Figure 9.1

      Studies of the San Andreas fault system using the Earth's gravity and magnetic fields began before 1960 but received their main impetus during the 1970's, when work on the possibility of predicting earthquakes on this system began in earnest. Early investigations focused mainly on short segments of the faults because only limited data were available. More extensive potential-field data sets that have been published in recent years now permit the gravity and magnetic expression of the entire fault system to be viewed in a regional context (fig. 9.1).

      Gravity and magnetic data reflect, respectively, the density and magnetization of the rocks beneath the surface; and, in many situations, these properties can be closely correlated with the rock types seen in outcrop. Anomalies in the Earth's gravity and magnetic fields - for example, local deviations of the measured fields from those predicted on the basis of simplified Earth models - primarily reflect lateral variations in density and magnetization that generally are not included in such simple models. These anomalies can be interpreted qualitatively to infer the general spatial distribution of rock types in the subsurface, and quantitatively, through the use of efficient computer-based modeling techniques (Saltus and Blakely, 1983; Chuchel, 1985; Blakely and Simpson, 1986), to determine the geometries and specific locations of concealed rock bodies. Although all such interpretations are nonunique, both because many different distributions of density and magnetization can give rise to identical anomalies and because density and magnetization do not uniquely define a specific rock type, the combined use of gravity and magnetic data with geologic, geochemical, and other geophysical data can be especially effective in limiting the number of acceptable interpretations.
      In the sections below, we first present regional gravity and magnetic maps covering the San Andreas fault system and briefly discuss the sources, compilation methods, and limitations of the data from which they were produced and, in general terms, the sources of the anomalies shown on them. We then summarize the results of individual studies of sections of the major faults in the system and attempt to synthesize these results in terms of the geometries of the faults, the structures and rock types in the surrounding areas that are related to the faults, and the properties of the fault zones. Next, we focus on studies that relate to movement on the faults, including constraints on total displacements. Finally, we discuss the plate-tectonic implications of potential-field investigations of the fault system.

      An isostatic residual gravity map of the region surrounding the San Andreas fault system is shown in figure 9.2. We have chosen to present the gravity data in this form rather than in terms of the more common Bouguer or free-air gravity because of the generally closer correlation between isostatic residual gravity and mapped geology (Jachens and Griscom, 1985; Simpson and others, 1986). Most long-wavelength anomalies (longer than approx 250 km) on a Bouguer gravity map are caused by deep-seated density distributions that buoyantly support the topography in a manner consistent with the principle of isostasy (Simpson and others, 1986). Bouguer gravity anomalies related to isostasy are prevalent in California because of the extreme topographic relief in the State (Oliver, 1980; Jachens and Griscom, 1985), and they are particularly strong near the coast, where an eastward to northeastward decrease in gravity reflects the transition from thin oceanic crust to thicker continental crust. In such areas as California, the Bouguer gravity anomalies associated with isostatic support of topography are so strong that they tend to distort or even mask the lower-amplitude anomalies caused by density distributions in the middle to upper crust, those anomalies most easily correlatable with rocks exposed at the surface (Jachens and Griscom, 1985). Our isostatic residual gravity map has these long-wavelength isostatic effects removed, at least to first order. We emphasize that the anomalies remaining on our map are predominantly caused by lateral density variations in the middle to upper crust and, as such, do not represent areas that are out of isostatic balance (Jachens and Griscom, 1985).

Figure 9.2 - Isostatic residual gravity map of the San Andreas fault system. Contour interval, 5 mGal. Faults simplified from Jennings and others (1977), McCulloch (1987), and Vedder (1987). Fault names and explanation in figure 9.4.

      Our isostatic residual gravity map (fig. 9.2) is based on the new isostatic residual gravity map of the conterminous United States by Simpson and others (1986), who presented a detailed discussion of the data sets and procedures used to generate this map. The basic gravity-data set was compiled for the "Gravity Anomaly Map of the United States" (Society of Exploration Geophysicists, 1982) and includes 1 million on shore and 0.8 million offshore gravity observations. These data were sampled on a rectangular grid with a grid spacing of 4 km, containing Bouguer gravity values onshore and free-air gravity values at sea (Godson, 1985). To produce our isostatic residual gravity map, the offshore free-air gravity values were converted to Bouguer gravity values. The gravitational effects of the deep density distributions that support the topography within 166.7 km of each grid intersection were computed according to the Airy-Heiskanen model of isostasy (Heiskanen and Moritz, 1967) using a 5- by 5-minute topographic-bathymetric data grid and model parameters as follows: topographic density, 2.67 g/cm³; crustal thickness at sea level, 30 km; and density contrast across the base of the model crust, 0.35 g/cm³. Combined isostatic and topographic effects for the region from 166.7 km to the antipode of each grid intersection were obtained from the maps by Karki and others (1961). This model gravity field was subtracted from each Bouguer gravity grid value to yield a grid of isostatic residual gravity values; the resulting grid was contoured by computer and displayed in color-band intervals of 10 mGal to produce figure 9.2. Limitations on the use of this map stem both from uncertainties in the point data from which the grid was constructed and from characteristics generated by the grid ding process. For onshore data, uncertainties in the point data values resulting from errors in observed gravity, elevation, terrain corrections, and isostatic reductions are estimated to be less than 2 to 3 mGal for most stations, possibly larger in areas of extreme topographic relief (Simpson and others, 1986). In offshore areas, the greatest uncertainty results from conversion of the original free-air gravity data to Bouguer gravity values, using the 5- by 5-minute average bathymetry. Where the sea-bottom topography is relatively gentle, this conversion probably results in uncertainties of about 5 mGal, but in such areas as parts of the California Continental Borderland (south of lat 34° N.) and over the edge of the Continental Shelf, where water depths change rapidly, errors of several tens of milligals are possible. These conversion errors generally appear as high-amplitude, nearly circular anomalies with diameters of as much as 40 km.
      Although gravity coverage along most of the San Andreas fault system is quite dense when viewed at the scale of figure 9.2, sampling of these data on a 4-km grid means that anomalies with characteristic dimensions less than several times the grid spacing are not faithfully portrayed. Our isostatic residual gravity map (fig. 9.2) is sufficient for qualitative and quantitative interpretation at the scale shown, but for more detailed interpretations, especially quantitative modeling, the reader is referred to the original data sources, such as Oliver and others (1980), Roberts and others (1981), Snyder and others (1982), and the other reports cited throughout this chapter.

