SAFT Project Proposal

I. Introduction

Subduction zones are perhaps the most important features in global geodynamics, since they represent the primary mechanism by which surficial materials are recycled back into the mantle. The mineralogical reactions that occur within a subducting slab are a key component of this system, since they may control a wide variety of processes, including the depths at which volatiles are released [Tatsumi, 1989; Thompson, 1992; Wood, 1995], the occurrence of intermediate and deep earthquakes [Abers, 1996; Green and Houston, 1995; Kirby et al., 1996a; Kirby et al., 1996b], and the distribution of density anomalies and stress within the slab [Bina, 1996; Yoshioka et al., 1997]. Despite this importance, fundamental questions remain about mineralogical reactions in the vicinity of slabs, particularly the basalt to eclogite transformation in subducted crust [Gubbins et al., 1994; Hacker, 1996], and the mineralogical transformations at the 410 and 670 km discontinuities [Bina and Helffrich, 1994; Stixrude, 1997]. Although high pressure experiments have helped us to understand these reactions, experiments have been unable to address key questions about the kinetics of these reactions in the Earth, and the behavior of these reactions in real systems containing impurities and volatiles. Thus primary seismological observations regarding these mineral transformations will be of great importance in the years to come.

In this proposal we request funds to deploy two arrays in the Southwest Pacific to study mineralogical transformations in the subducting Tonga slab. In the first part of the proposal we will discuss the use of an 11 element array in Tonga to study the interaction of seismic phases with the intermediate depth portion of the slab. We propose to use the array to disentangle the effects of multipathing, converted phases, and possible waveform dispersion related to the velocity structure of the subducting oceanic crust, and provide constraints on the basalt to eclogite transformation. In the second part of the proposal, we discuss the use of an array in Fiji to study the interaction of converted phases from local slab earthquakes with the 410 and 670 km discontinuities. The array will be used to increase the signal to noise ratio for these observations and to allow detailed mapping of the elevation and thickness of these discontinuities near the slab.

There are several reasons why the Tonga subduction zone represents the optimum region to study these effects:

1. High seismicity rate. The Tonga slab contains about 66% of the world's deep earthquakes, and also shows one of the highest seismicity rates at intermediate depths, with earthquake activity at all depths throughout the upper mantle.

Table 1: Annual seismic events per degree length of subduction zone

Region	0-100 km	100-200 km	200-300 km	> 300 km
Tonga-Kermadec	2.20	0.28	0.19	1.37
Marianas	0.44	0.12	0.05	0.10
Izu-Bonin	0.36	0.06	0.02	0.35
Japan	1.29	0.08	0.02	0.11
Aleutians-Alaska	0.62	0.04	0.01	0
Cascadia	0.07	0	0	0
Central America	0.84	0.04	0.02	0
Peru-Chile	0.48	0.22	0.06	0.08

Note: Events are those with seismic moment > 10²⁴ dyne-cm, over the 20 years history of the Harvard CMT catalog. Although we would be able to use events smaller than this size, the number of smaller events is generally proportional.
2. High subduction velocity. The Tonga slab shows a subduction rate of 21 cm/yr [Bevis et al., 1995], which is by far the fastest subduction velocity of any subduction zone. This fast subduction of old oceanic lithosphere ensures that the slab is cold, so that kinetic effects on mineral reactions will be more pronounced.
3. Ease of logistics. We have previously operated a smaller experiment in the Tonga-Fiji region, and have worked out many of the logistical details. We have excellent support from local scientific organizations. The timing of the deployment coincides with the existing Japanese SPANET deployment (1997-2002), and we have arranged a data exchange agreement with Dr. Daisuke Suetsugu, which will allow us access to data from six broadband stations in the region.

Although we are proposing dual arrays in Tonga and Fiji with somewhat different goals, both will be operated simultaneously as one project. One reason for this is that simultaneous operation of the arrays allows both to be used to locate earthquake sources, thus greatly improving event location, which is a significant uncertainty when using a single array. Additionally, it is much cheaper, in terms of both money and personnel time, to operate both arrays as a single project. The additional costs of operating a second array in this part of the world is not large, relative to the scientific benefit.

We operated a highly successful deployment of 11 broadband seismographs in Tonga-Fiji during 1993-1995 (see results from prior funding). That deployment was designed to study the large-scale structure of the region, and employed seismographs with an average spacing of several hundred kilometers. The large-scale velocity and attenuation models developed during that study [Roth et al., 1998, Koper, 1998; Xu and Wiens, 1997; Zhao et al., 1997] will be useful in the current project, which is designed to use much tighter arrays (spacings of 5-50 km) to study the details of the complex body waveform interactions with the subducting slab and other mineralogical boundaries.

II. Structure of the intermediate depth slab and the fate of oceanic crust

Previous work

Mineralogical reactions are expected to occur as a result of the increased pressure and temperature as material descends into the subduction zone. These reactions include the dehydration of oceanic crust and sediments [Anderson et al., 1978; Delany and Helgeson, 1978], and transformations associated with the basalt/gabbro to eclogite reaction [Hacker, 1996; Ringwood, 1982]. These reactions are fundamental to the formation of island arc magmas within the mantle wedge [Davies and Stevenson, 1992; Tatsumi, 1989]. These reactions, combined with a complicated thermal structure [Peacock, 1996], ensure that the seismic velocity structure of the top of the downgoing lithosphere and the mantle wedge above it are highly complex.

Seismological constraints on the structure of downgoing crust have generally involved either phases that convert at the interface, or phases that travel along the strike of the slab as guided waves. Observations using a variety of methods in different regions have produced a striking divergence of opinion about the state of subducting crust. One set of observations detail the observation of high frequency precursors at sites in New Zealand from earthquakes in Tonga-Kermadec [Ansell and Gubbins, 1986; Gubbins and Snieder, 1991; Smith et al., 1994; van der Hilst and Snieder, 1996]. These observations are interpreted as indicating fast propagation of a phase within a thin high velocity layer which may represent oceanic crust transformed to eclogite [Gubbins et al., 1994]. However, at other subduction zones such as Japan [Iidaka and Mizoue, 1991] and Alaska [Abers and Sarker, 1996], no high frequency precursor is seen; instead, a low frequency phase arrives first. Dispersion calculations for Alaska suggest that the waveforms are consistent with a thin low velocity layer at depths of 100-150 km [Abers and Sarker, 1996]. At other locations in Japan, two arrivals are observed, with the later arrival consistent with a guided phase within a thin low velocity layer at the top of the slab at depths down to 60 km [Fukao et al., 1983; Hori, 1990].

