Identification of phase-timing, magnitude estimation and earthquake detection in New Zealand using neural networks and novel wavelets and precursory signal nonlinear analysis and prediction

Abstract

New Zealand is abundant in earthquake activity. It is a region of plate convergence, with subduction of the Pacific Plate in the North Island and of the Australian Plate in the South Island.  The motion between these two plates is the main reason for the occurrence of about 14000 recorded earthquakes each year.  The high level of seismic activity in New Zealand and the limitation of detection methods employed are our motivation and need respectively to develop advanced signal processing techniques to process the New Zealand seismic data records.

There are two main aims to this research. The first aim of this research is to improve the detection and phase picking performance. Most of the current systems for detection and phase picking are based on classic methods such as short time average / long time average (STA/LTA). The issue with these systems is the dependency of their performance with respect to the signal to noise ratio (SNR). These systems do not perform well in low SNR conditions (i.e. the system may not detect the presence of the phases in cases of weak or distant earthquakes) as well as at high noise levels. Furthermore, these systems may produce false alarms when they process a noise burst originated from either natural or cultural sources. To enhance the performance in such cases, conventional filtering methods are usually applied in these systems. However, such filtering methods are not optimal since they do not consider the non-stationary nature of the seismic noise. To enhance the detection reliability and therefore the accuracy of magnitude and location estimations advanced signal processing strategies will be used to de-noise the seismic signals. Higher accuracies will be obtained by designing a novel robust algorithm for detection and phase picking. Such implementation should serve to reduce the false alarm rates as well as to detect the missed events in low SNR cases.

The second aim of this research is to explore the precursory signals that exist before an event. It is intended to search of hidden markers to predict an event. Recent studies have shown the presence of some precursory signals before the P waves in some earthquakes. Although the existence of such signals has been supported by some theoretical analysis, little investigation has been performed in this context. To identify the existence of dynamical behaviour in the precursory signals, non-linear signal analysis will be initially used. If it is found that dynamics exist in the data, wavelet techniques and further non-linear methods will be then applied to identify the hidden markers that pertain to the recognized dynamics. Such hidden markers will be then used to predict the earthquake events.

References:

Rastin, S.J.; Unsworth, C.P.; Gledhill, K.R.; Coghill, G.G.; Chadwick, M.; Robinson, R.;, "Comparison of Simplified Earthquake Event Waveforms derived from Mexican Hat and Haar Wavelets in Matata, New Zealand", 2009 AOTULE Postgraduate Student Conference, National Taiwan University, Taiwan, 2-3 December 2009.

Rastin, S.J.; Unsworth, C.P.; Gledhill, K.R.; Coghill, G.G.; Chadwick, M.; Robinson, R.; , "Iterative coupling of standardised earthquake detection & wavelet thresholding to determine simplified earthquake event waveforms (SEEW) ", Information Sciences Signal Processing and their Applications (ISSPA), 2010 10th International Conference on , pp.173-176, 10-13 May 2010,doi: 10.1109/ISSPA.2010.5605547

Rastin, S.J.; Unsworth, C.P. and Bennet, L.; , "Fast Automatic Analysis of Antenatal Dexamethasone on Micro-seizure Activity in the EEG", Engineering and Physical Sciences in Medicine and Australian Biomedical Engineering Conference (EPSM-ABEC) 2010, Melbourne, 5-9 Dec 2010.

Rastin S, “PhD provisional report : Identification of Phase-Timing, Magnitude Estimation and Earthquake Detection in New Zealand using Neural Networks & Novel Wavelets & Precursory Signal Nonlinear Analysis & Prediction” , Dept. of Eng. Sci. University of Auckland,24 pages, June 2009