Bulletin of the Seismological Society of America, Vol. 95, No. 3, pp. 844–860, June 2005, doi: 10.1785/0120040134
The Most Recent Large Earthquake on the Rodgers Creek Fault,
San Francisco Bay Area
by Suzanne Hecker, Daniela Pantosti, David P. Schwartz, John C. Hamilton,
Liam M. Reidy, and Thomas J. Powers
Abstract The Rodgers Creek fault (RCF) is a principal component of the San
Andreas fault system north of San Francisco. No evidence appears in the historical
record of a large earthquake on the RCF, implying that the most recent earthquake
(MRE) occurred before 1824, when a Franciscan mission was built near the fault at
Sonoma, and probably before 1776, when a mission and presidio were built in San
Francisco. The ﬁrst appearance of nonnative pollen in the stratigraphic record at the
Triangle G Ranch study site on the south-central reach of the RCF conﬁrms that the
MRE occurred before local settlement and the beginning of livestock grazing. Chro-
nological modeling of earthquake age using radiocarbon-dated charcoal from near
the top of a faulted alluvial sequence at the site indicates that the MRE occurred no
earlier than A.D. 1690 and most likely occurred after A.D. 1715. With these age
constraints, we know that the elapsed time since the MRE on the RCF is more than
181 years and less than 315 years and is probably between 229 and 290 years. This
elapsed time is similar to published recurrence-interval estimates of 131 to 370 years
(preferred value of 230 years) and 136 to 345 years (mean of 205 years), calculated
from geologic data and a regional earthquake model, respectively. Importantly, then,
the elapsed time may have reached or exceeded the average recurrence time for the
fault. The age of the MRE on the RCF is similar to the age of prehistoric surface
rupture on the northern and southern sections of the Hayward fault to the south. This
suggests possible rupture scenarios that involve simultaneous rupture of the Rodgers
Creek and Hayward faults.
A buried channel is offset 2.2 (Ⳮ1.2, ⳮ0.8) m along one side of a pressure ridge
at the Triangle G Ranch site. This provides a minimum estimate of right-lateral slip
during the MRE at this location. Total slip at the site may be similar to, but is probably
greater than, the 2 (Ⳮ0.3, ⳮ0.2) m measured previously at the nearby Beebe Ranch
The Rodgers Creek fault (RCF) is a major strand of the
right-lateral San Andreas fault system in the San Francisco
Bay area. The RCF is the northern continuation of the Hay-
ward fault across a right-releasing step beneath San Pablo
Bay (Fig. 1) and shows evidence of geologically recent ac-
tivity along at least 60 km of its length from the bay to near
the town of Healdsburg. The largest instrumentally recorded
earthquakes in the vicinity of the RCF are the 1969 M
and 5.7 Santa Rosa earthquakes (Fig. 2) (Bolt and Miller,
1975). Microearthquake activity, which has been monitored
since 1969, deﬁnes a steep northeast-dipping trend to a depth
of ⬃10 km along the fault north of Santa Rosa (Fig. 2b).
South of Sonoma Mountain, the fault is seismically quies-
cent, suggesting the presence of a seismic gap (Fig. 2a)
(Budding et al., 1991; Wong, 1991).
The RCF is expressed at the ground surface by abundant
youthful geomorphology and has been long recognized as
an active fault (e.g., Willis and Wood, 1922; Weaver, 1949;
Brown, 1970; Huffman, 1971; Herd, 1978; Hart, 1982,
1992). Only since the 1990s, however, has enough infor-
mation on the RCF been developed to begin characterizing
its slip rate, recurrence, and earthquake potential (Budding
et al., 1991; Schwartz et al., 1992) and to include the fault
in assessments of the probability of large earthquakes in the
region (Working Group on California Earthquake Probabil-
ities, 1990, 1999, 2003). Paleoseismic evidence from studies
at the Beebe Ranch site (Fig. 1) on the south-central reach
of the fault has provided a slip-rate estimate of 6.4–10.4 mm/
yr, an average recurrence interval of 131–370 years
(Schwartz et al., 1992), and a surface-offset measurement
The Most Recent Large Earthquake on the Rodgers Creek Fault, San Francisco Bay Area 845
Figure 1. Shaded relief map of the San Francisco
Bay region showing major faults and locations men-
tioned in the text. Faults are Holocene age from Jen-
nings (1994), except for the trace of the Rodgers
Creek fault north of Windsor Creek (commonly
known as the Healdsburg fault), which is according
to Hart (1998) and is provisionally shown here as Ho-
locene. The Triangle G Ranch trench site is ⬃1km
northwest of an earlier trench site at Beebe Ranch.
Other trench sites mentioned in the text are labeled as
follows: PR/FC, Panorama Reservoir and Foss Creek
Detention Basin (⬃1.5 km apart); MV, Mira Vista;
MT, Montclair; TL, Tyson’s Lagoon. A triangle lo-
cates the center of seismic energy release for the 1898
“Mare Island” earthquake based on available intensity
data, and a dashed oval bounds the epicentral region
at a 50% level of conﬁdence (Bakun, 1999). A star
locates approximate epicenter of the M⬃6.8 1868
earthquake on the Hayward fault (Bakun, 1999);
dashed line shows minimum extent of 1868 rupture.
for the most recent earthquake (MRE) of 1.8–2.3 m (Budding
et al., 1991). Budding et al. (1991) estimated a minimum
elapsed time since the MRE of 182 years using historical
records. At the Triangle G Ranch site, ⬃1 km north of the
Beebe Ranch site, Schwartz et al. (1992) identiﬁed possible
evidence of three events and concluded that the MRE oc-
curred after A.D. 1438–1654.
This article describes new data from the Triangle G
Ranch site, including new ﬁeld observations, radiocarbon
dating, and pollen analysis, that, together with the record of
historical seismicity, provide signiﬁcantly tighter constraints
on the timing of the MRE. The age of the MRE is an impor-
tant parameter for developing rupture scenarios and com-
puting probabilities of future major earthquakes. Constraints
on the geometry of a buried channel allow us to estimate a
minimum amount of slip during the MRE at the study site.
The Triangle G Ranch site lies in a valley ﬂat along
Rodgers Creek, where the fault is expressed as a smallnorth-
west-trending pressure ridge within a ﬂoodplain (Fig. 3). The
pressure ridge, which is underlain by a fault-bounded slice
of Tertiary volcanic deposits, is 3 to 8 m wide and less than
1 m high. The distal portion of an alluvial slope impinges
on the site from the southeast (Fig. 3). Rodgers Creek, which
parallels the fault upstream of the site, makes a right bend
around the northern nose of the pressure ridge and diverges
from the fault downstream of the site.
Schwartz et al. (1992) excavated seven trenches parallel
to and across the RCF to expose stratigraphic and structural
relations at the site. In addition to these, we excavated two
trenches in 1998–1999 (trench 9 and trench 5b, Fig. 3). Our
interpretations here are based primarily on trenches 3 and 9,
which provided the best evidence of the MRE. Other trenches
at the site showed evidence of older faulting events, but the
ages of these earthquakes are poorly constrained.
Stratigraphy at the site consists of ﬂuvial and alluvial-
slope deposits that lie in fault and depositional contact with
colluvium underlying the pressure ridge. The main colluvial
deposit, unit 9, is a poorly sorted gravelly sand in a clayey
silt matrix (Figs. 4 and 5). The deposit is brown with yellow,
red, and pink variegations and appears to be derived from
the late Tertiary Sonoma Volcanics, the main bedrock unit
along the length of the RCF. A very dark grayish-brown A
soil horizon, about 30 cm thick, has formed on unit 9. A
dark brown-to-black deposit of possible colluvial origin
(unit 8) lies in fault and depositional contact with unit 9
along the southwest side of the pressure ridge beneath de-
posits of Rodgers Creek (Fig. 4b,c). Unit 8 comprises poorly
sorted gravel and sand in a clayey matrix and is unstratiﬁed.
Alluvial deposits exposed in trenches 3 and 9 that pre-
date the MRE are grouped as unit 4 and consist of gravelly
846 S. Hecker, D. Pantosti, D. P. Schwartz, J. C Hamilton, L. M. Reidy, and T. J. Powers
Figure 2. (a) Seismicity around Rodgers Creek fault (1969–2003) showing earth-
quakes greater than magnitude 2. Data source: Northern California Earthquake Data
Center; data contributed by Northern California Seismic Network, U.S. Geological
Survey, Menlo Park, and Berkeley Seismological Laboratory, University ofCalifornia,
Berkeley. Locations of the 1969 magnitude 5.6 and 5.7 Santa Rosa events, which are
highlighted, are from table 2 of Wong and Bott (1995). (b) Cross section shows seis-
micity within 5 km on either side of AA⬘.
stream-channel and debris-ﬂow deposits interbedded with
clayey silt overbank and sheet-ﬂow sediments (Figs. 4 and
5). The channel deposits are typically sandy gravel in a
clayey silt matrix; the overbank and overland-ﬂow deposits
are massive or weakly bedded and have scattered clasts of
gravel and very coarse-grained sand that increase in size and
abundance close to the pressure ridge. In general, the ﬁne-
grained component of unit 4 is very dark gray to grayish
brown, although the lowest intervals of clayey deposits ex-
posed in trench 3 have a stronger chroma. Southwest of the
pressure ridge, some of the ﬁne-grained deposits contain
enough organic matter to have a granular soil structure.
