Stress orientations in northern and central ... - AGU Publications

0 downloads 0 Views 12MB Size Report
high angles of SH ($80°) found immediately adjacent to, as well as farther from, the creeping segment ... frictional strength along the San Andreas plate boundary system, J. Geophys. Res. .... quiet all the way up to Mendocino Triple Junction, except on ...... cation to the San Fernando earthquake sequence, J. Geophys. Res.
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. B3, 2175, doi:10.1029/2001JB001123, 2003

Stress orientations in northern and central California: Evidence for the evolution of frictional strength along the San Andreas plate boundary system Ann-Sophie Provost1 and Heidi Houston Department of Earth and Space Sciences, University of California, Los Angeles, California, USA Received 24 August 2001; revised 27 June 2002; accepted 9 December 2002; published 29 March 2003.

[1] We determined orientations of principal stresses around the San Andreas fault (SAF)

system in the greater San Francisco Bay Area and regions farther north along the strikeslip plate boundary. Stress orientations, as well as a ratio between principal stress magnitudes, were determined by inversions of 6000 earthquake fault plane solutions, divided into 100 groups based on the spatial distribution of seismicity with respect to major regional fault strands (i.e., the SAF, the Hayward-Rodgers Creek-Maacama fault zone, and the Calaveras-Green Valley-Bartlett Springs fault zone). The stress orientations, while spatially variable, reveal several features; to describe them, it is useful to distinguish between groups of events occurring on and off the major fault strands. First, for off-fault groups, the angle between major fault strands and the maximum horizontal compression SH decreases systematically to the north. This contrasts with the high angles of SH (80) found immediately adjacent to, as well as farther from, the creeping segment of the SAF in central California by Provost and Houston [2001]. Second, for on-fault groups, the angle that SH makes with major fault strands changes little along strike, averaging 50 to 55 in the creeping segment, the Bay area, and the northernmost part of our study area. Third, as in the vicinity of the creeping segment of the SAF in central California, the majority of both the off-fault groups, and the on-fault groups, are in a strike-slip, rather than thrust, tectonic regime. Finally, anomalous eastwest SH orientations are seen in the vicinity of Sutter Buttes. In the north, multiple strands of the strike-slip fault system have accumulated little slip, dip relatively shallowly and are composed of short, complex en e´chelon segments, suggesting that they originated as thrust faults in the accretionary prism associated with the Farallon subduction and have been subsequently reactivated in a strike-slip sense following the northward passage of the Mendocino triple junction. Our results, together with the geological context, suggest that the fault system is mechanically stronger with a greater effective frictional strength in the northern portion than in the creeping section of the SAF, with the Bay Area in an intermediate state. We interpret this situation to result from the evolution of the plate INDEX TERMS: boundary toward lower effective frictional strength with increasing slip. 7230 Seismology: Seismicity and seismotectonics; 7209 Seismology: Earthquake dynamics and mechanics; 8150 Tectonophysics: Evolution of the Earth: Plate boundary—general (3040); 8110 Tectonophysics: Continental tectonics—general (0905); KEYWORDS: fault strength, fault mechanics, San Andreas fault, fault friction, Bay Area, Sutter Buttes Citation: Provost, A.-S., and H. Houston, Stress orientations in northern and central California: Evidence for the evolution of frictional strength along the San Andreas plate boundary system, J. Geophys. Res., 108(B3), 2175, doi:10.1029/2001JB001123, 2003.

1. Introduction [2] The level of the frictional strength of the San Andreas fault (SAF) has been a point of contention for over two decades. Many researchers have proposed that the shear 1 Now at Laboratoire Dynamique de la Lithosphere, Universite´ Montpellier 2, Montpellier, France.

Copyright 2003 by the American Geophysical Union. 0148-0227/03/2001JB001123$09.00

ETG

stress at which the SAF slips must be very low compared to that implied by friction levels measured in laboratory experiments on many rock types [Byerlee, 1978]. Briefly, there are two main arguments for low frictional strength of the SAF. First, a heat flow anomaly localized around the SAF is expected according to steady state conductive models of frictional heating at Byerlee friction levels, but such an anomaly is absent [e.g., Brune et al., 1969; Lachenbruch and Sass, 1980]. Second, the maximum horizontal compressive stress in central California off the SAF has been found to lie at high angles to the fault, requiring the fault to slip at low

13 - 1

ETG

13 - 2 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

levels of shear stress compared to those needed in the adjacent crust [e.g., Mount and Suppe, 1987; Zoback et al., 1987]. A further line of evidence that many major faults are weak follows from the need for weak faults embedded in a strong crust in various numerical modeling studies [e.g., Bird and Kong, 1994; Bird, 1995; Che´ry et al., 2001]. [3] Questions remain about the implications of the lack of a localized heat flow anomaly on the SAF. The main issue is that advection of heat by groundwater flow may play a large role in removing frictionally generated heat. The rate of advection depends on the permeability of the crust, which is not well constrained and may vary spatially with rock type, as well as with scale. The presence of fault-related topography near the SAF is also a factor, tending to promote gravity-induced groundwater flow [e.g., Williams and Narasimhan, 1989]. Recently, these concerns were reiterated by Scholz and Hanks [2003]. Given this uncertainty, as well as the large and growing catalog of microseismicity in California, thorough analysis of stress orientations around the SAF system is called for. [4] Here, we mention a few particularly salient studies of stress orientations in the vicinity of the SAF. A key issue is orientation of the maximum horizontal compression SH adjacent to the fault, because this orientation indicates whether Anderson-Byerlee mechanics are allowed (i.e., whether the SAF is slipping in response to levels of shear stress compatible with Byerlee friction). [5] On the basis of data sets including fold axes, well bore breakouts, and individual P axes of off-fault earthquakes, Mount and Suppe [1987] and Zoback et al. [1987] found that SH is oriented nearly perpendicularly to the SAF in central California. They inferred from this that the SAF slips at low levels of shear stress. However, most of their data were from shallow depths relative to seismogenic depths, and only a few data were located very near the SAF. A different technique, inversions of groups of focal mechanisms, provides stress orientations at seismogenic depths and can be applied wherever sufficient seismicity has occurred. Hardebeck and Hauksson [1999] inverted 50,000 of focal mechanisms in southern California, grouping seismicity according to distance from the SAF. Townend and Zoback [2001] performed a similar analysis but grouped the seismicity in a different way. Far from the SAF, SH typically lies at a high angle (>60) to the fault. Close to the SAF, the angle that SH makes with the fault is difficult to interpret in some locations and is still controversial. [6] Recently, Provost and Houston [2001] inverted 2400 of focal mechanisms to determine stress orientations around the creeping segment of the SAF in central California. They found high angles of SH (80) immediately adjacent to the creeping SAF, implying that the creeping segment is a narrow mechanically weak zone, in which intrinsic friction is low or pore pressure is high. [7] In this paper, we extend our analysis of stress orientations to portions of the strike-slip plate boundary system farther north in California. In the San Francisco Bay Area, the SAF system branches into several major faults, which each carry significant portions of the slip, in contrast to the situation in central California [Brown, 1990]. Thus the relatively young plate boundary system in northern California spans a greater width than the mature boundary in central California, and, as we will show below, appears to

Figure 1. Location of seismicity used for inversions in this study, in northern California and the greater Bay Area. The events shown (6000) occurred between January 1969 and October 2000, have M  1.5, and were selected from the Northern California Earthquake Data Center (NCEDC) according to criteria for well-constrained focal mechanisms discussed in the text. The region was divided in two quadrilaterals covering the San Andreas plate boundary system in the greater Bay Area and Northern California. Faults are from Jennings [1992]. Abbreviations are BS, Bartlett Springs fault; C, Calaveras fault; Ga, Garberville fault; Gr, Greenville fault; GV, Green Valley fault; Ha, Hayward fault; He, Healdsburg fault; LM, Lake Mountain fault; M, Maacama fautl; PD, Point Delgada; RC, Rogers Creek fault; SAF, San Andreas fault; and SB, Sutter Buttes. be mechanically stronger, suggesting that the effective frictional strength of the SAF system varies along its strike.

