Abstract
Multi-proxy investigations on geological archives provide valuable information about environmental variations in the past. As opposed to single-proxy studies, the combination of several proxies can reveal more detailed information and strengthen subsequent paleoenvironmental reconstructions. However, there is still no consensus about how to deal with resulting highly dimensional datasets in a statistical manner. In many cases, the interpretations of multi-proxy datasets rely on visually matching several proxy records, which can lead to incorrect or insufficient interpretations. Here we report an innovative approach that combines the novel dimension reduction technique Uniform Manifold Approximation and Projection (UMAP) and the time series analysis R package asdetect to identify and characterize Holocene environmental phases and phase boundaries in a sediment core from Lake Chatyr Kol, southern Kyrgyzstan. Despite the fact that the Holocene climate evolution of Central Asia has been intensively studied during the last decades, knowledge about regional climate development during the Holocene and the underlying mechanisms is still relatively scarce. We particularly focus on phase transitions and differentiate between event-based shifts as opposed to gradual phase transitions. For this study, long-chain alkenones were used as a paleotemperature proxy and variations in long-chain alkyl diol distributions were ascribed to relative changes of algal input. The compound-specific stable hydrogen isotope compositions (δD) of individual n-alkanes were utilized as paleohydrological proxies, with the δD of mid-chain n-alkanes reflecting changes in the δD of the lake water and the δD of long-chain n-alkanes recording the δD of the meteoric water. We show the potential of modern analysis tools for data-driven paleoenvironmental reconstructions and advocate for their more frequent implementation in multi-proxy studies.
Highlights
Paleoenvironmental reconstructions are essential to elucidate the Earth’s natural environmental variability in the past
Biomarkers derive from distinct biotic sources and their abundance and composition is diagnostic for environmental parameters of their formation (Meyers, 2003)
Principal Component Analysis (PCA) is a multivariate statistical analysis that utilizes matrix factorization for projecting high-dimensional data on a lower dimensional subspace comprized of the so-called principal components, while retaining the maximum variability present in the original variables (Ringnér, 2008; Abdi and Williams, 2010; Naik, 2018)
Summary
Paleoenvironmental reconstructions are essential to elucidate the Earth’s natural environmental variability in the past. In a previous study we showed the effectiveness of applying UMAP to biomolecular data obtained from a lake sediment core, allowing to clearly divide the record into distinct phases in a data-driven manner (Schroeter et al, 2020) It is unclear whether phase boundaries ascertained by UMAP reflect true abrupt shifts in time series, often classified as tipping points (Lenton et al, 2008; Lenton, 2011; Scheffer et al, 2012), or if these changes can occur gradually within a determined phase. We utilized a multiproxy dataset of terrestrial and aquatic biomarkers obtained from a sediment core from Lake Chatyr Kol, Kyrgyzstan, Central Asia, reflecting hydrological, and environmental changes of the lake and its catchment area, respectively. Since the lake is not inhabited by fish, a considerable number of the amphipod species Gammarus alius sp. nov., which is mostly confined to continental freshwater/brackish habitats, have colonized aquatic plants up to a depth of 50 cm (Sidorov, 2012)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