Northeast of the pressure ridge, a thin, white ash from an in-
place burn forms a distinctive horizon within the otherwise
dark, massive sediments. This ash marks the position of a
paleoﬂoodplain near the top of unit 4 (Fig. 4). A gravelly
matrix-supported channel deposit exposed in both walls of
trench 9 (unit 4Y) parallels the pressure ridge and has a
gradient to the northwest toward Rodgers Creek. A similar
lenticular deposit exposed in trench 3 at a greater depth is
interpreted to be the same channeled debris ﬂow (Figs. 4 and
5). The source of the deposit is a valley-side drainage that
feeds the alluvial slope southeast of the site (Fig. 3). The
unit 4Y deposit is cut partly into unit 9, suggesting that the
debris ﬂow was diverted along and scoured into the northeast
side of the pressure ridge. The upper contact of unit 4, which
is formed by ﬂoodplain deposits, is ﬂat-lying, in general, and
everywhere appears to be depositional in origin. Sharp in-
ﬂections in the contact that coincide with mapped faults are
clearly tectonic (Fig. 4a, meter 6; Fig. 4b, meter 6.3). Gentler
rises in dip that correspond to similar dips in underlying de-
posits can be attributed to tilting within the zone of defor-
mation (Fig. 4a, meter 5–7; Figs. 4c and 5a, meter 6–8). Along
the southwest side of the pressure ridge, ﬂoodplain deposits
blanket pre-existing topography (Fig. 4b, meter 11–12).
Overlying unit 4 are ﬂoodplain sediments (unit 2/3) that
are lighter in color than those of unit 4. Unit 2/3 consists of
a brown silt to clayey silt with a few lenses of ﬁne gravelly
silt and a few scattered cobbles. The unit is thickest on the
northeast, upstream side of the pressure ridge, where it in-
cludes an upper subunit with many strong brown mottles.
To the southwest of the ridge, the unit consists of a thin layer
of silt with modest components of clay and sand. These de-
posits extend up the ﬂanks of the pressure ridge at the po-
sition of trench 9 and top the ridge at trench 3, where the
ridge is lower and closer to Rodgers Creek (Figs. 3 and 4).
Fault Structure and Evidence of the Most
Zones of faults bound the slice of unit 9 beneath the
pressure ridge (Fig. 4). Slip along the northeast zone occurs
The Most Recent Large Earthquake on the Rodgers Creek Fault, San Francisco Bay Area 847
Figure 3. Three-dimensional perspective model of
the Triangle G Ranch site looking southeast along the
trend of the Rodgers Creek fault (RCF). The general
path of the unit 4Y debris ﬂow is inferred from trench
exposures and geomorphic setting. Path is dashed
where covered by ﬂoodplain deposits. The present-
day geometry of the 4Y deposit is used to constrain
slip along the northeast side of the pressure ridge in
the most recent event (see text for discussion).
Trenches 3 and 9, which are 19 and 16 m long, re-
spectively, are the principal exposures discussed in
this article. Trench 5b provides auxiliary information
on the geometry of unit 4Y.
on closely spaced reverse-oblique faults (zone 1) that dip
toward and appear to converge with generally higher-angle
faults on the opposite margin of the slice (zone 2).
Faulting within zone 1 extends to the top of unit 4. Some
fault strands that lie within unit 9 appear to terminate at the
base of channel deposits (Fig. 4a, meter 7); other strands cut
channel deposits (Fig. 4a and 4b, meter 6–7; Figs. 4c and
5a, meter 7.5–8.5). Fault traces clearly extend to the upper
contact of unit 4 at only two locations (Fig. 4b, meter 6.5;
Fig. 4c, meter 8–8.5). Identifying fault discontinuities in the
massive clayey silt overbank sediments is difﬁcult, however,
and faults that appear to terminate at the top of channel de-
posits probably continue upward for another ⬃10–20 cm to
the top of unit 4 (e.g., Figs. 4c and 5a, meter 7.5). Faults
inferred to extend to the top of unit 4 are dashed and queried
on the trench logs.
Deformation of the upper contact of unit 4 is interpreted
as having occurred at the ground surface and, thus, this con-
tact deﬁnes the event horizon for the MRE. This deformation
is expressed as small, ⬃15- to 20-cm-high northeast-facing
buried scarplets and broader warping associated with fault
zone 1 (Fig. 4a and 4b, meter 5–7; Figs. 4c and 5a, meter
6–8.5). Unit 2/3 onlaps the fault zone and shows no evidence
of deformation. Warping of the top of unit 4 corresponds to
similar warping in underlying channel deposits (Fig. 4a, me-
ter 5–7; Figs. 4c and 5a, meter 6–8). A buried scarplet in
trench 3 (Figs. 4c and 5a, meter 6.5) lies slightly east offault
zone 1 and expresses fold-related deformation. Note that an
apparent scarp on the southeast wall of trench 9 at meter 7
is an artifact of a boulder mapped at the ground surface and
does not represent faulting in the MRE.
Although the earthquake event horizon is unambiguous
within the zone of deformation, the position of the horizon
is less clear northeast of fault zone 1. The burn ash near the
top of unit 4 rises slightly as it approaches the buried scarplet
in trench 3 (Figs. 4c and 5a, meter 6.2). The ash thins and
dies out on the margin of the deformation zone, but the ho-
rizon appears to project onto the scarp. This relation suggests
the possibility that the ash lies on or slightly above the event
horizon and that the ash and the ⬃5 cm of overlying unit 4
ﬂoodplain deposits postdate faulting. This interpretation is
tentative both because of the discontinuous nature of the ash
near the scarplet and because the apparent inﬂection may be
one of many local irregularities in the horizon and unrelated
to deformation of the ground surface. The discontinuous na-
ture of the ash deposit near the scarplet is consistent with
evidence from trench 9 that the ash-forming burn was re-
stricted, at least locally, to the ﬂoodplain north of fault zone
1 (Fig. 4a,b). Note that if the ash had originally extended
across the scarplet at the position of trench 3 and was sub-
sequently eroded, one would expect to see reworked ash in
deposits above the in-place burn.
Fault zone 2 on the southwest side of the pressure ridge
varies from a narrow, well-deﬁned steeply northeast-dipping
zone in trench 3 to a distributed, less distinct set of south-
west-dipping strands in trench 9. Two subparallel faults ex-
posed in trench 9 bound a slice of unit 8 between units 9
and 4 (Fig. 4b, meter 12–13). Neither of these faults could
be traced upward through the upper part of unit 4, which
laps onto the pressure ridge. However, a fault that splays
into unit 8 displaces the base of this onlapping section (Fig.
4b, meter 12.0–12.2). Because the unstratiﬁed, ﬁne-grained
nature of these sediments makes detection of faulting difﬁ-
cult, each of these strands may in fact extend to the top of
unit 4. Fault zone 2 exposed in trench 3 juxtaposes units 8
and 9 and clearly offsets upper unit 4 channel deposits, but,
like elsewhere, was not recognized in the upper ⬃20 cm of
massive silty overbank deposits (Fig. 4c, meter 12.8).
Slip during the Most Recent Event
We estimate the amount of right-lateral slip on fault
zone 1 during the MRE using the geometry of the fault-
parallel debris-ﬂow unit 4Y (Fig. 3) and the overlying in-
terval of unit 4 ﬂoodplain deposits. The unit 4Y channel in
Figure 4. Logs of trench exposures showing evidence of the most recent surface-rupturing earthquake. Evidence includes faults
that extend to the upper contact of unit 4 and scarplets buried by unit 2/3. Trench-wall grids are registered across trench 9 (logs a and
b), but not between trench 9 and trench 3. Horizontal reference line at meter 2 of trench 9 highlights difference in vertical position of
channel deposit 4Y across the trench. The geometry of this channel constrains the amount of horizontal slip on fault zone 1 in the most
recent event (see Fig. 6 and discussion in text). See text for unit descriptions. Ages of charcoal samples are listed in Table 1. Pollen-
sampling proﬁle (on log c) was taken in a pit adjacent to trench 3 after the trench was closed.
The Most Recent Large Earthquake on the Rodgers Creek Fault, San Francisco Bay Area 849
Figure 5. Detail of trench exposure on northeast side of the pressure ridge (trench
3, southeast wall). (a) Trench log enlarged from Fig. 4c. (b) Photograph showing char-
acter of deposits. Black lines are unit contacts; white lines are faults.
cross section is tilted by amounts comparable with the top-
of-unit-4 ground surface, indicating that the channel records
only one event (Fig. 4a, meter 5–7; Figs. 4c and 5a, meter
6.5–8). We primarily use the walls of trench 9 for slip re-
construction because exposures of unit 4Y can be conﬁ-
dently correlated across a known distance. Note that,
whereas unit 4Y is present on the northeast side of the fault
zone on the southeast wall of trench 9 (Fig. 4a, meter 5–
6.5), it is absent ⬃1 m away on the opposite wall (Fig. 4b).
This suggests that the channel is highly sinuous here and
therefore crosses the fault before reaching the opposite
trench wall. The channel crosses back to the northeast side
of the fault between trenches 9 and 3 (Fig. 4c).