      A magnetic anomaly map of the region surrounding the San Andreas fault system is shown in figure 9.3. This map is based on the magnetic anomaly map of the Western United States by Bond and Zietz (1987), which was compiled from hundreds of magnetic surveys with widely differing flight heights, flightline spacings, and sensor types.

Figure 9.3 - Magnetic map of the San Andreas fault system (from Bond and Zietz, 1987). Contour interval, 100 nT. Faults simplified from Jennings and others (1977), McCulloch (1987), and Vedder (1987). Same symbols as in figure 9.4.

      In contrast to the lengthy series of reduction steps that were required to convert the gravity observations to the form shown in figure 9.2, very little was done to the observed magnetic data to prepare them for compilation. Although the original data were collected at many different heights, no analytic procedures were used to continue them to a common elevation. Instead, the various surveys were referenced to the International Geomagnetic Reference Field (lG RF) adjusted for the date of the survey and an arbitrary zero datum, and then combined manually by inspection. Long profiles of magnetic data collected under the National Uranium Resource Evaluation (NURE) program of the U.S. Department of Energy and by the U.S. Naval Oceanographic Office (NOO) served as guides for determining the zero datum for the various surveys. The resulting data are presented at a color contour interval of 100 nT (gammas) in figure 9.3.
      Our magnetic anomaly map (fig. 9.3) is the most complete compilation available for the San Andreas fault system and is useful for qualitatively determining the location, shape, and regional setting of large magnetic bodies. However, because of the compilation methods used to construct this map and because the contour interval is relatively coarse (100 nT), it will not, in general, be adequate for detailed qualitative or quantitative examination of individual anomalies. Where detailed information is required, the reader is referred to the original sources from which our map was compiled; a comprehensive listing of these sources is given by Bond and Zietz (1987).
      A particularly valuable source of aeromagnetic data over the San Andreas fault system is the profile data collected under the NURE program. In general, these data were collected along long profiles oriented east-west at a nominal height of 120 m above terrain and spaced about 5 km apart. The wide flightline spacing and low altitude of the survey lines preclude constructing realistic contour maps from these data in most places, but the long profiles are well suited for quantitative modeling. These data are available in the form of atlas folios or digital tapes for individual 1° by 2° quadrangles from the U. S. Department of Energy, Grand Junction, Colo.
      When interpreting magnetic data, the inclination of the Earth's magnetic field must be taken into account because the magnetization induced in the magnetic source rocks by this field will have a similar inclination. Along the San Andreas fault, this inclination ranges from 58° to 64° downward toward magnetic north. Contoured magnetic anomalies over inductively magnetized or normally magnetized sources at these field inclinations will commonly display a dipole response, namely, magnetic lows associated with the north sides of magnetic highs. Inspection of our magnetic-anomaly map (fig. 9.3) indeed identifies numerous such magnetic lows on the north or northeast sides of major magnetic highs. In general, each magnetic low is located directly beyond the north or northeast contact of the causative magnetic mass.

      Conspicuous features of the gravity field over the San Andreas fault system are linear highs and lows that trend subparallel to the major faults in the system. Highs (≥10 mGal) generally occur over exposed crystalline rocks of the Salinian block southwest of the San Andreas fault, over mafic granitic and metamorphic rocks of the Sierra Nevada and the Mojave Desert, and over Mesozoic and Tertiary layered rocks of the Franciscan assemblage, particularly in areas containing large amounts of mafic volcanic rocks (generally part of an ophiolite belt) or high-pressure metamorphic-mineral facies. Most of the deepest lows are caused by thick accumulations of low-density Cenozoic sedimentary rocks that fill tectonic basins adjacent to the faults and in the surrounding areas. Shallower lows occur over certain large serpentinite bodies within the Franciscan assemblage, over felsic plutons in the granitic terranes of California, and over a young concealed granitic pluton associated with the Geysers geothermal area at lat 39° N., long 122°45' W. (Chapman, 1975; Isherwood, 1976).
      Magnetic anomalies in the vicinity of the San Andreas fault system typically are caused by anyone of three different rock types. The strongest anomalies generally reflect tabular bodies of serpentinite associated with the Franciscan assemblage and may also reflect the ophiolitic rocks, especially serpentinite, that locally lie above it. Mafic plutonic rocks, such as those exposed in the western Peninsular Ranges and along the west edge of the southern Sierra Nevada, can produce moderate to strong magnetic anomalies. Plutonic sources, not necessarily mafic only, probably account for most of the anomalies in the Salinian block, southwest of the San Andreas fault. Although younger volcanic rocks, in particular the mafic varieties, commonly are highly magnetic, such rocks do not cause significant magnetic features near the San Andreas fault as shown on our magnetic anomaly map (fig. 9.3) because magnetic volcanic rocks are volumetrically unimportant at the scale of this map.
      In most areas, sedimentary rocks are considered nonmagnetic because they fail to cause aeromagnetic anomalies. Along the San Andreas fault system, however, several sedimentary-rock units cause magnetic anomalies as large as 150 nT. These units include rocks of Mesozoic and Tertiary age; other units composed primarily of detrital serpentinite also produce anomalies of this magnitude. None of these sedimentary units are really large enough to produce magnetic anomalies visible at the scale of our magnetic anomaly map (fig. 9.3).
      Near the north end of the San Andreas fault, several magnetic anomalies project landward from the linear pattern of anomalies that characterizes the oceanic crust. These anomalies reflect remanent magnetization in the oceanic crust; the source rocks are primarily basaltic volcanic rocks.

      The usefulness of potential-field data along the San Andreas fault system is maximized where rock masses with differing physical properties are juxtaposed. Under these conditions, geophysical anomalies arise from which the location and attitude of the fault may be calculated (Blakely and Simpson, 1986). In general, the fault is expected to be situated at or near the steepest gradient of the anomaly. These sites are particularly helpful in areas where the fault trace or zone is concealed by young sedimentary deposits or by the Pacific Ocean. In addition, we have found that some of these data are useful in identifying the main strand of the fault zone where the presently active fault trace may not, in fact, be the original plate boundary. Some areas where the potential-field data define the locations of faults are shown on figure 9.4 and are discussed below.

