AGU Fall 2000

US Pacific Northwest GPS Velocity Field Inferred from Campaign Measurements: Evidence for Large-scale Rotation and Plate Locking

Rob McCaffrey, Anthony Qamar, Zuoli Ning, Maureen Long, and Charles A. Williams

Global Positioning System (GPS) measurements made by various groups over the past 9 years in Oregon and Washington states are being compiled and re-processed to estimate a regional velocity field for the southern Cascadia margin. Campaign data are from surveys by the US Gelogical Survey (1992-1999), by the Cascades Volcano Observatory (1991-1998), by a consortium of volunteer observers in Oregon and Washington through direction of the National Geodetic Survey (1998), by Rensselaer Polytechnic Institute and Oregon State (1996-1999), by the University of Washington (1994-1999), and by Rensselaer Polytechnic Institute and the University of Washington (2000). We also routinely process data from several continuous sites of the Western Canada Deformation Array, the Pacific Northwest Geodetic Array, and the US Coast Guard CORS array. GPS data are processed with the GAMIT/GLOBK software and velocities are estimated in the ITRF96 reference frame. Site velocities are put in the North American (NA) reference frame by removing the NA-ITRF96 rotation estimated by DeMets and Dixon (GRL, 1999). The predominant feature of the velocity field in the NA reference frame is ~1 ^o/Ma clockwise rotation about a pole near 46^oN, 119^oW. That this rotation is very close to that inferred by Wells et al. (Geology, 1998) based on geological evidence suggests that the rotation has been stable for the past 12 Ma. The rotating block includes but is not limited to the accreted Siletz basalt that makes up most of the coastal region of Oregon and Washington. The eastern boundary of the rotating block appears to fall in the Basin and Range and not along the Cascade arc. Removing the predicted instantaneous effects of this rotation from the GPS vectors reveals ENE directed motion of coastal sites that decays landward, consistent with strain due to locking of the subducting Juan de Fuca plate with the NW edge of North America. We are inverting the GPS results to estimate rotation and plate boundary strain simultaneously using both dislocation and stress-based finite element deformation models. Initial results from Oregon indicate that offshore plate locking is large, with a possible additional component of locking beneath the coast ranges that shows up in the dislocation models but not in the finite element models. See: http://www.rpi.edu/~mccafr/oregon.htm Download PDF file of poster

North Cascadia Margin Deformation from GPS Measurements

J.A. Henton, H. Dragert, R. McCaffrey, K. Wang, R.D. Hyndman

Recent densification of repeated GPS campaign measurements across Vancouver Island, British Columbia, is beginning to consolidate the pattern of crustal deformation for the northern Cascadia Subduction Zone (CSZ). Deformation is generally consistent with a fully locked subduction megathrust fault surface and no free-slip zones are resolved. As a result, the northern CSZ margin accumulates the full slip-deficit (i.e. at the rate of Juan de Fuca/North America relative plate convergence) along its length with no apparent abrupt changes in the width of the seismogenic zone. The observed horizontal velocity field fits the elastic dislocation model well to first order and shows no significant components of rigid block motion. The widths of the locked zones off central Vancouver Island and southern Vancouver Island are approximately 55 km and 85 km, respectively, assuming equivalent widths for their respective transition zones. Directions of observed principal strain axes are well aligned with the direction of convergence but the magnitude of the observed rate of margin normal shortening is about 1.5 time smaller than predicted by the dislocation models; i.e. observed velocities for the inland GPS sites are larger than predicted by the model. This discrepancy can be minimized by introducing time-dependent viscous deformation within mantle underlying the CSZ as demonstrated by the viscoelastic model developed by Wang and He. For northern Vancouver Island, the deformation measurements also provide evidence for crustal strain that is not fully accounted for by current elastic models of a locked subduction thrust fault. The northwesterly motion of the continuous GPS site HOLB, the reduced deformation motions in this region, and the small-scale margin parallel extension measured here mark a pronounced change in the character of the accumulating crustal strain in the margin overlying the Explorer Plate. This change in strain rates to the north of the locked subducting Juan de Fuca Plate may play a role in the origin of large crustal earthquakes on central Vancouver Island.

Estimation of the Rate of Stress Accumulation Along the Cascadia and Sumatra Subduction Zones Constrained by Geodetic Observations

Charles Williams and Rob McCaffrey

We use a two-dimensional elastic finite element model to create synthetic Green's functions that represent the stresses applied to the fault boundary and the base of the overriding plate at a subduction zone. These functions are then used to invert geodetic observations of surface displacement to obtain the rates of normal and shear stress accumulation along the boundaries of the overriding plate. We perform a linear least squares inversion with linear inequality constraints. The constraints allow us to insure that the predicted displacement field provides a reasonable solution outside the region where observations are available, and it also allows us to require positive values for the shear stress rate and fault-parallel displacement along the fault boundary. We use this inversion within a genetic algorithm that optimizes the parameterization of the problem so that we simultaneously minimize the condition of the inversion matrix and the chi-square misfit normalized by the number of degrees of freedom. We have applied this inversion technique to both the stress-based model and a standard dislocation model so that we may compare the results.

We have applied our model using horizontal GPS velocities in conjunction with uplift rate estimates from coral rings in Sumatra, while using horizontal GPS velocities and leveling results from Oregon and Washington. We find that the patterns and magnitudes of the predicted fault slip distributions are significantly different for the stress-based models when compared with results for dislocation models. In particular, the stress-based models generally require smaller amounts of fault slip, with a corresponding reduction in the rate of shear stress accumulation along the fault. The two models thus predict different Coulomb Failure Function rates along the fault. The resolution of this issue will therefore have important implications for earthquake hazard research.

Faulting at strike-slip boundaries: The effect of viscous strengthening/weakening and loading on fault patterns

Luc Lavier, Charles A. Williams, Roger Buck, Rob McCaffrey

We analyze numerical experiments of the spontaneous formation and evolution of faults at a strike-slip boundary. To do so we developed a 2+1D thermomechanical model of lithospheric deformation that allows us to study the predicted patterns of fault formation when an elastic-plastic (brittle) upper layer is coupled to a viscoelastic (ductile) lower layer for large deformation. The mechanical and thermal evolution of the model are coupled to calculate the viscous stresses using strain rate and temperature dependent creep laws. The formation of faults in the brittle layer is modeled as a reduction of strength as a function of strain.

We find that the faulting pattern is dependent on both the viscosity structure of the lower crust and the loading conditions at the plate boundary. In the case of a brittle layer overlying a viscous layer, we find that when the plate is loaded from below, faulting in the brittle crust occurs only if the lower crust has a very high viscosity. When the plate is loaded from the sides and the lower crust viscosity is high (plagioclase controlled) our model predicts the formation of multiple faults. For the same case and a low viscosity lower crust (quartz controlled) deformation leads to the formation of a single fault. For the case of multiple faults the fault spacing is dependent on the way viscosity decreases with depth.

