Geophysical
Constraints on Seismic Hazard and Tectonics in the Western Basin and Range
Dr. John
N. Louie/Dissertation Advisor
Three
studies show the effectiveness of shallow exploration geophysics in solving problems
related to tectonics and seismic hazard. In Chapter 2, 200 new gravity
measurements are used to find depth-to-bedrock in the Reno and Carson City
areas. Depth-to-bedrock and shallow shear-wave velocities have been shown to
have an effect on ground motion from earthquakes, prompting the study. Maximum
basin depth in Reno is 1.2 km and maximum depth in Carson City is 0.53 km.
Quaternary fill might only rarely exceed 200 m. An unexpected and previously
unsampled gravity trough in Reno may increase seismic hazard there relative to
other parts of the basins. In Chapter 3, I use a combination of seismic and
gravity techniques to prove that the 1954, Ms 6.8, Dixie Valley,
Nevada, earthquake produced slip on a low-angle (<30°) normal fault. The
fault maintains a low-angle and slightly listric geometry to the depth of constraint
(1.75 km seismic; 2.7 km gravity). As such, the Dixie Valley event may
represent the first large, low-angle normal earthquake on land recorded
historically. This result may increase the awareness of the potential hazard posed
by low-angle normal faults. In Chapter 4 I conclude, from the study of weak
ground motion spectral ratios and in situ velocity characterization, that the
continued existence of precarious rocks in their untoppled state near the San
Andreas fault is not due to local site or path effects. Velocity characterization,
using a combination of refraction-microtremor and seismic tomography methods,
shows that the precarious rock sites are slower than the National Earthquake
Hazards Reduction Program BC boundary from 0 to 10 m, and much faster from 10
to 45 m. As a result, synthetic spectral ratios of small earthquakes at the
precarious rock sites relative to the BC boundary show de-amplification below 5
Hz and amplification above 5 Hz. One possible conclusion is that the current
seismic hazard relations overestimate the hazard near large strike-slip
earthquakes.
Thanks above all goes to my parents, who have been most patient and supportive over the many years of schooling. I greatly appreciate the help offered by my committee and co-authors. John Louie, in particular, was extremely supportive and was everything a good advisor should be. Thanks to the many great friends and fellow graduate students that helped keep me sane (on average), especially: Aline Concha-Dimas, Mark Engle, Keith Weaver, Ana Cadena, Nancy Glenn, Mandy Johnson, and Mike Ossofsky.
Chapter
1: General Introduction
Chapter
2: Depth to bedrock using gravimetry in the Reno and Carson City, Nevada, area
basins
2.4
Discussion and Conclusions
2.4.1
Constraints From a Previous Carson City Study
2.4.2
Constraints From Well Data
2.4.3
Origin of the Deep Sub-Basin
2.4.4
Seismic Hazard Estimates
3.1.1
Regional and Tectonic Setting
3.1.2
The 1954 Dixie Valley Earthquake
3.1.3
Geologic Evidence for Low-Angle Dip on the Dixie Valley Fault
3.1.4
Previous Geophysical Work in Dixie Valley
3.2.2.1
Conventional processing and poststack migration
3.2.2.4
Gravity data processing
4.2.5
Calculation of Synthetics
4.3.1
Local Shear-Velocity Structure
4.3.2
Horizontal and Vertical Spectra
4.3.4
Synthetic Spectral Ratios with Piute Butte
4.3.5
Synthetic Spectral Ratios with Two Reference Models
4.4
Discussion and Conclusions
Chapter
5: Recommendations for Future Work
5.1
Recommendations for Future Reno-Carson City Basins Work
5.1.1
Increased Gravity Coverage
5.1.3
Finite Difference Modeling
5.1.4
Test Ground Motion Predictions Using New ANSS Stations
5.1.5
Shallow Velocity Measurements
5.2
Recommendations for Continued Dixie Valley Research
5.2.1
Static and Dynamic Stress Modeling
5.2.1.1
Differentiating among the possible mechanisms for fault weakening.
5.2.2
Electromagnetic Study of the Fault Zone
5.2.2.1
Dixie Valley Evidence Supporting an Elevated Pore-Fluid Model
5.2.3
Gravity Survey to Study Low-Angle to High-Angle Accommodation Zones
5.2.3.1
Modeling and Interpretation
5.3
Recommendations for Future Work in the Precarious Rock Study
5.3.1
Optimize the ReMi equipment
5.3.3
More ReMi Data in Varied Environments
5.3.5
Measurements at Precarious Rocks in Different Environments
Chapter 1: General Introduction
This dissertation is a collection of
three loosely related geophysical investigations. Chapter 1 is this
introduction. Chapters 2, 3, and 4 are reformatted versions of three publications.
Chapter 2 is published in Geophysics [Abbott and Louie, 2000]; Chapter 3 is
published in the Journal of Geophysical Research [Abbott et al., 2001a]; and
Chapter 4 is, as of this writing, in submission to the Bulletin of the Seismological
Society of America [Abbott et al., 2001b]. Chapters 2 and 3 are presented with
formatting changes only, and the respective journal articles are essentially
identical to the chapters herein. The reader may find it worthwhile to look for
the BSSA version of Chapter 4 for more information, however. Chapter 5 contains
recommendations for future work that will only be found in this dissertation. A
supplemental CD-ROM is included that contains digital copies of all figures in
PostScript and PDF formats, including color for Figures 2.1, 2.3-6, 2.9-12, 4.4-5,
and 5.3.
Although the proposals and research plans for the three projects presented in this dissertation were not written with each other in mind, the projects have much in common. It is what they have in common that drew my interest and kept my attention over the years creating this work. Although the three field areas can be loosely grouped in or near the Basin and Range province of the western United States, it is not their location that is the uniting thread. Nor is it the precise methodology of investigation used at each site, for the individual techniques are quite different. Rather, the three projects are united in that they show that the cheap and effective tools of exploration geophysics, properly applied, can answer very important and fundamental questions about the nature of tectonics and its potential impact on human beings.
Recent studies on how earthquake waves propagate in basins [e.g. Olsen, 1995] provided strong motivation for the study in Chapter 2. In that Chapter, I present a study in which I used variations in the earth’s gravitational acceleration in and around Reno and Carson City, Nevada, to discover the thickness of sediments in the basins. The interpretation of the complex gravity field was simplified by removing the known gravity effect of earth tides, latitude and longitude, elevation, etc., and making the very simple assumption that the remaining anomaly was due to changes in rock density below the measuring sites. Because basin fill is less dense than basement rock, the areal distribution of the gravity anomaly as measured at the surface revealed the subsurface basin geometry.
Advances in computing power in the past two decades has made the study of the propagation of earthquake waves through strong velocity contrasts feasible for everyone. Although no simulations of seismic waves through the Reno and Carson City basins have been made as of yet, my basin mapping may reveal basin geometries that have a significant effect on the seismic hazard of the region. I predict that basin studies of this sort will become increasingly common as more and more urban areas feel the need for increased knowledge of potentially increased earthquake damage within heavily populated areas.
The results of Chapter 3, although achieved using the methods more commonly associated with resource exploration, have far-reaching implications for tectonics, seismic hazard, and fault mechanics. I used seismic reflection, seismic refraction, tomography, and gravity techniques to prove that the Dixie Valley fault is characterized by subsurface dips as low as 25° at the surface. This result bolsters the argument of many geologists who have argued for the existence of many low-angle normal faults that have accommodated large amounts of extension in the upper crust. The results also weaken the arguments of many fault mechanics researcher who have claimed that low-angle slip is impossible due to frictional constraints. This study shows that low-angle slip must be occasionally possible (at least in the upper 2 km) and future research should focus on mechanisms that make low-angle extension possible.
The results of Chapter 3 also have seismic hazard implications, especially if the model proposed in Wenicke [1992] is correct. In the 1992 review paper, Wernicke derived using simple relations, an equation that suggest that the recurrence interval of normal-mechanism earthquakes as a function of dip θ is proportional to tan(θ)sin(θ). This relation predicts that low-angle normal earthquakes account for less than 10% of the number of earthquakes on faults dipping 30° to 60°. However, given the large down-dip dimension of low-angle faults (i.e. 15 km, the width of a standard seismogenic zone, divided by the sine of the dip), and the inherit strength of the fault geometry due to increased normal stress, earthquakes on low-angle faults can pose a serious seismic hazard [Wernicke, 1992].. This coupled with the fact that low-angle faults have more surface area near the surface of the earth, could mean that many urban areas have under-recognized faults posing an increased hazard.
Chapter 4 shows again how small-scale
investigations can have large tectonic and seismic hazard implications. In
contrast to Chapter 2, where I investigated the possibility that conditions in
the basin amplify seismic waves, I now investigate the possibility that local
de-amplification of seismic waves affects the overall hazard. The scale of the
two investigations bears mentioning. In Chapter 2, large-scale structures, such
as the basin-bedrock interface, several hundred meters distant, serve to focus
waves, increasing the hazard. In Chapter 4 I discuss the seismic hazard
implications of geology less than 100 m from the measuring station. In this
case, the hazard posed by large strike-slip faults was studied. Motivated by
the existence of precarious rocks (which Los Angeles Times Magazine once termed
“The Seismometers of the Gods”) within sight of the San Andreas fault, I used
seismic refraction-microtremor (ReMi) and tomographic techniques to determine
if local effects were responsible for the apparent decreased ground motions at
the sites. Ground motion generated by an earthquake is a convolution of source,
path, and site effects. By using the above techniques and small magnitude
earthquake spectral ratios, I conclude that site and path effects at the
precarious rock sites are minimal, and that current regressions appear to
overestimate the hazard inherit to large strike-slip events. To the millions of
people living near the San Andreas, North Anatolian, and other large
strike-slip faults this issue is of paramount importance.
Chapter 2: Depth to bedrock using gravimetry in the Reno and
Carson City, Nevada, area basins
Robert
E. Abbott and John N. Louie
The University of Nevada, Reno, Seismological
Laboratory and Dept. of Geological Sciences, Reno, Nevada 89557
Sedimentary
basins can trap earthquake surface waves and amplify the magnitude and lengthen
the duration of seismic shaking at the surface. Poor existing gravity and
well-data coverage of the basins below the rapidly growing Reno and Carson City
urban areas of western Nevada prompted us to collect 200 new gravity measurements.
By classifying all new and existing gravity locations as on seismic bedrock or
in a basin, we separate the basins' gravity signature from variable background
bedrock gravity fields. We find an unexpected 1.2 km maximum depth trough below
the western side of Reno; basin enhancement of the seismic shaking hazard would
be greatest in this area. Depths throughout most of the rest of the Truckee
Meadows basin below Reno are less than 0.5 km. The Eagle Valley basin below
Carson City has a 0.53 km maximum depth. Basin depth estimates in Reno are
consistent with depths-to-bedrock in the few available records of geothermal
wells, and one wildcat oil well. Depths in Carson City are consistent with
depths from existing seismic reflection soundings. The well and seismic
correlations allow us to refine our assumed density contrasts. The basin to
bedrock density contrast in Reno and Carson City may be as low as -0.33 g/cm3.
The log of the oil well, on the deepest Reno sub-basin, indicates that
Quaternary deposits are not unusually thick there and suggests that the
sub-basin formed entirely before the mid-Pliocene. Thickness of Quaternary
fill, also of importance for determining seismic hazard, below Reno and Carson
City may only rarely exceed 200 m.
Alluvial
basins can amplify the magnitude and lengthen the duration of seismic waves. In
Mexico City, for example, "basin site effects" due to waves trapped
in the low-velocity basin are cited as a primary reason for disastrously high
ground motion in the great 1985 Michoacan earthquake [Campillo et al., 1989;
Sanchez-Sesma et al., 1989]. Kawase and Aki [1989] showed that both basin shape
(i.e. depth-to-bedrock) and velocity contrasts within the alluvium were
essential parameters needed to model ground motion in the Mexico City basin. Frankel
and Vidale [1992] used water well depth-to-bedrock data to create a 3-D
simulation of seismic waves in the Santa Clara, California, basin. Efforts to
predict ground motions in basins in the Salt Lake City, Utah, and Los Angeles,
California, areas have required knowledge of sediment thickness as well as
bedrock topography [Olsen et al., 1995; Frankel, 1993]. The highest ground
motion amplification in Olsen's simulations of Salt Lake City occurred near the
edges of the deepest portions of the basin (rather than directly over the
deepest portion), where the depth gradient was steepest. In these areas, Olsen
states that particle motion can be up to 2.9 times higher than in bedrock
stations. In addition, the duration of the seismic signal is up to 40 times
longer [Olsen et al., 1995]. Although many additional factors must be
considered to estimate seismic hazard, to model wave propagation in western
Nevada population centers, accurate basin models are essential. It is with this
in mind that we undertook a detailed gravity survey of the urban centers of
Reno and Carson City, Nevada.
