TECTONICS,

VOL. 9,

PALEOSEISMICITY
TION
THE

ZONE:

OF

NO. 4,

THE

EVIDENCE

CASCADIA

FROM

OREGON-WASItINGTON

PAGES 569-583, AUGUST 1990

SUBDUC-

TURBIDITES

OFF

MARGIN

John Adams

GeophysicsDivision, Geological Survey of Canada,

Ottawa, Ontario

Abstract. Cores from Cascadia deep-seachannel contain

sequencesof turbidites that can be correlated and dated
by the first occurrenceof volcanic glass from the Mount

INTRODUCTION

Mazamaeruption(6845 4- 50 radiocarbonyrBP). Turbid-

disturbances

ity currents fi'om the tributaries appear to have occurred

synchronouslyto form single depositsin the main channel,
there being only 13 turbiditc deposits in the lower main
channel since the Mazama eruption, instead of the twice
as many expected if the tributaries had behaved independently. In addition to the Cascadia Channel, 13 postMazama turbidires have been deposited in the Astoria
Canyon and at two sites off Cape Blanco, sample locations that span 580 km of the Oregon-Washingtonmargin. Pelagic intervals deposited between the turbidites
suggestthat in each place the turbidity currents occurred
fairly regularly, every 590 4- 170 years on average. The
best explanation of the spatial and temporal extent of the
data is that the turbidity currents were triggered by 13

true nature of these events was not explained until the

great earthquakeson the Cascadiasubductionzone. The

variability of turbiditc timing is similar to that for great
earthquake cycles. The thickness of the topmost pelagic
layer suggests
the last eventwas300 4- 60 yearsago(from

Earthquakes have long been known to causesubmarine

sufficient

to break

submarine

cables.

The

classicpaper by Heezenand Ewing [1952] analyzedthe

sequentialbreakageof telegraph cablesfollowing the 1929
"Grand Banks" earthquake off Canada's eastern margin
and showed that a submarine debris flow and subsequent
turbidity current had travelled down the Laurentian Fan

at speedsup to 65 km/hr and depositeda turbidireon the

deep sea floor. Oceanographersin the 1950s and 1960s
cored many turbidites in all the world's oceans, showed
that they were chiefly associatedwith canyonsand channels on submarine fans, and speculated on their rates of
occurrence. Very high rates - every fev years - off deltas
such as the Magdalena and Congo suggestedthat a prime
causeof the turbidity currents was sediment instability due
to rapid sedimentation; the role of infrequent earthquakes
was less easy to assess.
A classicstudy in turbidites and deep sea-sedimentation
was carried out in the late 1960s by a group of students

three placesalongthe margin), but this numbermay be

(Griggs, Nelson,Carlson,and Duncan) at OregonState

a biased underestimate. It is, however, consistent with

the youngestsudden-subsidence
event on the Washington

University under the supervision of L. D. Kulm. The

most spectacular results came from the Cascadia Channel

coast.

off the Oregonand southernWashingtonmargin [Griggs,

1969; Griggs and Kulm, 1970]. Coresfrom the Cascadia

The

turbiditc

data

demonstrate

that

the near-term

hazard of a great earthquake on the Cascadia subduction

zone is of the order of 2 - 10% in the next 50 years.

Channel contained sequencesof turbidites that could be

correlatedand dated by the first (lowest) occurrenceof

Copyright 1990
by the American Geophysical Union.
Paper number 89TC03759.

0278-7407/90/89TC-03759510.00.

volcanic glassfrom the Mount Mazama eruption. The predominantly muddy sedimentation in the channel and the
similarity of the coressuggeststhat they represent a complete record of turbidity currents in the Holocene. Here,
20 years later, the results are used to address the problem
of great-earthquake hazard in the Pacific Northwest.

570

THE

Adams: Paleoseismicityof the CascadiaSubductionZone

EARTHQUAKE

PROBLEM

IN THE

PACIFIC

Washington and Oregon began to change [e.g., Riddihoughand Hyndman, 1976]. Earlier wor. had established

Klamath Mountains in southern Oregon. Much sediment

from the Oregon and Washington coastalrangesand from
Vancouver Island is trapped in coastal inlets, and sediment from the Fraser River has been trapped in the Strait
of Georgia for the last 8000 years.
At the present time, sediment from the Columbia River
is carried north along the shelf and is depositedon the shelf

that the Juan de Fuca plate was convergingtoward North

America. However the lack of seismicity on the plate in-

views and cites the literature and infers that of the 21 x 106

terface

t yr-x carried by the Columbia, 17% is depositedon the

N O RTHWES

About a dozen years ago, perceptions of the nature of

the subduction zone beneath southern British Columbia,

lead some to consider

that

the subduction

was oc-

curring extremely slowly or had stopped.