2. Distribution of Seismicity and Selection of Fault Plane Solutions [8] The area studied in this analysis covers two zones (Figure 1): first, a 150-km-wide zone centered on the greater San Francisco Bay Area, bounded to the south by the Loma Prieta aftershock zone and to the north by San Pablo Bay, and second, a 200-km-wide zone extending from San Pablo Bay to Point Delgada (south of the Mendocino Triple Junction). In these regions, the fault system is not as simple as in the central section of the SAF, where only one strand accommodates almost all the slip. Instead, it includes fault strands branching off the San Andreas fault including the Calaveras and Hayward faults in the Bay Area, as well as en e´chelon

PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

faults, subparallel to the SAF, such as the Rogers CreekMaacama and the Lake Mountain-Bartlett Springs fault zones in northern California. The presence of multiplestranded faults controls the spatial distribution of the seismicity. The San Andreas Fault itself is generally seismically quiet all the way up to Mendocino Triple Junction, except on the segment that was reactivated during the 1989 Loma Prieta event. Subparallel faults, like the Garberville-Maacama-Healdsburg-Rogers Creek and the Lake MountainBartlett Springs-Green Valley fault systems, are the loci of most of the seismicity in the area. Off-fault seismicity does occur in both areas, in between major fault trends (in the Bay Area and up to Point Delgada), and to the east of the fault system (in the Sacramento Valley). The majority of these earthquakes are between 4 and 12 km deep, but seismicity occurs as deep as 28 km farther north, in the region 50 km to the east of the Bartlett Springs fault system. These deep events are part of two lineaments that trend N-S and point toward the Cascadian volcanoes, Mount Shasta and Lassen Peak. Thus they may be spatially associated with volcanic activity similar to that occurring under relatively young volcanoes related to the Cascade subduction zone [Castillo and Ellsworth, 1993; Walter and Dzurisin, 1989]. [9] The focal mechanisms analyzed here were taken from the Northern California Earthquake Data Center (NCEDC) catalog and correspond to earthquakes with ML  1.5 occurring in the area of interest between January 1969 and October 2000. Most of the seismicity in the northernmost zone was recorded after 1980 because of the relatively late improvement of the seismic network coverage in this area, compared to the Parkfield area. [10] The events to be inverted are screened according to the quality of the fault plane solution. Because of the difference in data abundance and quality in the two regions, we used different selection criteria for each region. In the greater Bay Area we selected 3200 well-constrained focal mechanisms (out of 12,500 mechanisms) according to the following criteria: depth  2 km, solution misfit value  0.25 (unless ML  3.0), number of first motions  30, station distribution ratio  0.45, ratio of number of machine picks to number of hand picks  0.3, and maximum half width of the 90% confidence range of strike, dip, and rake  20, 35, and 30, respectively. Farther north (from Pablo Bay to the triple junction) we selected 2800 events (out of 4100) whose parameters satisfied the following less selective criteria: depth  1 km, solution misfit value  0.3, number of first motions  30, station distribution ratio  0.45, ratio of number of machine picks to number of hand picks  0.54 and maximum half width of the 90% confidence range of strike, dip, and rake  36, 46, and 54 respectively. These values were chosen because they tend to characterize better constrained fault plane solutions, while ensuring that sufficient numbers of events were retained. The parameters are described in the section of the NCEDC online manual under Fault Plane Solution Format ftp://quake.geo.berkeley.edu/pub/doc/cat5/ncsn.mech.y2k.5). [11] Once selected, the seismicity is sorted into groups of varying shapes and extent based on two principles; proximity to one another and, to the degree discernable, occurrence on vs. off a major fault; on-fault events were usually grouped in elongated bins only 3 to 8 km wide and off-fault events in more equant bins. None of the bins exceeding 20

ETG

13 - 3

by 45 km in size, and most would lie within a rectangle of 10 by 20 km. Many in the Bay Area are much smaller. The spatial distribution of these groups can be estimated from Figures 5 and 7, on which the direction of the maximum horizontal compression, SH, obtained after inversion of fault plane solutions in each bin, is plotted on the map at the average latitude and longitude of the events in the corresponding bin. These locations as well as the number of events in each bin are listed in Tables 1 and 2, for northern California and the Bay Area, respectively. To identify the bins on the maps, each bar is labeled with the corresponding bin number.

3. Stress Inversion Methods [12] In the past 30 years, several inversion codes have been developed in order to obtain information on the stress tensor from earthquake focal mechanism or slickenside data [Carey and Brunier, 1974; Angelier, 1975; Angelier and Mechler, 1977; Etchecopar et al., 1981; Gephart and Forsyth, 1984; Michael, 1984; Gephart, 1990; Michael, 1987a]. These methods are similar in that they give the directions of the principal stresses, S1, S2, and S3, and a ratio of their magnitudes s1, s2, and s3. Absolute magnitudes cannot be determined from focal mechanisms or slip data. The methods differ primarily in the procedures used to determine the trial tensors and to calculate the error estimated for the best tensor retained. They generally make four assumptions: the homogeneous part of the stress tensor is larger than the heterogeneous part over the volume containing the seismicity to be inverted; the material in which these earthquakes occur behaves isotropically; the focal mechanisms in a group show sufficient diversity; and slip on a fault plane occurs in the direction of the shear stress resolved on that plane. [13] Here we use an approach similar to that of Provost and Houston [2001], who inverted focal mechanisms in central California. We employ two inversion codes, and compare results. The first one, Fault Slip Analysis software (FSA), was developed by Ce´le´rier [1995], and employs a Monte Carlo search technique following Etchecopar et al. [1981] to determine trial stress tensors. The misfit for a given trial tensor is taken as the average of the angle between the observed slip on each fault plane in the data set and that predicted by the trial tensor. The trial tensor with the lowest misfit is taken as the best stress tensor. The second code, Slickenside Analysis software, was developed by Michael [1984, 1987b] based on a linear inversion algorithm. This code yields confidence limits on the best tensors using a bootstrap resampling method. [14] When inverting fault plane solutions it is usually not known which of the two nodal planes is the actual fault plane. We developed a well-defined procedure to choose the optimal fault plane, which is preferable to using both planes (which leads to difficulties in estimating confidence limits [Michael, 1987a]) or simply using the plane listed in an earthquake catalog (which is often in effect a random choice). As a first step we inverted both nodal planes in each bin with FSA, then we selected the nodal plane which would give the lowest misfit with the best tensor obtained. We then inverted only these selected fault planes with Michael’s [1984] code. As a check, in northern California

ETG

13 - 4 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

Table 1. Results of Stress Inversions in Northern California Using A. Michael’s Program and Selected Fault Planes Bin

Events per Bin

NCA1 NCA2 NCA3 NCA4 NCA5 NCA6a NCA7a NCA8 NCA9 NCA10a NCA11a NCA12 NCA13 NCA14 NCA15 NCA16 NCA17a,f NCA18a NCA19a NCA20 NCA21 NCA22 NCA23 NCA24 NCA25 NCA26 NCA27a NCA28a NCA29a NCA30 NCA31 NCA32 NCA33 NCA34 NCA35a NCA36a NCA37 NCA38 NCA39 NCA40 NCA41 NCA42a NCA43 NCA44 NCA45 NCA46 NCA47 NCA48 NCA49 NCA50 NCA51 NCA52 NCA53 NCA54f NCA55 NCA56 NCA57 NCA58f

11 11 8 14 5 29 11 15 13 44 6 23 6 6 51 4 29 10 9 16 34 32 6 7 20 15 25 26 26 21 13 71 27 10 7 15 24 31 237 9 10 34 31 17 68 6 14 7 16 8 11 9 11 12 13 23 102 7

Locationb Long, Lat

S1c Azimuth, Plunge

S2c Azimuth, Plunge

S3c Azimuth, Plunge

Form Ratio r0d

Mean Misfit (Angular Error)

123.62, 123.47, 123.13, 123.20, 123.65, 123.08, 123.17, 123.04, 123.09, 123.27, 123.54, 123.38, 123.15, 123.45, 123.04, 122.88, 122.87, 122.98, 123.07, 122.79, 122.94, 122.75, 122.93, 123.19, 122.69, 122.68, 122.86, 122.63, 122.75, 122.59, 122.65, 122.98, 122.78, 122.61, 122.67, 122.99, 122.40, 122.87, 122.78, 122.81, 122.62, 122.19, 122.66, 122.26, 122.43, 122.60, 122.05, 122.03, 122.22, 122.08, 122.02, 122.09, 121.80, 121.53, 122.43, 122.20, 121.89, 121.89,

163.3, 8.5 25.4, 13.8 17.8, 7.5 24.9, 1.3 151.5, 1.4 20.2, 1.3 33.1, 15.7 29.1, 5.7 35.1, 19.0 14.1, 2.5 0.6, 16.1 165.7, 10.1 27.4, 26.2 43.5, 24.1 8.9, 5.0 18.4, 4.9 62.9, 55.2 125.0, 8.1 178.5, 8.3 20.3, 24.1 29.7, 30.7 17.9, 23.9 168.2, 9.5 28.2, 33.0 170.0, 9.3 16.0, 7.0 25.8, 8.4 25.6, 5.9 13.7, 3.3 19.0, 14.7 147.4, 4.9 37.0, 2.0 20.5, 6.0 154.2, 17.3 29.2, 17.4 103.9, 35.6 30.3, 10.8 157.6, 11.4 26.6, 4.3 13.2, 6.8 15.6, 11.5 39.3, 8.0 28.3, 6.2 33.9, 7.0 25.3, 19.0 23.5, 10.6 74.6, 16.2 89.7, 15.6 84.6, 13.3 78.4, 3.8 4.3, 4.2 168.4, 15.4 26.8, 10.4 5.4, 56.4 18.8, 8.9 33.5, 1.7 27.2, 13.6 88.9, 41.8

