Download San Andreas offset markers.pmd

Using seismically-observed basement
structure as an offset marker
Cumulative offset of the San Andreas fault in central California:
A seismic approach
Justin Revenaugh*
Colin Reasoner
Institute of Tectonics, Earth Sciences, University of California, Santa Cruz, California 95064
ABSTRACT
Scattered-wave imaging of upper crustal heterogeneity along nearly 500 km of the San Andreas fault in central California is used to estimate cumulative offset of basement rocks in the
fault zone. Optimal cross-fault realignment of scattering patterns is achieved through removal
of ~315 km of right-lateral slip. This value agrees with most previous estimates of early Miocene
displacement, placing the initiation of movement on the San Andreas no earlier than ca.
23.1 Ma. Scattering along the fault correlates with segment boundaries established on the basis
of historic and paleo seismicity, corroborating evidence from southern California that the upper
crustal structures responsible for scattering are important in seismogenesis.
INTRODUCTION
Beginning with its recognition as a major fault
in the aftermath of the great earthquake of 1906,
study of offset along the San Andreas fault has mirrored the development of modern tectonic theory peppered with innovation, and major revisions of thought and controversy, often of a
persistent nature. Is displacement strike slip or dip
slip? Is it measured in kilometres, tens of kilometres or hundreds of kilometres? When did motion
begin? Has it been steady through the fault s lifetime? How has the fault s geometry changed? That
these questions have endured is a testament to the
difficulty of measuring fault offset. Marker formations hundreds of kilometres apart must be identified, mapped, and correlated; piercing points must
be extrapolated from irregular contacts and uncertain contours; and, ages must be determined.
In central California the San Andreas fault is
well exposed, linear at the scale of Figure 1A
and compared to southern California part of
a simple right-lateral strain system. Displacement
since the early Miocene is determined to be
300—320 km, on the basis of cross-fault correlation of the Pinnacles and Neenach volcanic formations (Matthews, 1976), fan deposits of the
Temblor Formation and the Vaqueros Sandstone
of the Santa Cruz Mountains (Graham et al.,
1989), and paleoshorelines recorded in the northernmost Gabilan Ranges and San Emigdio
Mountains at Pleito Hills (Huffman et al., 1973).
an estimate of San Andreas basement offset in
central California is still needed to delimit initiation of movement on the San Andreas (Stanley,
1987; Graham et al., 1989), constrain slip on the
San Gregorio—Hosgri and Rinconada-Reliz fault
systems (e.g., Graham and Dickinson, 1978), and
to provide an important boundary condition on
total slip in southern California (e.g., Powell,
1993; Matti and Morton, 1993).
METHOD
Revenaugh (1995a; 1995b) and Revenaugh and
Mendoza (1996) document a migration algorithm
capable of mapping crustal scattering variability
using teleseismic earthquakes recorded by a regional seismic network. Unlike tomography that
maps velocity perturbations, the method maps
variations in a non-dimensional indicator of scattering intensity. Specifically it estimates the local
significance of scattering, or scattering potential.
Over the scale length of the migration operator
(~60 km), the estimator is uniform, such that
closely spaced scatterers have potentials proportional to their relative scattering strengths. Scattering is the product of elastic heterogeneity, in particular, abrupt variation in shear and compressional
wave velocity and density. Metre-scale and larger
cracks and faults appear to play a prominent role in
crustal scattering (Revenaugh, 1995b; Aki, 1995),
but other structures, such as intrusive contacts and
tight folds, are important also.
the scattered-wave image is a vertical average of
scattering intensity within the upper ~15 km of
crust, the offsets are substantially for basement
rocks and should be cumulative.
Figure 1B displays teleseismic P to S scattering potential in the San Andreas fault zone of
central California. The image was derived from
analysis of 8295 seismograms of 215 earthquakes recorded by the Northern California and
Southern California Seismic Networks between
1980 and 1994. Along the San Andreas, station
coverage is good, but in places the coverage approximates a linear array resulting in some offfault circular blurring of the scattering image.
Nonetheless, typical resolution is sufficient to
distinctly image scatterer volumes separated by
as little as 10 km. We find little evidence of the
San Andreas fault in scattering, much as observed for the San Jacinto fault and the San Gorgonio Pass stretch of the San Andreas fault system in southern California (Revenaugh, 1995b).
