A Plate Boundary Observatory

Paul G. Silver1, Yehuda Bock2, Duncan C. Agnew2, Tom Henyey3, Alan T. Linde1 , Thomas V. McEvilly4 , Jean-Bernard Minster2, Barbara A. Romanowicz4 , I. Selwyn Sacks1, Robert B. Smith5, Sean C. Solomon1 , Seth A. Stein6

1Carnegie Institution of Washington, DTM; 2IGPP, Scripps Institution of Oceanography; 3University of Southern California/Southern California Earthquake Center; 4University of California, Berkeley; 5University of Utah; 6Northwestern University/UNAVCO

topographic map of Western USFigure 1. A topographic map of the western U.S., where the Pacific/North American plate boundary zone reaches its greatest width. Also shown are the locations of earthquakes above magnitude 6 that have occurred over the last 200 years (violet circles), and background seismicity over the last 10 years (violet dots). This particular plate boundary is very broad, containing about a third of the width of North America, although the seismic activity is concentrated towards the western edge of the boundary zone. Black arrows give GPS displacement vectors with respect to stable North America for most available GPS data (Bennett et al., 1998). Note large displacements along the San Andreas fault system and smaller, more diffuse deformation in the Basin and Range. Red lines denote plate boundaries.

A basic tenet of plate tectonics is that plates are rigid and that deformation is concentrated in narrow zones at their boundaries. We now know that plate boundary deformation zones can actually be quite broad, often extending thousands of kilometers into continental interiors, as illustrated by the Alpine-Himalayan chain and the western cordilleras of North and South America. They account for fully 15% of the Earth's surface (Gordon and Stein, 1992). Nearly all present-day tectonic activity and most non-meteorological natural hazards, particularly earthquakes and volcanic eruptions, are concentrated within these zones, making the plate boundary zone a critical area of study both from scientific and societal points of view. The segment of the Pacific-North American plate boundary zone found in the western United States shares these characteristics. It covers a third of the North American continent and includes such diverse features as the Rocky Mountains, the Basin and Range, the Coast Ranges and the Sierras. It also contains the seismogenic San Andreas fault system along its western edge.

The diverse tectonic processes found in these zones are ultimately due to the inexorable and quasi-steady relative motion of tectonic plates. An important constraint provided by modern geodesy is that spatially averaged decadal geodetic estimates of plate motion are, to first order, indistinguishable from geologic estimates based on million-year time scales. This "steadiness" provides a valuable framework for studying plate boundary deformation; it is also in marked contrast to the extremely variable tectonic response to this motion. This deformation spans at least 14 temporal and 3 to 5 spatial orders of magnitude, and includes processes that range from mountain building to earthquake occurrence. The study of plate boundary deformation is a rich research area that deserves increased attention from a broad spectrum of Earth scientists. There are several first-order unanswered questions that are nevertheless critical to understanding any tectonic process.

The central observational requirement for the study of plate boundary deformation is the characterization of the three-dimensional deformation field over the maximum ranges of spatial and temporal scales. The surface field can be measured geodetically; instrumentation must provide: (i) sufficient coverage of the plate boundary zone so as to capture an integral tectonic system, (ii) sufficient station density for detecting localized (e. g., fault-specific) phenomena, and (iii) the necessary bandwidth to detect plausible transient phenomena from fast and slow earthquakes to strain buildup and viscoelastic relaxation. For studying long-term, large scale tectonic processes, it is probably sufficient to examine spatial variations in steady-state strain rate, which may then be compared to geologically inferred deformation rates over the last few million years (Figure 1).

integrated plate boundary deformation networkFigure 2. The necessary components of an integrated plate boundary deformation network, and observed transients. Thresholds of strain-rate sensitivity (schematic) are shown for strainmeters, GPS, and INSAR as functions of period. The diagonal lines give GPS (green) and INSAR (blue) detection thresholds for 10-km baselines, assuming 2-mm and 2-cm displacement resolution for GPS and INSAR, respectively (horizontal only). GPS and INSAR strain-rate sensitivity is better at increasing periods, allowing, for example, the detection of plate motion (dashed lines) and long-term transients (periods greater than a month). Strainmeter detection threshold (red) reaches a minimum at a period of a week and then increases at longer period due to an increase in hydrologic influences. This is a conservative estimate which has been bettered in some situations. At long periods (months to a decade), GPS has greater sensitivity than strainmeters by one to two orders of magnitude. At intermediate periods (weeks to months), sensitivities are comparable, and at short periods (seconds to a month), strainmeter sensitivity is one to three orders of magnitude greater than for GPS. Combined use of both data sets provides enhanced sensitivity for detection of transients from earthquakes to plate motion. Also shown are several types of transients observed by strainmeters (red), GPS and equivalent (green), and INSAR (blue): Post-seismic deformation (triangles), slow earthquakes (squares), long-term aseismic deformation (diamonds), preseismic transients (circles), and volcanic strain transients (stars).

