Bulletin
of
the
Seismological
Society
of
America,
Vol.
74,
No.3,
pp.
933-941,
June
1984
SEISMIC
POTENTIAL
ASSOCIATED
WITH
SUBDUCTION
IN
THE
NORTHWESTERN
UNITED
STATES
BY
THOMAS
H.
HEATON
AND
HIROO
KANAMORI
ABSTRACT
Despite
good
evidence
of
present-day
convergence
of
the
Juan
de
Fuca
and
North
American
plates,
there
has
been
remarkably
little
historical
seismic
activity
along
the
shallow
part
of
the
Juan
de
Fuca
subduction
zone.
Although
we
cannot
completely
rule
out
the
possibility
that
the
plate
motion
is
being
accommodated
by
aseismic
creep,
we
find
that
the
Juan
de
Fuca
subduction
zone
shares
many
features
with
other
subduction
zones
that
have
experienced
great
earthquakes.
INTRODUCTION
In
this
paper,
we
compare
the
mode
of
subduction
of
the
Juan
de
Fuca
plate
beneath
the
North
American
plate
with
that
of
other
subduction
zones.
We
show
that
the
Juan
de
Fuca
subduction
zone
shares
many
features
with
other
subduction
zones
that
experience
great
earthquakes,
while
several
features
indicative
of
aseismic
subduction
are
absent.
General
reviews
of
characteristics
of
the
subduction
process
are
given
by
Kanamori
(1977a),
Uyeda
and
Kanamori
(1979),
Ruff
and
Kanamori
(1980),
and
Lay
et
al.
(1982).
They
demonstrate
the
existence
of
striking
correlations
between
the
nature
of
seismic
energy
release
and
the
physical
characteristics
of
subduction
zones.
In
general,
they
find
that
total
seismic
energy
release
rates
are
highest
along
subduction
zones
where
young
oceanic
crust
is
subducted
rapidly.
They
interpret
this
result
to
be
a systematic
variation
in
seismic
coupling which
is
related
to
buoyancy
of
the
subducted
lithosphere.
We
begin
by
summarizing
the
results
of
the
studies
mentioned
above.
We
then
discuss
the
physical
characteristics
of
the
Juan
de
Fuca
subduction
zone,
describe
some
of
the
similarities
between
it
and
other
subduction
zones,
and
make
some
inferences
about
the
expected
seismic
potential
of
the
area.
SEISMIC
COUPLING
AND
EARTHQUAKE
SIZE
Kanamori
(1977a)
points
out
that
the
seismic
energy
release
rate
along
subduction
zones
is
not
a simple
linear
function
of
convergence
rates.
Ruff
and
Kanamori
(1980)
show
that
the
seismic
energy
release
rate
is closely
related
to
the
size
of
the
maximum
observed
earthquake
along
any
subduction
zone.
That
is,
the
cumulative
energy
release
from
small
events
is
usually
negligible
compared
to
the
energy
released
by
the
largest
events
in
a region.
Kanamori
(1977b)
shows
that,
on
a world-
wide
basis,
the
cumulative
seismic
energy
release
rate
is
closely
related
to
the
occurrence
of
very
large
earthquakes.
It
follows
that
the
seismic
energy
release
rate
along
individual
subduction
zones
is
closely
related
to
the
size
of
the
maximum
earthquake
observed
along
that
zone.
In
general,
shallow
low-angle
thrust
events
are
the
dominant
factor
in
determining
seismic
energy
release
rates.
Kanamori
(1977a)
concludes
that
variations
in
the
seismic
energy
release
rates
(i.e.,
size
of
maximum
earthquake)
for
differing
subduction
zones
are
caused
by
differences
in
seismic
coupling.
Strong
seismic
coupling
implies
that
slip
occurs
only
during
earthquakes,
whereas
weak
seismic
coupling
implies
that
slip
occurs
mainly
in
the
form
of
aseismic
creep.
933
934
THOMAS
H.
HEATON
AND
HIROO
KANAMORI
SEISMIC
COUPLING
AND
SUBDUCTION
ZONES
We
now
summarize
the
results
ofKanamori
(1977a),
Uyeda
and
Kanamori
(1979),
Ruff
and
Kanamori
(1980),
and
Lay
et
al.
(1982),
who
correlated
physical
charac-
teristics
of
the
subduction
process
with
the
maximum
earthquake
size
for
most
of
the
major
subduction
zones.
