September/October 1998


A common problem in seismology can be illustrated by a simple analogy: Given the higher number of fatalities in this century due to grizzly bear attacks in Montana relative to New York, what can be said about the future hazard? Is New York a "bear gap", where attacks are overdue and hence likely, or is the hazard intrinsically less? Given no information beyond the short record, either assumption is tenable because of the ambiguity involved in reasoning from the nonoccurrence of an event. To resolve the issue, we need an independent method of looking for bears.

Using the earthquake record to estimate the recurrence and hazard of future large earthquakes raises a similar problem. The historical record is short by geological standards, and the instrumental record is even shorter. Hence a standard way to obtain additional insight is to assume that earthquake occurrence is grossly linked to plate motions or intraplate deformation such that strain which accumulates on long time scales is released seismically. Although it is unclear how this process works--seismicity along plate boundaries and in intraplate areas seems highly variable in space and time, as illustrated by paleoseismological studies showing that the mean recurrence time of earthquakes on the southern San Andreas Fault is approximately 132 + 105 years--some relation seems plausible. The models of plate motions used, however, typically average over the past three million years. Hence it has been unclear whether plate motions predicted by these models are relevant to earthquake recurrence, which involve time scales of a few hundred years.

Valuable insight into these issues is being provided by the rapidly evolving techniques of space-based geodesy. These techniques (Very Long Baseline radio Interferometry [VLBI], satellite laser ranging, the Global Positioning System [GPS], and DORIS [a French system similar to GPS but using ground transmitters]) have in common the use of space-based technologies to measure the positions of geodetic monuments to accuracies of better than a centimeter, even for sites thousands of kilometers apart. Hence, measurements of positions over time yield relative velocities to precisions almost unimaginable during the early days of plate tectonic studies. Moreover, these studies cover much larger areas than would have been practical with traditional geodesy, which is restricted to sites which are in view of each other.

Although the various techniques provide similar data, GPS is the system of choice for most tectonic applications. GPS relies on a constellation of satellites carrying very precise clocks, which continuously transmit time and position information. Radio receivers which track multiple satellites can thus locate themselves in a fashion similar to locating an earthquake. Although GPS was developed by the Department of Defense to provide military users with positions accurate to tens of meters, space geodesists have developed data-processing techniques to derive far more accurate positions. The required geodetic-quality GPS receivers, though larger and more expensive than handheld units widely used to derive approximate positions, are easily transported and relatively inexpensive (less than about $20,000). Despite the complexity of the technology, the resulting velocities can be used in tectonic studies with little consideration of how they are derived, much as earthquake mechanisms or seismic velocity profiles are used in other applications.

One of the most important results of space geodesy for seismology is that plate motions have remained generally steady over the past few million years, as shown by the striking agreement between motions measured over a few years by space geodesy and the predictions of NUVEL-1, a global plate motion model which averages over the past three million years. Only in a few places, for example the Caribbean, are significant discrepancies observed, perhaps because of limitations in the plate motion model. The general agreement is consistent with the idea that although motion at plate boundaries can be episodic, as in large earthquakes, the viscous asthenosphere damps out the transient motions (much like the damping element in a seismometer) and causes steady motion between plate interiors.

Space geodetic data helped resolve the troubling discrepancy between the 35 mm/yr slip rate on the San Andreas inferred for the past several thousands of years and the higher (48 mm/yr) 3-Myr average motion between the Pacific and North America plates predicted by NUVEL-1. Two explanations were possible: Either plate motions had slowed, with obvious implications for earthquake recurrence, or significant deformation occurs across a plate boundary zone much broader than the San Andreas itself. GPS and VLBI measurements show the latter to be the case, with the distributed motion summing to the expected plate motion. This steadiness implies that data on various time scales (plate motions, paleoseismology) can be combined for seismic slip and hazard analyses. In the bear analogy, a long-term history of the bear population can be compared to the record of bear attacks.

A second important application of space geodesy to seismology is to explore how the seismic and aseismic portions of the deformation vary in space and time in the diffuse deformation zones which characterize many plate boundaries. For example, GPS data at the Peru subduction zone show deformation of the overriding South American plate, indicating that about half the total plate motion is accumulating on the locked plate boundary thrust fault and should be released in future great earthquakes. The GPS approach avoids many of the difficulties in previous seismological estimates due to the limited earthquake history, the variability of large thrust earthquakes in space and time, and the possibility of slow or silent earthquakes or afterslip. (These difficulties are illustrated by the observation that the seismic slip rate along the Chile Trench estimated from slip in the great 1960 earthquake and the recurrence of major earthquakes exceeds the plate convergence rate, so either the typical Chilean subduction earthquake is smaller than the 1960 event, the average recurrence interval is greater than observed in the last 400 years, or both.)

