ML20046C130

App a to Seimic Margin Earthquake Study for Trojan Site

ML20046C130
Person / Time
Site:	Trojan File:Portland General Electric icon.png
Issue date:	08/05/1993
From:	PORTLAND GENERAL ELECTRIC CO.
To:
Shared Package
ML20046C128	List: ML20046C127 ML20046C130
References
NUDOCS 9308090230
Download: ML20046C130 (35)

Text

-

o

k -

p i

- Trojan Nuclear Plant Enclosure to PGE Letter Docket 50-344 dated August 5,1993, License NPF-1 Supplemental Information -

Seismic Margin Earthquake Study (34 Pages)

APPENDIX A to Seismic Margin Earthquake Study for the Trojan Site 1

l l

i I

I i

1 4

930B090230 930805

'E}

PDR-ADOCK 05000344 P

PDR a

u

1 Final Report to the Portland Generni Electric Company GEOMETRY OF THE CASCADIA SUBDUCTED SLAB AND ITS RELATIONSHIP TO THE SEISMIC HAZARD OF THE TROJAN POWER PLANT by John Nabelek and Xiao-Qing Li College of Oceanic and Atmospheric Sciences Oregon State University Corvallis, OR 97331 ABSTRACT The recent deployment of several broadband stations in western Oregon provides an opportunity to investigate geometry of the subducted Juan de Fuca plate beneath these stations using convened phases from teleseismic earthquakes. Three of the stations (COR, in Corvallis; CHE, southwest of Portland; and.SIU, west of Eugene) are located over the subducted plate, and their data am analyzed for the deep stmeture employing the receiver function deconvolution method.

We focus our analysis on data from COR, which has been in operation since December,1989. The azimuthal distribution of suitable events for the analysis is very good. The receiver functions for events with similar location are very consistent; however, the azimuthal variation is strong. The broadband data are dominated by high-frequency energy, indicating multiple sharp impedance discontinuities, and the signal from the slab (phase conversions at the subducted oceanic Moho) is not readily apparent. The character -

of the traces changes as they a : low-pass filtered with a progressively lower cutoff, indicating variable gradients in the stmeture above the slab. The slab signal becomes dominant at lower frequencies, with the prominent phases identified as Ps (7 s after Pp) and PpPms (16-19 s after Pp). Below frequencies of about 0.5 Hz, the data show sufficient coherence that they can be modeled using relatively simple cmstal models parameterized by planar interfaces. We concentrate our modeling on the low-frequency data (low-pass cut-off of 0.2 Hz) where the slab signal overwhelms other cmstal contributions. At these low frequencies even the simplest model consisting of a homogeneous crust overlying a mantle with a dipping Moho fits the data remarkably well.

The azimuthal variations of Ps and PpPms are consistent with an eastward dipping slab with a dip angle of about 10-15*. If we assume the average compressional wave velocity of the crust (Otav), the arrival times of Ps and PpPms with respect to Pp can be used to determine the slab depth (depth of oceanic Moho, hmoho) beneath the station and the average shear-wave crustal velocity (Poisson ratio, o ). For n of 6.5 km/s, a value av av indicated the refraction studies, hmoho is 43 km and o y is 0.33. For higher a, the slab a

av depth slowly increases and the Poisson ratio decreases. Adding a more realistic upper crustal stmeture based on the refraction data and a 6-km-thick low-velocity layer above the subducting mantle to the modelimproves the fit. For such a model the oceanic Moho beneath COR is at 40 km depth, dipping 10 eastward, and the average Poisson ratio of the overlying crust is 0.33. The high Poisson ratio indicates that the entire crust beneath COR is composed of ultramafic/ gabbroic rocks, which typically have Poisson ratio in the 0.30-0.33 rage. It is possible that our estimates of hmoho and cav based on planarinterface models are somewhat biased. A careful examination of the azimuthal amplitude variations of the major phases indicates that the oceanic moho beneath the station may be dipping at a lower angle than the average dip indicated by the planar model, while the dip further west w..

e

~-

--m

-.--.n...

i 2

may be steeper. In that situation, Gav would be somewhat lower and hmoho (below COR) would be somewhat greater than the values derived for the planar model. The low-velocity i

layer above the subducting mantle has a thickness and velocity indicating that it might be

~

the subducted oceanic crust. The top interface of this low-velocity zone is at appmximately the same depth as the deep reflecting surface imaged by COCORP in this region.

Only a few good records are available for CHE and SIU. Data from CHE indicate a i

greater depth (47-50 km) of the slab beneath this station than beneath COR. Tnis is expected, since CHE is funher areward fmm the trench axis. On the other hand, data from SIU indicate a slab depth similar to that found beneath COR.

This study was conducted on behalf of the Pordand General Electric Company.

)

i i

/

8f[43 i

John Nabelek Associate Professor Oregon State Universiny tel: (503) 737-2757 fax:(503) 737-2064 e-mail: nabelekjCeues.orst.edu 9

P E

I t$

t i

e

3 BACKGROUND Of the many factors influencing seismic hazard, the most important is the

[

proximity to the fault plane on which the earthquake occurs. Because of this relationship the seismic hazard in a subduction zone is largely controlled by the geometry of the subducted slab. The geometry of the subducted slab influences how far landward from the coast strong ground shaking from a large subduction zone earthquake can extend. When slab is steeply dipping, its distance from the surface increases rapidly and thus the area of strong ground shaking from earthquakes associated with the slab is confined to a relatively narrow mgion near the coast; on the other hand, for a gently dipping slab the hazard extends much further inland.

i One of the simplest ways to study the geometry of the subduction zone is to map out the seismicity. The seismicity in northwestern Washington indicates the slab is bowed up, with relatively low dip near the Canadian border and progressively steepening to the south (Crosson and Owens,1987; Weaver and Barker,1988). Unfortunately, the anomalously low seismicity associated with the subducted slab south of central western Washington does not allow to extend the seismicity map further to the south.

