FAULT ZONES

San Andreas and San Gregorio Fault Subsystem

The great 1906 earthquake, M_w~7.9, the predominant historic seismic event of the San Andreas fault system (SAFS) in northern California, ruptured all currently locked segments of the fault (A1 in Figure 2a), from near the Mendocino triple junction (MTJ) to San Juan Bautista (SJB), where the dominantly creeping segment (A7) begins southward to Parkfield [Thatcher and others, 1997]. The 1906 rupture ( A1) overlaps the independent subsegments A2 and A3, and southward to the point labeled END 1906. Current research into prehistoric events along the northern San Andreas fault indicates that a similar great event probably occurred most recently in the 17th century [Schwartz and others, 1997], thus it is reasonable that our model of the SAFS be structured around slip accumulation and release in great earthquake cycles that are about two centuries long or somewhat longer (e.g., we use 210 yr ~ 5.1 m/0.024 m/yr, the average 1906 slip north of the Golden Gate divided by the long-term slip rate, see below). The 1906 rupture on the San Andreas fault had distinct variations in coseismic slip along strike (Figure 3) based on the improved analyses by Thatcher and others [1997] of the historic triangulation data. The 330-km north coast subsegment of the fault, north of the Golden Gate, averaged 5.1 m of slip excluding the extremely high slip modeled near the MTJ. In contrast, south of the Golden Gate, 1906 slip averaged 3.4 m and 2.5 m for segments of the San Francisco Peninsula and Santa Cruz Mountains respectively. This reduction in 1906 slip at the Golden Gate has important implications for segmentation and for developing a recurrence model that is consistent with regional plate boundary conditions as well as local observations of events and slip rates.

Click for high-resolution image

Figure 2. Location map showing segmentation of faults labeled by codes (e.g., H2) that appear in Table A-1. Segment ends shown as X ; Great Valley blind thrusts as dashed lines with ticks; areal sources by dashed polygons labeled near centers. Faults with slip rate >1 mm/yr shown with thickness proportionate to slip rate. Locations: GG, Golden Gate; HOL, Hollister; MTJ, Mendocino triple junction; SJB, San Juan Bautista.

 

Figure 3. 1906 slip from analysis of triangulation by Thatcher and others [1997]. Lithic patterns indicate extent of hypothetical segments (A2 and A3) where independent rupture is assumed.

The abrupt drop in 1906 slip at the Golden Gate may be explained in part by a ~2-km right stepover here in the San Andreas fault [Cooper, 1973], because such extensional zones tend to inhibit propagation of a rupture. Additionally, the San Gregorio fault (A4) branches away from the San Andreas near the Golden Gate, offshore of the City of San Francisco. The San Andreas slip rate of ~ 22-24 mm/yr north of the Golden Gate drops to ~ 17 mm/yr to the south. We deduce that most or all of the San Andreas decrease of ~5 mm/yr to the south occurs as slip lost to the San Gregorio fault. Thus, the San Gregorio fault is effectively a branch of the San Andreas fault with regard to the accumulation of slip over multiple great earthquake cycles. A cartoon (Figure 4) shows a reasonable model of slip accumulation that satisfies our current knowledge of historical and prehistoric earthquakes on the two fault zones considered jointly. Detailed discussions of individual fault segments follow. The important idea that emerges is that of effective recurrence time (t_e) between earthquakes on a given fault segment, that is the time between earthquakes on that segment considered independently of great or multisegment (M~7.9) ruptures. Especially important is the consideration of how slip on the San Gregorio fault (A4) relates to the recurrence times of events on the Peninsula segment (A2) of the San Andreas fault because jointly they must keep pace with ~5 m of slip per event on the San Andreas fault north of the Golden Gate. Usually, recurrence time only accounts for how often fault-rupture occurs at a particular point along the fault, but ignores whether the magnitude of the earthquake is a 7 or an 8. Probabilistic seismic hazard analysis must consider the effective time( t_e) between events of a particular size on a particular segment.

Figure 4. Cartoon shows hypothetical sequence of slip accumulation along the joint San Andreas, San Gregorio and Sargent fault system.