The velocity structure at the top of the slab can also be inferred from converted phases. ScSp conversions in Japan are interpreted as indicating a low velocity layer at the top of the slab [Helffrich and Stein, 1993; Nakanishi et al., 1981]. In addition, PS converted phases in both Japan [Matsuzawa et al., 1986] and the Aleutians [Helffrich and Abers, 1997] provide strong evidence for a thin low velocity layer at the top of the slab to depths of at least 150 km.

Observations in Tonga

It might be thought that P-wave observations in Tonga would be similar to those in New Zealand, farther to the south along the same slab. However, a careful search of records from all stations in the 1993-1995 Tonga deployment by Gideon Smith, who carried out much of the New Zealand work, has disclosed no records with high frequency precursors, despite the fact that such precursors are also observed at AFI in Samoa to the north. Instead, most waveforms show low frequency initial arrivals, followed by a second high frequency arrival within 1-3 s of the first motion, similar to what has been observed in Japan and Alaska.

Although waveforms of this type found elsewhere have been modeled as dispersion effects from low velocity subducted crust [Abers and Sarker, 1996], we find evidence that these waveforms in Tonga are caused by multipathing. Figure 2 shows arrivals from along the main line of ocean bottom seismographs (see Figure 1 for locations), from an intermediate depth earthquake about 1000 km to the south. Ocean bottom seismographs to the east of the trench (OBS-30 and OBS-28) clearly show an initial low frequency arrival and a later high frequency arrival with different ray parameters. Other events where these arrivals propagate along the strike of the OBS line show that the high frequency arrival propagates horizontally in the oceanic lithosphere with a velocity of about 8.3 km/s, whereas the low frequency arrival shows the expected ray parameter for a PREM earth structure. The high frequency arrival shows systematically earlier arrival times (relative to P) as one goes eastward, with a continuous progression across the arc and into the backarc. In addition to the differences in ray parameter for the high and low frequency arrivals there is also a difference in back-azimuth for the arrivals. Polarization analysis of our 3-component data shows that the low frequency arrival is at the correct station-event azimuth; the high frequency arrival however arrives sytematically deviated from this path. It thus appears that the high frequency arrival represents a different raypath, perhaps a late arriving guided phase along the slab, rather than dispersion. If the two arrivals were caused by dispersion, one would expect the greatest dispersion for stations near the trench axis, whereas these stations show high frequencies arriving very shortly after the initial P phase. We interpret this phenomenon as multipathing between a high frequency guided wave and a lower frequency wave that follows a more conventional path.

Another set of provocative observations from our initial experiment concern secondary arrivals, primarily between the first arriving P and S phases. Some of these arrivals are extremely large, and are not easily interpreted as conventional P to S or S to P conversions at the upper slab interface. An example of one of these arrivals is shown in figure 3; the phase is very strong at station EUAT. This phase does not resemble PS or SP conversions from the top of the slab at other subduction zones [G. Helffrich, personal comm.]. The particle motion is nearly circular within the horizontal plane, and the motion is emergent and shows some evidence of dispersion. We have collected a significant number of these observations, but understanding of this phase is limited by the large station spacing of the previous deployment; as can be seen from figure 3, there is a large spatial variability (station NUKU, with only a suggestion of the phase, is only 30 km from EUAT). We suggest that understanding and modeling of this phase and other secondary phases we have identified will provide significant constraints on the velocity structure at the top of the downgoing slab and in the mantle wedge.

Proposed array observations

We propose to use a tight array of seismographs on the main Tongan islands of Tongatapu and 'Eua to identify the ray parameter and backazimuth of the phases and disentangle the very complex set of arrivals observed in the Tonga arc. In particular, this array will allow us to distinguish between multipathing and dispersion. In addition, we propose to use the array to identify various later phases, such as PS and SP conversions from the slab interface. Station spacing in the array will range from about 5 -15 km, and the dimensions of the array will be about 30 km x 50 km. For reasonable ray parameters, this gives a time difference of 3-8 seconds across the array, sufficient to determine the ray parameter and backazimuth at frequencies of 0.5 - 2 Hz.

Data for suitable earthquakes will be analysed with the array processing package in DATASCOPE. P- and S-waves will be stacked in limited frequency bands, and the backazimuths and slownesses used to distinguish arrivals. Ray tracing and finite difference modeling using slab models determined in our initial study will be used to interpret the results.

Waveform modeling of slab effects using finite-difference and reflectivity synthetics

We have made extensive tests with synthetic data to demonstrate the ability of the proposed arrays to study the converted, multipathed, and discontinuity phases as discussed in this proposal. To perform these tests we computed synthetic seismograms using 2D finite difference and reflectivity methods using the actual geometry of the array statsions proposed here. To make our simulations more realistic we incorporate random noise time series, observed during our previous Tonga-Fiji experiment, and process the seismograms using the techniques we plan to use in the actual experiment. In the real experiment it is normal to expect less than 100% operation of the instruments. In this preliminary feasability test we therefore only incorporate seismograms from a subset of our proposed stations to demonstrate that our experiment should be successful even when data is not captured across the whole array.

Finite-difference synthetics:

We propose to use elastic finite difference methods to investigate unusual late or converted arrivals, and the possible effects of frequency dispersion from a layered slab. The distinct advantage of finite difference methods is their ability to completely describe wave motion in media with any complex spatial variation of elastic properties. We have implemented the elastic finite-difference method of [Pitarka et al., 1994] to simulate P-SV and SH-wave propagation in a subduction zone setting. This method is similar to that of [Vidale et al., 1985] and can simulate 3-D P-SV and SH wave propagation for a double-couple point source in a 2-D heterogeneous structure.

