FOCAL MECHANISMS AND TRANSFORM-BOUNDARY KINEMATICS

see http://plate-tectonic.narod.ru/califcrustseismphotoalbum.html

Primary considerations in the selection of these events were (1) size -larger events were chosen where available, because they represent large-scale processes along major boundaries; (2) date of occurrence-the quality of data for focal-mechanism determinations improved significantly during the mid-1970''s; and (3) location-some larger events were omitted because they were redundant in terms of mechanism and location, and some smaller events were included because they occurred in regions of significant seismicity where no larger events were available. Most focal mechanisms were determined from first arrivals at stations in the northern and southern California seismic networks. Fault-plane solutions for the few large earthquakes on the list before the mid-1970''s were supported by observations from stations outside the California networks
Since the mid-1970''s, focal mechanisms have been determined for only a fraction of the events for which adequate local first-motion data were available. Therefore, in addition to the three considerations listed above, there was a fourth, the interests of the investigators who analyzed the data. These interests included topical studies of large earthquakes and aftershock sequences, analyses of regional traveltimes on the basis of M>4 earthquakes, and a special study of the focal mechanisms of earthquakes on or near the San Andreas fault in southern California (Jones, 1988).

Focal mechanisms discussed in the first two subsections below are for earthquakes in the contiguous Coast Ranges-Transverse/Peninsular Ranges-Mojave Desert region associated with the principal seismic expression of the San Andreas fault system, where the seismic networks are best developed. Outside that region, except for the Cape Mendocino area and the vicinity of Long Valley caldera, the few well-determined focal mechanisms that are available provide only limited information on tectonic processes.
STRIKE-SLIP KINEMATICS OF THE SAN ANDREAS FAULT SYSTEM
Most moderate and large (Mб▌3) earthquakes along the San Andreas fault and its major branches produce nearly pure right-lateral displacements along near-vertical planes that closely follow the surface traces of the respective fault segments. This relatively simple kinematic pattern holds for the great earthquakes that rupture "locked" sections of the fault every few hundred years (Sieh, 1981), as well as for nearly all the moderate earthquakes that rupture limited patches along persistently active segments of the fault system (Ellsworth and others, 1982; Jones, 1988). Displacements associated with these earthquakes dominate the kinematic pattern along the transform boundary in California. DeMets and others (1987) and Minster and Jordan (1987), for example, argued that the cumulative displacement from earthquakes along the faults in the San Andreas system, together with the contribution from aseismic slip along its creeping segments, accounts for 60 to 70 percent of the total displacement between the Pacific and North American plates.
. In central California, such mechanisms mark the San Andreas fault itself from San Francisco to Cholame (events 26, 36, 38, 45, 46), the Calaveras-Greenville fault (events 23, 28-34) and the Hayward fault (event 27). Farther north, such mechanisms occur along the Green Valley-Bartlett Springs fault (event 15) and the Rodgers Creek-Healdsburg-Maacama faults (events 16, 17, 19, 20). In southern California, such mechanisms mark the San Jacinto fault (events 78, 82-85) and the Imperial fault (event 89). Along the coast west of the San Andreas fault, similar focal mechanisms occur along the San Gregorio-Palo Colorado fault (events 39, 40) in northern California and along the Newport-Inglewood fault zone (events 62, 71), the Rose Canyon fault (event 73), and the San Clemente fault (event 70) in southern California.

Exceptions to this simple pattern for moderate (M>4) events along the San Andreas fault and its major branches appear to be limited to regions of unusual complexity, such as the major bends in the San Andreas near Cajon Pass (event 69) and San Gorgonio Pass (event 80). Jones and others (1986) attributed the July 8, 1986, earthquake (event 80) to right-lateral slip on the Banning segment of the San Andreas fault where it dips 45бу N. beneath the San Bernardino Mountains. The October 17, 1989, M=7.1 Lorna Prieta earthquake involved nearly equal amounts of right and reverse slip along a section of the San Andreas fault that takes a slight westerly bend through the Santa Cruz Mountains and dips 70бу SW. (see chap. 6). Smaller (M<4) events near, but probably not on, the fault show a great variety of focal mechanisms that reflect varying conditions along the fault; these mechanisms range from reverse or reverse-oblique slip on easterly-striking planes (events 37, 67), through right-lateral strike slip on planes parallel to the San Andreas fault (events 65, 66, 68, 87), to normal or normal-oblique slip on northerly-striking planes (events 77, 81).

