Seismicity rate change associated with the Hector Mine and Landers earthquakes.
Changes are quantified with the β-statistic calculated for spatially smoothed (20 km) seismicity (M≧2.0) cataloged by the Council for a National Seismic System (CNSS) using data from the Southern California Seismic Network (SCSN) and Northern California Seismic Network (NCSN).Rate change is calculated for a two-week period immediately after each main shock, relative to the background period 1987.0-1992.0.Earthquakes in post-main shock period are shown by plus signs. Blue area represents rate decrease, but is not significant because the post-main shock periods are so short. White indicates area with no cataloged earthquakes in post- and pre-main shock periods. a, Hector Mine response. Circles mark sites of triggered activity. Unrelated increase at 37.4 latitude consists of aftershocks of a M=5.3 earthquake there on Aug. 1, 1999.Increase near San Gregornio pass southwest of the aftershock zone (SG) began a week before Hector Mine earthquake, while the events near the Geysers (38.3 latitude) occurred 2 weeks after it. b, Landers response. Most remote triggering occurred to the north-northwest. Strong increase (yellow) at 36.7 latitude is the (triggered) M=5.6 Little Skull Mountain earthquake and its aftershocks.
Cumulative number of earthquakes in selected areas (25 km-radius circles centered on Indio, south end of the Salton Sea, Cerro Prieto and Mammoth Mountain, Long Valley ) during 1999 and October, 1999 listed in the SCSN, NCSN and RESNOM (at Cerro Prieto) catalogs. RESNOM is operated by the Centro de Investigacion Cientifica y Educacion Superior de Ensenada, Baja California, Mexico.Dashed lines indicate the time of the Hector Mine eartkquake.
Recorded E-W component seismic velocities for the Hector Mine (a) and Landers (b) earthquakes, superimposed on portions of the β-statistic maps of figure 1
We do not show the vertical component seismograms, which are mush smaller at all stations. Waveforms are plotted all at the same vertical and horizontal scale, and touch the triangles that mark the seismic stations that recorded them. Waveform data are derived from accelerations recorded at strong motion station operated by the California Division of Mines and Geology or are part of TriNet. Accelerations were filtered in the passband 0.05-0.5 Hz and integrated to velocity. Data were selected from all stations that recorded both earthquakes on scale and others to provide azimuthal coverage. Diagonal white lines indicate surface traces of the Hector Mine and Landers ruptures.
Earthquake triggering: A few good shakes is all it takes
Although nature rarely provides repeat experiments, the 1999, Mw=7.1 Hector Mine, California, earthquake very nearly did so. It confirmed inferences that seismicity rate increase were triggered remotely by transient, oscillatory dynamic deformations radiated as seismic waves from the 1992, Mw=7.3 Landers, California, earthquake. The close proximity and similarity of the two earthquakes permit direct comparisons of their deformation fields and testing of hypothesis about earthquake triggering not previously possible. To first order these earthquakes differed only in their magnitudes and rupture directions. The rupture direction and associated seismicity rate increase were northward for Landers and primarily southward for the Hector Mine earthquake. In this report we quantify the spatial and temporal patterns of the seismicity rate changes. The observations satisfy theoretical predictions that rupture directivity results in elevated dynamic deformations north and south of the Landers and Hector Mine faults, respectively. Recorded seismic velocity fields, which serve as strain proxies, document these asymmetries. Our analysis shows both dynamic and static stress changes to be important in the near-field with the later dominating at farther distances. Comparisons of the velocities for both earthquakes permit upper and lower bounds to be placed on dynamic triggering thresholds. These range between a few tenths and a few MPa in most places, depend on local site conditions, and exceed static thresholds by more then an order of magnitude. At some sites, the onset of triggering was significantly delayed after the dynamic deformations subsided. While not restricted to hydrothermal areas, triggering appears more likely in them.
We characterize the spatial seismicity rate change associated with the Hector Mine and Landers earthquakes using maps of smoothed β-statistics (Fig.1).These show that after the Landers earthquake, which ruptured northward, seismicity increased to the north. After the Hector mine earthquake, which ruptured southward, remote seismicity rate increases occurred to the south in three clusters; near Indio, the Salton Sea and Cerro Prieto. Rate increases were also observed after and to the north of both earthquakes at Long Valley.The post-Hector Mine increase there consists nearly entirely of M<2.0 events and was detectable only because a densely spaced seismic network operates at Long Valley.Time histories of cumulative seismicity reflected in the available earthquake catalogs (Fig.2) verify the Post-Hector Mine increases in these areas. The short delays suggest a causal link and their precise values provide important clues for differentiating between physical models of triggering. The first cataloged post-Hector Mines events occur 36 minutes, 1 hour, 1.35 days, and 40 minutes after the main shock at Indio, Salton Sea, Cerro Prieto, and Long Valley, respectively. To better constrain these delays examined continuously recorded ground motions at TriNet, US Geological Survey, and Berkeley Digital Seismic Network seismic stations, except near Cerro Prieto where no such stations exist (table 1). These data show a vigorous increase in local microearthquake activity near Salton Sea that commenced during the passage of Hector Mine waves, none at Indio prior to the first cataloged earthquake, and a delay of -10 minutes at Long Valley. The delays at Indio and Salton Sea are consistent with immediate triggering, give the average Post-Hector Mine inter-event times.We conclude that the delays at Cerro Prieto and Long valley are significantly longer than those expected by chance and longer than the duration of dynamic deformations experienced at each site.However, it should be noted that identification of early events can be difficult, and that the operational status of the network near Cerro Prieto during the study interval could not be verified.
