Mai Winstrup – Niels Bohr Institute - University of Copenhagen

Forward this page to a friend Resize Print Bookmark and Share

Niels Bohr Institute > Research > PhD theses 2016 > 2011 > Mai Winstrup

 


Mai Winstrup

A thesis submitted october 23, 2011 for the degree of Doctor of Philosophy and defended November 21, 2011.

The PhD School of Science
Faculty of Science

Centre for Ice and Climate
Niels Bohr Institute
University of Copenhagen

Academic advisor:
Anders Svensson 

Co-advisor:
Sune Olander Rasmussen

Download Thesis >>

Abstract

An Automated Method for Annual Layer Counting in Ice Cores - and an application to visual stratigraphy data from the NGRIP ice core

An accurate chronology is of fundamental importance for the interpretation of a paleoclimatic record. The high temporal resolution of the Greenland ice cores has allowed the construction of an annual layer counted chronology for these reaching back to 60 ka BP, the oldest part of which is based on the NGRIP ice core. But as the annual layers become thinner towards the bed, the annual signal in most components weakens, and the subjectivity involved in manual layer interpretation increases. To extend the layer counted chronology beyond 60 ka BP, a more objective methodology of layer detection is needed.

For this purpose, an automated layer detection algorithm has been developed. It is based on the statistical framework of Hidden Markov Models (HMMs), originally developed for use in speech recognition. Meticulously based on statistical considerations, the algorithm is able to determine the most likely annual layering in an entire data section at once. The fundamental strength of the algorithm lies in the way that it is able to imitate the manual procedures, while being based on purely objective criteria for annual layer recognition.

The algorithm has been implemented for the visual stratigraphy data from NGRIP, in which the annual signal is covered in noise, but maintained to great depths. The algorithm is tested for three sections: A cold period (GS-13), a warm period (GI-12), and the transition between the two. The algorithm has not yet been tuned to provide an accurate chronology, but the results look promising. The algorithm was e.g. able to obtain a good result when passing over the transition period with a corresponding halving in annual layer thicknesses over merely five meters.

Download Thesis >>