Finally we analyze the effect of viscous weakening (ductile shear) and of more complex crustal rheological layering on the distribution of strike slip faults for both types of loading (from below or from the sides).

Plumbing of the Toba Magma System: Petrologic and Geophysical Evidence of Two Shallow Reservoirs and Their Mantle Roots

D.A. Wark, Masturyono, R. McCaffrey, G.L.Farmer, Mawardi Rani, and R. Sukhyar

The 100 x 30 km Toba depression in northern Sumatra, Indonesia, which is often cited as Earth's largest Quaternary caldera, was the source of the 2500 km³ Youngest Toba Tuff (YTT) 74,000 yrs ago. With the goal of understanding the evolution of this large-volume, actively recharging caldera complex, and the present-day distribution of magma within it, we mapped seismic velocities in the underlying mantle and crust, and studied the petrology of lavas and tuffs that were extruded during the past ~1.2 Ma. Our results show that the Toba depression is underlain by two isolated magma reservoirs, each with its own mantle roots.

The main reservoir (MR), identified by a region 40 to 60 km across of low compressional wave velocities, underlies the southern two-thirds of the depression with the shallowest portions at depths of less than 10 km. The two largest rhyolite ash flows that erupted at 0.8 and 0.074 Ma from calderas within the Toba depression originated in the MR, as did younger dacite and rhyolite lavas. ENd is uniform (-9.9) among the tuffs and lavas, although 87Sr/86Sr varies (0.7131 to 0.7140), perhaps due to episodic influx of recharge melts. Evidence of recharge is provided by post-YTT lavas at Pusubukit volcano, on the MR's western margin. Pusubukit rocks include strongly contaminated (ENd is -11; 87Sr/86Sr is 0.7146-0.7168) andesitic lavas that formed by mixing of ascending melts (probably mafic andesite) with dacite from the margin of the MR. Extruded dacite at Pusubukit is chemically indistinguishable from other MR dacites, despite a stronger crustal isotope signature (ENd is -10.2; 87Sr/86Sr is 0.7151) that is attributed to interaction with the contaminated recharge melts. Ascent paths of these melts are defined by a tabular, low-velocity region containing long-period seismic events that extends below Pusubukit into the mantle to ~40 km depth.

The second, smaller magma reservoir is defined by low seismic velocities under the northern end of the Toba depression. This reservoir was the source of a relatively small-volume (<100 km³) tuff at 0.5 Ma. It has recently (since YTT eruption) been tapped by eruptions of basaltic andesite and dacite (and mixtures thereof) at Tandukbenua volcano. The 0.5 Ma tuff and the dacite have the same ENd (-10.8) but different 87Sr/86Sr (0.7154 and 0.7140, respectively), probably reflecting influx of recharge melts similar to extruded basaltic andesites, which are also highly contaminated (ENd is -10.4; 87Sr/86Sr is 0.7128). Chemically, rocks that originated in the northern magma reservoir can be distinguished from MR rocks to the south by their higher Ce/Yb, Ba/Y and Th/U ratios.

An anatectic origin has been suggested for the Toba rhyolites because of their exceptionally strong crustal isotopic signature. This signature is present in all magma types, however, indicating instead that the rhyolites have genetic roots in mantle-derived (but crustally contaminated) melts, probably like those extruded from the two young volcanoes that formed near the margins of the Toba depression after cataclysmic YTT eruption. The relatively mafic melts extruded at these volcanoes document the continued recharge of the shallow, silicic magma reservoirs beneath Toba, each of which appears to be in a period of repose between major eruptive events. Download PDF file of poster

AGU Spring 2000

Estimation of Forearc Rotation and Strain in Western Oregon and SW Washington Using GPS

Maureen D. Long, Rob McCaffrey, Cheryl K. Johnson

Global Positioning System (GPS) measurements from Oregon and SW Washington are being used to quantify the rotation and deformation of the Cascadia forearc relative to North America. We have compiled and re-processed campaign-style and continuous GPS data from a variety of sources, covering western Oregon and southwestern Washington from 42^oN to 47^oN. Campaign-style observations used in this analysis include those collected by Cascades Volcano Observatory (1992-1997), by the US Geological Survey (1991-1999), by local surveyors in collaboration with the National Geodetic Survey (1998), and by Rensselaer Polytechnic Institute and Oregon State University (1996-1999). Data from continuously operating reference sites in Oregon and southern Washington (Pacific Northwest Geodetic Array) since 1996 and regional sites are incorporated into our solutions. We have re-processed more than 200 days of campaign data and 350 days of continuous data using the GAMIT software. Site positions are estimated monthly in the ITRF96 reference frame using GLOBK. Site velocities are calculated in the ITRF96 reference frame by least-squares regression, then rotated to the North America reference frame using DeMets and Dixon's (GRL, 1999) NA-ITRF96 pole of rotation. The vectors reveal a dominant clockwise rotation of the Cascadia forearc about a pole of rotation located nearby in northeastern Oregon. This rotation is most evident in the vectors from 42^oN to 47^oN. Superimposed on the rotation is a relatively small component of eastward-directed contraction due to coupling with the Juan de Fuca plate subducting beneath the Cascadia forearc.

AGU Fall 1999

Modeling of Surface Deformation at Subduction Zones: Alternatives to Dislocation Models

Charles A. Williams, Robert McCaffrey

Surface deformation at convergent plate boundaries is generally modeled using elastic dislocations. These models are mathematically simple, making implementation straightforward, and they provide a set of easily-identified parameters that may be adjusted to fit observed surface deformation. Modified versions of such models have been reasonably successful at representing observed deformation patterns; however, they predict unrealistic stress distributions along the plate interface. In particular, these models predict high values of fault-parallel shear stress outside of the locked portion of the interface, contradicting the assumption that the unlocked portions of the interface should support only small amounts of shear, and the sense of shear along the unlocked sections is opposite to the expected direction. These difficulties are characteristic of dislocation models in general, and are a consequence of specifying the problem only in terms of kinematic conditions along the plate interface. We have explored alternative models of representing the surface deformation at subduction zones through the use of finite element methods. Our goal is to provide relatively simple models that are able to provide a reasonable fit to observed surface deformations while predicting physically realistic stress distributions along the plate interface. One such example is a model in which the plate convergence velocity is specified along the lower boundary of the subducting plate. The unlocked portions of the interface are represented as slip surfaces along which the fault-parallel shear stress is minimized. This type of model yields considerably lower levels of fault-parallel shear stress along the plate interface, and the shear stress is a maximum along the locked zone. The direction of shear is also consistent with plate movement. The surface deformation field is similar to that predicted by a modified dislocation model in some cases, but the gradients are generally much smaller. In particular, the model does not predict large variations in the horizontal strain across the locked region, as does the dislocation model. To test the validity of our models, we compare the results with geodetic observations from the Cascadia subduction zone.