Figure 9.4 - San Andreas fault system, showing fault dips calculated from gravity and magnetic data, locations of offset geophysical anomalies, and south border of the subducted Juan de Fuca plate. Note wells at lat 39° and 40° N. ND, Navarro discontinuity; PF, Pilarcitos fault.

      Although the location of the San Andreas fault between Point Arena and Cape Mendocino is concealed by the Pacific Ocean, the aeromagnetic data show a linear magnetic anomaly, striking northwest within the Pacific plate, that is inferred to be obliquely cut off by the fault about 20 km northwest of Point Arena. Farther north, the fault trace just south of Cape Mendocino has proved particularly difficult to locate because it may be too close to shore to be resolved by marine geophysical surveys. A recent aeromagnetic map of this problematic area has, indeed, displayed a magnetic boundary trending close to and along the shore, thus representing the likely location of the San Andreas fault (Griscom, 1980a).
      Aeromagnetic surveys over the Pacific Ocean at the entrance to the San Francisco Bay (Brabb and Hanna, 1981) show that the offshore extension of the Pilarcitos fault (an inactive fault branching westward from the San Andreas fault) is cut off by the offshore northward extension of the San Gregorio fault. The San Gregorio fault can be traced northward by using a detailed aeromagnetic map to the point where it intersects the San Andreas fault at Bolinas Lagoon, about 20 km northwest of the bay mouth (see McCulloch, 1987, fig. 15).
      From San Francisco southward to lat 35°15' N., the detailed gravity and magnetic data indicate that, in general, the westernmost strand of the main San Andreas fault zone is the major plate boundary. The layered Franciscan assemblage to the east may be less competent than the granitic basement of the Salinian block to the west, and new strands may be more likely to appear in the less competent rocks. An exception to this generalization is found at lat 36° N., where a thin fault sliver of hornblende-quartz gabbro occurs at Gold Hill (Ross, 1970) that has been used to estimate offset on the San Andreas fault. The magnetic anomaly associated with this gabbro body indicates that it is at most 10 km long by 2 km wide (U.S. Geological Survey, 1987).
      Farther south along the San Andreas fault, a linear magnetic high extends along the fault approximately between long 116° and 118° W. (fig. 9.3). On the basis of local model studies of this anomaly, Simpson and others (in press) show that this feature probably reflects the edge of an extensive block of magnetic rocks on the northeast side of the San Andreas fault, where the magnetic material is Precambrian igneous and metamorphic rocks, as well as Mesozoic plutonic rocks. Using detailed magnetic data (U.S. Geological Survey, 1979), the south border or magnetic boundary of this magnetic block (fig. 9.4) can be traced from west to east along a series of fault segments; from long 117°15' W., the boundary follows the southern fault trace to long 116°15' W., then crosses over to the northern trace along the short, east-west-trending fault segment, and finally continues eastward along the northern trace. These faults thus may represent the original fault boundary (now somewhat kinked) between the two plates. The geologic observation that rocks on the north side of these fault segments are native to the San Bernardino Mountains and contrast with compositionally different rocks on the south side (Matti and others, 1985) agrees with the magnetic interpretation. The magnetic boundary continues southeastward along the San Andreas fault in Coachella Valley to long 116°08' W. A possible farther continuation of this linear magnetic high extends south-eastward at a lower amplitude and diverges eastward from the present San Andreas fault, generally following and lying northeast of the Clemens Wells fault, a possible earlier strand of the San Andreas fault.
      In Coachella Valley, the San Andreas fault (North Branch or Coachella segment) is situated along the northeast side of a substantial linear gravity low caused by at least 4.7 km of low-density sedimentary rocks (Biehler, 1964) filling the valley. Gradient studies on the relatively detailed gravity data in this valley by one of us (Griscom) identify numerous fault strands, including the northern and southern branches of the San Andreas fault (the latter, the Banning fault), as well as several possible fault segments on the southwest side of the valley.
      The Garlock fault at long 118° W. changes direction and forms a zone as much as 8 km wide. Models of both the magnetic and gravity fields calculated normal to the fault indicate that here the main lithologic boundary is the most northerly fault (fig. 9.4); the granitic rocks farther north are more magnetic and less dense than those to the south.
      Near the intersection of the San Andreas and Garlock faults, gravity modeling by Andrew Griscom and K.G. Freeman (Griscom and Oliver, 1980) suggests that the fault dips 55° SW. to a depth of 6 km and thence vertically to a depth of at least 10 km (fig. 9.5B). About 60 km farther southeast, gravity data also indicate a southwesterly dip for the fault, but the angle of dip (30°-60°) is uncertain, owing to difficulty in interpreting the complex gravity field that results from large lateral density variations in the region southwest of the fault. Farther southeast, where the San Andreas fault splits into numerous branches (long 116°00' W.), gravity data on two branches indicate that both faults dip northeast, with Precambrian crystalline rocks in the upper plate overlying young sedimentary rocks and alluvium. The gravity models suggest dips of 15°-25° NE. to depths of 1.5 to 2.5 km but do not resolve the fault attitude at greater depth (fig. 9.5C). Geologic mapping, which shows part of the southern branch of the San Andreas fault as a north-dipping thrust fault (Matti and others, 1985), and a study of recent earthquakes in this area, which yielded fault-plane solutions of predominantly oblique-slip motion and including low-angle thrust solutions dipping 30° N. (Nicholson and others, 1986), both support the gravity interpretation of northeast-dipping faults in this area.

      The inferred fault attitudes shown in figure 9.4 suggest a relation between attitude and plan-view geometry. Faults tend to be vertical except where they undergo abrupt changes of strike. The sinistral bends in the San Andreas fault near its junction with the Calaveras fault and in the Big Bend region southeast of its junction with the Garlock fault create regions likely to be subject to compression due to relative southward movement of the North American plate with respect to the Pacific plate. The dipping fault planes in these regions may reflect a thrust component of fault movement that accommodated the compression. Similarly, the region around the broad dextral bend in the San Andreas fault north of Point Arena may have a component of extension parallel to the direction of relative plate motion, and the low- angle eastward dip of fault plane there may reflect accommodation of the extension by low-angle normal faulting.
      The number of examples on which the above speculations are based is quite limited, and further detailed investigations at critical sites along the San Andreas fault system are needed to test the relation between fault attitude, change of strike, and relative plate-motion direction.