Peninsula segment (A2)

The San Francisco Peninsula segment of the San Andreas fault has been previously considered for purposes of earthquake forecasting to extend from north of the prominent bend in the fault through the Santa Cruz Mountains to south of the Golden Gate [Working Group on California Earthquake Probabilities, (WGCEP) 1988; 1990]. The occurrence of a major earthquake in this vicinity in 1838 [Louderback, 1947] and the apparent deficiency of slip in 1906 suggested that M7 events were likely on the Peninsula between great earthquakes. Historical evidence of the 1838 earthquake does not strictly constrain the source to be the San Andreas fault, so the segmentation model is necessarily speculative. In approaching segmentation, the WGCEP [1990] did not have the advantage of Thatcher and others' [1997] analysis of 1906 slip variation that makes the Golden Gate a simple choice for a northern segment boundary. Arguments about the southern segment boundary become complicated by details of fault geometry, but for our purposes the northern end of the 1989 Loma Prieta rupture seemed a reasonable and cautious choice of segment boundary.

Hall and others [1995] measured slip of ~1.6 m on a buried offset stream channel at the Filoli trenching site that they associated with the penultimate earthquake on the Peninsula segment (event preceding 1906). Although the date of this penultimate event was not constrained stratigraphically, the 1838 earthquake is a permissible interpretation and seems reasonable. Because 1906 slip was ~2.5 m near Filoli, their data support the hypothesis of two distinct characteristic events for this segment (e.g., A1-type of M_w 7.9 and A2-type of M_w 7.1 as described in the database; Table A-1 in Appendix A). For a M_w 7.1, independent Peninsula segment rupture (in 1838?), using the area formula of Wells and Coppersmith [1994] and other relationships described in our standard methodology, we expect average slip of 1.5 m on an 88-km-long rupture.

Slip rate for this segment of the San Andreas fault is still not constrained as well as for some other major faults. However, the Filoli site has also yielded a minimum slip rate of 14.8 ± 2.7 mm/yr over a time of 2.1 ka [Hall and others, 1995]. As is common with trenching data, the offset features used to measure slip rates do not span the entire active fault zone, so that the full slip rate could be greater. A higher estimate of 19 mm/yr derives from dividing the dextral slip of ~1.6 m accompanying the 1989 Loma Prieta earthquake [Lisowski and others, 1990] by the 83 yr since the 1906 earthquake which is assumed to have fully relieved the dextral strain, but as we discuss below this incorrectly assumes that 1989 is a characteristic San Andreas fault event. Another approach is to take the well-determined value of slip rate for the central San Andreas fault of ~34 mm/yr [Sieh and Jahns, 1984] and deduct from it the ~17 mm/yr of strain that is conveyed to the southern Calaveras fault system [Savage and others, 1979], leaving ~17 mm/yr for the San Andreas fault northward of San Juan Bautista. We have adopted 17 ± 3 mm/yr for the segment A2 because we find it consistent within expressed limits of error considering a broad range of reasonable arguments from local data, regional kinematics and strain, and also plate boundary and VLBI constraints as will be discussed later. Recurrence time is not yet established by prehistoric evidence on the Peninsula, thus we had to rely on a hypothetical strain accumulation model (Figure 4). In this model, the Peninsula keeps pace with the SAF slip accumulation on the north coast by having an equal number of San Gregorio events each of ~2 m of slip and independent Peninsula San Andreas events of ~1.6 m at an effective recurrence time (t_e) for A2 type events equalling two recurrence times of great SAF earthquakes for each, hence ~ 400 yr.

Santa Cruz Mountains segment (A3)

We have adopted the extent of the M_w 6.9 Loma Prieta earthquake source region of 1989 to define the Santa Cruz Mountains segment, however much about this event does not fit the ideal of a characteristic earthquake, especially the sizable dip-slip component and the mismatch of the 6-17 km deep, dipping 1989 rupture surface determined from aftershock locations with a near-vertical orientation generally expected for ruptures associated with the main trace of the San Andreas fault that ruptured in 1906. It remains uncertain whether there really is a characteristic earthquake for this segment that is independent of great San Andreas fault earthquakes. The location of the large 1865 earthquake may not have been on the San Andreas fault [Tuttle and Sykes, 1993]. Schwartz and others [1997] find no evidence for any earthquake except the 1906 rupture on the main trace of the San Andreas fault in this segment since the 17th century. Nevertheless, a high long-term slip rate and low slip here in 1906 require that earthquakes occur frequently near here, either on the San Andreas fault or other nearby structures. Thus we assume an idealized event with the same dextral slip and rupture extent as the 1989 event but having an effective recurrence time (t_e) of 400 yr. Such events on a Santa Cruz Mountain segment could, taken in combination with San Gregorio fault events and great (e.g., 1906) events, keep this segment in synchrony with the cycles of ~5-m of strain accumulation and release that we assume for the San Andreas fault north of the Golden Gate. Slip rate of 14 mm/yr reflects a loss of ~3 mm/yr dextral slip from the ~17 mm/yr assumed available on the Peninsula segment of SAF to the Sargent fault and a complex of other known and unknown structures in the Santa Cruz Mountains.