Moderate earthquakes with strike-slip focal-mechanisms that are not located on major faults of the San Andreas system but yet are broadly associated with it commonly have right-slip planes, with strikes ranging from northwestward (event 42) to north-southward (events 35, 44, 74, 75, 76, 86). In most cases, these right-slip planes agree in strike with local mapped faults or with alignments of epicenters that strongly suggest active faults (events 75, 76).
CRUSTAL CONVERGENCE ADJACENT TO THE SAN ANDREAS FAULT SYSTEM
One of the more important results to emerge from high-resolution focal-mechanism studies in recent years is that earthquakes occurring even a short distance off faults of the San Andreas system can involve displacements that diverge sharply from local San Andreas strike-slip displacements. This pattern is particularly pronounced in the strong component of reverse slip at large angles (more than 60бу ) to the local strike of the San Andreas fault on both sides of the San Andreas fault system in both the Transverse and Coast Ranges.

North-south convergence within the Transverse Ranges is dominated by reverse slip on easterly-striking planes. The M = 7.7 Kern County earthquake of 1952 (event 64), which occurred on the south-dipping White Wolf fault along the north flank of the Transverse Ranges about 25 km north of the junction of the San Andreas and Garlock faults, and the M=6.6 San Fernando earthquake of 1971 (event 71), which ruptured a 20-km-Iong stretch of the northeast-dipping San Gabriel-San Fernando thrust faults (Whitcomb, 1971; Heaton, 1982), are two striking examples of this deformation. So, also, is the alignment of M=5-6 reverse-slip earthquakes (events 59, 60, 63) along the southern margin of the Transverse Ranges. The reverse slip on east-west-striking planes associated with these earthquakes suggests that the north-dipping Santa Monica-Cucamonga fault serves as an important convergent boundary between the Peninsular and Transverse Ranges.

The east-west-trending zone of convergence associated with these earthquakes curves northward near Santa Monica Bay and continues northwestward along the coast at least as far as Point Sal and probably as far as San Simeon. Focal mechanisms of earthquakes along this zone from Point Sal to Whittier (events 55-60, 63) are predominantly reverse slip, with slip directions nearly normal to the local trend of the zone. The focal mechanism of event 43 near San Simeon, which indicates right-oblique reverse slip on a northeast-dipping plane parallel to the coast, is intermediate between those of event 40 at Point Sur and event 55 at Point Sal.

Reverse-slip focal mechanisms for offshore events 18 and 41 in central California and for event 72 in southern California suggest that the offshore crust is undergoing compression normal to the coastline throughout the length of the San Andreas fault system. The Coalinga-Kettleman Hills earthquake sequence of 1982-85 (events 47-50) emphasizes the important role of crustal convergence along the southern Coast Ranges-Great Valley boundary in central California. The principal events in this sequence (event 48, Coalinga, and event 49, Kettleman Hills) involved reverse slip on subparallel planes at depths of 10 to 12 km that dip gently (approx 20бу) southwest. Much of the aftershock activity, however, occurred at shallower depths and involved high-angle reverse slip on planes dipping steeply (45бу-70бу) northeast (events f, g, i, o, q). Displacements associated with these earthquakes, which are nearly perpendicular to the San Andreas fault, represent a convergent process in which Franciscan melange on the west is being wedged between crystalline basement and the overlying Great Valley sedimentary sequence on the east (Wentworth and others, 1983; Eaton, 1990).