Evidence for Block Rotation in the World's Fastest Slipping Continental Shear Zone, Irian Jaya, Indonesia

Colleen Stevens, Robert McCaffrey, Manuel Pubellier, Yehuda Bock, Jeff Genrich, Cecep Subarya, S.S.O. Puntodewo

The island of New Guinea comprises the northern margin of the Australian plate, where it converges obliquely with the Pacific plate at a rate of about 110 mm/a at an azimuth of N68^oE. Yearly Global Positioning System measurements were made throughout the Irian Jaya region (western New Guinea) from 1991 to 1997. Results indicate that southern Irian Jaya is rigidly attached to the Australian plate. East of 138^oE, 80% of the Pacific-Australian plate convergence is accommodated on a fault or faults north of the northern coast of the island, including the New Guinea trench. Only a small amount of crustal shortening occurs at the New Guinea Highlands and Mamberambo fold-and-thrust belts. West of 138^oE, Pacific-Australia convergence is accommodated in a very different manner. Here convergence drives the rapid west-southwestward motion of the Bird's Head microplate relative to southern Irian Jaya, resulting in about 50 mm/a of subduction of the Bird's Head continental crust below the Banda arc at the Seram trough. Only a fraction of convergence occurs between Bird's Head and the Pacific plate. The motion of Bird's Head relative to southern Irian Jaya occurs along a 300 km wide shear zone that accommodates 80 mm/a of left-lateral slip, making this the world's fastest slipping continental shear zone. Due to sparse geographic coverage of GPS sites, it is not clear if this slip is confined to a single fault or distributed along a system of faults. Geologic mapping reveals one or two dominant features that could accommodate the fast slip. GPS sites within the deforming region have large radial velocity components (pointing towards or away from the pole about which the Bird's Head microplate rotates), suggesting the presence of rotating blocks, the sizes of which approach the width of the deforming region. Download PDF file of poster

Receiver Function Analysis of Broadband Data in Toba Caldera, Sumatra

Jeremiah Armitage, Steve Roecker, Masturyono, Rob McCaffrey, John Nabelek, Dave Wark

We use receiver function analysis to explore the existence and location of magma chambers beneath Toba Caldera, northern Sumatra, the largest known Quaternary caldera on Earth. Broadband teleseismic waveforms were collected during the 1995 Toba PASSCAL experiment from sites aligned both along- and across-strike. Receiver function processing interprets P to SV phase conversions from horizontal interfaces. Waveforms are rotated to the radial direction and deconvolved with a spiking filter. For each station in the network, waveforms are stacked over a range of depths to enhance the converted phases. The resulting seismogram is then compared to a synthetic seismogram, created using the Thomson-Haskell technique, and the depths of and constraints across the interface are adjusted to minimize the misfit between the two. We use this analysis to place constraints on the depth of the magma chamber. By mapping the absence or presence of a discontinuity below each station of the network, lateral constraints on the extent of the magma chamber are also obtained. Our analysis thus far shows a mixture of high and low frequecy waves. High frequency waves are present at azimuths between 276^o and 278^o for all stations. Events with azimuths near 240^o and 40^o vary from high to low frequency depending on station location. High frequecy waves at these orientations are located on the northwestern side of the caldera, suggesting the absence of any chamber in this area. Inspection of the low frequecy waves shows the presence of strong reverberations which are likely due to the top of the magma chamber.

GPS Measurements of Plate Coupling and Strain Partitioning in Northwestern Oregon

Rob McCaffrey, Chris Goldfinger, Colleen W. Stevens, Cheryl K. Johnson, Peter Zwick, Maureen Long, Charles Williams, Joel Johnson, Yongdong Zhou, John Nabelek, Mark H. Murray, Curtis L. Smith

Gobal Positioning System measurements in the northern half of western Oregon (44^oN to 46.5^oN, 121^oW to 124^oW; Eugene to Portland) are being used to explore the motion of the Oregon forearc relative to North America, plate coupling, and strain partitioning associated with oblique convergence between the Juan de Fuca plate and North America. We are utilizing GPS occupations made by the US Geological Survey (1992-1994), by the Cascades Volcano Observatory (1992-1997), by Rensselaer and Oregon State University (1996-1999) and by by an Oregon consortium of volunteer observers through direction of the National Geodetic Survey (1998). In addition, several continuous sites operate in this region, both part of the Pacific Northwest Geodetic Array (PANGA) and U.S. Coast Guard CORS sites. Preliminary results based on 1994-1997 measurements suggest that the Oregon forearc moves northward relative to North America at several mm/yr possibly with localized shear along the Cascades arc. The western edge of the network reveals significant strain rates, probably associated with subduction coupling with the Juan de Fuca plate. Site velocities that incorporate measurements through 1999 will be presented at the meeting. Download PDF file of poster

AGU Spring 1999

GPS determination of deformation and inferred plate coupling in Oregon, Pacific Northwest

Cheryl Johnson, Maureen Long, Rob McCaffrey, Chris Goldfinger, Peter Zwick, John Nabelek

We are using Global Positioning System (GPS) data from the area of convergence between the Juan de Fuca and the North American plates to measure the deformation of the Oregon forearc. Available data include three years of continuous GPS data from two sites in Oregon, as well as campaign data collected during the summers of 1996, 1997, and 1998. We will process these data in order to calculate and map the velocity field of this area relative to North America. Additionally, we will use the velocity field to infer the extent of plate coupling between the Juan de Fuca and North America plates. Preliminary results indicate that the locked zone between the two plates extends farther inland than previously thought. This finding has significant implications for the assessment of seismic hazard in the heavily populated areas of the Pacific Northwest.

Strain and Slip Partitioning at Oblique Convergent Margins

Robert McCaffrey and Peter Zwick

Strain and slip partitioning are commonly observed at oblique convergent margins. Half of modern subduction zones reveal margin-parallel translation and margin-parallel straining of their forearcs. The deflections of earthquake slip vectors away from their expected directions reveal the kinematics of the forearc translations. GPS measurements at several plate boundaries are also used to examine the degree to which strain is partitioned during oblique convergence. We estimate the directions of the principal strains within the leading few hundred kilometers of the upper plate from GPS vectors and compare them to the azimuths of earthquake slip vectors, plate convergence vectors, and the orientations of the deformation fronts. At some margins, such the Himalayas (Nepal), Nankai, and Sumatra, the contractional principal strain direction is perpendicular to the deformation front even though plate convergence is very oblique. In other places, such as along the Peru-Chile trench from 15^oS to 18^oS, where the obliquity is 35^o, in Alaska (Prince William Sound) where obliquity is 20^o, and in Ecuador where obliquity is 25^o , contractional strain is more aligned with the plate convergence vector than with the normal to the trench. Finite element modeling of oblique convergence is being done to try to understand the conditions under which strain is or is not partitioned. In some cases, the presence of an arc-parallel strike-slip fault may be partially responsible for the partitioning but in general, we find that the strike-slip fault need not exist for partitioning to occur. A local increase in margin-parallel shear strain within the overriding plate can arise from down-dip variations in the tractions along the dipping thrust interface. We suggest that this localization of shear strain may lead to the development of a margin-parallel strike-slip fault, as commonly observed inboard of oblique subduction zones.

