Parkfield Studies

T. V. McEvilly, R. M. Nadeau, R. W. Clymer, R. A. Uhrhammer, Collaborating Scientists: L. R. Johnson, C. G. Sammis, W. Foxall, V. Korneev

• Background
• Fault-zone Guided Waves
• Current Emphases
• References

Background

A high-resolution seismographic network (HRSN) for the Middle Mountain stretch of the San Andreas fault zone, where the M6 event is presumably nucleating, was installed in boreholes beginning in 1986. In November 1987, the Varian well vertical array was installed and the first VSP survey was conducted, revealing clear S-wave anisotropy in the fault zone. During 1988, the network was completed to ten 3-component 500 sps radio-telemetered stations, incorporating a deep (572 m) sensor in the Varian well string into the network. The Varian system was slaved in 1988, for about two years, to the Vibroseis control signals, allowing simultaneous recording of vibrator signals on both systems. In 1991, low-gain event recorders (from PASSCAL) were installed to extend the dynamic range to M_L about 4.5. In 1991, we determined a local 3-D velocity model by joint inversion, and data reduction methods were designed to handle the massive data sets in a monitoring mode. Since 1989, new analysis techniques gradually have been developed to extract and display the information contained in the high-resolution data from both Vibroseis and microearthquakes to support a variety of research efforts in the project. The daunting technical and data processing requirements for this project were met only through the participation of several graduate students and technical personnel at the Lawrence Berkeley National Laboratory Center for Computational Seismology and Geophysical Measurements Facility. The system has recorded more than 10000 microearthquakes in the Parkfield region along with many times that number of regional and teleseismic events. More than 5000 good quality 3-D locations make up the catalog for the 40 km stretch of the fault zone centered at Middle Mountain. More than 50 controlled-source data sets from the Vibroseis monitoring program have also been gathered from mid-1987 until its termination in 1997.

About two-thirds of all the Parkfield earthquakes are members of a few hundred sequences of as many as 20 or more characteristic events - regularly recurring earthquakes so similar that waveform coherency exceeds 0.98 over a 50Hz+ bandwidth for the entire wavetrain, Figure 15.1. Relative hypocenter locations for these types of events differ only by 5-10 meters, near the resolution of our analysis. This phenomenon may be common to creeping seismogenic fault zones in general, but there are not the high-resolution networks in place to test this at the -1 <M< 2 level where recurrence times are less than a few years. This new view of fault-zone process is fundamental to our research program, and we are completing the characterization of the fault zone up to the present time while also extending the characterization of fault seismicity to the entire creeping part of the full San Andreas system. 'Characteristic' implies more than just 'similar', a criterion easily satisfied by nearby events when observed at lower bandwidth. Very narrow moment and recurrence

  

Figure 15.1: Step diagram of seismic moments and recurrence intervals (top) and vertical component seismograms recorded at Parkfield on borehole station VCA (inoperative for event 18) (bottom) for a repeating earthquake sequence containing 19 events. Among events in repeating sequences, seismic moments, waveforms and locations are virtually identical and recurrence times are quasi-periodic.

interval distributions are characteristic of the repeating events we are working with, Figure 15.2.

  

Figure 15.2: Distributions of sequence-normalized seismic moments (top) and recurrence intervals (bottom) for 160 repeating earthquake sequences. Coefficients of variation (COVs) for moments and intervals are 0.22 and 0.49, respectively, for the full data set (filled black bars). COVs for recurrence intervals are 0.42 and 0.32, respectively, for subsets (right scale) of larger (magnitude > 1.0) events (broken line) and events occurring before June 1992 (solid gray line). These distributions characterize the size and recurrence time variability for all 160 repeating sequences and provide the basis for a variety of unique analyses now possible using repeating sequences. These include slip rate estimates at depth, the augmentation and testing of statistically based earthquake recurrence and forecast models, and the estimation and mapping of fault healing rates in nature.

There are immediate consequences of characteristic recurrence. One striking implication follows from the near-constant moment release rate in a sequence. If we can independently estimate the associated slip rate on the fault at the sequence, an entirely new approach to source parameter estimation is possible that is independent of spectral corners, yielding a set of scaling relations among moment, area, slip and stress drop that seem to hold for the range 0 <M< 6 on the San Andreas fault in central California. An inescapable result is a strong moment dependence for stress drop, with kilobar+ values implied for the small events. There is supporting petrological evidence in exhumed fault zones for these extreme hypocentral conditions.

Systematic changes in recurrence intervals are seen in some sequences, equivalent to changes in moment release rates, implying change in slip rate. The characteristic sequences thus become strain meters within the fault zone, and we have begun to use this idea to map the changing slip rate throughout the fault zone. The results are generally consistent with changes expected to accompany the 10/92-12/94 M4.5+ major earthquake sequences, and there is good correlation of recurrence-derived slip rates with surface-based measurements of deformation.