The boundary between the Coast Ranges and Great Valley is marked by reverse-slip earthquakes throughout much of its length: event 54 southeast of the Kettleman Hills, the Coalinga-Kettleman Hills sequence, event 21 near Winters east of Lake Berryessa, and event 14 west of Oroville. The similarity in focal mechanism of event 21 near Winters to the Coalinga and Kettleman Hills main shocks suggests that the convergent process acting in the southern Coast Ranges is common to the entire eastern margin of the Coast Ranges. Indeed, the strong earthquakes that shook the Winters-Vacaville-Dixon area in 1892, just south of event 21, resemble the Coalinga-Kettleman Hills sequence in both setting and intensity distribution. Focal mechanisms of smaller earthquakes along the Coast Ranges-Great Valley boundary in central California studied by Wong and others (1988) also suggest convergence across that boundary.
Convergence normal to the strike of the San Andreas fault is not limited to the coast and the Coast Ranges-Great Valley boundary described above. In a detailed examination of the focal mechanisms of aftershocks of the 1984 Morgan Hill earthquake, Oppenheimer and others (1988) concluded that the direction of maximum compression immediately adjacent to the Calaveras branch of the San Andreas fault is at an angle of about 80бу to the N. 10бу W. strike of the fault. Along the entire stretch of the San Andreas fault from Parldield to the Salton Sea, Jones (1988) found a constant angle of 65бу between the strike of the fault and the maximum-principal-stress direction for earthquakes occurring off the fault.

This evidence from earthquake focal mechanisms and other stress indicators (such as borehole breakouts and fold axes) that the maximum principal compressive stress may be oriented at a high angle to the local strike of the San Andreas fault seems to contradict long-accepted ideas for brittle failure in the crust based on laboratory experiments in rock mechanics. Zoback and others (1987) and Oppenheimer and others (1988) suggested that these relations can be explained by an exceptionally low average shear strength for the San Andreas fault system. As pointed out by Lachenbruch and McGarr in chapter 10, however, the strength and state of stress along the San Andreas fault are still matters for discussion.
EAST-WEST EXTENSION IN THE SIERRA NEVADA
The three moderate-earthquake focal mechanisms for the Sierra Nevada and its western foothills shown in figure 5.11A all indicate normal faulting on northerly-striking planes and suggest pervasive east-west extension throughout the Sierra Nevada. Event 52 is in a dense north-south-trending band of epicenters about 15 km east of the Kern River canyon, and event 53 is in a north-south-trending band of earthquakes about 10 km west of the Sierra frontal fault near Walker Pass. These relations suggest that the east-west spreading and associated normal faulting on northerly-striking faults of the Great Basin are encroaching into the southeast corner of the Sierra Nevada block (Jones and Dollar, 1986).

Event 13 is the main shock (M=5.7) of an earthquake sequence on a north-south-striking, west-dipping normal fault near the Oroville Dam that occurred in 1975. The uplift of the Sierra Nevada relative to the Great Valley to the west indicated by the focal mechanism of this event is also visible in the Chico monocline, which marks the Sierra Nevada-Great Valley boundary northwest of Oroville.
CONJUGATE FAULTING IN THE SIERRA NEVADA-GREAT BASIN BOUNDARY ZONE
The Sierra Nevada-Great Basin boundary zone is represented in figure 5.11A by three focal mechanisms. Events 11 and 12 lie northwest of Lake Tahoe along the edge of a minor gap in the band of seismicity along the east edge of the Sierra Nevada. Both events appear to have resulted from left-lateral slip along steeply dipping, northeast-striking faults; both events had aftershock regions that were elongate northeast-southwest. About 250 km southeast, the M=6.4 July 21, 1986, Chalfant Valley earthquake (event 51) resulted from right-lateral strike-slip displacement on a north-northwest-striking surface dipping 60бу SW. An M=5.7 foreshock on July 20 resulted from left-lateral strike slip on a northeast-striking, northwest-dipping surface. These two conjugate slip surfaces merge at their north ends (Cockerham and Corbett, 1987; Smith and Priestly, 1988).