The central creeping zone and adjacent Parkfield
segments are associated with slightly elevated
mean scattering levels, but are not otherwise apparent. In interpreting this, it must be remembered that the relation of scattering potential to
absolute scatterer strength is local. High potential
scatterers are locally strong, but in regions of low
overall scattering, a high potential scatterer need
not be strong in a global sense. A consequence of
this behavior is that any long expanse of fault associated with high scattering strength will be
marked by scattering potential highs at the ends
rather than a continuous high. Thus if scattering
is strong throughout the creeping zone it is unlikely that we would image it as an extended zone
of high scattering potential. What we image are
the shorter wavelength (≤30 km) variations.
Profiles of scattering potential along the southwest and northeast sides of the San Andreas fault
were obtained by averaging scattering potential
within boxes measuring 15 km in the fault-normal
Using seismically-observed basement
structure as an offset marker
Method: observe scattering potential profile along both sides of
SAF, then autocorrelate profiles (that is, line them up optimally).
•
Uses teleseismic P to S wave scattering.
•
Scattering potential—high if there are many variations of
seismic velocity in a small area.
•
NOT tomography. This is more like migration, for those of
you who know/care.
Figure 1. A: Physiographic feature map of central California. BP = Big Pine fault; BT = Butano
fault; CV = Calaveras fault; GL = Garlock fault;
HW = Hayward fault; OR = Ortigalita fault; PC =
Pilarcitos fault; PL = Pleito fault; RR = Rinconanda-Reliz fault system; SCO = South
Cuyama-Ozena fault system; SG/H = San Gregorio–Hosgri fault system; SJ = San Juan
fault; WW = White Wolf fault; ZVG = ZayanteVergeles fault. Triangles are stations of the
Southern and Northern California Seismic
Networks. B: Depth slice at 10 km of the P to S
(teleseismic P scattered into upgoing S) scattered-wave image obtained from ~8300 recordings of 215 teleseisms representing geographic variation of scattering potential in
upper crust. Scattering potential near unity
(light yellow) implies locally strong scattering
within the 0.02° by 0.02° cell, whereas zero
(light blue) implies very little or no reradiated
energy. White denotes regions of poor azimuthal station coverage or fewer than 200
contributing seismograms. Dashed green line
follows center of simplified fault zone used to
compute scattering profiles; bars alternate
colors at 50 km intervals. Black dots are epicenters of ML ≥ 2 seismicity since 1969.
at the southeastern and northwestern ends of the
mapped fault zone and linearly interpolating offset between them. We tested right-lateral offsets
between 0 and 400 km at 2 km intervals, limiting
the increase (or decrease) of offset along the fault
to 200 km to avoid excessive profile stretch. The
preferred model, i.e., the model yielding the greatest correlation coefficient (r = 0.65), has a mean
offset of 315 km; displacement decreases from
320 km at the southernmost extent of the study
area to 311 at the northern limit of overlap (Figure 2A). Before bounding the uncertainty of this
estimate, we first document its significance since
many offset models were tried and the possibility
of spurious correlation is real.
Although removal of a running mean largely
eliminated autocorrelation peaks at large lag distances, the profiles remain autocorrelated at short
lags, such that there are fewer degrees of freedom
than scattering potential pairs in the correlation.
To estimate the former, we divided the length of
offset profile overlap by 6 km the variogram
range (e.g., Journel and Huijbregts, 1978). By
this method, the preferred offset model has 28¡ of
freedom and a correlation significance in excess
of 99.99% (<1 in 10 000 chance of occurrence in
uncorrelated data). By comparison, the secondhighest correlation peak (r = 0.45; mean offset of
279 km) attains 98.8% significance while correlation at zero offset (r = 0.26) falls just shy of
98% (Fig. 2B). Although both are highly significant, they are more than two orders of magnitude more likely to occur by chance than the peak
correlation and are entirely in keeping with the
number of independent offset models tested. We
conclude that if the peak correlation is chance, it
is a rare chance.
124
GEOLOGY, February 1997
Using seismically-observed basement
structure as an offset marker
Figure 2. A: Comparison of
southwest and northeast faultblock scattering potential profiles following removal of
~315 km of right-lateral offset,
corresponding to peak correlation model. Profiles are
obtained by averaging scattering potential in 15-km-wide by
2-km-long boxes on either side
of fault. Offset increases towards the southeast (dotted
line; right side axis), varying by
~10 km over a distance of
170 km. Along-fault distance is
appropriate for northeastern
fault block (southwestern
block is shifted). B: Correlation
significance for all offset models tested. Models within 2σ of
peak significance are shaded
at one-half standard-deviation
intervals. Offset is made to
vary linearly along fault zone
by interpolating between values specified at the southeast and northwest ends of sampling. Contour lines indicate mean offset (km) within overlapping region. They curve because overlap is
shorter than sampled fault length. Peak correlation is obtained for 294 and 320 km offset at northwest and southeast ends, respectively (black dot; model in A). Offset differentials >200 km require
unrealistic extension or contraction of fault blocks and are not shown.
