Clock skew corrections of a large-aperture OBS array in the Atlantic Ocean reveal the need to include non-linear drift corrections

Accurate timing corrections for seismic data recorded by ocean bottom seismometers (OBS) are essential for a wide range of applications. The synchronization of internal OBS clocks with Global Positioning System (GPS) is only possible prior and subsequently to deployment in the seafloor. Thus, untracked, possibly nonlinear clock errors in seismic data may accumulate over the deployment period. The measurement of the clocks offset from GPS at retrieval, referred to as `skew’, can be used to correct the data solely under the assumption of a uniform rate of clock drift throughout the whole deployment. We estimate OBS clock drift curves for 40 OBSs of the large-scale UPFLOW amphibian array in the Madeira-Azores-Canaries region that lasted ~14 months in 2021-22. We use the relative shift of daily Empirical Green Functions obtained from seismic ambient noise recorded by all available data channels to track clock error. We find that 95% of our OBS clock drift observations have a substantial non-linear component, notably in the first months of the deployment. Overall, our estimated time-dependent clock drift curves accurately predict the final skew, with an average difference of ~160 ms to GPS skew values obtained at retrieval. We test our skew curves by using them to correct examples of recordings of teleseismic earthquakes and of local-regional seismicity. Uncertainty analysis of the skew curves gives a mean skew error of ~110 ms, indicating the suitability of the corrected data for future seismological studies such as for seismic tomography, seismicity analysis and moment tensor inversions. A dedicated open-source Graphical User Interface toolbox in Python 3 has been developed to facilitate OBS clock synchronization using seismic ambient noise.