Seismology Research
Automatic signal analysis
Reinoud Sleeman
The increase in number of seismographs, on global scale as well as in the Netherlands, and recent developments
in real-time data exchange (e.g. over Internet) has led to a huge growth of seismic waveform data available in
real-time. It requires efficient and accurate procedures and algorithms to handle the large amount of real-time
data and to process the data in order to pick seismic phases, to calculate hypocentral locations and to
estimate earthquake magnitudes.
Signal analysis research at the KNMI focuses on methods to automatically detect and identify seismic energy in
continuous recordings. Algorithms to detect seismic energy often uses amplitude information only, mostly within
a certain frequency range. Identification of seismic phases is often based on polarization and correlation
techniques. The combination of detecting and identifying is called a 'picker'. The KNMI developed and implemented
a 'picker' to automatically estimate the onset time of a P-phase. Automatic triggers of P-phases are provided
using an algorithm based on auto-regressive modelling of the P-phase. A detailed description of the method is
described by Sleeman and van Eck (1998). The present implementation uses real-time data from seismic station HGN
(Netherlands) and results are being visualized in our so-called 'Live Seismograms' at Internet.
Seismic registration of a P-wave (upper signal) and the P-picker (lower signal). The time that belongs to the
maximum in the P-picker is the (automatic) estimation of the p-wave arrival time.
Another approach to 'pick' the onset time of seismic energy is by using amplitude and frequency information
simultaneously. Along this line we have developed an algorithm for an automatic S-phase picker. Polarization
analysis of the estimated P-arrival (see above) is used for rotating the raw data into radial and transversal
components. We apply the wavelet transform on the rotated components and apply an adaptive noise-reducing filter
to select the frequency bands (scales) that contain S energy. This step reveals the type of seismic signal:
local, regional or teleseismic. To identify and pick the S-phase onset we apply a time-varying characteristic
function on the selected scales. This characteristic function is composed of the degree and direction of
polarization and the amount of transverse energy, and theoretically becomes 1 for S-phases
(Oonincx et al., 2001; Sleeman and van Eck, 2003).
Registration of a teleseismic event. The first 3 signals are the vertical and two horizontal components. The lower
signal is the S-picker. The picker is not sensitive for P-waves in the first part of the registration (between
0 en 4 minutes) and shows a maximum at the arrival time of the S-wave (t = 10 minutes).
References
- Oonincx,P., R. Sleeman & T. van Eck, 2001. An application of the DWT in Seismic Data Analysis. in: Wavelets in
Signal and Image Analysis, eds. A.Petrosian, F. Meyers. Kluwer, Dordrecht, 2001, 479-500
- Sleeman, R, T van Eck, 2003. Single station real-time P and S phase pickers for seismic observatories.
In: Methods and applications of signal processing in seismic network operations. Springer,
Lecture notes in Earth sciences, 98.