Figure 4 shows the application of this finite difference method to Tonga. The waveform simulation uses a model including the subducting Tonga slab, as shown in Figure 6 (left). For the velocity model, we used a slab model which is based on the inferred thermal and mineralogical characteristics of the Tonga slab, and is consistent with observed Tonga traveltimes. This model was developed by a recent Washington University Ph.D. Keith Koper. We also placed an eight km thick low velocity crustal layer at the top of the downgoing slab, and examined arrivals at the array from intermediate depth earthquakes in the slab. The structural model is shown in Figure 4 (left).

Waveforms interacting with the low velocity crust produce distinct secondary arrivals in the seismograms (figure 4), although these arrivals are not generally identifiable on individual seismograms. The array allows the data to be transformed to the τ???p domain, where the arrivals can be distinguished by their arrival time and ray parameter. In particular, we observe an S to P slab conversion with a ray parameter of about 0.02 s/km following the P phase, a P to S slab conversion with a ray parameter of -0.10 shortly before the S phase, and an S-wave reverberation off the top of the subducting crust at large negative ray parameters (-0.15 s/km) after the S arrival (figures XXX1 and XXX2). The amplitude of such reflections and conversions in the real data will place important constraints on the mineralogy of the subducting crust.

Reflectivity synthetics:

In Figure 6 two later phases are clearly visible between P- and S-waves. The first later phase comes about 5 s after the first P-wave, which is considered to be the S to P converted wave at the boundary between the low-velocity layer and the mantle wedge [Matsuzawa et al., 1986]. The second later phase comes about 10 s after the first P-wave, which is believed to be the P to S converted wave at the slab boundary [Matsuzawa et al., 1990]. The amplitudes of the later phases change when the source location, mechanism, and details of the structural model change. These different phases can be seen in Figure XXXX. This shows the computed synthetics for the Fiji array, and added noise for the same event and CMT focal mechanism. The resulting synthetics show the mantle discontinuity, but it is not easily identified on individual records, particularly since it arrives only shortly after crustal reverberations in the P-wave coda. However, the discontinuity phase can easily be identified in the p domain by its delay and ray parameter (figure XXXX). In fact, another mantle discontinuity phase (P660P) is also observed. The discontinuity phases can be distinguished by their relative ray parameters - the ray parameter of s400P is larger than the direct S-wave (it arrives with a lower incidence angle), whereas the ray parameter of P660P is smaller than direct P (see proposal figure 5 for diagram). Crustal reverberations have a ray parameter similar to the direct P and can easily be distinguished from the mantle discontinuities. These phases could not be sorted out without an array such as the one proposed.

This first example has a near optimal focal mechanism, since the P arrival is nodal at the array. We therefore investigated another case with a less optimal focal mechanism (figure XXXY). The depth in this case is also 50 km deeper, such that the P660P arrives prior to the s400P. These discontinuity phases, although fainter, are also easily identified by the array stacking procedure.

Once the discontinuities have been identified using the arrays, the arrivals can be stacked to enhance the signal to noise ratio, and the frequency content of the mantle discontinuity phases compared to the direct P-wave to infer the sharpness of the discontinuities. Locally observed P phases from the discontinuities are the only phases that can resolve the sharpness of the mantle discontinuities on a 1-3 km scale, since these scales require observations at 2-3 Hz. The timing and amplitude of the phases will also be used to map out discontinuity elevation changes near the slab.

We propose to investigate the relationship between the observed multipathing and later arriving waveforms and the detailed structure of the subducting crust using finite difference modeling. Models will begin by including the known overall velocity difference between the slab and the mantle [Koper et al., 1998] and incorporating a thin layer of subducted oceanic crust. We will use this forward modeling approach along with a range of models for the subducted oceanic crust to develop insight into the waveform effects of various structures and to replicate the arrival times and amplitudes of the different arrivals.

III. Mantle discontinuity structure and the mineralogy of deep slabs

Mineral physics constraints

Upper mantle discontinuities are expected to show changes in elevation, velocity contrast, and sharpness near subducting slabs. The temperature dependence of the isochemical phase transformations from olivine to denser phases should cause changes in the elevation of the 410 and 670 km discontinituies [Akaogi et al., 1989; Bina and Helffrich, 1994]. In general, the 410 km discontinuity, having a negative Clapyron slope, should be uplifted near cold subducting slabs, whereas the 670 km discontinuity should show an increased depth. A decrease in the sharpness of the 410 km discontinuity may also be expected due to the thickening of the divarient phase loop at lower temperatures [Helffrich and Bina, 1994], and increased presence of water in the top part of the slab [Helffrich and Wood, 1996; Wood, 1995]. Conversely, kinetic effects [Solomotov and Stevenson, 1994] and the presence of untransformed phases [Stixrude, 1997] may cause a sharpening of the interface. It is clear that a careful seismological mapping of the elevation and thickness of the discontinuities as a function of position near a subducting slab could help to constrain several important factors, including variations in temperature and water content. The question of whether subduction zones transport large quantities of water into the mantle clearly has very important implications for the volatile budget of the earth and mantle mineralogy in general [Thompson, 1992].

Seismological observations

During the past decade, a large number of seismological studies have noted changes in the elevation of the 410 km and 670 km mantle discontinuities near subduction zones [Bina, 1991; Castle and Creager, 1997; Collier and Helffrich, 1997; Flanagan and Shearer, 1998; Niu and Kawakatsu, 1995; Revenaugh and Jordan, 1991; Roth et al., 1996; Shearer, 1993; Shearer and Masters, 1992; Vidale and Benz, 1992; Wicks and Richards, 1993]. These studies generally show an uplift of the 410 km discontinuity and a depression of the 670 km discontinuity in the vicinity of subduction zones, in agreement with the mineral physics data. However, most studies suggest greater topography on the 670 km discontinuity, in apparent conflict with mineral physics data which would predict greater topography on the 410 [Bina and Helffrich, 1994].