Figure 6 illustrates our approach to estimating right-
lateral slip. Essentially, we determine the original fault-
parallel gradient of the channel (Fig. 6b) and then backslip
850 S. Hecker, D. Pantosti, D. P. Schwartz, J. C Hamilton, L. M. Reidy, and T. J. Powers
Figure 6. Perspective view toward south of the
walls of trench 9 as logged on the northeast side of
the pressure ridge (simpliﬁed from Fig. 4a, b; north
direction, approximate), depicting a geometric and
trigonometric approach to restoring offset of unit 4Y
during the most recent event. For clarity, trenchwidth
is exaggerated (⬃1.5 ⳯) and postfaulting deposits
(unit 2/3) are omitted. (a) Present conﬁguration of
deposits. (b) Conﬁguration of deposits with vertical
component of deformation removed. Note that the top
of unit 4 on the southeast wall is restored to horizontal
primarily by removing the tilt in the underlying unit
4Y channel. Restoration returns channel to its original
gradient. (c) Inferred preslip position of channel ex-
posures. Lateral slip (x⬘) is restored by sliding the
northeast section of channel left-laterally until it
aligns with the gradient-projected position of the
channel on the other side of the fault (location marked
with a star). See text for deﬁnition of parameters and
discussion of channel-gradient and slip determina-
one side of the fault until the channel exposed in cross sec-
tion intersects the gradient-projected position of the channel
on the other side of the fault (Fig. 6c). This lateral translation
restores the channel to its prefaulting position.
Conceptually, the ﬁrst step is to remove the small ver-
tical (pressure-ridge-forming) component of slip represented
by deformation of the originally ﬂat-lying ﬂoodplain depos-
its that form the upper contact of unit 4 (Fig. 6a,b). The
vertical separation of channel 4Y across the fault zone is
partly the result of vertical slip and partly the result of jux-
taposition of deposits translated laterally along gradient. The
up-to-the-southwest sense of vertical separation that remains
after vertical slip is removed (Fig. 6b) is consistent with
right-lateral offset of a northwest-ﬂowing channel (Fig. 3).
The vertical component of deformation is greater on the
southeast wall (⬃35–40 cm) than on the northwest wall
(⬃20–25 cm) of trench 9. This along-strike difference in
uplift, which acts to exaggerate the gradient of the channel,
is consistent with the cumulative pattern of uplift represented
by the pressure ridge (Fig. 3).
The gradient (
) of channel 4Y is estimated trigono-
metrically using the distance between trench walls (x) and
the difference between walls in the thickness of unit 4 above
the base of the channel on the southwest side of the fault
zone. That is,
are the above-channel thicknesses of unit 4
on the downslope and upslope trench walls, respectively
(Fig. 6b). Our best estimate of
is 10⬚NW (using x⳱
1.05 m and y
⳱0.19 m). Incorporating uncertainty
in xof Ⳳ0.1 m and uncertainty in y
of Ⳳ0.08 m
that ranges from 6⬚to 16⬚NW.
The gradient along this 1-m reach of channel can be
compared with the gradient along the ⬃4-m reach between
the southeast walls of trench 9 and trench 3 (Fig. 3). The
base of the channel section exposed in trench 3 is buried
⬃0.62 m deeper below the top of unit 4 than the section on
the same (northeast) side of the fault in trench 9 (Fig. 4a,c).
These coordinates (x⳱⬃4 m and y
⬃9⬚, similar to the ⬃10⬚estimated between the walls
of trench 9.
The Most Recent Large Earthquake on the Rodgers Creek Fault, San Francisco Bay Area 851
Figure 7. Log of part of trench 5b showing an
exposure of channel unit 4Y that, as a supplement to
other exposures (Fig. 4), was used to estimate the gra-
dient of the channel to constrain the amount of slip
on fault zone 1 during the most recent event (see dis-
cussion in text).
In a trench ⬃4 m southeast of trench 9 (trench 5b;
Fig. 3), a section of the channel that is caught up in the fault
zone is not buried by ﬂoodplain deposits (Fig. 7). If the base
of the channel at this location, with respect to the pressure-
ridge side of the fault, is level with the top of unit 4, then
the gradient between here and trench 9 is ⬃5–6⬚NW. This
is similar to the ⬃6–7⬚W slope of the active, unburied al-
luvial surface southeast of the pressure ridge (Fig. 3). The
steeper (6–16⬚NW) slope of the buried channel between the
walls of trench 9 and trench 3 may reﬂect proximity to a
broad paleochannel of Rodgers Creek.
The amount of right-lateral slip (x⬘) indicated by ex-
posures of channel 4Y in trench 9 can be calculated from
the channel gradient (
) and the difference across the fault
in the thickness of unit 4 that overlies the base of the channel.
where and y
are the above-channel thicknesses of unit 4y⬘
on the downthrown and upthrown sides of the fault, respec-
tively (Fig. 6c). Substituting equation (1) into equation (2)
Our best estimate of x⬘is 2.2 m (using y⬘ⳮy⳱0.39 m
and the parameter values for a 10⬚gradient). Incorporating
the ranges of uncertainty in our measurements (which in-
cludes Ⳳ0.03 m for ) yields x⬘that ranges from 1.4y⬘ⳮy
to 3.4 m.
This estimate of 2.2 (Ⳮ1.2, ⳮ0.8) m of right-lateral
slip along the northeast side of the pressure ridge is a min-
imum for the site. An unknown amount of additional slip
occurred on fault zone 2 ﬂanking the southwest side of the
ridge. Total slip at the Triangle G Ranch site during the MRE
may be similar to, but is probably greater than, the 2 (Ⳮ0.3,
ⳮ0.2) m of surface offset measured 0.7 km to the south at
the Beebe Ranch paleoseismic site (Budding et al., 1991).
Timing of the Most Recent Event
Spanish mission records indicate that no large-magni-
tude earthquakes have occurred on the RCF since at least the
early nineteenth century (Budding et al., 1991). Initial pa-
leoseismic evidence from the Triangle G Ranch site, in the
form of radiocarbon-dated detrital charcoal collected from
unit 4, indicates that surface rupture occurred on the fault
sometime after A.D. 1438–1654 (Schwartz et al., 1992).
Herein, we present new radiocarbon results that signiﬁcantly
improve the maximum-age constraint and narrow the time
window for the MRE. In addition, we consider the robustness
of written records for constraining the minimum age of the
MRE and present data on the arrival of exotic pollen at the
site with respect to timing of the earthquake.
The Written Record. The recency of the MRE is con-
strained by the preinstrumental historical record. Franciscan
missions, located near the coast between San Diego and Son-
oma, kept detailed accounts of damaging earthquakes, until
about 1834 when the missions were secularized. Earth-
quakes that did not damage the missions were rarely reported
(Ellsworth, 1990). These written records, along with some
government correspondence, are the primary sources of in-
formation on earthquakes that occurred during the period
before the 1849 gold rush (Toppozada et al., 1981).
The oldest mission close enough to the RCF to be af-
fected by a major earthquake on the fault is Mission Dolores
in San Francisco, founded in 1776 (Fig. 1). The Presidio of
San Francisco was founded the same year, with the construc-
tion of a quadrangle and living quarters. The initial church
at Mission Dolores was built of wood; construction of the
adobe church that stands today was begun in 1782 and com-
pleted by 1791 (Engelhardt, 1924). This church was so well
built that it was not signiﬁcantly damaged by the 1906 earth-
quake on the San Andreas fault, although it was damaged
by an earthquake in 1838 generally assumed to have been
on that fault (Toppozada and Borchardt, 1998). Whether
Mission Dolores would have been damaged by a large earth-
quake ⬃50 km away on the RCF is unclear. The earliest
earthquakes reported at either the Mission or the Presidio is
a series of eighteen shocks over a period of a month in 1808
that damaged some adobe walls at the Presidio (Townley
and Allen, 1939; Toppozada et al., 1981). It is noteworthy
that the quality of construction at the Presidio during this
852 S. Hecker, D. Pantosti, D. P. Schwartz, J. C Hamilton, L. M. Reidy, and T. J. Powers
time was poor and buildings were in constant need of repair
(Langelier and Rosen, 1992). The 1808 earthquakes were
not reported as damaging at Mission Dolores (Toppozada
and Borchardt, 1998). Because the location of the 1808
earthquakes is unknown, Budding et al. (1991) allowed, but
did not conclude, that the source could have been the RCF,
thereby limiting the recency of the MRE to 1808. The ap-
parent absence of a dominant mainshock in the 1808 series
argues against the occurrence of a large, surface-rupturing
earthquake accompanied by an aftershock sequence at that
time. Toppozada and Borchardt (1998) similarly concluded
from the account of damage that the 1808 events did not
indicate a large earthquake, but instead may have been a
nearby moderate earthquake and its aftershocks. No other
damaging earthquakes are noted in the years before a mis-
sion was built in Sonoma, only 7 km east of the RCF,in
1824. The earliest earthquake mentioned in a chronicle of
the Sonoma Mission is a damaging event in 1868 (Smilie,
1975, in Budding et al., 1991), which was an estimated mo-
ment-magnitude (M)⬃6.8 earthquake ⬃70 km to the south-
east on the southern Hayward fault (Fig. 1) (Bakun, 1999).
An earthquake that occurred in 1898, the “Mare Island”
earthquake, caused substantial damage along the north mar-
gin of San Pablo Bay and to the northwest in communities
along the valleys of the Petaluma River and Sonoma Creek.