The density contrast between bedrock and unconsolidated or poorly consolidated sedimentary rocks allows the study of bedrock structures underlying sedimentary basins. With good gravity data coverage, only changes in rock density affect the shape of any gravity anomaly. Basin shape and depth can be inferred from the spatial distribution of the anomaly. Examples of this general technique can be found in West [1992, p. 200-209]. Schaefer [1983] modeled the Dixie Valley, Nevada, basin using a similar technique; many researchers have used this method for hydrologic, geothermal, mineral, and oil exploration. Jachens and Moring [1990] mapped Cenozoic thickness across Nevada with this principle.
Our study area is situated along the western edge of the Basin and Range geologic province, in the western United States. The cities of Reno and Carson City lie within the fault-bounded basins of the Truckee Meadows and Eagle Valley, respectively (Figure 2.1). The two basins are bordered on the west by the Carson Range of the Sierra Nevada Mountains and on the east by parts of the Pah Rah and Virginia Ranges, and the Pine Nut Mountains. The Carson Range is predominantly Mesozoic granite with older metamorphic rocks, and the other ranges generally consist of Tertiary volcanic rocks. The basin fill consists of Quaternary and Tertiary alluvial and lacustrine deposits, and outwash from the most recent glacial epochs [Bell et al., 1989]. A significant portion of the Truckee Meadows is underlain by low density diatomaceous sediments.
The subsurface geology in this region is poorly understood. Existing gravity coverage, as compiled from the 1994 NOAA Gravity CD-ROM [Hittelman et al., 1994], is too sparse to adequately resolve basin structure. Thompson and Sandberg [1958] conducted a gravity survey of the Virginia City, Nevada, and Mt. Rose, Nevada, quadrangles in 1952. However the average station spacing of 1 station per 2 mi2 (5 km2) was inadequate to characterize the basins. Very few of the existing gravity measurements
were made over the basins. As a result, basin details are not revealed in the Bouguer anomaly gravity maps of Plouff [1992] and Saltus and Jachens [1995].
Hess [1996], Garside and Schilling [1979], and associated, recently updated databases of geothermal and oil wells present some information on 56 boreholes that are over 150 m deep in the Reno basin. Table 2.1 summarizes data from a few of these that we use to constrain our basin models. We selected the 26 wells for Table 2.1 because we could find some minimal location and total depth information for them. In those cases where a group of wells are on the same property or in very close proximity to one another, we only list the deepest well and/or the well with the best logs of the group. All but a few are clumped in the 5 km2 "Moana Hot Springs" area on the southwest side of the basin. Four deep geothermal wells there logged bedrock at more than 300 m depth. The bedrock there is the Tertiary Kate Peak formation andesite (part of “Consolidated Basement Rocks”, Figure 2.2.) Garside and Schilling [1979] report a single wildcat oil well in the Reno-Truckee Meadows basin. The well was drilled in 1908 on the western side of the basin, with a log interpreted by Anderson [1910]. The 1890 ft (576 m) hole encountered only sedimentary rocks, giving a minimum basin thickness in that area. The other wells outside these limited areas are domestic water wells with only total depths known. These provide some corroborating minimum basin depth constraints. Seismic studies of Reno basin velocities are underway, but will not be adequate for describing bedrock geometry.
Arteaga [1986] mapped depth-to-bedrock in Eagle Valley using a combination of seismic reflection, seismic refraction, and gravity techniques. Their gravity results provide independent corroboration of our technique, and their seismic depth soundings allow for more accurate density calibration.
The seismic hazard of western Nevada is high, with many faults capable of producing magnitude 7 and greater earthquakes [dePolo et al., 1996]. USGS seismic hazard maps [Frankel et al., 1996] do not include any evaluation of basin amplification effects. They show that for both Reno and Carson City there is a 2% probability of ground motions exceeding 0.6 g in the next 50 years.
As of 1995, approximately 400,000 people live in the Reno, Carson City and surrounding areas. A hypothetical magnitude 7+ earthquake would represent a tremendous potential for loss of life and property. Identifying those areas susceptible to greatest ground movement would be of use to emergency planning personnel.
We made approximately 200 gravity measurements with a LaCoste and Romberg model G gravity meter. The measurements generally follow north-south or east-west roads in the urban Reno and Carson City, Nevada, areas (Figure 2.2). In Reno, vertical control was provided by a geodetic-quality GPS. In Carson City, an electronic distance-measuring theodolite was used for vertical control. The surveys were tied to international gravity (IGSN 1971) at a gravity base station in Reno (ACIC 0454-1). Local base stations were re-occupied on a regular basis to monitor tidal variations to gravity as well as instrument-related drift. Terrain corrections (using 2.67 g/cm3) were estimated by eye in the field from 1 meter to 54 meters horizontally (Hammer zones B-C) and computed by algorithm from 54 m to 167 km, using 90-meter digital elevation models. The data were reduced to complete Bouguer anomaly using a reduction density of 2.67 g/cm3. The curvature correction to the Bouguer slab equation was applied when calculating terrain corrections beyond 18 km.
Existing gravity coverage [Hittelman et al., 1994] was merged into the dataset to complete our coverage, since we took fewer measurements outside the basins. The terrain corrections from 54 m to 167 km were re-computed and re-applied to the existing data, along with the curvature correction. The total coverage included 600 points.
To differentiate gravity effects due to small-scale basins from broader, regional anomalies, a "bedrock gravity" value was removed from the data set. Following Jachens and Moring [1990], all gravity stations are classified as "bedrock" or "basin" stations using geologic maps [Bonham and Rogers, 1983; Trexler, 1977; Bell and Garside, 1987; Bonham and Bingler, 1973]. We considered measurements on or near Tertiary Kate Peak formation andesitic rocks (part of “Consolidated Basement Rocks”, Figure 2.2) to be on bedrock for our purpose of differentiating low-seismic-velocity sedimentary fill from relatively high-velocity seismic bedrock. Similarly, points on or near Tertiary Hunter Creek formation sandstones (Figure 2.2) were considered to be basin fill. Density measurements by Thompson and Sandberg [1958] indicate that the density of the Kate Peak formation averages around 2.61 g/cm3. A single density measurement on the Hunter Creek formation (Truckee formation of Thompson and Sandberg [1958]) indicates a density of 1.76 g/cm3, although there is evidence from well-log data [Anderson, 1910] and this study to indicate that the density of this formation varies widely.
Bedrock gravity values were computed by kriging the complete Bouguer anomaly of those gravity stations known to be in areas where basin fill is minimal, or non-existent. The bedrock gravity is subtracted from the complete Bouguer anomaly of all measurement points. By removing the perturbations to gravity caused by bedrock density contrasts, basin structure is emphasized and the gravity effect of deep density variations below the surrounding mountain ranges is attenuated. The gravity effect of a basin extends beyond the basin boundaries, however, and these are subtracted as part of our “bedrock gravity” estimate. Thus, basin depths subsequently estimated will be minima.
Unlike
Jachens and Moring [1990], initial basin depth estimates were accomplished
using the infinite slab approximation. We simply scale the basin gravity
anomaly value at each measurement point by a factor that assumes the anomaly
results from one or more slabs of constant density contrast and infinite
lateral extent, to find the total sediment thickness. This estimate is similar
to reversing the Bouguer slab calculation, and produces a smoothed basin-depth
profile, with the deepest depths being underestimated.
We
initially used the sediment compaction model given in Table 2.2 to find the
alluvium-basement density contrast. The sediment compaction model is the same
used by Blakely et al. [1998] and Jachens and Moring [1990], and represents a
regional average for basins within the Basin and Range province. Depth was
calculated by applying the slab approximation for the shallowest (-0.65 g/cm3,
200 m) slab. If the gravity anomaly caused by this slab is less than the
observed anomaly, deeper layers were taken into account. It should be noted
that the infinite slab approximation works best when the slab thickness is much
less than the lateral extent of the basin. Errors in depth calculations can
occur when nearing the basin edge, where this approximation fails.
We also forward modeled 2.5-dimensional selected linear transects in Reno and Carson City using the GM-SYS software package, developed by Northwest Geophysical
Associates. Sediment “blocks” were modeled as extending 3.5 km north and south of the Truckee River east-west transect in Reno . In Carson City, sediment blocks extend 2 km north and 6 km south of the 5th Street east-west transect . Care was taken that the transects followed the trend of the measuring stations as closely as possible. Information from well data (in Reno) and seismic data (in Carson City) were used to constrain parameters in basin modeling. In Carson City, due to the complete lack of local density and lithology information, we made use of average regional density contrasts (Table 2.2).
The repeat error of LaCoste and Romberg gravity measurements is estimated to be 0.03 mGal. This is higher than would be expected if the measurements were taken in a quiet environment under controlled conditions. However, most measurements were taken along busy urban streets where traffic and other urban vibrations caused measurement errors. Base stations were carefully chosen to be in quiet, controlled environments. For those measurements, a repeat error of 0.01 mGal is estimated. GPS and EDM theodolite locations, accurate to plus or minus one meter, allow us to neglect latitude correction errors. Vertical position is accurate to within 0.3 meters, as confirmed by GPS or EDM theodolite re-occupations of sites. Inner-ring terrain corrections, estimated by eye, rarely approached 0.1 mGal and were 0.01 mGal on average. Still, in areas of high relief, a 20% error in estimating inner ring terrain effects is possible. In these rare instances, an error of 0.02 mGal could have been introduced. All considered, an error in observed gravity of plus or minus 0.08 mGal is possible. Measurements of repeated points from different surveys in the existing data [Hittelman et al., 1994] exhibit a maximum error of plus or minus 0.5 mGal. This is the limiting factor in the dataset. Given the magnitude of the anomaly in Reno (15-20 mGal) and the coarse contour interval, we view this as an acceptable amount of error and that the benefits of their inclusion outweigh the problems caused by decrease in accuracy. The depth error in the infinite slab approximation caused by a 0.5 mGal error is 36 m using a –0.33 g/cm3 bedrock-alluvium density contrast. In the forward models, our tolerance fit levels mean that the 0.5 mGal error between measurement campaigns is essentially invisible.
Our density approximations are the principal source of error in our analysis. The well logs available in Reno lack density measurements or analyses. This lack of density data leads to highly speculative density values. With upper and lower limits for basin-bedrock density contrast set at 0.65 g/cm3 and 0.30 g/cm3, a 50% depth error is conceivable.
Several products result from our data analysis: (1) Complete Bouguer anomaly maps derived from all stations; (2) Complete Bouguer anomaly maps derived from bedrock stations; (3) Basin anomaly gravity maps; (4) Basin depth maps derived from the infinite slab approximation; and (5) 2.5-D forward models of selected linear transects.
The anomaly maps of Reno (Figures 2.3, 2.4, and 2.5) show an extended, asymmetrical gravity low over the Truckee Meadows. The gravity low represents the density contrast of bedrock and sediments. The western side of the basin shows the steepest gravity gradients and the most negative anomaly. The maximum local anomaly of -16 mGal yields a basin depth of 1160 m using the infinite slab approximation (Figure 2.6) if we assume an alluvium-bedrock density contrast of –0.33 g/cm3. There is evidence that the residual gravity separation may not have completely succeeded near this sub-basin. Figure 2.4, the bedrock gravity grid, also shows a gravity low over this area. Note
that the infinite slab approximation underestimates basin depth for a given density contrast. Because we used well-log information to calibrate depth at certain areas, the density contrast required in the infinite slab approximation was underestimated to compensate. Sediment density is likely to be less than the 2.34 g/cm3 we used.
The western-most elongation of the basin represents the east-west trending Tertiary Verdi basin. This basin is underlain by the Miocene-Pliocene Hunter Creek sandstone formation. The sandstone has a lower average density than the alluvium of the Truckee Meadows. As such, the depth of the basin may be slightly shallower than indicated on the depth-to-bedrock maps (Figures 2.6 and 2.7). A sub-basin in the Steamboat Springs area is represented by another gravity low to the southwest, with -6 mGal local anomaly, corresponding to a depth of approximately 430 meters.
The east-west cross-section along the Truckee River in Reno (Figure 2.8) yields a maximum basin depth of 1000 m. This profile shows a striking structural trough in the western portion of the basin. The maximum basin depth in this model is under West McCarran Boulevard. A second trough in the eastern portion of the basin is separated from the western trough by a bedrock ridge that comes within 200 meters of the basin surface near the Reno/Tahoe International Airport.
