ANTEVS: a quantitative varve sequence cross-correlation technique with examples from the Northeastern USA

Abstract

Varve correlation by hand was successfully applied by Ernst Antevs to establish the New England Varve Chronology, which has since been updated to form the North American Varve Chronology (NAVC). Although these methodologies are successful, numerical techniques can assist in finding and evaluating correlations. A quantitative numerical method for varve correlation using time-series Fourier analysis and cross-correlation is proposed and implemented in the computer program ANTEVS (Automatic Numerical Time-series Evaluation of Varve Sequences). The technique is demonstrated by correlating several varve sequences in the northeastern USA. Tests on NAVC data from the Hudson and Connecticut River Valleys show strong positive local and regional cross-correlations, confirming the method's validity. Guidelines for the evaluation of the correlation are determined by cross-testing NAVC sequences, suggesting minimum values for the cross-correlation statistic r, and z-score, a measure of its variation. Field relationships and careful examination of the data graphs and correlograms, however, must accompany numerical analysis. We then apply the method to previously uncorrelated sequences. A Champlain Valley varve sequence at Whallonsburg, NY, is compared with the NAVC, and to another Champlain Valley sequence at Keeseville, NY. No match is found with the NAVC, although none was expected as the sequences are of slightly different ages. A weak correlation is found between the two Champlain Valley sequences. This correlation is not significant and disagrees with the stratigraphic interpretation of the sites. We suggest that an overly strong local sedimentary signal at one of the sites masks the regional signal necessary for positive cross-correlation.

Keywords::

• varve chronology
• Fourier analysis
• time-series
• glacial
• Laurentide
• ice retreat

Acknowledgements

We would like to acknowledge and thank David Barclay (SUNY Cortland) for discussions on statistical methods of tree-ring analysis. We also thank the SUNY New Paltz students who helped test and improve ANTEVS: Alexis Wright, Christopher Callinan, Aileen Force and Mary Walsh. We especially thank Jack Ridge for his insightful advice and assistance. Thoughtful reviews of this manuscript were provided by Jack Ridge, David Franzi and Andrew Brekenridge. Comments and suggestions from Thomas Lowell are also greatly appreciated.