44.1, 72.9 163.5, 71.8 126.0, 67.2 115.2, 13.9 61.0, 21.9 114.5, 73.4 139.3, 44.9 125.4, 47.9 157.0, 70.5 108.8, 62.2 151.1, 71.6 82.0, 64.6 162.9, 55.3 150.8, 33.7 175.2, 84.8 113.6, 47.0 156.1, 28.2 137.0, 43.8 77.8, 51.8 174.4, 63.5 168.0, 58.0 170.2, 63.4 67.0, 73.6 151.7, 40.3 71.0, 43.8 109.1, 24.2 119.6, 24.8 146.5, 78.5 109.1, 58.4 169.9, 73.2 29.1, 79.8 131.3, 65.4 151.7, 80.9 62.2, 68.7 149.1, 57.8 154.9, 15.0 136.3, 55.4 92.1, 59.7 131.5, 73.8 107.5, 32.4 109.7, 19.9 79.2, 73.7 133.8, 67.9 83.8, 75.2 66.8, 6.3 135.5, 63.5 169.1, 15.2 90.5, 74.3 36.3, 65.3 169.9, 22.7 120.9, 80.9 58.6, 50.9 65.5, 12.9 175.6, 33.6 157.2, 78.2 78.6, 85.4 131.7, 46.3 132.8, 39.8

104.3, 14.6 67.5, 11.6 75.0, 21.3 70.4, 76.0 114.8, 68.0 70.2, 16.5 70.9, 40.9 65.9, 41.5 56.1, 3.8 77.2, 27.6 91.8, 8.5 71.4, 23.0 73.4, 20.9 74.5, 46.4 81.1, 1.2 76.1, 42.5 55.8, 18.5 26.9, 45.0 85.1, 36.9 74.3, 10.2 65.0, 8.0 77.0, 10.9 76.0, 13.2 85.9, 32.1 90.5, 44.7 88.9, 64.6 81.5, 63.6 65.4, 9.7 78.3, 31.3 73.0, 7.7 121.7, 8.9 53.9, 24.5 70.2, 6.7 60.5, 11.9 69.6, 26.1 46.1, 50.3 66.6, 32.3 61.6, 27.6 64.5, 15.5 87.1, 56.7 102.6, 66.7 131.3, 14.1 64.0, 21.1 125.5, 12.9 174.2, 69.8 71.2, 24.0 59.8, 67.5 171.5, 44.7 179.6, 20.3 20.5, 66.9 86.3, 8.3 90.4, 34.8 154.6, 73.3 85.3, 0.5 72.3, 7.7 123.5, 4.2 74.7, 40.5 22.7, 22.4

0.51 0.55 0.56 0.81 0.74 0.52 0.16 0.57 0.61 0.58 0.47 0.50 0.51 0.74 0.59 0.81 0.67 0.1 0.45 0.47 0.46 0.50 0.48 0.50 0.56 0.83 0.82 0.74 0.47 0.60 0.26 0.72 0.53 0.92 0.87 0.26 0.83 0.51 0.57 0.47 0.67 0.72 0.58 0.90 0.71 0.63 0.58 0.56 0.42 0.49 0.39 0.62 0.67 0.55 0.63 0.51 0.84 0.58

15.33 10.72 14.60 20.06 15.73 9.32 11.89 14.80 16.25 22.77 10.15 22.25 9.78 20.47 22.72 15.65 31.27e 38.33e 15.50 14.47 24.44 25.47 36.78e 11.53 46.56e 24.24 26.77 17.25 18.51 18.71 13.85 21.66 12.62 31.20e 17.20 10.31 23.48 35.26e 37.99e 18.81 21.14 35.80e 18.22 20.43 24.11 8.60 55.30e 23.38 18.17 28.77 25.33 49.23e 20.62 37.36e 11.63 15.19 19.95 20.43

38.82 39.17 38.76 38.95 39.56 39.02 39.19 39.00 39.21 39.38 39.72 39.57 39.45 39.83 39.36 39.24 39.35 39.49 39.65 39.28 39.48 39.28 39.52 39.83 39.81 38.37 38.62 38.36 38.52 38.38 38.45 38.76 38.69 38.56 38.55 38.89 38.35 38.83 38.81 38.97 38.79 38.33 38.93 38.55 38.75 38.99 39.05 39.12 39.22 39.45 39.57 39.72 39.84 39.46 38.11 38.03 38.10 38.64

a

Correspond to on-fault bins. Location is given by the centroid of the longitude and latitude of the events contained in each bin. S1, S2, and S3 are the directions of the maximum, intermediate, and least principal stresses. Azimuths were measured clockwise from north. d Form ratio r0 = (s1  s2)/(s1  s3) = 1  R, where s1, s2, and s3 are the absolute magnitudes of the principal stresses and R is the form ratio as defined by Angelier [1975]. f For this bin the maximum horizontal compression was taken to be S2 because S1 was dipping very steeply. e Correspond to tensors with a angular error larger than 35. b c

only, we also inverted the selected fault planes with FSA. The results of the three inversions are, in general, quite similar, and are displayed in Figure 2, which shows histograms of the different SH directions obtained from the three

inversions. It is clear from the narrow peaks of the histograms around zero that the orientation of SH does not differ greatly from one inversion to the other. This is also seen by comparing the azimuth of SH in Tables 3 and 4 in the

PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

ETG

13 - 5

Table 2. Results of Stress Inversions in the Bay Area Using A. Michael’s Program With the Selected Fault Planes Bin

Events Per Bin

BAn1 BAn2 BAn3 BAn4 BAn5a BAn6a BAn7 BAn8a BAn9 BAn10 BAn11 BAn12 BAn13 BAn14a BAn15 BAn16 BAn17 BAn18 BAc1 BAc2 BAc3a BAc4a BAc5 BAc6 BAc7a BAc8 BAc9 BAc10 BAc11 BAc12 BAc13 BAc14 BAs1 BAs2 BAs3 BAs4a BAs5a BAs6a BAs7 BAs8a BAs9a BAs10a BAs11a BAs12a BAs13a BAs14a BAs15a BAs16a BAs17a BAs18a BAs19a BAs20 BAs21a BAs22a BAs23 BAs24 BAs25a BAs26 BAs27 BAs28a BAs29 BAs30 BAs31 BAs32 BAs33 BAs34 BAs35 BAs36a BAs37a BAs38 BAs39a BAs40

23 10 7 5 15 31 12 38 9 6 48 22 7 17 10 20 13 7 14 9 7 29 28 5 97 16 15 22 34 22 66 7 14 16 14 35 34 36 13 10 58 61 57 18 10 7 29 85 152 187 35 15 141 93 19 53 85 37 86 103 7 8 15 13 19 16 14 60 106 10 420 33

Locationb Long, Lat

S1c Azimuth, Plunge

S2c Azimuth, Plunge

S3c Azimuth, Plunge

Form Ratio r0d

Mean Misfit (Angular Error)

122.57, 122.48, 122.34, 122.42, 122.30, 122.23, 122.19, 122.13, 122.21, 122.08, 121.99, 121.98, 122.05, 122.03, 121.98, 121.86, 121.93, 121.89, 122.29, 122.26, 122.09, 122.04, 122.09, 121.96, 121.89, 121.83, 121.97, 121.79, 121.95, 121.69, 121.76, 121.63, 122.23, 122.03, 121.90, 121.98, 121.98, 121.88, 121.80, 121.95, 121.86, 121.86, 121.73, 121.74, 121.82, 121.74, 121.71, 121.65, 121.67, 121.60, 121.63, 121.48, 121.60, 121.47, 121.41, 121.43, 121.46, 121.48, 121.33, 121.49, 121.37, 121.27, 121.08, 121.74, 121.63, 121.77, 121.54, 121.56, 121.77, 121.70, 121.67, 121.66,

15.0, 41.2 157.8, 33.7 28.8, 44.3 13.7, 20.7 6.5, 17.8 9.7, 15.7 18.6, 15.9 10.0, 10.6 19.7, 15.6 40.2, 21.3 30.6, 5.1 32.6, 17.5 161.4, 6.5 1.6, 28.4 20.9, 10.5 22.9, 12.6 18.4, 11.0 153.2, 1.4 138.4, 17.6 53.2, 9.2 145.7, 6.8 31.0, 3.9 36.0, 3.2 37.0, 34.4 15.8, 7.4 142.0, 1.0 154.9, 23.6 38.9, 2.8 24.9, 9.8 40.9, 14.6 152.5, 3.0 36.1, 5.5 145.3, 5.2 150.2, 19.1 152.4, 3.9 40.8, 19.4 18.5, 4.7 153.1, 13.5 6.6, 7.7 29.2, 12.1 169.8, 20.1 14.0, 16.7 174.8, 5.4 161.9, 22.6 20.9, 9.5 166.0, 19.8 6.7, 15.6 172.8, 5.3 13.8, 30.6 1.2, 1.4 2.2, 11.1 8.5, 15.8 13.2, 12.9 175.2, 1.5 150.2, 1.0 156.0, 39.9 154.1, 2.8 154.0, 6.2 165.4, 0.6 161.4, 8.3 28.3, 27.7 157.1, 19.5 3.1, 32.8 128.5, 4.5 40.6, 6.3 25.6, 3.3 44.6, 3.2 151.9, 5.2 168.5, 3.1 30.1, 6.9 9.9, 10.2 158.2, 13.8

