Retrieving displacement from seismic acceleration records is often difficult because unknown small baseline offsets in the acceleration time series will contaminate the doubly integrated record with large quadratic errors. One-hertz Global Positioning System (gps) position estimates and collocated seismic data are available from the 2003 Mw 8 Tokachi-Oki (Hokkaido) earthquake. After a process of correcting for possible misorientation of the seismic sensors, an inversion method is used to simultaneously solve for ground displacement with both data sets as input constraints. This inversion method takes into account the presence of unknown offsets in the acceleration record, and the relatively large uncertainties in the estimated 1-Hz gps positions.
In this study, 117 channels of seismic data were analyzed. Only 5% of the time does the static displacement retrieved from traditional baseline correction processing without gps information agree with the absolute displacement measured with 1-Hz gps to within the errors of the gps data. In solving simultaneously for constrained displacements that agree with both the seismic and gps data sets, an optimal solution was found that included only one- or two-step functions in the acceleration records. Potential explanations for the offsets are analyzed in terms of tilt of the sensor or electronic noise. For nine stations, clear misorientations of the seismic sensors of more than 20 deg from the reported orientation were found. For this size event, the 30-sec sampled gps solutions were also a sufficient constraint for establishing the offset errors and recovering reliable displacements. The results significantly extend the frequency band over which accelerometer data are reliable for source inversion studies.
Online material: Plots of constrained seismograms with metadata.