Active tectonics of southern Tibet: Oblique convergence, strain partitioning, and margin-parallel extension

Robert McCaffrey

The active tectonics of southern Tibet can be understood in terms of the process of strain partitioning that accompanies oblique convergence at most of the world's convergent margins. As seen at many subduction zones, interplate thrust earthquake slip vectors at the Himalayan front are nearly perpendicular to the plate margin. The radial slip vectors indicate that there is permanent extension within the overriding forearc and that the upper plate can withstand very little lateral shearing stress elastically. The upper-plate extension is expected to occur over the curved, dipping plate boundary. Upper plate trench-parallel extension is presently observed below sea level at some oblique subduction zones. Accordingly, the normal faulting in southern Tibet should not be used as a proxy for the uplift history of the plateau. The upper plate deformation also typically includes a component of translation by strike-slip faulting above or landward of the downdip edge of plate coupling. The Karakorum - Jiali fault zone serves to translate the southern half of Tibet and the Himalayas along strike. GPS results of Bilham et al. (Nature, 1996) and Larson et al (JGR 1999) for Nepal and Lhasa are consistent with strain partitioning in southern Tibet and the Himalaya in a manner similar to Sumatra and the Aleutians. Considering that India subducts a distance of at least 100 km beneath S Tibet, we can make the analogy between the Himalayan front and an oblique subduction zone. The radial slip vectors and principal strain directions in Nepal, extension in S Tibet on N-trending normal faults, and right-lateral strike-slip faulting along the Karakorum - Jiali fault zone are reminiscent of patterns of strain partitioning in subduction zones. If so, this implies that some part of Indian lithosphere extends well under S Tibet, possibly as far as the Karakorum - Jiali fault zone.

SSA 1999

GPS Constraints on Plate Coupling in central Western Oregon

GOLDFINGER, C. College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR 97331, gold@oce.orst.edu; MCCAFFREY, R., Department of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, NY, 12180; MURRAY, M, Department of Geophysics, Mitchell Building, B76, Stanford University, Stanford, CA 94305; ZWICK, P., Department of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, NY, 12180; NABELEK, J., College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR 97333,SMITH, C.L., NOAA, National Geodetic Survey, P.O. Box 12114, Salem, OR 97309; JOHNSON, C., Department of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, NY, 12180

We have been using GPS data from permanent sites and campaigns in 1992 and 1994 by USGS, and in 1996, 1997, and 1998 by OSU, RPI, and NGS to examine the variability and landward extent of interplate coupling in central Oregon. We established two permanent sites and a campaign network in central Oregon to investigate potential strong variability of the coupling signal suggested by earlier leveling studies. The earlier results showed little or no landward tilt of the coast range at 45 deg. N, while other arc-normal transects showed landward tilts. The earlier data have been variously interpreted as either poor coupling at this latitude, or as a coupled zone offshore, with the lack of tilt falling within the survey error. New GPS results indicate a probable locking signal in the central Oregon corridor, with station vectors consistent with an elastic signal from JDF-NOAM coupling. Vectors are also rotated toward arc-parallel from the NUVEL 1A vector, suggesting motion of a forearc sliver. GPS measurements also suggest that rapid surface displacement related to plate coupling extend further landward than would be expected from a locked zone lying entirely offshore. Preliminary elastic dislocation models suggest that plate coupling may extend beneath the Oregon Coast Range. The anomalous lack of landward tilt in earlier uplift data might be related to broader distributed coupling beneath the coast range/Siletzia terrane. The relative lack of uplift in the same corridor is supported by geologic evidence spanning several time scales, suggesting both an elastic and anelastic response of the upper plate to coupling stress.

MEASUREMENT AND INTERPRETATION OF CONTEMPORARY CRUSTAL STRAIN ALONG THE CASCADIA MARGIN

DRAGERT, H., Pacific Geoscience Centre, Geological Survey of Canada, Sidney, B.C., V8L 4B2, dragert@pgc.nrcan.gc.ca;

QAMAR, A., Geophysics Program, University of Washington, Seattle, WA 98195-1650, tony@geophys.washington.edu;

MCCAFFREY, R., Dept. of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, mccafr@rpi.edu;

GOLDFINGER, C., College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR 97331, gold@oce.orst.edu;

MILLER, M., Dept. of Geology, Central Washington University, Ellensburg, WA 98926, meghan@cwu.edu.

Beginning with tide gauge monitoring, precise gravity measurements, and leveling and trilateration surveys, geodetic techniques have been used along the coastal margin of the Cascadia Subduction Zone for more than 30 years in an effort to measure present-day crustal motions in this active seismic region. The objectives of such measurements are to provide a better understanding of regional tectonics and to augment the limited seismic and paleoseismic information on strain rates associated with damaging earthquakes. Results from these earlier surveys, that focused more on vertical deformation, provided somewhat sparse but consistent evidence for the existence of a locked subduction thrust fault underlying the continental slope from northern California to central Vancouver Island. The routine availability of extremely precise GPS orbits since circa 1992/93 revolutionized the monitoring of crustal motions on a global scale because it allowed relative horizontal positioning at the level of a few millimeters over distances of hundreds of kilometers. As elsewhere, this new technology is being exploited in the Pacific Northwest through the establishment of continuous GPS tracking networks as well as GPS campaigns repeated over more densely spaced sites at intervals of several years. Results from the most recent GPS measurements have provided more precise and more comprehensive estimates of crustal motions that not only confirm the presence of the locked subduction zone but also indicate spatial variations in strain rates that are not explained by current elastic slip-dislocation models. These newly resolved motions may provide key information addressing the north-south compression inferred from crustal seismicity, the motions and deformation of crustal blocks in the forearc, the variability of plate coupling along the strike of the subduction zone, the partitioning of strain across the margin of this oblique convergent zone, and the possible temporal nature of the accumulating strain.