The GPS approach, which does not rely on the seismic history, does however face the difficulty of accounting for possible deformation of the upper plate due to strain diffusion for tens to hundreds of years after large earthquakes. Postseismic deformation is increasingly being observed, so extended periods of data will provide a better understanding of this process and of the rheology of the lithosphere and asthenosphere. Similarly, GPS data across the foreland thrust belt directly measure the rate of crustal shortening which is building the Andes and show that this rate is much faster than inferred from seismic moments, either because most occurs aseismically or because of the short seismic record. As similar data accumulate from various plate boundary zones, we may find that significant aseismic deformation is the norm rather than the exception. Hence, studies around the world may provide valuable insight into the process of aseismic deformation and thus help address the ongoing debate over a possible seismic moment deficit in Southern California. This question has obvious implications for earthquake recurrence and hazards. Using the bear analogy, many bears may not attack humans.

A third interesting application is study of rare, but sometimes large, intraplate earthquakes, like the 1811 and 1812 earthquakes in the New Madrid seismic zone. Plate motions give no a priori idea for how often such earthquakes occur, beyond the trivial prediction that there is no deformation within ideally rigid plates, so the earthquakes should not have occurred (no bears expected). At least at New Madrid, space geodetic data show very small (or zero) deformation, suggesting that the recurrence time for large earthquakes may be much longer than previously assumed. If these data are confirmed by further measurements, which will give better rate estimates due to the longer time series, they have interesting implications. Perhaps they are consistent with paleoseismological studies indicating that several large earthquakes occurred between A.D. 400 and 1811, if the earthquakes were precursors to the great earthquakes which are unlikely to recur for a very long time. If so, the seismic hazard at New Madrid may be considerably less than often assumed (the bear has gone away), whereas strain may be accumulating on other preexisting weak features within the continental interior. Such studies of intraplate seismic zones should shed light into the stresses within continental interiors and how they cause deformation and earthquakes.

Space geodetic data will become even more valuable as the technology advances. Two particularly important developments are the improvements in the accuracy of vertical measurements and the availability of seafloor geodetic systems, which combine acoustic measurements underwater with shipboard GPS. Adding vertical and seafloor data to the horizontal data, which are the staple of present studies, should provide better constraints on issues such as the geometry of the plate interface at subduction boundaries. Similarly, vertical data will give powerful insight into the poorly understood rates of a variety of vertical tectonic processes, such as mountain building.

In years to come, space geodetic studies will be giving superb data about the kinematics, or motions, within many important plate boundary zones and regions of intraplate deformation. The results from the dense geodetic networks in California and Japan illustrate what can be done in important tectonic environments around the world. We have the opportunity to do much more than simply confirm our earlier ideas about the motions. The challenge is to use these data, together with seismological and other geophysical and geological data, better to understand the fundamental processes within these complex areas. For example, comparison of the geometry and time history of faulting inferred from geodetic data to seismological results is already showing that significant afterslip is common and is providing constraints on the rheology of the lithosphere and asthenosphere. Comparison of geodetic, seismological, and geologic estimates of deformation rates within active regions should lead to an improved understanding of the partitioning between seismic and aseismic deformation. We can expect improved understanding of fundamental plate tectonic processes such as ocean-continent convergence (e.g., the Andes and Alaska), continental collision (e.g., the Himalayas), continental rifting (e.g., East Africa), and backarc spreading (e.g., Western Pacific). All these applications will require new ideas and models to make full use of the new data. The new geodetic data alone cannot resolve the many unsolved fundamental problems in tectonics and earthquake studies, any more than did the new generations of seismometers. We have a new tool to look for bears, but we're not out of the woods yet.

Seth Stein
Department of Geological Sciences
Northwestern University
Evanston, IL 60208

Timothy H. Dixon
Rosenstiel School for Marine and Atmospheric Sciences
University of Miami
Miami, FL 33149

To send a letter to the editor regarding this opinion or to write your own opinion, contact Editor John Ebel by email or telephone him at (617) 552-8300.

Posted: 14 September 1998