Multi-scale compressional wave velocity structure of the San Gregorio Fault zone



[1] Understanding fault architecture at multiple scales is crucial to delineate in situ fault zone physical properties and rupture dynamics through modeling and geophysical imaging/monitoring. An exposure of the active large-offset, strike-slip San Gregorio Fault at Moss Beach, CA provides a unique field site to relate the well-mapped fault zone architecture with compressional wave velocity (Vp) structure measured at centimeter to meter scales. Laboratory ultrasonic velocities of fault zone samples, adjusted for fluid-related frequency and structural dispersion, indicate that (i) a seismic velocity reduction of ∼30% characterizes the central smectite-rich clay gouge relative to the rocks 100 m away in the relatively undeformed host rocks, and (ii) the across-fault velocity profile trends for the seismic to ultrasonic bandwidth correlate almost exactly to the previously mapped macroscale fault zone structure. These results highlight the value of conducting multiscaled investigations when measuring fault zone properties defined by physical elements at multiple scale lengths.

1. Introduction

[2] Characterization of the in situ spatial variation of lithologic composition, structural properties (e.g., nature and extent of fracturing, fabric development, and cataclasis), constitutive relations, fluid content, and pore fluid pressure in fault zones is required to understand the controls on fault strength, earthquake source mechanics, and temporal change during the earthquake cycle. These fault zone properties have been shown to correlate with wave propagation phenomena such as low seismic velocities, strong seismic reflectivity, and anomalous Vp/Vs ratios.

[3] The complex architecture and hydrologic properties observed in exhumed fault zones [Chester and Logan, 1986; Caine et al., 1996; Rawling et al., 2001], however, indicate that key in situ physical properties are under-resolved and/or non-uniquely determined in the most detailed seismic imaging of kilometer-deep structures. Even when near-surface seismic experiments are designed to approach the required resolution [e.g., Pratt et al., 1998; Li et al., 2001], the resulting seismic structure may be overprinted by exhumation artifacts, such as diagenesis, saturation heterogeneity, and near-surface sedimentation.

[4] In this paper, we report results of an unique high-resolution, multi-scale laboratory (cm) and field measurement (m) campaign to determine the P-wave velocity structure at the surface (<2 m depth) of the San Gregorio Fault (SGF) at Moss Beach, CA (Figure 1). After briefly describing the SGF and laboratory analyses, we present a simplified theoretical method to link the two spatial scales to predict seismic velocities from ultrasonic counterparts. Finally, a comparison of these predicted seismic velocities with both (i) the previously detailed geologic mapping of the fault zone exposure, and (ii) a 1-m resolution velocity reconstruction developed from the field experiments, highlights the advantages of studying a complicated geologic structure by measuring physical properties at multiple scales.

Figure 1.

Field map of the San Gregorio Fault exposed at Moss Beach, CA [modified from Lohr et al., 1999]. The fault zone is defined by seven structural elements. Inset map shows study site location and regional fault zones (SGFZ - San Gregorio Fault Zone; SAF - San Andreas Fault; HF - Hayward Fault).

2. Near-Surface Ultrasonic P-Wave Velocity and Physical Property Estimates

[5] The San Gregorio Fault is part of the regional strike-slip San Andreas Fault system in west central California, with estimated slip rates of 5–11 mm/yr and horizontal displacement of 90–150 km [Simpson et al., 1997]. At Moss Beach, the 85oNE-dipping SGF exposure juxtaposes Miocene-aged Purisima Formation mudstones against sandstones of the same formation. Lohr et al. [1999] show seven structural elements over a ∼200-m distance perpendicular to the fault strike (Figure 1). On either side of the 17 m-wide foliated clay gouge at the fault zone core, tectonically mixed breccias (∼8-m thick in the NE block sandstones and ∼50-m thick in the SW block mudstones) transition to fractured versions of the intact host units. The steep dip, minimal sand cover, and continuous seawater saturation of the actively uplifting and eroding wave-cut platform containing the fault zone enhances preservation and exposure of structural components and thus provides an ideal environment for directly interpreting wave propagation experiments in terms of the fault zone architecture. Both rock sampling and seismic experiments were conducted below mean low tide to minimize desaturation artifacts and complications from beach sand cover.

