At present, We have developed a two-stage workflow of identifying and picking PmP waves in a semiautomatic way (Li et al., 2022). Briefly speaking, the two-stage workflow includes two parts: At the first stage, high-quality PmP waves are automatically picked on selected seismograms. At the same time, a visual check on three-component waveforms is conducted to confirm that the chosen signals indeed come from Moho reflection. At the second stage, the volume of the PmP dataset is expanded by involving other waves traveling along similar paths as those picked at the first stage. By utilizing the newly developed two-stage workflow, we have built the first PmP database with 10,192 PmP waves from the broadband vertical-component seismic data (from January 2000 to December 2018) retrieved from southern California Earthquake Data Center (SCEDC).
By utilizing the high-quality PmP dataset (10,192 manual picks) in southern California, we further develop PmPNet (Ding et al., 2022), a deep-neural-network-based algorithm to automatically identify PmP waves efficiently. PmPNet applies similar techniques in the machine learning community to address the unbalancement of PmP datasets. The trained optimal PmPNet can efficiently achieve high precision and high recall simultaneously to automatically identify PmP waves from a massive seismic database. Applying the pre-trained PmPNet to the seismic database from January 1990 to December 1999 in southern California, we obtain nearly twice more PmP picks than the original PmP dataset.
More detailed information on PmP database and our developed workflow/code to identify PmP phase can be found at PmPWorld documentation.
- PmP database built by two-stage workflow (Li et al. 2022).
- PmP database built by PmPNet (Ding et al. 2022).
- Codes for PmPNet and PmP-traveltime-Net (Ding et al. 2022).
- Example data for PmPNet and PmP-traveltime-Net.