A number of studies have addressed the question of the sharpness of the mantle discontinuities [Benz and Vidale, 1993; Neele, 1996; Vidale et al., 1995; Yamazaki and Hirahara, 1994]. These studies have found that both the 410 and 670 km discontinuities must be narrow in depth, with most of the velocity increase occurring over a depth range of 4 km or less. This was somewhat unexpected, since the width of the divarient loop in the α-β phase diagram should be 8-19 km [Akaogi et al., 1989]. Several factors have been invoked to explain the sharpness of the 410 km discontiniuty [Helffrich and Wood, 1996; Stixrude, 1997]. However, it is not clear how much variability there may be in discontinuity sharpness, and no study has successfully mapped variations in discontinuity thickness near a subduction zone, where large variations in temperature and water content may be expected.

All of the previous studies have relied on teleseismic data, either long period precursors to phases like SS [Shearer and Masters, 1992], or short period converted or reflected phases [Vidale and Benz, 1992]. Although these observations have been of great importance, the teleseismic paths limit the ability of these phases to precisely map the discontinuity elevation and thickness. High frequencies are depleted in the teleseismic waveforms due to attenuation along the long path length, and the necessity of choosing larger earthquakes, with longer source durations. The lack of high frequencies is particularly a problem for thickness estimates, which rely on the frequency content of the reflected or converted phase. Since most teleseismic methods are limited to frequencies no higher than 0.5 - 1 Hz, thickness estimates of 5-10 km are generally minimum estimates, and the actual thickness is not resolved. These approaches also result in large Fresnel zones where the waves interact with the discontinuities, limiting the spatial resolution.

Proposed array observations of discontinuity structure in the Tonga slab

In this study we propose to use regional waveforms from deep Tonga earthquakes, recorded at an array of broadband seismographs in Fiji, to map out the elevation and sharpness of upper mantle discontinuities near the Tonga slab. This strategy has several advantages over teleseismic approaches. The large number of deep earthquakes in the Tonga-Fiji region provides many high frequency regional sources with a variety of paths.

If the 410 and 670 km discontinuities represent sharp discontinuities, prominent P to S and S to P converted phases should be visible between the main P and S arrivals on broadband records of deep Tonga earthquakes recorded at stations in Fiji. These phases will arrive at various arrival times and ray parameters, depending on the distance and depth of the earthquake (figure 5). We have identified several observations of these phases using our sparse 1993-1995 network, and an example is shown in figure 6. In this case a s410p phase is observed about 15 s following the main P arrival, and the identity of the phase is confirmed by polarization observations, indicating that the observed phase shows a strong linear polarization (figure 6, middle) with a ray parameter and backazimuth appropriate for the phase. The phase arrives at the time predicted by reflectivity synthetics (figure 6, lower) calculated for a standard earth model, suggesting very little elevation anomaly for the 410 at this conversion point. Comparison of frequency content between the 410 conversion and the direct P phase suggests that the highest frequencies (∼ 1 Hz) are absent in the conversion, consistent with a measurable discontinuity thickness of about 6 km. Identification of such conversions is facilitated by finding focal mechanisms which excite strong conversions relative to initial P-wave amplitude; this is possible in Tonga because of the large number of deep earthquakes.

We propose to study the sharpness of the discontinuities by modeling the waveforms and frequency content of the converted phases, as identified using an array of about 10 stations on the main island of Fiji. P-wave arrivals in Fiji commonly have energy at frequencies up to about 4 Hz, thus deep S-P conversions should have similar frequency content if the interfaces are sharp. The waveforms are sensitive to even fairly minor deviations from a sharp interface. The phases will be modeled using a finely discretized reflectivity model to determine the thickness of the discontinuities in various locations. Regional variations in the discontinuity thickness will be mapped, as well as regional variations in the elevation and velocity contrast.

The array will enable enhanced signal to noise, as we expect to be able to coherently stack for the conversions and reflections at frequencies down to at least 1 Hz. The aperture of the array, about 100 km, combined with the ray parameters for the converted phases, typically 0.05 - 0.15 s/km, give arrival time variations of 5-15 s across the array, such that ray parameter and backazimuth are readily determined.

Given the high rate of deep seismicity in the Tonga subduction zone, we expect that deployment of the Fiji array for a year will enable us to collect a large number of measurements of thickness, impedence contrast, and elevation of the 410 and 670 km discontinuities as a function of bouncepoint distance from the slab. This will enable us to investigate the dependence of discontinuity elevation and thickness on factors such as the temperature variation and possible release of water from the slab.

We also propose to use the Fiji array to investigate the possible presence of other discontinuities, such as those near 520 km [Shearer, 1996] and 210 km [Vidale and Benz, 1992] depth. Both discontinuities have been observed in the Tonga-Fiji region previously, but little is known about their thickness, impedence contrast, and elevation variation near the slab.

IV. Array observations of deep earthquakes

Since the Tonga slab contains 2/3 of the world's deep earthquakes the installation of seismic arrays in the Tonga region presents a unique opportunity to study the deep earthquakes in greater detail. The operation of two high resolution seismic arrays in the Tonga -Fiji region will allow us to extend our recent work on the configuration of active faults in deep slabs, the occurrence of repeating earthquakes in deep slabs, the variation of source parameters such as stress drop along these faults, and the prevalence of deep earthquake aftershocks.

Detailed fault geometry of deep earthquakes

Detailed analysis of records from the 1993-1995 Tonga deployment shows that deep earthquakes align along fault-like features of exceptionally high seismicity. Indications of these structures existed before [Giardini and Woodhouse, 1984], but we are now able to map these features down to a resolution of 1-2 kilometers. Waveforms from different earthquakes within these structures generally show a high degree of similarity. We have implemented a waveform cross-correlation location method [Beroza et al., 1995; Poupinet et al., 1984; Vidale et al., 1994] to resolve the detailed structure of these regions. The cross correlation method provides locations with 95% uncertainty regions of ± 1-2 km, about five times more accurate than standard arrival time picking. A view of the relocated seismicity of one highly seismic region is shown in figure X, along with planes denoting the CMT solution for the region.

