Subsections

Remote Triggering of Fault-Strength Changes on the San Andreas Fault at Parkfield

Taka'aki Taira, Paul G. Silver (Carnegie Institution of Washington), Fenglin Niu (Rice University), and Robert M. Nadeau

Introduction

Figure 2.17: (a) Map view of the study area. $D$($t$) for sequence K3 during (b) 1987-2008 and (c) 2004-2008. (d) $\hat{T}_{r}$ and $M_{0}$ normalized by its average for sequence K3. (e) Average $D$($t$) using the lowest-noise stations VCA, LCCB, and CCRB, for sequences K2 and K5.
\begin{figure*}\begin{center}
\epsfig{file=taira09_1_1.eps, scale=1}\end{center}\end{figure*}

Fault strength is a fundamental property of seismogenic zones, and its temporal changes can increase or decrease the likelihood of failure and the ultimate triggering of seismic events. While changes in fault strength have been suggested to explain various phenomena, such as the remote triggering of seismicity, there has been, to our knowledge, no means of actually monitoring this important property in situ. Here we argue that $\sim$20 years of observation (1987-2008) of the Parkfield area at the San Andreas Fault have revealed a means of monitoring fault strength. We have identified a long-term change in fault strength most likely induced by the 2004 $M_{w}$ 9.1 Sumatra-Andaman earthquake (SM04) (Taira et al., 2009). The change possessed two manifestations: temporal variations in the properties of seismic scatterers - likely reflecting the stress-induced migration of fluids - and systematic temporal variations in the characteristics of repeating-earthquake sequences that are most consistent with changes in fault strength.

Fault-Strength Change

The time-varying properties of seismic scatterers have recently been used to probe stress-induced changes in the San Andreas Fault zone near Parkfield in central California (Niu et al., 2003; Taira et al., 2008). As a measure of temporal scatterer behavior from one earthquake to the next, we use the decorrelation index $D$($t$)=1- $C_{\mathrm{max}}$($t$) derived from the cross correlation of two seismograms, where $C_{\mathrm{max}}$($t$) is the maximum cross correlation (Niu et al., 2003). Using tightly-clustered repeating microearthquakes (Nadeau and McEvilly, 1999) recorded by the High-Resolution Seismic Network (Figure 2.17a), we have been able to track the behavior of a group of time-dependent scatterers (the target scatterer) that was first identified by Niu et al. (2003), for a 22-year period (1987-2008) (Figure 2.17b). The target scatterer has been interpreted as fluid-filled fractures, and their temporal variations as due to the stress-induced migration of fluids near the target scatterer.

There are three excursions in $D$($t$). The $1^{\mathrm{st}}$ excursion is coincident with the 1993 Parkfield Aseismic Transient (PAT93). It initiated around 1993 (Figure 2.17b), peaked in the mid 1990s and slowly decayed over about a subsequent $\sim$7-year period. The $2^{\mathrm{nd}}$ excursion (September 2004) is associated with the 2004 Parkfield earthquake (PK04) (Figure 2.17c) and decayed back to the pre-earthquake level after about 2-3 months. The $3^{\mathrm{rd}}$ excursion in $D$($t$) occurred about three months after the 2004 Parkfield earthquake (Figure 2.17c). The magnitude of change in $D$($t$) is comparable to that observed for the other two transients. The increase in $D$($t$) takes place over three months, after which time $D$($t$) decays slowly over a subsequent $\sim$1 year period (Figure 2.17c).

There are also changes in repeating-earthquake properties that accompany this $3^{\mathrm{rd}}$ excursion in $D$($t$). Following the 2004 Parkfield earthquake, there is a characteristic increase in recurrence interval $T_{r}$ as is typically observed postseismically. This trend, however, is interrupted roughly three months after the 2004 Parkfield earthquake. In 6 of the available 13 sequences, there is a systematic reduction in $T_{r}$ that reaches a minimum about 6 months after the 2004 Parkfield earthquake. To explore this apparent disruption further, we removed the post-seismic effect by the 2004 Parkfield earthquake, assuming Omori's law, $T_{r}$($t$)=$at^{p}$ where $t$ is time after the earthquake and $a$ and $p$ are constants to be estimated, and computed the residual recurrence interval $\hat{T}_{r}$($t$)=$T_{r}$($t$)/($at^{p}$).

