Discriminating Landslide Waveforms in Continuous Seismic Data Using Power Spectral Density Analysis

Discriminating landslides from other events in seismic records is challenging due to unclear phases and overlapped frequency content. We analyze the seismic waveform power spectral density (PSD) and its skewness to discriminate landslides from earthquakes and background noise. By comparing PSDs of landslides with small-magnitude earthquakes and noise in the Alaskan region, we find distinct power decay trends in the 0.01-5 Hz frequency range. The method was successfully tested on the seismic waveforms of seven global landslides. Further, the statistical significance of the approach was tested on 835 landslide waveforms using probability density, skewness and crosscorrelation of waveform PSD. This novel integration of seismic waveform PSDs and their skewness analysis is found to be robust and statistically significant for automatic landslide detection in continuous seismic data, with vast potential for early warning through real-time seismic networks.