Near real-time noise removal for the Monterey Ocean Bottom Broadband (MOBB) seismic station data

The Monterey Ocean Bottom Broadband (MOBB) observatory, located 40 km offshore central California, at a water depth of 1000 m, provides important complementary coverage of the San Andreas Fualt system to the land-based network. First installed in 2002, it is arguably the longest lived ocean bottom broadband seismic station. It includes a three-component broadband Guralp CMG-1T seismometer and a collocated differential pressure gauge (DPG) to measure the local water pressure continuously, as well as a current meter. After 7 years of autonomous operation, in February 2009, MOBB was successfully connected to the MARS cable (http://www.mbari.org/mars), and the data have been available in real time at the Northern California Earthquake Data Center (Romanowicz et al., 2009). However, the usage of MOBB data has been limited because of the noisy character of the data, in particular at periods of interest for regional moment tensor studies (20-100 sec), due to the ocean infragravity waves. Crawford and Webb (2000) demonstrated that there is a strong correlation between the water pressure and the vertical component of seafloor ground velocity in the infragravity wave band. Applying this to MOBB vertical component data, a transfer function (TF) was determined and utilized to successfully deconvolve the pressure-correlated noise from the vertical component of MOBB seismograms (Dolenc et al., 2007) in the period band 20-200 sec. Romanowicz et al. (2003, 2009) presented examples of how the cleaned MOBB data contribute to the determination of source parameters and regional structure. These past efforts, however, have been mostly case studies for illustration purpose. In this study, we systematically process all the available MOBB data since 2009 (because the cable was trawled, about a year of data is missing from February 2010 to June 2011). We calculate the TF over time and find that it is generally very stable, except for one change in 2010 due to an instrument replacement. Two universal TF's (one for the period before the change and one after) are therefore defined and utilized for systematic noise removal. We demonstrate the effectiveness of the technique by applying the cleaned MOBB data to moment tensor inversion of all Mw4.0+ and many Mw 3.5+ events in Northern and Central California in 2009-2012. In addition, we also try to process the MOBB data before 2009. Although the TF is less coherent during the autonomous period of operation due to various problems with one or the other of the instruments, there is potential for application to at least part of the data. Because the vertical component TF is so stable, it can be computed in advance, and the noise removal can be done routinely in near real-time ( with ~8 minutes delay), which is sufficient for routine regional moment tensor determination. This procedure is in the process of implementation in the northern California real time earthquake notification system.