Picking of P and S waves is a fundamental process in seismology, and various kinds of picking techniques have been developed. Seismic waveforms change dramatically depending on the magnitude, the mechanism of the earthquake, and the positional relationship between the hypocenter and the seismic station.
The availability of various picking techniques is supposed to be helpful for appropriately dealing with a variety of seismic records. Hence, in addition to the revision of conventional techniques, the development of new picking techniques is worthwhile. In the present study, we developed a new stochastic technique to detect P and S waves based on the statistical amplitude distribution in the seismic record amplitude.
In the proposed method, the probabilistic density function (PDF) of the amplitude is calculated for each segment of seismic records, and the similarity between the PDF of the amplitude and that of the Rayleigh or Gaussian distribution is evaluated by divergence. Because Rayleigh and Gaussian distributions are typically found in amplitude distributions of highly random waves, such as coda waves, the divergence indicates the randomness of the seismic records. P and S waves are found by tracing the temporal change of the divergence.
We tested the proposed method using local seismic records for a series of seismic events that occurred before and after the 2016 Kumamoto earthquake. The mean absolute errors for picking P and S waves are and , respectively. The proposed method is a simple and new statistical picking method that enables automatic detection of P‐ and S‐wave arrivals.