The highly structured seismicity at Parkfield stands in stark contrast to the much less organized picture of earthquake rupture and fault deformation painted by conventional lower resolution data sets. When high-resolution data are processed using advanced techniques, the appearance of diffuse uncorrelated seismicity largely resolves itself into an organized pattern of repeating earthquake groups having the same size, location, waveforms and constant recurrence interval. In addition, the hypocenters of most earthquakes become restricted to a narrow planar core of seismic strain release within which the repeating earthquake sites cluster on several scales - clusters of clusters. Fractal analyses of the highly resolved hypocenters show a considerably lower fractal dimension, (D $\sim$ 1), than that observed and predicted for Parkfield based lower resolution data (typically, D = 2 to 3). In addition, the fractal pair-correlation function shows discrete behavior (undulations) indicating a strong heterogeneity in the spatial distribution of earthquakes on the SAF at Parkfield. This discrete behavior is masked, however, by uncertainties in location when less well resolved catalogs such as NCSN are used, Figure 15.3.

  

Figure 15.3: 3D two-point correlation sum verses hypocenter separation distance for Parkfield earthquakes recorded on the High Resolution Seismic Network, HRSN (filled circles), and the Northern California Seismic Network, NCSN (open triangles). The slopes of the curves provide estimates of the D2 fractal dimension for the distribution of hypocenters. The discrete fractal behavior and low fractal dimension seen using the HRSN data are masked by large location errors typical of routine high quality catalogues such as the NCSN.

At Parkfield, repeating earthquakes comprise a large part of the earthquake population, yet their fraction of the total seismicity is not constant through space and time. Rather, this fraction exhibits systematic variations which appear to be closely related to fault slip and the earthquake process. Some subsections of the fault at Parkfield, for example, show a cyclic pattern of decreasing amplitude in the time evolution of the fraction of repeating earthquakes, Figure 15.4.

  

Figure 15.4: Fraction of repeating earthquakes through time for a 14 by 6 km subsegment of the SAF enclosing the nucleation region of the 1966 M6 earthquake at Parkfield. The figure shows that the relative number of repeating to non-repeating events is not constant through time but varies systematically with a 1.5 to 2 year cyclicity and decreasing peak amplitudes occurring subsequent to the late 1992 increase in M4.5+ activity at Parkfield.

We have found differences among seismogenic fault segments in central California, apparently related closely to slip rate. For example, this new measure of fault activity varies from about 70$\%$ at Parkfield, using the borehole network data base (48$\%$ for the same region using NCSN waveforms only) to 13$\%$ and 6$\%$ respectively for NCSN data on the southern Hayward/Mission and northern Hayward faults. At Parkfield, the fraction, as well as the total seismicity, drops rapidly as the locked part of the fault is approached. We have begun a more intensive investigation of this phenomenon in order to characterize the San Andreas and subsidiary faults from Parkfield to the Bay Area, where we can again tie in the NCSN/borehole performance difference using the developing Hayward Fault Borehole Network. The goal of this proposed effort is to test the utility of the repeating earthquake statistic and the waveform similarity characterizations to help delineate fault properties and fault segment boundaries based more quantitatively on what appears to be a measure of fault slip rate.

We have begun a detailed characterization of waveform similarity with earthquake separation that we have named the 'Cluster Signature' (see figure in the Hayward Fault section) Highly similar repeating earthquakes are co-located to a spatial resolution proportional to the signal bandwidth - about 5-10 m for the 100 Hz borehole data at Parkfield. To map the Cluster Signature throughout the San Andreas system in central California, we used the NCSN (surface sensors) waveforms and catalog for events recorded at Parkfield and compared these results to the higher-resolution picture from the borehole network to calibrate the NCSN sensitivity. We then apply the method to characterize the central San Andreas and Bay Area faults using the more broadly distributed NCSN. The NCSN-determined Cluster Signature is sufficient to characterize some important aspects of waveform similarity and clustering in the fault zone and it shows distinct differences in the seismic signature of these zones both in the degree of hierarchical clustering and in the proportion of clustered earthquakes. These differences imply that potentially useful information on the variable character of California faults can be detected and related to features indicative of fault stress such as hierarchical clustering using the Cluster Signature approach.

Fault-zone Guided Waves

There has been a lot made of so-called fault-zone guided waves (FZGW). Much of it has been directed toward modeling wave propagation in relatively simple, vertical low-velocity structures in order to match discrete observations of the late, low frequency arrivals sometimes recorded near the fault trace. We have begun to explore this problem from a somewhat different approach, using the extensive observations of these waves in the Parkfield network, the 3-D P- and S-velocity model for the fault zone, and our Vibroseis results that place an apparent strain-related zone of changing wave propagation parameters within the shallow (the upper 3 km) part of the fault zone. In this two-fold investigation we have begun to characterize the distribution throughout the fault zone of source-receiver paths that produce strong FZGW signals from earthquakes. The goal of this part of our research is to be able to first determine the patterns of generation and propagation of FZGW, to characterize the wavefield in terms of velocity and particle motion relative to the fault zone, and finally, to model the phenomenon numerically using new 3-D guided- wave algorithms under development. Our initial work on this is suggesting that the strong generation as well as the wave propagation is a shallow feature of the fault zone. We would like to take this study farther as part of our ongoing Parkfield research plan.

