Calaveras Fault Subsystem

Southern Calaveras fault (C1)

At its south end, the southern Calaveras fault (C1) diverges from the fully creeping central San Andreas fault (A7) gaining nearly half of the central San Andreas' 34 mm/yr slip rate. The southern Calaveras branches into the Northern Calaveras and Hayward faults by the south end of Calaveras Reservoir (CR, Figure 5). The southernmost Calaveras-Paicines fault zone, ~80-km long, that extends from San Benito to Coyote Reservoir has previously been assumed not to have large earthquakes because its creep rate is high (12-17 mm/yr) and matches its long-term or geologic slip rate within the limits of uncertainty [Bakun and others, 1986; Harms and others, 1987; Perkins and Sims, 1988; Sims, 1991]. South of Hollister the Calaveras is located only 3-5 km away from and parallel to the San Andreas fault, thus may not be an independent source of large earthquakes. We suggest that the 1984 Morgan Hill earthquake, M_w 6.2, is a reasonable maximum magnitude event to occur in these segments, so rather than apply a detailed segmentation model we chose to assume that such an event has an equal likelihood of occurring anywhere along the entire southern Calaveras fault.

Northern Calaveras fault (C2)

The Northern Calaveras fault (C2) extends from Calaveras Reservoir (CR, Figure 5) to its north end, a right stepover to the Concord fault (C3). Based on simple geometric interpretation of the surface trace, this fault can be divided in three segments of roughly similar lengths [Simpson and others, 1993]. Thus, we have considered the reasonableness and practical impact of assuming that shorter single segment ruptures occur in addition to ruptures of the entire fault (C2).

The San Ramon segment, C2c, has had at least one sizable earthquake historically in 1861 of roughly M6 (±0.5) and an apparently-associated ground rupture of about 13 km length from near Elworthy Ranch (ER, Figure 5) [Rogers and Halliday, 1993] to near Dublin Canyon (DC, Figure 5). Northward from Elworthy Ranch the fault has a less distinct geomorphic expression and trenching near Alamo (CA, Figure 5) [Simpson and Lettis, 1994] reveals no distinct evidence of Holocene slip on the main trace in well-stratified deposits of late Pleistocene to Holocene age. The northern end of the fault is a right step over of a few kilometers to the Concord fault (C3). Figure 5 shows recent microearthquakes that mark possible subsurface connections between the Northern Calaveras and Concord faults (M³2, 1989-1995). Creep rate on this segment (Figure 5, triangle labeled SF-19 ) has increased significantly since the 1989 Loma Prieta earthquake, from near zero (0.4 ± 0.1 mm/yr, 1981-1989) to 2.7 ± 0.2 mm/yr (1989-1996) (J. S. Galehouse, writ. comm., 1996), similar to the ~3 mm/yr creep rates on the Concord fault and the Calaveras fault in Sunol.

The middle segment of the fault (C2b) may have ruptured in a M~6 earthquake in 1864, but the location of the event is poorly known and there was no report of surface rupture as there was for the 1861 earthquake. The length of this segment is controlled by two distinct right stepovers in the fault and thus could be expected to have its own earthquakes of M~6.

The southern subsegment of the Northern Calaveras, near Sunol Valley and Calaveras Reservoir (CR), appears to be the dominant source segment of much larger ground-rupturing earthquakes at Leyden Creek site (LC in Figure 5; Kelson and others, 1996). We assume that this Sunol Valley subsegment only ruptures along with the San Ramon (C2c) and Amador (C2b) segments, hence we do not have a separate entry in Table A-1 for an independent C2a source. Greater strength of this segment may be attributable to a 0.7-0.9 km left stepover located between Leyden and Welch Creeks that could act as a compressional asperity. A cartoon (Figure 7) illustrates how slip could accumulate at a rate of 6 ± 2 mm/yr over centuries as a combination of larger slip in major earthquakes that break the entire fault length of ~50 km and contributions of lesser events, such as the 1861 earthquake that break a smaller part of the fault. Our database includes both types of events with the effective recurrence of each weighted to reflect a combination of events that accumulates slip evenly along the fault and agrees with the current understanding of each segment's behavior from paleoseismologic and historical observations.

Click for high-resolution image

Figure 7. Hypothetical slip accumulation along Northern Calaveras fault. See text for further discussion.

Recurrence time between major events could be as long as 550 ± 300 yr at Leyden Creek [Kelson and others, 1996], but our methodology yields 170 yr. Bonilla and Lienkaemper [1990] showed that even large historic surface ruptures sometimes cannot be recognized in trenches because of a variety of factors such as the contrastiness of geologic materials and effects of soil forming processes. To address this serious difference in recurrence estimates, a rounded intermediate value of 400 yr was adopted as a cautious estimate that reflects our considerable uncertainty in the fault's actual behavior. In the cartoon ( Figure 7), two M_w6.1 events on both the San Ramon and Amador segments can occur for each M_w7.0 occurring on the entire Northern Calaveras fault. This model is only one of many possibilities that agree with the sparse and uncertain historical and paleoseismological data. An alternative approach for hazard analysis is to distribute the same seismic moment rate over a range of magnitudes and recurrence times [Frankel and others, 1996; Petersen and others, 1996a, b].

Concord-Green Valley faults

Although the overall zone is highly complex and includes some significant active secondary traces such as the Cordelia fault [Harlan Tait Associates, 1994], most of the working group felt that the Concord-Green Valley fault zone is likely to fail in one major event, but a few were concerned that a stepover capable of stopping ruptures might exist under Suisun Bay. A single event model (C34, Table A-1) involves a M_w6.9 every 180 yr, but the opposite alternative of independent Concord (C3) and Green Valley (C4) fault events would produce a M_w6.5 every 110 yr and a M_w6.7 every 150 yr respectively. In view of the absence of events on the two faults in the ~150+ yr historic record, the latter alternative seems somewhat suspect. The model given in our database is an average of these two opposing assumptions. Applying a Gutenberg-Richter distribution using a single segment model (C34) for maximum magnitude M_w6.9 every 180 yr achieves a similar result to our model [Frankel and others, 1996; Petersen and others, 1996a, b].