2.1. Sampling and Methods

[6] Laboratory ultrasonic velocities were measured on forty-two ∼38-mm diameter core plugs extracted from twenty-five 300–800 cm3 samples which represent the principal characteristics of each architectural component. Where possible, sets of mutually orthogonal cores were extracted from a sample to test for directional dependencies in petrophysical and elastic properties. Care was taken to insure that no observable fractures penetrated any of the cores and that in situ saturation was maintained after sampling. All cores were subjected to 24 hours (minimum) of vacuum-assisted seawater resaturation prior to laboratory testing.

[7] Ultrasonic P-wave velocity measurements, at atmospheric (and elevated) pressure, were made following standard pulse transmission techniques. Absolute velocity errors are 0.5-2%, controlled primarily by core face parallelism and length measurements (±0.01 mm). Porosity (±0.01) and density (wet bulk, ρbw, and dry, ρd, with uncertainty of ±0.02 g cm−3), at atmospheric pressure, were measured on sample splits [Blum, 1997].

2.2. Results

[8] Table 1 summarizes the near-surface physical properties and Vp within each structural element of the fault zone. The ultrasonic velocities (VPultra) within any particular element show variability between 0.15–0.45 km s−1. Decreasing VPultra with increasing deformation toward the fault core is clearly apparent. The key result is the large velocity reduction of >30% between the foliated clay gouge and intact mudstones, consistent with but significantly greater than that inferred at the similarly scaled Punchbowl Fault [Li et al., 2001].

Table 1. Near-Surface (<2 m) Physical Property and P-wave Velocity Measurements at the San Gregorio Fault Exposure at Moss Beach, CAa
Structural memberρbw (g cm−3)ϕ (%)VPultra (km s−1)ΔVfl(%)ΔVstr(%)Vpr. seis (km s−1)
  • a

    The range for each measurement includes all cores (or subsamples thereof) within a structural element, and results from [Tobin and Lohr, 1998].

int. mudstone1.85–2.1035–202.45–2.80<1<22.35–2.70
frac. mudstone1.80–2.0535–202.10–2.45<1<22.00–2.35
br. mud. gouge1.75–1.9030–202.05–2.20<2<21.95–2.10
fol. clay gouge1.75–1.9040–301.85–2.00<na01.85–2.00
sand. breccia      
  - fine gr.1.80–2.1040–252.00–2.35<22–101.75–2.25
  - coarse gr.1.90–2.1535–202.90–3.10
frac. sand stone1.75–2.0540–201.90–2.30<10−2–151.45–2.10
  pebble/shell cong.2.10–2.3025–153.00–3.30
int. sandstone1.75–2.3035–152.15–2.60<10−2–151.65–2.55

3. Upscaling in Frequency and Space

[9] In this section, we first present methods to adjust ultrasonic velocity measurements to their seismic equivalents by directly accounting for two potential frequency-dependent (dispersive) effects. Dispersion between seismic and ultrasonic frequencies primarily depends on (i) viscoelastic effects given the pore fluid types and relative saturations [Winkler, 1985; Mavko and Mukerji, 1998], and (ii) sampling scale, especially related to fine-scale layering [Hovem, 1995; Mukerji et al., 1995] and microfracture intensity [Peacock et al., 1994; Hudson et al., 2001]. The predicted seismic velocities are then compared to the velocity structure developed from field experiments.