      Millions of years of strike-slip movement along faults of the San Andreas system have produced, in many places, a narrow fault zone in which physical properties differ from those in the surrounding rock masses. These differences are due to the presence within the fault zone of fractured or pulverized rock, exotic rock slivers that have been transported along the fault from other places, and such mobile materials as fluids and serpentinite that have migrated along the fault zone. A few investigators have used gravity and magnetic data to study the properties of this zone.
      Although Stierman (1984) and Wang and others (1986) sought to explain gravity lows along the fault as the result of a substantial increase in porosity by fracturing, gravity lows not directly associated with basins filled by Cenozoic sedimentary rocks along the faults are very uncommon. These rare lows amount, with few exceptions (such as the 10-12-mGallow studied by Stierman, 1984), to amplitudes of only a few milligals. Feng and McEvilly (1983) and Trehu and Wheeler (1987) inferred from seismic data that zones of low seismic velocity 5 to 10 km wide and more than 10 km deep are associated with the San Andreas fault zone and, presumably, with fractured rocks. The low-velocity zone of Trehu and Wheeler (1987), however, has no associated gravity low, even though calculations by Andrew Griscom indicate that this zone might be expected to produce a gravity anomaly of about -25 mGal and more than 10 wide, using the standard velocity-density relations of Hill (1978). An explanation for this unexpected result can be found in the borehole gravity and seismic-velocity results (Schmoker, 1977; Stierman and Kovach, 1979) from a 600-m-deep borehole in diorite located 1.2 km from the San Andreas fault. For the lower half of this borehole, the seismic velocity averages only 3.1 km/s (although saturated core samples measured 6.6 km/s in the laboratory), and the average computed rock densities are as follows: bulk density from cores, 2.72 g/cm³; borehole density from gravity measurements, 2.60 g/cm³; and computed density from borehole velocities (density-velocity relations of Hill, 1978), 2.36 g/cm³. Correcting for a nearby low-density sedimentary section that causes a gravity gradient along the hole raises the borehole density (from gravity measurements) closer to the bulk density. The results described above indicate that macrofractures can cause large decreases in seismic velocity but much smaller decreases in density than those predicted from standard velocity-density relations.
      Allen (1968) pointed out a possible relation between the style of fault movement and the presence of serpentinite within fault zones of the San Andreas, Calaveras, and Hayward faults. He noted that serpentine is common within the fault zone along the creeping section of the San Andreas fault between Hollister and Cholame, whereas it is absent along the locked segments to the north and south. Irwin and Barnes (1975) noted the same relation between serpentinite and fault creep and discussed the possible role of metamorphic fluids on the seismic behavior of fault segments. Hanna and others (1972) studied aeromagnetic data along the San Andreas fault between San Francisco and San Bernardino and found that the creeping segment of the fault is characterized by broad aeromagnetic anomalies, which they interpreted as reflecting large concealed masses of serpentinite. Linear magnetic anomalies that most likely reflect serpentinite also are present along the creeping section of the Hayward fault east of the San Francisco Bay (fig. 9.3). These magnetic data support the speculation that appreciable amounts of serpentinite contained within a fault zone can influence the style of movement on the fault.

