Silent Slip on the Cascadia Subduction Interface
Geodetic satellites of the Global Positioning System (GPS) now permit continuous recording of surface motions around earthquake faults and volcanoes with millimeter precision. Data from distinct points on Earth's surface can be combined to infer the locations of the sources of deformation at depth in the crust. These data are radically altering our understanding of earthquake processes, justifying ambitious new sensor arrays to image active deformation sources in Earth's crust. In recent years, large continuous GPS arrays have been deployed in Japan and southern California, and smaller networks have been installed in other seismically active regions. On page 1525 of this issue (1), Dragert et al. provide a glimpse of the kinds of insights we can expect from these arrays.
Dragert et al. use data from a 14-station continuous GPS net located
in southwestern British Columbia and northern Washington to detect an episode
of silent (aseismic) slip on a major fault that surfaces offshore and dives
eastward beneath the continent. This fault defines the Cascadia subduction
zone, which separates the large, ~100-km-thick blocks of the Juan de Fuca
and North American plates (see the figure immediately below). Geological
studies (2) have shown that large earthquakes occur along the Cascadia
plate boundary roughly every 600 years. Modeling of geodetic data (3 )
indicates that the shallow, upper ~20 km of the plate boundary fault is
currently "locked" (not slipping) because of frictional resistance on the
fault interface. Continued relentless motion of the plates and aseismic
slip increase stresses on the overlying earthquake fault. These stresses
will eventually be relieved by abrupt shallow slip in a large earthquake.
The Cascadia subduction zone.
Subdiction of the Juan de Fuca plate beneath the North American plate results in the formation of the Cascade Range.
The Cascadia slip episode studied by Dragert et al. is unique and important because it occurred beneath an apparently completely locked plate boundary fault and was unrelated to any local earthquake activity. Aseismic events reported elsewhere occurred at the transition between locked and freely sliding fault segments (4), and many of them were triggered by transient adjustments after large earthquakes (5, 6 ). In contrast, the Cascadia event occurred on the downward extension of the locked earthquake-generating fault, where aseismic slip or distributed ductile shearing (7) was expected to occur at a uniform speed. The physical processes responsible for all such deformation instabilities are not much studied or well understood. If similar episodes are identified elsewhere, they will doubtless stimulate new work directed toward understanding their mechanisms.
Aseismic events of the kind identified by Dragert et al. have potentially important implications for earthquake occurrence. Each aseismic slip episode perturbs the local stress field and may bring a fault closer to failure. No single deformation episode is necessarily an earthquake precursor, but if the locked fault is near the end of its stress buildup cycle, an episodic slip event may be sufficient to trigger a large earthquake (8).
The Cascadia slip episode, which took about 35 days, increased stresses
across the shallower, earthquake-generating part of the plate boundary
fault (white line segment in the figure immediatley below). This stress
increase associated with the event is very small, equivalent to about half
a year of steady stress buildup, bringing the fault very slightly closer
to failure. The maximum change is about a factor of 10 smaller than the
stress changes caused by earthquake slip that have triggered subsequent
earthquake events (9).
A small step toward failure. Cross section showing elastic stress changes due to silent slip as inferred by Dragert et al. on fault segment shown in black. Positive stress changes (warm colors) bring earthquake faults like that shown by the white line closer to failure.
The Cascadia subduction zone last experienced a great earthquake in 1700 (10 ) and so may be only about halfway through its ~600-year earthquake cycle. Therefore, even events substantially larger than the 1999 aseismic slip episode may not soon push this fault over the brink. However, other regional faults may be closer to failure and such episodic events could lead to large earthquakes. Continuous monitoring of these faults, in Cascadia and elsewhere, is thus of major importance.
Much remains to be learned about earthquake stress buildup. We do not yet know whether aseismic episodes are rare or common, large or small. For earthquakes, the numbers of events increases roughly tenfold for each unit decrease in earthquake magnitude. For aseismic events, we must determine the relation between frequency of occurrence, slip, and slipped area to understand what causes them and evaluate whether such episodes will trigger large earthquakes. For the Cascadia episode, the ratio of fault slip to fault area was quite small, about two orders of magnitude less than that typical for earthquake slip. This ratio is proportional to the stress change caused by the event, so the magnitude of the effects like those shown in the figure is less important when the ratio is small.
Recent work (1, 4-6 ) has revealed a rich spectrum of aseismic behavior in seismically active regions but has also raised many new questions. Further progress in understanding these processes depends on the deployment of dense networks. Focused instrument clusters are needed to spatially resolve buried sources of deformation, many of which could be much smaller or more localized than that identified with the relatively sparse array used by Dragert et al. These arrays should also include ultrastable borehole strain meters, which are increasingly more sensitive than continuous GPS for time intervals of a month or less. Parallel developments are needed in refining analysis and modeling techniques aimed at extracting the maximum information from these large data sets (11).
Local prototype arrays of continuous GPS and borehole strain meters have recently been installed by the U.S. Geological Survey at Parkfield on the central San Andreas fault and at Long Valley caldera, a region of volcanic unrest in eastern California. Additional clusters in a range of geological environments will ensure timely acquisition of the kinds of data needed to capitalize on the capabilities of continuous GPS and borehole strain meter technology and rapidly expand our understanding of how earthquakes occur and why volcanoes erupt.
The author is at the U.S. Geological Survey,
Menlo Park, CA 94025, USA. E-mail: email@example.com