The anomaly maps of Carson City (Figures 2.9, 2.10, and 2.11) show an elongate north-south gravity low over Eagle Valley. The anomaly closely approximates the anomaly shape of Ateaga [1986], which he mapped using a combination of gravity and seismic techniques. The magnitude of the local anomaly, -7 mGal, is much smaller than in the Truckee Meadows, suggesting a shallower basin depth. This corresponds to a 520 m depth with the infinite plate approximation (Figure 2.12, assuming a –0.33 g/cm3 density contrast). The northeast-trending contours in the northern part of the basin are poorly constrained and may be an artifact of the poor data coverage in the area. A sub-basin to the northwest is separated from the main basin by the subsurface expression of a northwest-southeast trending ridge of Triassic metavolcanic rocks. This formation outcrops at Lone Mountain in northern Carson City [Trexler, 1977].
An east-west cross-section along 5th Street in Carson City (Figure 2.13) yields a maximum basin depth of 530 m. The 5th Street transect shows Eagle Valley to be a more symmetrical basin in which the depth increases fairly smoothly to 0.53 km before returning to bedrock on either side of the basin. The density scheme used is from Table 2.2. The maximum basin depth along this transect is located 1.5 km east of US Highway 395 (Figure 2.13).
Contours indicating negative bedrock depths outside basins allow the estimation of errors in depth-to-bedrock calculations caused by shallower bedrock density contrasts. The +2 mGal contour on the western margin of Eagle Valley (Figure 2.10) would correspond to a -140 meter depth-to-bedrock with the infinite slab calculation (Figure 2.11), where negative depth would mean bedrock above actual elevation. Therefore, our estimates of bedrock gravity (Figures 2.3 and 2.9) may be in error by 2 mGal, and depth to bedrock in the basins cannot be considered more accurate than ±140 meters. The cause for this is unclear, but the poor coverage of bedrock gravity measurements, isolated bedrock density variations such as hidden intrusions, measurement errors, or isostatic effects near the Sierra Nevada Mts. are possible. Based on our Reno bedrock gravity estimate (Figure 2.4), our basin gravity difference (Figure 2.5), and basin depth (Figures
2.6 and 2.7) maps, we estimate a depth uncertainty of 250 meters for the Truckee Meadows.
To constrain absolute depths, accurate density measurements need to be obtained. Specific knowledge of how density increases with depth, especially from outside the geothermal fields, would be particularly useful. We can only use the depth to bedrock logged in a few of the wells to check our overall density assumptions. In particular, the thickness of low-density diatomaceous deposits in the Hunter Creek formation varies widely from location to location. Currently, density uncertainty is the overriding cause of depth uncertainty. Using our reasonable end-member values for density contrasts, 50% error in depth calculation is possible, if no other factors, such as seismic depth soundings, were taken into account. The error in our analysis is probably significantly less than this maximum value, however.
Arteaga’s [1986] hydrological study of Eagle Valley included some depth-to-bedrock calculations based mostly on seismic depth soundings, supplemented by gravity measurements. Absolute comparison of gravity values is impossible because the study did not publish complete Bouguer values. The only published result was the gravity residual, obtained by subtracting out the regional gradient. The shape of gravity residual obtained by Arteaga matched extremely well with our basin anomaly residual.
Arteaga’s seismic depth soundings predicted a 620 m maximum basin depth, compared to our 530 m maximum. The data quality of the seismic depth soundings where charaterized as only being “fair” in Arteaga’s [1986] study. Overall agreement in depth-to-bedrock is generally within approximately 25%. The location of the three measurements closest to the deeper sub-basins is plotted on Figure 2.12, along with the associated depths in meters.
Garside and Schilling [1979] and more recent associated databases [e.g. Hess, 1996] provide some well data that corroborate our basin depths for Reno (Figures 2.6, 2.7, and 2.8; Table 2.1). Neither of these latest databases show any boreholes of record in Carson City-Eagle Valley. The well casing from the 1908 deep oil prospect [Anderson, 1910] was located by the Nevada Bureau of Mines and Geology before housing development overtook the site. The location is thus known to within 30 m and appears near the 800 m basin thickness contour from the slab calculations on Figure 2.7 (labeled 1). Anderson [1910] interpreted all but the top few meters of the 576 m total depth drilled as partly penetrating the middle-Tertiary "Truckee formation" of sands, shales, and diatomites. The formation is equivalent to the Miocene-Pliocene "sandstone of Hunter Creek" mapped in the area by Bonham and Rogers [1983]. This deepest boring into basin deposits in Reno is only 1.5 km south of our lowest basin gravity anomaly. Garside and Schilling [1979] and the more recent records show an additional 55 wells drilled to depths greater than 150 m in the Reno basin. Table 2.1 lists the 25 of these that provide the best constraints throughout the basin. Locations of most of these wells are given by partial sections or permittee addresses, and could easily be 300 m in error. The 30 wells not in Table 2.1 either lacked any reliable depth or location information, or were not as deep or as well-logged as another well on the same property or very nearby. Of the 55 wells, only twelve are outside the immediate area of the "Moana Hot Springs" geothermal district (the concentration of well locations at the lower center of Figure 2.7). Eleven of these appear on Figure 2.7; the twelfth lies off the map to the south.
South of the 200 m depth contour, along South McCarran Blvd., Figure 2.7 shows basin thicknesses of 100 m or less, corroborated by domestic wells such as the Talsma and the Peterson (labeled 20 and 18 on the Figure), which logged Kate Peak volcanic bedrock at 49 m and 91 m, respectively. Depths increase rapidly to the north and to the northwest, in concert with the depth contours derived from gravity. The Pennington domestic and Warren Estate Geothermal Well No. 3 (labeled 19 and 24 on Figure 2.7) logged Kate Peak at 283 and 317 m, respectively. The Pennington well is located near the 400 m depth contours on the Figure. The Warren Estate #3 is located by partial sections, where the first quarter-section may be stated in error in the database; its location on Figure 2.7 may be 1 km southeast of the true location.
At the north end of the Moana Hot Springs area, the deep Kohlenberg Domestic Injection Well No. 1, Peppermill Well No. 4, and Salem Plaza Injection Well No. 1 (labeled 16,15, and 21 on Figure 2.7) log Kate Peak depths increasing from 310 m, to 344 m, and to 418 m, respectively, from east to west over a 1 km distance across US 395 Business. The Peppermill No. 4 well is the deepest hole in the region, with logs to 1008 m (n.b.: logged lateral deviation of the hole is less than 100m). These bedrock depths closely approximate the depth contours on Figure 2.7.
Every logged well in Reno shows Hunter Creek sandstones and diatomites extending through 20%-90% of the section above the Kate Peak bedrock, averaging about 80% (Table 2.1). Anderson [1910] summarized observations of similar diatomites throughout the Great Basin, and proposed that every Miocene basin in the region might host them. The diatomites consist of an open network of silica microfossils having porosities as high as 70%, and air-filled samples from the surface near the oil well (labeled 1 on Figure 2.7) will float on water (mentioned by Anderson, 1910]. These diatomites, with pores filled with water, would have a maximum density of 1.7 g/cm3. It is likely that the diatomites have less porosity at depth and/or have pore spaces filled with mineralization, thus increasing the formation’s density.
In Reno
outside the Moana Hot Springs district, the 12 deeper wells listed by Garside
and Schilling [1979] have total depths but no logs on file. Figure 2.7 locates them with white circles,
having ‘≥’ marked above their total depths. Given that these are
all domestic water sources, it seems unlikely that they would have been drilled
far into any bedrock formations, although that possibility cannot be ruled out.
Assuming that these wells provide minimum basin thicknesses from their total
depths, one of them (labeled 3, on Figure 2.7) does not match the
slab-gravity-derived depth contours. The other water wells for which there are
no logs are all shallower than or close to the infinite-slab depth contours. The
"negative depth" contours on Figure 2.7 surrounding the interchange
suggest bedrock density contrasts that are not constrained by the bedrock data,
and have artificially pushed the depths to more shallow levels. A 198 m well
drilled in 1958 (labeled 11 on Figure 2.7) in a sub-basin to the southeast, is
very close to bedrock outcrops and may have been drilled partially into bedrock.
A novel result of our work is our definition of the 16 mGal gravity low on the west side of Reno. This anomaly, about 7 km in diameter, had not been sampled by the previously very sparse gravity coverage of the basin. Within the limits of our data coverage, Figures 2.5, 2.6, and 2.7 show that this low defines a north-south trending trough about 5 km long and 3 km wide, and up to 1.2 km deep, that we call the West McCarran sub-basin. It should be noted that the extent of the sub-basin is poorly constrained to the southwest. The sub-basin is twice as deep as any other sub-basin below Reno or Carson City, and identifies the location of what could be the largest basin effects on earthquake ground motion in the western Nevada urban areas.
The 576 m well drilled in 1908 shows that this deep sub-basin was formed entirely in Miocene and Pliocene time. The well sits nearly atop the deepest part of the basin (labeled 1 in Figure 2.7), only 1.5 km south of the line of depth section modeled in Figure 2.8. Anderson [1910] mapped the fully exposed Truckee formation (Hunter Creek) section, inspected the well during drilling, and interpreted the driller's log. He proposed that the entire borehole had penetrated just the middle diatomite-dominated member of the Hunter Creek sandstones, perhaps with some sampling of the lower sandy member. Given the exposed 200 m thickness of each of these members in sections compiled throughout the region, and the lesser thicknesses in the Moana Hot Springs well logs [Garside and Schilling, 1979, and recent associated databases], the Hunter Creek sandstones appear to be thickened in the deep sub-basin by a factor of two or three. This thickening suggests the sub-basin was actively subsiding during deposition of the diatomite member near the Miocene-Pliocene boundary (age taken from Bonham and Rogers, 1983], and probably initiated during deposition of Hunter Creek basal members, or earlier.
Since the entire 1.2 km-deep sub-basin is filled by early Pliocene and older sediments, all of the related, basin-forming vertical deformation must have occurred by the early Pliocene. Thus the existence of the West McCarran sub-basin requires no Quaternary deformation. The 274 m thickness of Quaternary deposits logged by the driller in the Peppermill Well No. 4 (Table 2.1, number 15) may be overstated. Re-interpretation of this log, and comparison to nearby logs by the authors suggest only 168 m of Quaternary fill above the Hunter Creek sands. This depth thus represents the maximum observed Quaternary vertical deformation in the Reno basin. All other logged wells in Table 2.1 show less, between 30 and 165 m. While the total thickness of low-velocity Miocene through Quaternary sediment varies greatly among the Reno sub-basins, the maximum thicknesses of Quaternary deposits in these basins may well be less than 200 m.
Our less than 30% error in depth calculation (based on negative depth contours) should have little effect on seismic modeling, except possibly in one key area. Currently, modeling of seismic waves in basins is rarely done for frequencies greater than 1 Hz. Any change in depth equal to or less than one-quarter seismic wavelength may be undetectable. Work in progress, not presented here, by the University of Nevada, Reno Seismological Laboratory, Kyoto University, and the Shimizu Corporation of Japan is employing the microtremor analysis of Horike [1985]. Shear wave velocities for depths below 100 m at a test site near the Reno/Tahoe Airport (Figure 2.7), are on the order of 2 km/s. At 1 Hz, this would correspond to a 2 km seismic wavelength. Therefore 500 meter resolution is required for seismic hazard modeling. Only in the deepest section of the Truckee Meadows would a 30% maximum error in depth even approach this limit. In the shallower sections of both basins, even 30% depth error will be much less than 500 meters, and will have a smaller effect on a smaller seismic hazard. Therefore, the depth error in the shallow sections may be insignificant. It is the deepest, most poorly-characterized sub-basin that has the most seismic hazard potential, and this is the area with the most error.
Depth gradient maps, not presented here, suggest in the manner of Olsen et al. [1995] that the areas that could most experience basin effects might be near the north and south corners of West McCarran Blvd. (Figure 2.7). In these areas over the western sub-basin, the combination of a deep basin and high gradients may produce the ground motions most amplified by surface waves trapped in the basin. High gradients also exist at the eastern edge of the Truckee Meadows. However, the basin is not as deep in this area, and therefore ground motion amplification may be less. Eagle Valley, with a very muted basin structure as compared to the Truckee Meadows, should show less ground amplification due to basin site effects altogether. The difference in site effects might be of significant amplitude; and seismic hazard maps may have to be re-evaluated for these two areas. The seismic hazard for Reno may increase with respect to the seismic hazard of Carson City.
The
authors would like to thank C. Mann, J. Ollerton, and M. Herrick for their
tireless and conscientious fieldwork. Thanks to Kennecott Exploration Company
for generously loaning their geodetic-level GPS, and to J. McKinney and C. Lide
of Zonge Geoscience Inc. for their help in applying terrain corrections to the
data. We also thank P. Cashman and J. Trexler for their thought provoking insights.