Bakun (1999) analyzed seismic intensity data and obtained
a moment magnitude of M⬃6.3 and a source location cen-
tered in northeast San Pablo Bay (Fig. 1). Uncertainty in the
location allows that the earthquake could have occurred on
any of several faults, including the southern end of the RCF,
the West Napa fault, the Concord-Green Valley fault, the
north Hayward fault, or faults beneath fold structures eastof
Mare Island (Bakun, 1999). Toppozada et al. (1992) inter-
preted the intensity data as indicating that the epicenter was
probably on the southern RFC (southeast of the Triangle G
Ranch trench site) and the magnitude was about 6.7. Sub-
sequently, Toppozada revised the magnitude downward to
6.4 (Toppozada et al., 2000; Toppozada and Branum, 2002).
Wesson et al. (2003) used an approach that evaluates the
association of speciﬁc earthquakes with speciﬁc faults to de-
termine that the northern Hayward fault is more likely to be
the source of the 1898 earthquake than the RCF. Parsons et
al. (2003) cited accounts of a small tsunami as evidence of
coseismic vertical displacement of the bayﬂoor and, using
historical hydrographic surveys to identify bayﬂoor subsi-
dence and a hydrodynamic model to constrain rupture pa-
rameters, concluded that the earthquake occurred on a nor-
mal fault within the step-over between the RCF and Hayward
fault. Regardless of its location, the 1898 event appears to
be too small to be the MRE recognized at the Triangle G
Ranch and Beebe Ranch trench sites. Empirical relations in-
dicate that M⬍6.5 earthquakes on strike-slip faults have
surface displacements that are generally less than a fewdeci-
meters (Wells and Coppersmith, 1994), considerably smaller
than the ⬃2-m displacements observed at both trench sites.
Toppozada et al. (1981) consider that by 1800, the rec-
ord of earthquakes of magnitude ⬃7 or greater is probably
complete within about 100 km of the coast between San
Diego and Sonoma counties. Almost certainly, the record of
large-magnitude earthquakes is complete for the RCF by
1824, when the mission at Sonoma was established. The
absence of damaging-earthquake accounts in documents
from Mission Dolores and the Presidio indicates that the
record for the RCF may be complete back to circa 1776.
Toppozada and Borchardt (1998) concluded that there is no
historical evidence for any major earthquakes in the San
Francisco Bay area before the 1838 earthquake on the San
Andreas fault, back to the founding of the Mission and the
Presidio in 1776.
The Pollen Record. The position of the MRE horizon rela-
tive to the earliest appearance of nonnative pollen in the
stratigraphic record at the Triangle G Ranch site provides an
additional constraint on the recency of the MRE. The arrival
of nonnative pollen in a region serves as a marker for the
beginning of the historical period and is playing an important
role in determining the timing of large prehistoric and early
historic earthquakes in California (e.g., Lienkaemper et al.,
2002; Stone et al., 2002). Erodium cicutarium, a weedy
Mediterranean annual, accompanied the ﬁrst Spanish settle-
ments. In San Francisco, the ﬁrst appearance of E. cicutar-
ium (in lake cores) is estimated to be 1800 (Ⳳ20 years)
(Reidy, 2001). In southern California, the plant’s arrival has
been precisely dated and determined to have occurred
slightly in advance of Spanish settlement (Mensing and
Byrne, 1998). E. cicutarium pollen identiﬁed in varved-
marine-sediment cores from the Santa Barbara Channel
dates from the period 1750–1765, which is at least two de-
cades before the founding of the Santa Barbara mission in
1786. Mensing and Byrne (1998) argue that E. cicutarium
dispersed northward through natural propagation from Baja
California, where it was probably introduced by livestock.
They estimate a migration rate of ⬃30 or 50 km/yr, depend-
ing on when and where the plant was ﬁrst established. If
migration continued northward at the same rate, E. cicutar-
ium would have reached our trench site on the RCF (⬃490
km of Santa Barbara) in ⬃1770–1775. However, the cooler,
moister climate and greater vegetation cover of northern
California may have slowed the advance of E. cicutarium.
In addition, San Francisco Bay may have been a barrier to
migration. Indeed, preliminary analysis of a core from Bol-
inas Lagoon, north of the San Francisco Peninsula, indicates
that E. cicutarium may have arrived in that area not long
before 1850 (Byrne, 2002; R. Byrne, personal comm., 2003).
We sampled for E. cicutarium pollen at the Triangle G
Ranch site in a pit adjacent to trench 3 along a vertical proﬁle
east of fault zone 1 where unit 4 is warped (Fig. 4c). We
sampled at 1-cm intervals through the uppermost part of unit
4 and most of unit 2/3 to document the ﬁrst appearance and
depth distribution of E. cicutarium pollen grains. Standard
pollen extraction and preparation procedures were followed
(Faegri and Iversen, 1975), and a minimum 800-mm
The Most Recent Large Earthquake on the Rodgers Creek Fault, San Francisco Bay Area 853
Figure 8. First arrival and abundance of E. cicu-
tarium pollen in the stratigraphic record relative to
the earthquake event horizon. Pollen samples were
taken at 1 to 5-cm intervals from the southeast wall
of a pit dug adjacent to trench 3 (see Fig. 4c for lo-
cation of samples). Slides made for pollen-grain
counts were carefully prepared to ensure consistency
in sample volumes. For each sample, a minimum area
of 800 mm
was scanned at 10⳯magniﬁcation.
of each slide was scanned at 10⳯magniﬁcation for the
large, easily identiﬁable E. cicutarium grains.
E. cicutarium pollen ﬁrst appears in deposits that lie
above the earthquake horizon at the base of unit 2/3 (Fig. 8).
The shape of the pollen-depth curve in Figure 8 is similar to
curves from lake and marsh pollen studies (e.g., Cole and
Liu, 1994; Reidy, 2001), suggesting that the depositional
record is fairly continuous. The ﬁrst appearance and gradual
increase in E. cicutarium pollen likely document the arrival
and expansion of E. cicutarium locally. The subsequent de-
cline and plateau in abundance likely reﬂect increasingcom-
petition from other nonnative weeds and grasses and then
quasistabilization of the population.
The ﬁrst appearance of E. cicutarium lies 1 cm above a
prominent decrease in organic-matter content that deﬁnes the
contact between unit 4 and unit 2/3. We infer that this cor-
respondence reﬂects disturbance in the local environment
related to the introduction of livestock. In a study of envi-
ronmental changes on Santa Cruz Island in southern Cali-
fornia, Brumbaugh (1980) found similar evidence of two
distinct geomorphic regimes: one represented by ﬁne-
grained alluvium containing charcoal fragments that reﬂects
a stable, vegetated landscape and the other represented by
coarser alluvium without charcoal that reﬂects increasedhill-
slope erosion due to livestock grazing. An increase in quartz
sand grains in the upper half of unit 2/3 (above a depth of
22 cm in our proﬁle) may reﬂect increased hillslope erosion
at our site.
Grazing near the trench site probably began about 1830,
when the ﬁrst land grants were established in the region (G.
Burch, California State Parks, personal comm., 1998). Thus,
the evidence provided by E. cicutarium pollen reinforces our
conclusion that the earthquake did not occur after the nearby
Mission Sonoma was founded in 1824. If the arrival of E.
cicutarium actually predates local grazing (our inferences
concerning geomorphic changes notwithstanding), and E. ci-
cutarium dispersed through natural propagation from south-
ern California, the timing constraint may be similar to that
provided by the founding of Mission Dolores and the Pre-
sidio in 1776.
Radiocarbon Constraints and Chronological Modeling
Dating of detrital charcoal from the upper ⬃0.5 m of
unit 4 (Table 1; Figs. 4 and 9) provides us with the oppor-
tunity to place well-constrained limits on the maximum age
of the MRE. Charcoal samples were collected from overbank
and channel deposits on both sides of the pressure ridge; 20
samples were submitted for radiocarbon analysis. Most of
the samples analyzed were near faults and were clearly
within the upper part of the faulted section. Four of the sam-
ples are in situ charcoal recovered from the burn-ash hori-
zon. All the samples collected appeared to be part of the
original deposit and not introduced by postdepositional pro-
cesses. We took care in the ﬁeld to avoid sampling charcoal
that may have been part of a root, was located in any feature
resembling a burrow, or showed signs of having been bio-
turbated into place.
Although the radiocarbon ages of the samples analyzed
range up to ⬃2000
C years B.P., the ten youngest dates
cluster between 110 and 230 (Ⳳ40)
C years B.P. (Fig. 9).
We regard this group of young dates, which span most of
the sampled interval, as a reliable constraint on the age of
the stratigraphy and the older dates as the ages of wood from
long-lived trees or charcoal that was reworked and is sig-
niﬁcantly older than the surrounding deposit.
One surprising result is that the burn-ash horizon, which
lies just a few centimeters below the top of unit 4, yielded
some of the oldest dates (Fig. 9). Because charcoal from an
in situ burn has not been subject to reworking and redepos-
854 S. Hecker, D. Pantosti, D. P. Schwartz, J. C Hamilton, L. M. Reidy, and T. J. Powers
Radiocarbon and Calendar Dates of Charcoal Samples from Upper Part of Unit 4
vert; horiz (m)
Depth Below Top
of Unit 4 (m)
Radiocarbon Age, yr B.P.