In the absence of slab seismicity, an effective means of determining the position of the slab is the so-called receiver function method. This method utilizes conversions of seismic waves arriving from distant earthquakes as they pass through the slab. The method provides information about the slab depth and its dip beneath the station. The analysis requires high-quality 3-component seismic stations with good response over a broad band of frequencies. In this study we utilize data from three stations in Oregon that are well suited for such an analysis, the OSU/ IRIS station in Corvallis, which is a part of the global network and has been in the operation with a new state-of-the-art digital seismic instrumentation since December 1989, aniPwo temporary broadband digital stations operated by the University of Oregon in 1991 and 1992; one is located west of Eugene and the other at Chehalem Hills, southwest of Portland. All three stations are at about the same i

distance from the trench axis as Trojan (Figure 1) and hence an estimate of the slab position beneath these stations is directly relevant to the seismic hazard of the power plant.

i PAST WORK WITH CONVERTED PHASES Waveform modeling of converted phases of teleseismic P waves has been used effectively to study deep structure at isolated stations. Body waves from distant earthquakes travel through the lithosphere at steep angles (typically less than 20 from the vertical). Interaction of the incident teleseismic body wave with large impedance interfaces beneath the recording station generates various converted phases and multiples that can be observed in the recorded waveform (Figure 2). Because the P to S conversions dominate the receiver function waveform, the receiver function is primarily sensitive v. the shear wave velocity stmeture.

The concept of using converted phases and reverberations within the lithosphere as recorded in the signature of a telescismic P wave to characterize the lithosphem dates to Haskell (1962) and was first used in the spectral domain for this purpose by Phinney (1964). Burdick and Langston (1977) moved the analysis to the time domain, and it is in I

this form that the technique has found most ofits success. Teleseismic P-wave aceiver functions have been used successfully to study lithospheric structure in a wide variety l

i

-m--=

.w

-.e-c

4 tectonic settings (e.g., Langston,1977a; Langston,1977b; Langston,1979; Langston, i

1981; Owens 1984; Owens et al.,1987; Owens et al.,1988).

Detailed analysis of the P-wave receiver function requires estimation and removal of the time function of the P wave incident at the base of the lithosphere. Techniques for this are described by Langston (1979) and Owens et al. (1984). The waveforms recorded at the receiver are a convolution of receiver structure, instmment response and incident wavefonn. The incident waveform (let us call it the source) and instrument msponse (for matched instmments) are common to the radial and vertical components. The i

deconvolution of the radial P waveform by the venical P waveform removes the effects of source and instmment. Because the vertical lithospheric response is nearly an impulse l

function, the resulting waveform, the P receiver function, is predominantly characterized by P to S conversions and multiples within the lithosphere.

For horizontally layered structure as in Figure 2, the receiver functions are independent of azimuth and are radially polarized. When the stmeture is dipping, the mam difference from the horizontally laye ed case is that the receiver functions are no longer radially polarized and become strong functions of azimuth. The azimuthal dependence is strong both for the pnmary conversions and for the multiple reflections; it affects the observed amplitudes and polarities of the phases as well as the period of the reverberations.

This is illustrated in Figures 3 and 4. The predictable amplitude dependence of the primary conversions on the dip of the stmeture and on the azimuth of the incident P wave makes the identification of the phases possible. Using ray tracing codes, the primary converted phases can be modeled and the structure, including the dip of the interfaces, can be estimated. An effect of the the dipping stmeture is that the sampling volume by the crustal phases is not symmetric beneath the station, as is the case for the the horizontal stmeture, but is shifted updip (Figure 5).

PREVIUS RECEIVER FUNCTION STUDIES IN THE PACIFIC NORTHWEST The subducted slab in the western Washington and Oregon has provided an l

obvious target for testing receiver function techniques. In his seminal work, Langston (1977a,1981) used the former WWSSN station COR to investigate the structure beneath this station. He argued that in order to explain the large Ps conversion observed in the data a low-velocity zone is required in a depth mnge between 25 and 45 km. The lower interface, according to Langston,is sharp and is dipping to the east at an angle of about 20.

He suggested that this sharp velocity discontinuity represents the crust-upper mantle interface of the subducting oceanic cmst. Langston concentrated only on the deeper stmeture, assuming the structure in the upper part of the crust based on Berg et al.'s (1966) model. The recent cmstal studies in the mgion (Trehu et al.,1992) indicate that Berg et al.'s model, which has a very shallow Moho discontinuity, is not appropriate for the area near Corvallis and therefore the details of Langston's model are subject to substantial

(

uncertainties.

The more recent work in the Pacific Northwest using receiver functions has concentrated on the geometry of the subduction zone in the Puget Sound region where the analysis has helped to delineate the arched stmeture (Owens et al.,1988; Lapp et al.,1990).

One of the best studied the ponions of the Juan deFuca subduction zone is the Vancouver l

Island region where the LITHOPROBE profiles in combination with the just-completed receiver function work by Cassidy (1991) trace the subduction interface from the near-1 u

5 surface at the trench to a depth of about 70 km beneath westem British Columbia.

t Cassidy's work involved modeling of receiver functions from 3 broadband stations aligned perpendicularly to the trench axis. The analysis reveals several dipping low-velocity layers, i

the deepest one presumably corresponding the subducted oceanic crust.

-i DATA DESCRIPTION AND ANALYSIS PROCEDURE l

We focus our analysis on data from COR, which has the most extensive data set.

i The time period from which we chose suitable events for the analysis spans from i

December,1989 to October,1992. Using PDE catalogue we chose potentially useful events from various azimuths surrounding the station. These were examined and those showing good recordings were included in the data set. Generally, only events with magnitude greater than 6 provided good quality data. The chosen events are listed in Table t

1. The azimuthal distribution is very good (Figure 6); the largest two gaps in the coverage are 20 - 105* and 170 - 230. The seismicity rate in those azimuths is low and it may y

take many years to fill in those voids.

l 1

For each event, the set of 3-component broadband P-wave recordings was converted to the receiver function through the following steps. First, the P waves were isolated from the original broadband recorded seismograms. The time window for the extraction was adjusted for each event. In general, the start point of a window was about

't 10-30 s (about 20% of the total window length) before first P arrival. The end point was chosen as late as possible so that the P energy on the venical component decayed to a low l

level, but before the arrival of other major phases, such as PP. For some deep focus.

