Inverting ground motion from a seismometer array to obtain the vertical component of rotation: A test using data from explosions

To study factors that affect reliability of the derived vertical rotational ground motions (about vertical axis) from translational array waveforms, we have conducted a field test using explosion. The dataset was collected on March 4, 2008 from an active source experiment of the Taiwan Geodynamics Research project (TAIGER, 2008), in which 750 kg of explosives were set off about 500 m away from a specially designed seismic array in free field consisting of 8 tri-axial rotational sensors, 13 tri-axial accelerometers, and 12 six-channel, 24-bit dataloggers with GPS receivers. Our results show that the waveform fits were greatly improved when the explosion-induced crustal strains were less than 2 × 10−5 for our station configuration. As a result, our modeling for the large amplitude part of the mainshock was not entirely successful. However, we have successfully recovered the first order features of the later part of the vertical rotational ground motions using translational velocity waveforms in the 0.5–20 Hz bandwidth using the Spudich and Fletcher’s procedure (2009). Moreover, we used a modified Jaeger’s (1969) formula and derived similar rotational synthetics. We found that the best rotational rate waveform fits were derived using a station configuration of a dense, small-aperture, translational array combined with a few large-offset stations. A small-aperture array is needed to fulfill the uniform rotation assumption. However, inverting data from an array of small spatial dimension requires waveforms of high signal to noise ratio and high sampling rates. Surprisingly, large-offset translation stations provided additional constraints in time and space, helping to fit the low frequency waveforms. Using the inverted result, we forward predicted the vertical translational waveforms not used in the inversion and got excellent fits. This suggests that the inverted results are of good quality for predicting the complete translational ground motions, even though only horizontal components were used in the inversion. Array-derived or directly measured rotational ground motions can be used to predict translational ground motions in the near-by region. Lessons learned from this study could be helpful for future studies using translational ground motions to derive dynamic ground strains, tilts, and torsions.

► An explosion was recorded by an array of dense translational and rotational sensors.
► Translational waveforms were inverted for synthetic rotational waveforms.
► Two different inversions methods gave similar and good results.
► Difficult to fit the large amplitude parts of the rotational waveforms.
► Preferred network has a dimension of 1/2 to 1/8 of the dominant seismic wavelength.