(95.4% Probability Level)
3S-3 CAMS-51766 1.84; 12.78 0.36 110 ⴣ40 1670 AD–1780 AD (33.1%) Detrital
1800 AD–1960 AD (62.3%)
3S-54 CAMS-51769 2.13; 6.74 0.16 120 ⴣ40 1670 AD–1780 AD (36.2%) Detrital
1800 AD–1960 AD (59.2%)
5bN-3 CAMS-58487 2.19; 2.93 0.06 140 ⴣ50 1660 AD–1960 AD (95.4%) in situ
3N-46 CAMS-51772 2.11; 7.07 0.18 170 ⴣ40 1650 AD–1890 AD (79.2%) Detrital
1910 AD–1960 AD (16.2%)
3N-47 CAMS-51773 2.18; 7.96 0.28 170 ⴣ40 1650 AD–1890 AD (79.2%) Detrital
1910 AD–1960 AD (16.2%)
3S-57 CAMS-51770 1.88; 7.60 0.5 180 ⴣ40 1640 AD–1710 AD (20.1%) Detrital
1720 AD–1890 AD (59.5%)
1910 AD–1960 AD (15.9%)
3N-48 CAMS-51774 2.29; 7.21 0.04 190 ⴣ50 1640 AD–1890 AD (80.7%) Detrital
1910 AD–1960 AD (14.7%)
9N-4 CAMS-51763 1.66; 12.87 0.12 200 ⴣ40 1640 AD–1700 AD (25.3%) Detrital
1720 AD–1820 AD (50.8%)
1830 AD–1880 AD (4.9%)
1910 AD–1960 AD (14.4%)
3S-4 CAMS-51767 1.84; 12.54 0.38 230 ⴣ40 1520 AD–1580 AD (7.3%) Detrital
1620 AD–1690 AD (38.4%)
1720 AD–1820 AD (40.3%)
1920 AD–1950 AD (9.4%)
9N-35 CAMS-51765 1.85; 4.41 0.24 230 ⴣ50 1510 AD–1600 AD (13.2%) Detrital
1610 AD–1700 AD (32.5%)
1720 AD–1820 AD (36.7%)
1830 AD–1880 AD (3.2%)
1910 AD–1960 AD (9.8%)
3N-43 TO-2927 1.54; 12.95 0.62 340 Ⳳ60 1440 AD–1660 AD (95.4%) Detrital
9N-26 CAMS-50077 2.02; 2.25 0.04 350 Ⳳ40 1450 AD–1640 AD (95.4%) in situ
3S-20 CAMS-51768 1.86; 7.35 0.52 380 Ⳳ40 1440 AD–1530 AD (55.1%) Detrital
1540 AD–1640 AD (40.3%)
3N-61 CAMS-51776 1.52; 13.30 0.66 400 Ⳳ40 1430 AD–1530 AD (67.3%) Detrital
1550 AD–1640 AD (28.1%)
3N-10 CAMS-51771 1.60; 12.74 0.56 440 Ⳳ50 1400 AD–1530 AD (81.0%) Detrital
1560 AD–1630 AD (14.4%)
9S-34 CAMS-51762 2.01; 4.44 0.07 480 Ⳳ50 1320 AD–1360 AD (6.2%) in situ
1380 AD–1520 AD (87.4%)
1590 AD–1620 AD (1.8%)
3S-18 TO-2767 2.04; 6.95 0.28 640 Ⳳ60 1270 AD–1420 AD (95.4%) Detrital
9N-25 CAMS-50576 1.98; 2.11 0.1 710 Ⳳ90 1150 AD–1430 AD (95.4%) in situ
3N-59 CAMS-51775 1.70; 13.48 0.46 970 Ⳳ40 990 AD–1170 AD (95.4%) Detrital
3S-16 TO-2766 2.06; 6.90 0.26 1960 Ⳳ70 160 BC–130 BC (1.4%) Detrital
120 BC–230 AD (94.0%)
*Includes trench number (north or south wall), and sample no. See Figure 4 for locations (except for sample 5bN-3, not shown).
Ages from the Center for Accelerator Mass Spectrometry (CAMS), Lawrence Livermore National Laboratory, are by AMS analysis; ages fromIsoTrace
Radiocarbon Laboratory at the University of Toronto (TO) are conventional.
Ages corrected for d
C (assumed value of ⳮ25, appropriate to type of material) (Stuiver and Pollach, 1977). Ages in bold form young cluster shown
in Figure 9.
Calibrated with OxCal Program, ver. 3.8 (Bronk Ramsey, 2002). Calibration curve from Stuiver et al. (1998).
ition, we expected samples from the horizon to be among
the youngest and closest in age to the surrounding deposit.
One possibility is that much of the wood that produced the
charcoal was detrital from old trees, possibly redwoods, and
was considerably older than the ground surface on which it
was lying when it burned. Radiocarbon dating determines
the age of the wood itself, not when the wood burned and
became part of the deposit. The general antiquity and range
of ages obtained from the young burn illustrates how sig-
niﬁcant this distinction can be. One of the burn-horizon sam-
ples (5bN-3) has a young radiocarbon age (140 Ⳳ50 years
B.P.) similar to the youngest detrital charcoal ages. This sug-
gests that some of the wood that burned to produce the ash
was not much older than the underlying ground surface.
The ages of the youngest charcoal samples, when con-
verted from radiocarbon to calendar ages, have distributions
The Most Recent Large Earthquake on the Rodgers Creek Fault, San Francisco Bay Area 855
Figure 9. Depth versus age distribution of charcoal samples from unit 4. Bars rep-
resent one standard deviation conﬁdence in the analytical ages. Arrows adjust depths
of samples taken from an inset channel-ﬁll sequence (southwest side of pressure ridge,
trench 3) to better reﬂect the true stratigraphic position of the host deposits. Dashed
line demarcates group of young dates (to the left of line) used to constrain timing of
MRE (see text for discussion). Note break in timescale between 1000 and 2000 years.
that are broad and multimodal (Fig. 10, distribution out-
lines). Because the distributions overlap, incorporating
stratigraphic and other chronological ordering constraints
can reduce the likely range of dates (Biasi and Weldon,
1994). The MRE can be treated as an undated event in the
sequence and thus the age of the earthquake can be more
precisely and probabilistically determined. We accomplish
this using the radiocarbon-calibration program OxCal,
which uses Bayesian statistics to analyze and incorporate
stratigraphic information (Bronk Ramsey, 1995, 2001, 2002).
This approach has been applied in developing detailedpaleo-
earthquake chronologies elsewhere along the San Andreas
fault system (e.g., Fumal et al., 2002; Lienkaemper et al.,
2002; Lindvall et al., 2002).
At the Triangle G Ranch site, uncertainties in determin-
ing relative stratigraphic position between trench walls,
across faults, and across the pressure ridge preclude placing
the entire set of samples in stratigraphic order. Even a se-
quence of only a few dates, however, tightens the constraint
on the age of the earthquake. Historical limits on the recency
of the earthquake can be included to further narrow and more
accurately deﬁne a probability distribution for the date of the
The stratigraphic sequence of charcoal samples is inter-
preted by using cross-cutting relations between channel and
overbank deposits, evidence of layering within the overbank
deposits, and, where layering is not apparent, depth below
the top of unit 4. Individual overbank deposits are probably
thin, such that charcoal samples vertically separated bymore
than a few centimeters are likely to be from different-age
deposits. The ⬃5-cm thickness of unit 4 sediments above
the paleo ﬂoodplain demarcated by the burn ash is consistent
with this concept.
We identify a stratigraphically ordered sequence of
three samples collected from the northeast side of the pres-
sure ridge on the southeast wall of trench 3 (samples 3S-20,
54, and 57) and the uppermost sample collected on the op-
posite wall (sample 3N-48; Figs. 4c and 5a). The latter sam-
ple is from 4 cm below the top of unit 4 and is the highest
in the faulted section. Two of the samples (3S-57 and 3S-
20) were located only 2 cm apart, but separated by a change
in deposit color and grain size that likely represents a layer
boundary. The lowest sample (3S-20) is not from the group
of young dates, but is included to determine whether older
dates may have an effect on the model.
The results of the chronological model for this sequence
of samples (Fig. 10a) indicate that the MRE likely occurred
no earlier than A.D. 1715 (at the 95-percentile level). The
model uses A.D. 1776, the year that Mission Dolores and the
Presidio were founded in San Francisco, as the beginning of
the historical earthquake record. This single-year constraint
produces an age distribution for the event that is highly
asymmetric, with the probability weighted toward the his-
torical period (such that the age range at 68 percentile begins
at 1750). If a younger date is substituted for the historical
cutoff, the age distribution broadens and shifts to younger
856 S. Hecker, D. Pantosti, D. P. Schwartz, J. C Hamilton, L. M. Reidy, and T. J. Powers
Figure 10. Results of chronological modeling of
constraints on the timing of the MRE (using the pro-
gram OxCal, v. 3.8, Bronk Ramsey, 2002). Open
curves show probability distributions of calibrated
radiocarbon ages prior to running the model. Shaded
portions of curves are the model distributions that re-
ﬂect limits imposed by other age constraints, includ-
ing the start of the historical period. Solid black
curves are model distributions for the date of the MRE
(lower limiting age at 95% is shown). We made al-
ternative runs to address the uncertainty in construct-
ing a chronological model from our knowledge of the
stratigraphy: (a) Model that we prefer, because it in-
corporates the largest number of radiocarbon-dated
samples that can be conﬁdently placed in stratigraphic
order and uses the likely historical constraint of 1776.
