A Graph-Based Approach for Resolving Conflicts in Underwater Acoustic Event Association for Systematic and Improved Trilateration

Underwater passive acoustics enables the detection of a wide range of sounds from diverse sources, including seismic T-waves. To locate these events, hydrophone arrays are deployed, requiring arrivals to be identified across multiple synchronized sensors. While manual picking is reliable, it is labor-intensive, time-consuming, and prone to human bias. Automated picking using neural networks significantly increases the number of detected events but introduces challenges such as false positives. The next challenge is to associate detections of the same events to locate their source by trilateration.

The recently published TAPAAS algorithm (Raumer et al., 2025) generates candidate associations for these detections. However, among the produced candidates, there is a substantial number of overlapping and subsets of detections leading to different event locations. Relying solely on least-squares cost or RMSE is insufficient to address these conflicts. To overcome this, we developed a systematic approach combining relocation and a graph-based review of conflicting events, which we validated using synthetic event datasets.

When applied to real data from OHASISBIO and CTBTO arrays of hydrophones in the Indian Ocean, our method successfully built comprehensive catalogs of events. In addition, with a sufficient number of hydrophones (> 4), it is possible to check and improve the instrument synchronization. Generally, autonomous instruments are synchronized with a GNSS clock at deployment and upon recovery, allowing to measure the clock drift (~1-2s/yr); but when hydrophones run out of battery before recovery, the final time information is lost (i.e. drift unknonw). We also found that, sometimes, real-time hydrophone clocks are reset. Our approach allowed us to identify and determine these unknown clock drifts or clock offsets, then correct the arrival-times, and iteratively improve the trilateration.

Raumer et al. (2025). Geochem., Geophys., Geosyst. https://doi.org/10.1029/2025GC012572