Automated ice-core layer-counting with strong univariate signals

• Published in: Climate of the Past
• Main Authors: Wheatley, J. J., Blackwell, P. G., Abram, Nerilie, McConnell, J. R., Thomas, E. R., Wolff, E. W.
• Format: Article in Journal/Newspaper
• Language: unknown
• Published: Copernicus GmbH
• Subjects: Keywords: data set; geochronology; hydrogen peroxide; ice core; reconstruction; signal; sulfur; Antarctic Peninsula; Antarctica; West Antarctica; Antarctic; Greenland; Antarc*; Greenland ice core; Greenland Ice core Project; NGRIP; North Greenland; North Greenland Ice Core Project
• Online Access: http://hdl.handle.net/1885/69420; https://doi.org/10.5194/cp-8-1869-2012; https://openresearch-repository.anu.edu.au/bitstream/1885/69420/5/Automated_ice-core_Wheatley_Abram_etal_2012.pdf.jpg; https://openresearch-repository.anu.edu.au/bitstream/1885/69420/7/01_Wheatley_Automated_ice-core_2012.pdf.jpg

We present an automated process for determining the annual layer chronology of an ice-core with a strong annual signal, utilising the hydrogen peroxide record from an Antarctic Peninsula ice-core as a test signal on which to count annual cycles and explain the Methods. The signal is de-trended and normalised before being split into sections with a deterministic cycle count and those that need more attention. Possible reconstructions for the uncertain sections are determined which could be used as a visual aid for manual counting, and a simple method for assigning probability measures to each reconstruction is discussed. The robustness of this process is explored by applying it to versions of two different chemistry signals from the same stretch of the NGRIP (North Greenland Ice Core Project) ice-core, which shows more variation in annual layer thickness, with and without thinning to mimic poorer quality data. An adapted version of these Methods is applied to the more challenging non-sea-salt sulphur signal from the same Antarctic Peninsula core from which the hydrogen peroxide signal was taken. These Methods could readily be adapted for use on much longer datasets, thereby reducing manual effort and providing a robust automated layer-counting methodology.