Note that the ordering constraint has no effect on the
model distribution of the oldest sample. (b) Model
with the same suite of ordered samples, but with an
historical constraint of 1824. (c) Model with all the
young charcoal samples (see Fig. 9 and Table 1) con-
sidered to be a coherent group older than the earth-
quake and whose relative-age sequence is unknown.
For completeness, we include sample 5bN-3 from the
burn ash, whose stratigraphic position relative to the
event horizon is uncertain (see text discussion). Omit-
ting this date from the group has no effect on the
model. (d) Model with only the youngest radiocarbon
date used to constrain the timing of the earthquake.
dates. For the certain constraint of 1824 (when the nearby
Sonoma Mission was founded), 95% of most likely results
are no older than A.D. 1740 (Fig. 10b).
The accuracy of the model depends on whether the
stratigraphic sequence of the charcoal samples reﬂects their
chronological order. The prior calibrated-age distributions
for samples in the young group (Fig. 10c, distribution out-
lines) are so similar that they mask any differences in actual
ages. Because of this and because detrital charcoal can be
older than the host deposit, individual samples in the model
may in fact be out of chronological order. However, if all
the young samples in the section lacked stratigraphic order,
we would not expect their ages to form such a cluster. We
infer that the concentration of radiocarbon ages within a
⬃100-year period (Fig. 9) is evidence of an age progression
that is simply unresolvable at the scale of radiocarbondating.
To test the sensitivity of the model to the number and
choice of input dates, we used alternative suites of samples
that appear to be in stratigraphic order (samples 3N-48, 9N-
4, 3S-3, and 3S-4, altogether and with 3N-48 or 9N-4 omit-
ted) and obtained minimum-limiting ages for the MRE rang-
ing from A.D. 1710 to A.D. 1736 (at the 95- percentile level).
The modeling results indicate that if as few as three of the
young samples from unit 4 are in chronologically correct
order, we can conclude that the earthquake likely occurred
after the beginning of the eighteenth century. We have great-
est conﬁdence in the stratigraphic order of samples shown
in Fig. 10a and thus prefer an event-age distribution that
begins at A.D. 1715 (at the 95-percentile level).
The Most Recent Large Earthquake on the Rodgers Creek Fault, San Francisco Bay Area 857
To avoid inferring the relative ages of samples, we can
model the group of young dates as a “phase” within an over-
all sequence (Bronk Ramsey, 1995, 2002). This structure
produces a likely earliest date for the earthquake of A.D.
1690 (Fig. 10c). A model can also be run that includes the
age of only a single charcoal sample. Selecting the sample
with the youngest radiocarbon age (110 Ⳳ40
C years B.P.,
sample 3S-3) provides the best constraint. This simple model
yields a likely earliest date for the earthquake of A.D. 1695
(Fig. 10d), similar to the group result.
In summary, we conclude that the MRE on the RCF oc-
curred no earlier than A.D. 1690 and probably no earlier than
A.D. 1715. The earthquake may have occurred shortlybefore
the beginning of the historical period, inasmuch as the age
distribution for the event is skewed toward young dates. For
the preferred model, the range for 68% of most likely results
begins at A.D. 1750. Also, if the group of young radiocarbon
ages is more chronologically ordered than represented byour
models, then the corresponding event age would verge even
more closely on the beginning of the historical period.
The timing of paleoearthquakes is fundamental to un-
derstanding the space-time behavior of faults and provides
critical input for seismic hazard estimates. We have de-
scribed ﬁeld relations, radiocarbon dating, pollen evidence,
and the record of historical seismicity that allow us to con-
strain the age of the most recent large surface rupture on the
RCF to after A.D. 1715 and before 1776, with some conﬁ-
dence, and to after A.D. 1690 and before 1824, with cer-
tainty. The age of this event can be compared with that of
recent earthquakes on the Hayward fault to better understand
possible fault-rupture patterns. In addition, the elapsed time
since the MRE can be compared with the average recurrence
interval for the RCF to gauge the imminence of a future
Rupture Scenarios and Earthquake Size
The timing of the MRE on the RCF has been used to-
gether with similar information for the Hayward fault to de-
velop segmentation models and potential rupture scenarios
for the Hayward–Rodgers Creek fault system (Working
Group on California Earthquake Probabilities, 1999, 2003;
hereafter referred to as WG02). Segmentation is an impor-
tant concept for characterizing the spatial distribution of
earthquake sources and the magnitudes of expected events.
WG02 divides the Hayward–Rodgers Creek into three dis-
tinct fault segments: the Rodgers Creek, north Hayward, and
south Hayward segments. The northern extent of rupture in
the 1868 earthquake (Fig. 1) deﬁnes the boundary between
the south Hayward and north Hayward segments. The 6-km
releasing step-over from the north Hayward to the RCF be-
neath San Pablo Bay forms a likely structural and geometric
barrier to rupture and thereby deﬁnes another segment
boundary. However, the timing of prehistoric earthquakes
suggests that these segments may not always rupture inde-
The MRE on the north Hayward fault, documented at
the Mira Vista site in El Cerrito (Fig. 1), occurred after A.D.
1640 (Hayward Fault Paleoearthquake Group, 1999). A
study of historical documents by Toppozada and Borchardt
(1998) indicates that an earthquake in 1836, commonly be-
lieved to be on the north Hayward fault, actually occurred
south of the San Francisco Bay area and that the most recent
large earthquake on the north Hayward fault occurred before
the beginning of the written record in 1776.
The south Hayward fault ruptured in 1868 and in late
prehistoric time as well. In a study at Tyson’s Lagoon in
Fremont (Fig. 1), Lienkaemper et al. (2002) used chrono-
logical modeling to estimate a maximum age for the pen-
ultimate earthquake of A.D. 1650. Like the MRE on the RCF
at the Triangle G Ranch site, the penultimate event on the
south Hayward fault at Tyson’s Lagoon lies just below the
appearance in the stratigraphic record of the nonnative pol-
len, E. cicutarium. In addition, Lienkaemper and Williams
(1999) postulate that the penultimate earthquake recorded at
the Montclair site near the north end of the 1868 rupture in
north Oakland (Fig. 1) corresponds to the 1640–1776 event
documented farther north at the Mira Vista site.
The general correspondence in timing of surface rupture
on the RCF and the north and south Hayward faults (after
A.D. 1640 and before circa 1776, Fig. 11a) suggests that
segments of the fault system may have failed together in a
single event or in separate events that were closely timed.
WG02 considered this information to deﬁne four possible
rupture scenarios (Fig. 11b). In developing models of fault-
rupture behavior, WG02 regarded the San Pablo Bay step-
over as a strong segmentation point and assigned a low
weight (⬍⬃10%) to simultaneous rupture of the RCF and
the north (or entire) Hayward fault. Evidence that the 1995
M6.9 Kobe, Japan, earthquake nucleated in a 4-km-wide
extensional step and ruptured bilaterally (e.g., Wald, 1996)
supports these low-probability scenarios and suggests the
possibility that San Pablo Bay may be a nucleation point for
bilateral rupture on the RCF and Hayward fault (WG02).
Where rupture terminates to the north on the RCF is
uncertain. Budding, et al. (1991) and Wong and Bott (1995)
identiﬁed an area of complex faulting near Santa Rosa where
the fault takes a ⬃1-km right step as a possible barrier to
rupture. Two damaging earthquakes (M
5.6 and 5.7) oc-
curred on 1 October 1969 in the step-over area (Wong and
Bott, 1995) (Fig. 2) and a M4.3 event occurred there on 25
May 2003 (Northern California Earthquake Data Center). To
the north of Santa Rosa, the RCF is known also as the
Healdsburg fault. WG02 considers the north end of potential
rupture on the RCF to be the north end of geomorphologi-
cally recognized Holocene faulting at Windsor Creek (Bry-
ant, 1982; Hart, 1992) (Fig. 1). Exploratory trenches at two
sites on fault strands north of Windsor Creek in Healdsburg
(the Panorama Reservoir and Foss Creek Detention Basin
858 S. Hecker, D. Pantosti, D. P. Schwartz, J. C Hamilton, L. M. Reidy, and T. J. Powers
Figure 11. Earthquake ages and rupture scenarios for the Hayward–Rodgers Creek
fault system. (a) Shaded rectangles represent the 95-percentile range and darker ellipses
represent the 68-percentile range for preferred ages of the most recent event on the
Rodgers Creek and north Hayward faults and the penultimate event on the south Hay-
ward fault. Note that we use 1776 as a nominal cutoff for the age of paleoearthquakes
along the fault zone, although Lienkaemper et al. (2002) incorporate the uncertainty in
this historical constraint to obtain a modeled age for the event on the south Hayward
fault that extends to 1790 (at 95-percentile conﬁdence). The bar indicates the 1868
earthquake on the south Hayward fault. Not shown is the prepenultimate event on the
south Hayward fault, which has a modeled age of A.D. 1530–1740. (b) Rupture sce-
narios proposed by W02. Small solid rectangles represent segment boundaries, or rup-
ture end points, between the Rodgers Creek (RC), north Hayward (NH), and south
Hayward (SH) faults (modiﬁed from Working Group on California Earthquake Prob-
abilities, 2003, ﬁgure 3.2). The correspondence in earthquake ages shown in (a) sup-
ports the possibility of simultaneous rupture of the RC and NH (scenario 2), the NH
and SH (scenario 3), or the RC, NH, and SH (scenario 4).
sites; Fig. 1) exposed evidence suggestive of Holocene dis-
placement (Harlan Tait Associates, 1996; S. Hecker, ﬁeld
observations, 1996; Kleinfelder, Inc., 1996; Hart, 1998).
