Automatic detection and rapid determination of earthquake magnitude by wavelet multiscale analysis of the primary arrival

B. Dando, F. J. Simons
University College London

Richard M. Allen
University of California Berkeley

AGU Fall Meeting 2006
Eos Trans. AGU, 87(52), Fall Meet. Suppl., Abstract S11B-02

Earthquake early warning systems save lives. It is of great importance that networked systems of seismometers be equipped with reliable tools to make rapid determinations of earthquake magnitude in the few to tens of seconds before the damaging ground motion occurs. A new fully automated algorithm based on the discrete wavelet transform detects as well as analyzes the incoming first arrival with unmatched accuracy and precision, estimating the final magnitude to within a single unit from the first few seconds of the P wave. The curious observation that such brief segments of the seismogram may contain information about the final magnitude even of very large earthquakes, which occur on faults that may rupture over tens of seconds, is central to a debate in the seismological community which we hope to stimulate but cannot attempt to address within the scope of this paper. Wavelet coefficients of the seismogram can be determined extremely rapidly and efficiently by the fast lifting wavelet transform. Extracting amplitudes at individual scales is a very simple procedure, involving a mere handful of lines of computer code. Scale-dependent thresholded amplitudes derived from the wavelet transform of the first 3--4 seconds of an incoming seismic P arrival are predictive of earthquake magnitude, with errors of one magnitude unit for seismograms recorded up to 150 km away from the earthquake source. Our procedure is a simple yet extremely efficient tool for implementation on low-power recording stations. It provides an accurate and precise method of autonomously detecting the incoming P wave and predicting the magnitude of the source from the scale-dependent character of its amplitude well before the arrival of damaging ground motion. Provided a dense array of networked seismometers exists, our procedure should become the tool of choice for earthquake early warning systems worldwide.

© Richard M Allen