DECONVOLUTION OF WHOLE-CORE MAGNETIC REMANENCE DATA BY ABIC MINIMIZATION

A new, objective deconvolution of whole-core magnetic remanence data is proposed on the basis of Bayesian statistics. The deconvolution is conducted as a smoothness-constrained least squares method with optimum smoothness determined by minimizing a Bayesian Information Criterion (ABIC). The deconvolution scheme was applied to whole-core magnetic remanence data of a sediment core measured at intervals of 5 mm. 5-mm-thick specimens were cut from the core to measure the individual magnetizations. These two results showed good agreement in each axis. Synthetic data with different magnitudes of Gaussian noise were also generated; the deconvolution showed more detailed structure with increased signal-to-noise ratio.