i earthquakes, pP was also eliminated from the window. The NS and EW components wem rotated to radial and transverse and tapend by a 20% cosine taper before transforming them into the frequency domain. The radial and transverse receiver functions were obtained by dividing the spectrum of the radial and transverse component, respectively, by the spectrum of the vertical component, and then transforming back to time domain. The deconvolution procedure is similar to that described by Ammon (1992). To assure stability of the deconvolution, one generally sets two parameters which smooth the waveforms, a " low water level" parameter c and a low-pass Gaussian filter parameter g. In this smdy we set c to be an average of the vertical-component spectral amplitude between 5-6 Hz, resulting in c-values between 10 5 o 10-8. The g-value was 10. These values gave t

1 a comer frequency of about 2 Hz. At the low frequency end, the stability was controlled by a high-pass filter with a comer at 0.02 Hz. For good high-frequency events the derived l

receiver functions are broadband Sm 0.02 Hz up to about 2 Hz. The deconvolutions for events which did not prodac grong high-frequency energy are generally noisy at the highest frequencies, but these were subsequently low-pass filtered into the bands suitable for modeling as discussed below. In this study we consider only the radial receiver.

functions because they are generally much more stable than the transverse.

)

In order to test the stability of the deconvolution procedure, we compare receiver functions in different frequency bands for two events with similar locations (Figum 7).

These are some of the best recorded events and we see consistency of the waveforms even at frequencies as high as 2 Hz. Most events with mb > 6 appear to produce stable results at frequencies lower than 1 Hz. The broadband data are dominated by high-frequency energy, indicating multiple sharp impedance discontinuities, and the signal from the slab (phase conversions at the subducted oceanic Moho) is not readily apparent. The character of the traces changes as they are low-pass filtered with a progressively lower cutoff,

.l

6 indicating variable gradients in the structure above the slab. The slab signal becomes dominant at lower frequencies, with the prominent phases identified as Ps (7 s after Pp) and PpPms (16-19 s after Pp). This will become apparent in the subsequent figures.

Figures 8 and 9 show the observed receiver function plotted as a function of azimuth of the incident waves for frequencies below 0.5 Hz and 0.2 Hz, respectively.

Below frequencies of about 0.5 Hz, the data show sufficient coherence that they can be modeled using relatively simple crustal models parameterized by planar interfaces. We concentrate our modeling on the low-frequency data (low-pass cut-off of 0.2 Hz) where the slab signal overwhelms other crustal contributions. Because of the large sampling volume (Figure 5) by the reflected phases modeling procedures based on homogeneous layers and planar interfaces are not adequate (and would not be expected to be adequate) for frequencies higher than 0.5 Hz.

MODELING OF CORVALLIS RECORDS The structure beneath the station is inferred from the observed receiver functions by comparison with theoretical receiver functions for a model structure. The model structure j

is adjusted until the predicted receiver functions match the observed. The theoretical l

receiver functions are calculated using a raytracing procedure developed by Owens (1984)

-I following the theoretical development of Langston (1977b). The procedure is relatively simple and allows only a class of models for which the medium within the layers is homogeneous and the interfaces, although allowed to dip in arbitrary direction, must be l

planar.

A comparison of the observed waveforms shown in Figure 9 with the theoretical seismograms shown in Figure 4 indicates that at low frequencies the character the observations is quite well represented by such a very simple model. We therefore start modeling the observations with this simple parameterization of the structure. The parameters that specify this structural model are the dip angle and strike of the dipping interface,its depth beneath the station, and the medium compressional and shear velocities and densities. In this model the continental and oceanic cmst are lumped into one, and are represented by the material above the interface. The material beneath the interface represents the mantle of the subducted oceanic lithosphere. For the purposes of the presentation, we will refer to this interface as Moho. The dominant phases for this structure are Ps and PpPms (Figure 4). Modeling of these phases will give us a rough estimate of the structural parameters. We can then refine the estimates by adding complexity to this model as requited by the data and by providing constraints from other geophysical observations.

One of the easiest parameters to estimate is the azimuth of the dip direction of the interface. Figure 4 shows that the character of the receiver functions is very sensitive to the relative azimuth between the dip direction and the angle of the incident waves. A rough estimate of the dip azimuth can be made by sliding Figure 4 over Figure 9. A more detailed analysis, which takes into account precise angular relationships for each receiver function, indicates that the slab beneath Corvallis dips along an azimuth of 90* - 100'. We assume 90 in all subsequent calculations.

Next we investigate what the receiver function waveforms indicate about the steepness of the interface. Figure 10 shows how the waveforms vary for three different

--m--

+.%

-.~+-

7 i

dip angles: 10,15, and 20. Focusing on the azimuthal amplitude variations of the major phases of the observed records (also shown in Figure 10), we see that on average a dip of 10-15' fits the best, however, a lower dip seems to be more appropriate for Ps, and a higher dip for PpPms. Since Ps sample the region under the station, while the largest-amplitude PpPms have the reflection points more than 20 km to the west (see Figure 3),

this may indicate that the Moho dip is smaller under the station, and increasing to the west.

The relative arrival time between Pp, Ps and PpPms (3 observables) is controlled by four parameters, the depth of the interface, its dip, and the crustal compressional and shear velocities. We therefore cannot determine all four of them from these observations alone. Fonunately, the average crustal compressional velocity is quite well know from the recent refraction experiments (Trehu et al.,1992); a value of 6.5 km/s appears to be well resolved. Assuming this value, we can solve for the depth, dip, and shear velocity (i.e., the Poisson's ratio).

Figure 11 shows how these values are resolved by the observed data. The figure shows observed traces from four different azimuths which are compared to the theoretical waveforms for combinations of three different values of dip and four different values of Poisson's ratio. The depth of Moho for each set of theoretical traces was adjusted so that Ps arrives at the time of the observed. We notice, that the PpPms arrives approximately on time only for dips between 10 -15 and Poisson's ratios of 0.31 - 0.33. The best fit is obtained for the dip of 10 and Poissun's ratio of 0.33. For these values, the depth of Moho is about 41 km. The high value of the Poisson's ratio is interesting, because, together with the rather high average compressional velocity determined from the refraction data, it implies that the nearly entire crust beneath Corvallis is probably composed of ultramafic/ gabbroic material (Holbrook et al.,1992). The nominal uncertainty in the depth estimate is 1.7 km, reflecting the precision with which the relative timing of Pp and Ps can be determined. As was said before, the estimated depth depends on the assumed compressional velocity in the cmst (n ). Because n was taken from an independent av av study, we have also calculated how the estimates of the Moho depth and Poisson ration would change if a were different. A conversion to other values of a can easily be av av made using Figure 12.

