of 26
Bulletin
of
the
Seismological
Society
of
America,
Vol.
72,
No.
6,
pp.
2037-2062,
December
1982
THE
1971
SAN
FERNANDO
EARTHQUAKE:
A
DOUBLE
EVENT?
BY
THOMAS
H.
HEATON
ABSTRACT
Evidence
is
presented
which
suggests
that
the
1971
San
Fernando
earthquake
may
have
been
a
double
event
that
occurred
on
two
separate,
subparallel
thrust
faults.
It
is
postulated
that
the
initial
event
took
place
at
depth
on
the
Sierra
Madre
fault
zone
which
runs
along
the
base
of
the
San
Gabriel
Mountains.
Rupture
is
postulated
to
have
occurred
from
a
depth
of
about
15
km
to
a
depth
of
about
3
km.
A
second
event
is
thought
to
have
initiated
about
4
sec
later
on
another
steeply
dipping
thrust
fault
which
is
located
about
4
km
south
of
the
Sierra
Madre
fault
zone.
The
surface
trace
of
this
fault
coincides
with
the
San
Fernando
fault
zone
which
was
the
principal
fault
associated
with
surface
rupture.
It
is
postulated
that
rupture
propagated
from
a
depth
of
8
km
to
the
free
surface.
The
moments
of
the
first
and
second
events
are
approximately
0.7
x
1026
dyne-cm
and
1.0
x
1026
dyne-cm,
respectively.
This
model
is
found
to
explain
the
combined
data
sets
of
strong
ground
motions,
teleseismic
P
and
S
waveforms,
and
static
offsets
better
than
previous
models,
which
consist
of
either
a
single
fault
plane
or
a
plane
having
a
dip
angle
which
shallows
with
decreasing
depth.
Nevertheless,
many
features
of
the
observed
motions
remain
unexplained,
and
considerable
uncertainty
still
exists
regarding
the
faulting
history
of
the
San
Fernando
earthquake.
INTRODUCTION
The
purpose
of
this
study
is
to
search
for
models
of
the
1971
San
Fernando
earthquake
which
are
compatible
with
both
local
and
teleseismic
observations.
This
may
seem
at
first
glance
to
be
an
unexciting
problem.
After
all,
in
the
I0
yr
since
the
occurrence
of
this
earthquake,
there
have
been
dozens
of
studies
of
the
local
and
teleseismic
records
of
the
San
Fernando
earthquake.
However,
none
of
these
studies
has
addressed
one
major
problem:
no
single
model
proposed
to
date
adequately
explains
both
local
and
teleseismic
observations
of
this
earthquake.
Furthermore,
it
is
crucial
to
our
understanding
of
seismic
sources
that
such
models
be
found.
Unraveling
the
story
contained
within
these
records
is
our
only
hope
for
discovering
the
time
history
of
the
faulting
which
was
responsible
for
this
earthquake.
Since
the
San
Feruando
earthquake
has
one
of
the
most
complete
data
sets
to
date,
failure
to
obtain
a
consistent
model
from
different
types
of
data
seriously
undermines
our
ability
to
interpret
earthquakes
with
less
complete
data
sets.
I
briefly
review
some
significant
findings
of
several
previous
studies
which
provide
motivation
for
this
work.
From
studies
of
the
surface
rupture
(Bonilla
et
al.,
1971;
Kamb
et
al.,
1971),
focal
mechanism
(Whitcomb,
1971),
and
hypocentral
location
(Allen
et
al.,
1973),
it
has
long
been
recognized
that
a
constantly
dipping
fault
plane
which
is
compatible
with
P-wave
first-motion
observations
would
not
intersect
both
the
hypocenter
and
the
surface
rupture
(Allen
et
al.,
1973).
An
obvious
and
simple
solution
to
this
problem
is
to
assume
that
the
fault
changes
dip
angle
somewhere
between
the
hypocenter
and
the
surface
rupture.
