Automatic determination of OBS sensor orientation using direct water waves generated by airgun shots: a prelude to seismological analyses, northern Hikurangi subduction margin, New Zealand

Abstract:

The northern margin of Hikurangi subduction zone experiences a range of slip behaviors including tsunami earthquakes [Eiby, 1982], repeating slow slip events, [Wallace et al., 2016], and tectonic tremor [Todd et al., 2018]. To deepen our understanding of physical processes along the subducting plate interface and upper crust, we have deployed a dense Ocean Bottom Seismometer (OBS) array as part of the NZ3D seismic experiment. Ninety-seven 3-component OBSs were placed in a 60 km x 14 km area, approximately 2 km apart, during December 2017 to March 2018. An active-source seismic survey was undertaken during January 2018, as well as passive seismic observation in the remaining period.

The data recorded through the NZ3D experiment offers tremendous opportunities to study the Hikurangi subduction margin. However, many analysis methods capitalizing on the 3-component recordings, cannot be used without first correcting for the orientation of the two horizontal OBS sensor components. Using direct water arrival phases generated by airgun shots, the orientation of sensors at the seafloor can be estimated with a precision of less than 1˚ [Anderson et al., 1987; Duennbier et al., 1987]. However, it requires accurate pick of the direct phases; a tasked usually achieved manually in most active-source seismic surveys. The large number of receiver and sources (97 OBSs and >140,000 shots) used in the NZ3D experiment rules out manual picking of direct phases as a viable method to determine sensor orientation.

To overcome this issue, we have developed an automatic algorithm to pick the direct phase and determine orientation of OBS sensors. The direct phase is picked by analyzing 3-component waveforms using a frequency index [Buurman and West, 2010] and a recursive kurtosis [Poiata et al., 2016]. The orientation is determined by fitting a linear curve to the particle motion of the picked direct phase using least squares. To test the performance of the developed scheme we spatially down-sampled the full data volume using only 2,094 airgun shots to determine the orientation. As a result, the orientations of the OBS sensors are determined with mean standard deviation of 5.76˚. The whole process can be executed within one day. The orientation corrections now pave the way for further advanced analyses of the OBS dataset acquired in the NZ3D experiment.