Paul G. Silver

Observing Transient Deformation

Transient deformation in the Earth's crust is one of the important sources of information about the processes that trigger seismic events. In collaboration with postdoctoral fellow Stephen Gao and Staff Members Alan Linde and Selwyn Sacks, I have been examining two very interesting transients that may shed light on the earthquake-triggering process.

As discussed at greater length above by Selwyn Sacks, it is not unusual to observe slow multiyear deformation following a large earthquake, as the crust slowly adjusts to the change in the stress field brought about by such an event. It is much more unusual to observe such slow deformation unrelated to a large earthquake. We have observed one such event, and it produced an increase in the slip velocity along the San Andreas Fault that began in 1993 and continues to the present. This transient was observed in the Parkfield region of central California -- a site of an expected magnitude 6 earthquake. This expectation is based on a past sequence of earthquakes that ruptured the same patch of the fault every 22 years on average. While the anticipated Parkfield earthquake has not yet occurred, the resulting concentration of many different kinds of strain instrumentation has produced an unprecedented data set spanning 15 years of observations. The instrumentation used includes two-color electronic distance meters (EDMs), which are similar in precision to the better-known GPS receivers; creepmeters, which measure motion locally across the fault; borehole strainmeters, including those produced at DTM; and borehole seismometers for detecting microearthquake activity, which is a useful measure of strain in the seismogenic zone beneath the surface. The integration of data from all these instruments reveals a fascinating deformation event.

The most easily observable aspect of the transient, from the EDM and creep data, shows a speeding up of the San Andreas Fault slip from 1 cm/yr to 1.3 cm/yr beginning in 1993. The increase in slip rate corresponds temporally to a dramatic increase in general seismic activity, including the occurrence of four M ~ 4 earthquakes that occurred along a 6-km segment of the fault just northwest of the EDM network. There was also a synchronous change in borehole shear strain rate. Such a change in slip and strain rate requires that the conditions on the fault have changed: either the resistance to slip on the fault has been reduced, or the stress driving the fault slip has increased.

Modeling of the EDM and borehole strain data provides a means for distinguishing between these two possibilities. Indeed, the data suggest that the stress has increased as a result of even larger slow aseismic slip to the north that loaded the southern segment. This larger event was probably triggered by the occurrence of the four seismic events. The existence of this northern slip transient is not directly visible from surface observations, but it has been inferred from careful analysis of the rate of microearthquake activity. Thus, the examination of these various sources of data reveals a complex episode of transient deformation and stress redistribution. Such events have only occasionally been observed, and we presently know relatively little about them. They nevertheless represent a phenomenon that could significantly increase our understanding of the redistribution of stress and strain in the crust and its relation to the occurrence of earthquakes.

Another intriguing transient involves the M_w = 7.3 Landers earthquake of June 28, 1992. Since the occurrence of this earthquake, there has been heightened interest in the topic of triggered seismicity, because this event was followed by intense seismic activity for several weeks over a broad area encompassing much of the western United States. To see if there was a long-term effect, we examined all earthquakes, including the very smallest. We found that the short-term triggering was only the beginning of a previously unrecognized longer-term (five-year) trend, consisting of a persistently elevated, but decaying, microearthquake rate (Fig. 13). This finding suggests that the changes brought about by this earthquake were both widespread and semipermanent, essentially sensitizing a large area to small stress perturbations.

Perhaps even more surprising, this seismicity excess was modulated by an annual cycle, in which there was a maximum of earthquakes in the fall and a minimum in the spring (Fig. 13). Much of this signal comes from volcanic/ hydrothermal areas (Fig. 14) where the short-term triggering was clearly observed. There are several possible environmental sources of stress that might give rise to this cycle, including tidal loading, precipitation, and barometric pressure. For a variety of reasons, barometric pressure appears to be the most plausible of the three. Whichever is in fact the cause, it implies that very small environmental stresses of order tens of millibars are capable of triggering small seismic events.

Fig. 13. A time series of seismicity rate (events/day) and amplitude of the annual cycle is shown here: (a) number of events per day; (b) monthly average of number of events per day (solid line), fit to a decaying sinusoid with an annual period (dotted line); (c) amplitude of the annual cycle (and uncertainty) calculated using different cutoff magnitudes. The amplitude is significantly different from zero for events less than magnitude 1.7.

Fig. 14. This map shows the spatial distribution of the annual cycle (defined as the difference in the total number of events between the second half and the first half of the year) for the post-Landers period. Dark areas (at lat 39°, long -123° lat 37°, long -119° and lat 36°, long -118°) show the largest annual cycle, with more events in the fall than in the spring. Dots denote events that occurred within 10 days after Landers and reflect regions of short-term triggering. Note that the annual cycle is seen where there was short-term triggering.