A comparison of earthquake coda waves at surface versus subsurface seismometers

Abstract

The coda waves of 21 Parkfield earthquakes were recorded simultaneously at the surface and at a depth of 198 m. We characterize these codas by computing (1) the time dependence of the integral of squared particle velocity, (2) their frequency content, (3) time‐frequency coda decay planes, (4) the frequency dependence of root mean square ellipticity, and (5) the difference in their normalized decay rates. These five measures reveal significant differences in the early portion of the uphole versus downhole codas. Our analysis solves for coda Q as an explicit function of both time and frequency. Average apparent coda Q for the 21 events observed both at the surface and downhole is Qa(f, τ) = 39(f/f0)0.43 (τ/τ0)0.37, with the reference frequency and time defined at 1 Hz and 1 s, respectively. Coda Q exhibits a trend in its value over the 1‐month period of the study. The trend correlates with a spatial variation in source locations and is not produced by temporal variations of the Q of the medium. The spectral ratio between the uphole signal and the downhole signal varies with time through the coda. The early portion of the uphole coda experiences a pulse of low‐frequency energy not seen in the downhole coda. To explain this observation, we propose a model of the coda that includes body wave to surface wave scattering near the site. In this model, the near‐surface acts as both a filter and a scattering waveguide. The coda at the surface is the sum of a coda produced by scattering deep in the lithosphere and a coda produced by scattering in the near‐surface. The scattering in the near‐surface is made up of near‐vertically traveling multiples and horizontally travelling surface waves. The guided waves are recorded in the surface coda but are absent from the coda recorded at depth. The time‐frequency structure of the surface waves in the uphole coda is determined by the velocity, scattering, and attenuation characteristics of the site. By fitting the model to our data, we have found that the fractional conversion factor of body waves to surface waves travelling toward the receiver is 0.03 at the Vineyard Canyon site. The relatively rapid damping of these trapped waves can be modelled by assuming a quality factor of QT = 50. The principles of this model are general enough to be of use at other locations.

Attenuation models (QP and QS) in three dimensions of the southern California crust: Inferred fluid saturation at seismogenic depths

“…The inverse relationship between t *( Q S ) residuals and the M L corrections is consistent with the results of Blakeslee and Malin [1990]. Using borehole records, they demonstrated that at lower frequencies the early coda recorded at the surface has more low‐frequency content than coda recorded at depth.…”

Local wave propagation near the San Jacinto Fault Zone, southern California: Observations from a three‐component seismic array

“…The term "site amplification" appears, however, to be a bit of a misnomer in that the near-surface environment imparts far more complexity than can be accounted for by a simple amplification factor. These array observations concur with results from other studies which suggest that the effects of the near-surface environment (at some sites) represent a combination of a filtering effect due to the low Q nature of the near-surface material, an amplification effect due to the impedance contrast between the basement and near-surface materials, and the excitation of resonance modes that remain trapped in a surficial waveguide [Blakeslee and Malin, 1990]. These three-component array observations add to the growing body of observational evidence showing that neither the P nor the S codas for local earthquakes are entirely consistent with the conventional, random scattering model for coda excitation.…”

“…First, much of the "incoherent" coda is composed of waves that can in fact be modeled as coherent plane waves on a tens-of-meters scale [Wagner and Owens, 1993]. Second, a comparison of the borehole records shown above the BTRs in Plates 1 and 2, and the surface records shown in Figure 3 (compare also the borehole and surface records in Figure 4) reveals that the surface records are not simply the borehole records amplified by some factor owing to the impedance contrast between the basement and near-surface materials, but instead that the surface records contain a considerable amount of energy that is not observed at depth [Blakeslee and Malin, 1990]. These two features lead to the almost inescapable conclusion that the incoherent component of the coda at PFO is related to localized near-surface effects as opposed to scattering from distant heterogeneity [Fletcher et al, 1990;Aster and Shearer, 1991b ; Anderson, 1993; Al-Shukri et al, 1995].…”

Source array analysis of coda waves near the 1989 Loma Prieta, California, mainshock: Implications for the mechanism of coseismic velocity changes

“…They found that the early coda was dominated by waves with slowness near the direct arrivals. Blakeslee and Malin [1990] Ellsworth et al [1992] analyzed 21 multiplet earthquakes that occurred on the Calaveras and San Andreas faults near the Loma Prieta mainshock and that spanned the occurrence of the mainshock temporally. The authors identified path-averaged coseismic velocity changes of up to 0.8% on source-receiver paths traversing the aftershock zone, and in every case the velocity was lower after the mainshock.…”