      Strike-slip movement on the faults of the San Andreas system has produced offsets in formerly continuous geophysical anomalies. As might be expected, on those faults where the geologic offset is at most a few tens of kilometers, it is generally easy to identify corresponding magnetic or gravity features that are offset by similar distances. Examples of such faults are the Elsinore fault and the rectilinear system of minor strike-slip faults in the Mojave Desert block northeast of the San Andreas fault. In figure 9.4, the two or more piercing points of an offset geophysical anomaly are labeled with the same letter, and the specific points being described are designated with subscripts, numbered consecutively from northeast to southwest across the fault system.
      Offset along the San Andreas fault system in southern California is believed to be approximately 300 km in a right-lateral sense, based on offset of the Pelona-Orocopia schist belts, together with associated characteristic Precambrian and Triassic rock assemblages of the thrust plate overlying the schist belts (Crowell, 1962; Clarke and Nilsen, 1973; summarized in Hamilton, 1978). Isostatic gravity highs are associated with the Orocopia Schist (point A₁ at lat 33°35' N., northeast side of fault), with the Pelona Schist of the Sierra Pelona/Soledad area (points A₂, A₃ at lat 34°35' N., between the San Andreas and San Gabriel faults), and adjacent to the south side of the Pelona Schist of the Tejon/Garlock area (point A₄ at lat 34°50' N., south of the San Andreas fault and west of the San Gabriel fault). Point A₄ is not well determined. The offsets of the gravity highs are, respectively, 240 km along the San Andreas between the first two highs (A₁ and A₂) and 60 km along the San Gabriel between the second two highs (A₃ and A₄), for a total of 300 km along the San Andreas fault system, in agreement with Crowell (1962). The source of the gravity highs is not obvious and may not be any of the rocks exposed at the surface (Griscom, 1980b), both because the density of the schist coring the antiforms is probably similar to or slightly lower than that of the surrounding Precambrian crystalline rocks and because other large areas of Pelona/Orocopia schist do not display associated gravity highs.
      The schist is marine in origin, predominantly metagray-wacke of low metamorphic grade (Haxel and Dillon, 1978), and may be underlain by subducted oceanic crust. The gravity highs may indicate relatively uplifted oceanic crust beneath these specific antiformal exposures of the schist, or else the proportion of greenstone interbedded with schist may increase here with depth.
      As mentioned above in the subsection entitled "Plan View," a linear magnetic high that extends along the San Andreas fault from long 116° to 118° W., a distance of about 200 km, indicates that a large area north of the fault in this region is composed of magnetic rocks, predominantly Mesozoic granitic plutons; the northwest limit of this magnetic area (J₁) is shown in figure 9.4. A similar large area of magnetic basement, also predominantly Mesozoic granitic rocks, that extends along the southwest side of the San Andreas fault is displaced from the former area right-laterally approximately 300 km; the northwest limit of this correlative area (J₂) is also shown in figure 9.4. This second area of magnetic rocks does not produce a significant magnetic high directly at the fault because the fault is on the northeast side of the magnetic mass and a magnetic low should occur for this geometry.
      Several significant geophysical anomalies are found along the central section of the San Andreas fault north of its junction with the Garlock fault. A pronounced gravity high is located on the northeast side of the fault at lat 34°55' N., where the southern "tail" of the Sierra Nevada is exposed. The associated rocks are hornblende-quartz gabbro and anorthositic gabbro (Ross, 1970, 1984) that also produce a substantial aeromagnetic high (point B₁, fig. 9.4). Similar rocks (Ross, 1970) are found within the San Andreas fault zone at Gold Hill (point B₂ at lat 35°50' N., too small to show at this scale) and at Logan (point B₃ at lat 36°52' N.), where magnetic anomalies (U.S. Department of Energy, 1981; U.S. Geological Survey, 1987) indicate that the gabbro bodies are thin slivers within the fault zone. The Logan outcrops are offset about 290 km from the gabbro of the Sierran "tail." A major northwest-trending magnetic anomaly extends northwestward of Logan near the coast (from point B₄ at lat 37°08' N.). The source rocks for this magnetic anomaly are interpreted to be gabbro, similar to that exposed near Logan (Hanna and others, 1972a), because the anomaly requires a source body several kilometers thick. These corresponding offset geophysical anomalies support the geologic correlations implying about 300 to 320 km of offset.
      The additional 100 km of granitic rocks extending northward from Logan to Montara Mountain (lat 37°35' N.) along the southwest side of the San Andreas fault does not have any correlative rocks exposed on the northeast side of the fault north of the gabbro of Sierran "tail," but the concealed crystalline basement rocks beneath the sedimentary rocks of the Great Valley may be correlative. Indeed, recent work on the Tertiary sedimentary rocks that overlie this additional 100 km of granitic rocks on the San Francisco peninsula suggests a lithologic and paleogeographic correlation with similar sedimentary rocks of the Great Valley (San Joaquin Basin) that are relatively offset 320 to 330 km to the southeast (see fig. 3.4; Stanley, 1987).
      Movement on the San Andreas fault north of San Francisco (Griscom and Jachens, 1989) is complicated by right-lateral displacement added by the presently active San Gregorio-Hosgri fault, which intersects the San Andreas fault at San Francisco and provides an additional 115 km (Graham and Dickinson, 1978) or 150 km (Clark and others, 1984; Ross, 1984) of offset. The total offset on the San Andreas fault system here is further complicated by movement on branch faults to the east (Calaveras and Hayward fault systems of unknown offset) and, more importantly, by past movement along the Pilarcitos fault, the presently inactive fault strand branching westward from the San Andreas fault on the San Francisco peninsula. Probably most of the 300 to 320 km of displacement on the San Andreas fault has taken place along this strand because the presently active strand of the San Andreas fault that lies directly east of the Pilarcitos fault demonstrates only about 26 km of offset of a characteristic limestone belt within the Permanente terrane of the Franciscan assemblage (Bailey and others, 1964, p. 69; M.C. Blake, Jr., oral commun., 1987). The Pilarcitos fault (now truncated to the northwest by the San Gregorio fault) may have its former extension on the ocean side of the San Gregorio-San Andreas fault at about lat 38°30' N. (see fig. 9.4), as proposed by Graham and Dickinson (1978). This proposed extension may have granitic rocks on the southwest side (more than the additional 100 km already discussed) that have no correlatives northeast of the San Andreas fault, unless the total offset on the San Andreas system substantially exceeds 300 km or unless granitic rocks underlie thrust blocks of Franciscan assemblage near the south end of the Great Valley (see preceding paragraph). There may be other, unidentified faults within the Salinian block that allow for this additional offset.
      Offsets of geophysical anomalies along the San Gregorio- Hosgri fault support a total right-lateral movement of about 100 to 130 km that has been added to the total offset on the San Andreas fault system north of its junction with the San Gregorio fault. An offset gravity high (Silver, 1974) is located on the northeast side near Point Sur (point C₁ at lat 36°30' N.) and on the southwest side at Ano Nuevo (point C₂ at lat 37°15' N.), with an offset of 105 to 130 km as remeasured by Graham and Dickinson (1978). We prefer an offset of 105 km (max 115 km) because any larger displacement will place the offset extension of the Pilarcitos fault on land north oflat 38°30' N., where no such fault is known.