174.7, 46.9 3.4, 54.8 167.7, 44.4 178.4, 68.6 176.0, 72.1 167.8, 74.2 168.5, 73.9 152.3, 76.7 151.9, 74.2 168.0, 66.1 178.4, 84.1 154.8, 59.4 41.6, 77.1 177.1, 61.5 176.3, 78.4 160.7, 73.2 127.7, 76.8 67.8, 88.1 103.4, 56.0 147.5, 25.2 54.2, 13.1 122.0, 14.3 126.7, 12.7 170.7, 45.3 118.9, 60.1 125.6, 65.2 5.8, 63.0 59.4, 71.4 167.6, 79.9 176.1, 71.8 92.9, 82.9 144.7, 73.4 53.3, 21.3 105.5, 35.2 108.5, 66.2 162.0, 55.7 92.0, 76.7 55.6, 28.6 119.6, 77.1 124.3, 22.5 78.3, 4.3 109.4, 17.5 65.7, 74.0 95.7, 27.0 69.7, 4.3 93.9, 25.7 169.8, 74.3 1.2, 84.6 132.9, 54.7 101.7, 83.9 116.3, 64.4 100.4, 49.0 158.7, 76.9 77.7, 78.5 115.4, 76.7 7.9, 48.9 36.3, 87.1 83.4, 78.4 60.9, 87.4 80.7, 72.5 141.0, 61.8 24.3, 62.4 156.2, 55.4 37.9, 9.4 134.4, 31.3 116.4, 14.8 141.2, 64.5 43.4, 73.9 93.3, 68.6 71.7, 59.5 163.9, 79.7 29.5, 68.5

84.0, 10.2 106.1, 8.8 69.4, 8.4 78.2, 5.1 83.8, 0.9 99.9, 0.6 71.9, 1.9 82.5, 7.9 110.2, 2.2 53.8, 10.3 59.5, 2.8 65.5, 24.2 107.2, 11.0 89.3, 1.8 69.9, 4.7 69.5, 10.8 109.7, 7.1 63.2, 1.2 38.8, 28.0 55.2, 62.9 97.1, 75.1 74.0, 75.1 68.1, 77.0 71.4, 24.7 78.2, 28.7 51.6, 24.7 109.4, 12.3 129.7, 18.3 65.4, 2.1 51.8, 10.5 62.2, 6.4 55.4, 15.6 111.4, 68.0 37.2, 48.4 60.8, 23.4 59.5, 26.9 109.4, 12.3 94.4, 57.8 97.9, 10.2 87.0, 64.0 23.3, 69.3 116.8, 65.3 93.5, 15.0 37.8, 53.5 175.9, 79.5 43.0, 56.5 96.9, 0.9 97.0, 0.9 113.4, 15.8 91.3, 5.9 92.4, 22.7 110.6, 36.6 102.6, 1.7 94.3, 11.4 60.0, 13.2 107.1, 8.0 64.1, 0.5 63.0, 9.6 104.4, 2.5 69.1, 15.3 120.5, 4.4 105.9, 18.6 99.3, 9.7 116.3, 79.5 59.4, 57.8 76.4, 74.8 46.9, 25.2 116.5, 15.1 77.4, 21.1 124.0, 29.5 100.0, 1.1 107.6, 16.1

0.08 0.21 0.62 0.59 0.55 0.30 0.60 0.62 0.43 0.56 0.60 0.66 0.32 0.42 0.45 0.62 0.89 0.36 0.82 0.53 0.83 0.70 0.78 0.52 0.58 0.71 0.57 0.68 0.53 0.62 0.42 0.53 0.82 0.92 0.91 0.77 0.85 0.48 0.51 0.84 0.76 0.48 0.73 0.87 0.74 0.60 0.59 0.61 0.37 0.60 0.62 0.94 0.46 0.50 0.45 0.37 0.47 0.56 0.51 0.47 0.45 0.37 0.36 0.57 0.75 0.90 0.67 0.54 0.46 0.69 0.53 0.62

18.18 18.78 17.59 0.93 11.32 16.77 8.50 12.05 7.88 7.61 16.88 21.85 10.16 11.61 9.18 31.37e 3.57 4.63 15.08 10.10 17.14 13.32 11.61 17.28 20.99 33.35e 19.09 10.62 14.78 17.87 34.55e 7.85 13.86 10.28 15.70 23.22 31.27e 29.12e 22.62 14.33 62.57e 31.30e 13.33 23.17 11.15 8.43 16.21 16.47 9.37 12.56 16.27 20.24 16.11 16.51 18.14 9.33 10.43 22.07e 22.07e 14.00 6.07 14.13 13.09 14.52 12.77 35.10e 14.86 17.68 14.06 11.26 10.23 17.82

37.76 37.62 37.85 38.10 37.93 37.85 37.90 37.74 38.02 37.91 37.87 37.83 37.95 37.97 37.91 38.05 38.15 38.15 37.07 37.38 37.20 37.19 37.31 37.28 37.53 37.55 37.73 37.66 37.81 37.61 37.82 37.70 36.97 37.09 36.96 37.13 37.18 37.04 36.95 37.15 37.06 37.11 36.97 37.01 37.11 37.04 37.00 36.92 36.93 36.87 36.96 36.86 36.97 36.91 36.88 36.94 36.97 37.02 36.88 37.06 36.97 36.91 36.98 37.25 37.19 37.38 37.11 37.15 37.42 37.32 37.29 37.29

ETG

13 - 6 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

Table 2. (continued) Bin

Events Per Bin

BAs41a BAs42 BAs43 BAs44 BAs45 BAs46 BAs47 BAs48

32 21 12 28 230 14 7 11

Locationb Long, Lat

S1c Azimuth, Plunge

S2c Azimuth, Plunge

S3c Azimuth, Plunge

Form Ratio r0d

Mean Misfit (Angular Error)

121.58, 121.75, 121.49, 121.61, 121.68, 121.72, 121.55, 121.27,

23.7, 22.7 127.9, 15.0 45.6, 3.0 38.3, 3.4 138.9, 2.2 138.3, 13.0 36.4, 15.2 20.8, 9.4

130.5, 65.0 73.6, 73.9 61.4, 79.8 70.4, 79.5 122.8, 75.0 95.6, 68.5 171.5, 72.9 69.4, 1.7

117.7, 9.7 36.4, 5.6 136.1, 9.7 128.8, 9.9 48.4, 14.8 44.4, 16.7 55.7, 7.6 169.6, 80.4

0.55 0.82 0.38 0.62 0.46 0.46 0.45 0.82

13.43 17.17 11.13 12.32 14.10 9.80 7.33 26.91e

37.19 37.43 37.20 37.39 37.50 37.48 37.48 37.54

a

Correspond to on-fault bins. Location is given by the centroid of the longitude and latitude of the events contained in each bin. c S1, S2, and S3 are the directions of the maximum, intermediate, and least principal stresses. Azimuths were measured clockwise from north. d Form ratio r0 = (s1  s2)/(s1  s3) = 1  R, where s1, s2, and s3 are the absolute magnitudes of the principal stresses and R is the form ratio as defined by Angelier [1975]. e Correspond to tensors with a angular error larger than 30. b

auxillary material1. On the basis of the similarity of the orientations obtained by three different inversions, we infer that the inversion procedure is robust with respect to the stress orientations obtained despite some degree of ambiguity in the selection of the fault planes. [15] Confidence limits are obtained with a bootstrap resampling procedure in which the original data in a group are randomly selected. The algorithm uses ‘‘selection with replacement’’ so each randomly selected set has the same number of data (focal mechanisms) as the original. To account for uncertainties in the selection of the fault plane from the two nodal planes, we use the option developed by Michael [1987b] of substituting the auxiliary plane for the fault plane for a specific fraction of the events. Each set is then inverted for its best stress tensor. The resulting distribution of tensors gives the confidence limits on the best tensor for the original data set; for example, the 80% of the resampled stress tensors that are closest to the best tensor for the entire data set delineate the 80% confidence limit on the best tensor. The fraction of events for which the auxiliary plane is substituted for the selected fault plane (in the bootstrapping inversions) was taken to be 30%, because, following the argument of Michael [1987a], it was deemed likely that 50% of the fault planes would be chosen correctly by random chance and at least another 20% would be chosen correctly by our procedure for selecting fault planes; to test the consequences if that assumption is optimistic, we used 50% instead of the 30% in several test cases; this resulted in only slightly larger confidence limits, which would not affect our ability to interpret the results. [16] Important point concerns the diversity of the fault plane solutions in each group. As mentioned in section 2, the segregation of the data in groups was made according to the fault geometry and the distribution of the seismicity, respecting as much as possible the distinction between on- and offfault events. The poles of the fault planes used in the inversions for each group are plotted on lower hemisphere stereonets on Figures 3 and 4 for northern California and the greater Bay Area, respectively. Most of the bins contain a rather diverse group of fault plane orientations. However, it 1 Auxillary Tables 3, 4, and 5 are available via Web browser or via Anonymous FTP from ftp://ftp.agu. org, directory ‘‘append’’ (Username = ‘‘anonymous’’, Password = ‘‘guest’’); subdirectories in the ftp site are arranged by paper number. Information on searching and submitting electronic supplements is found at http://www.agu.org/pubs/esupp_about.html.

