SOUTHWEST PACIFIC SEISMIC EXPERIMENT

Introduction

Deployment

Instruments were deployed at 12 sites on 11 different islands. The sites were chosen to maximize coverage of the northern part of the Tonga slab as well as for logistical reasons. All of the sites are reachable by scheduled air service, except Niuafo'ou Island, which was deployed and picked up from ships servicing the OBS deployment, and Tofua Island, with is reachable by a charter seaplane which lands in a caldera lake.

Eight Streckeisen STS-2 and three Guralp CMG3-ESP sensors were deployed, with each sensor connected to a Reftek 24 bit DAS equipped with GPS timing and 1.0 or 1.2 Gb disks. At most sites, the sensors were housed inside 1 meter diameter fiberglass cylinders, which were cemented to bedrock or coral and equipped with a fiberglass lid and filled with Styrofoam insulation. The sensors were placed on a small slab of concrete inside the cylinders and about 1 meter below the ground surface. The electronic equipment and two automobile batteries were placed inside wooden field boxes near the vaults, and powered by two 30 watt solar panels (Figure 2 and Figure 3).

Because of the high frequency-dependent seismic noise levels on the islands, the Reftek triggering algorithm misses many smaller events. Therefore, we record only a single, continuous data stream at 20 or 25 samples/sec. The stations are serviced at 3 month intervals, when disks are swapped and the data is downloaded on a field computer in Suva, Fiji. A service trip to all the stations from North America takes more than one month and entails about 30 separate plane flights. Kiti Draunidalo of the Fiji Mineral Resources Department services the stations on approximately one-half of the service runs. A copy of all the data is retained in Suva for use by the local scientists in seismic hazard studies. Data recovery over the first year of operation is about 85-90 percent.

The bankruptcy of Polynesian Airlines mid-way through the experiment left Niue Island with only one weekly scheduled flight from anywhere in the world, such that servicing this one station takes about nine days. We are therefore cooperating with the Lamont Instrument Center in testing an ARGOS satellite state-of-health communications system. Data will be returned by using local people to swap out disks and return them to the US by air cargo, and the station state-of -health will be monitored via the ARGOS satellite. This system should render it necessary to visit the island only in the case of equipment failure.

Initial Results

We have now processed data from December, 1993 through June, 1994, which is enough to make some initial studies. Over 1000 earthquakes were located in the Tonga-Fiji region during the first seven months of the experiment, most of which were not detected by the PDE or other global earthquake compilations.

Deep Earthquakes. The March 9, 1994 large deep earthquake (Mw 7.6, depth 570 km) was the largest deep earthquake in 20 years, and even though now exceeded by the deep Bolivian event (June 9, 1994, Mw 8.3), it promises to provide a wealth of detail on the source processes of deep earthquakes. Previous to this event, a true aftershock sequence had not been observed for any deep earthquake, and rupture zones of deep earthquakes were poorly constrained due to the lack of aftershocks to delineate the fault planes. The March 9 event was unusually prolific in producing aftershocks compared to other large deep events; we found a sequence of 83 aftershocks ranging from mb 3.8 to 6.0 and extending for at least 40 days. The aftershocks show a power law decay with time similar to shallow aftershock sequences, and the number and magnitude distribution of the aftershocks is similar to that observed for typical shallow earthquakes. The contrast between this strong aftershock sequence and the weak aftershock sequence of the Bolivia event is particularly striking.

Most of the well-located aftershocks locate along a steeply dipping plane (figure 4) consistent with one of the main shock nodal planes and appear to delineate a 50 km by 65 km main shock rupture zone. Inversion of broadband body waveforms, recorded on-scale by six regional stations, also suggests this plane denotes the fault plane. Both the region of moment release and the aftershock zone cuts entirely through the active seismic zone and extends about 20 kilometers into the surrounding aseismic region. Thus there must be a mechanism for producing both rupture and aftershocks within the normally aseismic region surrounding the active slab. The width of the rupture zone is hard to reconcile with predictions of the transformational faulting hypothesis for the origin of deep earthquakes, which suggests that deep earthquakes should be confined to a thin zone of metastable olivine (see report in Dec 8, 1994 issue of Nature).

Several other deep earthquakes with Mw > 5.5 have occurred during the experiment. Most of these events have showed several aftershocks, including one Mw 6.4 event which showed 9. These results, combined with the March 9 aftershock sequence, suggests that Tonga deep earthquakes show more active aftershock sequences than deep earthquakes in other subduction zones.

Attenuation Structure.The broadband body waveforms show first order differences in attenuation between paths following the slab and various paths within the backarc region. We are developing a 3 dimensional Q model for the upper mantle beneath the Lau backarc using a differential attenuation method which determines the dt* along the raypaths of regionally propagating P and S waves. The dt* measurements and the raypaths can then be inverted to determine the lateral and depth dependent Q variations. Initial results suggest that Q increases rapidly with depth. A low Q region is found in the upper several hundred kilometers beneath the Lau backarc spreading center, possibly suggesting the presence of partial melt, and much higher Q is found beneath the South Fiji Basin, an older, inactive region.

Anisotropy. Shear wave splitting is readily observed on records of intermediate and deep earthquakes recorded in Fiji, and we are currently investigating anisotropy beneath the Lau backarc and Fiji platform in collaboration with Karen Fischer (Brown University). All Fiji stations generally show about 1 s. of splitting, with an average fast direction of about N60W. There appears to be little variation in the amount of splitting with event depth or path length for sources between depths of 400 to 650 km, and several teleseismic SKS phases show similar splitting results. This suggests that the splitting occurs largely within the upper 400 km of the mantle beneath the station. These results are consistent with a 1% azimuthal anisotropy uniformly distributed in the upper 400 km or the mantle, or greater anisotropy if the splitting occurs at shallower depths. The fast direction of the anisotopy is approximately parallel with the convergence direction of the Pacific plate and the spreading direction in the Lau backarc, suggesting the anisotropy may be produced by counterflow induced within the backarc by the subducting slab or the back arc extension.

Other studies of the velocity structure of the subducting slab and backarc basin using arrival time data and waveforms are also underway, as well as source parameter studies of regional earthquakes. The Southwest Pacific Seismic experiment will produce a better understanding of the seismic structure and geodynamnics of island arc - backarc systems, as well as a better understanding of the rupture processes of deep earthquakes. The data will be made available to the entire community through the IRIS data center one year after the end of the experiment.

Figure 1: Map showing the locations of the broadband stations of the Southwest Pacific Seismic Experiment (red dots with station codes). Seismographs of the Lau Basin OBS Deployment (LABATTS), conducted in collaboration with Leroy Dorman, Spahr Webb, and John Hildebrand, are denoted with numbered green dots. Colors denote seafloor bathymetry.

Figure 2: A typical field installation, this one located in Ha'apai, Tonga. The sensor is deployed in the fiberglass cylinder at the rear, and the batteries and recording gear in the plywood box, with solar panels on top. Washington University graduate student Erich Roth inspects the site.

Figure 3: A view of a typical field box, showing the batteries and gear.