Many
model
studies
have
been
based
upon
the
premise
that
rupture
proceeded
unilaterally
up-dip
on
a
fault
plane
which
decreases
dip
as
the
depth
decreases.
This
assumption
has
been
so
often
repeated
that
it
has
become
somewhat
dogmatic.
Unfortunately,
there
is
no
direct
evidence
that
the
San
Fernando
earthquake
occurred
as
a
single
rupture
along
a
:2037
2038
THOMAS
H.
HEATON
variably
dipping
fault
plane.
I
later
show
that
an
alternative
model
of
fault
geometry
is
one
which
assumes
that
the
earthquake
was
actually
two
events
on
two
subparallel
faults.
I
now
trace
the
history
of
events
which
led
to
this
alternate
hypothesis.
Langston
(1978)
proposed
a
finite-fault
model
of
the
San
Fernando
earthquake
which
explained
many
of
the
features
of
both
the
long-
and
short-period
teleseismic
body
waves
recorded
for
this
event.
The
model
is
relatively
simple
and
consists
of
a
uniform
rupture
which
propagates
up-dip
at
a
rupture
velocity
of
1.8
km/sec.
Langston
tried
several
models
which
differed
in
the
dip
angle
of
the
top
part
of
the
fault.
He
found
that
the
teleseismic
long
periods
fit
best
if
the
dip
of
the
upper
fault
is
29
°.
A
similar
conclusion
was
reached
by
Bache
and
Barker
(1978)
from
their
study
of
teleseismic
short-period
P
waves.
Studies
of
both
strong
ground
motions
(Boore
and
Zoback,
1974;
Bache
and
Barker,
1978;
Heaton
and
Helmberger,
1979)
and
static
vertical
offsets
(Alewine,
1974)
have
also
managed
to
explain
features
of
this
earthquake
using
a
variably
dipping
fault
assumption.
However,
none
of
these
studies
has
demonstrated
that
a
single
model
adequately
explains
both
local
and
teleseismic
data
simultaneously.
In
an
earlier
study
of
strong
ground
motion
(Heaton
and
Helmberger,
1979),
we
produced
synthetic
strong
ground
motions
for
a
model
which
was
similar
to
the
one
Langston
(1978)
used
to
explain
teleseismic
body
waves.
We
found
that
this
model
could
not
adequately
explain
the
observed
strong
ground
motions.
However,
by
changing
such
parameters
as
rupture
velocity
and
the
distribution
of
faulting,
we
found
a
new
model,
Norma
163,
which
was
much
more
compatible
with
observed
strong
ground
motions.
In
this
earlier
study,
we
did
not
investigate
the
teleseismic
body
waves
which
would
be
produced
from
our
local
observation-based
model,
Norma
163.
We
reasoned
that
since
Norma
163
was
not
greatly
different
from
Langston's
teleseismic
model,
we
expected
that
our
model
would
also
explain
the
teleseismic
records
fairly
well.
Unfortunately,
we
have
since
discovered
that
Norma
163
does
a
poor
job
of
explaining
teleseismic
long-period
body
waves.
Upon
further
investigation,
we
discovered
that
it
was
extremely
difficult
to
simultaneously
explain
both
the
teleseismic
body-wave
data
and
the
strong-motion
data
with
the
type
of
model
consisting
of
rupture
along
a
single
variably
dipping
fault
plane.
We
further
discovered
that
if
we
assume
that
the
earthquake
occurred
on
two
subparallel
steeply
dipping
faults,
then
it
is
possible
to
explain
more
adequately
the
combined
teleseismic
and
local
observations
of
the
San
Fernando
earthquake.
In
this
paper,
I
will
show
why
I
believe
that
a
model
with
a
variably
dipping
fault
plane
is
inconsistent
with
the
data.
I
will
then
present
synthetic
records
for
a
model
consisting
of
two
steeply
dipping
faults.
Although
I
believe
that
such
a
model
offers
an
attractive
interpretation
of
the
data,
it
will
be
clear
that
there
is
still
much
to
be
learned
about
this
earthquake.