The installation of the two new arrays will allow us to extend the waveform correlation techniques to smaller events. This will allow us to better define the geometry of the active faults in the Tonga slab, the recurrence relations of earthquakes on these faults, and the relationship of the faults delineated by small earthquakes to the rupture areas of larger events.

Repeating deep earthquakes

We also observe several cases of repeating earthquakes where the same section of fault is reactivated. Figure X showS-waveforms from repeating earthquakes in the Tonga slab recorded in Fiji which show exceptional similarity down to the very minor details of the coda. Location of these events with the cross-correlation method shows that the centroids of these events are located about 1 km apart along the fault plane. Deconvolution of these traces using an Empirical Green's Function method shows that the rupture dimensions of these events are on the order of 5-8 km, demonstrating that these events involve a recurrence of rupture at the same location. This observation is difficult to reconcile with the transformational faulting hypothesis for the mechanism of deep earthquakes [Green and Houston, 1995; Kirby et al., 1996a], which involves a reversible process. Empirical Green's function analysis of co-located events provide an excellent opportunity for high resolution studies of the temporal variation in rupture parameters such as stress drop [Hough, 1997; Vidale et al., 1994]. The proposed arrays will allow us to extend these studies to smaller events. Similar studies of shallow repeating earthquakes have provided important data on fault healing rates and the relationship between temporal earthquake occurrence and loading [Marone, 1998; Nadeau and McEvilly, 1999; Schaff et al., 1998]. We propose to compare our results from deep earthquakes with the shallow earthquake data to provide important constraints on the differences in mechanical properties between shallow and deep faults.

The prevalence of deep earthquake aftershocks

Prior to the 1994 Tonga and Bolivia deep earthquakes, it was thought that deep earthquakes showed very few aftershocks, and that this represented a key difference between deep and shallow earthquakes [Frohlich, 1989]. It now seems that deep earthquake aftershocks are much more widespread than previously thought. The aftershock characteristics of deep earthquakes are variable, and show systematic variation as a function of the subduction zone, with Tonga providing the greatest aftershock productivity [Wiens and Gilbert, 1996]. A survey of moderate-sized intermediate and deep Tonga earthquakes during the 1993-1995 experiment demonstrates that all events with M_w > 6.0 exhibited some aftershock activity [Wiens et al., 1997]. However, determining the aftershock characteristics of Tonga earthquakes is hampered by extremely high noise levels and high detection thresholds. The installation of the two arrays will enable us to perform array detection procedures following any intermediate or deep earthquake, and provide more complete aftershock catalogs and accurate studies of statistical parameters such as the aftershock temporal decay and b-values. This will allow us to address key questions such as whether these parameters indicate any fundamental differences between intermediate-depth and deep earthquakes, and what factors control the variability of aftershock occurence in the Tonga slab.

V. Proposed work and logistics

Instrumentation and station distribution

We propose to install 25 seismograph systems in the Southwest Pacific and record continuously for one year. The reason we chose the Tonga-Fiji region for this experiment is primarily the extremely high seismicity rate of intermediate and deep earthquakes in this region; given the high seismicity rate, a one year duration is sufficient that we can expect good earthquake sources in all regions of the slab.

For our study we require a good array of sources deeper than 100 km, encompassing a variety of source depths and approach azimuths to the sensor array, in order to ensure that phase conversion interfaces and discontinuity depths and sharpnesses can be mapped out in an experiment of only one year duration. The number of intermediate and deep earthquakes in Tonga per length is four to five times larger than any other subduction zone, such that experiments of four years duration or longer would be necessary at those sites to obtain the source distribution that we need.

Most of the seismographs will comprise two arrays; a tight array in Tonga, with average station spacing of about 10 km, and a less dense array with average station spacing of about 40 km on the main island of Fiji (figure 7). We propose to use a mixture of traditional broadband instruments and newer "semi-broadband" or intermediate period instruments. The broadband instruments will provide recording out to longer periods with multiple data streams for more widely dispersed sites, whereas the "semi-broadband" instruments are smaller, more readily available from PASSCAL, and well suited to more closely spaced sites where recording at periods greater than ∼ 30 s is not necessary.

Most of the seismographs in each of the arrays will be Guralp CMG-40T instruments coupled with REFTEK 3-channel 24 bit recording systems. At very quiet sites the 40T instruments do not resolve ground noise at frequencies of less than about 0.07 Hz, but at noisy island sites there is probably no difference between the 40T and traditional broadband instruments out to much smaller frequencies. Since the purpose of the dense arrays is to study body wave energy at frequencies of 0.1 - 10 Hz, these instruments are ideal for the array deployments and offer greater portability without the problem of a 1 Hz corner, as would be the case for L-4 sensors.

In addition to the main Fiji and Tonga arrays, we will install several stations on outlying islands to help with locating smaller events and to comprise a larger, sparse outer array. These stations will have complete broadband systems including Streckeisen STS-2 instruments and REFTEK 6-channel 24 bit recorders. In most cases these instruments will reoccupy sites that we instrumented in our sparse 1993-1995 experiment, so little site work will be necessary.

There are significant advantages to carrying out the experiment in the 2000-2001 time frame, since the deployment will coincide with the sparse broadband Japanese SPANET (South Pacific Broadband Seismic Network) deployment installed by Daisuke Suetsugu, which was deployed in 1998 and will be deinstalled in 2001. These stations consist of Guralp CMG-3TEBB sensors and 24 bit data loggers. We have a cooperative relationship with Dr. Suetsugu, as we aided him in setting up contacts in Fiji and Tonga and the SPANET stations in Vava'u, Tonga and Labasa, Fiji use installations we built in 1993. Therefore, we will have access to SPANET data through a data sharing, enabling us to use SPANET stations and avoid the difficulty of maintaining stations on remote islands such as Niue. Most of the SPANET stations in the region are shown on Figure 7.