GSA Penrose conference on subduction to strike-slip transitions, 1999

Strain partitioning and development of upper plate strike-slip faults during oblique convergence: An example from Sumatra

Rob McCaffrey

GPS measurements at several plate boundaries are examined to examine the degree to which strain is partitioned during oblique convergence. We estimate the directions of the principal strains within the leading edge of the upper plate from the GPS vectors and compare them to the azimuths of earthquake slip vectors, plate convergence vectors, and the orientations of the deformation fronts. At some margins, such the Himalayas (Nepal), Nankai, and Sumatra, the contractional principal strain direction is perpendicular to the deformation front even though plate convergence is very oblique. Margin-parallel shear strain is then localized landward of the strongest contractional strain. In other places, such as the large bend in the Peru-Chile trench from 15^oS to 18^oS, where the obliquity is 35^o, in Alaska (Prince William Sound) where obliquity is 20^o, and in Ecuador where obliquity is 25^o, the contractional strain is aligned with the plate convergence vector. Finite element models are being used to try to understand the conditions under which strain is partitioned. In particular, we are addressing the factors that can cause the principal strain directions rotate by up to 45^o between the forearc and the arc regions of convergent margins. In some cases, the presence of an arc-parallel strike-slip fault may cause the partitioning but in general, we find that the strike-slip fault need not exist for full partitioning to occur. Our work suggests that the requirements for partitioning of strain are that the plate boundary thrust fault supports shear stress well below the hanging wall and that the hanging wall be strong enough to transmit this stress to the upper crust. GPS measurements at the oblique subduction zone of north Sumatra reveal that the strain associated with the plate motion is fully partitioned between trench-normal convergence within the forearc and arc-parallel shear strain at the Sumatra fault. The strain partitioning is consistent with kinematic inferences made from earthquake slip vectors on the thrust and strike-slip faults although the slip vectors indicate slightly less partitioning. Partitioning in Sumatra also includes a component of arc-parallel stretching of the forearc. Numerical models are being used to understand the conditions under which the principal strain directions rotate by up to 45^o (e.g., in Sumatra) between the forearc and the arc regions of subduction zones. In the Sumatra case, the presence of the Sumatra fault may cause the partitioning but in general, the strike-slip fault need not exist for full partitioning to occur. We are exploring how strain partitioning is influenced by the down-dip distribution of shear stress on the thrust fault and by thermal conditions in the crust of the overriding plate.

Fall AGU 1998

Shortening Parallel to Convergent Plate Boundaries at Cusps Between Broad Bends Toward the Lower Plate

Rob McCaffrey and Leonardo Seeber

The shape of convergent plate boundaries is generally asymmetric. They typically form broad festoon-like curves convex toward the lower plate that come together at very sharp bends or cusps convex toward the upper plate. Margin-parallel transport of forearc rocks occurs along most modern margins where the relative convergence vector is oblique to the margin's orientation. Interplate shear forces also tend to produce stretching parallel to the margin in the upper plate. At the cusps, the translating forearc rocks converge forming transverse mountain belts that trend at high angles to the margin. Examples are the Hidaka Mountains on Hokkaido and their offshore extension into the forearc at the junction of the Kurile and Japan trenches, the Nanga Parbat antiform in Pakistan at the junction of the Himalayan and the Hazara thrusts, and possibly the Olympic mountains at the bend in the Cascadia trench and the Bjeshkt e Nemuna mountains at the western end of the Aegean arc in northern Albania.  These mountain belts typically expose high grade metamorphic rocks possibly exhumed by thrusting with vergence perpendicular to the main convergent plate boundary. Mechanically the topography of these mountains balances the basal forces driving the forearc sliver along strike. Slab flexure and underplating are alternative mechanisms for forming forearc highs at cusps.  Unlike these, margin-parallel transport predicts flow of material toward the cusps which is easily tested with geodetic methods. Accordingly, such transverse mountain belts should not be evident at broad concave-seaward bends where there is not also some evidence of margin-parallel transport. One good example of this is the bend in the South American margin at 20^oS.
 

GPS Constraints on Forearc Sliver Motion, Plate Coupling, and Strain Partitioning in Northwestern Oregon

R. McCaffrey, C. Goldfinger, M. Murray, P. Zwick, J. Nabelek, C. Smith, C. Johnson

Repeated Gobal Positioning System measurements along a forearc-crossing transect in NW Oregon between Eugene and Salem (44N to 45N, 121W to 124W) are being used to explore the motion of the Oregon forearc relative to North America, plate coupling and strain partitioning associated with oblique convergence between the Juan de Fuca plate and North America. We are utilizing GPS occupations made by the US Geological Survey in 1992 and 1994, by Rensselaer and Oregon State in 1996, 1997, and 1998 and by the National Geodetic Survey in 1998. (The August 1998 survey is part of a statewide reference survey conducted by the NGS and many local county and private surveyors.) In addition, we have operated 2 continuous sites in this corridor since early 1996 and a third was installed by Central Washington University in July, 1998, all forming part of the Pacific Northwest Geodetic Array (PANGA). Preliminary results based on 1994, 1996, and 1997 data suggest that the Oregon forearc moves north relative to North America at several mm/yr possibly with localized shear along the Cascades arc.  We see evidence for distributed margin-parallel shear strain across the forearc but uncertainties are still too large to allow resolution of spatial variations in it. The overall principal contraction direction is rotated from the plate convergence direction toward the trench-normal, a sign of strain partitioning. These results suggest that plate coupling extends farther inland at this latitude than is interpreted from uplift data alone. Similar rotation of the principal contraction direction occurs at the Nankai trough near Shikoku where convergence is oblique. Results from Oregon that incorporate the 1992 USGS and 1998 RPI/OSU measurements into the velocity determinations will be presented at the meeting. We will also discuss the expected relationships between plate coupling variations, rotations of the principal strains, and the across-forearc distribution of margin-parallel shear strain based on finite element models.

Rupture zone of the Biak earthquake of February 17, 1996 inferred from aftershocks, coseismic deformation, and waveform analysis

Masturyono, Rob McCaffrey, John Nabelek, Colleen Stevens, Berndt Shur, Fauzi, Gunawan Ibrahim, Prijanto, Herry Saroso, Yehuda Bock, Jeff Genrich

Following the great Mw=8.1 Biak earthquake of February 17, 1996, we re-occupied a regional GPS network, measured subsidence of shorelines, and deployed a temporary seismic network of 10 PASSCAL RAMP seismographs to record aftershocks for about 3 weeks. Each seismograph station was equipped with 2 three-component seismographs, both short-period and accelerometers. From more than 650 locatable events, we selected good quality locations to infer the geometry of the Biak earthquake rupture zone. The aftershocks are distributed over more than 300 km along strike and 150 km across strike (0o to 1.5oS latitude and 134.5oE to 137.5oE longitude). The hypocenter distribution clearly reveals a gently south dipping (about 5o dip) rupture zone consistent with the dip of New Guinea Thrust. The aftershocks extend deeper than 40 km depth to the south. The distribution of the aftershocks is consistent with the co-seismic rupture plane inferred from GPS measurements and observations of co-seismic subsidence of Biak Island (up to 2 meters). Several events located below Yapen Island indicate that the depth of Yapen trench-parallel strike-slip fault zone is about 20 Km. The relatively low activity recorded along the Yapen fault zone during our observation time suggests that the Yapen fault was not seriously activated by the Biak earthquake although it produced a large early aftershock.
 