3.1. Fluid Effects

[10] We address fluid-related dispersion as described by Mavko and Mukerji [1995] applied to macroscopically homogeneous media as defined by Brown and Korringa [1975]. We predict the velocity at seismic frequencies by jointly analyzing saturated and dry ultrasonic velocity-porosity datasets measured during elevated confining (isotropic) pressure tests on each core. Dry ultrasonic and seismic velocities are equivalent in our laboratory tests since all wavelengths over the ultrasonic bandwidth are much larger than the typical scale of heterogeneity (e.g., a pore and immediately surrounding grain matrix) [Winkler, 1985]. Dry ultrasonic velocities were converted to effective bulk (K) and shear moduli, producing an unique K/Ko vs. porosity dataset, where Ko is the effective mineral bulk modulus. The Kdry/Ko-porosity dataset defines a unique line, extending through (0, 1) for the appropriate Ko. The Biot-Gassmann model [e.g., Winkler, 1985] was then used to predict the low frequency, seawater-saturated K-and hence, seismic velocity-at each porosity/pressure state. VPultra data dictate an upper limit to the predicted seismic velocity for any porosity or effective pressure state. We estimate the near-surface, fluid-related dispersion (ΔVfl in Table 1) as the difference between VPultra and the Biot-Gassmann saturated seismic velocity at atmospheric pressure.

[11] The benchtop and elevated-pressure VPultra represent wave propagation through continuously seawater-saturated, non-macroscopically fractured media. Following the saturated pressure tests, we allowed the cores to dry out, over a period of 5–10 days, to a partial saturation state in equilibrium with the ambient room humidity of 10–20%. These room-dry cores were then tested at elevated pressures to obtain the equivalent “dry” ultrasonic/seismic velocity measurements.

[12] Axial strain during elevated pressure testing was measured by a calibrated linear potentiometer with resolution of ±0.25 μm so that the relative velocity uncertainty between two pressure states is ∼0.1%. At each elevated pressure, porosity and density were estimated from the axial strain measurement assuming that the axial: radial strain ratio was 1.00 ± 0.05. This range was supported by measurements of post-experiment total inelastic axial:radial strain ratios. This uncertainty results in estimated physical property differences of less than ±0.01 and ±0.02 g cm−3 at ∼30 MPa for porosity and wet bulk density, respectively.

[13] Results from representative samples of the NE block (sandstone, breccia) and the SW block (mudstone) are in good agreement with the bulk modulus-porosity Biot-Gassmann model. The effective Ko and predicted Δ Vfl for the sandstone, sandstone breccia, and mudstone elements (37 GPa, <10%; 29 GPa, <2%, and 25 GPa, <2%; respectively) are consistent with the bulk mineralogy, petroscopic evidence, and macroscale deformation features. We note that at low effective pressures, the distribution and intensity of microfracturing in the feldspathic breccia and mudstone cores appears to create a double-porosity fracture/matrix medium [e.g., Berryman and Pride, 2002] which converges to a macroscopically homogeneous medium above 2–6 MPa. Given the mechanically weak fracture porosity, low matrix porosity, and small pore size of these fine-grained materials, we have concluded that the predicted near-surface seismic velocities are artificially high relative to VPultra measurements, and have thus used an upper bound on the fluid effect (2%) reported in Table 1.

3.2. Structural Effects

[14] Material heterogeneity (e.g., sedimentary layering or macrofractures in 1D media) may induce an apparent velocity dispersion controlled by the λ/a ratio, with λ the wavelength of investigation and a the characteristic heterogeneity scale length [Mukerji et al., 1995]. Previously published maps [Lohr et al., 1999] and our geologic reconnaisance suggest that significant lithologic and/or macroscopic deformation variation occurs in almost every 1-m interval, with characteristic heterogeneity lengths of 0.05–0.50 m. Quantification of the potential dispersion may thus be crucial for making in situ near-surface seismic velocity predictions on the scale of the high-resolution field experiment conducted across the fault zone (Figure 1).

[15] Structural dispersion estimates (ΔVstr in Table 1) are derived for 1-m 1D synthetic representatives of each structural element. We use an anelastic propagator matrix formulation [after Kennett, 1974], assuming constant-Q for the seismic frequency bandwidth, to model compressional waves through layered media with layer properties parameterized by (i) the range of predicted seismic velocities (VPultra corrected for ΔVfl) for a specific lithology (Table 1), (ii) bulk density range as given in Table 1, and (iii) a series of layers with thicknesses of 0.05–0.50 m, generally organized as observed in the field.