Results:
•
Possible mean offset of 306 km to 319 km
•
The correlation has less than 1/10,000 chance
of occuring in random data.
Offset Cretaceous mafic rocks
Gabbroic rocks at Eagle Rest Peak, Gold Hill, and Logan
are similar.
124
o
o
120
o
42
K/Ar dates
(state-of-the-art methods for 1970)
Cape
Mendocino
Logan
Gold Hill
Eagle Rest Peak
156± 8, 172± ? Ma
144± 7 Ma
134± 4, 165± 4, 207± 10 Ma
Offsets
LOGAN
Point Reyes
38
o
Logan-Gold Hill
Logan-ERP
n
Sa
s
ea
dr
An
ul
Fa
GOLD HILL
EAGLE REST PEAK
t
a
G
ault
kF
c
rlo
34
Los Angeles
N
100 km
o
160 km
320 km
Provenance of Cretaceous mafic clasts
Conglomerate and sandstone at Gualala appears to be composed
of clasts of Logan/Gold Hill/Eagle Rest Peak rock.
Implication: 560 km of offset
124
o
o
120
o
42
Cape
Mendocino
GUALALA
LOGAN
Point Reyes
38
o
n
Sa
s
ea
dr
An
t
ul
Fa
GOLD HILL
EAGLE REST PEAK
a
G
ault
kF
c
rlo
34
Los Angeles
N
100 km
o
Pinnacles-Neenach
•
The most famous, most convincing offset marker.
•
Early Miocene (23.5 Ma) volcanic sequence.
•
Ten rock types found in the sequence.
124
o
o
120
o
42
Cape
Mendocino
Field, petrographic, and geochemical
characteristics are identical at the
two sites.
Point Reyes
38
o
PINNACLES
n
Sa
s
ea
dr
An
o
ul
Fa
116
t
a
G
lt
Fau
ck
o
rl
SAN JOAQUIN BASIN
NEENACH
Los Angeles
N
100 km
34
o
Pinnacles-Neenach
Cross section
sandstones,
shales, and
interbedded
basalts
andesite,
rhyolite,
dacite
Rock type
La Honda basin: Santa Cruz Mountains
San Joaquin basin: Temblor Range
Neenach: Los Angeles Co. near Gorman
Pinnacles: Pinnacles National Monument,
Highway 146, San Benito Co.
General location of outcrops
quartz-rich gabbro,
anorthosite
23.5 Ma
Age
Gualala conglomerate (located between Fort Ross
and Point Area) contains clasts that may be derived
from the Eagle Rest Peak - Gold Hill - Logan
Gabbro; if this is true, it records ~560 km of
displacement.
~150 Ma
~320 km
~210-280 km
~315-320 km
~315 km
Displacement
Ross, 1970
Ross et al., 1973
Crowell, 1962
Haxel and Dillon, 1978
Jacobsen et al., 2002
Jacobsen and Sorensen, 1986
Graham et al., 1989
Critelli and Nilsen, 2000
Matthews, 1976
Selected references
The Pelona - Orocopia Schist is correlated to the
Rand Schist and Catalina Schist and other rocks
located in California and Arizona. A total of 15
schist/graywacke/gabbro bodies believed to be
derived from the same source are displaced by faults
in California.
Paleocene or late
Cretaceous metamorphism;
uplifted and exhumed by
23-22 Ma
Oligocene - Miocene
Extra goodies
Logan: Coast Ranges, near Vergeles fault
Gold Hill: Coast Ranges near the town of
Cholane
Eagle Rest Peak: San Emigdio Mountains
Butano - Point of Rocks Sandstone is a similar
unit (the same?) as the displaced sub-marine
fans in the San Joaquin and La Honda Basins
and records ~ 300 km of displacement.
Eagle Rest Peak - Gold
Hill - Logan Gabbro
Pelona Schist: San Gabriel Mountains
Pelona - Orocopia Schist pelitic schist,
metagraywacke and
gneiss at greenschist Orocopia Schist: Mecca Hills (state
to amphibolite facies highway 195), Gavilan Hills
San Joaquin - La Honda
basin sub-marine fans
Pinnacles - Neenach
volcanics
Name
Selected geologic markers of offset on the San Andreas fault