We refine our structural model by including a multi-layered upper cmst based on the refraction results published by Trehu et al. (1992). This allows us to focus on the lower cmst and the interface between the subducting and overriding plates. Our final model, shown in Figure 13, has a 6-km-thick low-velocity layer paralleling the oceanic Moho.

The thickness and velocity of this layer indicate that it migh; be the subducted oceanic crust.

The top interface of this low-velocity zone is at appmximately the same depth as the deep reflecting surface imaged by COCORP in this region (Keach,1986).

Figure 14 shows the fit to tb-data without the low-velocity layer. We see that the amplitude of the convened phases are substantially lower than actually seen. For this model, the impedance contrast betweca the oceanic mantle and the overlying cmst is too low. One could increase the mantle velocity, but the required values would be beyond the acceptable values. By adding a low velocity layer (Figure 15), the amplitude match for the main phases improves substantially. Figure 16 shows the theoretical receiver functions for the final model in the same form as the observations shown in Figure 5. Although our modeling did not explicitly involve the higher frequency receiver functions, we have also calculated the theoretical waveforms for the final model for the frequency band shown in i

I

8 Figure 4. The higher frequency theoretical waveforms (Figure 17) predict the character of the observations surprisingly well. We fee. that this model matches the data as well as can be expected for the class of stmetures we can model using ti e available modeling techniques. Some improvement could be clearly echieved by changing the dip of the structure near the surface and by including the deep complexi.y shown in Figure 13 c.

Such changes are, however, difficult to properly implement into the modeling code and would not significantly alter the fundamental conclusions we have derived.

DATA FROM CHE AND SIU Unfortunately the data from CHE and SIU are inadequate for a detailed analysis (Figure 18). The reason is two fold. The time frame over which these stations operated was too short to collect enough data and the stations were run with low-gain settings for which the amplifiers in the recording package are noisy at low frequencies. Nevenheless, we can draw some very uncenain conclusions based on a comparison of the few available records with those recorded in Corvallis. Such comparison indicates that Ps and PpPms at CHE arrive later than at COR, indicating a deeper depth of the Moho interface. If we assume the same average crustal compressional wave velocities as for COR, the observed arrival times imply the Moho depth of about 47-50 km under CHE. On the other hand, the records from SIU do not seem to differ significantly from the COR records, indicating a similar Moho depths. CHE is about 25 km further inland from the deformation front than COR (Figure 1). The increase in the Moho depth is consistent with the slab dip of about 15.

MAIN CONCLUSIONS The depth of the oceanic Moho beneath Corvallis is about 40-45 km, assuming an average cmstal velocity of 6.5 km/s.

The average Poisson ratio of the crust beneath and west of Corvallis is 0.31-0.33, indicating that the entire crust is composed of ultramafic material. The thickness is comparable to that of oceanic plateaus.

The average dip of the oceanic Moho beneath Corvallis is 10-15. The actual geometry may involve a lower dip (could be horizontal) directly beneath Corvallis and a higher dip (could be as high as 20 ) beneath the Coastal range west of Corvallis.

Paralleling the Moho, there appears to be a low-velocity layer whose thickness and velocity indicate that it might be the subducted oceanic cmst. The top interface of this low-velocity zone is at approximately the same depth as the deep reflecting surface imaged by COCORP in this region (Keach,1986).

The Corvallis data do not indicate continental Moho at a shallower depth. If present, it would need to be deep (roughly the same depth as the Moho we identify as oceanic).

Our results are in a broad agreement with earlier receiver function studies (Langston, 1977), refraction studies (Trehu et al.,1992), inferences based on gravity and magnetics (Finn,1990), and magneto-telluric measurements (Wannamaker et al.,

1989) in this region.

~

h Ma.mm --

~

.---_r_

9 FINAL COMMENT We are currently conducting a receiver function study involving over 40 temporary broadband stations distributed in a linear array from the coast near Waldport, across the Car

• arc, to the back arc region near Redmond. The recording started in April,1993 and will continue until December,1993. We believe that the results of this study will provide an extremely detailed image of the subduction zone beneath the array. Preliminary data from the array seem to support the conclusions reached in this report.

i h

9 h

f f.

i P

t 1

1 l

_-..1

10 REFERENCES Ammon, C. J., A comparison of deconvolution techniques, submitted to Pure and Applied Geophys.,1992.

i BerF, J.W., Trembley, L., Emilia, D.A., Hutt, J.R., King, J.M., Long, L.T., McKnight, W.R., Sarmah, S.K., Souders, R., Thiruvathukal, J.T., and Vissler, D.A., Cmstal seftaction profile, Oregon Coast Range: Bulletin SeismologicalSociety ofAmerica, v.

56, p.1357-1362,1966.

Burdick, L. J., and C. A. Langston, Modeling crustal-stmeture through the use of converted phases in teleseismic body-waveforms, Bull Seismol Soc. Am., 67, 677-691,1977.

Cassidy, J. F., Teleseismic receiver function analysis of the cmst and upper mantle of southern British Columbia, Ph.D. thesis, Univ. of British Columbia, p. 174,1991.

j Crosson, R.S. and Owens, T.J., Slab geometry of the Cascadia subduction zone from earthquake hypocenters and teleseismic converted waves, Geophys. Res. Lett., v.14, p.824,1987.

Finn, C., Geopgysical Constraints on Washington Convergent Margin Structure, Journal of Geophysical Research, 95,19533-19546,1990.

Holbrook, W.S., W.D. Mooney, and N.I. Christensen, The velocity stmeture of the deep continental crust, Continental Lower Crust, D.M. Fountain, R. Arculus, and R. Kay, eds., Elsevier, Amsterdam, 1-43,1992.

Keach, R.W., Cenozoic active margin and shallow Cascades structure: Cocorp results from western Oregon, M.S. Thesis, Cornell University,1986.

Langston, C. A., Corvallis, Oregon, crustal and upper mantle receiver structure from teleseismic P and S waves, Bull Seis. Soc. Amer., 67, 713-724,1977a.