R. Blakely and P. Milligan provided careful and thorough reviews; J.G.
Anderson, G. Biasi, and A. Cadena helped refine the manuscript. R. Abbott was
supported for this work by a Nevada Seismological Laboratory Fellowship. Research
supported by the U.S. Geological Survey (USGS), Department of the Interior,
under USGS award number 1434-HQ-97-GR-03041. The views and conclusions
contained in this document are those of the authors and should not be
interpreted as necessarily representing the official policies, either express
or implied, of the U.S. Government.
Chapter 3: Geophysical confirmation of
low-angle normal slip on the historically active Dixie Valley fault, Nevada
Robert E. Abbott and
John N. Louie
Seismological Laboratory and Department of Geological Sciences, University of Nevada, Reno, Nevada
S. John Caskey
Department of Geosciences, San Francisco State University, San Francisco, California
Satish Pullammanappallil
Optim L.L.C., Seismological Laboratory, University of Nevada, Reno, Nevada
The
December 16, 1954, Dixie Valley earthquake (MS=6.8) followed the
nearby Fairview Peak earthquake (MS=7.2) by 4 min, 20 s. Waveforms
from the Fairview Peak event contaminate those from the Dixie Valley event,
making accurate fault plane solutions impossible. A recent geologic study of
surface rupture characteristics in southern Dixie Valley suggests that the
Dixie Valley fault is low angle (<30°) along a significant
portion of the 1954 rupture. To extend these observations into the subsurface,
we conducted a seismic reflection and gravity experiment. Our results show that
a portion of the Dixie Valley ruptures occurred along a fault dipping 25° to
30°. As such, the Dixie Valley event may represent the first large, low-angle
normal earthquake on land recorded historically. Our high-resolution seismic reflection
profile images the rupture plane from 5 to 50 m depth. Medium-resolution reflections,
as well as refraction velocities, show a smoothly dipping fault plane from 50
to 500 m depth. Stratigraphic truncations and rollovers in the hanging wall
show a slightly listric fault to 2 km depth. Gravity profiles conservatively
constrain maximum basin depth and define overall geometry. Extension along the
low-angle section may have occurred in two phases during the Cenozoic. Current
fault motion postdates a 13 to 15 Ma basalt, imaged in the hanging wall, and
inherits from a fault formed during an earlier extensional pulse, concentrated
at 24.2 to 24.4 Ma. The earlier extension suggests extraordinary slip rates as
high as 18 mm/yr, resulting in the formation of the low-angle fault break.
Sections of the Dixie Valley fault where there is no evidence for current
low-angle slip correlate well with areas where no pre-15 Ma slip has been
documented.
Despite
growing geological and geophysical evidence arguing for the existence of
low-angle normal faults that have accommodated large amounts of extension, the
paradox of the near-complete absence of low-angle normal-mechanism earthquakes
in the seismic record remains. Slip on low-angle normal faults is not predicted
in Andersonian theory [Anderson,
1942] and studies of earthquake focal mechanisms, both
regional [Doser and Smith, 1989] and global [Jackson,
1987], show an extreme scarcity of large (M>5.5) normal
fault mechanisms with dip <30°. However, several researchers [e.g., Abers et al., 1997; Hatzfeld et
al., 2000; Johnson and Loy, 1992] have presented compelling evidence that substantial
extension has occurred along low-angle normal faults in the brittle upper
crust.
Theories
to resolve the seismicity paradox fall into two categories: those that do not
require brittle slip at low angles (e.g., “rolling hinge” models [Wernicke and Axen, 1988] and flexural rotation [Buck, 1988]), and those that argue either for long earthquake
recurrence intervals [Wernicke,
1995], or a current rarity of active low-angle faults [Burchfiel et al., 1992].
Compelling
evidence that brittle slip is possible on at least one low-angle normal fault
would have important ramifications for both fault mechanics theory and seismic
hazard calculations. Here, we present the results of a seismic reflection and
gravity experiment to test whether part of the December 16, 1954, Dixie Valley
earthquake (MS=6.8) produced slip on a low-angle normal fault.
Dixie
Valley, Nevada, lies in the north central portion of the Basin and Range
province (Figure 3.1). The Basin and Range is a region that has experienced a
large amount of intraplate extension in the Cenozoic. Much of the extension is
accomplished by high-angle (50°-70°) normal faulting, with several large
seismic events recorded historically (e.g., 1915 Pleasant Valley, 1954 Fairview
Peak-Dixie Valley, 1983 Borah Peak). The faulting has created predominately
north-south trending mountain ranges and sedimentary basins. Dixie Valley is
one such basin, bounded by the Stillwater Range on the west and the Clan Alpine
Range on the east (Figure 3.1). The Dixie Valley fault, site of the 1954 Dixie
Valley earthquake, is the east dipping range-bounding normal fault along the
eastern front of the Stillwater Range.
The 1954
fault trace lies along the southern portion of the Stillwater Range and
separates Mesozoic and Tertiary footwall rocks from late Tertiary and
Quaternary basin fill. Miocene and Oligocene volcanic rocks and granitic
plutons related to the Stillwater caldera complex [John, 1995] and Mesozoic metasedimentary rocks represent the
“geophysical basement.” The basin fill at the surface consists of alluvial fan
and lacustrine deposits [Wilden
and Speed, 1974].
The
December 16, 1954, Dixie Valley earthquake was the last of a series of large
earthquakes that took place within a period of 6 months in central Nevada. The
preceding events were the Rainbow Mountain sequence (MS=6.6 and 6.4
on July 6, 1954, MS=6.8 on August 24, 1954) and the Fairview Peak
earthquake (MS=7.2 on December 16, 1954). The Fairview Peak event
preceded the MS=6.8 Dixie Valley earthquake by 4 min and 20 s. Focal
mechanisms for the Fairview Peak and Rainbow Mountain events indicate NNW
striking normal-oblique faults with dips ranging from 60° to 78°.
Fault
plane solutions for the Dixie Valley event are poorly constrained because the
arrivals are obscured by waveforms from the Fairview Peak event. Doser [1986]
used waveform modeling to determine fault geometry with a strike of N10°W and a
dip of 60°E; however, owing to contamination of the Dixie Valley waveforms
“large changes in strike and dip did not significantly change the waveform
shape” [Doser, 1986, p.12,583]. Similarly, Okaya and Thompson’s [1985] solution of
N11°W, 62°E cannot be relied upon. Okaya and Thompson [1985, p.116] noted that
“Of the four focal parameters (depth, dip, strike, and slip), only changes in
depth are significant; changes in fault plane strike, dip, or slip have
negligible effect.” The Dixie Valley fault plane solution (N8°E, 49°E) of Hodgkinson
et al. [1996] using leveling and triangulation benchmarks suffers from a
paucity of data (very few prerupture survey bench marks) in the rupture region,
such that the triangulation network is unable to document slip along most of
the rupture.
Caskey et
al. [1996] conducted the most recent and detailed study of the surface-faulting
characteristics of the Fairview Peak and Dixie Valley earthquakes. They noted
substantial geologic evidence for low-angle dip for the Dixie Valley fault
along an ~20-km-long portion of the rupture zone. Geologic evidence for a low
dip angle lies between The Bend in the north, to just north of Coyote Canyon in
the south (Figure 3.1). Geological evidence for low-angle dip at the surface
from Caskey et al. [1996] include (1) three-point fault plane reconstructions,
(2) shallow subsurface modeling of the rupture-trace graben, (3) fault-parallel
fracture sets in the footwall, and, (4) geometry of the Stillwater range front.
Dixie
Valley has been the subject of numerous geophysical investigations. The studies
primarily focused on the northern part of the valley, around 40 km north of our
study area. Okaya and Thompson [1985] combine seismic reflection and gravity
data to model northern Dixie Valley as a half graben with one major normal
fault dipping 50°E to the northwest (the Dixie Valley fault), and three
smaller, west dipping normal faults to the southeast. This model is
inconsistent with Smith’s [1967] aeromagnetic study north of The Bend
(Figure 3.1). Smith maps pre-Tertiary basement under Dixie Valley as a graben
within a graben.
Schaefer [1983]
collected widespread gravity data throughout Dixie Valley. A portion of Schaefer’s
[1983] Bouguer anomaly map is reproduced in Figure 3.2. As can be seen in
Figure 3.2, two 4-5 mGal local gravity lows are near the latitudes of Coyote
Canyon and Wood Canyon, consistent with a more moderately dipping range front
fault (or uplifted bedrock) between these two latitudes. It should be noted
that the shape of these anomalies is very poorly constrained and the gravity
lows may have other explanations unrelated to a change in range front fault
dip. An east-west linear transect of 24 gravity points near the latitude of
Little Box Canyon (Figure 3.3) is consistent with a low-angle Dixie Valley
fault, assuming reasonable bedrock-alluvium density contrast.
Meister
[1967] and Herring [1967] conducted seismic refraction experiments near the
latitude of IXL Canyon (Figure 3.3). Meister [1967] interpreted southern Dixie
Valley as a composite graben, based on several short refraction lines parallel
to the rangefront, and an east-west cross-valley profile. The east dipping
Dixie Valley fault was interpreted to be a combination of high-angle normal
faults and flat terraces, resulting in a “staircase-like” fault geometry. These
data allow for an alternate interpretation, however. In our evaluation of
Meister’s [1967] data a single low-angle dip normal fault can replace the
previous stair-step structure. Herring’s [1967] experiment assumed, a priori,
high-angle dip and the assumed geometry was used to test a
“sideswipe-refraction” technique.
Figure
3.3 shows the location of our geophysical transects. The Cattle Road profile
consists of two seismic reflection profiles and a gravity profile, while the
Scarp profile consists solely of gravity coverage. The trend of the Cattle Road
profile is within 10°
of the gradient of the gravity field near the range front. If the gravity
gradient reflects the direction of true dip of the Dixie Valley fault, then apparent
dips measured from the geophysical profiles will be underestimated by no more
than 0.5°.
The medium-resolution Cattle Road profile extended 3.6 km into the basin and
utilized 8-Hz geophones. It was composed of four stationary setups of 48
receivers with 15.2-m spacing. The 132 source points were 2-7 kg of high
explosive buried 2 m below the surface. Off-end and longer offset shots were
recorded to increase fold and improve deep velocity information. Maximum
source-receiver offset for the medium-resolution line was 2.8 km, and maximum
fold was 24.
The
high-resolution Cattle Road profile overlapped the medium-resolution profile
close to the 1954 rupture. It was conducted within 150 m of the range front
scarp using 100-Hz geophone groups with 2-m spacing. Sixty-seven sledge-hammer
source points were rolled through the array. Six inline geophones per group
were used to reduce noise from ground roll. Maximum source-receiver offset for
the high-resolution line was 124 m. Maximum receiver fold was 32.
Gravity
data along Cattle Road were acquired with a LaCoste and Romberg Model G gravity
meter. Gravity coverage started at the scarp and extended eastward 12.5 km into
the basin at a 250-m average station spacing. Vertical control was supplied by
a geodetic quality GPS.
Gravity
data were also collected parallel to the rangefront scarp (Scarp gravity,
Figure 3.3). The Scarp profile was located along a line where depth-to-bedrock
was assumed to be approximately constant, and therefore any gravity variations
would be largely due to density variations within the geophysical basement. In
this way, errors in depth to bedrock calculations from our 2.5-D forward model
along the Cattle Road profile can be estimated.
3.2.2.1 Conventional processing and
poststack migration
Conventional
seismic data-processing techniques were used to remove noisy traces, mute
direct waves, and attenuate other unwanted arrivals. The medium-resolution line
was band-pass-filtered (6-8 Hz, 100-120 Hz trapezoidal filter), and then
filtered with a polygonal, 48 trace, 500 ms f-k domain filter. The f-k filter
was designed to eliminate 400 m/s Rayleigh waves. After automatic gain control,
constant velocity stacks at 200 m/s intervals were used to pick stacking
velocities. Common depth points were binned at 15.2-m intervals, with no
amplitude variation with fold. The subsequent common depth point stack was then
Stolt-migrated at a constant 2000 m/s velocity.
The
high-resolution seismic line was band-pass-filtered (8-12 Hz, 200-202 Hz) and
f-k domain filtered (48 trace, 125 ms). Stacking velocities were chosen using
the same method as with the medium-resolution line. Common depth points were
binned at 2-m intervals, and the resulting CDP stack was Stolt-migrated using
the root-mean-square stacking velocities.