Perhaps faulting north of Windsor Creek occurs less fre-
quently or as discontinuous, distributed slip as part of the
right step-over to the Maacama fault.
The size of an earthquake on the RCF depends on
whether the Hayward fault also ruptures. WG02 calculated
earthquake magnitudes for alternative rupture scenarios us-
ing a weighted combination of several area-magnitude re-
lations wherein fault-plane areas are reduced to account for
the effect of creep (an issue for the Hayward fault). Given
rupture that is restricted to the RCF (with the north end at
Windsor Creek and the south end in the center of San Pablo
Bay), WG02 calculations provide a magnitude of M6.8–
7.1. If rupture occurs on the RCF and the north Hayward
fault, the expected magnitude is M6.9–7.3. In the event that
rupture involves the south Hayward fault as well, the esti-
mate increases to M7.1–7.4.
We can calculate Mdirectly from seismic moment (M
M⳱(2/3 log M
)ⳮ10.7 (Hanks and Kanamori, 1979),
⳱luA, and where lis the shear modulus of
elasticity (taken to be 3 ⳯10
) and uis the av-
erage displacement over the ruptured fault plane with area
A. We do this for the case where earthquake rupture is re-
stricted to the RCF by assuming that the 2 (Ⳮ0.3, ⳮ0.2) m
of surface slip observed at Beebe Ranch approximates the
average slip on the fault in a typical event. Using the length
(53–73 km) and width (10–14 km) adopted by WG02 for
the RCF, we calculate M7.0–7.2.
Earthquake Recurrence and Elapsed Time
Timing constraints discussed in this article deﬁne the
elapsed time since the MRE on the RCF. Given a date for the
MRE of between A.D. 1715 and 1776, the elapsed time (as
of 2005) is between 229 and 290 years. Considering the most
conservative age interval for the MRE (no earlier than A.D.
1690 and no later than 1824), the elapsed time is certainly
no longer than 315 years and no shorter than 181 years. The
elapsed time, together with the average recurrence interval,
is critical for calculating conditional probabilities of the next
large earthquake on the fault.
Recurrence intervals estimated for the RCF encompass
the elapsed time since the MRE. Schwartz et al. (1992) de-
rived an average recurrence interval of 131 to 370 years
(with a best estimate of 230 years) using a slip-rate estimate
of 6.4–10.4 mm/yr for the past 750 years and event-slip in-
formation from the Beebe Ranch site. The earthquake model
of WG02 computes a recurrence interval for the RCF of 136
to 345 years (with a mean of 205 years), similar to the value
determined from geologic data. Because the elapsed time on
the RCF appears to have reached, or to be approaching, the
average repeat time for large earthquakes on the fault, a large
event may be likely in the near future. WG02 estimated the
mean probability of a Mⱖ6.7 event on the RCF in the next
The Most Recent Large Earthquake on the Rodgers Creek Fault, San Francisco Bay Area 859
30 years at 17%, greater than that on any other fault segment
in the San Francisco Bay region.
The timing of the most recent surface-rupturing earth-
quake on the south-central RCF is well determined from
chronological modeling using radiocarbon and historical
constraints to be no earlier than A.D. 1690 and probably no
earlier than A.D. 1715. Overlap in radiocarbon dates, com-
bined with information on stratigraphic order, of abundant
charcoal from the top of the faulted section at the Triangle
G Ranch site allows good resolution of the maximum earth-
quake age. The presence of nonnative pollen associated with
livestock grazing above the faulted section indicates a pre-
settlement date (before circa 1830) for the most recent rup-
ture. The lack of evidence in the written record of a large
earthquake on the RCF implies that the surface-rupturing
event occurred before 1824, when a mission was built near
the fault, and likely before 1776, when the written record
began regionally in San Francisco. The timing of rupture on
the RCF is similar to prehistoric rupture on the north and
south segments of the Hayward fault (A.D. 1640–1776). This
permits the possibility that an earthquake ruptured both the
RCF and the Hayward fault across the San Pablo Bay step-
over. The elapsed time since the MRE on the RCF is probably
between 229 and 290 years and certainly between 181 and
315 years, which is comparable with a calculated average
recurrence interval for the fault of 131 to 370 years (best
estimate, 230 years). Slip in the MRE is estimated to be at
least 2.2 (Ⳮ1.2, ⳮ0.8) m at the Triangle G Ranch site, simi-
lar to or greater than the 2 (Ⳮ0.3, ⳮ0.2) m of total slip
measured at the nearby Beebe Ranch site.
We thank the Triangle G Ranch for graciously granting us permission
over the years to trench on their property. We thank Bill Bryant, Tom
Fumal, Jim Lienkaemper, and Pat Williams for their careful reviews and
suggestions that improved the manuscript. We also thank Bob Simpson for
so agreeably providing illustrations of the region’s seismicity and Pauline
Curiel for formatting the references and Table 1. We wish to acknowledge
Koji Okumura and Francesca Cinti, who contributed their insights as part
of the ﬁrst ﬁeld crew at the Triangle G Ranch site in the early 1990s.
Bakun, W. H. (1999). Seismic activity of the San Francisco Bay region,
Bull. Seism. Soc. Am. 89, 764–784.
Biasi, G. P., and R. J. Weldon, II (1994). Quantitative reﬁnement of cali-
C distributions, Quat. Res. 41, 1–18.
Bolt, B. A., and R. D. Miller (1975). Catalog of Earthquakes in Northern
California and Adjoining Areas: 1 January 1910–31 December 1972,
Seismographic Stations, University of California, Berkeley, 567 pp.
Bronk Ramsey, C. (1995). Radiocarbon calibration and analysis of stratig-
raphy: the OxCal Program, Radiocarbon 37, 425–430.
Bronk Ramsey, C. (2001). Development of the radiocarbon calibration pro-
gram OxCal, Radiocarbon 43, 355–363.
Bronk Ramsey, C. (2002). OxCal Program Ver. 3.8, Radiocarbon Accel-
erator Unit, University of Oxford, Oxford, United Kingdom,
www.rlaha.ox.ac.uk/oxcal/oxcal.htm (last accessed September 2003).
Brown, R. D., Jr. (1970). Faults that are historically active or that show
evidence of geologically young surface displacement, San Francisco
Bay region—a progress report, October 1970, U.S. Geol. Surv. Open-
File Map (Basic Data Contribution 7), map scale 1:250,000.
Brumbaugh, R. W. (1980). Recent geomorphic and vegetal dynamics on
Santa Cruz Island, California, in The California Islands: Proceedings
of a Multidisciplinary Symposium, D. M. Power (Editor), Santa Bar-
bara Museum of Natural History, Santa Barbara, California, 139–158.
Bryant, W. A. (1982). Chianti, Healdsburg, Alexander, Maacama and re-
lated faults, Sonoma County, Calif. Div. Mines Geol. Fault Evaluation
Rept. 135, 18 pp., map scale 1:24,000.
Budding, K. E., D. P. Schwartz, and D. H. Oppenheimer (1991). Slip rate,
earthquake recurrence, and seismogenic potential of the Rodgers
Creek fault zone, northern California: Initial results, Geophys. Res.
Lett. 18, 447–450.
Byrne, R. (2002). Comments on the Bolinas Lagoon Restoration Project
Draft Environmental Impact Statement (DEIS), Environmental Impact
Report (EIR), Draft Feasibility Report (DFR), and their Technical
Appendices, University of California, Berkeley, 27 pp.
Cole, K. L., and G. Liu (1994). Holocene paleoecology of an estuary on
Santa Rosa Island, California, Quat. Res. 41, 326–335.
Ellsworth, W. L. (1990). Earthquake history, 1769–1989, The San Andreas
Fault System, California, R. E. Wallace (Editor), U.S. Geol. Surv.
Profess. Pap. 1515, 153–188.
Engelhardt, Z. (1924). The Missions and Missionaries of California: San
Francisco or Mission Dolores, Franciscan Herald Press, Chicago, Il-
linois, 432 pp.
Faegri, K., and J. D. Iverson (1975). Textbook of Pollen Analysis, Third
Ed., Hafner Press, New York, 295 pp.
Fumal, T. E., M. J. Rymer, and G. G. Seitz (2002). Timing of large earth-
quakes since A.D. 800 on the Mission Creek strand of the San Andreas
fault zone at Thousand Palms Oasis, near Palm Springs, California,
Bull. Seism. Soc. Am. 92, 2841–2860.
Hanks, T. C., and H. Kanamori (1979). A moment-magnitude scale, J.
Geophys. Res. 84, 2348–2350.
Harlan Tait Associates (1996). Fault and geotechnical investigation,
planned Panorama Resevoir, Healdsburg, California, report prepared
for the City of Healdsburg, Project No. 697.431, 22 pp.