The zone of intense seismicity in the vicinity of Long Valley caldera and the Sierra Nevada block to the south produced 11 M> 5.5 earthquakes from 1978 through 1984 (Savage and Cockerham, 1987), as well as many thousands of smaller events and numerous earthquake swarms. Most of the larger events occurred in the Sierra Nevada block south of Long Valley caldera, involving left-lateral slip along near-vertical, north-south- to north-northeast-striking faults. One of four Mб╓6 events that occurred on May 25-27, 1980, however, was located within the south moat of the caldera along the west-northwest-striking fault zone that produced most of the earthquake swarms (see Hill and others, 1985a, b).
FRAGMENTATION OF THE SOUTHEAST CORNER OF THE GORDA PLATE
The 1980 Eureka M=7.2 earthquake occurred along a fault break that extended from the continental slope 40 km west of the coastline at lat 41бу N. for a distance of 140 km southwestward to the MFZ, virtually cutting off the southeast corner of the Gorda plate. Focal mechanisms of the main shock and largest aftershock (events 1, 3, fig. 5.11A) both indicate left-lateral strike-slip displacement along a vertical fault that coincides with the line of aftershocks. Some early aftershocks, including event 4 and other moderate events farther east along the MFZ, have focal mechanisms that indicate right-lateral slip along the MFZ. Although the main shock occurred beneath the Continental Shelf, it seems clear that the 1980 earthquake primarily involved the Gorda plate because the fault broke well beyond the base of the continental slope and the edge of the North American plate. Moreover, left-lateral slip along the 1980 break stimulated right-lateral slip along the adjacent part of the MFZ. Ongoing right-lateral displacement along the MFZ is also indicated by event 5 (Dec. 1983).

Two moderate earthquakes near Cape Mendocino in 1981 and 1987 (events 6 and 7, respectively) had focal mechanisms similar to that of the 1980 Eureka earthquake, indicating left-lateral strike-slip displacement on steeply dipping, northeast-striking planes. Aftershocks of the 1987 M=5.8 event outlined a narrow, steeply dipping, northeast-trending, 20-km-Iong zone between about 15- and 25-km depth that extended southwestward from the shoreline just north of Cape Mendocino to the MFZ. This aftershock zone appears to cut off the southeasternmost corner of the Gorda plate just north of the abrupt eastward termination of intense seismicity along the MFZ, at a point that might be taken as the Mendocino triple junction from the viewpoint of seismicity.

Relative horizontal extension at seismogenic depths is suggested by events 2 and 8. Event 2 (Nov. 10, 1980; 7 km deep) was the largest in a detached cluster of shallow aftershocks 20 km east of the 1980 main shock, and event 8 (Apr. 9, 1987; 26 km deep) occurred about 100 km east of the 1980 main shock in the zone of seismicity associated with the subducting Gorda plate.
DISCUSSION
The Pacific plate moved northwestward with respect to the North American plate by 300 to 400 mm during the 7-yr interval 1980-86. Earthquakes occurring along the San Andreas fault system during the same interval, however, accommodated only a small fraction of this relative plate motion. Only four earthquakes of M>5 occurred along branches of the San Andreas fault system during 1980-86: the pair of M=5.9-5.3 Livermore earthquakes (events 29, 30, fig. 5. 10A) on the Greenville fault (Jan. 24-27, 1980), the M=6.2 Morgan Hill earthquake (event 33) on the Calaveras fault (Apr. 24, 1984), and the M=5.6 North Palm Springs earthquake (event 80) on the Banning segment of the San Andreas fault (July 8, 1986). Each of these moderate San Andreas earthquakes ruptured fault segments limited to 20 to 30 km in length, with average displacements over the respective rupture surfaces of 100 to 200 mm (see Hartzell and Heaton, 1986). As is typical of earthquakes along the San Andreas fault system, each of these events involved nearly pure right-lateral strike-slip displacement coincident with the local strike of the fault. As is also typical of San Andreas earthquakes, slip on the first three events occurred on near-vertical fault planes with a northwestward to north-northwestward strike. The North Palm Springs earthquake, which ruptured a section of the east-west-striking Banning fault in the structurally complex San Gorgonio bend in the fault system at the southern margin of the Transverse Ranges, represents an important deviation from typical San Andreas earthquakes. Although its displacement was dominantly right-lateral strike slip, it occurred along a plane that dips 45бу N. (Jones and others, 1986) and included a small but significant component of reverse slip (Mendoza and Hartzell, 1988). With the arguable exception of the North Palm Springs earthquake (arguable because of the complex section of the fault system in which it occurred), however, none of these M>5 earthquakes ruptured the main trace of the San Andreas fault. Indeed, the two most recent M>5 earthquakes to clearly do so were the M=6 Parkfield earthquake of 1966 (Bakun and McEvilly, 1984) and the M=7.1 Lorna Prieta earthquake of 1989.