is possible that the diversity of some of the on-fault bins could have been inadvertently increased by the fault plane selection procedure described above. Given that on-fault focal mechanisms tend to be an inherently less diverse group, a mistaken choice of the auxiliary plane by the selection procedure would increase the apparent diversity of the fault geometries. To check this effect, we tested another selection procedure on several on-fault bins; which consisted in selecting the nodal plane closet to the strike of the fault. As we found in the creeping section of the SAF [Provost and Houston, 2001], a significant amount of diversity remains in most of the on-fault bins even with this alternate selection procedure. This is likely due in part to complex internal fault plane solutions (see section 2). The orientations of the principal stresses given by inversion of the alternately selected bins lay within the original confidence limits, and the new confidence limits were very similar in size. [17] Additionally, some bins contain few events and the tensors obtained by inversions of these events could be poorly resolved. On Figures 5 and 7, however, the orientations of SH corresponding to these groups are similar to those of larger nearby groups of the same type (i.e., on- or off-fault). Moreover, the misfit obtained for the bins with few events using the different inversions, are not large (Tables 1 and 2, as well as the tables in the auxillary material). To verify our results on bins with few events, seismicity was regrouped so that bins with few data were combined together or regrouped with larger bins. Our original grouping has four bins west of the Maacama fault zone containing 5 to 14 events (see Figure 5). We recombined the two northernmost (2 and 5) and the two southernmost (3 and 4) bins and inverted these for the best tensors. The resulting orientations of the maximum horizontal compression, SH, correspond to an average of the initial ones. Similar results were obtained for other recombinations of events. The recombinations were the following: 11 and part of 12, 14 and part of 12, and 24 and 23 on the Garberville and the Lake Mountain-Bartlett Springs fault systems, and, n9 and n10, and s29 and s30 in the Bay Area (see Figures 5 and 7). Other instances of regrouped bins yielded tensors with SH similar to the original ones. These situations correspond to a few events incorporated into a much larger group, thus not strongly influencing the larger group (i.e., 43 and 46 in northern California and c5 and c6 and s10 and s14, in the Bay Area). Furthermore, most of these recombinations give larger misfit on the new tensors

PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

ETG

13 - 7

[18 ] In summary, the stress orientations have been robustly determined for the most part by our methods, but the accuracy of the confidence limits is less certain. In some bins with few events or low diversity or where much of the diversity is produced by errors in the fault plane solutions, confidence limits may be somewhat larger than our method estimated it. However, due to the spatial consistency and redundancy of a large number of bins, as well as fairly regional differences, this uncertainty in some of the confidence limits does not materially affect our subsequent interpretation of the stress orientations.

4. Results 4.1. Orientations of the Maximum Horizontal Compression [19] The results are shown on maps in Figures 5 and 7, for northern California and the San Francisco Bay Area respectively. Bars show the orientation of the maximum horizontal compression, SH, obtained for each bin using Michael’s [1984] method on the selected fault planes. Light bars correspond to on-fault bins and dark ones correspond to off-fault bins. The plate boundary in the northern region is much broader, with a lower level of seismicity, leading to a smaller number of bins in the north. [20] On a broad scale, in northern California, the orientation of the maximum horizontal compression averages N26E, making an approximate angle with the regional strike of the SAF system of 55. On an intermediate scale, the angle between SH and the strike of major faults ranges from 40 to 70, systematically increasing from north to south. This phenomenon is due both to a slight counterclockwise rotation of the fault trends and a clockwise rotation of SH as one goes toward the south. On a more refined scale, some differences appear geographically, as the fault system encompasses a larger region and as fault segments interact with each other. In northern California there is not a large difference between the orientation of the on-fault and off-fault SH. Only in the north of that region (between 39.5 and 40 latitude, around the Garberville fault) is a rotation of the orientation of SH noticable, from outside of the fault zone to the on-fault group. [21] In the greater Bay Area, the orientations of the maximum horizontal compression are more complex, but still show two features (see Figure 7). Parts of the Calaveras fault and the Hayward fault show SH rotations to smaller angles with the fault as one gets closer to the fault zone similar to observations made in the creeping zone [Provost

than on the original ones. These observations lead us to think that recombination of bins is not necessarily beneficial to the analysis, as we lose spatial coverage and detail in the stress orientations, and as the best tensors found are not of better quality than the original ones.

Figure 2. (opposite) Histograms of the angular difference in degrees of the azimuth of SH obtained with three inversion procedures. (a) Angular difference between SH obtained with Ce´le´rier’s program using both nodal planes and that obtained using the selected fault planes. (b) Angular difference between SH obtained with Ce´le´rier’s program using both nodal planes and that obtained with Michael’s program using the selected fault planes. (c) Angular difference between SH obtained with Ce´le´rier’s program using the selected fault planes and that obtained with Michael’s program using the selected fault planes. The three histograms peak at zero and are relatively narrow showing the closeness of the tensors obtained with the three different procedures.

ETG

13 - 8 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

Figure 3. Lower hemisphere stereonet projections of the poles of the fault plane, chosen as described in section 3, for each .earthquake in each bin, in the northern California region. The names of the bins correspond to the numbering of the bins plotted on the map in Figure 5. The wide range of poles demonstrates the diversity of the fault planes used in the inversions.

and Houston, 2001]. The segment of the SAF, where the Loma Prieta event took place, shows a spatial variation of the orientation of the stress from south to north of the main shock surface rupture [Gephart, 1990; Michael et al., 1990]. 4.2. Orientations of the Principal Stresses [22] For a more detailed display of the results, Figures 6 and 8 show, for northern California and the San Francisco Bay Area, respectively, the best tensors obtained with both programs, as well as confidence limits on the tensors obtained with Michael’s [1987a] method, plotted on lower hemisphere stereonet projections. The best tensors obtained from the inversions with one fault plane per event using the method of Michael [1987a] are generally quite similar to the best tensors obtained with both nodal planes using the method of Ce´le´rier [1995] (see also Figure 2, as well as tables in the auxillary material). In these figures, we highlighted in light gray the stereonets that correspond to poorly constrained tensors with a broad distribution of the 80% confidence limit. They constitute about a quarter of the ensemble of the tensors. Three quarters of these correspond to bins in northern California, where the seismicity was sparse and the selection criteria for the focal mechanisms was less stringent compared to the ones in the Bay Area. However, in general the orientations are similar to those of nearby better constrained tensors. The confidence limits thus obtained could be slightly larger considering the

ambiguity in the chosen fault plane and the lack of diversity in the focal mechanisms in some cases. [23] Tables 1 and 2 list various parameters concerning the data and the tensors, for northern California and the Bay Area respectively, using Michael’s inversion code. The following information is given for each group of data: the number of events per bin, the location of the bars on the map (longitude and latitude), the orientations of the principal stresses, S1, S2 and S3, the form ratio, r0, between their magnitudes (r0 = (s1  s2)/(s1  s3)), and the mean misfit for the best tensor. 4.3. Ratio of Magnitudes of the Principal Stresses [24] Variations in the form ratio, r0, are shown in Figure 9, versus the distance along the SAF system and the distance perpendicular to it, for both on- and off-fault bins (solid and open circles, respectively). Here we took as a reference fault line the Maacama fault, continuing south onto the Hayward fault and ending on the San Andreas Fault in the creeping zone. The distance is measured along the reference line from the point on the Maacama fault, marked by a star on the map in Figure 10. There is a broad distribution of the r0 values along the strike of the reference line, with most between 0.4 and 0.9. The higher values of r0 (0.8 to 1.0) indicate that s2 and s3 are close in magnitude; the tectonic regime is thus close to switching from strike slip to thrust, because, in most cases, one of the principal

PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

ETG

13 - 9

Figure 4. Lower hemisphere stereonet projections of the poles of the fault planes, chosen as described in section 3, for each earthquake in each bin, in the greater Bay Area region. The names of the bins correspond to the numbering of the bins plotted on the map in Figure 7. The wide range of poles demonstrates the diversity of the fault planes used in the inversions.

axes is near vertical. Figures 6 and 8 indicate that the majority of the bins are in a strike-slip regime. [25] There is no clear variation between on- and off-fault r0 values along the fault (Figure 9a), whereas across the fault, off-fault r0 seems to average to 0.5 in the far field, about 50 to 100 km from major faults (Figure 9b). These observations may imply that the local tectonic regime is more uniform in the far field than closer to the faults where interaction between en e´chelon fault segments may perturb it.