Table 2: Regional broadband stations from which we will access data

Array characteristics

Each of the two arrays has been designed based on the frequencies and ray parameters we plan to observe and the constraints imposed by island geography. The optimum aperature (and thus average station spacing) is generally a trade-off between the competing effects of better resolution for ray parameter and azimuth, which suggests wider aperature, and the desire for coherence, which generally favors smaller aperatures. Optimally, a signal will be coherent across an array such that it can be stacked, whereas noise will be incoherent and thus reduced upon stacking. Experience gained with the Tararua array in New Zealand is important for defining the optimum station spacing in an island arc setting. Figure 8 shows the correlation results for both noise and signal from the Tararua array [Smith, 1994; Stuart et al., 1995]. Figure 8a shows that 0.12 - 0.5 Hz noise is uncorrelated for station spacings larger than 5 km, whereas for the same frequencies, the correlation of the signal (figure 8b) is still well correlated for station spacing as large as 40 km. Figure 8c shows that station spacings of 10 km still show some correlation for a band of 0.5 - 1.0 Hz. Signals at frequencies higher than 2 Hz require station spacings no larger than 2-3 km for correlation, but the ray parameter and azimuth can still be determined for high frequency phases from the progression of the phase onset across the array [Smith, 1994], even if stacking cannot enhance the signal to noise ratio.

The Tonga array is designed to study the ray parameters and approach azimuths of various phases traveling up the low attenuation slab, with frequencies ranging from 0.2 to 5 Hz. A relatively small array is desired to allow stacking for phases in the 1 Hz frequency range, yet adequate aperature is needed to resolve ray parameter and azimuth. The array that we have designed has station spacings ranging from 5-15 km (figure 7), and dimensions of about 30-50 km, with the long direction roughly perpendicular to trench strike.

The Fiji array is located in the backarc, and arrivals are generally well attenuated, with peak signal to noise occurring at about 1 Hz. Arrivals are generally quite coherent relative to subduction zone settings, with even stations several hundred km apart showing good correlation at 0.5 Hz. We have designed this array with station spacings of 25-50 km, and anticipate that we will be able to stack arrivals to enhance signal in the 0.4 - 1.0 Hz range.

Deployment plan

Jim Fowler (PASSCAL program manager) estimates that instruments for this deployment should be available in the year 2000. This is considerably sooner than would be the case for a 25 instrument deployment using fully broadband/ 6-channel REFTEK instruments. We envision the following project timetable:

August, 2001: Trip to South Pacific to meet local investigators and find sites (1 person)
December, 2001: Installation of seismographs (4 people)
March, 2002: service trip #1 (1 person)
June, 2002: service trip #2 (1 person)
September, 2002: service trip #3 (1 person)
December, 2002: deinstallation trip (1 person)
2003-2004: write up results and forward data to the IRIS-DMC

The field personnel will consist of Douglas Wiens (PI), Gideon Smith (co-PI and postdoctoral associate), Patrick Shore (staff scientist and field deployment manager), and a graduate student. The deployment will be aided by the extensive experience we have acquired in installing and maintaining similar equipment at remote sites during the Southwest Pacific Seismic Experiment (SPASE) and the current Seismic Experiment in Patagonia and Antarctic (SEPA). Information on both projects is available at our web site. Gideon Smith also has extensive experience installing the Leeds Tararua array in New Zealand. A Washington University graduate student (possibly Stacey Robertson, who also has experience in the Antarctic project) will accompany the initial deployment and be available to make further trips if needed.

Each sensor will be placed on a small concrete slab cemented to bedrock (if available) at a depth of at least 0.5 meters below ground surface. The sensor will be housed in small plastic or fibreglass enclosures, and insulated with styrofoam. The DAS will be enclosed in an adjacent box above the ground surface. The system will be powered by two 12V marine deep cycle batteries, which will be charged by 30 W solar panels.

Data will be recorded at 40 sps onto 2 Gb disks. Stations will be serviced and these disks will be swapped at 3 month intervals, and data will be downloaded to tape at a field computer in Fiji, as in our last experiment. The 3 month service interval is adequate for obtaining a high data return rate; we used a 3 month interval on our last experiment and obtained an 85% return rate. Most of the data outages were caused by insufficient solar recharging during cloudy months (May-August); we plan to remedy this by using more solar panels this time. If spread-spectrum radio technology is available from PASSCAL by the time of this experiment we intend to investigate its use, but we are tentatively planning on using traditional site visits and disk-swapping for data return.

VI. References

Abers, G.A., Plate structure and the origin of double seismic zones, in Subduction Top to Bottom, edited by G.E. Bebout, D.W. Scholl, S.H. Kirby, and J.P. Platt, pp. 223-228, AGU, Washington DC, 1996.

Abers, G.A., and G. Sarker, Dispersion of regional body waves at 100-150 km depth beneath Alaska: In situ constraints on metamorphism of subducted crust, Geophys. Res. Lett., 23, 1171-1174, 1996.

Akaogi, M., E. Ito, and A. Navrotsky, Olivine-modified spinel-spinel transitions in the system Mg₂SiO₄ · Fe₂SiO₄: calorimetric measurements, thermochemical calculation, and geophysical application, J. Geophys. Res., 94, 15671-15685, 1989.

Anderson, R.N., S.E. DeLong, and W.M. Schwarz, Thermal model for subduction with dehydration in the downgoing slab, Journal of Geology, 86, 731-739, 1978.

Ansell, J.H., and D. Gubbins, Anomalous high-frequency wave propagation from the Tonga-Kermedec seismic zone to New Zealand, Geophys. J. Roy. astron. Soc., 85, 93-106, 1986.

Benz, H.M., and J.E. Vidale, Sharpness of upper mantle discontinuities determined from high-frequency reflections, Nature, 365, 147-150, 1993.

Beroza, G.C., A.T. Cole, and W.L. Ellsworth, Stability of coda wave attenuation during the Loma Prieta, California, earthquake sequence, J. Geophys. Res., 100 (B3), 3977-3988, 1995.