Owing to
complex geometry, strong lateral velocity variation, and steep dips (for
seismic imaging) in the subsurface at Dixie Valley, we chose to compute
prestack depth migrations. For input into our prestack migration, we obtained a
detailed velocity image of the subsurface by performing a nonlinear
optimization on first arrivals picked off raw shot gathers. The optimization
technique employs a generalized simulated annealing algorithm [Pullammanappallil
and Louie, 1994] to invert first arrivals for subsurface velocity structure. We
used a commercial package, SeisOpt @2D™ (copyright Optim LLC, 1998-1999), that
implements this method.
The
simulated annealing algorithm is a Monte Carlo based estimation process that
has the property of being independent of the starting model and has the ability
to find the global minimum (i.e., solution) for a highly nonlinear problem.
These characteristics make the algorithm a very effective tool for velocity
estimation. Travel time inversion is a highly nonlinear problem because any
perturbation in the velocities alters the path of the ray propagation, changing
the travel times recorded at the surface geophones. This nonlinearity makes
linear methods dependent on the starting model; that is, the accuracy of the
final velocity model is dependent on a good initial guess. The method employed
by SeisOpt @2D “tests” several thousand models before arriving at the optimized
velocity model. The only inputs required by the algorithm are the first arrival
picks and survey geometry (source and receiver coordinates). In addition to the
final velocity model, SeisOpt @2D outputs a ray coverage or “hit count” plot
that shows what parts of the model were sampled by the seismic array. The
algorithm outputs only the velocities in the subsurface that have been sampled
by the rays.
A total
of 6117 first-arrival picks from 134 shot gathers were used for the
optimization. SeisOpt @2D can handle only two-dimensional array geometry.
Hence, in order to overcome a bend in the medium-resolution seismic line, we
project the source locations to a straight line while maintaining the true
offsets of the source-receiver pairs. As a result of this projection, the
optimized velocities might show some lateral smearing in the vicinity of the
bend in the profile. The resulting velocity model was used in a prestack
migration algorithm to image the seismic reflectors directly.
In order
to use the optimized velocities to perform a prestack Kirchhoff migration, we
first extended the velocities down to a depth of 2.0 km. Like Pullammanappallil
and Louie [1994] and Chavez-Perez et al. [1998] we extended the optimized
velocity models for migration by finding the maximum constrained velocity value
in each column of the velocity model and substituted that value into the column
of the model everywhere below the depth of the maximum velocity. We then
performed a severe lateral smoothing below the depth of first arrival
constraint.
The
resulting velocity model was used in a prestack migration algorithm to image
the seismic reflectors directly. The prestack Kirchhoff summation algorithm was
originally used to image the San Andreas fault zone by Louie et al. [1988] and
modified by Louie and Qin [1991] to account for reflection ray paths that may
bend significantly through strong lateral velocity variations. In Dixie Valley
we did not attempt to image near-vertical structure as previous work did, but
we needed to account for strong ray bending through the velocity contrasts at
the edge of the basin and through any lateral variability in the Tertiary
basalts within the basin section. So we added to the algorithm a dip-dependent
obliquity factor.
Another
addition is the migration-operator antialiasing criterion of Lumley et al.
[1994], which leads to high-cut filtering of the seismic traces. Completely
preventing the spatial aliasing of the migration operator leads to
discontinuous coverage of depth points for our medium-resolution survey and
detracts from the lateral continuity of reflections. Thus, for the aliasing
calculation we used a receiver spacing that is half of the actual 15.2-m
spacing. This apparent spacing yields a mild operator antialiasing effect that
only removes the worst aliased frequencies of the most steeply-dipping
structure, while retaining the lateral continuity of the unaliased
near-horizontal structure.
We
migrated the f-k filtered data also used for stacking, with additional 8-100 Hz
band-pass filtering, locations projected to a straight line, and AGC with a
0.5-s window for amplitude balancing. Along with migration of the data we estimated
noise data in the manner of Harlan et al. [1984] by resampling to destroy
prestack trace-to-trace coherency. Harlan et al. [1984] provided a Bayesian
prestack coherency measure computed through statistical comparison of the data
migration against the noise migration. We screened the data migration through
the coherency image to yield an enhanced structural image. The enhancement
emphasizes those structures that produce the most coherent reflections in the
prestack shot gathers.
3.2.2.4 Gravity data processing
We reduced
gravity data to simple Bouguer anomaly using standard techniques. Terrain
corrections out to 54 m (Hammer rings A-C) were estimated by eye in the field.
We used a density value of 2.67 g/cm3 for both the Bouguer slab
correction and for terrain correction. Elevations are accurate to within 5 cm,
as seen in site reoccupations. Analysis of loop closures indicates a maximum
drift of 0.16 mGal, forming the limiting error of these data. This amount of
error is acceptable, given the magnitude of the basin anomaly (~35 mGal).
Figure
3.4 is a no vertical exaggeration common-depth-point stack of our
high-resolution line near the 1954 scarp. The Dixie Valley fault is the very
prominent reflector dipping to the east from 40 to 160 ms, at 28°. This is
interpreted as the bedrock-alluvium contact. The fault surface is ~6 m below
the surface at the range front scarp. This depth to fault is nearly identical
to that predicted by the balanced geological cross section of Caskey et al.
[1996, Figure 12d]. The balanced section is reproduced at the location of the
graben in Figure 3.4. There is no evidence for staircase-like fault geometry at
this scale. Surface-parallel reflections are seen early in the section (above
60 ms) and represent layering in the alluvial fan above the fault. The
coherency of these reflections is disturbed in the vicinity of the 1954 scarp
(at 25 m, 30 ms), possibly in relation to the formation of the rupture graben.
Fault
geometry can be inferred from arrivals seen in several raw shot gathers. The
headwave arriving at the same time at all stations in the gather of Figure 3.5
limits the range of possible fault dips. If we use our optimization velocity
model (Figure 3.6) as a guide, maximum and minimum bedrock-to-alluvium velocity
ratios can be placed at 2.1:1 and 1.4:1. After adjusting for elevation changes
along the array, fault dips of 21° to 39° could result in the vertically
propagating headwave seen in Figure 3.5. Using a more reasonable velocity
estimate than the minimum and maximum results in a fault dip of 29° to 30°.
Using the same reasonable velocity estimation, but assuming fault dips of 21°
and 39°, results in the synthetic arrivals represented by dashed lines in
Figure 3.5.
These results are in general agreement with Meister’s [1967] velocity model of The Bend area, generated from his Dixie and IXL Canyon refraction lines. His bedrock-alluvium velocity contrast of 1.9:1 results in a fault dip of 25°. The coherency of the first arrival on Figure 3.5 indicates that the fault is smoothly varying and relatively planar down to 500 m depth. This is in contrast to the segmented first arrivals that would be generated by refraction and diffraction along a staircase-like fault geometry. Meister’s [1967] model in which he modeled a staircase-like fault geometry had insufficient resolution (cross-basin spacing of approximately one geophone per 400 m as compared to one geophone per 15 m in this study) to resolve downdip fault geometry at a fine scale.
A
postmigrated stack of our medium-resolution Cattle Road profile is presented in
Figure 3.7. The data were Stolt time migrated at a constant velocity of 2000
m/s. The interpreted fault plane reflector, dipping eastward at 25° to 30°, can
be traced to the Dixie Valley fault reflector seen in the high-resolution
profile (Figure 3.4). Reflections subparallel to the fault can be seen in the
footwall, suggesting the Caskey et al. [1996] foliation in the footwall
bedrock. The fault reflections are obscured below a highly reflective hanging
wall layer at ~400 ms. This first, strong basin reflector has been interpreted
to be a “capping” basalt layer by Hastings [1979] in a reflection profile 25 km
northwest, in the Carson Sink (Figure 3.1). This interpretation is supported by
drilling logs that intersected the profile. Okaya and Thompson’s [1985]
reflection profile near the Dixie Valley Geothermal Field (Figure 3.1) also
shows a similar reflection, which they interpret to be from the same reflector
seen in Hastings [1979]. This basalt is the culmination of a Tertiary volcaniclastic
sequence that is seen in all the ranges surrounding the Carson Sink and Dixie
Valley. The sequence appears to be locally thickened in the southern Stillwater
Range, with thicknesses approaching 500 m [Page, 1965]. We interpret the strong
series of reflections starting at 500 ms and ending at 900 ms to be originating
from this sequence.
Below the capping basalt, hanging wall stratigraphy can be traced to its termination against the fault at depth. Mapping the terminations allows us to extend our fault plane interpretation to 1.25 s. Many of the reflections show increased westward dip close to the fault, forming rollover anticlines. The rollover anticlines form in response to listric fault geometry.
Figure
3.8 shows the enhanced pre-stack migration that results from imaging the
medium-resolution data through the extended 8 m velocity model shown in Figure
3.6. Because of the low-pass filtering inherent in the antialiasing criterion
this section does not have as high a vertical resolution as the poststack
migrated section and may not image more steeply dipping structures such as
anticline flanks. As demonstrated in Death Valley, California, by Chavez-Perez
et al. [1998], however, the prestack migration through an optimized velocity
model does assure that this section puts structures at their true depths. Thus
any dips interpreted from Figure 3.8 will be accurate to within a few degrees.
The top
of the Tertiary basalt at 0.5 km depth, and its bend into a west dipping
rollover anticline against the Dixie Valley fault ~1 km east of the 1954 scarp,
can be interpreted in Figure 3.8. The prestack image shows a reflection
sequence at 0.8 km depth, which the stacked data could not image without
velocity pull-down effects. These reflections also show some evidence of
rollover and may originate at the bottom of the Tertiary volcanic sequence at
the top of an earlier-Tertiary basin-fill sequence. The image shows only parts
of the Dixie Valley fault plane itself, at 0.5, 1.4, and possibly 1.7 km
depths, dipping east from the 1954 scarp at 28°. The low-frequency response of
the antialiasing obscures the fault plane above 0.5 km depth. The prestack
migration, like the poststack migration, failed to clearly image the fault
plane below the strong reflectivity of the capping basalt. Truncations of the
deeper basin stratigraphy do support the linear, shallowly dipping fault
geometry.
The imaging of a flat basin fill sequence at 1.1 km, in on-lap relation to the Dixie Valley fault, suggests that no appreciable rotation of the fault plane has occurred since its
formation. It also suggests that significant
extension had occurred before the initiation of volcanism. The flanks of any
rollover anticlines at this depth could have been hidden by the migration
antialiasing criterion. Alternatively, a constant average fault dip could have
avoided the formation of rollover anticlines in the deep basin. The two upwarps
in the fault profile suggested by Figure 3.7 between 0.5 and 1 s two-way time
may have resulted in rollovers forming only updip of each. Although the fault
reflector itself cannot be seen below the capping basalt reflector, the Figure
3.8 imaging of a continuation of on-lap truncations along the projected fault
plane to 1.5 km depth supports our interpretation of an early, now deeply
buried, but unrotated episode of basin filling.
Bouguer
gravity along Cattle Road agrees with Schaefer’s [1983] study. The
2.5-dimensional gravity modeling (Figure 3.9) is consistent with a shallowly
dipping fault. In the cross section the Dixie Valley fault is modeled as
dipping 26°. Density values generally follow those of Speed [1976] and Thompson
[1959] as summarized by Okaya and Thompson [1985]. A value of 2.5 g/cm3
was chosen for the volcaniclastic units. Higher in the section, an average
value of 2.3 g/cm3 was chosen for volcanic units intercalated within
sedimentary units, as seen in Hastings’ [1979] well log in the Carson Sink. The
gravity effect of topography (terrain correction) was modeled in the plane of
the cross section.
Maximum basin depth is ~2700 m, using our density scheme. The Dixie Valley fault merges with the steeper Clan Alpine range-bounding fault. It is likely that the Clan Alpine range-bounding fault shown in Figure 3.9 is actually a series of faults as shown by Okaya and Thompson [1985], rather than the one large fault modeled. Owing to our lack
of accurate density information or seismic profiles
in this area, an effort was made to produce the simplest model that fits the
anomaly. Although a staircase-like fault geometry to the east cannot be ruled
out solely on the basis of gravity data, the absence of short-wavelength anomalies
suggests no rapid changes in fault dip. Results from the Scarp profile (Figure
3.2), not shown, indicate that intrabasement density contrast is negligible
along this reach of the fault.
Evidence
for low-angle faulting is seen at several scales, from observations at the
surface to over 2.5 km depth. Moving from shallow to deep, the evidence
consists of the following:
1.
Geologic evidence from Caskey et al. [1996], in the form of rupture mapping and
geologic cross-sections is valid from 0 to 10 m depth.
2. Our
high-resolution profile (Figure 3.4) confirms the geologic observations and
extends the smooth, low-angle fault plane to 75 m depth.