For short-term processes and their related deformation, such as earthquakes and volcanic eruptions, temporal and spatial resolution becomes much more important. Good sensitivity is needed across the sub-second-to-decade period band. The sub-second to hour range is readily covered by seismological observations. At longer periods, geodetic techniques are needed, but presently, there is no one geodetic technique that spans this broad temporal range (5 orders of magnitude) with roughly uniform strain-rate sensitivity. It will thus be necessary to utilize several techniques, including strainmeters, GPS, and interferometric synthetic aperture radar (INSAR) ัthe first being most useful from an hour to a month and the latter two (including non-continuous campaign measurements) for periods longer than a month (Figure 2). The published observations of transient phenomena reveal a variety of temporal scales that span this entire range (Figure 2). The post-seismic deformation of the 1992 Landers earthquake provides an excellent example (Figure 3a). Transients with three distinct time constants have been detected by these three instrumentation types: 5 days by strainmeters (Wyatt et al., 1994), 48 days by GPS (Shen et al., 1994) and 3 years by INSAR (Massonnet et al., 1996). In addition, remotely triggered seismicity from the Landers event at Long Valley was accompanied by a 6-day deformation pulse observed clearly on two strainmeters (Figure 3b, see Linde et al., 1994). These diverse post-seismic transients are suggestive of multiple deformation mechanisms. Two other examples illustrate the potentially broad range of transient behavior. The first is a slow earthquake (duration ~10 days) on the San Andreas fault near San Juan Bautista that was detected on two strainmeters, and was accompanied by increased seismic activity (Figure 3c, Linde et al., 1996). The other is a long-term (multi-year) aseismic transient in San Andreas fault slip that was observed on 2-color geodimeters, strainmeters, and creepmeters, and was coincident with an increase in seismicity (Figure 3d, Gao et al., 1998). Clearly, a crucial task in utilizing a multicomponent system is the integration of these geodetic techniques. There are ongoing efforts by investigators to incorporate at least two of these techniques into an internally consistent measure of the surface strain field: GPS and INSAR (Bock and Williams, 1997), and GPS and strainmeters (Gao et al., 1998).

a) b)

Figure 3: Examples of transients: a) Post-seismic deformation for the 1992 Landers earthquake from a laser strainmeter (LSM) and GPS, illustrating 5-day and 48-day time constants, respectively (after Wyatt et al., 1994). (b) Landers-triggered strain transient produced at Long Valley. Produced increase in seismicity (blue) and observed by a dilatometer (green) and tiltmeter(green) 20km apart (After Linde et al, 1994). (c) Slow (10-day) earthquake detected along the San Andreas fault at San Juan Bautista (south of Bay Area) and accompanied by elevated seismicity (Linde et al., 1996). (d) Multiyear aseismic transient in San Andreas fault slip at Parkfield, observed on 2-color geodimeters (stack of fault-crossing lines, GPS equivalent), dilatometers and tensor strain (not shown), creepmeters, and accompanying increased seismicity. The increase and decrease in line length (top panel) between 1991-1993 and 1993-1996 corresponds to a decrease and increase in fault slip rate, respectively (Gao et al.,1998).

Determining strain at depth is a less straightforward but crucial task. Deformation within the seismogenic zone, for example, may provide vital information on the triggering of seismic events. Strain indicators rather than calibrated strainmeters, must be used, however. For example, microearthquake activity can be interpreted as the radiated component of deformation in the seismogenic zone. Recent results of cluster analysis of microearthquakes at Parkfield have demonstrated the power of this type of technique for furthering our understanding of fault zone processes (Nadeau et al., 1995). Another important approach is imaging spatial and temporal variability in crustal structure. Images of faults can be obtained by the use of fault-zone guided waves (e.g., Li et al., 1997). The seismological detection of strain-induced opening and closing of fluid-filled cracks can be achieved through characterizing temporal variations in crustal structure. Mantle deformation is also accessible by seismic imaging, through constraints on the thermal (tomography) and strain (anisotropy) fields.

mapFigure 4. Existing and planned GPS (green circles), strainmeter installations (red circles), and three-component broadband seismographs (crosses). Pink zones denote most seismogenic part of the plate boundary (see Figure 1). These zones, shaded dark and light pink, correspond to areas of high and low population density, respectively. The PBO would be concentrated in these areas, factoring in population density in deployment priorities. The rest of the plate boundary (brown) would have more sparsely distributed instrumentation.

