Broad-band array analysis of the Puget Sound Region 1434-95-G-2605 Kenneth C. Creager and Stephen D. Malone, P.I.s Dept. of Earth and Space Sciences University of Washington Seattle, WA 98195 (206) 543-8020 e-mail: kcc@ess.washington.edu, steve@ess.washington.edu

Program Element: II.1

Investigations Undertaken

In 1994 we deployed sixteen three-component, broad-band seismometers in a linear, 100 km long, east-west array stretching from the Olympic Mountains through Seattle to the Cascade Mountains (see Figure 1). These stations recorded continuous data at 20 samples per second for about two and a half months. We have converted most of the data to SEED format and are archiving it at the IRIS Data Management Center in Seattle.

Figure 1. Crustal geology of the Puget Lowlands (after Johnson et al., 1994), and station locations of our temporary array.

Results

Our primary modeling work to date is analyzing the receiver functions calculated for these data. A receiver function is the deconvolution of the vertical component of a teleseismic P wave and its coda from the radial component. In principle, the structure of the P-waveform caused by the source complexity and by wave interactions near the source are the same for both components and are therefore removed by the deconvolution, leaving a waveform that depends primarily on structure near the receivers. These waveforms are modeled by comparing them with similarly deconvolved synthetic seismograms. An advantage of analyzing receiver functions of closely-spaced seismometers is that several features of the receiver function are correlated across the array, facilitating their interpretation.

Two features in the observed receiver functions that are very clear across the array are apparently caused by impedance contrasts associated with the eastward dipping subducted slab, and with the southward dipping base to the Seattle Basin.

• The Seattle Basin: Our work to date has focused on forward modeling of the entire array using simple 3-D models with one or two dipping interfaces. A striking feature of the first 5 seconds of 1-Hz receiver functions is a pronounced positive peak at about 2 sec. and the absence of a peak at zero lag for stations that lie within the "Seattle Basin" east of the Hood Canal fault and west of the Cascade foothills, and The peak at 2 sec. disappears entirely and the zero-second-lag arrival appears for stations to the west of Hood Canal, and in the Cascades. Both of these features can be readily explained by synthetics calculated from a model inferred from a north-south reflection profile taken along Puget Sound. The reflection line shows a very strong reflector at 3.5 sec. two-way travel time, which has been interpreted as a sharp boundary between the Crescent Basalts below and sedimentary rocks above. This boundary is dipping 15 degrees down to the south. Synthetics calculated from a 3-D model consistent with this interpretation, and with a large impedance contrast, explain the main features in the first few seconds of the observed receiver functions. It is especially interesting to note that the Seattle Basin does not appear to have any significant east-west dip, and seems to continue at the same depth all the way to Hood Canal where the basin disappears abruptly. The Seattle Basin appears to be fault-bounded to the east by an extension of the Whidbey Island fault.

Broad-band, transverse-component seismograms from an intermediate-focus earthquake in Mexico, 40 degrees away, display large-amplitude reverberations after body-wave phases such as S, sS, and SS for stations within the basin. The reverberations are peaked at a period of about 10 sec., and are often larger in amplitude than the direct body waves. We are modeling this part of the seismogram using a Direct Solution Method [Cummins et al., 1994] which calculates complete synthetic seismograms in a spherical geometry at periods of 4 sec. and larger. The qualitative features of the SH seismograms can be modeled with a simple low-S velocity basin below Seattle, but a detailed waveform match may require a 2-D calculation.

• Slab Structure: The Mw 7.6 deep-focus Fiji earthquake of 9 March 1994 was recorded by each of our seismometers, and provides a source with good signal-to-noise at long periods. We are analyzing long-period receiver functions from this event to image the crust-mantle boundary of the subducting slab, which lies at a depth of 40-60 km along our array. A prominent arrival, which can be traced across our entire array, and varies systematically in time from 15 to 25 sec., is consistent with synthetics calculated from a discontinuity dipping at 16 degrees to the east, and is interpreted as the Moho of the subducted oceanic lithosphere. We hope to model this arrival in more detail to be able to determine the precise depth of the subducted Moho relative to the seismicity so we can determine whether the seismicity is in the subducted crust or subducted mantle, a question with important implications regarding the physical mechanisms for intermediate-focus earthquakes. Before this question can be answered we must improve the model of the upper crust (Seattle Basin) because the large S-wave velocity anomalies associated with this basin significantly affect the travel times of phases used to image the depth to the slab.

Pullen, J.L., K.C. Creager, S.D. Malone, Broadband array study in the Puget Sound region, Washington, EOS (abs) 1995 in press