Bevis, M., F.W. Taylor, B.E. Schutz, J. Recy, B.L. Isacks, S. Helu, R. Singh, E. Kendrick, J. Stowell, B. Taylor, and S. Calmant, Geodetic observations of very rapid convergence and back-arc extension at the Tonga arc, Nature, 374, 249-251, 1995.

Bina, C.R., Mantle discontinuities, Rev. Geophys., 29, 783-793, 1991.

Bina, C.R., Phase transition buoyancy contributions to stresses in subducting lithosphere, Geophys. Res. Lett., 23 (24), 3563-3566, 1996.

Bina, C.R., and G.R. Helffrich, Phase transition Clapyron slopes and transition zone seismic discontinuity topography, J. Geophys. Res., 99, 15853-15860, 1994.

Castle, J.C., and K.C. Creager, Seismic evidence against a mantle chemical discontinuity near 660-km depth beneath Izu-Bonin, Geophys. Res. Lett., 24 (3), 241-244, 1997.

Collier, J.D., and G.R. Helffrich, Topography of the "410" and "660" km seismic discontinuities in the Izu-Bonin subduction zone., 1997.

Davies, J.W., and D.J. Stevenson, Physical model of source region of subduction zone volcanics, J. Geophys. Res., 97, 2037-2070, 1992.

Delany, J.M., and H.C. Helgeson, Calculation of the thermodynamic consequences of dehydration in subducting oceanic crust to 100 kb and 800 C, Am. Jour. Sci., 278, 638-686, 1978.

Flanagan, M.P., and P.M. Shearer, Global mapping of topography on transition zone velocity discontiniuties by stacking SS precursors, J. Geophys. Res., 103, 2673-2692, 1998.

Frohlich, C., The nature of deep-focus earthquakes, Ann. Rev. Earth Planet. Sci., 17, 227-254, 1989.

Fukao, Y., S. Hori, and M. Ukawa, A seismological constraint on the depth of the basalt-eclogite transition in a subducting oceanic crust, Nature, 303, 413-415, 1983.

Giardini, D., and J.H. Woodhouse, Deep seismicity and modes of deformation in Tonga subduction zone, Nature, 307, 505-509, 1984.

Green, H.W.I., and H. Houston, The mechanics of deep earthquakes, Ann. Rev. Earth Planet Sci., 24, 169-213, 1995.

Gubbins, D., A. Barnicoat, and J. Cann, Seismological constraints on the gabbro-eclogite transition in subducted oceanic crust, Earth Planet Sci. Lett., 122, 89-101, 1994.

Gubbins, D., and R. Snieder, Dispersion of P-waves in subducted lithosphere: Evidence for an Eclogite layer, J. Geophys. Res., 96, 6321-6333, 1991.

Hacker, B.R., Eclogite formation and the rheology, buoyancy, seismicity, and H₂O content of oceanic crust [H₂O], in Subduction Top to Bottom, edited by G.E. Bebout, D.W. Scholl, S.H. Kirby, and J.P. Platt, pp. 337-346, AGU, Washington DC, 1996.

Helffrich, G., and G.A. Abers, Slab low velocity layer in the eastern Aleutian subduction zone, Geophys. J. Int., 130, 640-648, 1997.

Helffrich, G., and C. Bina, Frequency dependence of the visibility and depths of mantle seismic discontinuities, Geophys. Res. Lett., 21, 2613-2616, 1994.

Helffrich, G., and S. Stein, Study of the structure of the slab-mantle interface using reflected and converted seismic waves, Geophys. J. Int., 115, 14-40, 1993.

Helffrich, G.R., and B.J. Wood, 410 km discontinuity sharpness and the form of the olivine α-β phase diagram: resolution of apparent seismic contradictions, Geophys. J. Int., 126, F7-F12, 1996.

Hori, S., Seismic waves guided by untransformed oceanic crust subducting into the mantle: the case of the Kanto district, central Japan, Tectonophysics, 176, 355-376, 1990.

Hough, S.E., Empirical Green's function analysis: Taking the next step, J. Geophys. Res., 102 (B3), 5369-5384, 1997.

Iidaka, T., and M. Mizoue, P-wave velocity structure inside the subducting Pacific plate beneath the Japan region, Phys. Earth Planet Inter., 66, 203-213, 1991.

Kirby, S., E.R. Engdahl, and R. Denlinger, Intermediate-depth intraslab earthquakes and arc volcanism as physical expressions of crustal and uppermost mantle metamorphism in subducting slabs (overview), in Subduction Top to Bottom, edited by G.E. Bebout, D.W. Scholl, S.H. Kirby, and J.P. Platt, pp. 195-214, AGU, Washington DC, 1996a.

Kirby, S.H., S. Stein, E.A. Okal, and D. Rubie, Deep earthquakes and metastable mantle phase transformations in subducting oceanic lithosphere, Rev. Geophysics, 34, 261-306, 1996b.

Koper, K.D., D.A. Wiens, L.M. Dorman, J.A. Hildebrand, and S.C. Webb, Modeling the Tonga slab: Can travel time data resolve a metastable olivine wedge?, J. Geophys. Res., 103, 30079-30100, 1998.

Marone, C., The effect of loading rate on static friction and the rate of fault healing during the earthquake cycle, Nature, 391, 69-72, 1998.

Matsuzawa, T., T. Kono, A. Hasegawa, and A. Takagi, Subducting plate boundary beneath the northeastern Japan are estimated from SP converted waves, Tectonophysics, 181, 123-133, 1990.

Matsuzawa, T., N. Umino, A. Hasegawa, and A. Tagaki, Upper mantle velocity structure estimated from PS converted waves beneath the northeastern Japan arc, Geophys. J. Roy. astr. Soc., 86, 767-787, 1986.

Nadeau, R.M., and T.V. McEvilly, Fault slip rates at depth from recurrence intervals of repeating microearthquakes, Science, 285, 718-721, 1999.

Nakanishi, I., K. Suyehiro, and T. Yokota, Regional variations of amplitudes of ScSp phases observed in the Japanese Islands, Geophys. J. R. Astron. Soc,, 67, 615-634, 1981.

Neele, F., Sharp 400-km discontinuity from short-period P reflections, Geophys. Res. Lett., 23 (5), 419-422, 1996.