Thus, aside from the displacement accommodated by steady aseismic slip at a rate of 32 to 37 mm/yr along the creeping section of the fault in central California, most relative plate motion across the San Andreas transform boundary during this 7-yr interval accumulated as elastic shear strain. Accordingly, the earthquakes plotted in figures 5.3 through 5.9 are symptomatic of accumulating strain along the San Andreas fault system rather than of effective strain release. The latter requires rupture with a major earthquake along one of the locked stretches of the San Andreas fault.
SEISMICITY PATTERNS AND THE EARTHQUAKE CYCLE
What changes in spatial-temporal patterns of earthquake occurrence might we expect to see as the next great earthquake on the San Andreas fault approaches? Both historical and instrumental seismicity records indicate that the spatial distribution of earthquakes in California changes only slowly over periods of decades to centuries, although the intensity of activity within this distribution fluctuates year to year (Ellsworth and others, 1981; Hill and others, in press; Hutton and others, in press). Temporal fluctuations in activity during the interval 1980-86, for example, were dominated by a short-lived aftershock sequence following the 1980 Eureka M=7.2 earthquake and by the long-lived aftershock sequence following the 1983 Coalinga M=6.7 earthquake. The overall spatial distribution of earthquakes in California, however, remained nearly stationary throughout this 7-yr interval. Furthermore, the spatial pattern defined by 1980-86 seismicity is much the same as that outlined by the record of M>5 earthquakes that extends back nearly 200 yr.