Langston, C. A., The effect of planar dipping stmeture on source and receiver responses for constant ray parameter, Bull Seis. Soc. Amer., 67,1029-1050,1977b.

Langston, C. A., Structure under Mount Rainier, Washington, inferred from teleseismic body waves, J. Geophys. Res., 84, 4749-4762, 1979.

3 Langston, C.A., Evidence for the subducting lithosphere under southern VancouverIsland and westem Oregon from teleseismic P wave conversions: Joumal of Geophysical l

Research, v. 86, p. 3857-3866,1981.

Lapp, D.B., T. J. Owens, and R. S. Crosson, P-waveform analysis for local subduction l

geometry south of Puget Sound, Washington, Pure Appl Geophys., J33, 349-365, 1990.

Owens, T.J., Determination of crustal and upper mantle structure from analysis of broadband teleseismic waveforms, Ph.D. thesis,146 p., Univ. of Utah, Salt lake City, Utah,1984.

Owens, T.J., S. R. Taylor, and G. Zandt, Crustal structure at Regional Seismic Test 3

Network Stations determined from inversion of broadband teleseismic P waveforms, i

Bull Seismol. Soc. Am., 77, 631-662,1987.

Owens, T.J., Crosson, R.S., and Hendrickson, M.A., Constraints on the subduction j

geometry beneath western Washington from broadband teleseismic waveform modeling, Bull Seismol Soc. Am., v. 78, p.1319-1334,1988.

j o

Phinney, R. A., Structure of the Earth's crust from spectral behavior oflong period body waves, J. Geophys. Res., 69, 2997-3107,1964.

Trdhu, A., J. NdbElek, et al., A cmstal cross-section across the Cascadia subduction zone in central Oregon, Trans. Amer. Geophys. Un.,73,43,391,1992,

P 11 Wannamaker et al., Magnetotelluric Observations across the Juan de Fuca subduction system in the EMSLAB project, JGR, v. 94,14111-14126,1989.

Weaver, C.S. and Baker, G.E., Geomel,f cf the Juan de Fuca plate beneath Washington i

and northern Oregon from seismicity, Bull. Seismol. Soc. Am., v. 78, p. 264-275, 1988.

P l

r

}

4 I

5 i

l I

1

-1

.i I

i a

= - -

s-12 Table 1. The list of the prameters for the events recorded at COR.

Event # Date Time Lat.

Lon.

Depth mb Ms A BAZ Codet.2 (h:m:s)

()

()

(km)

()

(*)

01 19900620 21:00:10.0 36.957N 49.409E 18.5 6.4 7.7 98.2 5.9 I

44 19920720 07:46:47.8 78.691N 5.007E 10.0 5.4 6.6 533-11.2 M

02 19900530 10:40:06.1 45.841N 26.668E 893 6.7 85.8 20.5 I

G4 19920525 16:55:05.1 19.620N 77.820W 33.0 6.0 7.0 45.1 108.4 I

51 19921018 15:11:59.0 7.067N 76.870W 10.0 6.7 7.2 55.0 118.6 I

50 19921017 08:32:39.8 6.836N 76.873W 10.0 6.2 6.6 55.1 118.7 I

063 19910422 21:56:51.8 9.685N 83.073W 10.0 63 7.6 49.2 122.7 I

08 19900325 13:22:55.6 9.919N 84.808W 22.2 6.2 7.0 47.9 1243 I

09 19900508 00:01:40.0 6.905N 82.622W 9.6 6.2 63 51.6 124.4 I

11 19901017 14:30:13.1 10.970S 70.776W 598.8 6.7 73.0 125.5 E

133 19910524 20:50:55.8 16.506S 70.701W 127.7 63 77.6 128.7 m

14 19910623 21:22:28.9 26.802S 63349W 558.0.

6.4 89.9 129.4

.I-15 19900821 14:13:04.8 27.487S 104.266W 10.5 6.0 53 74.1 162.5 I

16 19910903 11:56:163 17.8355 116.034W 10.0 5.8 5.8 62.8 172.2 184 19920512 18:05:44.9 16.518S 172.325W 33.0 63 6.7 75.4 228.4 I

19 19910609 07:45:02.1 20.252S 176.218W 265.5 6.1 80.8 2293 I

39-19920711 10:44:20.5 22311S 178.510W 381.0 6.1 83.5 229.9 I

l 48 19900303 12:16:28.0 22.122S 175.163E 33.2 63 7.4 87.1 234.6 I

.i 23 19900812 21:25:22.0 19.435S 169.132E 140.4 63 88.7 240.7 I

i 49 19921017 02:51:543 19.129S 169.426E 33.0 5.7 6.5 88.1 240.8 I

46 19921011 19:24:15.5 19.077S 168.906E 33.0 6.6 6.9 88.4 241.2 I

24 19900727 12:37:59.5 153555 167.464E 125.7 6.4 86.7 244.6 m

25 19910814 19:15:06.6 13.548S 167.574E 33.0 6.06.6 853 245.7 M

264 19920527 05:13:42.9 10.592S 165.268E 33.0 6.0 7.0 84.5 249.4 1

573 19910209 16:18:58.4 9.929S 159.139E 10.0 6.4 6.9 883 254.2 27 19911014 15:58:14.1 9.0875 158.471E 33.0 6.2 7.1 88.1 255 3 I

q i

1

~

13 I

28 19901230 19:14:18.9 5 0975 150.967E 178.6 6.6 90.6 263.4 I

29 19900405 21:12:35.6 15.125N 147.596E 11.4 6.6 7.5 78.8 2803 I

43 19920718 08:36:59.2 39383N 143388E 33.0 5.9 6.8 65.6 302.1 11 3

30 4 19920313 16:01:033 52.741N 178.885W 181.0 6.2 36.8 303.1 M

i 32 19901106-20:14:29.7 53.452N 169.871E 24.8 6.3 7.0 43.0 306.7 m

41 19920713 15:34:05.5 51.159N 157.675E 54.0 5.7 51.0 3073 m

i 33 19900512 04:50:08.7 49.037N 141.847E 605.7 6.5 60.6 311.4 I

f 34 19910221 02:35:34.0 58.427N 175.450W 20.2 6.2 6.5 34.2 312.7 m

35 19910308 11:36:28.4 60.904N 167.023E 13.0 6.4 6.6 42.8 317.9 I

i 45 19920819 02:04:36.9 42.096N 73.540E 24.0 6.8 7.5 92.4 347.5 1

{

.i 36 19910714 09:09:11.9 36334N 71.119E 212.9 6.4 98.0 3483 I

1.

