Unraveling multiple changes in complex climate time series using Bayesian inference

Online Access: Get full text
http://meetingorganizer.copernicus.org/EGU2016/EGU2016-11748.pdf
Author(s): Berner, Nadine; Trauth, Martin H.; Holschneider, Matthias
Author Affiliation(s): Primary:
Gesellschaft für Anlagen- und Reaktorsicherheit, Garching, Germany
Other:
Universität Potsdam, Germany
Volume Title: European Geosciences Union general assembly 2016
Source: Geophysical Research Abstracts, Vol.18; European Geosciences Union general assembly 2016, Vienna, Austria, April 17-22, 2016. Publisher: Copernicus GmbH on behalf of the European Geosciences Union (EGU), Katlenburg-Lindau, Germany. ISSN: 1029-7006
Note: In English
Summary: Change points in time series are perceived as heterogeneities in the statistical or dynamical characteristics of observations. Unraveling such transitions yields essential information for the understanding of the observed system. The precise detection and basic characterization of underlying changes is therefore of particular importance in environmental sciences. We present a kernel-based Bayesian inference approach to investigate direct as well as indirect climate observations for multiple generic transition events. In order to develop a diagnostic approach designed to capture a variety of natural processes, the basic statistical features of central tendency and dispersion are used to locally approximate a complex time series by a generic transition model. A Bayesian inversion approach is developed to robustly infer on the location and the generic patterns of such a transition. To systematically investigate time series for multiple changes occurring at different temporal scales, the Bayesian inversion is extended to a kernel-based inference approach. By introducing basic kernel measures, the kernel inference results are composed into a proxy probability to a posterior distribution of multiple transitions. Thus, based on a generic transition model a probability expression is derived that is capable to indicate multiple changes within a complex time series. We discuss the method's performance by investigating direct and indirect climate observations. The approach is applied to environmental time series (about 100 a), from the weather station in Tuscaloosa, Alabama, and confirms documented instrumentation changes. Moreover, the approach is used to investigate a set of complex terrigenous dust records from the ODP sites 659, 721/722 and 967 interpreted as climate indicators of the African region of the Plio-Pleistocene period (about 5 Ma). The detailed inference unravels multiple transitions underlying the indirect climate observations coinciding with established global climate events. [Copyright Author(s) 2016. CC Attribution 3.0 License: https://creativecommons.org/licenses/by/3.0/legalcode]
Year of Publication: 2016
Research Program: ODP Ocean Drilling Program
Key Words: 12 Stratigraphy, Historical Geology and Paleoecology; Alabama; Arabian Sea; Atlantic Ocean; Bayesian analysis; Cape Verde Rise; Case studies; Cenozoic; Climate; Climate change; East Mediterranean; Indian Ocean; Leg 108; Leg 117; Leg 160; Mediterranean Sea; Models; Neogene; North Atlantic; ODP Site 659; ODP Site 721; ODP Site 722; ODP Site 967; Ocean Drilling Program; Pleistocene; Pliocene; Probability; Quaternary; Statistical analysis; Tertiary; Time series analysis; Tuscaloosa Alabama; Tuscaloosa County Alabama; United States
Coordinates: N330100 N333700 W0870500 W0875000
N180437 N183438 W0210134 W0210135
N164038 N164039 E0595147 E0595146
N163718 N163719 E0594746 E0594745
N340411 N340411 E0324331 E0324331
Record ID: 2019065084
Copyright Information: GeoRef, Copyright 2019 American Geosciences Institute. Reference includes data from European Geosciences Union, Munich, Germany