5. Discussion 5.1. Summary of Results [26] The stress tensors obtained in the northern portion of our study zone show angles of 40 to 70 angles to the major fault trends and little systematic difference between the orientation of maximum compressive stress for groups of mostly on-fault compared to mostly off-faults events. These results contrast with those found in the creeping section [Provost and Houston, 2001], and suggest stronger, less well developed faults in this area, consistent with the en e´chelon, discontinuous, east dipping segments east of the SAF in this region, noted by Castillo and Ellsworth [1993]. This is discussed further in section 5.3. [27] The results obtained in the Bay Area are more complex, and include some SH orientations at high angles to major faults. However, in general they show an inter-

mediate state between that of the creeping zone [Provost and Houston, 2001] and the regions farther north. The complexity of the orientation of SH in this area is probably due to the fault distribution and intersections in the area, as the SAF branches into the Calaveras fault and then the Hayward fault. [28] The stress orientations found in this analysis are combined with those found around the creeping section of the SAF by [Provost and Houston, 2001]. Figure 10 shows the orientations of SH obtained from central to northern California. A difference in the pattern of the stress orientations is clearly visible from the creeping section where SH is almost perpendicular to the fault trend, to the GarbervilleLake Mountain fault zone in the north where SH lies closer to the trends of the major faults. Figure 11 compares the angles between SH and the major trends of faults on histograms for the three regions studied for on- and off-fault bins. For onfault bins, SH makes angles of 40 to 60 with major fault trends for all three regions considered. For the off-fault bins, SH makes smaller angles to the major fault trends, from south to north along the plate boundary fault system. The off-fault stress orientations are about 80 from the SAF in central California, but 55 from the major fault trends in northern California. Figure 12 shows these orientations versus the distance along the reference line defined previously (from the Maacama fault in the north to the creeping zone in central California, see caption of Figure 10) for bins within 30 to 40 km of this line. Least squares fits show the increasing

ETG

13 - 10 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

Figure 5. Orientation of the maximum horizontal compression in Northern California, in the vicinity of the Maacama-Healdberg-Rogers Creek, and Bartlett Springs-Green Valley fault systems. Light gray bars represent stress orientation for events occurring on-fault; dark gray ones represent the stress for events occurring off-fault. The bars are plotted at the centroid latitude and longitude of the events contained in each bin. Numbers next to the bars correspond to the names of each bin. In this inversion we used 2800 events with M  1.5 that occurred between January 1969 and October 2000. Faults are from Jennings [1992]. angle between off-fault SH and major fault trends, from northern California to the creeping zone. As in the creeping section [Provost and Houston, 2001], most (three quarters) of the tensors correspond to a strike-slip regime while a smaller proportion correspond to a thrust regime. 5.2. Implications for Variations in Frictional Strength [29] The stress orientations obtained from inverting groups of fault plane solutions can provide some constraints on the strength of the fault system in each area studied. These observations taken alone can only constrain

the upper bounds of the coefficient of friction on the major fault segments. The variation of the orientations of SH from the creeping zone to northern California (Figures 10, 11, and 12) implies that this upper bound varies with latitude in central and northern California, if the same pore pressure is assumed on all major faults. In the creeping zone, Provost and Houston [2001] found SH oriented at high angles (65 to 85) to the SAF adjacent to the fault and away from it. These observations imply a low upper bound on the frictional strength (m  0.2 according to Lachenbruch and McGarr [1990, figure 10.10]). In northern California we

Figure 6. (opposite) Lower hemisphere stereonet projection of best tensors and the 80% confidence limit for each bin in northern California. Open symbols represent the best tensor found with Ce´le´rier’s [1995] program using both nodal planes, and solid symbols represent the one found with the method of [Michael, 1987a] using the selected fault planes. The orientations of the three principal stresses, S1, S2, and S3, are represented by circles, diamonds, and triangles, respectively. The distribution of dots shows the 80% confidence limit on S1 for each tensor (each dot represents the result of one bootstrap-resampling inversion). As plotted, these tensors do not necessarily follow the right-hand rule, as only vectors on the lower hemisphere could be shown for clarity. Shaded stereonet projections correspond to poorly constrained tensors with broad 80% confidence limits.

PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

ETG

13 - 11

ETG

13 - 12 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

Figure 7. Orientation of the maximum horizontal compression in the Greater Bay Area, in the vicinity of the San Andreas, and the Calaveras-Hayward fault systems. Light gray bars represent stress orientation for events occurring on-fault; dark gray ones represent the stress for events occurring off-fault. The bars are plotted at the centroid latitude and longitude of the events contained in each bin. Numbers next to the bars correspond to the names of each bin. In this inversion we used 3200 events with M  1.5 that occurred between January 1969 and October 2000. Faults are from Jennings [1992]. find that SH is oriented at smaller angles to the trend of major faults (40 to 70) and there is no major difference between the off-fault and the on-fault orientations. These observations alone imply an upper bound on the frictional strength of the major strike-slip faults in northern California of higher value (m ffi 0.6) without excluding the possibility that they could be weak. On the other hand, because off-fault stress orientations cannot distinguish between a low coefficient of friction and high pore pressure (as by Rice [1992]) on the fault, and because our on-fault orientations are less well constrained (see section 3), our results permit an alternative explanation in which m is everywhere near 0.6 –0.8 but pore pressure is higher on the SAF in central California than on the faults of the SAF system in northern California. That is, the severe misorientation of the SAF in central California could be due to a low coefficient of friction or high pore pressure or a combination of these factors. [30] In order to draw further implications from these stress orientations it would be necessary to make some strong simplifying assumptions about the state of stress in California. One possibility is that SAF-parallel thrust faults could be regulating the normal tractions on planes parallel to the SAF, so that they are equal in northern and central California. If that is assumed, as well as the same normal

traction across planes perpendicular to the SAF, faultparallel velocity component, deviatoric stress, and pore pressure, then one can infer that variations in the orientations of the principal stresses closer to the fault must be due to variations in effective frictional strength of the fault system. Specifically, a simple Morh-Coulomb construction shows that the change in orientations of off-fault SH relative to the trends of major faults implies an increase in the effective frictional strength of the SAF system toward the north. The increase in m is a factor of about 2.5, given typical SH orientations at 80 to the SAF in central California and at 55 to the SAF system in the north. The geological observations discussed in section 5.3, regarding contrasts between central and northern California in the geometrical configuration of the strikeslip fault system, suggest that such regional variations in the effective frictional strength exist. 5.3. Relation to Tectonic History [31] The systematic decrease toward the north in the angle between SH and the major faults (Figures 10, 11, and 12) may be related to the relative youth and multiplestranded nature of the northernmost portions of the SAF system compared to the central SAF. The plate boundary system appears progressively stronger as one goes farther

ETG

Figure 8. Lower hemisphere stereonet projection of best tensors and the 80% confidence limit for each bin in the Greater Bay Area. The arrangement is similar to that in Figure 6.

PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

13 - 13

ETG

13 - 14 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

north. The younger, discontinuous less developed fault strands in the north are believed to be thrust fault structures associated with subduction of the Farallon/Gorda plates, now reactivated in a mostly strike-slip sense [Castillo and Ellsworth, 1993]. These structures dip northeastward and get steeper toward the south becoming nearly vertical in the Bay Area (Hayward-Calaveras fault zone). Near-vertical planes are more favorably oriented for strike-slip than more shallowly dipping ones. Thus the plate boundary appears to evolve, with increasing slip, toward a geometry, and possibly a rheology, that more readily accommodates strike-slip plate motions. Although northern California may have a greater frictional strength than the creeping zone, even SH as low as 40 to the fault still does not exhibit the optimal angle near 30 predicted for brittle failure under the Byerlee friction law [Byerlee, 1978]. [32] On the basis of their 3-D thermomechanical modeling of northern California, Furlong et al. [1989] concluded that the lithospheric boundary between the Pacific and North American plates, as well as the crustal structure in the area, must evolve as the Mendocino triple junction travels northward. They inferred that the plate boundary will tend to migrate eastward to the Maacama and Bartlett Springs fault zones. [33] The stress states found in northern and central California in this analysis show differences as mentioned above, but also may contrast with the ones in southern California. Hardebeck and Hauksson [1999] determined stress orientations by inverting focal mechanisms of southern California earthquakes. Across some of their transects, most notably one across the Big Bend of the SAF near Fort Tejon, they found a broad zone (20 – 40 km wide) with SH rotated to lower angles as the SAF is approached, which they interpreted as a 20– 40 km wide weak zone surrounding the SAF. Scholz [2000] found that interpretation untenable because such a wide weak zone would rapidly extrude vertically, which is not observed. He reinterpreted Hardebeck and Hauksson’s observation of a broad zone of rotated SH around the Big Bend using a transpressional plate boundary model. By making the lower crust and upper mantle very weak, this model essentially requires the upper crust to be strong. Thus Scholz’s [2000] interpretation has a mechanically strong SAF in the Big Bend region and by extension elsewhere in southern California, although the rotated zones of SH are somewhat equivocal elsewhere in southern California away from Big Bend. However, Townend and Zoback [2001] reanalyzed the data used by Hardebeck and Hauksson [1999] and found that the rotation of SH may be an artifact of the gridding scheme, which involved projecting and grouping earthquakes over large distances. Townend and Zoback [2001] used a more equant gridding scheme. In their interpretation, no systematic rotation of SH near the SAF is found; they contend that SH lies at and average of 60 to 65 to the strike of the SAF in the near field (within 5 km) for at least 400 km along the fault in southern California. [34] Slip partitioning and slip rates estimated along the fault system from southern California to the Bay Area, also indicate variations in the strength of the SAF along its strike [Jones and Wesnousky, 1992]. Wright [2000] segments the SAF system from north to south, according to slip rates, fault types and spacing, and the depth and distribution of

