On Baseline Corrections and Uncertainty in Response Spectrafor Baseline Variations Commonly Encounteredin Digital Accelerograph Records

Abstract

Most digital accelerograph recordings are plagued by long-period drifts, best seen in the velocity and displacement time series obtained from integration of the acceleration time series. These drifts often result in velocity values that are nonzero near the end of the record. This is clearly unphysical and can lead to inaccurate estimates of peak ground displacement and long-period spectral response. The source of the long-period noise seems to be variations in the acceleration baseline in many cases. These variations could be due to true ground motion (tilting and rotation, as well as local permanent ground deformation), instrumental effects, or analog-to-digital conversion. Very often the trends in velocity are well approximated by a linear trend after the strong shaking subsides. The linearity of the trend in velocity implies that no variations in the baseline could have occurred after the onset of linearity in the velocity time series. This observation, combined with the lack of any trends in the pre-event motion, allows us to compute the time interval in which any baseline variations could occur. We then use several models of the variations in a Monte Carlo procedure to derive a suite of baseline-corrected accelerations for each noise model using records from the 1999 Chi-Chi earthquake and several earthquakes in Turkey. Comparisons of the mean values of the peak ground displacements, spectral displacements, and residual displacements computed from these corrected accelerations for the different noise models can be used as a guide to the accuracy of the baseline corrections. For many of the records considered here the mean values are similar for each noise model, giving confidence in the estimation of the mean values. The dispersion of the ground-motion measures increases with period and is noise-model dependent. The dispersion of inelastic spectra is greater than the elastic spectra at short periods but approaches that of the elastic spectra at longer periods. The elastic spectra from the most basic processing, in which only the pre-event mean is removed from the acceleration time series, do not diverge from the baseline-corrected spectra until periods of 10-20 sec or more for the records studied here, implying that for many engineering purposes elastic spectra can be used from records with no baseline correction or filtering.

Estimating coseismic deformations from near source strong motion records: methods and case studies

“…Iwan et al (1985) for the IS604 north-south component data of the ICEARRAY. Solid lines with different colours correspond to different values of t FITb 4.2 Method of Boore (2001) and improvements by Akkar andBoore (2009) Boore (2001) generalized the correction procedure of Iwan et al (1985) by making t 1 and t 2 as free parameters, not determined by a threshold of shaking, and subject to the constraints t 1 > 0, t 2 > t 1 and t 2 < t END , where t END corresponds to the end of the record. In a further simplification he proposed that t 2 be chosen as the time where a linear fit to the later portion of raw velocity becomes zero.…”

“…In addition, the selection of different correction time parameters, namely t 1 and t 2 , were found to give widely differing results (Boore 2001). Akkar and Boore (2009) proposed additional parameters t BLb (time of base line beginning), t BLe (time of baseline end), and t FITb (time after which a straight line is fitted to the velocity). These parameters were intended to enforce additional constraints on t 1 and t 2 .…”

“…They also proposed iterative schemes to estimate these parameters. Detailed definitions of the parameters and the iterative procedure to estimate them will not be mentioned here, and interested readers are referred to Akkar and Boore (2009).…”

“…Such applications are generally not feasible, and the accuracy of derived information is often in doubt due to the presence of baseline shifts and distortions as well as superimposed longperiod noises. Baseline shifts are small variations in acceleration zero-level, often occurring as near-instantaneous shifts (Akkar and Boore 2009). Baseline shifts in acceleration result in nonzero velocities at the end of the record, which are unphysical (Boore 2001;Boore et al 2002).…”

Déplacements et rotations du sol lors de forts séismes à proximité de failles actives : apports des capteurs accélérométriques

“…Par ailleurs, il apparaît ici que soustraire la rotation permanente des accélérogrammes ne modifie pas de façon significative le spectre de réponse en déplacement jusqu'à des périodes d'environ 20 s, ce qui est cohérent avec les études antérieures (Akkar et Boore, 2009 ;Paolucci et al, 2008 ;Wang et al, 2007).…”