Cluster analysis for time series based on organic geochemical proxies. (July 2020)

• Record Type: Journal Article
• Authors: Dauner, Ana Lúcia L.; Mollenhauer, Gesine; Bícego, Márcia C.; Martins, César C.
• Is Part Of: Organic geochemistry. Volume 145(2020)
• Publication Date: 2020-07
• Subjects: Statistical analysis -- Temporal analysis -- Sediment cores -- SAX representation -- Constrained cluster -- Paleoceanography -- Organic matter; Organic geochemistry -- Periodicals; Biogeochemistry -- Periodicals; Géochimie organique -- Périodiques; 553.205
• DOI: 10.1016/j.orggeochem.2020.104038
• Languages: English
• ISSNs: 0146-6380

Abstract

Highlights: Cluster analyses were applied to time-series data of organic geochemical proxies. Symbolic Aggregate approXimation cluster identified proxies with similar sources. Constrained clusters identified temporal zones for each organic matter source. This approach simplifies interpretation of large data sets with multiple compounds. Abstract: Depositional and paleoenvironmental studies using organic geochemical proxies often present the temporal evolution of several compounds. Despite the importance of using several proxies to understand how the surrounding environment changed through time, this large amount of data usually hampers interpretations. In this scenario, the use of statistical tools for time series analysis can help simplify and interpret large data sets, even if they were not initially developed for molecular marker data. In this study, we show the benefits of using two different cluster analyses in order to: (i) group compounds with similar sources; and (ii) identify temporal zones. Cluster analysis using SAX (Symbolic Aggregate approXimation) representation groups together different proxies with similar sources (whether anthropogenic or natural, autochthonous or allochthonous), based on their temporal evolution. Temporal zones, on the other hand, can be identified by using a constrained cluster analysis, in which samples (sediment layers) are grouped according to the temporal variability of the organic compounds. These two approaches were successfully applied to organic proxy datasets from two sediment cores, retrieved from distinct environments and with distinct temporal recoveries. Based on these analyses, we were able to identify the probable source of compounds with multiple sources, and to show how the terrestrial and marine organic matter presented distinct patterns over time. These techniques do not replace the study of the temporal evolution of compounds individually but synthesize a large amount of information and may indicate which compounds of an assemblage yield the most robust information in environmental studies.