[16] Our modeling produced a suite of phase velocities which was used to define the minimum and maximum structural dispersion for a 1-m thick layered sequence typical of each fault element (ΔVstr in Table 1). We note that the propagator matrix approach produces phase velocities in excellent agreement with ray theory (Vrt) and effective medium (Vem) limits by Mukerji et al. [1995] if the acoustic impedance variation is small (<15%) over the 1-m interval [Gettemy et al., 2001; Pride et al., 2002]. For large boundary impedance contrasts, however, the phase velocities may exhibit dispersive effects up to four times greater than that defined by Vrt – Vem. The transition frequency below which P-waves propagate at the Vem equivalent occurs between 300–500 Hz, for all 1-m synthetics, consistent with both (i) the maximum usable frequency obtained from our field recordings (<200 Hz), and (ii) modeling results of Pride et al. [2002]. In general, ΔVstr are concluded to be small (Table 1) except in the NE block where strong acoustic impedance contrasts between the fine- and coarse-grained materials in the breccia, or between sandstone-pebble conglomerates, are observed.

3.3. Seismic Velocity Profile Comparison

[17] Applying the Biot-Gassmann and structural effect adjustments (ΔVfl and ΔVstr, respectively) to the benchtop ultrasonic measurements produces a fault-perpendicular seismic velocity structure (Vpr.seis in Table 1 and Figure 2) consistent, in trend, with a 2D tomographic reconstruction of first arrival traveltimes [Sayed, 2001]. While the Vpr.seis is higher in absolute terms, the two profiles show that (i) the structural boundaries correlate almost exactly with changes in P-wave velocity measured at both scales, and (ii) meter-scale variability of the field experiments can be explained by the range of Vpr.seis within each fault zone element. The low tomographic velocities over the mixed breccia mudstone, central foliated clay gouge, and mixed breccia elements appear to show the combined influences of the 0.30–1.3 m of sand deposited over this region of mechanically weak, severely wave-eroded rock, and large seismic propagation wavelengths (∼4 m in the partially saturated sand, ∼20 m in the underlying rock).

Figure 2.

Across-fault P-wave velocity profiles corresponding to saturated benchtop ultrasonic measurements (VPultra range - gray bands; core measurements - black rectangles), seismic frequency predictions after adjusting VPultra for fluid- and structure-related dispersion (Vpr.sseis, cross-hatch), and travel-time tomographic reconstruction from field experiments (solid line) [after Sayed, 2001]. The VPultra core measurement range corresponds to the minimum and maximum VP for the mutually-orthogonal cores at a given sample location.

4. Conclusions

[18] Multi-scale mechanical and hydrologic processes controlling strength, state of stress, and seismogenic behavior of crustal fault zones must be better understood to develop a more robust, perhaps predictive, knowledge of fault properties and ultimately seismic risk and hazard assessment. The first step toward gaining this understanding is through improvements of in situ multi-scale seismic characterization of the fault zone itself. Our analyses illustrate that centimeter-scale laboratory ultrasonic velocities can be upscaled to their seismic equivalents by correcting for velocity dispersion through a combined application of the Biot-Gassmann model and 1D layered structure modeling using the propagator matrix method. Our results at the San Gregorio Fault exposure at Moss Beach, CA, also demonstrate that variations in petrophysical measurements at the laboratory scale correlate to the well-defined structural architecture of the ∼200-m wide fault zone, with the central foliated clay gouge imaged as a ∼30% low velocity anomaly relative to the undeformed host rock. This multi-scale approach should be integral to planned fault zone studies (e.g., SAFOD-Parkfield surface and borehole experiments, IODP-Nankai seismogenic zone drilling) that will require scaling analyses to interpret broadband (seismic to ultrasonic) wave propagation phenomena in terms of spatially complex dynamic fault zone processes and architectures.


[19] This work was funded by NSF-EAR-9909270 to Tobin and Hole. We thank the staff of the Fitzgerald Marine Reserve for allowing access and coordination for the field experiments and rock sampling. We also thank Laurel Goodwin (NMIMT) whose expertise in field geology guided our sampling endeavors, and Steve Franklin and Sunny Baer for assistance in core drilling and preparations.