Variations in the historical rate of moderate to large (M>5) earthquakes in central California before and after the 1906 San Francisco earthquake appear to mimic those described by Fedotov (1965) and Mogi (1968) for the earthquake cycle associated with great, subduction-zone earthquakes in Japan, Kamchatka, and the Kurile Islands (see chap. 6; Ellsworth and others, 1981). The history of instrumentally recorded M <5 earthquakes in California is too short, however, to indicate whether we might expect to see distinctive changes in the seismicity pattern a short time (months to years) before the next great earthquake on the San Andreas fault. We have yet to see, for example, whether the quiescent (locked) segments of the San Andreas fault remain aseismic except for the rupture of a great earthquake, or whether these segments become active with small to moderate earthquakes as foreshock activity to great earthquakes.
DISTRIBUTED SEISMICITY AND DEFORMATION OF THE PLATE MARGINS
The two largest earthquakes in California during the interval 1980-86 occurred off the faults of the San Andreas system, and their occurrence emphasizes the importance of deformation within the plate margins along the San Andreas transform boundary. The M=7.2 Eureka event (Nov. 8, 1980), for example, involved deformation internal to the Gorda plate; and the M=6. 7 Coalinga event (May 2, 1983) involved crustal shortening with reverse slip perpendicular to the San Andreas fault. These two earthquakes and the many smaller, "off fault" events reflect local deviations from the simple rigid-plate approximation of plate tectonics.
DEFORMATION OF THE CORDA PLATE
As the small, youthful Gorda plate is subducted obliquely northeastward beneath the North American plate, it is being subjected to north-south compression in response to a component of convergence between the larger, older Juan de Fuca plate to the north and the Pacific plate to the south (Jachens and Griscom, 1983; Wilson, 1986). Distorted marine magnetic anomalies within the Gorda plate indicate that it has undergone progressive internal deformation over the past 5 Ma in response to this compression (Silver, 1971), and current seismicity within the plate (fig. 5.4) indicates that this deformation continues to the present. The 1980 Eureka M=7.2 earthquake emphasizes that part of this deformation occurs with left-lateral slip on northeast-striking faults within the plate. The seismicity map and cross sections (fig. 5.4) demonstrate that deformation associated with the Gorda plate terminates abruptly against the Pacific plate in a steeply north-dipping zone of interaction along the MFZ, which can be followed on shore beneath the North American plate as a gently east-dipping, subhorizontal zone of widely scattered small to moderate earthquakes. Thus, convergence between the Gorda and Pacific plates across the MFZ apparently occurs by crushing and thickening of the southern margin of the Gorda plate as it is jammed against the anvil-like mass formed by the thicker and colder Pacific plate. Diminished east-west stress in the Gorda plate resulting from the subducting limb of the plate farther east serves to increase the difference between the maximum (north-south) and minimum (east-west) compressive stresses within the plate, leading to left-lateral strike-slip displacements along northeast-striking faults, as in the M=7.2 Eureka earthquake. This process accommodates the convergent component of Gorda-Pacific plate motion along the east end of the MFZ at the expense of fragmentation and eastward expansion of the Gorda plate north of the MFZ.
THE SAN ANDREAS DISCREPANCY
Much of the seismicity adjacent to the San Andreas fault system is attributable to differences between the long-term slip rate and direction (slip vector) along the San Andreas fault system and that predicted for relative motion between the Pacific and North American plates along the San Andreas transform boundary on the basis of global models of plate motion. Minster and Jordan (1978, 1987) predicted that the direction of dextral slip between the Pacific and North American plates along the San Andreas transform boundary in central California is N. 35бу W. The main trace of the San Andreas system, however, strikes N. 41бу W. through central and northern California and N. 65бу-70бу W. through the Transverse Ranges in southern California. DeMets and others (1987) concluded that the marine magnetic anomalies at the mouth of the Gulf of California constrain the slip rate to an average of 49 mm/yr over the past 3 to 4 Ma. Both long-term geologic offset data and geodetic data measured over the past several decades, however, indicate that the average slip rate along the San Andreas fault system is only about 35 mm/yr. The contribution to deformation of the western margin of the North American plate from spreading across the Basin and Range province is about 10 mm/yr in a N. 56бу W. direction (Minster and Jordan, 1987). Ellsworth (see chap. 6) suggests that most of the San Andreas discrepancy can be explained if the component of dextral slip associated with historical Basin and Range earthquakes reflects a long-term trend superimposed on the N. 56бу W. spreading direction. If so, then the residual component of Basin and Range extension perpendicular to the San Andreas fault system is approximately balanced by convergence across the Coast Ranges and continental margin.
CONVERGENCE NORMAL TO THE SAN ANDREAS FAULT SYSTEM
Focal mechanisms of earthquakes occurring off the San Andreas fault system suggest that the component of the San Andreas discrepancy normal to the fault system may, indeed, be accommodated by distributed brittle deformation on either side of the fault system. These mechanisms range from dextral strike slip on planes subparallel to the San Andreas fault, through oblique-reverse slip, to nearly pure reverse slip with a slip direction perpendicular to the San Andreas fault.