GPS Measurements of Interseismic, Coseismic, and Postseismic Slip at the Central Ramu-Markham Fault in Papua New Guinea

C Stevens, EA Silver, L Wallace, R McCaffrey, R Jackson, R Little, P Pasen, S Hasiata, W Loratung, P English, H Davies

In Papua New Guinea, convergence between the Finisterre Range and the Australian plate is accommodated at the Ramu-Markham fault. EDM measurements were made at six sites crossing the fault in 1973 and 1975, and in 1993 we remeasured these sites with GPS. No large earthquakes occurred here during this 20 year period, and a comparison of EDM to GPS measurements suggests the interseismic rate of convergence is about 4 mm/yr. In October of 1993, two months after the sites were observed with GPS, four large (M=6.3 to 6.9) thrust earthquakes occurred 30 km north of the geodetic network. We remeasured the sites with GPS in January of 1994, and results indicate 20 cm of coseismic slip. Waveform modeling and CMT solutions suggest that the earthquakes had very gentle dips and occurred at depths of about 20 km. Together, the coseismic slip and earthquake source parameters suggest that the Ramu-Markham fault is a low-angle, mid-crustal detachment fault that connects to a steep ramp fault that comes close to the Earth's surface near the Markham Valley. In April 1997, we remeasured the sites with GPS. Comparison of 1994 to 1997 measurements suggests that up to 10 cm of slip occurred during this time period. We infer this to be postseismic creep, because no large earthquakes occurred during this period. In July of 1998, we again measured these sites with GPS, and a preliminary comparison of 1997-1998 measurements indicates that the fault is now locked again.
 

GPS Measurements Across the eastern Ramu-Markham Fault, Papua New Guinea

C Stevens, EA Silver, L Wallace, R Little, R Jackson, R Curley, P Pasen, S Hasiata, R McCaffrey, W Loratung, P English, P Tregoning, H Davies

The Ramu-Markham fault of Papua New Guinea accommodates convergence between the Finisterre island arc terrane and Australia. In 1993 we established a 13 site GPS network in the eastern Markham Valley to study strain accumulation at the Ramu-Markham fault. These sites and two sites that cross the fault at the western end of the vally were measured with GPS in 1993, 1997, and 1998. Preliminary results show a gradient in convergent rate along strike in agreement with Tregoning et al. (JGR, v.103, p.12,181- 12,204, 1998). GPS measurements at the western end of the valley suggest that the convergent rate is about 8 mm/yr. This rate increases to about 40 mm/yr at the eastern end of the valley, and the convergence direction is everywhere normal to the fault. Furthermore, results indicate that convergence across the valley at its eastern end occurs within a very narrow zone (< 13 km). In June of 1998, we installed and measured with GPS 17 new sites spanning the Finisterre Range, Highlands fold-and-thrust belt, and part of the Adelbert Range.
 

Strain rates along the Sumatra Fault from triangulation and GPS

Linette Prawirodirdjo, Y. Bock, R. McCaffrey, J. Genrich

The island of Sumatra in Indonesia experiences active deformation due to oblique subduction of the Australian plate beneath Eurasia. The oblique convergence results in an arc-parallel component of motion which is accommodated by the Sumatra Fault (SF). In the late 19th century the Dutch colonial government established a triangulation network that covered most of the island and spanned the entire length of the SF.  Although the network was not designed to detect plate motion, deformation due to earthquakes on the SF necessitated resurveying segments of the network. We obtain strain rates for those segments of the triangulation network which were surveyed more than once. Interseismic deformation is dominated by shear strain at the SF and strain rates are consistent with rates calculated from GPS data (gamma1 = 8.8E-08 /yr at 0.6 deg S and 1.4E-08/yr at 1 deg S latitude). Coseismic displacements are estimated for the 1892 Tapanuli earthquake and 1926 Padang Panjang earthquake. Deformation rates are also obtained by performing a solution combining the triangulation survey with results from GPS geodetic surveys performed this decade on the old triangulation sites. The results yield strain rates over a time period of over 100 years at three segments of the Sumatra Fault. Coseismic displacements are also estimated for the 1987 Tarutung earthquake. Information regarding interseismic and coseismic deformation derived from historical geodetic data serve to constrain the long term, spatial distribution of strain at a region which experiences complex tectonic deformation.

Spring AGU 1998

Strain partitioning at convergent margins: GPS constraints and numerical modeling

R. McCaffrey and P. Zwick

GPS measurements at several plate boundaries are used to examine the degree to which strain is partitioned during oblique convergence. We estimate the directions of the principal strains within the leading few hundred kilometers of the upper plate from GPS vectors and compare them to the azimuths of earthquake slip vectors, plate convergence vectors, and the orientations of the deformation fronts. At some margins, such the Himalayas (Nepal), Nankai, and Sumatra, the contractional principal strain direction is perpendicular to the deformation front even though plate convergence is very oblique. Margin-parallel shear strain is localized landward of the strongest contractional strain.  In other places, such as along the large bend in the Peru-Chile trench from 15S to 18S, where the obliquity is 35deg, in Alaska (Prince William Sound) where obliquity is 20deg, and in Ecuador where obliquity is 25deg, contractional strain is aligned with the plate convergence vector. Finite element models are being used to try to understand the conditions under which strain is or is not partitioned. In particular, we are addressing the factors that can cause the principal strain directions to rotate by up to 45deg between the forearc and the arc regions of convergent margins. In some cases, the presence of an arc-parallel strike-slip fault may be partially responsible for the partitioning but in general, we find that the strike-slip fault need not exist for full partitioning to occur. Our work suggests that the requirements for partitioning of strain are that the plate boundary thrust fault shear the base of the hanging wall a few hundred km landward of the deformation front and that the hanging wall is able to transmit this stress to the upper crust.