San Gregorio fault (A4, A5)

The north end of the San Gregorio fault segment (A4) is generally accepted as a branching connection to the San Andreas near the Golden Gate. Because most of this fault zone is offshore, the map details of the active trace are obscure but allow that the trace may be generally straight for the 129 km from the Golden Gate branch point south to Monterey Bay where a 2-3 km right stepover exists. Although slip rate data are not available for the San Gregorio fault in the Sur region (A5), the southernmost San Gregorio fault system or Hosgri fault zone near San Simeon are believed to have a lower rate than segment A4. Because the Monterey Bay-Tularcitos fault zone (L04) appears to branch off near this right stepover, we assume that it is kinematically reasonable that the southward decrease in slip on the San Gregorio fault zone may occur at least partially near this branch point and that farther southward some of this slip may next transfer to the Rinconada fault (L01).

Quaternary and Holocene slip rates along the San Gregorio fault have been difficult to constrain narrowly, partly because much is offshore and because much of the fault has highly complex geometry. For the northern segment (A4) we assume ~5 mm/yr because preliminary work at Seal Cove on two different-aged markers (see Table A-1 for references) seems to force this result. Other sites permit a broader range of slip rate, so we must regard this rate as preliminary. The southern segment (A5) has no slip rate data, but to the south data support a rate of ~3 mm/yr (see Table A-1).

Estimates of average recurrence time of earthquakes, loosely constrained by fault offset on archaeological layers at the Seal Cove sites, range from 350 to 680 yr for the last two events (G. D. Simpson and others, 1995, writ. com.) Horizontal offset of these layers suggest that events have coseismic slips of >2 m. We use a hypothetical 400 yr effective recurrence time estimated from our hypothetical San Andreas-San Gregorio slip accumulation cartoon (Figure 4). Our standard methodology yields recurrence of 330 yr and slip of 1.7 m. Our hypothetical recurrence time is a moderate value and errs on the side of caution, but we recognize that considerably longer recurrence times also satisfy existing data. For the Sur region segment (A5) we determined all segment parameters by our standard methodology and by comparison to the adjacent segments: the estimated 400-yr recurrence is coincidentally in phase with the segment to the north (A4) and we use the ~3 mm/yr slip rate of the Hosgri segment which lies to the south of a major structural discontinuity (Table A-1).

Related faults

The Sargent fault (A6) has ~3 mm/yr dextral creep rates measured (see Table A-1) in its central part where traces are distinctly Holocene and dominantly right-lateral in their geomorphic appearance [Bryant and others, 1981]. Evidence for recency is more obscure in the steep and heavily vegetated terrain to the north where a larger dip-slip component is required to accommodate volume problems at the junction with the San Andreas fault. Our chosen source model is for a hypothetical event using a maximum length and applying our standard methodology. Our hypothetical Sargent event is intended in part as an approximate representation of additional hazard that probably exists in the region where the San Andreas fault bends and splays in the Santa Cruz Mountains. This additional hazard is not strictly confined to the main trace of the San Andreas. Some minor faults such as the Zayante (L10) and Shannon-Monte Vista (L07) also address this issue, but are of nearly negligible impact compared to the Sargent and related faults. Preliminary values of slip rate (0.6 mm/yr) and recurrence time (1200 yr) have been suggested by Nolan and others [1995] from a paleoseismic investigation on the southernmost Sargent fault, however both values derive from speculative interpretations using vertical separations and a broad range of low-angle rakes on slickensides to infer slip vectors. We use the larger slip rate inferred from measured surface creep as a minimum and the resulting short recurrence time from our standard methodology because they are not excluded by the preliminary paleoseismic work and are a more cautious assumption for hazard analysis.

Preliminary versions of our database assumed a short segment of the San Andreas between the Santa Cruz Mountains and the southern end of the 1906 rupture at San Juan Bautista. A M~6 earthquake occurred in this region in 1890, but was not indisputably located on the San Andreas. Such an event has little impact on the hazard analysis because it is in the range that will be reasonably modeled as background seismicity. Hence this speculative segment was deleted as a source of independent large earthquakes. In 1836 a sizable earthquake, M~6.5 ± 0.5, apparently happened on some fault in the region around San Juan Bautista; this event had been mistakenly associated with the Hayward fault [Toppozada and Borchardt, 1997]. The San Andreas south of this segment is treated as fully creeping (A7) and events less than about M~6 will be adequately modeled as background seismicity.