      Displacements along the San Andreas fault north of lat 38°30' N. have proved difficult to measure, both because the rocks exposed southwest of the fault near Point Arena have no obvious correlatives on the opposite side of the fault and because most of the fault trace is concealed beneath the Pacific Ocean (Griscom and Jachens, 1989). The rocks cropping out southwest of the fault near Point Arena are Upper Cretaceous and Tertiary marine sedimentary rocks, together with some older spilitic volcanic rocks that may be part of the Franciscan assemblage (Wentworth, 1968). Little basement information from rock samples is available in the shelf areas west of the San Andreas fault between Point Arena and the Mendocino fault. An important well 20 km west of Point Arena (fig. 9.4) recovered quartz-mica schist and slate basement cuttings (Hoskins and Griffiths, 1971) at a depth of about 1.43 km. This description resembles rocks either from the eastern metamorphosed Franciscan assemblage south of San Francisco or from roof pendants in the Salinian block, implying that a major strike-slip fault is located between the well and Point Arena. The proposed Pilarcitos fault extension is thus interpreted to lie here between the well and the coastline on a major fault shown by McCulloch (1987). Location of this proposed Pilarcitos fault extension farther northwest than Point Arena is uncertain, although the fault presumably continues to the former triple junction. McCulloch (1986; 1987, fig. 2b) described a boundary, termed the "Navarro discontinuity," trending east-west from the Point Arena area to the lower continental slope, on the basis of regional differences in magnetic pattern and physiography; this boundary may be the fault extension or an earlier strike-slip fault of this system. Griscom and Jachens (1989) also hypothesized a more northwestward extension, approximately colinear with the fault segment south of Point Arena, following a fault trace interpreted from seismic-reflection profiles (McCulloch, 1987, fig. 14).
      Distinctive gravity and magnetic anomalies characterize the poorly known shelf area lying north of Point Arena and between the San Andreas fault and the proposed Pilarcitos extension (figs. 9.2, 9.3). The sources of these anomalies lie in the basement, with their upper surfaces at the basement interface below Tertiary sedimentary rocks, according to geophysical models and basement-depth calculations. A major gravity high (+ 20 mGal) is located near lat 40° N. (point D₂, fig. 9.4). We believe that the high-density basement rocks which cause this high extend southward along the west side of the San Andreas fault at least as far as at Point Arena (E₂, fig. 9.4), even though the gravity values on the map fall below 0 mGal along the southern part of this reach. The basement along the postulated southern part of the high is mantled by 1 to 3 km of Tertiary sedimentary rocks (Hoskins and Griffiths, 1971), which probably cause gravity lows (-15 to -30 mGal) that here mask the gravity high caused by the basement. Two magnetic anomalies on the shelf are truncated by the San Andreas fault at point F₂ and at a place a few kilometers south of point G₂ (fig. 9.4), which is located where the steepest gradient on the northeast side of the second anomaly is truncated by the fault (see McCulloch, 1987, fig. 17).
      Our search for geophysical anomalies or features matching points D₂, E₂, F₂, and G₂ on the opposite (northeast) side of the San Andreas fault (see Griscom and Jachens, 1989) began with the observation that gravity highs are not characteristic of much of the Franciscan assemblage and are observed only extending along the San Andreas fault between approximately lat 37° and 38° N. (fig. 9.2). We have selected points D₁ and E₁ (fig. 9.4) as the approximate limits of the gravity highs on the northeast side and propose to correlate these points and their connecting strip of high gravity with the corresponding points D₂ and E₂ and associated gravity high discussed in the previous paragraph. The positions of these points (D and E) along the fault vary in reliability but are probably no more accurate than ±20 km; point E₂ is the most uncertain. The total offset of the gravity high by the San Andreas fault is thus about 250±40 km. We have used the gravity results to explore our magnetic-anomaly map (fig. 9.3) for additional correlations. Only one correlation was found within an offset range of 200-300 km. We suggest that point F₁, marking the end of a truncated magnetic high passing through San Francisco, correlates with point F₂ and that point G₁, the truncated end of a magnetic gradient more than 50 km long, correlates with the other truncated gradient at point G₂. The locations of points F₁, F₂, and G₂ along the fault are accurate to within about ±5 km. Point G₁ is located a few kilometers too far to the southeast because a short northwestward extension of the feature was recently cut off by the young segment of the San Andreas fault in the San Francisco peninsula area and now lies between the San Andreas and Pilarcitos faults. The offset of points F₁ and F₂ is 250 km; the offset of points G₁ and G₂ is 263 km. The magnetic anomalies truncated at points F₁ and G₁ are associated with northwest-striking belts of mafic and ultramafic rocks within the Franciscan assemblage and are best shown on the more detailed maps by Brabb and Hanna (1981) and Griscom and Jachens (1989).
      We conclude that the total offset of the pairs of corresponding magnetic features is approximately 250±10 km. Of this offset, about 105 km is attributable to the San Gregorio-Hosgri fault, leaving only 145 km for the San Andreas fault south of its junction with the San Gregorio fault. Because the total San Andreas offset south of the San Francisco peninsula is considered to be much larger, namely, about 300 km, we suggest that the missing 155 km is predominantly accounted for by former movement on the Pilarcitos fault and its proposed north-westward extension, which is thought to intersect the San Andreas at about lat 38°30' N., as described above. This early Pilarcitos fault was thus formerly the main strand of an earlier San Andreas fault system that lay to the west of both magnetic features F and G (that is, before they were offset by faulting). Note that this interpreted fault-movement history and the subsequent plate-tectonic analysis all depend on the correctness of the correlation between the pairs of offset magnetic and gravity anomalies on the San Francisco peninsula and northwest of Point Arena. The magnetic and gravity anomalies northwest of Point Arena and west of the San Andreas fault are such conspicuous features and so obviously truncated by the San Andreas fault that we would expect to find their counterparts somewhere on the opposite side of the fault. Although we can find no alternative correlations for these anomalies other than those indicated in figure 9.4, we are aware that they may not correlate with any anomalies on the opposite side of the fault, although we consider this noncorrelation to be unlikely.
      Additional information on offset along the proposed Pilarcitos fault extension is provided by interpretation of two strong magnetic anomalies on the northeast side of the San Andreas fault in central California (H₁ at lat 35°30'-35°40' N. and lat 36°00'-36°15' N., respectively). The source bodies for both anomalies appear to be truncated by the fault, and interpretations of the gravity and magnetic fields over both bodies suggest that they are composed of serpentinite (Hanna and others, 1972; Griscom and Jachens, in press). The most likely candidates for corresponding magnetic features on the southwest side of the fault system are the poorly defined anomalies at points H₂ and I₂ west of Point Arena. The magnetic field is poorly known in this area, and so anomaly locations and shapes may not be accurate, but the offset is approximately 435 km from points H₁ and I₁. This distance can be obtained by summing an assumed 320 km for offset on the San Andreas fault south of San Francisco plus 115 km offset on the San Gregorio- Hosgri fault. The location of point I₂ supports the Navarro discontinuity as a possible continuation of the proposed Pilarcitos fault extension, or some earlier continuation. We suggest that the large magnetic-high area bounded by the 500-nT contour and located 25 km south of point I₂ (fig. 9.3) may represent a southerly extension of anomaly I, which north of point I₁ extends for 400 km along the east side of the Coast Ranges. There appear to be no satisfactory alternative anomalies for correlation with points H₁ and I₁ along the southwest side of the present San Andreas fault near Point Arena.