Figure 9. Variations in the ratio r0 (= (s1  s2)/(s1  s3)) (a) along the San Andreas fault system and (b) away from it. As a reference fault line, along which to measure distance, we took a straight line from the Maacama fault, through the Rogers Creek and the Hayward faults, to the SAF-Calaveras fault junction, which then bends to follow the strike of the San Andreas fault in the creeping section. The distance is taken from the reference point on the Maacama fault zone, marked by a dark star on the map in Figure 10. Open circles correspond to values for off-fault groups and solid circles correspond to on-fault ones.

seismicity. He identified five regions: the northern region, from the Mendocino triple junction to San Pablo Bay (‘‘Juvenile’’), the San Francisco Bay region (‘‘Adolescent’’), the SAF in central California (‘‘Mature’’), the Big Bend region (‘‘Confounded’’), and the southern region from the Mojave to the Salton trough (‘‘Expanding’’). It is interesting that his segmentation of the SAF system is quite consistent with the variations in stress orientations seen by the present study, Provost and Houston [2001], and Hardebeck and Hauksson [1999]. [35] In summary, these results together suggest that the San Andreas plate boundary is a rather complex fault system with spatially varying frictional strength that evolves over geological time as slip accumulates and as bends in the major faults develop. 5.4. Comparison With Other Studies [36] The orientations of SH in bins n1 and n2 on and just offshore of the northern San Francisco Peninsula are somewhat anomalous compared to those farther south on the SAF (Figure 7). Seismicity in this region was relocated and analyzed by Zoback et al. [1999], who concluded that an en e´chelon step in the trace of SAF just offshore of the

PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

ETG

13 - 15

Figure 10. Orientations of the maximum horizontal compression in central and northern California, from the Parkfield region to near the Mendocino triple junction. The results from our previous study [Provost and Houston, 2001] are combined with the ones found in this analysis to allow a better view of the variation of the state of stress around the San Andreas plate boundary system. Faults are from Jennings [1992]. Seismicity has been omitted here for better clarity. Light gray bars represent the stress orientation for known on-fault seismicity; dark gray bars represent the rest of the data. The solid star (40 latitude) corresponds to the reference point used to plot Figures 9 and 12. northern SF Peninsula creates an extensional stress regime. The stress regimes we determined in bins n1, n2, c2, and c5 are consistent with the focal mechanisms reported by Zoback et al. [1999]. The direction of S3 in bins n1 and n2 (Figure 8) is similar to the typical T axes of the focal mechanisms that Zoback et al. determined in this area [see Zoback et al., 1999, Figure 4]. The directions of S1 and S2 in bins n1 and n2 are not well-constrained and the roles of s1 and s2 are reversed, consistent with the low r0 values (Figure 8 and Table 2, as well as Table 5 in the auxillary material). This switch from a normal regime in bin n1 to a strike-slip regime in bin n2 is consistent with the preponderance of normal mechanisms offshore and strike-slip

mechanisms onshore seen in Plate 5 of Zoback et al. [1999]. Interestingly, the stress tensor for bin n3 is similar to that for bin n2 (Figure 8). [37] Townend and Zoback [2001] inverted fault plane solutions in a relatively large box (about 15 km by 30 km) along the SAF on the southern San Francisco Peninsula. Our stress orientations in bins c2, c3, c4, c5, and s5 are consistent with the result they obtained. 5.5. Implications for the Creeping Segment of the Hayward Fault [38] The seismicity on the Calaveras fault corresponds in large part to aftershocks of the Morgan Hill earthquake

ETG

13 - 16 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

Figure 11. Histograms of orientations of SH relative to fault trends for (left) on-fault events and (right) off-fault events, for the three studied regions: the creeping zone in central California, the greater Bay Area, and northern California. Mean values of the angles are given for each case.

of May 1984. The moderate angles between the maximum horizontal compression and the Hayward fault (bins n5, n6, n7, and n8), where some creep has been measured [e.g., Burgmann et al., 1997], contrasts with the high angles adjacent to the SAF in the creeping zone [Provost and Houston, 2001]. This suggests that the Hayward fault is not creeping as freely as the creeping section of the SAF, although the relatively short length of the creeping Hayward segment suggests that edge effects could make adjacent stress orientations more difficult to interpret. 5.6. Anomalous Stress Orientations Near Sutter Buttes [39] One of the most enigmatic features we find involves anomalous SH orientations in the northeast sector of our study zone. To the east of the fault system in northern California (around 39.5 latitude and – 122 longitude) three groups (bins 48, 49, and 50, Figure 5) have SH oriented roughly east-west, which is nearly perpendicular to typical values of SH for the entire region.

[40] Figure 13 shows focal mechanisms of events in these bins on a map of the area and their depth distribution on a cross section along latitude. The P axes of the fault plane solutions have, in general, an orientation close to N-S in the north, and close to E-W in the southern part, showing the origin of the difference in the stress tensors. As shown by the cross section there is no major difference in the depth range between these two groups. Except for bin 54 which contains shallow events (1 – 12 km), the seismicity lies mostly between 12 and 25 km depth. So the large difference in the orientations of the maximum horizontal compression occurs for only some of the anomalously deep events. [41] Bins 50 and 51 are particularly intriguing in that they show a very different orientation of SH even though they are located less than 20 km from each other. When combined in one group, the resulting tensor gives a SH oriented N40E, an intermediate azimuth between that of the two bins taken separately. The angular error on this resulting stress tensor is very large (56), indicating that the focal mechanisms

PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

contained in this recombined group can not be activated by the same stress tensor. This justifies segregating the data into the original two bins 50 and 51. [42] These stress orientations appear to represent the state of stress in the lower crust in this particular area, as the events’ depths concentrate between 15 and 28 km, in contrast to the shallow seismicity on the major strike-slip fault strands to the west. These anomalous orientations could be caused by subduction-related volcanism that has persisted after the Mendocino Triple Junction has passed, or by the presence of detached fragments of the Gorda slab [Benz et al., 1992]. An example of these volcanic processes is given by the presence of the volcanic Sutter Buttes, east of the fault system (Figure 5) near the anomalous stress orientations. The anomalous orientations could also be due to asthenospheric flow around of the edge of the subducting Gorda plate, locally disrupting the state of stress in the lower crust. Although the position of this slab edge is uncertain, several studies conclude that the northward migration of the slab stalled 4 Ma ago, with the edge at 39.5 [Wilson, 1989; Atwater and Stock, 1998] just north of Sutter Buttes in the vicinity of the anomalous stress orientations.

6. Conclusions [43] Systematic analysis of stress orientations from inversions of earthquake focal mechanisms reveals spatial variations in the mechanical behavior of a major plate boundary system. The orientations of maximum horizontal compression surrounding the San Andreas plate boundary system from central to northern California are summarized in Figures 10, 11, and 12. In the north, the maximum horizontal compression makes an average angle of 55 with major fault trends, and there is no clear difference between on- and off-fault stress orientations. These results

Figure 12. Orientations of the maximum horizontal compression relative to major fault trends versus distance along the fault. The reference line and reference point from which the distance is estimated are the same ones that of Figure 9. Open circles correspond to on-fault bins and solid squares correspond to off-fault bins. The dashed and the straight lines are linear fits using the least squares method for the on-fault and the off-fault groups, respectively.