THE
DATA
I
will
employ
three
different
data
sets
in
this
study:
long-period
teleseismic
body
waves,
strong
ground
motions,
and
static
vertical
ground
deformations.
Long-period
teleseismic
body
waveforms
are
copied
directly
from
Langston's
(1978)
study.
He
reviewed
WWSSN
and
Canadian
network
data
and,
using
the
criteria
of
good
signal-
to-noise
ratio
and
azimuthal
coverage,
he
selected
17
long-period
vertical
P
wave-
forms,
6
long-period
SH
waveforms,
and
8
long-period
radial
SV
waveforms.
A
full
description
of
this
data
set
is
given
in
Langston's
paper.
The
locations
of
most
of
the
accelerometers
which
recorded
the
San
Fernando
earthquake
are
shown
in
Figure
1.
From
this
large
set
of
locations,
I
have
chosen
THE
1971
SAN
FERNANDO
EARTHQUAKE:
A
DOUBLE
EVENT?
2039
five
stations
on
the
basis
of
distance
to
the
fault
and
azimuthal
coverage.
The
five
stations
we
have
chosen
are
Pacoima
Dam
(PAC),
Holiday
Inn
(HLI),
Jet
Propulsion
Laboratory
(JPL),
Lake
Hughes
Array
Station
4
(LKH),
and
Pear
Blossom
Pumping
Plant
(PRB).
These
stations
are
indicated
in
Figure
1
by
the
codes
C041,
C048,
Gll0,
J142,
and
F103,
respectively.
These
codes
refer
to
the
cataloging
system
used
eEO71
eF
104
~-"
'~'~A
,,fm~;~,
u
O(;ll0
ACCELEROGRAPH
SITES
MAJOR
FA
U'LT
S
EXPOSED
BASEMENT
ROCK
,,~~
....
IO
---
BASEMENT
ELEVATION
IN
THOUSANDS
OF
FEET
~,@
SEA
LEVEL
DATUM
I
~,;~o(~
o
Io
20
30
114
I
km
SANTA
SUSANNA
MTS
00XNARD
SAN
SANTA
MONICA
MTS
(])AREA
I
®AREA
2
(~)AREA
3
'~
E
0/5
C
054
D
059
E
083
F
089
R
249
D065
F
098
I
1:~4
P217
G
112
I
151
S
265
K
157
N
188
$266
K
159
J
148
R
255
PACIFIC
~
OCEAN
SANTA
C
:
SAN
GABRIEL
Io~
~
r
£fL~
~
eHI21
---'ALLEY
\
~'~
""
......
/
,
~)
O
ONTARIO/~
062,
F,O92-~-v-'-
#
i
V~
/
/
ANGELE~S~%
"'--
30
I---
',~-.x
""
"
"~"
~
~"
....
""
-2o
"
~As'NeH
"
W
J°0
~eF~7124
~
.
,
P220
N
195
FIG.
1.
The
site
locations
of
accelerometer
recordings
of
the
San
Fernando
earthquake
are
superim-
posed
on
the
gross
geological
and
structural
features
of
the
area.
The
encircled
cross
is
the
Allen
et
al.
(1973)
epicenter,
and
the
arrows
point
to
stations
which
are
studied
in
this
paper
(modified
from
Hanks,
1975).
in
the
series
of
strong-motion
data
reports
published
by
the
Earthquake
Engineering
Research
Laboratory
of
the
California
Institute
of
Technology.
It
is
from
these
reports
that
I
have
taken
observed
ground
motion.
Because
I
am
interested
in
explaining
the
overall
faulting
history,
I
have
chosen
to
model
the
longer
period
parts
of
the
strong
ground
motions.
Thus
for
most
stations,
I
have
chosen
to
model
the
ground
displacement
history.
Since
very
long-period
ground
motions
generally
2040
THOMAS
H.
HEATON
cannot
be
recovered
from
accelerograms,
the
strong
ground
motions
have
been
subjected
to
parabolic
baseline
correction
and
Ormsby
filtering.
