Ocean bottom seismometer clock correction using ambient seismic noise

Ocean-bottom seismometers (OBSs) are equipped with seismic sensors that record acoustic and seismic events at the seafloor. One critical parameter for obtaining accurate earthquake locations is the absolute time of the recorded seismic signals. It is, however, not possible to synchronize the internal clocks of the OBSs with a known reference time, as GNSS signals do not reach the sea bottom. We address this issue by introducing a new method to synchronize the clocks of large-scale OBS deployments. Similar to some previous approaches, our method leverages the theoretical time-symmetry of time-averaged cross-correlations of ambient seismic noise: broken time-symmetry is attributed to clock drift. A non-uniform surface wave illumination pattern, however, can also break the time-symmetry. Existing noise-based synchronization techniques usually ignore the latter, but we do address it by means of a weighted least-squares inversion (based on station-to-station distances). The weighted least-squares inversion mitigates the adverse effect of a nonuniform surface wave illumination on the time-symmetry. Furthermore, our method includes a unique feature: it estimates and corrects for an initial clock error introduced at the deployment time. This initial clock error can be attributed to either (i) a wrong initial time synchronization or (ii) the temperature shock during deployment. The methodology is implemented in an open-source Python package named OCloC and was tested with OBS recordings acquired around the Reykjanes peninsula, southwest Iceland. Our results indicate that all OBSs experienced a clock drift, and that a significant number of them were subject to an initial clock error at the deployment time. This study provides a substantial improvement in the inherent quality of OBS data, laying a solid foundation for more robust seismic data analysis