3. The no
a priori assumption velocity optimization (Figure 3.6) shows a surface of
increasing velocity dipping shallowly to 480 m depth.
4. Raw
shot gathers (Figure 3.5, for example) constrain the fault to be relatively
planar to 500 m and, given reasonable velocity estimations, suggest low-angle
dip.
5. The
medium-resolution time migration reflection profile (Figure 3.7) shows direct
fault plane reflections from 50 to 750 m. In addition, truncations in hanging
wall stratigraphy seen in the medium-resolution profile allow the
interpretation of a slightly listric low-angle fault plane to ~1.0 km depth.
6. The
same character is seen in the Kirchhoff depth migration of the
medium-resolution profile (Figure3. 8), extending observations to 1.75 km
depth.
7.
Gravity mapping by Schaefer [1983] (Figure 3.2) and this study (Figure 3.9) is
valid from the surface to the maximum depth of the basin, ~2.7 km. The gravity
data, too, are consistent with a low-angle fault geometry.
The
recognition of an early basin fill sequence below the mid-Tertiary capping
basalt (Figure 3.8) suggests that extension in the southern Stillwater Range
started earlier than extension to the north. Early sediments have been mapped
in the Stillwater Range at the latitude of our transects by John [1995].
Seismic reflection profiles near the Dixie Valley geothermal field [Okaya and
Thompson, 1985] and the northern Carson Sink [Hastings, 1979] record no such
sequence. The pervasive capping basalt layer, dated as 8 ± 4 Ma
[Hastings, 1979] and 13 to 17 Ma [Nosker, 1981], was interpreted to be preextension.
In southern Dixie Valley a rapid, but spatially limited, pulse of extension is
recorded by tilted fault blocks in the Stillwater caldera complex [John, 1995;
Hudson et al., 2000]. The tuffs, flows, and plutons associated with the complex
have tilts of 60° to 70°. Similar deposits in the southern Clan Alpine
Mountains and northern Stillwater Range dip < 30°, suggesting very localized
uplift and tilting. Hudson et al. [2000] estimate over 200% extension in a
brief period from a balanced cross-section near IXL Canyon (Figure 3.2). The extension
is well constrained to have started at 24.2 to 24.4 Ma [Hudson et al. 2000].
Parry et al. [1991], on the basis of fluid inclusion and alteration mineral
studies, also estimate extension starting at 20 to 25 Ma. The extension direction
was, in general, perpendicular to current structural strike, although some
vertical axis rotation has occurred since early Miocene time [Hudson and
Geissman, 1987; Hudson et al., 2000].
Two
contrasting models of early Miocene extension have been suggested. Parry et al.
[1991] suggest the current Dixie Valley fault is the major fault accommodating
extension and localized uplift and has been active since the early Miocene.
Similarly, King and Ellis [1990] cite the Dixie Valley fault as a potential field
example of a fault accommodating dramatic, localized uplift, starting in the
early Miocene. The rotated crustal blocks mapped in the Stillwater Range
represent rotation of footwall blocks of the Dixie Valley fault in this interpretation.
In contrast, Hudson et al. [2000] consider it unlikely that the current Dixie
Valley has been active longer than 13 to 15 Ma (postdating the capping basalt).
As evidence, they point to a sharp accommodation zone between two dip domains
in the Stillwater Range [Hudson et al. 2000]. The sharply defined accommodation
zone separates domains of east and west dip in the rotated caldera deposits,
requiring two large normal faults of opposing dips to form. The current Dixie
Valley fault cuts across both the east and west dip domains. Instead, they
favor a model in which extension along a west dipping (now inactive) detachment
rotated the caldera deposits, which are synthetically and antithetically
dipping in the hanging wall of the low-angle detachment [Hudson et al., 2000].
Current Basin and Range extension subsequently cut the resulting structure in
the middle Miocene [Hudson et al., 2000].
We favor
early Miocene initiation of the southern Dixie Valley fault from our reflection
profiles and evidence for low-angle dips. Evidence for early extension comes
from the basin fill and volcanic sequences seen in Figures 3.7 and 3.8 and
mapped in the Stillwater Range [John, 1995]. This flat-lying sequence is
directly below a thickened Tertiary (early to middle Miocene) volcanic section.
The volcanic section is locally thickened in the southern Stillwater Range
[Page, 1965]. The volcanic sequence may be thickened in our profiles because it
filled a preexisting basin at this latitude. The reflection profiles are
adjacent to the west dipping blocks in the Stillwater Range, which require a
large east dipping normal fault in precisely the same location as the current
Dixie Valley fault.
Formation
at low-angle may have been in response to rapid extension. Abers et al. [1997]
found a correlation between low-angle faulting and increasing strain rate in
the Woodlark-D’Entrecasteaux rift system, where strain rates of up to 10-8 s-1
are seen. Hudson et al. [2000] find minimum strain rates in the early Miocene
of 10-13 s-1 in the southern Stillwater Range. That strain rate may
be underestimated because the exact duration of extension is not known.
Nevertheless, strain rates at least an order of magnitude greater than usually
encountered may have played a role in forming the low-angle structure. Current
Basin and Range extension along the low-angle reach may have inherited the same
fault that accommodated early Miocene extension. North and south of the
low-angle section that formed in the early Miocene, extension had no favorably
oriented structures to inherit and formed a new steeply dipping normal range
front fault at ~13 to 15 Ma.
Estimates
for fault slip from our seismic lines can be broken into two phases (Figure
3.10). One phase is constrained by the thickness of the early Miocene basin,
and the other is constrained by the current elevation of the capping 13 to 15
Ma basalt. Our gravity results indicate a maximum basin depth of 2.7 km. The
bottom of the Tertiary volcanic sequence and top of the early Miocene basin is
interpreted to be at 1.1 km depth from our Kirchhoff migration (Figure 3.8).
Using dates and tilts of caldera deposits, Hudson et al. [2000] estimate that
this phase of extension lasted at most 4 Myr. The majority of the rotation was
most likely accomplished in ~0.2 Myr. The 1.6 km of vertical offset along a 30°
fault in ~0.2 Myr corresponds to slip rates of 18 mm/yr (extension rate of 9
mm/yr). This is an extraordinarily fast slip rate, but it has a current analog
in the Woodlark-D’Entrecasteux extensional province, which has extensional
rates of over 15 mm/yr and as high as 40 mm/yr in an area of low-angle normal
faulting [Abers et al., 1997]. However, using the most conservative estimate of
extension duration from Hudson et al. [2000] yields a more typical slip rate of
0.9 mm/yr in the early Miocene. Errors in the gravity model and the duration of
extension can make these fault slip rates uncertain, however.
The
second phase of extension postdates the middle Miocene capping basalt (13 to 15
Ma). The basalt is seen at 400 m below the surface (650 m elevation) in Figure
3.8 and at an elevation of 2500 m in the Stillwater Range. Accommodating the
1850 m of vertical offset along a 30° fault in 13-15 Myr requires a slip rate
of 0.28 to 0.32 mm/yr. These slip rates compare favorably with Meister’s [1967]
estimation of 0.3 mm/yr over 15 Myr for southern Dixie Valley. Okaya and
Thompson [1985] found a fault slip rate in northern Dixie Valley of 0.47 mm/yr
along a 50° fault (using Hasting’s [1979] basalt age of 8 Ma). This would be
equivalent to a 13 Myr slip rate of 0.29 mm/yr. Bell and Katzer [1990] report a
Holocene vertical slip rate of 0.5 mm/yr in the Bend area.
Our
results indicate that slip along a section of the December 16, 1954, Dixie
Valley earthquake rupture took place along a fault plane of unusually low dip
(<30°). In this regard, it is the first large historical earthquake on land
for which slip on a low-angle normal fault has been documented. Evidence for
the low-angle fault plane is seen at several different scales, from 0 to 2.7 km
depth. A computed velocity model, with no a priori assumptions, supports a
low-angle hypothesis, as does gravity modeling. Our results suggest extension
in the southern Stillwater Range had two distinct phases. The first period is
marked by rapid extension that initiated a low-angle fault. The second period
of extension began at 13-15 Ma and inherited a portion of the previous Dixie
Valley fault along the low-angle section. Fault sections with steep dip formed
at 13 to 15 Ma.
This project was funded by the National Science
Foundation under grant EAR-9706255 to J. Louie, S. J. Caskey, and S. Wesnousky.
The W. M. Keck Foundation donated seismic equipment, computers, and modeling
software. K. Miller and G. Abers provided careful reviews and helpful comments.
The 1998 Geophysical Applications class at the University of Nevada, Reno
performed all geophysical fieldwork. Class participants were A. Cadena, T.
Rabe, M. Herrick, M. Johnson, A. Rael, T. Blechen, and E. Hobson. C. Mann, J.
Ollerton, and J. Oswald rendered additional field assistance.
Chapter 4:
Weak ground motion and amplifications predicted from shear-wave
velocities at precarious rocks, near the 1857 rupture of the San Andreas fault
Robert E. Abbott, John
N. Louie, James. N. Brune, Abdolrasool Anooshehpoor
The University of Nevada, Reno, Seismological Laboratory and Dept. of Geological Sciences, Reno, Nevada 89557.
Sathish Pullammanappallil
Optim L.L.C., Seismological Laboratory, University of Nevada, Reno, Nevada
The
effects of local geology at fields of precarious rocks are inadequate to
explain the persistence of untoppled rocks near the San Andreas fault, given
current attenuation relationships. As evidence, we present weak ground motion
spectral ratios at six locations from 61 earthquakes, in situ velocity
measurements using both refraction-microtremor and tomographic techniques, and
synthetic spectral ratios based on our velocity measurements. We compare the
“known” toppling acceleration of the rocks and the accelerations predicted by
current USGS/CDMG probabilistic seismic hazard maps at the rock locations. We
modify the predicted USGS/CDMG accelerations by accounting for the measured
shallow shear-wave velocity structure at the rock sites and the assumed
velocity structure used for the maps. Comparing the synthetics generated using
our velocity modeling to synthetics generated using the soft rock/soil
interface velocity structure (NEHRP BC boundary), we find ground motions at the
precarious rock sites 0.75 to 1 times the ground motion at an assumed BC
boundary location at 3 to 5 Hz. Above approximately 5 Hz, the ground motion at
the precarious rock sites is amplified above the ground motions at the BC
boundary site. Velocity modeling shows that the precarious rock sites are
characterized by velocities significantly slower the than the BC boundary from
0 to 10 meters, and significantly greater below 10 m. Strong impedance
contrasts at the precarious rock sites in the upper 10 m caused by seismically
slow dry sand over fast fractured granite explains the amplification pattern.
The
geographical distribution of fields of precariously balanced rocks is gaining
increased acceptance as a potential ground motion constraint in seismic hazard
models. “Precarious” rocks (defined as rocks with toppling accelerations of 0.1
to 0.3 g) and “semiprecarious” rocks (0.3 to 0.5 g) are located as close as 14
km to the Mojave section of the San Andreas fault [Brune, 1999]. This section
of the fault ruptured in the M8, 1857, Fort Tejon earthquake. Sieh [1978], at a
trench site approximately 14 km due south of the precarious rocks, documented 8
large earthquakes in the last 1400 years. At least 6 of those events exhibit
ground deformation similar to the 1857 event [Sieh, 1978]. The continued
existence of such rocks in their current configuration for many thousands of
years [Bell et al., 1998] is incompatible with most ground motion prediction
models [Brune, 1996], since the rocks have remained untoppled for many cycles
of great earthquakes.
There are
several theories to explain the discrepancy. One line of reasoning is that
current attenuation relationships overstate the hazard of large magnitude
earthquakes. Strong-motion recordings of magnitude 7 and greater events at
short distances are rare, resulting in attenuation curves near great
earthquakes that are extrapolated by parameters found from smaller earthquakes
recorded at greater distances. Recent observations of ground motion from the
1999 Kocaeli, Turkey, MW 7.6, earthquake and the 1999
Chi-Chi,Taiwan, MW 7.6, earthquakes have dramatically increased the
number of strong motion recordings near large ruptures. Peak accelerations
recorded at these stations less than 100 km from their respective ruptures are
significantly less than predicted. Anderson et al. [2000] summarize these data
and conclude that it would be unsatisfactory to conclude that the two
best-recorded earthquakes are anomalous, and that current attenuation relationships
may overstate the hazard.
Brune
[1999] suggested that great earthquakes on the San Andreas fault have
“characteristic” spectra, repeated each earthquake cycle. This alternative
hypothesis is supported by the observation of Wesnousky [1988] that showed a
negative correlation in the “complexity” of a fault (measured by the number of
steps or fault segments per km length) and total fault offset. Longer, more
developed faults have fewer steps and other impedances to rupture, resulting in
slower earthquakes with less high frequency ground motion components. Building
upon this, Anderson and Brune [1999a] discuss a thought experiment created by
limiting or eliminating the statistical variation of repeated large earthquakes
and find a better fit to the precarious rock data.