      In the previous sections, we have discussed how potential-field data provide information on the three-dimensional configuration of the San Andreas fault and on the various offsets along member faults of the San Andreas system in relation to plate tectonics. Here, we interpret the potential-field expression of the two ends of the San Andreas fault system in relation to plate tectonics and lithospheric thickness.

      At the north end of the San Andreas fault off Cape Mendocino, three lithospheric plates (the Juan de Fuca, Pacific, and North American) meet at the Mendocino triple junction, where a trench meets two transform faults, the San Andreas and Mendocino faults. Along this trench to the north, the Juan de Fuca plate is subducting eastward beneath the North American plate. The geometry of this subducted plate has important implications (Jachens and Griscom, 1983) for an understanding of the Mendocino triple junction and its effects on the tectonics of California. During approximately the past 29 Ma, this triple junction has been migrating relatively northwestward along the coast of California from a latitude near Los Angeles (see fig. 3.11; Atwater, 1970; Atwater and Molnar, 1973). As this incipient San Andreas transform fault lengthened over time, eastward subduction continued to the north of the migrating triple junction.
      During 29-23 Ma, the major fault of the San Andreas system was probably situated near the base of the continental slope, where an accreted wedge of Miocene(?) sedimentary rocks (McCulloch, 1987) accumulated between lat 35° and 40° N., presumably because of oblique subduction from transpressive forces between the plates. This now-inactive fault forms a contact between oceanic and continental crusts (fig. 9.4) that have major differences in magnetic properties (fig. 9.3). The oceanic crust displays the typical oceanic lineated or striped magnetic pattern striking north-south and northeast, with interruptions striking east-west or southeast that are caused by transform faults. The continental crust adjacent to this inactive fault is magnetically rather smooth and featureless. The magnetic boundary between oceanic and continental crust west of the San Andreas fault is very abrupt in comparison with active subduction zones (compare the magnetic expression of the Cascadia subduction zone off Oregon in Bond and Zietz, 1987); the oceanic stripes terminate at the base of the continental slope, even though reflection profiles show oceanic crust continuing farther east beneath the slope (McCulloch, 1987). The low convergence rate of oblique subduction and the time available since the fault became inactive may have allowed the concealed or subducted oceanic crust to heat up sufficiently beneath the continental margin to destroy the remanent magnetization that causes the stripes.
      During early Miocene time (23 Ma), the motion along the transform must have been essentially strike slip and was substantially transferred to the present San Andreas fault system in central California. Without subduction east of the elongating transform, an ever-enlarging triangular hole or window (Dickinson and Snyder, 1979) developed in the slab of lithosphere subducted beneath the continent. This window model is also applicable to the time interval (29-23 Ma) but needs modification to include effects of transpression along the earlier San Andreas fault. The north boundary of this window is the subducted south edge of the Juan de Fuca plate, and hot upwelling asthenospheric material presumably occupies the window. The south edge of the Juan de Fuca plate lies beneath the North American plate at the shore about 20 km south of Cape Mendocino and can be identified by an east-west magnetic anomaly (Griscom, 1980a), as well as by the distribution of seismicity (Hutchings and others, 1981). This position coincides with a steep gravity gradient (here called the Cape Mendocino gravity anomaly) that slopes downward into a large gravity low (-50 mGal) to the north and east. The spatial coincidence between the position of the Cape Mendocino gravity anomaly at the coast and the place where the south edge of the Juan de Fuca plate passes beneath the coastline strongly suggests that this gravity anomaly reflects the south edge of the subducted plate (fig. 9.4). At least three other characteristics (Jachens and Griscom, 1983) of the anomaly support this interpretation. (1) The southeastward trend of the gravity anomaly and then its change to easterly are consistent with the directions of present and past relative motions between the Juan de Fuca and Pacific plates (Nishimura and others, 1981; Wilson, 1986). (2) The gravity anomaly broadens and is less steep toward the southeast, suggesting that its source progressively deepens in this direction; calculated depths along the anomaly to the end of the southeast-trending segment define a line plunging approximately 9° SE. with a depth of only 6 km at the coastline corresponding well to the 8-km depth estimated from aeromagnetic data (Griscom, 1980a). (3) A cross section across this anomaly, using the above depths together with reasonable densities and thicknesses for the subducted Juan de Fuca plate and the asthenospheric window fill to the southwest, produces a calculated gravity model (Jachens and Griscom, 1983) in good agreement with the observed gravity field. We draw the following conclusions from the gravity data (Jachens and Griscom, 1983).

      Interpretation of geologic and geophysical data for the San Andreas fault system north of San Francisco (Griscom and Jachens, 1989) suggests that eastward migration of the plate boundary from its presumed original position at the base of the continental slope to its present position at the San Andreas transform fault may have occurred by means of a series of eastward jumps of the Mendocino triple junction covering a total distance of about 150 km during the past 29 Ma. Our general model for the history of this triple junction is one of successive eastward jumps, with sustained periods at each position while significant strike-slip motion occurred on the various transform fault systems, including the San Andreas fault. We are aware, however, that the picture in detail may have been far more complex. The present position of the San Andreas fault north of San Francisco is thus interpreted to be relatively youthful. The triple junction was initially situated near the base of the continental slope at the northwest end of the Miocene(?) accreted wedge (but far to the south of its present latitude); the basal fault (McCulloch, 1987) below the subduction complex is shown as a toothed line in figure 9.4 because of the thrust component in this transform fault. The triple junction is interpreted to have been subsequently situated at the north end of the proposed Pilarcitos fault extension and then to have jumped eastward a minimum of about 100 km to the present San Andreas fault trace at what is now approximately lat 38°20' N. on the North American plate. When this jump occurred, the three faults that formed the junction all had to readjust; the simplest scenario is as follows: (1) The Mendocino fault was extended on strike farther eastward, for the distance of the jump, about 100 km; (2) a new segment of the San Andreas fault broke obliquely through the Franciscan assemblage to the northwest (severing the correlated geophysical anomalies described above) and extended from the new triple junction to the junction of the newly formed (or soon to be formed) San Gregorio fault with the Pilarcitos fault, a distance of about 250 km; and (3) the surface trace of the subduction zone north of the triple junction also jumped eastward 100 km, thus abruptly isolating a thin triangular slab of Franciscan assemblage (probably less than 15 km thick) from the North American plate. This postulated triangular slab of rocks is now gone, most likely subducted away. Further complexity is provided by the King Peak subterrane of the King Range terrane (McLaughlin and others, 1982), which is an elongate mass of turbidites, about 45 km long, just south of Cape Mendocino (fig. 9.4) that is believed to have been obductively accreted from the west during the early Pleistocene (McLaughlin and others, 1986). The King Peak subterrane may have been detached and transported northwestward from the San Francisco area (just south of lat 38°20' N.) as part of the Pacific plate and then reattached to the North American plate by a very recent local jump of the triple junction westward less than 35 km (McLaughlin and others, 1982); this explanation may account for the anomalously higher thermal metamorphism of this subterrane relative to the terranes that are now adjacent to it. Recent work suggests that the triple junction may be on shore at Cape Mendocino (Clarke, 1988; McLaughlin and others, 1988); if so, the King Peak subterrane may still be essentially part of the Pacific plate. The tectonic interpretation detailed above also requires that the San Gregorio- Hosgri fault first began moving and joined the present San Andreas fault at approximately the same time as or shortly after the eastward jump of the triple junction, and thus cut off the proposed northward extension of the Pilarcitos fault, after which the extension became inactive.
      The proposed 150-km eastward movement of the triple junction can also explain the submarine topography near Cape Mendocino, where the Continental Shelf south of the Mendocino fault extends about 130 km farther west than that directly north of the fault.
      The timing of the jump can be estimated from the horizontal offset of the paired geophysical anomalies, about 250 km, which translates to an age of about 5 Ma, assuming combined strike-slip rates of 4.8 cm/yr (DeMets and others, 1987; Minster and Jordan, 1978) for the San Andreas and San Gregorio faults. This age estimate is crude because it assumes that no other faults were absorbing the relative motion between the two plates. For example, simultaneous movement on the Hayward-Calaveras fault system will cause the computed age of offset to be too young. The eastward jump of the triple junction appears to be associated with a change in stress orientations in this region. The north end of the San Gregorio- Hosgri fault trends about 20° clockwise relative to the older fault traces. In addition, the northward-migrating triple junction subsequently traced out a major right-lateral bend, as shown by the present position of the San Andreas fault north of Point Arena. The central section of this bend is about 100 km long and trends 20° clockwise to the older trace. This change may correlate with the gradual change in absolute motion of the Pacific plate that occurred between 5 and 3.2 Ma (Cox and Engebretson, 1985; Pollitz, 1986), producing a change from strike slip to transpression in this region and a clockwise rotation of 20° (Harbert and Cox, 1986) in the relative-velocity vector for the plate pair, the same angle as the anomalous change in direction for both the San Gregorio fault and the right-lateral bend in the San Andreas fault north of Point Arena. This change in relative motion probably correlates with a change in strike direction of the subducting south edge of the Juan de Fuca plate, as deduced from gravity data (Jachens and Griscom, 1983). Before the jump, this strike was east-west, thus permitting eastward extension of the Mendocino transform fault without interference; after the jump, the strike of the subducting plate edge changed to S. 60° E., making later eastward fault extension more difficult.
      Stratigraphic evidence for the postulated eastward jump of the triple junction about 5 Ma may be sought in the late Miocene and Pliocene stratigraphy of Deep Sea Drilling Project (DSDP) Site 173 (fig. 9.4). Depositional hiatuses occur at 5 and 4.3-3.2 Ma (Barron, 1989), whereas a study of both micropaleontology and tephra beds indicates a hiatus from about 4.4 to 2.8-Ma (Sarna-Wojcicki and others, 1987). McCulloch (1987, fig. 25) believed that the middle Pliocene deformation and minor erosion interpreted from reflection profiles correlate with this 4.4-2.8-Ma hiatus at Site 173. We suggest that the eastward jump of the triple junction about 5 Ma was shortly followed by the middle Pliocene deformation and by the hiatus at Site 173. These two correlative events were thus caused both by the jump and by the simultaneous change in the direction of relative plate motion.