I: both low-pass 0.2 and 0.5 Hz data were used.

l II: only low-pass 0.2 Hz data were used.

III: only low-pass 0.5 Hz data were used.

2.

Events in each of the following groups were stacked: 50 and 51; 06,08 and 09; 11 and 13; 18,19 and 39; 23 and 46; 24 and 25; 32 and 41; 45 and 36.

3.

Event was also used for station SIU.

4.

Event was also used for station CHE.

i 1

l 4

.I 1

e m

I

14' FIGURE CAPTIONS i

Figure 1. The location of the stations used in this study. COR is in Corvallis, CHE is southwest of Ponland, and SIU is west of Eugene. The location of the Trojan power plat is also indicated.

Figure 2. Simplified ray diagram illustrating the reverberations which may be generated by a P-wave incident at a boundary (Owens,1984).

Figure 3. a) Eanh model used to illustrate the effect of horizontal and dipping structure on j

the convened phases from the incident P wave. b) Azimchal variation in radial and transverse receiver functions. P repmsents the dimet P arrival; D and H denote Ps conversion from an eastward dipping and a horizontal interface respectively. (Cassidy, 1991)

Figure 4. Theoretical receiver functions as a function of event azimuth for a homogeneous cmst over a mantle and a dipping Moho. The model parameters are: dip = 15 E, hmoho l

av = 6.5 km/s, o y = 0.33. For all azimuths the event epicentral distance is

= 43 km, a a

90. The main arrivals are indicated. The waveforms were filtered with a low-pass filter with a cutoff at 0.2 Hz.

i Figure 5. In this study we use converted phases from incident teleseismic P waves to investigate the structure beneath the stations. The top figure shows in the plan view of the areal sampling of the structure amund COR by the principal convened phases Ps and PpPms for events with an epicentral distance of 90 from the station. The bottom figure shows the ray-paths in and east-west oriented cross-section. Note that due to the dip of the oceanic Moho to the east, the sampling area lies beneath and west of Corvallis.

Figure 6. The distribution of the earthquakes used in this study.

[

Figure 7. Example of the radial " receiver functions" (radial component of teleseismic P wave deconvolved by vertical component) for events 11 (red) and 13 (blue) shown in Fig. 2. The top traces are broadband; the traces below are low-pass filtrations with a progressively lower cutoff (2.0,1.0,0.5, and 0.2 Hz). The filter used is a zero-phase,3-pole Butterworth. The pulse at t-O is Pp, followed by other converted phases and reverberations. The broadband data are dominated by high-frequency energy, indicating strong heterogeneity at shon wavelengths. The Ps conversion from oceanic Moho (7 s after Pp) dominates the signal at low frequencies. The fact that it is much less pronounced at high frequencies indicates that the Moho transition is not sharp. These two events are from a similar epicentral region and show very good coherence across the frequency band, indicating stability of the deconvolution procedure. However, because of the rather large sampling area (Fig. 5) by the convened phases and the complicated structure, our modeling procedures are inadequate for frequencies higher than 0.5 Hz.

Figure 8. The observed radial receiver functions for station COR (low-pass 0.5 Hz) plotted as a function of event azimuth. In the time frame of 0 to 20 s the data show good coherence.

I e

eh,,,

7NN uw A ;nses-

15 Figure 9. The observed radial receiver functions for station COR (Iow-pass 0.2 Hz) plotted as a function of event azimuth. In the time frame of 0 to 25 s the data show very good coherence. In this frequency mnge simple models should be adequate. The large positive arrival about 6-7 s after Pp is identified as Ps from the oceanic Moho, while the positive arrival at 16-19 s corresponds to PpPms. The rather large negative arrivals at about 5 and 15 s seem to indicate the presence of a deep-seated low-velocity layer.

Figure 10. Sensitivity of the receiver functions to a variation in the Moho dip; other model parameters are the same as in Fig. 4. O' mrved - red, theoretical - blue. On average, a dip of 10-15 fits the best, however, a wwer dip seems to be more appropriate for Ps, and a higher dip for PpPms. Since Ps sample under the station, while the largest-amplitude PpPms have the reflection points more than 20 km to the west, this may indicate that the Moho dip is smaller under the station, and increasing to the west.

Figure 11. Effect of the Moho dip and Poisson ratio on the receiver functions for four representative azimuths. The compressional velocity is 6.5 km/s, and the Moho depth for each case was adjusted so that the theoretical Ps would arrive at the time of the observed Ps. For this crustal compressional velocity the best fitting Poisson ratio is 0.33, and the dip is 10-15.

Figure 12. Acceptable solutions for various Moho depths, crustal compressional velocities and Poisson ratios. For the plausible comprminnal velocities (around 6.5 km/s), the Moho depth is about 40-43 km and the Poisson ratio is about 0.33.

Figure 13. (a, b) The best fitting model satisfying the low-frequency receiver functions at Corvallis. The compressional velocity structure of the upper crust (above 30 km) is based on refraction data. The receiver functions provide a constraint on the lower structure and the average Poisson ratio in the crust. The misfit for the planar model suggests an improvement shown in (c).

Figure 14. The fit to the low frequency data using a more detailed compressional velocity model of the upper crust based on the refraction data. Observed - md, theoretical -

black. This is the same model as shown in Fig.13, except that it does not involve a low-velocity zone above the oceanic Moho.

Figure 15. The fit to the low frequency data using model model shown in Figure 13.

Observed - red, :heoretical - black. Adding the low-velocity layer above the Moho improves the fit compared to Figure 14 especially for western azimuths.

Figure 16. The theoretical receiver functions for the final model (Figure 13) plotted in the same form as in the observed in Figure 9.

Figure 17. The theoretical higher frequency receiver functions (low-pass cut-off of 0.5 Hz) for the best-fitting low-frequency model. Many important features seem to be matched (see Fig. 8).