The Coalinga-North Kettleman Hills earthquake sequence provides clear evidence for crustal convergence perpendicular to the San Andreas fault system in the Coast Ranges. The several smaller events with similar mechanisms to the north along both the eastern and western (coastal) margins of the Coast Ranges (fig. 5.11) suggest that the convergence responsible for the Coalinga earthquake may be active the length of the Coast Ranges (Wong and others, 1988; Eaton and Rymer, 1990). The subparallelism of fold axes within the Coast Ranges with the San Andreas fault indicates that fault-normal convergence has been important for the past 3 Ma in central California ( Page and Engebretson, 1984). N amson and Davis (1988) proposed that the entire system of Coast Range folds may be genetically related to Coalinga-like earthquake sequences and low-angle (blind) thrust faults that are rooted in a decollement near the base of the seismogenic crust. The reverse focal mechanisms for earthquakes associated with offshore faults along the western margin of the Coast Ranges suggest that, here, convergence involves westward thrusting of the Coast Ranges over oceanic crust of the Pacific plate.
The pronounced discrepancy in the strike of the San Andreas fault through the Transverse Ranges with respect to the Pacific-North American plate slip direction provides an obvious source of local crustal convergence (Hill and Dibblee, 1953; Atwater, 1970), and the associated structural complexities serve to distribute brittle deformation (seismicity) much more widely about the San Andreas fault system in southern California than about the relatively straight sections of the fault system in central and northern California. The largest earthquake in California since the great 1906 San Francisco earthquake occurred near the northern margin of this convergent regime; this M=7.7 Kern County earthquake ruptured some 35 km of the southeast-dipping White Wolf fault with left-oblique reverse slip on July 21, 1952.

The focal mechanisms of larger Transverse Range earthquakes, together with the mapped attitudes of major faults with Holocene offsets, show that much of this convergence occurs with slip on north-dipping thrust faults within and along the southern margin of the central Transverse Ranges (fig. 5.11A). For earthquakes in the western Transverse Ranges, the direction of reverse slip is more southwestward, consistent with thrusting of the western Transverse Ranges over the Pacific plate similar to that in the Coast Ranges to the north.
EXTENSIONAL DEFORMATION AND THE SOUTHERN SECTION OF THE SAN ANDREAS FAULT SYSTEM
The fault-normal convergence that dominates deformation adjacent to the San Andreas fault system through both the Coast Ranges and Transverse Ranges gives way rather abruptly to the extensional regime of the Salton Trough near the southern margin of the intensely active San Gorgonio bend in the fault. Focal mechanisms of earthquakes occurring on secondary structures adjacent to the seismically quiescent Indio segment of the San Andreas fault, for example, show a mix of strike- and dip-slip mechanisms. As is the case farther north, however, P-axes for these earthquakes tend to be oriented at a high angle (60бу-65бу) to the fault, suggesting that the Indio segment of the fault may also be relatively weak (Jones, 1988).

One particularly noteworthy aspect of seismicity south of the Transverse Ranges is the, tendency for earthquakes to occur along conjugate strike-slip structures. Recall that the Sierra Nevada-Great Basin boundary zone also shows this tendency and that both regions are subject to extensional deformation, earthquake swarms, and late Quaternary volcanism. Earthquake sequences within the southern section of the San Andreas fault system commonly produce epicenter lineations that intersect at nearly a 90бу angle with the northwest-striking right-slip plane and the northeast-striking left-slip plane. Earthquake-swarm sequences in the Brawley seismic zone, for example, typically occur along northeast-striking lineations normal to the trace of the adjacent Imperial fault (Johnson, 1979), and the M=5.7 Westmorland earthquake of 1981 involved left-lateral slip along several subparallel, northeast-striking planes (Johnson and Hutton, 1982). The diffuse lineations of epicenters spanning the area of the Peninsular Ranges between the San Jacinto and Elsinore faults also tend to be orthogonal to these two branches of the San Andreas fault system (fig. 5.10A). An impressive recent example of this orthogonal conjugate pattern is the M=6.2 and 6.6 Superstition Hills earthquakes of November 24, 1987 (Magistrale and others, 1988).

