Open AccessStatistical Classification of Weak Seismic Signals and Noise at the NORSAR Array
Author(s) -
Fyen Jan,
Husebye Eystein,
Christoffersson Anders
Publication year - 1975
Publication title -
geophysical journal of the royal astronomical society
Language(s) - English
Resource type - Journals
eISSN - 1365-246X
pISSN - 0016-8009
DOI - 10.1111/j.1365-246x.1975.tb05876.x
Subject(s) - wavelet , noise (video) , amplitude , test statistic , statistic , mathematics , statistics , signal (programming language) , sign (mathematics) , similarity (geometry) , statistical hypothesis testing , sign test , seismic array , signal to noise ratio (imaging) , acoustics , computer science , physics , mathematical analysis , artificial intelligence , optics , wilcoxon signed rank test , image (mathematics) , programming language , mann–whitney u test
Summary Dominant characteristics of P ‐waves recorded by NORSAR are signal similarity and an approximately lognormal amplitude variation across the array. This type of information, extracted from the subarray beam traces, is used in three different test statistics for statistical classification of weak seismic signals and noise wavelets at the NORSAR array. The events in question are wavelets which trigger the array's automatic signal detector with a signal‐to‐noise ratio so small that it is not obvious whether they represent true seismic signals or just correlated noise. One test statistic considered is the Sign‐bit Semblance test which checks on the signal similarity between subarrays. Another is a Binomial test which checks on the sign distribution of least square amplitude weights calculated for the subarray beam traces. The third test statistic is a Student's t ‐test which checks on the distribution and size of subarray beam amplitude weights. The results obtained indicate that the performance of these tests may result in an improvement of the array's event reporting performance of around 10 per cent.