Figure 18. Data from CHE and SIU. Inadequate azimuthal coverage precludes detailed modeling of these data. If we assume the same average crustal compressional wave i

a

16 velocities as for COR, the arrival times of Ps and PpPms indic.ste the Moho depth of about 47-50 km under CHE, and about 43 km under SIU.

i 1

I i

I 5

I

~

a p

0 i

-. j 9

l

)

l O

a

.en-mem y - mm.. -hww..

m.

}

a -

2.

2 h

17 l

I I

~,.. si...

]6

@ T#*i**'

t ii

)

1 l

I I

f I

l I

3 e cac 1

i a-I i

3 I

i i

1

}-

t s c I "!

i con h

l 6

1,

o

=

t ti i

ii l

t I

0 f 1 1

i-'

3 u

f.

}

e scu e rwo l

g j

I A

1 e

Neweeny h one t

I e

n

(

I g

1 A c, t., uw I

f 1

e Figure L 1

l l

- 1

- I t

l l

' l i

m-

,g---

LJ h -

wrW k6N muuma - -

• EE

18 Converted Ray Phase Diagram Seismic Station E

- - - S-wave

/

\\

/

V, - 6.5 km/s P-wave

/ / \\\\

/

V. - 3 8 knvs

, ;,\\\\

P

• 2 0 GICM'

, Ig

\\\\ g

\\

/

I Moh 30 km

////

1. EI V, - 8.0 km/s

[#,44,e 4 4'

g',-

V. - 4.6 knvs

• o,

y 3

p - 3.3 g/cm After Owens,1984 i

Figure 2.

=%

me w ri+ %,

yg s, df

Seismic g Station g

w 1

20 km V,

v.

p 3

km/s km/s g/cm

{

5.0 2.89 2.53 20 km Strike - 0' 6.5 3.76 2.80 Dip - 15*

8.0 4.63 3.30 O

ijiiijiiiji jiii;iiiji i

Radial Transverse P

A

^D H

BAZ D

H p

^

000

^

v AA A

043 v

I

^

090 A

^

135

^

y

^

^

180

^

v

^

225

^

^

y

^

270

^

315

^

^

y b

.i.,,i,..I, i.,,1,,,i, 0

4 8

0 4

8 Eme (s)

Time (s)

Figure 3.

s w.

_--.-.h

-e=m

m 5

5

.a c

,0 5

u gn "0

,45 9

=

A 3

,0 4

3 0

=

o

,5 3

E

'5 1

,0 3

=p s

i 7-m d

h

_ 47

~ G_=

7 5

7 S

2 a

y p

l

)

s 7

P s s

s

(

m n

h*A

• A 1

,0 e AiP 2

o m

o i

T P

t i

cn u

,5 p

f m

1 r

e 7

Z Z

^

A P

v p

'bA^A*

i

,0 e

T P

c 1

a gg e

r N

s T-

,5 l

a P

i d

r

-E T

- E -

- E

- {

la p

c

, 0 P

i te r

o E

E

]

e h

, 5 T

7 0

_7 7

i

,1 5

1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 4 2 6 4 2 0 8 6 4 2 0 8 0 4 2 0 8 6 3 3 3 3 2 2 2 2 2 1 1

1 1

1 g$ 3$]So$

1

21 30 km A:90" 20 Top view 10 COR WO ps B

E l

i i

i i

i i i

-20i l

l l3 i

8 3

i i

g 8

i i

i 6

v&

v r

-44.5

-33.0

-21.9

-I 1.7 -3.6 6.2 km S

~I 3 COR W

g E

i i

Side view 42km i

_. _ _ _ /5 *

/P.Pms

__1_

Ps k

slab Ps PpPms Figure 5.

a

3 l

6 e

r

)

s u

ig F

e

' i 1

0 o

f 0

~

0 00 05 5

0 1

s 3 0 0

30 0

1 2 0 v0 1-50 9

1 b

5 e

' w

'.~6 0

9 40 1

C 40 0

6 5 0

3 d.

R5 g

E0 T 1 N

E C 0

/

00

.L 0

. /'"*

A 5 H 0 T

5 1

.Y-I U

M ZS AEL G

v I

N N OA

/

T g

y

\\

CW EO J D O N RI ff PW

i l

e E

a o

a a

de P

~

~

0 r

4 e

7 t

h e

l i

r f

h u

s V

g s

i

/

5 F

i a

0 I

t t

f N

P p

b w

)9 o

O l

d t

/

i i

t i

N n

AM a

d t

n I

a N

a l

9 5

i I

b 9

d 4

\\

a 1

o 4

t I

I i

N r

O I

b 0

8 s

)

t N

s n

(

e o

v i

V i

I Nt 5 e I

e 4

m 1

/

h i

o h

T w

t m

l h' 1

I I

I 0

1 o

r f

sn i Ni I

f i

b o

h i

tcnu f

I i

I I

b h

re f

v i

c t MI l

I 1

h e

i e

r M

la i

d 0

a

~

~

1 R

eg 5

5 8

cr 0

2,'

^

u

. 5 g

e_*m

~ W~_

i 4

~

~

i F

1-

^

i 4 5

h m-

~

^ ^*c

- w~

. 3 0

^

^

I

^

5 C

h a

y_m__L~

~

- I e

. 0 3

t a

H As ___' " ^

d M

. 2 z

s AA.W^~-

N h~

1 1

5 c-5

)

0 s

(

s 0

ie P

2 m

ap iT o

e

__ 4 f w

C

Y~ m__

~_,-

L 5

^

l d

_ n

^

^mA A- "+-

4*g.

i 1 R

O 0

c =-

^

^

24 -

i5 A

A i0

^

.D

~-

4~O'G ^m #

,5 E-

, 1 S

E

~

0 m

5 1

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 2 6 4 2 0 8 0 4 2 0 8 6 0 8 6 3 3 3 3 2 2 2 2 2 1 1

3 1

G$ c.