Hart, E. W. (1982). Rodgers Creek fault, Sonoma County, Calif. Div.Mines
Geol. Fault Evaluation Rept. 141, 20 pp., map scale 1:24,000.
Hart, E. W. (1992). Recently active traces of the Rodgers Creek fault,
Sonoma County, California, Calif. Div. Mines Geol. Open-File Rept.
92-07, 14 pp., map scale 1:24,000.
Hart, E. W. (Compiler) (1998). Fault number 31, Healdsburg fault, in Qua-
ternary fault and Fold Database of the United States, Ver. 1.0, U.S.
Geol. Surv. Open-File Rept. 03-417, http://qfaults.cr.usgs.gov.
Hayward Fault Paleoearthquake Group (1999). Timing of paleoearthquakes
on the northern Hayward fault—preliminary evidence in El Cerrito,
California, U.S. Geol. Surv. Open-File Rept. 99-318,33pp.
Herd, D. G. (1978). Map of Quaternary faulting along the southern Rodgers
Creek fault zone, Glen Ellen, Petaluma River, and Sears Point 7.5-
minute quadrangles, California, U.S. Geol. Surv. Open-File Rept. 78-
306, map scale 1:24,000.
Huffman, M. E. (1971). Geology for planning in the Sonoma Mountain and
Mark West-Riebli Road areas, Sonoma County, California, Calif. Div.
Mines Geol. Open-File Rept. 72-25, map scale 1:24,000.
Jennings, C. W. (1994). Fault activity map of California and adjacent areas,
Calif. Div. Mines Geol., Geol. Data Map No. 6, map scale 1:750,000.
Kleinfelder, Inc. (1996). Geologic investigation and fault hazard evaluation
report, Foss Creek Detention Basin, Healdsburg, California, report
prepared for the City of Healdsburg, Project No. 41-3100-01, 20 pp.
Langelier, J. P., and D. B. Rosen (1992). Historic Resource Study, El Pre-
sidio de San Francisco: A History under Spain and Mexico, 1776–
1846, Presidio of San Francisco, Golden Gate National Recreation
860 S. Hecker, D. Pantosti, D. P. Schwartz, J. C Hamilton, L. M. Reidy, and T. J. Powers
Area, California, U.S. Department of the Interior, National Park Ser-
vice, Denver Service Center, 181 pp.
Lienkaemper, J. J., T. E. Dawson, S. F. Personius, G. G. Seitz, L. M.Reidy,
and D. P. Schwartz (2002). A record of large earthquakes on the
southern Hayward fault for the past 500 years, Bull. Seism. Soc. Am.
Lienkaemper, J. J., and P. L. Williams (1999). Evidence for surface rupture
in 1868 on the Hayward fault in north Oakland and major rupturing
in prehistoric earthquakes, J. Geophys. Res. Lett. 26, 1949–1952.
Lindvall, S. C., T. K. Rockwell, T. E. Dawson, J. G. Helms, and K. W.
Bowman (2002). Evidence for two surface ruptures in the past 500
years on the San Andreas fault at Frazier Mountain, California, Bull.
Seism. Soc. Am. 92, 2689–2703.
Mensing, S., and R. Byrne (1998). Pre-mission invasion of Erodium cicu-
tarium in California, J. Biogeogr. 25, 757–762.
Parsons, T., R. Sliter, E. L. Geist, R. C. Jachens, B. E. Jaffe, A. Foxgrover,
P. E. Hart, and J. McCarthy (2003). Structure and mechanics of the
Hayward–Rodgers Creek fault step-over, San Francisco Bay, Cali-
fornia, Bull. Seism. Soc. Am. 93, 2187–2200.
Reidy, L. M. (2001). Evidence of Environmental change over the last 2000
years at Mountain Lake, in the northern San Francisco Peninsula,
California, M.A. Thesis, University of California, Berkeley, 108 pp.
Schwartz, D. P., D. Pantosti, S. Hecker, K. Okumura, K. E. Budding, and
T. Powers (1992). Late Holocene behavior and seismogenic potential
of the Rodgers Creek fault zone, Sonoma County, California, in Proc.
of the Second Conference on Earthquake Hazards in the Eastern San
Francisco Bay Area, G. Borchardt, S. E. Hirschfeld, J. J. Lienkaem-
per, P. McClellan, and I. G. Wong (Editors), Calif. Div. Mines Geol.
Spec. Publ. 113, 393–398.
Smilie, R. S. (1975). The Sonoma Mission—San Francisco Solano de Son-
oma, Valley Publishers, Fresno, California, 149 pp.
Stone, E. M., L. B. Grant, and J. R. Arrowsmith (2002). Recent rupture
history of the San Andreas fault southeast of Cholame in the northern
Carrizo Plain, California, Bull. Seism. Soc. Am. 92, 983–997.
Stuiver, M., and H. A. Pollach (1977). Discussion: Reporting of
Radiocarbon 19, 355–363.
Stuiver, M., P. J. Reimer, and T. F. Braziunas (1998). High-precision ra-
diocarbon age calibration for terrestrial and marine samples, Radio-
carbon 40, 1127–1151.
Toppozada, T. R., J. H. Bennett, C. L. Hallstrom, and L. G. Youngs (1992).
1898 “Mare Island” earthquake at the southern end of the Rodgers
Creek fault, in Proc. of the Second Conference on Earthquake Haz-
ards in the Eastern San Francisco Bay Area, G. Borchardt, S. E.
Hirschfeld, J. J. Lienkaemper, P. McClellan, and I. G. Wong (Editors),
Calif. Div. Mines Geol. Spec. Publ. 113, 385–392.
Toppozada, T. R., and G. Borchardt (1998). Re-evaluation of the 1836
“Hayward fault” and the 1838 San Andreas fault earthquakes, Bull.
Seism. Soc. Am. 88, 140–159.
Toppozada, T. R., and D. Branum (2002). California earthquakes of M ⱖ
5.5: their history and the areas damaged, in International Handbook
of Earthquake and Engineering Seismology, Part A, W. H. Lee, H.
Kanamori, and P. Jennings (Editors), Academic Press, New York,
Toppozada, T. R., D. Branum, M. Petersen, C. Hallstrom, C. Cramer, and
M. Reichle (2000). Epicenters of and areas damaged by M ⬎5 Cali-
fornia earthquakes, 1800–1999, Calif. Div. Mines Geol. Map Sheet
49, scale 1:1,500,000.
Toppozada, T. R., C. R. Real, and D. L. Parke (1981). Preparation of iso-
seismal maps and summaries of reported effects for pre-1900 Cali-
fornia earthquakes, Calif. Div. Mines Geol. Open-File Rept. 81-11
SAC, 182 pp.
Townley, S. D., and M. W. Allen (1939). Descriptive catalog of earthquakes
of the Paciﬁc coast of the United States 1769 to 1928, Bull. Seism.
Soc. Am. 29, 1–297.
Wald, D. J. (1996). Slip history of the 1995 Kobe, Japan, earthquake de-
termined from strong motion, teleseismic, and geodetic data, J. Phys.
Earth 44, 489–503.
Weaver, C. E. (1949). Geology of the Coast Ranges immediately north of
the San Francisco Bay region, Geol. Soc. Am. Memoir, Vol. 35,
242 pp., map scale 1:62,500.
Wells, D. L., and K. J. Coppersmith (1994). New empirical relationships
among magnitude, rupture length, rupture width, rupture area, and
surface displacement, Bull. Seism. Soc. Am. 84, 974–1002.
Wesson, R. L., W. H. Bakun, and D. M. Perkins (2003). Association of
earthquakes and faults in the San Francisco Bay area using Bayesian
inference, Bull. Seism. Soc. Am. 93, 1306–1332.
Willis, B., and H. D. Wood (1922). Fault map of the state of California,
Seismological Society of America, scale 1:506,880.
Wong, I. G. (1991). Contemporary seismicity, active faulting, and seismic
hazards of the Coast Ranges between San Francisco Bay and Healds-
burg, California, J. Geophys. Res. 96, 19,891–19,904.
Wong, I. G., and J. D. Bott (1995). A new look back at the 1969 Santa
Rosa, California, earthquakes, Bull. Seism. Soc. Am. 85, 334–341.
Working Group on California Earthquake Probabilities (1990). Probabili-
ties of large earthquakes in the San Francisco Bay region, California,
U.S. Geol. Surv. Circular 1053,51pp.
Working Group on California Earthquake Probabilities (1999). Earthquake
probabilities in the San Francisco Bay Region: 2000–2030, U.S. Geol.
Surv. Open-File Rept. 99-517,55pp.
Working Group on California Earthquake Probabilities (2003). Earthquake
probabilities in the San Francisco Bay Region: 2002–2031, U.S. Geol.
Surv. Open-File Rept. 03-214.
U.S. Geological Survey
345 Middleﬁeld Rd., M.S. 977
Menlo Park, California 94025
(S.H., D.P.S., J.C.H., T.J.P.)
Istituto Nazionale di Geoﬁsica e Vulcanologia
Sezione Sismologia e Tettonoﬁsica
Via di Vigna Murata 605
I-00143 Roma, Italy
Department of Geography
507 McCone Hall, #4740
University of California
Berkeley, California 94720
Manuscript received 9 July 2004.