Fall AGU 1997

Imaging the Toba Caldera Magma System using  P-Velocity Tomography

Masturyono, Robert McCaffrey, Peter Zwick, David A. Wark, Steven Roecker

We are imaging the magma system of Toba caldera, the largest known Quaternary caldera in the world, using seismic velocity tomography. We developed a one dimension P-wave velocity model by inverting arrival time data recorded by 30 short period stations of a temporary network of PASSCAL seismographs. Test results of several 1D models indicate that the depth of the Moho is about 40 km with a P-wave velocity of about 7.5 km/s underlying a low-velocity lower crust. Preliminary results for a 3D model indicate that a low P-velocity region throughout the crust and uppermost mantle is associated with Toba Caldera. The low velocity region is about 30 km wide in the upper crust (5-15 km depth), narrows to about 15 km in the middle crust (15-25 km depth), and widens again in the lower crust. We will present further refinements of the 3D modeling at the meeting.
 

Geodetic Evidence for Full Strain Partitioning in North Sumatra

Rob McCaffrey, Peter Zwick, Yehuda Bock, Jeff Genrich, Linette Prawirodirdjo, S.S.O. Puntodewo, C. Subarya

GPS measurements at the oblique subduction zone of north Sumatra reveal that the strain associated with the plate motion is fully partitioned between trench-normal convergence within the forearc and arc-parallel shear strain at the Sumatra fault. The strain partitioning is consistent with kinematic inferences made from earthquake slip vectors on the thrust and strike-slip faults although the slip vectors indicate slightly less partitioning. Partitioning in Sumatra also includes a component of arc-parallel stretching of the forearc. Earthquake slip vectors suggest that such strain partitioning occurs at many subduction zones. Numerical models are being used to understand the conditions under which the principal strain directions rotate by up to 45deg (e.g., in Sumatra) between the forearc and the arc regions of subduction zones. In the Sumatra case, the presence of the Sumatra fault may cause the partitioning but in general, the strike-slip fault need not exist for full partitioning to occur. We are exploring how strain partitioning is influenced by the down-dip distribution of shear stress on the thrust fault and by thermal conditions in the crust of the overriding plate.

Spring AGU 1997

Strain partitioning as applied to Southern Tibet

Robert McCaffrey

Strain partitioning occurs to some degree at most of the world's subduction zones. It is most commonly revealed by interplate thrust earthquake slip vectors that are more perpendicular to the plate margin than is the plate convergence vector. Kinematically, the slip vector deflections indicate that there is some permanent deformation of the overriding forearc and the most common type is arc-parallel translation and extension. The upper-plate extension is expected to occur over the curved, dipping plate boundary and the translation occurs by strike-slip faulting above or landward of the downdip edge of plate coupling. At some subduction zones, earthquake slip vectors remain perpendicular to the trench through large changes in its orientation, which probably means that the upper plate can withstand very little lateral shearing stress elastically. Recent GPS results from Sumatra (Prawirodirdjo et al., Science, submitted)  show complete partitioning of the strain, including arc-parallel extension, in accordance with inferences made from slip vectors. Similarly, slip vectors for thrust earthquakes at the Himalayan deformation front are nearly perpendicular to it throughtout its length. GPS results of Bilham et al. (Nature, in press) for Nepal reveal strain partitioning and block-like behavior of S Tibet, similar to Sumatra and the Aleutians. Considering that India subducts a distance of at least 100 km beneath S Tibet, we can make the analogy between the Himalayan front and an oblique subduction zone. The radial slip vectors and principal strain directions in Nepal, extension is S Tibet on N-trending normal faults, and right-lateral faulting along the Karakorum - Jiali fault zone are reminiscent of patterns of strain partitioning in subduction zones. If so, this implies that some part of Indian lithosphere extends well under S Tibet, possibly as far as the Karakorum - Jiali fault zone.
 

Fall AGU 1996

The February 17, 1996, Mw=8.2 Biak earthquake: GPS measurements, co-seismic sea-level changes, and aftershocks

C. Stevens, P. Zwick, R. McCaffrey, Masturyono, Fauzi,  J. Nabelek, B. Schurr, Y. Bock, J. Genrich, S.S.O. Puntodewo, C. Subarya

On February 17, 1996, a Mw=8.2 thrust earthquake occurred between Biak Island and the New Guinea trench in Irian Jaya, Indonesia. Historically, seismicity near the New Guinea trench has been sparse, with the largest earthquakes occurring on the left-lateral strike-slip Yapen fault that runs parallel to and approximately 130 km south of the New Guinea trench. Large aftershocks show thrust, normal, and strike-slip mechanisms. In March of 1996, following the Biak earthquake, we remeasured a regional Irian Jaya GPS network, made measurements of co-seismic subsidence from sea-level changes on Biak and Yapen Islands, and conducted a RAMP aftershock survey to better understand the area that ruptured during the earthquake. GPS measurements indicate over 1 meter of NNE co-seismic displacement of Biak Island and approximately 80 cm of horizontal extension between sites on Biak and Yapen Islands. Subsidence measurements show strong gradients southward and westward across Biak Island, with as much as 2 meters on the northern coast of the island and as little as 0.1 m of subsidence along the southern coast. The geodetic data alone do not resolve the fault plane ambiguity. Models for both a steep, north-dipping fault and a gently, south-dipping fault fit the data equally well. The deformation gradients across Biak Island suggest that one edge of the rupture plane is beneath Biak Island. Preliminary locations of aftershocks indicate that the rupture surface was probably the south-dipping nodal plane. Aftershock locations from the RAMP network and results from dislocation models constrained by these data will be discussed.
 

Role of oblique convergence in the active deformation of southern Tibet

R. McCaffrey, J. Nabelek, P. Zwick

The active deformation of southern Tibet can be explained by a late Cenozoic change in the obliquity of convergence between it and India. In this scenario, the region of active extension on normal faults striking perpendicular to the plate boundary in S Tibet is bounded below by India and in the north by the Karakorum - Jiali fault zone (KJFZ), an east-west trending, right-lateral shear zone in the upper plate. The extension is due to a gradient in the horizontal component of basal shear stress acting on S Tibet resulting from the westward increase in the westward component of India's motion relative to Tibet. Deformation models suggest that the KJFZ probably lies over the region where the horizontal shear stress between the upper and lower plates diminishes to zero. If so, some part of the Indian lithosphere may underlie most of southern Tibet, as far as 500-km north of the present deformation front. The onset of the present pattern of faulting in S Tibet roughly coincides with the start of motion on the Altyn Tagh fault. Motion along the Altyn Tagh fault rotates Tibet relative to Eurasia and India. This rotation produces nearly normal convergence between India and Tibet at the eastern end of the Himalaya and increases the gradient in convergence obliquity by a factor of 3 over that between India and Eurasia. The increase in the obliquity gradient, which probably also produces a similar increase in the gradient in the lateral component of the basal shear stress, may have triggered the normal faulting in S Tibet and the strike-slip at the the KJFZ.
 