E.y$ao t

t U

5 5

9 0

e

,q ru g

i F

5

^

k

~

O N

~

~

~

5

~

~

~

7_

7 AA1 a

~-

0 t

a

^

d z

T-H

~

5 1

_7 2

-w

~

)

0 s

o A a g~M"^_~

A

(

s N

s O e a

m p

i T

w 5

i 1

l R

Z O

7

-C

_g~

0 7

1

~

_ ; E ~

b

~

g-A A

b

~

5 0

1 5

- ~-

1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 4 2 6 4 2 0 8 0 4 2 0 8 6 0 0 6 3 3 3 3 2 2 2 2 2 1 1

1 0f gyQSO$

60

' 26 A

360 -

r?1 o00 -

, =-

1 r

e 240 -

wg f*4 w

J 180 -

Q n

120 -

60 60 40 60,

60 180 -

A_

y 120 -

~~

~~

60 60 f

O g

5 180 -

^

^

0 120 -

P C~ _ - f --

m 60

, 10 -5 0

5 10 15 20 25 30 35 40 45 50 55 Time (s)

Figure 10.

0 s

5 1

3 5 5

1 3 5 s

s 2 3 3 3 b 2 3 3 3 b 1

0 0 0 0 o

0 0 0 0 o

i 0

e i

, 4 r

n g

i 0

F

, 3 I

, Ad_/_Q-

%%h i 2 e

• 8

)

4 s

3 0(

=

ZA m

i 1 I

\\_

0i T

\\

. 0 0

4 p~

. 4 0

2

. A-3 3

)

6 s

=

0(

. 2 e Z

m A

I 0i I

i 1 T i 0

)

0 i

i 4 iiV_

i 2 e 0

i 3 2

)

6 s

1 i

0(

=

Z A

m 0i II i

. 1 T i

i 0 0

i-i 4 i y-i 0

e 3

8

)

1 s

1 i

0 (

Z i 2 e A

m I

0i I

i i 1 T i

i e 0

,yg g

1_

l i

28 56 54 -

52 -

50 -

48 -

E o=0.31 6 46 -

5 c.

ma 44 -

o=0.32 42 -

c=0.33 40 -

38 -

36 -

i 6.3 6.4 6.5 6.6 6.7 6.8 6.9 7.0 7.1 7.2 7.3 P-velocity (km/s) 1 Figure 12.

O 3

1 er i

i R

g Om F

i r

C

)c 02

(

0 i

8 0

)

~

i s 7

/

(L m0 3

k i

6 3

(

p 0

V 0

=

i r

5 c

0

)

i 4

m

)

k b

(

~

h

(

t 0

5 0

5 0

5 0

p 2

2 3

3 4

e 1

1 d

s

/

m k

5 7 7 5 3 6

5 3

4 6

7.

9 5

1 1

5 6

6 0

6 6

7 6

8

=

p V

m k

R O.

0 0 0

0 0

0 1

O5,5 3

3 5

5 4

9 F

C

=

h O

/

)

a

(

g 5

5 4

1 er 0

u 5

g

' v~

^ 2

~

i F

^==

v m^

5 k

7 0

k k

_^

~ -

m

~

5 3

0 Z Y 4

=

h 0

v_

^ m^ -

3 0

N 1

=

p V

M 5

Z

- vvv

~

i d

2

)

s

(

0 e L

@x N'V 2

m tu i

o T

h t

5 i

1 w

d

' e_

e

(

l o

-3 0

1 n

~-

m l

a b

iF b

~ =

4

=

i h

~__r S

^

-= _

0 N

1 5

.1 0 0 0 0 0 0 0 0 0 0 0 o 0 0 0 0 0 6 4 2 4 2 0 8 6 4 2 0 4 2 0 8 6 3 3 3 2 2 2 2 2 1 1 1

^

a g$ @hm

e

~~~

5 5

5 1

e 0

r s

5 u

~

g L^

i

^

F

~

v

^

5 O

h vU-m^ '-

2

~

0 h

h

' v~x^^m m

^

w T

5 k

3 r Ce'~ V<e 04 n

=

2 0

3 0

1

=p 5

i C vv d

)

s Z

N

(

V Ax@

T 0 e L

2 m

h i

t i

w 5

le 1

m N~

d o

l 0

F 1

a n

i b

b M~

=

1 =

=

^

5

- ^_r b

0 1-5 1

0 0 0 0 0 0 0 0 0

0. 0 0 0 0 0 0 0 0 0 6 4 2 6 4 2 0 8 6 4 2 0 8 6 4 2 0 8 6 3 3 3 3 2 2 2 2 2 1 1

1 1

1 v Eh yhmoo

Final model with LVZ, dip =10, h=40 km, LP.2 60 40 -

20 -

A-

^

360 -

__ A_+

- h A

320 -

^ 300 -

^

"D 280 -

v c 260 -

~

$ 240 -

[

5E y

3 220 -

~

$ 200 -

$ 180 -

160-

~

140 -

C-n 120 -

=

2 22 ---

80 -

60 -

-15

-10

-5 0

5 10 15 20 25 30 35 40 45 50 55 g

Time (s)

Figure 16.

?'

l' i

1 g

5 5

0 7

i 5 1

eru g

i F

i 5 4

0 4

5 PL 53 m

k i 3 0

0 4

=

{

_ - b h

i 5 0

y"_y 2

1 y

z@

=

)

p g

^

s

(

i d

g

_ _b

• ^

i 2 0 e 2

m Z

A

^- " A Aa w

V iT L

= x i 5 h

~

=

g=

=

1 t

i w

7 2-l d

=

i 1 e

0 o

l A

4;_"

^.r AA

_ ~

m la 7

b_A p=

==

5

~

=

~

n j

=

i

^

^

F 0

7

.5 -

0 1

5 1-0 0 0 0 0 0 0 0 0 0_ 0 0 0 0 0 0 0 0 0 6 4 2 6 4 2 0 8 6 4 2 0 8 6 4 2 0 8 6 3 3 3 3 2 2 2_ 2 2 1 1 1 1 1

^

a* c.

_j a

.$$sD(

l l

~

34 Low-pass 0.2 Hz data for stations COR, CHE, and SIU 60 360 -

300-

^

^

~

1. ~,

e

^

240 -

SIU 180 -

120 -

^

^

^

60 -

60 7

360 -

300 -

^ --

A m

i d

240 -

M vgE

- CHE i

i 180 -

120 -

i t

60 -

60 0-COR 7 180 -

^

u t

120 -

i

~r^_ -

60 -

, 10 -5 0

5 10 15 20 25 30 35 40 45 50 55 i

Time (s)

Figure 18.

I i

I

.. -.