The static stress change represented as the Coulomb Failure Function (ΔCFF) and its relationship to seismicity rate changes have been well studied for the Landers earthquake. We calculated ΔCFF for the Hector Mine earthquake on optimally oriented and other plausibly oriented planes. In the vicinity of Salton Sea, Cerro Prieto and Long Valley, the maximum absolute value of ΔCFF does not exceed 0.002MPa, which is smaller than peak-to-peak tidal variation (～0.003 MPa) and static ΔCFF triggering thresholds (～0.01MPa) inferred in previous studies .Moreover, ΔCFF is nearly spatially symmetric, whereas, with the exception of Long Valley (see below), the more remote seismicity increases occurred only to the south of the rupture. We conclude that static stress changes along could not have caused the remote seismicity rate change.＜p＞The correlation between the direction of rupture and seismicity rate change for both the Landers and Hector Mine earthquakes suggests a relationship between radiated seismic deformation and triggered seismicity. Indeed, superposing recorded seismic ground motion velocity records on theβ-static maps corroborates this correlation (fig.3).Seismic velocities have been shown to be approximately proportional to dynamic strains, both theoretically and empirically. The southern narrow band of seisimicity increase after Hector Mine coincides in azimuth with the large, pulse-like ground velocities that result from the southward rupture directivity (Fig. 3a). A similar coincidence is observed for Landers, but in the northward direction (Fig.3b). Some of the large amplitudes south of both ruptures undoubtedly reflect site-amplification due to thick Imperial Valley sediments. Nevertheless, the directivity is still apparent in the Hector Mine waveforms, which more pulse-like and large to the south then those for the Landers earthquakes. This is especially noteworthy given the large magnitude of the Landers earthquake.
The close proximity and similar source mechanisms of the Landers and Hector Mine earthquakes permits useful comparisons of their associated deformation fields cancel.
Because direct measurements come to be made at seismogenic depths, we assume that the relative magnitudes of surface deformations for both earthquakes are approximately the same as those at depth.
Measurements of peak velocities combined with estimates of seismicity rate change associated with each site (table 1) allow us to test the hypothesis that a dynamic triggering threshold exists, to consider whether it varies from site to site, and provide bounds on its magnitude. If true, this hypothesis would imply that at sites of Hector Mine-only seismicity increases Hector Mine peak strain (velocities) would exceed those from Landers. At remote distance, where static deformations should be insignificant, the velocity observations, albeit scant, are consistent with this hypothesis. Closer in, both static and dynamic deformations may encourage triggering. At the five Landers-only triggered sites that recorded both mainshocks, Landers peak velocities exceed those from Hector Mine at the two remote sites (GSC, PAS) but only at one of the three closer stations (Table 1). Of the six Hector Mine-only triggered or possibly triggered sites, both mainshocks were recorded at only two stations and Hector Mine peak velocities exceed those from Landers at both of these (Table 1).
Peak velocities from one earthquake along provide only upper or lower bounds on a triggering threshold at sites with or without seismicity rate increases, respectively.
By comparing peak velocities for both earthquakes, we obtain both upper and lower bound estimates (table 1).Although a single threshold value cannot satisfy the bonds at all sites, the observations suggest that thresholds span a small range between a few cm/s and a few tens of cm/s (10 cm/s corresponds to ～1 MPa and ～30 microstrain, assuming a rigidity of 3×10¹ MPa and a constant horizontal wave velocity of 3 km/s).
The measurements at Long Valley are an exception to this, indicating that the triggering threshold must be less than 1 cm/s (0.1 MPa or 3 microstrain). The reason for this extraordinary sensitivity is a subject for further research, but may explain why both the Landers and Hector Mine earthquakes triggering seismicity at Long Valley. All these threshold values must only be interpreted as suggestive, recalling that the data are at the surface, we do not resolve the deformations onto fault planes, and deformations at depth undoubtedly differ.
The seismicity associated with the 1999, Mw=7.1 Hector Mine, California provided a rare repeat experiment on dynamic triggering. The correlation between the direction of rupture, of enhanced seismic ground velocities, and increased seismicity rate for both earthquakes provides strong evidence of a causal relationship. It also demonstrates that the remote triggering that followed the Landers earthquake was not a fluke. The proximity and similarity of these two earthquakes allowed us to compare the observed deformation fields directly, thereby providing additional support, independent of model calculations, that the dynamic deformations were likely triggering agents. It also allowed us to estimate the magnitudes of dynamic triggering thresholds. The dynamic thresholds we estimate are generally consistent with peak stresses estimated at sites of seismicity rate increase following the Landers earthquake, and exceed △CFF triggering thresholds by more then an order of magnitude. Unlike static stress changes, transient deformations cannot permanently alter the applied stress and thus can only enhance the likelihood of failure by altering some physical property of or near the triggered fault. Our intuition and theoretical calculations suggest that such alteration requires large deformations, of the order of the dynamic thresholds estimated here. Our observations provide new evidence to support the hypothesis that transient dynamic deformations can trigger earthquakes, but the mechanism(s) by which they do so remain to be determined. Evidence presented here for a possible triggering threshold and for significant time delays between the shaking and earliest triggered earthquakes provide clues for elucidating and constraining the triggering mechanism.