# A better automatic body-wave picker with broad applicability

TL;DR: Using the component energy correlation method, the mathematical and practical advantages of the correlation operator are demonstrated and this operator is applied to the S¯T/L¯T and R¯P/ L¯P methods to reduce the dependence on user-defined time-scale parameters.

Abstract: For robust earthquake analysis, we need efficient and reliable automatic body-wave recognition methods. To do this, we combine the advantages of standard methods in an innovative and generalized approach. Using the component energy correlation method, we demonstrate the mathematical and practical advantages of the correlation operator and apply this operator to the S¯T/L¯T and R¯P/L¯P methods. We also implement multi-scale versions of these methods to reduce the dependence on user-defined time-scale parameters. We compare our results systematically to different methods, propose an optimal approach and demonstrate its reliability.

## Summary (2 min read)

### INTRODUCTION

- Whether analyzing global tectonic events or micro-earthquakes (natural or from human activity), the authors require estimates of bodywave arrival times from seismic records.
- The ratio of two running means (one short and one long time scale) increases with amplitude changes.
- In addition to amplitude changes, body-wave arrivals are also associated with frequency and polarization changes that may not be detected by ¯ST/L̄T .
- These methods aim to improve arrival time data by increasing their time precision and the number of detected arrivals, using pre-existing and potentially incomplete arrival time datasets from uninformed methods as prior information.

### METHOD

- Component energy correlation method P-waves produce displacements primarily along the propagation direction, while S-waves produce displacements primarily perpendicular to the propagation direction.
- Seismic noise by contrast, is a stochastic signal (Wentzell, 1981) and as such, exists on all polarizations.
- The correlation window (TC) defines the dominant period of the P-waves with best sensitivity (Nagano et al., 1989).

### Generalized correlation method

- In these C f , the ratio operator could be substituted by the correlation operator with two advantages: 1) the correlation outputs normalized values, whereas the R̄P/L̄P method gives values dependent on the input data, 2) correlation is a robust approximation of deconvolution.
- The correlation produces a finite, normalized probability; allowing us to combine several correlation-based C f for improved sensitivity.

### Multi-scaling

- The signal is first decomposed using high pass filters which are then used for running mean and rms calculations.
- The authors take advantage of the running mean filter capabilities in the time domain for reducing low frequency signals while retaining a sharp step response (Smith, 1997).
- Using correlation and multi-scaling, the authors can combine methods and frequency bands for a robust general detector of bodywave arrivals.
- The threshold over which a body-wave arrival might be considered is the only required parameter.

### Artificial tests

- The authors use what they call artificial data (because it does not aim to simulate a realistic Earth response) to study individually three properties of body-waves (changes in frequency, amplitude, and polarization) in the presence of seismic noise.
- The authors second key result shows that correlation-based methods have the best sensitivity to pure frequency changes.
- The response of ¯ST/L̄T and R̄P/L̄P, shown by test A in Figure 1 to all three properties (described in the caption) have similar precision and sensitivity.
- Using test B and C, the authors compare the ratio and the correlation methods using rms (C frms).

### Data test

- The use of real data allows visual comparison of manual picking data with automatic methods.
- The consistency between the onsets of tested C f and the manual P-wave arrival picks is significant through the entire distance range .
- The data noise level is the most important factor affecting the results, but it is not explained by body wave attenuation as function of distance.

### DISCUSSION

- The methods based on rms calculations, polarization changes, and correlation showed the best sensitivities to body-wave arrivals and they can all be implemented in a multiscale approach for application to a wide range of distances and magnitudes.
- The worst responses are obtained with noisy data independent of distance.
- This method is based on recursive algorithms and does not require high-order statistics.
- Time derivative and Kurtosis have been shown as reliable onset indicators for seismic wave detection (Baer and Kradolfer, 1987; Lomax et al., 2012); their maximum in the vicinity of the probability peak should give best estimates of arrival times.

### CONCLUSION

- The authors obtained a body-wave recognition tool with broad applicability from industrial surveys to earthquake monitoring.
- The authors defined an optimal calculation scheme and the related threshold that can then be used for event detection with their approach.
- The authors discussed its limitations, which are related to data quality and to the need for 3-dimensional seismic records.

Did you find this useful? Give us your feedback

...read more

##### Citations

26 citations

2 citations

2 citations

### Cites background from "A better automatic body-wave picker..."

...One way to do this is to use characteristic functions (see Massin and Malcolm [2016] and references therein)....

[...]

2 citations

### Cites methods from "A better automatic body-wave picker..."

...Massin [15] used the component energy correlation method to identify microseismic body waves....

[...]

##### References

413 citations

### "A better automatic body-wave picker..." refers methods in this paper

...They often rely on variants of the Short Term - Long Term averages ratio algorithm ( ̄ ST/L̄T is used in Earthworm and SeisComP3; Allen, 1982; Baer and Kradolfer, 1987; Lomax et al., 2012)....

[...]

...The ¯ST/L̄T uses the averaged absolute value of the signal (Allen, 1982) or its envelope (Baer and Kradolfer, 1987)....

[...]

...uses the averaged absolute value of the signal (Allen, 1982) or its envelope (Baer and Kradolfer, 1987)....

[...]

...They often rely on variants of the Short Term - Long Term averages ratio algorithm ( ¯ST/L̄T is used in Earthworm and SeisComP3; Allen, 1982; Baer and Kradolfer, 1987; Lomax et al., 2012)....

[...]

345 citations

342 citations

### "A better automatic body-wave picker..." refers background or methods in this paper

...The ¯ST/L̄T uses the averaged absolute value of the signal (Allen, 1982) or its envelope (Baer and Kradolfer, 1987)....

[...]

...Time derivative and Kurtosis have been shown as reliable onset indicators for seismic wave detection (Baer and Kradolfer, 1987; Lomax et al., 2012); their maximum in the vicinity of the probability peak should give best estimates of arrival times....

[...]

...They often rely on variants of the Short Term - Long Term averages ratio algorithm ( ̄ ST/L̄T is used in Earthworm and SeisComP3; Allen, 1982; Baer and Kradolfer, 1987; Lomax et al., 2012)....

[...]

...uses the averaged absolute value of the signal (Allen, 1982) or its envelope (Baer and Kradolfer, 1987)....

[...]

...They often rely on variants of the Short Term - Long Term averages ratio algorithm ( ¯ST/L̄T is used in Earthworm and SeisComP3; Allen, 1982; Baer and Kradolfer, 1987; Lomax et al., 2012)....

[...]

147 citations

### "A better automatic body-wave picker..." refers methods in this paper

...Significant work has been done with informed methods (using autoregression, Wenner and match filters, see Song et al., 2010,for a application to hydrofracture induced microseismic) and open-source standard libraries (obspy, Megies et al., 2011) are now available to combine their advantages....

[...]

125 citations

### "A better automatic body-wave picker..." refers background in this paper

...Seismic noise by contrast, is a stochastic signal (Wentzell, 1981) and as such, exists on all polarizations....

[...]