Current Emphases

The current emphases in the Parkfield project can be illustrated with excerpts from several recent papers:

• Nadeau and McEvilly (1999) devised a method for estimating fault slip rate "... recurring highly similar microearthquakes at Parkfield, California provide a means for inferring slip rate throughout the active fault surface from the time intervals between sequence events. ...an 11-year high-resolution microseismicity record revealed systematic spatial and temporal changes in the slip rate that were synchronous with earthquake activity and other independent indicators of fault-zone slip", Figure 15.5, which provide "...the basis for a new method to
  
    

  Figure 15.5: (Top) Seismicity rate smoothed with a 2-month running window. (Center) Variations in recurrence-derived slip rate (solid) and sequence-normalized seismic moment (dashed) for 160 repeating microearthquake sequences at Parkfield, CA. The left scale is absolute slip rate, and the right scale is relative change in slip rate and moment. (Bottom) Surface measured slip rate from the two-color laser geodimeter (Langbein et al., in press) at Parkfield

  
  
  monitor the changing strain field throughout the seismogenic fault zone, adding the depth dimension to conventional surface-constrained measurements of deformation."

• Korneev et al. (1999) provide a quantitative model of the Vibroseis travel-time anomaly: "... one element of the Parkfield, CA Prediction Experiment, the borehole seismographic network there was illuminated routinely by a large shear-wave Vibroseis from several source points... Clear and progressive travel-time changes of up to 50 msec were detected, most prominent in the S-wave coda, and localized to propagation paths through the fault zone southeast of Middle Mountain, the section of the fault where previous M6 earthquakes have initiated. We model the observations successfully as interaction (reflection and transmission) of the shallow wavefield with a 200-meter-wide low-velocity fault zone in which the velocity increases by 6$\%$ due, we hypothesize, to hydrological changes accompanying a significant pulse in slip rate and seismicity."

• Ellsworth et al. (1999), propose a new approach to earthquake probabilities: " A physically-motivated model for earthquake recurrence based on the Brownian relaxation oscillator is introduced. ..... Application of this model to the next M 6 earthquake on the San Andreas fault at Parkfield, California suggests that the annual probability of the earthquake is between 1:10 and 1:13."

• Sammis et al. (1999) use the Parkfield data in developing a fractal asperity model: "...Nadeau and Johnson [1998] found that the smallest events occurred on patches having a linear dimension of the order of 0.5 m, displacements of about 2 cm, and stress drops of the order of 2000 MPa, roughly 10 times larger than rock strengths measured in the laboratory. ... A hierarchical fractal asperity model is presented, which is based on recent laboratory observations of contact distributions in sliding friction experiments. ... The spatial distribution of hypocenters in the Parkfield area is shown to be consistent with this simple fractal model and with a hierarchical clustering of asperities having a fractal dimension of D = 1 and discrete resealing factor of about 20."

• Wenk et al. (1999) see evidence in psuedotachylites for faulting processes inferred from Parkfield microearthquakes: "...The pseudotachylites formed during a late brittle event (56-62 My) and postdated the large ductile mylonitic deformation in the Santa Rosa mylonite zone (65-87 My) as ascertained by 4OAr/39Ar ages. ... From the sizes, energies required for melting estimated to range between 105 J for smaller veins and 1010 J for larger ones. Interestingly the energy distribution and geometry of these pseudotachylites corresponds closely to energy distributions of current microseismic events along the San Andreas fault at Parkfield, suggesting that basic mechanisms may be similar and due to intrinsic mechanical properties of rocks."

References

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

Korneev, V.A., T.V. McEvilly and E.D. Karageorgi, Seismological Studies at Parkfield VIII: Modeling the Observed Controlled-Source Waveform Changes, Bull. Seism. Soc. Am. (submitted), 1999.

Ellsworth, Matthews, Nadeau, Nishenko, Reasenberg and Simpson, A Physically- Based Earthquake Recurrence Model for Estimation of Long-Term Earthquake Probabilities, Geoph. Res. L. (submitted), 1999.

Wenk, H-R., L.R. Johnson and L. Ratschbacher, A Large Psuedotachalyte Zone in the Eastern Peninsular Ranges of California, Tectonophysics (submitted), 1999.

Sammis, C.G.,R.M. Nadeau and L.R. Johnson, How strong is an Asperity, J. Geoph. Res., 104, 10609-10619, 1999.

Langbein, J.O., R. Gwyther, R. Hart, M.T. Gladwin, Slip rate increase at Parkfield in 1993 Detected by High-Precision EDM and Borehole Tensor Strainmeters, Geo-Phys. Res. Lett., in press

The Berkeley Seismological Laboratory, 202 McCone Hall, UC Berkeley, Berkeley CA 94720
Questions and comments to www@seismo.berkeley.edu
Copyright 1999, The Regents of the University of California.