In
another line of reasoning, Anderson and Brune [1999b] find that seismic hazard
maps that limit earthquakes to mapped faults (i.e. no “area” sources) are
compatible in most cases for precarious rocks in Nevada. Incomplete knowledge of
active fault distribution, as evidenced by the Landers, Hector Mine, and
Northridge earthquakes, suggests that this particular result of Anderson and
Brune [1999b] would not be appropriate for all situations. It would be most
appropriate for situations for areas where the hazard is dominated by a single,
well- characterized, fault, as is probably the case for the precarious rock
sites in this study.
Ground
motion is a convolution, of course, of source processes, effects on the seismic
waves along the path from the source to the site, and effects caused by the
very local geology of the site. This project specifically studies the problem
not at the source, but rather at the sites of the rocks themselves. It is
currently not known if these precarious rocks experience the same ground
motions as the surrounding regions they are located in, or if they define small
“islands of stability” in a general ground motion field. Namely, we aim to test
whether or not de-amplification of ground motion at the precarious rock sites
studied herein relative to “representative” sites is sufficient to decrease the
level of shaking to below the rocks’ toppling accelerations, using the
currently accepted attenuation curves. We accomplish this with a combination of
earthquake spectral ratios, in situ velocity measurements, and calculation of
synthetic ground motions.
In
October 1999, we deployed matched, calibrated, 3-componant digital stations
with 1 Hz L4 and S13 sensors at 6 sites near the San Andreas fault (Figure 4.1;
Table 4.1) in preparation to record blasts from the Los Angeles Regional
Seismic Experiment, Phase 2 [Similia et al., 2000]. We calibrated the sites by
applying a known voltage in the field and modeling the resulting waveform
(following a damped, harmonic oscillator) with three parameters: free period,
damping, and magnification. Four of the sites (Lovejoy Buttes, Piute Butte,
Alpine Butte, and near Black Butte) are located at or near fields of precarious
or semi-precarious rocks. Precarious rock are less than 200 m away at Lovejoy
Buttes, Piute Butte, and Alpine Butte. A thorough search for precarious rocks
has not been made near the Black Butte site, but the outcrop of rock where the
station is located is near an outcrop where
other precarious rocks exist. We also installed a station (Mill Creek Summit)
near the USC strong motion site MCS and one near the town of Llano, California.
Fortuitously, we recorded numerous aftershocks of the October 16, 1999, Hector
Mine earthquake during our 48 hour occupation of the sites (Table 4.2). The
poor signal-to-noise ratio of the blasts compared to the aftershocks led us to
focus on the earthquake data.
The
parameters of the 61 earthquakes in the analysis are presented in Table 4.2.
Table 4.2 also lists which sites recorded each event. The event magnitudes
ranged from 1.8 to 5.1 ML. Figure 4.2 shows a sample seismogram from
an aftershock. We tried to include events that have different azimuths to
create, as much as possible, a path-independent amplification scheme. Due to
our short occupation time at the sites, however, the sources are predominately
near the Hector Mine rupture area. Each event was examined and only events with
good signal-to-noise ratio on the majority of the stations were accepted for
further analysis.
We
computed the spectra of ground-velocity for each component of each record. The
Fourier spectra were calculated for a 5 second time window, starting one-half
second before S-wave arrival. This window was chosen to best contain most of
the high amplitude direct S-wave energy. As pointed out in Bonilla et al.
[1997] and Satoh et al. [2001] using longer times may result in better spectral
resolution but also contain energy from surface waves and other phases.
For each 5-second seismogram, the mean value was removed, and a 5% Hanning taper was applied. We accounted for attenuation using a frequency-independent Q value of 1000, based on when the high frequency spectra flattened out. This regional Q is similar to that found in Adams and Abercrombie [1998]. The ratios are, for the most part, insensitive to Q, except for ratios at MCS, the station farthest removed from the cluster of precarious rock stations. After FFT and Q removal, the two horizontal spectra for each
record were averaged to form one horizontal spectrum.
The resulting vertical and horizontal spectra were smoothed with a three point
moving average and then logarithmically stacked.
One of
the most common methods of estimating site response is the spectral ratio
technique [e.g. Borcherdt, 1970]. It is computed by dividing the Fourier
spectrum of one station by the Fourier spectrum of another. The spectral ratio
technique assumes that site effects can be deconvolved from the others if: a)
the instrument response is known for each station; b) the same sources are
used; and c) the path from source to station is the same. By using the same set
of earthquakes, with the hypocentral distance being much greater than the distance
between the stations, differences in source and path effects will most likely
be minimal. Local site conditions, therefore, cause the remaining differences
in the spectra.
In
general, spectral ratios are computed by dividing the horizontal spectra of a
station for which you wish to know the site response, by the horizontal spectra
of a nearby reference station on rock. The reference station, by virtue of
being on rock, is assumed to have minimal site effects, although there is
evidence that this is not necessarily true [e.g Steidl et al., 1996; Humphrey
et al., 1992, Tucker et al., 1984]. We chose Piute Butte as our reference site
because tomography results (discussed below) show it to be the site with the
fastest subsurface velocity.
We also
measured shallow shear-wave velocities at four of the sites for which we have
3-componant seismograms (LJB, PB, LLA, and MCS). Standard seismic refraction
equipment was used with 24 8-Hz vertical sensors and linear spread lengths of
~184 meters. The transects were as close as possible to the 3-component
stations as the local topography and access allowed. The resulting analysis
yielded the shear-wave velocity structure at the sites to a depth of 45+ m. The
data was acquired using a combination of refraction tomography and
refraction-microtremor (ReMi) techniques. The ReMi methodology is discussed in
detail in Louie [2001], who demonstrated that the resulting shear-wave velocity
depth profiles agree with borehole measurements to within 20% at other
locations.
To
supplement these data, we also used a commercial refraction velocity
optimization software, SeisOpt®@2D™ (© Optim LLC, 2001), to derive
P-wave velocity information from 1st arrival picks. This software uses a
non-linear optimization technique called simulated-annealing (Pullammanappallil
and Louie, 1994) to map subsurface velocities from first-arrival picks using no
a priori assumptions We picked at least 64 P-wave first arrivals along each
array, using hammer source points. After inversion for P-wave velocity, we
assume a constant Poisson’s ratio of 0.25, which allowed us to estimate the
shear-wave velocity and compare results from the two fundamentally different
methods.
We
computed synthetic transfer functions and synthetic spectral ratios for the
four for which we determined velocities. The synthetic ground motions were
calculated using the STK5 program (courtesy of John G. Anderson). STK5
calculates the response of a stack of sediments on a spectrally flat,
vertically incident, SH-wave. The algorithm calculates reflection and
transmission coefficients at each velocity interface for both the up-going and
down-going waves using the technique described by Luco [1983] and Apsel [1983].
The solution at the surface includes waves traveling in both directions. We
used our velocity models, as well as the velocity profiles for the NEHERP BC
boundary [Frankel et. al, 1996] and “generic rock” [Boore and Joyner, 1997], to
compute the synthetics. Also, to facilitate comparisons with our earthquake
spectral ratios, we ratio the synthetic spectra at each site with the synthetic
spectra from the Piute Butte velocity profile.
Shear-wave
velocities (Figure 4.3) from modeling of refraction-microtremor dispersion
curves (Figure 4.4) and refraction tomography (Figure 4.5) show that all sites
except MCS are in the NEHRP site class B (> 760 m/s) or even A (> 1500
m/s) range. In the upper 10 m, however all the sites are slower than the BC
boundary model used to generate the seismic hazard maps [Frankel et al., 1996].
The generic rock profile from Boore and Joyner [1997] matches the shallow depth
profiles much more closely. This is not surprising, given that the generic rock
profile is an average of velocity profiles from 57 rock boreholes. Our results
show that the precarious rock sites have higher velocities below 20 m than an
“average” rock site.
30-m
average velocity from the two methods (Table 4.3) match very well at MCS but agreement
is not good at the other sites. This is because the ReMi method emphasizes
layered geometry (all stations were modeled with 3 or fewer layers), while the
tomography method emphasizes velocity gradients. Also, all the models are not
unique, in that a number of plausible subsurface velocity structures can be
postulated that all fit the data equally well. In general, the two methods
agree better when deeper averages (i.e. averages over 50 or 60 m instead of 30
m) are used. Comparisons of the velocities at the maximum depth of penetration
show that the two methods agree to within 35%, even with the assumed 0.25
Poisson’s ratio.
Velocity
spectra for both the vertical and averaged horizontal components are presented
in Figure 4.6. As can be seen the horizontal spectrum is almost always greater
than or equal to the vertical spectrum. Also notice that for almost all the
stations, the maximum separation of the two components happens around 4-7 Hz.
Steidl et al. [1996], noted that spectral ratios using rock sites were often
underestimated as the rock sites had their own maximum site effect in this
approximate frequency range.
Stations
LLA and MCS have a slightly different character. The horizontal component at
both sites is elevated for the entire frequency range. A correlation seems to
exist between the relative amount of energy in the vertical and horizontal
components and the subsurface velocity. Sites with lower velocities (LLA and
MCS) have more energy in the horizontal, perhaps because the incoming S-wave is
refracted more toward vertical incidence near the sites. Alternatively,
contamination by reflected S and surface wave energy, due to both regional and
local geology may cause the elevated horizontal energy. Satoh et al. [2001]
compare P-wave, P-wave coda, S-wave, S-wave coda, and microtremor spectral
ratios and attribute the differences among the results to the relative strength
of surface waves. The surface waves are caused by conversion of phases at discontinuities
local to the site [Satoh et al., 2001]. A possible cause of the elevated
horizontal spectrum at LLA is its close proximity (< 2 km ) to the San
Andreas fault. Using the group velocities for low frequencies suggested in our
ReMi dispersion curves (Figure 4.4) shows that conversions at the SAF could
easily traverse the 2 km and contaminate our 5-s time window. Even if the
conversion of phases are not at the SAF itself, close proximity to the fault is
sure to cause some pervasive subsurface deformation. Similarly, tectonic
deformation the likely cause of the elevated low frequencies at station MCS,
located on pervasively fractured Precambrian rocks in the San Gabriel
Mountains. We ran separate analyses on subsets of the data to see if our goal
of a path-independent amplification scheme was achieved. Spectra from three
subsets of earthquakes were examined: the Hector Mine aftershock area, the Big
Bear Lake aftershock area, and the northern Mojave area (only two events here).
The spectra from all three source areas (not shown) are quite similar, leading
us to conclude that path induced affects are relatively minor. Very fine
structure in the spectra generated at the source might be averaged and
smoothed, however.
Traditional
spectral ratios at the precarious rock sites, with Piute Butte as the reference
site, show spectral ratios that are generally flat, with amplification factors
near unity (Figure 4.7). Again, the spectral ratios for MCS and LLA (Figure
4.7c-d) are different, probably due to being on highly fractured rock. The
spectra there have a “step-like” geometry with amplification relative to the
precarious rock site from 0.4 to 4 Hz, and de-amplifications relative to the
precarious rock site from 4 to 20 Hz. These results are reminiscent of those in
Stirling et al. [2001], who have studied the spectra of PB and LJB relative to
three TRINET stations on NEHRP class B sites. They find negative residuals
below about 4 Hz (meaning de-amplification at the precarious rock sites relative
to NEHRP B) and positive residuals above 4 Hz (amplification relative to NEHRP
B). If we consider LLA to be a NEHRP class B site, as suggested by the
tomography but not the refraction-microtremor, our results are in agreement.
Indeed, we show the same pattern of de-amplification of precarious rock sites
at low frequencies and amplification at high frequencies relative to our NEHRP
class C site, MCS. The flat spectral ratios with ALP and BB suggest similar
site conditions as LJB and PB and add robustness to the conclusions of Stirling
et al. [2001].
Synthetic
spectral ratios, based on the velocity models, show some success predicting
overall spectral shape (Figure 4.7). We feel confident that the velocity models
for all stations except MCS include information about the deepest relevant
basement interface (relevant in terms of having a velocity very near true
basement velocity with no significant impedance interfaces at greater depths).
The tomography model at MCS (Figure 4.5d) shows widely varying velocities at
the deepest depth of penetration. This suggests that the average velocity along
the 184-m array that both the tomography and the ReMi methods measure is too
low. As such, the synthetic transfer function generated for MCS is not
constrained at the lowest frequencies. Indeed, the synthetic ratios at MCS show
the most disagreement of all stations at low frequencies.