As
was
just
discussed,
this
correlation
is interpreted
in
terms
of
seismic
coupling,
which
is
related
to
the
buoyancy
of
the
subducted
lithosphere.
The
following
features
seem
well
correlated
with
observed
maximum
earthquake
size.
Convergence
rate
and
age
of
subducted
lithosphere.
In
Figure
1,
we
show
the
12
Chile
10
9.0
Kamchatka
0""'
8.1
9.1
C.
America
8
Aleutians
0
7.9
0
"'
....
Sumatra
>-
0
'
•
E
6
u
Q)
-+-
0
0:::
4
2
Age
m.y.
FIG.
1.
Relation
of
maximum
energy
magnitude,
Mw,
to
convergence
rate
and
age
of
subducted
lithosphere
for
major
subduction
zones.
The
contours
of
Mw
are
the
predicted
maximum
earthquake
magnitudes
resulting
from
linear
regression
of
observed
maximum
earthquake
magnitude
against
the
other
two
variables.
Dots
and
circles
are
subduction
zones
with
and
without
active
back-arc
basins,
respectively
(modified
from
Ruff
and
Kanamori,
1980).
relation
between
the
maximum
observed
energy
magnitude,
Mw,
and
the
conver-
gence
rate
and
age
of
subducted
lithosphere
for
the
major
subduction
zones.
Ruff
and
Kanamori
(1980)
performed
a linear
regression
of
convergence
rate
and
lith-
ospheric
age
against
the
maximum
observed
moment
magnitude,
and
the
solid
diagonal
lines
represent
the
best
linear
least-squares
fit.
It
is clear
that
the
maximum
observed
earthquake
size
increases
with
increasing
convergence
rate
and
decreasing
lithospheric
age.
In
Figure
2,
we
show
this
same
correlation.
In
this
figure,
however,
the
observed
maximum
energy
magnitude
is
plotted
against
the
energy
magnitude
predicted
from
the
regression
analysis
and
convergence
rate
and
lithospheric
age.
Ruff
and
Kanamori's
analysis
indicates
that
the
maximum
energy
magnitude
is
well
fit
by
the
following
relationship.
Mw
=
-0.00889T
+
0.134
V
+
7.96,
(1)
SEISMIC
POTENTIAL:
SUBDUCTION
IN
THE
NW
U.S.
935
where
Tis
the
age
of
the
subducting
plate
in
millions
of
years,
Vis
the
convergence
rate
in
centimeters/year,
and
the
standard
deviation
of
the
observed
Mw
around
the
predicted
value
is 0.4.
Presence
of
active
back-arc
basins.
In
Figures
1
and
2,
subduction
zones
with
and
without
active
back-arc
basins
are
plotted
as
dots
and
circles,
respectively.
Subduc-
tion
zones
without
active
back-arc
basins
are
clearly
associated
with
the
occurrence
of
large
shallow
subduction
earthquakes.
Thus,
the
absence
of
an
active
back
-arc
basins
seems
to
be
a good
indication
of
relatively
strong
seismic
coupling.
Depth
of
seismicity.
Ruff
and
Kanamori
(1980)
show
a good
inverse
correlation
between
the
maximum
depth
of
observed
seismicity
and
the
age
of
the
subducted
Mw=-
0.00889T
+0.134V+
7.96
C.
Chile
Peru
0
0
C.
America
0
Kuriles
RyuKyu
o
0
Sumatroo
o
NE.
Japan
SW.
/
0
•Tonga
New
Hebrides•
Kermadec
•Java
/
•
Scotia
/
•
New
Zealand
•
Izu-
Bonin
•
Caribbean
8
Observed
Mw
0
Aleutians
0
0
Aiaska
Japan
9
10
FIG.
2.
Maximum
observed
energy
magnitudes
plotted
against
maximum
energy
magnitudes
predicted
from
regression
analysis
shown
in
Figure
1.
T
is
the
age
of
the
subducted
plate
in
million
years,
and
V
is
the
convergence
rate
in
centimeters/year.
Dots
and
circles
are
subduction
zones
with
and
without
active
back-arc
basins
(modified
from
Kanamori,
1983).
plate,
but
the
corresponding
correlation
between
depth
of
seismicity
and
conver-
gence
rate
is poor.
Consequently,
there
is a weak
correlation
between
the
maximum
depth
of
seismicity
and
the
maximum
observed
earthquake
size.
