Applying a Change-Point Detection Method on Frequency-Magnitude Distributions

Abstract

A method based on nonparametric statistics (hereafter called "Median-Based Analysis of the Segment Slope,(MBASS)) is applied for the detection of change-points in Frequency Magnitude Distributions (FMD). The determination of the lowest magnitude for which the Gutenberg-Richter relation still applies is a key point for the computation of reliable b-values. The change-point detection method presented here is used to determine automatically this threshold magnitude (called m0 in this study) for large subsets extracted from four seismic catalogues. Results are successfully compared to those of a previous benchmark study from other authors. Moreover, the MBASS procedure is able to detect a magnitude artifact in the FMD of a regional catalogue. The results of the MBASS procedure confirm that a break in slope in a cumulative frequency distribution may be misleading when FMDs are analyzed by eye.