The
synthetics from both tomography and ReMi predict a change from amplification
relative to Piute Butte, to de-amplification relative to Piute Butte in the 5
to 20 Hz range, agreeing with the earthquake spectral ratios. We do not expect
the earthquake and synthetic ratios to be in perfect agreement because of any
number of assumptions that could be violated. Discrepancies between synthetics
and data could be due to a number of causes, including, but not limited to:
surface wave contamination, incorrect velocity models, and path-induced effects
on the input spectra below each site.
Comparisons
to the BC boundary model [Frankel et al., 1996] and generic rock [Boore and
Joyner, 1997] give seismic hazard context to our data (Figure 4.8). Care must
be taken analyzing the synthetics as they still have the same frequency
limitations and caveats as the ratios with Piute Butte. However, the data still
show the consistent result of de-amplification at precarious rock sites at low
frequencies and amplification at high frequencies, seen in this study and in
Stirling et al. [2001]. The NEHRP class C site, MCS, shows the least
amplification above 5 Hz, while the faster rock sites show much more.
The
synthetic ratios show amplification factors relative to the BC boundary model
for the precarious rocks of ~0.75 at 3 Hz, and ~1 at 5 Hz. Shake table results
(not shown) demonstrate that the peak ground acceleration threshold at which
models of precarious rocks first topple is linearly proportional to the
spectral accelerations at 3 and 5 Hz. That is, plots of PGA at the minimum
toppling threshold of PGA versus 5 Hz acceleration have a linear trend (with
some scatter). The scatter is greater at 3 Hz, with 1 Hz and below having no
relationship to PGA. This relationship implies that we can scale the probabilistic
seismic hazard PGA at the precarious rock sites (0.6 g with a 10% probability
of exceedance in 250 years [Frankel et al., 1996]) by the amplification factors
at 3 and 5 Hz. The rocks at LJB and PB have toppling peak accelerations of ~0.4
g [Brune, 1999]. Applying the amplification factors at 3 and 5 Hz results in
revised PGA estimates of 0.45 to 0.6 g. Given the mild de-amplification, and
rocks that have been balanced for around 40 times the exposure time of the PSH
estimate, we conclude that the persistence of the rocks in their current
configuration can not be attributable to de-amplification due to local shear
velocity.
Ground
motions at the precarious rock sites in the Mojave are de-amplified below 5-8
Hz and amplified from 10 to 20 Hz. Therefore the results of this experiment
indicate that site effects due to local geology cannot be used to explain the
continued existence of balanced rocks. Velocity modeling suggests the cause of
the amplifications are extremely strong impedance contrasts in the upper 20 m.
The precarious rocks are formed as the hard granitic rocks are differentially
eroded, leaving balanced outcrops. The current arid, erosional environment at
the precarious rock sites leads to the juxtaposition of dry, unconsolidated
granitic sands against hard, fractured rock. Even though the average velocity
to 30-m depth indicates NEHRP class A or B, the upper 10 m is slower than an
average soft-rock/soil interface site. It is the travel time, however, and not average
velocity, that is the critical amplification parameter and even very thin
layers can have large effects on the seismic travel time [Boore and Joyner,
1997]. The Mojave desert sites, due to the local climate and geology, have
slower than average velocities in the upper 10 m, and therefor greater than
average amplification factors above ~5 Hz. This may not be true for precarious
rock sites in different environments, such as those found in the Sierra Nevada
Mountains or near the Nevada Test Site.
We are
guilty, however, of using smaller earthquake data at greater distances and
applying it to a large San Andreas fault earthquake. It is widely recognized
that deep basin geometry can be a dominant factor in determining amplification
at sites within sedimentary basins [e.g., Gao et al. 1996, Hartzell et al.
1997; and Davis et al., 2000]. Subsurface focusing of seismic energy
propagating through velocity contrasts at basin margins is thought to be a
mechanism producing the amplifications. If this is true, than there must be
areas of de-focusing and de-amplification. It is possible that the fields of
precarious rocks are located in such an area for a San Andreas event. By
necessity, we essentially examined ground motions from sources near the
October, 2000, Hector Mine earthquake area and the amplification patterns will
be different for a San Andreas event. 3-D basin amplification patterns can
change dramatically given different earthquakes on the same fault. Even
changing the rupture direction of an earthquake can have dramatic results in
finite difference simulations [Olsen, 2000] and the effects of rupture
directivity have lead some to propose changes to the empirical attenuation
relations [Sommerville et al., 1997]. Given that the rocks have been in their
current configuration for as many as 10 great events on the San Andreas fault,
some sort of rupture pattern would probably have resulted in sufficient ground
motion to topple the rocks if current attenuation parameters are correct.
We thank
Glenn Biasi, John G. Anderson, and Mark Stirling for their valuable input and
assistance. This research was supported by the Southern California Earthquake
Center. SCEC is funded by NSF Cooperative Agreement EAR-8920136 and USGS
Cooperative Agreements 14-08-0001-A0899 and 1434-HQ-97AG01718. The SCEC
contribution number for this paper is 601.
Chapter
5: Recommendations for Future Work
In this chapter I will briefly describe my recommendations on how the work and conclusions presented in the previous three chapters can be strengthened through additional work. Each project will be treated separately.
In Chapter 2 I showed that quick and accurate basin mapping was possible using gravity measurements in key locations and very simple approximations. Subsequent to the publication of Abbott and Louie [2000], I became aware of two new wells that have penetrated bedrock (both in the “Verdi Basin” area. Examination of well logs and cuttings at Summit Engineering in Reno, NV, revealed that in these two wells, my depth-to-bedrock predictions were consistent to within 10%. There are several ways other than using new borehole results to test the data and to improve seismic hazard prediction.
As noted
in Chapter 2, there are several areas of insufficient gravity coverage in key
areas. These areas manifest themselves most clearly when negative depth-to-bedrock
values are predicted. Examples of these areas are I80-US395 interchange in Reno
(Figure 2.6) and the area just east of the 5th Street transect in
Carson City (Figure 2.12). These are the most obvious areas, but locations of
sparse gravity coverage near sharp gravity gradients are also areas of concern.
The southwest portion of the West McCarran gravity trough stands out as an
example of this.
Washoe County has recently acquired many new gravity measurement points between Verdi and Reno that would greatly improve coverage in this area [Mike Widmer, Washoe County, personal communication]. In the Reno area, I would also suggest transects along Sutro Street and Clear Acre Lane as being particularly useful.
Carson City coverage suffers from a general lack of measurements in the surrounding margins of the basin. Additional measurements will be difficult due to access problems, but just a few measurements in keys areas would be of use in making the bedrock gravity field more accurate.
One major weakness in our gravity modeling was our incomplete knowledge of density contrasts. We tried to make use of the nearby measurements of Thompson and Sandberg [1958] and the regional model of Blakely et al. [1999], but measurements within the local outcrops of the Hunter Creek and Kate Peak formations would have been particularly useful. The Hunter Creek sandstones would probably have to be sampled in more locations because it seems apparent that the density of the different members of the formation varies widely. This would allow the bracketing of the density contrasts between likely minima and maxima for “average” Hunter Creek for the one-dimensional approximation, or even the introduction two- or three-dimensional models.
Although the goal of Chapter 2 was to predict basin depth, it was part of a larger project whose goal was to estimate seismic hazard due to basin amplification of seismic waves. Finite difference modeling in the manner of Olsen et al. [1995] has not yet been undertaken. It is quite possible that the modeling will reveal a significant basin amplification contribution to the seismic hazard in the West McCarran area of Reno. The modeling should include the addition of several model earthquakes, such as magnitude 7+ earthquakes along the Genoa Fault and other possibilities listed in dePolo et al. [1995]. As is shown in Olsen [2000], the effects of basin geometry on seismic waves can be very different for different source azimuths, so many models should be run to get a full picture of the potential hazard.
In the coming years, the urban areas of Reno and Carson City will become home to many new quality digital seismic stations as part of the Advanced National Seismic System, including a proposed 200 new stations for Reno alone [http://quake.utah.edu/anss/Strawman.html]. Sites for the new stations are picked with such criteria as geologic formation and topography in mind (John Anderson, personal communication). I would suggest using depth-to-bedrock as a selection criteria as well. Sampling a number of different gravity gradient and depth environments (e.g., shallow and flat, deep and flat, deep and steep, etc.) and comparing actual seismic shaking to that predicted using finite differencing would be extremely rewarding.
As mentioned briefly in Chapter 2 and more extensively in Chapter 4, prediction of amplification of seismic waves also requires the knowledge of local shear-wave velocity. Measurements in a few different geological units (picked using the maps of Bell and Garside [1987], Bonham and Bingler [1973], and Trexler [1977], could help in creating a 30-m shear wave velocity map for Reno and Carson City as Wills et al. [2000] did for the state of California. The Wills et al. [2000] map has recently been correlated with some success with ground motions observed in southern California [Field, 2000]. This recommendation would be particularly helpful if all the existing seismic stations and proposed ANSS sites were measured so that detailed comparisons could be carried out.
In Chapter 3, I have proved that a portion of the slip on the 1954 Dixie Valley earthquake was accommodated along a low-angle reach of the fault. The strongest area of proof is directly underneath the geophysical transects, with some additional information coming from both the gravity survey of Schaefer [1982] and the geologic mapping of Caskey et al. [1998]. Current evidence points to approximately 50% of the fault’s strike-length being characterized by low-angle dip, but this value is uncertain. Neither do we have a firm handle on the mechanisms that allowed the system to slip at low angle. As the only truly accessible active low-angle normal fault (as opposed to being on the bottom of the sea or being a re-activated basement thrust), I believe the Dixie Valley fault warrants additional study. Below are some financially attractive ideas to study both the kinematics and the dynamics of the fault.
With the
1992 M=7.3 Landers, California earthquake came increased recognition that
dynamic strains associated with distal earthquake ruptures could trigger
seismicity [e.g. Hill et al., 1993; Anderson et al., 1994]. Statistically
significant increases in seismicity were noted up to 1250 km (17 fault lengths)
from the source, ruling out static stress changes as a possible mechanism [Hill
et al., 1993]. I propose to study the likelihood of Fairview Peak earthquake
waveforms dynamically triggering the Dixie Valley earthquake.
The
geometry and source-time functions of the Fairview Peak earthquake are known from
the body wave inversions of Doser [1987]. The aftershock studies of Stauder and
Ryall [1967] further constrain the dimensions of the Fairview Peak fault. We
also know the geometry of the Dixie Valley fault from Abbott et al. [2001].
This presents the opportunity to resolve dynamic strains caused by the Fairview
Peak event on the plane of the ‘known’ Dixie Valley fault. Modeled velocity and
reflectivity contrasts along the fault will partition the incoming disturbance.
As a first step, I propose the re-evaluation of the work of Caskey and Wesnousky [1997]. They modeled the static stress implications of the 1954 earthquake sequence using the Coulomb3D program of Ross Stein. Caskey and Wesnousky [1997] did not, however use the proper dip of the Dixie Valley fault in their model (the low-angle nature of the fault being merely proposed at the time). The model should be changed to examine the static stress implications of waves from the Fairview Peak sequence impinging upon a low-angle Dixie Valley fault.
For the
dynamic triggering situation, existing algorithms could be used to first model
the time-varying strain field in 1-dimension, and subsequently in 3 dimensions.
For input into the 3-d inversions, the depth to bedrock data compiled in
Blakely et al. [1999] could act as a starting point in the velocity modeling.
5.2.1.1 Differentiating among the possible mechanisms for fault weakening.
Armed
with the dynamic models, it will be possible to conclude which mechanisms might
be responsible for allowing low-angle normal slip. Dynamic mechanisms that have
been proposed are:
High pore flujid pressures
Transient pore-fluid pressure [e.g. Sleep and
Blanpied, 1994; Axen, 2000]
Gouge Liquefaction
Acoustic Fluidization [e.g. Melosh, 1997]
Frictionless Rollers [e.g. Scott, 1996; Anooshehpoor
and Brune, 1996]
Opening Modes [Brune, 1997]
Anomalous low frequency pulse [Anderson et al., 1994]
An
extensive body of literature has been written on each of these mechanisms. Strain modeled on the Dixie Valley fault will
provide some evidence for and against the individual mechanisms. For example,
acoustic fluidization in the fault zone requires high amplitude, high frequency
strain [Melosh, 1996]. In contrast, the mechanism Anderson et al. [1994] proposed
for Landers-triggered seismicity was a long period (>10 s), high amplitude
pulse. This type of modeling will not uncontrovertibly solve the problem of
dynamic triggering by any means. Rather, it would provide some additional
evidence for this very interesting and difficult problem.