Discussion
of
this
processing
can
be
found
in
Trifunac
(1971),
Trifunac
et
al.
(1973a,
b),
and
Hanks
(1975).
Stations
very
close
to
faulting,
such
as
Pacoima,
may
experience
significant
static
offset
during
the
course
of
strong
shaking.
Unfortunately,
parabolic
baseline
correc-
tions
not
only
remove
such
static
offset,
but
they
can
also
introduce
character
which
has
little
resemblance
to
actual
ground
motion.
For
this
reason,
I
will
also
model
observed
ground
velocity
at
Pacoima.
These
velocity
records
are
obtained
by
direct
integration
of
the
digitized
accelerograms
and
do
not
have
parabolic
baseline
corrections.
The
static
vertical
uplift
data
used
in
this
study
are
taken
directly
from
a
compilation
of
such
data
assembled
by
Alewine
(1974)."He
collected
over
100
ver-
tical
displacement
data
points
and
constructed
a
contour
map
which
summarizes
these
data.
In
this
study,
I
assume
that
the
vertical
deformation
is
adequately
represented
by
Alewine's
contour
map.
Unfortunately,
the
horizontal
deformation
data
seem
less
complete
and
reliable
(Alewine,
1974)
and
will
not
be
used
in
this
study.
The
model.
The
primary
objective
of
this
study
is
to
produce
synthetic
displace-
ments
for
both
local
and
teleseismic
observations
using
a
consistent
source
model
throughout
the
study.
My
model
consists
of
a
three-dimensional
finite
fault
located
within
a
half-space
(P-wave
velocity
--
6.2
km/sec,
S-wave
velocity
=
3.5
km/sec,
density
=
2.7
gm/cm3).
I
assume
that
a
circular
rupture
front
propagates
at
a
given
rupture
velocity
from
the
hypocenter.
I
also
assume
that
the
slip
angle
and
dislo-
cation
time
history
are
uniform
everywhere
on
a
fault
plane.
I
further
specify
the
absolute
size
of
dislocation
to
be
some
arbitrary
function
of
position
on
the
fault.
I
perform
a
summation
of
point
source
responses
to
compute
synthetic
ground
motions.
These
point
sources
are
located
on
a
gridwork
which
covers
the
area
of
the
fault.
In
the
case
of
the
strong-motion
modeling,
the
Cagniard-de
Hoop
technique,
together
with
a
linear
interpolation
scheme,
is
used
to
compute
responses
to
individual
point
sources.
Exact
Cagniard-de
Hoop
solutions
are
used
to
compute
responses
for
our
closest
observation
point,
Pacoima.
Fifth
order
asymptotic
ap-
proximations
of
modified
Bessel
functions
are
used
in
the
computations
of
responses
at
the
other
four
strong-motion
stations
which
are
investigated.
More
complete
descriptions
of
these
techniques
can
be
found
in
Heaton
and
Helmberger
(1979)
and
particularly
in
Heaton
(1978).
Since
1400
point
sources
(130
of
which
are
computed
directly
with
Cagniard-de
Hoop)
are
summed
to
construct
synthetics
for
each
station,
computation
of
synthetic
strong-motions
is
both
time-consuming
and
ex-
pensive.
The
computation
of
synthetic
teleseismic
body-wave
motions
for
complex
source
models
is
surprisingly
simple
and
inexpensive
when
compared
to
the
strong-motion
problem.
Once
again,
a
Green's
function
technique
(point
source
summation)
is
employed.
However,
in
this
case,
some
very
useful
approximations
can
be
made.
These
approximations
and
the
nature
of
the
resulting
solutions
are
discussed
in
the
Appendix.
Theoretical
static
offsets
for
an
arbitrary
three-dimensional
finite
fault
located
in
a
half-space
are
also
computed
by
using
a
Green's
function
integration
technique.
Once
again,
the
fault
is
subdivided
into
a
gridwork
(1-km
squares)
where
the
dislocation
is
homogeneous
within
each
grid,
but
the
dislocation
varies
from
one
grid
to
another.