Niu, F., and H. Kawakatsu, Direct evidence for the undulation of the 660-km discontinuity beneath Tonga: Comparison of Japan and California array data, Geophys. Res. Lett., 22 (5), 531-534, 1995.

Peacock, S.M., Thermal and petrologic structure of subduction zones (overview), in Subduction Top to Bottom, edited by G.E. Bebout, D.W. Scholl, S.H. Kirby, and J.P. Platt, pp. 119-134, AGU, Washington DC, 1996.

Pitarka, A., H. Takenaka, and D. Suetsugu, Modeling strong motion in the Ashigara Valley for the 1990 Odawara, Japan earthquake, Bull. Seism. Soc. Am., 84, 1327-1335, 1994.

Poupinet, G., W.L. Ellsworth, and J. Frechet, Monitoring velocity variations in the crust using earthquake doublets: An application to the Calavaras fault, California, J. Geophys. Res., 89, 5719-5731, 1984.

Revenaugh, J., and T. Jordan, Mantle layering from ScS reverberations: 2. The transition zone, J. Geophys. Res., 96, 19763-19780, 1991.

Ringwood, A.E., Phase transformations and differentiation in subducting lithosphere, J. Geol., 90, 611-643, 1982.

Roth, E., D.A. Wiens, P.J. Shore, and Y. Xu, Transition zone discontinuities beneath Tonga from ScS reverberations, EOS Trans. Am. Geophys. Un., 77, 472, 1996.

Roth, E.G., D.A. Wiens, L.M. Dorman, S.C. Webb, and J. Hildebrand, Seismic attenuation tomography of the Tonga-Fiji region using phase pair methods, J. Geophys. Res. , submitted, 1998.

Schaff, D.P., G.C. Beroza, and B.E. Shaw, Postseismic response of repeating aftershocks, Geophys. Res. Lett., 25 (24), 4549-4552, 1998.

Shearer, P.M., Global mapping of upper mantle reflectors from long-period SS precursors, Geophys. J. Int., 115, 878-904, 1993.

Shearer, P.M., Transition zone velocity gradients and the 520 km discontinuity, J. Geophys. Res., 101, 3053-3066, 1996.

Shearer, P.M., and T.G. Masters, Global mapping of topography on the 660 km discontinuity, Nature, 355, 791-795, 1992.

Smith, G., D. Gubbins, and W. Mao, Fast P-wave propagation in subducted Pacific lithosphere: Refraction from the plate, J. Geophys. Res., 99 (B12), 23,787-23,800, 1994.

Smith, G.P., Structure of the Tonga-Kermadec subducted slab from investigation of precursive seismic phases, Ph.D. thesis, University of Leeds, Leeds, UK, 1994.

Solomotov, V.S., and D.J. Stevenson, Can sharp seismic discontinuities be caused by nonequilibrium phase transformations?, Earth Planet. Sci. Lett., 125, 267-279, 1994.

Stixrude, L., Structure and sharpness of phase transitions and mantle discontinuities, J. Geophys. Res., 102 (B7), 14,835-14,852, 1997.

Stuart, G.W., D. Francis, D. Gubbins, and G. Smith, Tararua Broadband Array, North Island, New Zealand, Bull. Seism. Soc. Am., 85, 325-333, 1995.

Tatsumi, Y., Migration of fluid phases and genesis of basalt magmas in subduction zones, J. Geophys. Res., 94, 4697-4707, 1989.

Thompson, A.B., Water in the earth's upper mantle, Nature, 358, 295-302, 1992.

van der Hilst, R., and R. Snieder, High-frequency precursors to P-wave arrivals in New Zealand: Implications for slab structure, J. Geophys. Res., 101, 8473-8488, 1996.

Vidale, J., D.V. Helmberger, and R.W. Clayton, Finite difference seismograms for SH waves, Bull. Seism. Soc. Am., 75, 1765-1782, 1985.

Vidale, J.E., and H.M. Benz, Upper-mantle seismic discontinuities and the thermal structure of subduction zones, Nature, 356, 678-683, 1992.

Vidale, J.E., X.-Y. Ding, and S.P. Grand, The 410-km-depth discontinuity: A sharpness estimate from near-critical reflections, Geophys. Res. Lett., 22 (19), 2557-2560, 1995.

Vidale, J.E., W.L. Ellsworth, A. Cole, and C. Marone, Variations in rupture process with recurrence interval in a repeated small earthquake, Nature, 368, 624-626, 1994.

Wicks, C.W.J., and M.A. Richards, A detailed map of the 660 km discontinuity beneath the Izu-Bonin subduction zone, Science, 261, 1424-1427, 1993.

Wiens, D.A., and H.J. Gilbert, Effect of slab temperature on deep-earthquake aftershock productivity and magnitude-frequency relations, Nature, 384, 153-156, 1996.

Wiens, D.A., H.J. Gilbert, B. Hicks, M.E. Wysession, and P.J. Shore, Aftershock sequences of moderate-sized intermediate and deep earthquakes in the Tonga subduction zone, Geophys. Res. Lett., 24, 2059-2062, 1997.

Wood, B.J., The effect of H₂O on the 410 km seismic discontinuity, Science, 268, 74-76, 1995.

Xu, Y., and D.A. Wiens, Upper mantle structure of the Southwest Pacific from regional waveform inversion, J. Geophys. Res. , submitted, 1997.

Yamazaki, A., and K. Hirahara, The thickness of upper mantle discontinuities as inferred from short period J-array data, Geophys. Res. Lett., 21, 1811-1814, 1994.

Yoshioka, S., R. Daessler, and D.A. Yuen, Stress fields associated with metastable phase transitions in descending slabs and deep-focus earthquakes, Phys. Earth Planet. Int., 104, 345-361, 1997.

Zhao, D., Y. Xu, D.A. Wiens, L. Dorman, J. Hildebrand, and S. Webb, Depth extent of the Lau back-arc spreading center and its relationship to the subduction process, Science, submitted, 1997.

 back to SAFT home