Automated S-Wave Arrival Timing in GNSS Instantaneous Velocity Data Without Frequency-Domain Pre-Filtering

Seismic monitoring depends on accurately identifying P-wave and S-wave arrivals, which are critical for earthquake localization and Earthquake Early Warning (EEW). In EEW networks, a Global Navigation Satellite System (GNSS)-driven geodetic component enhances lead-time estimation and ground-shaking assessment, particularly for large earthquakes (Mw > 6.0). Advancing S-wave detection algorithms is essential to providing fast and reliable warnings to communities.

This study presents a sliding-window-based algorithm designed to detect the first time of arrival (ToA) of S-waves in high-rate (<1 Hz) GNSS instantaneous velocity time series without frequency-domain pre-filtering. The algorithm employs a three-phase process: (1) preprocessing, (2) statistical analysis and hypothesis testing for extracting ground-shaking disturbances, and (3) S-wave picking. It is implemented in an open-source Python-based toolbox, which also provides auxiliary seismic data, including ground-shaking duration, component-wise Peak Ground Velocity (PGV), and waveform energy.

The algorithm’s performance was evaluated using data from the 2016 Mw 6.2 Norcia and 2023 Mw 7.7 Kahramanmaraş-Gaziantep earthquakes. Results showed root mean square errors (RMSE) of 1.8 seconds and 3.8 seconds, respectively, when compared to ground-truth S-wave arrivals derived from P-wave readings on seismic waveforms recorded within 5 km of the GNSS sensors using the auxiliary Pphase-Picker software. The P-wave readings were extrapolated to GNSS sensor locations assuming equal P-wave velocities and a P-to-S-wave velocity ratio of 1.5.

Severe ground-shaking durations of up to 80 and 250 seconds, along with short S-P times of 3-5 seconds for the town of Norcia and the city of Gaziantep, highlight the severity of these events. This study demonstrates the new algorithm’s potential to enhance GNSS-based S-wave detection.