| Description |
This dataset is derived from over 2 million vertical-component seismograms in the FDSN PKP database (1990–2024). Following preprocessing, an optimized deep learning model is employed for quality assessment and PKP precursor classification, enabling the separation of high-quality records with and without precursors, along with the computation of the corresponding PKIKP piercing point coordinates. The attachment has two CSV files: pre_merged_info.csv and nopre_merged_info.csv. Both have three columns: (1) an event-station ID in the format {origin_time}_{Mw}_{Station}.{Network}.{Channel}.{Location} (for example, ‘19980119182400_6.2_RES.CN.BHZ.’), (2) Longitude, and (3) Latitude. In pre_merged_info.csv, each event–station pair appears twice, one for the station and one for the event, because piercing coordinates at both sides have their own longitude and latitude. This file includes only pairs that contain PKP precursor phases. nopre_merged_info.csv uses the same structure but lists pairs with good-quality waveforms with no PKP precursors. |
