Detection of free Earth oscillations using the GNSS VADASE algorithm: results for the 2011 Tohoku-Oki earthquake

It is well know that strong earthquakes generate observable low-frequency seismic waves that propagate many times around the Earth. These day-scale Earth's free oscillations have traditionally been detected by underground instruments such as horizontal pendulums (e.g. [Bolt & Marussi, 1962]), strainmeters, gravimeters and seismometers, to investigate Earth's internal structure, geodynamics and source properties of earthquakes. The Fourier spectra of observed seismic waves show many peaks in the very low-frequency range (<5 mHz), which correspond to Earth normal modes. Such spectra contain important information about the large-scale structure of the Earth (e.g [Dziewonski and Gilbert, 1971] and [Woodhouse et al., 1986]) and are also an essential element in the study of earthquakes to determine the overall moment of very large events (e.g. [Park et al., 2005]).

A pioneering study using data collected along 17 hours after the 2011 Tohoku-Oki megathrust earthquake [Mitsui and Heki, 2012] shows that the dense GPS arrays can truly detect both spheroidal and toroidal fundamental modes in 3D displacements. In that study the GPS observations acquired by the very dense Japanese permanent network GEONET were analysed using the standard Precise Point Positioning approach.

Here we propose to use the much simpler VADASE (Variometric Approach for Displacements Analysis Standalone Engine) algorithm, analyzing velocities instead of displacements time-series in order to detect the Earth normal modes. VADASE was successfully applied to estimate ground velocities and displacements induced by earthquakes, and a relevant application was just the Tohoku-Oki earthquake itself [Branzanti et al., 2012]. Hence, each considered GPS receiver works in standalone mode and the 3D velocities are estimated through VADASE; then, the 3D velocities are FFT transformed in frequency domain and a stacking method must be applied in order to improve the signal-to-noise ratios and have the chance to detect the free Earth oscillations.

We compared different data sets (IGS and the SONEL networks) and two different spectra stacking techniques. In particular we have processed 24 IGS stations for a period of 96 hours and more than 100 SONEL stations for a period of 24 hours, using the amplitude stacking and the product stacking methods. The encouraging results are compared and discussed and some conclusions and future prospects are outlined.