The
analytic
expressions
of
Mansinha
and
Smylie
(1971)
are
used
THE
1971
SAN
FERNANDO
EARTHQUAKE:
A
DOUBLE
EVENT.
9
2041
to
calculate
the
surface
offset
due
to
each
fault
grid.
A
linear
summation
of
all
of
the
fault
grids
yields
the
final
solution.
Variable-dip
single-fault
model.
In
this
section,
I
will
describe
the
problems
I
encountered
when
trying
to
explain
both
local
and
teleseismic
data
sets
using
models
which
consist
of
a
single
fault
whose
dip
shallows
as
depth
decreases.
A
cross-
sectional
view
of
the
fault
geometry,
which
was
proposed
by
Langston
(1978)
from
teleseismic
studies
and
which
we
later
assumed
in
our
strong-motion
study
(Heaton
and
Helmberger,
1979),
is
shown
in
Figure
2.
Recall
that
Langston
found
that
this
type
of
model
could
explain
the
teleseismic
body
waves
using
the
simple
assumption
that
the
size
of
dislocation
does
not
vary
with
depth
along
the
fault.
He
also
demonstrated
that
a
rupture
velocity
of
1.8
km/sec
would
best
fit
teleseismic
records
if
such
a
fault
geometry
was
chosen.
Our
study
of
strong
ground
motions,
however,
indicated
that
this
simple
rupture
model
is
not
compatible
with
observed
ground
motions.
In
particular,
we
demonstrated
that
uniform
shallow
faulting
beneath
the
strong-motion
station,
PAC,
would
produce
large
vertical
displacements
at
PAC,
Son
Fernando
Fault
jf
f
Pacoima
N__4
~
Depth
(k
m)"~
4
~.~
km
8
,o
12
ypocenter
14-4
16
J
FIG.
2.
Cross-sectional
view
of
fault
geometry
which
consists
of
a
single
fault
whose
dip
shallows
as
depth
decreases.
The
geometry
was
proposed
by
Langston
(1978)
and
assumed
in
our
earlier
study
of
strong
ground
motions
(Heaton
and
Helmberger,
1979).
but
these
cannot
be
seen
in
the
data.
In
order
to
alleviate
this
problem,
we
proposed
a
model
in
which
most
of
the
dislocation
on
the
shallow-dipping
portion
of
the
fault
plane
is
concentrated
near
the
free
surface
and
to
the
south
of
PAC.
This
conclusion
was
in
agreement
with
Alewine's
(1974)
modeling
of
the
vertical
static
offsets
produced
by
the
San
Fernando
earthquake.
In
order
to
explain
the
timing
of
pulses
at
PAC,
we
also
proposed
that
the
rupture
velocity
was
2.8
km/sec
on
the
steeply
dipping
portion
of
the
fault
and
1.8
km/sec
on
the
shallow-dipping
portion.
This
is
a
significantly
higher
rupture
velocity
than
that
derived
from
the
teleseismic
records
by
Langston.
Although
at
the
time
of
our
earlier
study,
we
recognized
that
our
model,
Norma
163,
might
not
explain
the
teleseismic
records
as
well
as
Langston's
model,
we
felt
that
Norma
163
would
probably
do
an
acceptable
job,
since
the
fault
geometry
and
the
relative
moments
of
the
steep
and
shallow-dipping
parts
of
the
fault
were
roughly
the
same
as
those
used
by
Langston.
We
did
not
anticipate
just
how
poorly
Norma
163
would
fit
the
teleseismic
records.
Strong
ground
motions.
In
Figure
3,
I
show
the
model
Norma
163
along
with
a
2042
THOMAS
H.