ETG

13 - 17

Figure 13. Details about events contained in seven bins to the northeast of the northern region. (a) Cross section along the latitudes. The numbers next to the symbols label each bin, and the key is as follows: solid circles, 48; open circles, 49; solid squares, 51; open squares, 50; solid diamonds, 52; open diamonds, 53; and open triangles, 54. (b) Zoomed map of the focal mechanisms contained in these bins, with their corresponding bin number. (c) Lower hemisphere stereonet projections of the P and T axes of events contained in bins 48, 49, and 50 and bins 51, 52, and 53. contrast with the ones found in the creeping zone [Provost and Houston, 2001]. Stress orientations in the greater Bay Area appear to be in an intermediate stage between central and northern California. By themselves, these observations provide an upper bound (that increases with latitude) on the effective frictional strength of the fault system; the SAF system could be as weak in northern California as it must be in central California [Provost and Houston, 2001]. However, the SAF system has evolved, following the northward passage of the Mendocino triple junction, from a young multiple-stranded plate boundary in the north, to a mature strike-slip plate boundary in the creeping zone where more total slip has occurred. On the basis of our results, together with differences in fault continuity, dip and total slip between northern and central California, it appears most likely that the effective frictional strength and mechanical behavior of the SAF system varies significantly along strike, with different characteristics in northern California, the Bay Area, the creeping segment, and southern California. In many tectonic analyses it may prove important to take this possible variation in frictional strength into account. [44] Acknowledgments. We thank Bernard Ce´le´rier and Andy Michael for providing their stress inversion programs, Peter Bird for insightful comments about the interpretation of our data, and Andy Michael, Bill Ellsworth, and Fred Pollitz for suggestions about the anomalous stress orientations in northern California. We also thank the referees, Jeanne Hardebeck and John Townend, and the anonymous AE for many thoughtful comments. The data used in this study were provided by

ETG

13 - 18 PROVOST AND HOUSTON: STRENGTH OF THE SAN ANDREAS FAULT SYSTEM

the NCEDC. Research was supported by the U.S. Geological Survey (USGS), Department of the Interior, under USGS award 01-HQ-GR-0120.

References Angelier, J., Sur l’analyse de mesures recueillis dans des sites failles: L’utilite´ d’une confrontation entre les me´thodes dynamiques et cine´matiques, C. R. Acad. Sci., 281, 1805 – 1808, 1975. Angelier, J., and P. Mechler, Sur une me´thode graphique de recherche des contraintes principales e´galement utilisable en tectonique et en sismologie: La me´thode des die`dres droits, Bull. Soc. Geol. Fr., 19(6), 1309 – 1318, 1977. Atwater, T., and J. Stock, Pacific-North America plate tectonics of the Neogene southwestern United States: An update, Int. Geol. Rev., 40(5), 375 – 402, 1998. Benz, H. M., G. Zandt, and D. H. Oppenheimer, Lithospheric structures of northern California from teleseismic images of the upper mantle, J. Geophys. Res., 97, 4791 – 4807, 1992. Bird, P., Lithosphere dynamics and continental deformation, U.S. Natl. Rep. Int. Union Geod. Geophys., 1991 – 1994, Rev. Geophys., 33, 379 – 383, 1995. Bird, P., and X. Kong, Computer simulations of California tectonics confirm very low strength of major faults, Geol. Soc. Am. Bull., 106, 159 – 174, 1994. Brown, R. D. J., Quarternary deformation, in The San Andreas Fault System, California, edited by R. E. Wallace, U.S. Geol. Surv. Prof. Pap., 1515, 83 – 113, 1990. Brune, J. N., T. L. Henyey, and R. F. Roy, Heat flow, stress, and rate of slip along the San Andreas fault, California, J. Geophys. Res., 74, 3821 – 3827, 1969. Burgmann, R., P. Segall, M. Lisowski, and J. Svarc, Postseismic strain following the 1989 Loma Prieta earthquake from GPS and leveling measurements, J. Geophys. Res., 102, 4933 – 4955, 1997. Byerlee, J. D., Friction of rocks, Pure Appl. Geophys., 116, 615 – 626, 1978. Carey, E., and B. Brunier, Analyse theorique et numerique d’un modele mecanique elementaire applique a l’etude d’une population de failles, C. R. Acad. Sci., 279, 891 – 894, 1974. Castillo, D. A., and W. L. Ellsworth, Seismotectonics of the San Andreas fault system between Point Arena and Cape Mendocino in northern California: Implications for the development and evolution of a young transform, J. Geophys. Res., 98, 6543 – 6560, 1993. Ce´le´rier, B., Tectonic regime and slip orientation of reactivated faults, Geophys. J. Int., 121, 143 – 161, 1995. Che´ry, J., M. D. Zoback, and R. Hassani, An integrated mechanical model of the San Andreas fault in central and northen California, J. Geophys. Res., 106, 22,051 – 22,066, 2001. Etchecopar, A., G. Vasseur, and M. Daignieres, An inverse problem in microtectonics for the determination of stress tensors from fault striation analysis, J. Struct. Geol., 3(1), 51 – 65, 1981. Furlong, K. P., W. D. Hugo, and G. Zandt, Geometry and evolution of the San Andreas fault zone in northern California, J. Geophys. Res., 94, 3100 – 3110, 1989. Gephart, J. W., Stress and the direction of slip on fault planes, Tectonics, 9(4), 845 – 858, 1990. Gephart, J. W., and D. W. Forsyth, An improved method for determining the regional stress tensor using earthquake focal mechanism data: Application to the San Fernando earthquake sequence, J. Geophys. Res., 89, 9305 – 9320, 1984. Hardebeck, J. L., and E. Hauksson, Role of fluids in faulting inferred from stress field signatures, Science, 285, 236 – 239, 1999. Jennings, C. W., Preliminary fault activity map of California, Calif. Div. of Mines and Geol., Sacramento, 1992.

Jones, C. H., and S. G. Wesnousky, Variations in strength and slip rate along the San Andreas fault system, Science, 256, 83 – 86, 1992. Lachenbruch, A. H., and A. McGarr, Stress and heat flow, in The San Andreas Fault System, California, edited by R. E. Wallace, U.S. Geol. Surv. Prof. Pap., 1515, 261 – 277, 1990. Lachenbruch, A. H., and J. H. Sass, Heat flow and energetics of the San Andreas fault zone, J. Geophys. Res., 85, 6185 – 6222, 1980. Michael, A. J., Determination of stress from slip data: Faults and folds, J. Geophys. Res., 89, 11,517 – 11,526, 1984. Michael, A. J., Use of focal mechanisms to determine stress: A control study, J. Geophys. Res., 92, 357 – 368, 1987a. Michael, A. J., Stress rotation during the Coalinga aftershock sequence, J. Geophys. Res., 92, 7963 – 7979, 1987b. Michael, A. J., W. Ellsworth, and D. Oppenheimer, Coseismic stress changes induced by the 1989 Loma Prieta, California earthquake, Geophys. Res. Lett., 17, 1441 – 1444, 1990. Mount, V. S., and J. Suppe, State of stress near the San Andreas fault: Implications for wrench tectonics, Geology, 15, 1143 – 1146, 1987. Provost, A.-S., and H. Houston, Orientation of the stress field surrounding the creeping section of the San Andreas fault: Evidence for a narrow mechanically weak fault zone, J. Geophys. Res., 106, 11,373 – 11,386, 2001. Rice, J. R., Fault stress states, pore pressure distributions, and the weakness of the San Andreas fault, in Fault Mechanics and Transport Properties of Rock, Int. Geophys. Ser., vol. 51, edited by B. Evans and T.-F. Wong, pp. 475 – 503, Academic, San Diego, Calif., 1992. Scholz, C. H., Evidence for a strong San Andreas fault, Geology, 28, 163 – 166, 2000. Scholz, C. H., and T. C. Hanks, The strength of the San Andreas fault: A discussion, in Rheology and Deformation of the Lithosphere at Continental Margins, edited by G. D. Karner et al., Columbia Univ. Press., in press, 2003. Townend, J., and M. D. Zoback, Implications of earthquake focal mechanisms for the frictional strength of the San Andreas fault system, in The Nature and Tectonic Significance of Fault Zone Weakening, edited by R. E. Holdsworth et al, Geol. Soc. Spec. Publ., 186, 13 – 21, 2001. Walter, S., and D. Dzurisin, The September 1988 earthquake swarm at Medecin Lake Volcano, northern California, Eos Trans. AGU, 70, 1189, 1989. Williams, C. F., and T. N. Narasimhan, Hydrogeologic constraints on heat flow along the San Andreas fault: A testing hypotheses, Earth Planet. Sci. Lett., 92(2), 131 – 143, 1989. Wilson, D. S., Deformation of the so-called Gorda plate, J. Geophys. Res., 94, 3065 – 3075, 1989. Wright, T. L., The evolving San Andreas fault system: Nascent tectonic secession of California?, in Proceedings of 3rd Conference on the Tectonic Problems of the San Andreas Fault System, edited by G. Bokelmann and R. L. Kovach, et al., Stanford University Publications, Stanford Univ., Stanford, Calif., 2000. Zoback, M. D., et al., New evidence on the state of stress of the San Andreas fault system, Science, 238, 1105 – 1111, 1987. Zoback, M. L., R. C. Jachens, and J. A. Olson, Abrupt along-strike change in tectonic style: San Andreas fault zone, San Francisco Peninsula, J. Geophys. Res., 104, 10,719 – 10,742, 1999. 

H. Houston, Department of Earth and Space Sciences, University of California, Los Angeles, 3806 Geology Building, Box 951567, Los Angeles, CA 90095-1567, USA. ([email protected]) A.-S. Provost, Laboratoire Dynamique de la Lithosphere, Universite´ Montpellier 2, cc060 Place E. Bataillon, F-34095 Montpellier cedex 5, France. ([email protected])