However,
3 of
the
4 subduction
zones
that
have
produced
earthquakes
of
Mw
~
9.0
have
maximum
depths
of
seismicity
of
less
than
200
km.
Depth
of
oceanic
trench.
Uyeda
and
Kanamori
(1979)
suggest
that
strongly
coupled
subduction
zones
are
accompanied
by
shallow
oceanic
trenches,
whereas
weakly
coupled
subduction
zones
are
accompanied
by
very
deep
oceanic
trenches.
Similarly,
they
conclude
that
free-air
gravity
anomalies
tend
to
be
larger
for
those
trenches
with
weak
seismic
coupling.
Dip
of
Benioff-
Wadati
zone.
Uyeda
and
Kanamori
(1979)
conclude
that
strong
936
THOMAS
H.
HEATON
AND
HIROO
KANAMORI
seismic
coupling
is usually
associated
with
subduction
zones
having
relatively
gently
dipping
Benioff-
Wadati
zones.
The
uppermost
part
of
strongly
coupled
subduction
zones
generally
dips
between
10°
and
20°.
They
also
conclude
that
strongly
coupled
subduction
zones
are
characterized
by
the
presence
of
well-developed
fore-arc
basins,
which
are
believed
to
be
accretionary
prisms
of
sediments
that
develop
on
the
landward
wall
of
trenches.
Furthermore,
they
note
that
this
style
of
subduction
is
often
accompanied
by
crustal
uplift
and
compression
in
the
overriding
plate.
These
features
are
schematically
shown
in
Figure
3.
Chilean
-type
Mariana-
type
FIG.
3.
Diagrams
showing
characteristics
of
strongly
coupled
(Chilean
type)
and
weakly
coupled
(Mariana
type)
subduction
zones
(modified
from
Uyeda
and
Kanamori,
1979).
Topography
of
the
subducted
slab.
Kelleher
and
McCann
(1977)
and
Lay
et
al.
(1982)
both
note
that
very
large
subduction
earthquakes
are
more
likely
to
occur
in
regions
where
the
subducted
plate
has
smooth
topography.
That
is,
the
subduction
of
plates
with
transforms,
ridges,
or
numerous
seamounts
is
rarely
associated
with
great
earthquakes.
Lay
et
al.
(1982)
suggest
that
the
subduction
of
irregular
topog-
raphy
results
in
heterogeneous
strength
distributions
along
the
subduction
zone
and
thereby
inhibits
the
occurrence
of
earthquakes
of
large
dimension.
Seismic
quiescence.
Although
weakly
coupled
subduction
zones
may
display
a
total
absence
of
major
earthquakes,
they
still
have
relatively
high
seismic
activity
at
small
magnitudes.
Benioff-
W
adati
zones
capable
of
truly
great
earthquakes,
however,
often
show
significant
periods
of
extremely
low
seismicity
(Lay
et
al.,
SEISMIC
POTENTIAL:
SUBDUCTION
IN
THE
NW
U.S.
937
1982).
This
pattern
may
be
somewhat
analogous
to
the
seismicity
observed
along
the
San
Andreas
fault
in
California.
The
central
creeping
portion
of
the
fault
is
characterized
by
relatively
high
seismicity,
but
no
large
earthquakes.
However,
the
portions
of
the
fault
that
are
capable
of
great
earthquakes
(1857
and
1906
breaks)
are
almost
devoid
of
present-day
seismicity.
THE
JUAN
DE
FUCA
SUBDUCTION
ZONE
We
have
seen
that
there
are
systematic
differences
between
subduction
zones
that
are
capable
of great
earthquakes
and
those
that
are not.
The
Juan
de
Fuca
subduction
zone
has
been
ignored
in
the
studies
that
established
these
differences.
The
Juan
de
Fuca
subduction
zone
has
been
considered
somewhat
anomalous
because
there
has
been
virtually
no
shallow
thrust
seismicity
of
the
type
we
usually
associate
with
active
subduction
zones.
There
are
several
possible
explanations
for
this
low
level
of
seismicity:
(1)
the
North
American
and
Juan
de
Fuca
plates
are
no
longer
converging;
(2)
the
plates
are
converging
but
slip
is accommodated
aseismi-
cally;
and
(3)
the
northwestern
United
States
is
a major
seismic
gap
that
is locked
and
presently
seismically
quiescent,
but
that
will
fail
in
great
earthquakes
in
the
future.
Each
of
these
possibilities
is discussed
below.