HEATON
Ap~
-
Sur
Rup,u,c
I
P
A'
I
I
I
5
cm
o
I
I
I
1
I
I
I
I
I
I
I
t
I
[
I
Pacoimo
(PAC)
_
oo;o
I
I
i
I
q
'
L'ake'Hulghe~
#14
I
I
(LKH)
I
I
S(3M
~
N21°E
]
FS
Down
cm
o
f
r
I
I
I
I
P
6
12
sec
P8
24
I
I
I
I
J
I
I
0
6
J2
sec
F8
24-
'
'/~'
JPL'
'
~I
F
t
0
6
12
sec
18
24
FIG.
3.
Summary
of
the
model
Norma
163,
proposed
in
our
earlier
study
(Heaton
and
Helmberger,
1979).
A
contour
map
of
the
assumed
dislocation
(in
meters)
is
shown
in
the
upper
left.
The
details
of
the
rupture
process
are
given
in
Table
1.
Also
shown
are
comparisons
of
observed
and
synthetic
displacements
for
the
stations
PAC,
JPL,
and
LKH.
The
top
trace
is
the
synthetic
ground
motion
(SGM);
the
middle
trace
is
this
synthetic
ground
motion
after
baseline
correction
and
Ormsby
filtering
(FS),
and
the
bottom
trace
is
the
observed
displacement
which
has
also
been
filtered
and
baseline
corrected
(IA).
THE
1971
SAN
FERNANDO
EARTHQUAKE:
A
DOUBLE
EVENT?
2043
comparison
of
synthetic
and
observed
strong
ground
motions
at
the
stations
PAC,
JPL,
and
LKH.
The
contours
signify
lines
of
equal
dislocation
in
meters.
The
rupture
front
spreads
radially
from
the
hypocenter
at
a
velocity
of
2.8
km/sec
and
then
slows
to
1.8
km/sec
as
it
ruptures
the
shallow-dipping
part
of
the
fault.
The
time
derivative
of
the
time
history
of
slip
for
each
point
on
the
fault
is
an
isosceles
triangle
with
a
duration
of
0.8
sec.
A
full
description
of
the
parameters
assumed
for
Norma
163
is
given
in
Table
1.
Comparisons
of
synthetic
and
observed
strong
ground
motions
for
Norma
163
are
also
shown
in
Figure
3.
Two
synthetic
records
are
shown
for
each
component
of
motion.
The
top
trace
is
computed
ground
motion
and
the
middle
trace
is
this
synthetic
motion
with
a
baseline
correction
and
high-pass
Ormsby
filter
applied.
We
used
the
baseline
correction
described
by
Nigam
and
Jennings
(1968),
and
an
8-sec
Ormsby
filter.
This
filter
is
described
by
Hanks
(1975).
A
detailed
description
of
the
seismic
phases
which
comprise
these
synthetics
can
be
found
in
our
earlier
paper
(Heaton
and
Helmberger,
1979).
Although
it
is
clear
that
the
synthetics
do
not
match
all
of
the
arrivals
present
in
the
data,
it
is
also
clear
that
this
model
explains
many
of
the
major
features
of
these
TABLE
1
SOURCE
PARAMETERS
FOR
NORMA
163
Lower
Upper
Segment
Segment
Depth
of
hinge
(km)
5.0
Strike
-75
°
-75
°
Dip
53
°
29
°
Rake
76
°
90
°
Rupture
velocity
(km/sec)
2.8
1.8
Rise
time
(sec)
0.8
0.8
Moment
1026
dyne-cm)
0.8
0.6
Hypocentral
longitude
118.41°E
118.33°E
Hypocentral
latitude
34.44°N
34.42°N
Hypocentral
depth
(km)
13
13
records
much
better
than
our
starting
model
which
had
relatively
uniform
faulting
throughout.
Teleseismic
body
waves.
Comparisons
of
synthetic
and
observed
teleseismic
long-
period
vertical
P
waveforms
for
Norma
163
are
shown
in
Figure
4.
All
amplitudes
have
been
normalized
such
that
the
number
listed
beside
each
record
is
the
moment
(x
1026
dyne-cm)
which
produces
identical
peak
amplitudes
for
the
observed
and
synthetic
records.