Plate kinematics and Crustal Deformation in Southeast Asia from GPS measurements

L. Prawirodirdjo, E. Calais, Y. Bock, J. Genrich, R. McCaffrey, C. Stevens, Fauzi, P.Zwick, J. Beavan, P. Tregoning, F. K. Brunner, S. S. O. Puntodewo, C. Subarya, J. Rais

The Eurasian, Australian, Pacific, and Philippine Sea tectonic plates interact in Southeast Asia, resulting in a broad zone of active deformation. We present an updated velocity field for a large portion of Southeast Asia from five GPS surveys conducted from 1989 to 1994. The data includes sites on each of the four plates, with a denser network on the Indonesian archipelago. The results come from reprocessing our entire data set. Recent improvements in the processing software have yielded a better defined reference frame and velocity estimates that are internally more consistent. This reprocessing also includes newly estimated velocities for several sites. Our main results are:  (1) Motion of the two Australian plate sites in the Indian ocean differ significantly from the Nuvel1-A prediction. They also differ significantly from each other in their agreement with the model. The motion of Christmas island, the site closer to the trench, might be contaminated by elastic strain accumulation.  (2) Newly estimated velocities for several sites on the Sundaland block confirm that it is not extruding eastward relative to stable Eurasia but suggest that it might be undergoing internal deformation. (3) Velocities for sites on Sumatra are now better constrained in the global reference frame. The oblique convergence between Australia and Eurasia is partitioned between normal convergence along the trench and right-lateral slip along the Sumatra fault. There is also evidence for motion between the fore-arc islands and the west coast of Sumatra, suggesting slip along the Mentawai fault.
 

Spring AGU 1996

Aftershocks survey of the Feb.17,1996 Irian Jaya Earthquake

Masturyono, Fauzi, B. Schurr, J. Nabelek, C. Stevens, Rob McCaffrey, Suhardjono, Wandono, Sunaryo, Sukarman, Priyanto

Utilizing 10 seismographs of RAMP program of IRIS, we are embarking on an aftershock survey of Biak, Irian Jaya earthquake. We expect to record for 3 weeks starting in early March. This event provides a rare opportunity to record afthershocks directly above the probable rupture plane of great subduction earthquake. The aftershock recording will complement GPS and regional uplift measurements. The goal is to determine whether it was gently, south dipping plane, consistent with thrusting at New Guinea thrust (NGT), or the steep, north-dipping plane that the ruptureed during Mw = 8.2 earthquake. If it was south-dipping plane, then we may be able to place accurate bounds on the down-dip extent of rupture. Another interesting aspect of this quake is the occurrence of large (Mw= 6.4) stike-slip aftershock that may have occured on the Yapen fault, a left-lateral, strike-slip fault that strikes nearly parralel to NGT and is about 120 to 150 km south of it. The slip vector of main shock, according to Harvard CMT solution, was nearly perpendicular to NGT but about 50o from the Pacific-Australia convergence direction. Hence this earthquake sequence showns active slip partitioning between NGT and Yapen fault. Our seismograph array will cover the possible interaction of the NGT and Yapen fault at depth and may provide details on the geometry of the slip partitioning process in the crust.

The Toba Seismic Experiment: Imaging a Large Silicic Magma System

Peter Zwick, Masturyono, Fauzi, Ibnu Purwana, Sunarjo, Suhardjono, Wandono, Kristanto, Sjafra Dwipa, Hendra Gunawan, Mawardi, Sukhyar Rifwar, Sofyan, Sunardi, Hendra Suarta, Suharyadi, Yanuar, Colleen Stevens, Robert McCaffrey, Dave Wark, John Nabelek, Paul Friberg

The Toba Caldera in northern Sumatra formed at 74 ka during the eruption of about 2800 km 3 of silicic magma in one of Earth's largest documented eruptions. Although post-74 ka eruption tapped magma of inter mediated composition from central vent along caldera margin, it is likely that a large volume of more differentiated, cristal rich magma still resides in the crust. Evidence that magma is still present includes strong attenuation of seismic waves, post-eruption resurgence of the caldera floor, and boiling temperature hot spring. The unique tectonic setting for Toba include it's association with a magma chamber, a possible tear in downgoing lithospheric slab beneath it, and proximity of the Great Sumatra Fault, a 1600 km long strike slip fault passing within 5 km of the SW wall of the caldera. In a collaborative project among several Indonesian and US institutions, we are attempting to characterize the distribution of the magma beneath the caldera using seismic tomography and to determine the relationship of magma to both the crustal structure(including the Sumatran Fault) and the subducting slab. We deployed a network of 30 short-period and 10 broadband PASSCAL instruments over an area of about 250 by 250 km for 4 months in early 1995. Over 140 teleseisms, 130 regional and 1500 local earthquakes were detected during that time. Preliminary modelling of receiver functions indicates extremely low velocities (Vp=3.0 to 3.5 km/sec) in upper 10 km of crust beneath the sites on the resurgent dome, interpreted to indicate the presence of melt.
 

GSA Spring 1996

THERMAL VIEW OF THE AGE-RATE DEPENDENCE OF SUBDUCTION ZONE SEISMICITY: HOW CASCADIA FITS IN

McCAFFREY, Robert

The seismic potential for the Cascadia subduction zone is often inferred from comparisons of its properties to those of other subduction zones with better known earthquake histories. In particular, linear extrapolation of the Ruff and Kanamori global correlation of subduction zone earthquake size with convergence rate and subducting plate age suggest that Cascadia may produce magnitude 9 earthquakes. The positive correlation of earthquake size and convergence rate can be explained by faster subduction lowering repeat times, so that faster subduction zones should have larger earthquakes within any time period. The slope of Mw vs. convergence rate found by Ruff and Kanamori is predicted by the recurrence model. Hence, in this context convergence rate has little influence on the size of the largest possible earthquake, but does control its average repeat time. Steady-state temperature on subduction thrust faults depends on the age and dip of the subducting lithosphere and the trench-normal convergence rate. In a global comparison, inferred average temperatures from 10 to 60 km depth on thrust faults at most subduction zones are within 100C of one another. The Cascadia subduction zone is considerably hotter than any other, even without considering insulation by the thick sediments on the Juan deFuca seafloor. As a thermal outlier, the Cascadia subduction zone is useful to study the effects of temperature on subduction zone seismicity. Warmer than most, but still much cooler than Cascadia, are Mexico and the Chile trench south of the Chile Rise, where young lithosphere subducts slowly. Neither of these are known to have earthquakes larger than about Mw=8 and are mostly aseismic. The largest observed earthquake and the seismic coupling coefficient at subduction zones correlate negatively with inferred temperature on the thrust fault. Anomalously high temperatures on the Cascadia thrust fault may explain why it is aseismic whereas models based on other properties of the world's subduction zones predict it to have a high rate of seismic moment release.