Using all available sequences, we find that the log of the residuals increases by roughly a factor of 2 beginning three months after the 2004 Parkfield earthquake. This increased variability suggests an additional perturbation to $T_{r}$ and a temporal change in the mechanical properties of the fault. An interesting feature of this variability is that there is a positive correlation between $\hat{T}_{r}$ and seismic moment $M_{0}$ (Figure 2.17d). Such a correlation is consistent with a slip-predictable model for earthquake occurrence (Shimazaki and Nakata, 1980), where the stress drop and $T_{r}$ are both determined by the failure strength of the fault for constant loading rate and constant minimum stress. Assuming rupture area is constant among members of a repeating-earthquake sequence (Nadeau and Johnson, 1998), then fault slip will be proportional to $M_{0}$ which should in turn be proportional to stress drop. The most dramatic correlation that we found is for sequence K3 (5 km depth) (Figure 2.17d). This sequence also has the least variability in event location, so that the assumption of constant fault area should be the most valid. A reduction in $\hat{T}_{r}$ (and $M_{0}$) around 6 months after the 2004 Parkfield earthquake indicates a temporary weakening of the fault.

We have localized the onset time of the $3^{\mathrm{rd}}$ excursion in $D$($t$), utilizing all of the repeating-earthquake sequences that displayed a change in $D$($t$). We find that the excursion must have initiated between December 21 (sequence K2) and December 26, 2004 (sequence K5) (Figure 2.17e). The most dramatic tectonic event to occur within the 5-day time window is the 26 December 2004 $M_{w}$ 9.1 Sumatra-Andaman earthquake, whose origin time is 7 hours before the end of the interval. The timing strongly suggests that the dynamic stresses from this earthquake, estimated to be about 10 kPa [based on the amplitude of long period surface waves ($>$30 s) which are likely to have the strongest impact on fluid flow], induced fluid flow that caused both a structural change in the fault zone region [i.e. changes in $D$($t$)] and, through variations in pore pressure, consequent changes in the strength of the fault. It is now well documented that such dynamic stresses are capable of remotely triggering seismicity. The present study suggests that these same dynamic stresses can actually produce long-term ($\sim$1 year for our study) changes in fault strength.

Conclusions

We showed that temporal changes in seismic scatterer properties and characteristics (frequency and magnitude) of repeating microearthquakes constitute a proxy for changes in fault strength, proposing that they provide a means of continually monitoring fault zone strength which is an important parameter in assessing earthquake potential of a fault.

Acknowledgements

The present study was supported by the National Science Foundation EAR-0337308, EAR-0408947, EAR-0409024, EAR-0453638, EAR-0537641, EAR-0544730, and DTM-2025-01 and by the Department of Terrestrial Magnetism, Carnegie Institution of Washington. The Parkfield High-Resolution Seismic Network is operated by University of California, Berkeley Seismological Laboratory with financial support from the U.S. Geological Survey through National Earthquake Hazards Reduction Program award 07HQAG0014.

References

Nadeau, R.M. and L.R. Johnson, Seismological studies at Parkfield VI: Moment release rates and estimates of source parameters for small repeating earthquakes, Bull. Seismol. Soc. Am., 88, 790-814, 1998.

Nadeau, R.M. and T.V. McEvilly, Fault slip rates at depth from recurrence intervals of repeating microearthquakes, Science, 285, 718-721, 1999.

Niu, F., P.G. Silver, R.M. Nadeau, and T.V. McEvilly, Migration of seismic scatterers associated with the 1993 Parkfield aseismic transient event, Nature, 426, 544-548, 2003.

Shimazaki, K. and T. Nakata, Time-predictable recurrence model for large earthquakes, Geophys. Res. Lett., 7, 279-282, 1980.

Taira, T., P.G. Silver, F. Niu, and R.M. Nadeau, Detecting seismogenic stress evolution and constraining fault zone rheology in the San Andreas Fault following the 2004 Parkfield earthquake, J. Geophys. Res., 113, B03303, doi:10.1029/2007JB005151, 2008.

Taira, T., P.G. Silver, F. Niu, and R.M. Nadeau, Remote triggering of fault-strength changes on the San Andreas Fault at Parkfield, Nature, doi:10.1038/nature08395, 2009 (accepted for publication).

Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail: www@seismo.berkeley.edu
© 2007, The Regents of the University of California