A
lower
hemispherical
projection
of
the
locations
of
stations,
as
well
as
the
P-wave
nodal
planes
for
the
two
parts
of
the
hinged
fault,
are
shown.
The
most
obvious
feature
of
this
figure
is
the
fact
that
there
is
a
poor
match
between
the
synthetics
and
the
data.
Furthermore,
the
matches
attained
by
Langston
(1978)
are
far
superior.
The
major
deficiency
in
the
synthetics
is
the
lack
of
a
strong
secondary
peak
which
is
universally
present
in
the
data.
Further
analysis
of
the
synthetics
produced
by
Langston's
model
reveals
that
this
second
peak
is
produced
by
the
shallow-dipping
portion
of
the
fault.
It
can
also
be
shown
that
most
of
the
energy
in
this
second
arrival
comes
from
the
part
of
the
fault
which
lies
between
depths
of
5
and
3
km.
For
a
shallow
dip-slip
fault,
there
is
a
strong
destructive
interference
between
the
direct
P
phase
and
the
combined
pP
and
sP
phases.
When
2044
THOMAS
It.
ttEATON
the
source
depth
is
very
small,
these
phases
all
arrive
very
closely
in
time
and
almost
total
annihilation
occurs.
Thus,
it
is
virtually
impossible
to
synthetically
produce
a
large
second
arrival
with
very
shallow
(less
than
3
km)
thrust
faulting.
I
am
faced
with
a
dilemma.
Modeling
of
strong
ground
motions
indicates
that
only
small
dislocations
are
acceptable
on
that
portion
of
the
fault
which
lies
beneath
PAC
at
depths
between
5
and
2.5
km.
Modeling
of
teleseismic
long-period
P
waves,
however,
indicates
that
significant
faulting
in
just
this
depth
range
provides
a
reasonable
interpretation
of
the
data.
It
appears
that
it
is
very
difficult
to
satisfy
both
the
strong-motion
and
teleseismic
data
using
the
fault
geometry
chosen
by
Langston
and
also
used
in
our
earlier
study.
Despite
this
difficulty,
the
teleseismic
M
o
(xlO
2s,
dyne
cm)
San
Fernando
Vertical
TRIA
/'~"~
0.59
P
Waveforms
Normu
165
Nodal
Planes
"
O
1
0
2
8
55
°
29
°
To.
068
90°
-'//_
-8o0
~"~
~8yn
..c
,-,,,
.%
c
sc-B-A
/
\../"
FIG.
4.
Observed
(top)
and
synthetic
(bottom)
long-period
vertical
P
waveforms
at
17
WWSSN
and
Canadian
network
stations
for
the
fault
model
Norma
163.
Lower
hemispherical
l~rojections
of
station
locations
are
shown,
as
well
as
the
nodal
planes
of
both
the
bottom
(solid
lines)
and
top
(dashed
lines)
sections
of
the
fault
model.
The
numbers
above
each
record
are
the
moments
10
~
dyne-cm)
which
produce
equal
peak
amplitudes
of
synthetic
and
observed
records.
The
observed
records
have
been
copied
from
Langston
(1978).
and
strong-motion
data
both
indicate
that
some
second
source
is
important
in
this
earthquake.
The
strong-motion
data
and
the
static
offset
data
also
indicate
that
there
was
significant
shallow
faulting,
and
this
shallow
faulting
is
the
most
likely
candidate
for
the
second
source.
How
can
we
construct
a
model
for
which
the
second
source
produces
the
large
shallow
faulting
and
also
extends
to
a
great
enough
depth
to
be
seen
teleseismically,
but
does
not
involve
significant
faulting
at
very
shallow
depths
beneath
PAC?
I
believe
that
changing
the
fault
geometry
offers
the
simplest
way
out
of
this
dilemma.
If
I
assume
that
the
dip
of
the
shallow
source
is
significantly
steeper
than
29
°,
then
it
is
possible
for
the
second
source
to
extend
to
greater
depths
without
passing
beneath
PAC
at
a
shallow
depth.