Present-day
convergence.
The
geometry
of
important
plate
boundaries
and
the
seismicity
in
the
Pacific
Northwest
are
shown
in
Figure
4.
Delaney
et
al.
(1981)
note
that
there
appears
to
be
43
km
of
new
oceanic
crust
formed
on
the
Juan
de
Fuca
ridge
since
the
700,000-yr-old
Brunhes-Matuyama
magnetic
reversal,
yielding
a half-spreading-rate
of
about
3 em/yr.
The
fact
that
the
oldest
crust
found
in
the
Juan
de
Fuca
plate
is
on
the
order
of
10
m.y.
old
(Atwater,
1970)
indicates
that
subduction
has
occurred
in
the
past.
The
average
convergence
rate
for
the
past
5
m.y.
has
been
estimated
from
magnetic
reversal
data
to
be
3.5
cm/yr
by
Riddihough
(1977),
4.2
cm/yr
by
Chase
et
al.
(1975),
and
3.0
cm/yr
by
Nishimura
et
al.
(1984).
Hyndman
and
Weichert
(1983)
show
that
historic
seismicity
can
account
for
slip
rates
expected
from
magnetic
reversal
data
on
all
plate
boundaries
between
the
Pacific
plate
and
the
North
American
plate
except
on
the
Juan
de
Fuca
subduction
zone.
Furthermore,
it
seems
difficult
to
concoct
a model
of
plate
motions
which
has
3.5
cm/yr
slip
rates on
faults
both
north
and
south
of
the
Juan
de
Fuca
subduction
zone,
but
with
no
convergence
on
the
subduction
zone
itself.
It
thus
appears
that
presently
available
evidence
supports
present-day
plate
convergence
of
3
to
4
em/
yr
across
the
Juan
de
Fuca
subduction
zone.
Physical
features
of
the
Juan
de
Fuca
subduction
zone
and
seismic
coupling.
The
subducted
part
of
the
Juan
de
Fuca
plate
appears
to
be
very
young,
probably
between
10
and
15
m.y.
old.
We
earlier
said
that
subduction
of
young
oceanic
crust
usually
is
associated
with
strong
coupling.
The
subduction
rate,
3
to
4 cm/yr,
however,
is
not
particularly
high,
and
this
rate
does
not
in
itself
indicate
particularly
strong
coupling.
If
we
insert
these
values
into
equation
(1),
then
we
predict
a maximum
moment
magnitude
of
8.3
±
.5.
This
high
value
is supported
by
inspecting
Figure
1.
We
see
that
strong
coupling
is
associated
with
every
subduction
zone
where
the
subducted
plate
is less
than
40
m.y.
old.
The
notion
of
strong
seismic
coupling
along
the
Juan
de
Fuca
subduction
zone
is
further
supported
by
the
fact
that
there
is
clearly
no
active
back-arc
basin
in
the
northwestern
United
States.
A
cross
section
of
seismicity
in
the
Puget
Sound
region
is
shown
in
Figure
5.
Although
a clear
Benioff-Wadati
zone
can
be
seen,
seismicity
deeper
than
100
km
has
not
been
observed.
The
dip
of
the
Benioff-
Wadati
zone,
as
defined
by
the
axis
of
the
trench
and
the
pattern
of
seismicity
beneath
Puget
Sound,
is
between
10°
•
•
•
•
N~
foll.nctoelna
Fr-=tur•""fa,..
•
I
\
•
.
.
·
.
•.•
+·
•
.·
o.=:.,
~
0
.,
'i'rv
~.
~
··---·--:;;
--·--~-----~
.
IT.
"\•
~.
·•
••
..
•
J
:.
•
EXPLANATION
•
2:
7.0
M~~gnitudtt
•
8.0-8.9
M~gnltude
•
-++-
5.0
- 5,9
Mtgnitude
4.0-
4.9
Magnitude
Relative
fMJit
movement
Direction
of
sprttedlng
at
ridge
Approximate
bne
of
Continental
Slope
0
_
__._
_
__!_~
I
km
300
FIG.
4.
Major
tectonic
features
and
associated
seismicity
of
the
Pacific
Northwest
(modified
from
Washington
Public
Power
Supply
System,
1983).
t.O
w
CP
>-3
::t
0
is:
>
w
::t
::t
tr:l
~
0
z
>
z
0
::t
;;
0
0
~
>
z
>
is:
0
~