hydrogeologi: 467; VDP::Mathematics and natural science: 400::Geosciences: 450; VDP::Matematikk og Naturvitenskap: 400::Geofag: 450; Arctic; Barents Sea; Climate change; Paleo-Arctic; Phytoplankton; Sea ice
• Online Access: https://hdl.handle.net/10037/15039; https://doi.org/10.1016/j.orggeochem.2018.08.014

Accepted manuscript version, licensed CC BY-NC-ND 4.0. Source at: http://doi.org/10.1016/j.orggeochem.2018.08.014 The development of various combinative methods for Arctic sea ice reconstruction using the sympagic highly-branched isoprenoid (HBI) IP 25 in conjunction with pelagic biomarkers has often facilitated more detailed descriptions of sea ice conditions than using IP 25 alone. Here, we investigated the application of the Phytoplankton-IP 25 index (PIP 25 ) and a recently proposed Classification Tree (CT) model for describing temporal shifts in sea ice conditions to assess the consistency of both methods. Based on biomarker data from three downcore records from the Barents Sea spanning millennial timescales, we showcase apparent and potential limitations of both approaches, and provide recommendations for their identification or prevention. Both methods provided generally consistent outcomes and, within the studied cores, captured abrupt shifts in sea ice regimes, such as those evident during the Younger Dryas, as well as more gradual trends in sea ice conditions during the Holocene. The most significant discrepancies occurred during periods of highly unstable climate change, such as those characteristic of the Younger Dryas–Holocene transition. Such intervals of increased discrepancy were identifiable by significant changes of HBI distributions and correlations to values not observed in proximal surface sediments. We suggest that periods of highly36 fluctuating climate that are not represented in modern settings may hinder the performance and complementary application of PIP 25 and CT-based methods, and that data visualisation techniques should be employed to identify such occurrences in downcore records. Additionally, due to the reliance of both methods on biomarker distributions, we emphasise the importance of accurate and consistent biomarker quantification.