      The San Andreas fault terminates to the southeast in a buried spreading center at the south end of the Salton Sea, where a row of five small siliceous volcanic domes ("buttes") protrude above recent sedimentary deposits of the Salton Trough. These domes, in addition to being associated with a local northeast-striking magnetic high, are situated on the crest of a larger, northwest-trending magnetic high (outlined on fig. 9.4) that is interpreted (Griscom and Muffler, 1971) to be caused by a magnetic mass, 30 km long, 3 to 12 km wide, and about 4 km thick, with its top buried more than 2 km below the surface. This magnetic high is associated with a similarly shaped gravity high (Biehler and Rotstein, 1979), the source of which may partly be the magnetic mass but may also be the relatively high density metamorphosed sedimentary rocks associated with the geothermal area (Elders and others, 1972). The Salton Buttes spreading center probably strikes northeast because the row of domes, the local aeromagnetic and gravity anomalies, and the geothermal area all coincide and strike northeast; (2) this proposed position for the center bisects the larger northwest- trending magnetic high into approximately equal parts interpreted to be new "oceanic" crust; and (3), ideally, a spreading center should trend approximately normal to an associated transform fault. In apparent contradiction, the Brawley seismic zone strikes S. 20° E. from the Salton Sea (Johnson and Hill, 1982) and consists of shallow earthquakes (Severson and McEvilly, 1987) located mostly within the valley fill; this seemingly anomalous direction may be due to accommodation of these overlying, partly decoupled materials to a series of short northeast-trending spreading centers between the Salton Sea and Cerro Prieto, Mexico (see fig. 3.6), on strike S. 20° E. and 100 km distant (Fuis and Kohler, 1984; Sibson, 1987). The large, northwest-trending magnetic mass is interpreted to reflect about 30 km of northwestward spreading along its long axis, in which the spreading was associated with intrusive activity that built up a 30- km-long strip of magnetic mafic rocks and new crust in the lower section of and below the sedimentary fill. This magnetic feature may not be directly comparable to oceanic-crustal anomalies because slow cooling beneath the fill probably results in weak remanent magnetization, unlike the situation for oceanic crust. This anomaly thus may be predominantly caused by induced magnetization.
      The gravity field of the Salton Trough, which is filled with great thicknesses of Cenozoic sedimentary rocks, varies systematically from north to south. An elongate gravity low of -30 to -40 mGal is associated with the sedimentary rocks northwest of the Salton Sea (beyond lat 33°20' N.). Southward along the axis of the trough, the gravity field increases rapidly until the south end of the Salton Sea, where maximum values of 0 mGal are obtained over the presumed spreading center described above. Farther southeast, to the United States-Mexico border, gravity values range from only -10 to -20 mGal, an initially surprising observation because the 3.5 km or more of young, unmetamorphosed sedimentary deposits in this area might be expected to produce anomalies lower than -40 mGal (Biehler, 1964; Griscom, 1980c, p. 20), similar to the gravity expression northwest of the Salton Sea. Biehler (1964) offered two explanations for the missing low: thinner crust or local high-density basement beneath the trough. Seismic-refraction studies (Fuis and others, 1982, fig. 17A) confirm the second explanation and show a deep "subbasement" (density, 3.1 g/cm³) in the trough that extends below about 12-km depth. Using this refraction model as a constraint, a gravity model (Fuis and others, 1982, fig. 20) indicates that the crust beneath the trough is no thinner than that of the bordering mountains a few kilometers to the northeast.

REFERENCES CITED