The kinematics of these conjugate structures remains a matter of conjecture. Dextral slip along through going faults of the San Andreas system must certainly dominate deformation, and the shorter, northeast-striking structures must play only a secondary role. Nicholson and others (1986) proposed that the northeast-striking lineations represent the boundaries between blocks rotating clockwise much like roller bearings, between subparallel pairs of dextral strike-slip faults. Hill (1977) and Weaver and Hill (1978/79) suggested that within local spreading centers, such as the Brawley seismic zone, conjugate strike-slip structures form miniature triple junctions with a dike or normal fault that subtends the acute angle between the conjugate strike-slip faults.
MAXIMUM FOCAL DEPTHS AND THICKNESS OF THE SEISMOGENIC CRUST
Maximum focal depths of earthquakes beneath the San Andreas transform boundary range from less than 5 km beneath the Geysers geothermal field in the northern Coast Ranges to more than 20 km beneath the Transverse Ranges, the eastern margin of the Coast Ranges, and the San Jacinto and Elsinore faults in southernmost California. Beneath relatively straight segments of the San Andreas fault system through central California, maximum focal depths range from 12 to 15 km. Sibson (1983) pointed out that these variations in maximum focal depth along the San Andreas fault system are inversely correlated with surficial heat flow, and he argued that the maximum depth of earthquakes coincides with the temperature-dependent transition from brittle failure in the upper crust to aseismic, quasi-plastic flow in the lower crust and upper mantle. For quartz-bearing rocks typical of the upper crust and deformation rates typical of the San Andreas fault system (1x10step -14 to 1x10step -13 sstep -l), this brittle/ductile transition occurs at about 300 буC (Sibson, 1983). By this interpretation, the thin seismogenic crust beneath both the Geysers and Brawley geothermal fields in northern and southern California, respectively, reflects elevated temperatures in the shallow crust, whereas the relatively thick seismogenic crust beneath the Transverse Ranges and the eastern margin of the Coast Ranges reflects depressed temperatures in the midcrust associated with crustal convergence. Although temperature may dominantly influence the thickness of the seismogenic crust, local variations in rock composition (particularly the presence or absence of modal quartz and structural water) and in strain rate can also be important. These variations, for example, may help explain isolated clusters of deep earthquakes, such as the 20- to 24-km-deep events north of San Pablo Bay in central California (see cross secs. F-F'', G-G'').

In any case, the thickness of the seismogenic crust beneath the San Andreas transform boundary seems to be much more strongly related to temperatures in the crust than to the structural thickness of crust defined by the depth to the Moho. This relation is strikingly illustrated by the twofold increase in thickness of the seismogenic crust beneath the rootless Transverse Ranges.
DECOLLEMENT AT THE BASE OF THE SEISMOGENIC CRUST?
A theme common to models of crustal convergence along the San Andreas fault system involves low-angle reverse slip on decollement surfaces near the base of the seismogenic crust (Wentworth and others, 1983; Webb and Kanamori, 1985; Namson and Davis, 1988; Eaton and Rymer, 1990). A natural extension of this theme leads to a view of the seismogenic crust as a conglomeration of relatively rigid blocks interacting by frictional slip along weak preexisting faults (block boundaries) in response to regional stresses transmitted through both the brittle crust and quasi-plastic deformation in the underlying lithosphere (Hill, 1982). However, the nature of a decollement surface at the base of the brittle crust and the relation of the seismogenic San Andreas fault system to the aseismic transform boundary in the underlying lithosphere remain speculative. It is not yet clear, for example, whether the San Andreas fault continues below the seismogenic crust as a narrow, near-vertical boundary (possibly offset a substantial distance from the seismogenic fault by slip on the horizontal decollement surface) that slips by quasi-plastic, mylonitic deformation or whether it broadens rapidly with depth into a wide shear zone spanning, say, the entire width of the Coast Ranges (Sibson, 1983).