Correlation and Stacking of Relative Paleointensity and Oxygen Isotope Data
Abstract
The transformation of a depthseries into a timeseries is routinely implemented in the geological sciences. This transformation often involves correlation of a depthseries to an astronomically calibrated timeseries. Eyeball tiepoints with linear interpolation are still regularly used, although these have the disadvantages of being nonrepeatable and not based on firm correlation criteria. Two automated correlation methods are compared: the simulated annealing algorithm (Huybers and Wunsch, 2004) and the Match protocol (Lisiecki and Lisiecki, 2002). Simulated annealing seeks to minimize energy (crosscorrelation) as "temperature" is slowly decreased. The Match protocol divides records into intervals, applies penalty functions that constrain accumulation rates, and minimizes the sum of the squares of the differences between two series while maintaining the data sequence in each series. Paired relative paleointensity (RPI) and oxygen isotope records, such as those from IODP Site U1308 and/or reference stacks such as LR04 and PISO, are warped using known warping functions, and then the unwarped and warped timeseries are correlated to evaluate the efficiency of the correlation methods. Correlations are performed in tandem to simultaneously optimize RPI and oxygen isotope data. Noise spectra are introduced at differing levels to determine correlation efficiency as noise levels change. A third potential method, known as dynamic time warping, involves minimizing the sum of distances between correlated point pairs across the whole series. A "cost matrix" between the two series is analyzed to find a leastcost path through the matrix. This leastcost path is used to nonlinearly map the time/depth of one record onto the depth/time of another. Dynamic time warping can be expanded to more than two dimensions and used to stack multiple timeseries. This procedure can improve on arithmetic stacks, which often lose coherent highfrequency content during the stacking process.
 Publication:

AGU Fall Meeting Abstracts
 Pub Date:
 December 2012
 Bibcode:
 2012AGUFMGP12A..07L
 Keywords:

 0520 COMPUTATIONAL GEOPHYSICS / Data analysis: algorithms and implementation;
 1165 GEOCHRONOLOGY / Sedimentary geochronology;
 1520 GEOMAGNETISM AND PALEOMAGNETISM / Magnetostratigraphy;
 1522 GEOMAGNETISM AND PALEOMAGNETISM / Paleomagnetic secular variation