With these issues in mind, we recommend that the scientific community consider establishing a strain observatory along the Pacific/North American plate boundary (hereafter referred to as the Plate Boundary Observatory or PBO). The PBO should measure deformation over a broad spectrum of spatial and temporal scales and provide sufficient spatio-temporal resolution to constrain any transients associated with short-wavelength phenomena such as earthquakes. We propose that, where such phenomena are most prevalent, namely the most seismogenic areas of the boundary, 10-km spacing of instruments be achieved (Figure 4). This portion of the plate boundary is also where the greatest temporal resolution is needed. A close integration of seismometers, strainmeters, GPS, and INSAR is necessary to provide uniform strain-rate sensitivity, at plate-motion strain rates, across the temporal band from several Hertz to a decade. On the order of 1000 observing sites would be required. For the broader plate boundary, it would be possible to use coarser spacing, and to utilize GPS and INSAR exclusively, since these techniques are most successful for detecting long-period or steady-state strain. Constraints on the subsurface deformation field would be supplied by studies of strain indicators: microeathquake activity and crustal and mantle structure (including possible temporal variations). The seismological component would require both an augmentation of permanent seismic instrumentation in the plate boundary zone and transportable array deployments to map out particular regions in detail.

It would not be necessary to start from scratch in this effort, since some pieces of the PBO are already, or will soon be, in place (Figure 4). The most advanced component consists of geodetic-quality arrays of continuous GPS stations in southern (SCIGN) and northern (BARD) California, northern (NBAR) and eastern (EBAR) Basin and Range, and the Pacific Northwest (PANGA) (Figure 4). There are presently about 200 GPS receivers deployed in the proposed area, and 250 more should be installed within the next 1 to 2 years. The strainmeter component is much less advanced, since there are only about 20 strainmeter sites along the entire San Andreas fault system. Regarding INSAR, images are being acquired by non-U.S. satellites over western North America and are available to U.S. investigators, although issues of data access remain.

The establishment of a fully capable plate boundary observatory will require progress in four areas: (i) A more effective integration of strainmeters and GPS for a truly broadband plate boundary observatory. This integration concept should first be tested on a smaller scale, limited to a region of the plate boundary, where there are GPS receivers and strainmeters in roughly equal numbers. (ii) The densification of geodetic and seismic instrumentation along the northern San Andreas fault system for increased spatial resolution. (iii) The linking of the northern and southern San Andreas zones, to cover the seismogenic part of the plate boundary. (iv) Improving access to INSAR data for more effective integration. Present efforts involve operating a downlink facility in cooperation with the European Space Agency, and/or launching a SAR satellite that would collect data over western North America on a regular basis.

While the main focus of the PBO would be to gain a basic understanding of plate boundary processes, the PBO would also provide information of immense practical value. In particular, we would be in a position to detect precursory strain transients that may prove practical for the forecasting of earthquakes and volcanic eruptions. Such precursors exist for volcanic eruptions and have already been used to make predictions. Whether such precursors exist for earthquakes as well is something we still have to find out. The answer to this question would be crucial knowledge for society.


Bennett, R.A., J.L. Davis, B.P. Wernicke, The present-day pattern of western U.S. Cordillera deformation, Geology, 1998, submitted.

Bock, Y., and S. Williams, Integrated satellite interferometry in southern California, EOS Trans. AGU, 78, pp. 293, 299-300, 1997.

Gao, S., P. G. Silver, and A. T. Linde, A comprehensive analysis of deformation data at Parkfield, California: Detection of a long-term strain transient, J. Geophys. Res., submitted, 1998.

Gordon, R. G. and S. Stein, Global tectonics and space geodesy, Science, 256, 333-342, 1992.

Li, Y. G., K. Aki and F. L. Vernon, San Jacinto fault zone guided waves; a discrimination for recently active fault strands near Anza, California, J. Geophys. Res., 102, 11,689-11,701, 1997

Linde, A. T. I. S. Sacks, M. Johnston, D. Hill and R. Bilham, Increased pressure from rising bubbles as a mechanism for remotely triggered seismicity, Nature, 371, 408-410.

Linde, A. T., M. T. Gladwin, M. J. S. Johnston, R. L. Gwyther, and R. G. Bilham, A slow earthquake sequence on the San Andreas fault, Nature, 383, 65-68, 1996.

Massonnet, D., W. Thatcher, and H. Vadon, Detection of postseismic fault-zone collapse following the Landers earthquake, Nature, 382, 612-616, 1996.

Nadeau, R. M., W. Foxall, and T. V. McEvilly, Clustering and periodic recurrence of microearthquakes on the San Andreas Fault at Parkfield, California, Science, 267, 503-507, 1995.

Shen, Z., D. D. Jackson, Y. Feng, M. Cline, M. Kim, P. Fang, and Y. Bock, Postseismic deformation following the Landers earthquake, California, 28 June 1992, Bull. Seismol. Soc. Am. 84, 780-791, 1994.

Wyatt, F. K., D. C. Agnew, and M. Gladwin, Continuous measurements of crustal deformation for the 1992 Landers earthquake sequence, Bull. Seismol. Soc. Am., 84, 646-659, 1994.

Return to: IRIS Newsletter Information
Return to: Title Page and Table of Contents
Continue to: Next Article