U005-04
Semi-automatic, autocorrelation-based varve counting and sediment chronology: the open source countMYvarves toolbox
Tuesday, 8 December 2020: 16:14
Maximillian S. Van Wyk de Vries, University of Minnesota, Department of Earth and Environmental Sciences and Limnological Research Center, Minneapolis, MN, United States; University of Minnesota, Saint Anthony Falls Laboratory, Minneapolis, MN, United States, Emi Ito, University of Minnesota Twin Cities, Earth and Environmental Sciences and Limnological Research Center, Minneapolis, MN, United States, Mark Shapley, University of Minnesota, Earth and Environmental Sciences, and the Continental Scientific Drilling and Coring facility, Minneapolis, MN, United States and Guido Brignone, Universidad Nacional de Córdoba, Centro de Investigaciones en Ciencias de la Tierra (CICTERRA-CONICET), Córdoba, Argentina
Abstract:
Determining the chronology of sedimentary sequences is crucial to the study of paleoclimate, past volcanism, cryospheric change and more. Annual periodicity sediment layers, known as varves, can provide continuous and high-resolution chronologies of sedimentary sequences. Varve counting has a more than 100 year history and is not burdened with the high laboratory costs of many geochronological analyses. However, manual varve counts are time consuming, error prone, and can be subjective or challenging to reproduce. Several partially and fully automated numerical varve counting tools have been created to overcome these limitations. These are based on two main strategies: i) counting peaks or inflection points along a transect of digital varve image colour-intensity, and ii) machine learning techniques in which the computer is trained to identify varve boundaries. However, existing tools present limitations for semi-automated counting of complex or very fine scale varves.
Here we present a novel varve counting methodology, based on the use of sliding-window image autocorrelation to count repeated patterns in core scans or outcrop photos. Image autocorrelation can efficiently detect even complex or disrupted periodic sediment patterns and is insensitive to the scale of varves. Unlike machine learning approaches, it does not require an external training dataset. The scale or repeated patterns is used to build an annually resolved record of sedimentation rates, which are depth-integrated to provide ages. We compare model outputs to four manual counts of a high sedimentation rate lake with biogenic varves (Herd Lake, USA), and a low sedimentation rate glacial lake (Lago Argentino, Argentina: Ito et al., EP010, Van Wyk de Vries et al., EP010). In both cases the autocorrelation-based varve chronology is consistent with manual counts, is fully reproducible, and provides a better record of uncertainties. This varve counting toolbox, named countMYvarves, is open source and can be run directly from a graphical user interface.