More

recent studies of the deformation

front at the base

of the slope [e.g., Barnard, 1978], of geodeticdeformation rates on land [Ando and Balazs, 1979;Reilingerand
Adams, 1982; Savageand Lisowski, 1988], and of onshoretectonicdeformationsuch as warped terraces(see
for example,Adams [1984],but also]Vest and McCrumb
[1988a,b]and Atwater [1988a])have confirmedthat the
Oregon-Washington margin is being deformed at rates as
rapid as those above other active subduction zones. The
conundrumis this: with suchhigh deformation rates, why
have there been no large thrust earthquakeson the plate
interface?

the

Cascadia

subduction

zone

and

other

zones

worldwide,Heaton and Kanamori [1984]showthat physical conditionsimply that the subductionzone could generate earthquakesof magnitude 8.3 q- 0.5. Although the
magnituderange they obtain is large, their study strongly
suggeststhat infrequent great earthquakes are a distinct
possibility,even though none have occurredhistorically.
Independently of Heaton and Kanamori, I suggested

[Adams,1984;1985]that the simplestwayto reconcilethe

discrepanciesbetween long-term and short-term evidence
for deformation, strain accumulation, and stressdirections
was if there was a great-earthquake cycle of long duration. I discussedhow an analysis of landslides, landslidedammed lakes, drowned trees below sea level, and uplifted
beachescould be used to determine evidencefor past great
earthquakes, and mentioned the turbidites in the Cascadia

Channeldescribedby Griggsand Kulm [1970]as providing a minimum recurrenceinterval for such earthquakes.
The present paper analysesthe turbidite record further,
and showsthat it providesgood evidencefor earthquakes
about every 600 years. Such conclusions are important

for the current debate [Heatonand Hartzell, 1987]on the

level of seismic hazard in the Pacific Northwest, and are
an important complement to the large body of onshore
research into paleoseismicitythat has occurred since 1986

[e.g.,Atwater, 1987a,b].
SEDIMENT

upper slope, most of it in the upper reachesof the Willapa,

Grays, and Quinault canyons,which are tributaries of the
Cascadia Channel, and some in the head of the Astoria
Canyon. The substantial storage in the fluvial system and

on the shelfis ableto averageout the short-term(10x 102yr) climaticvariations

in supply.
The sediment accumulateson the slope until submarine
failure occurs and the newly-accumulated sheet of sediment then sweeps dovn the channels as a density flow or

turbidity current(Figure2). Griggsand Kulm [1970]suggest that each major turbidity current takes about two
days to travel the 735 km length of the Cascadia Channel

and carriesabout 525 1106 ms of sedimentin a flow up

In a stimulating paper that addressessimilarities between

and the edgeof the slope(Figure 1). Sternberg[1986]re-

DISPERSAL

OFF

THE

CASCADIA

MARGIN

To understand sedimentation off Oregon and Washington, it is necessary to understand the sediment sources,
the processesby which sediment is carried to the deep-sea
floor, and the temporal changesthat have affected these
processes. The chief source of sediment is the Columbia
River, with lesser amounts of sediment coming from the

to 17 km wide and 100 m high. Low deposition rates in

the middle

channel

indicate

that

most of the sediment

is

carried as far as the Blanco Fracture zone, where it ponds,

forming individual turbidite layers up to several meters

thick [Griggsand Kulm, 1970].

The present dispersal pattern described above has lasted
only for the past 6000-7000 years. Before about 7000 years
ago, sea level was lower and still rising as water from the
melting ice sheetswas added to the oceans. During lower

sea levels, the sediment was not carried north along the
shelf to the same degree, but travelled directly to the deep
sea down the Astoria canyon and channel. Cores on the
middle Astoria Channel show little or no turbidite deposi-

tion in the last 7000 years[Nelson,1976],althougha core

from the mouthof the AstoriaCanyon(6502-PC1)shows
many thin turbidites, presumablygeneratedfrom the sediment spill-overfrom the Columbiashownon Figure 1.
A typical turbidite in the Cascadia Channel system consists of an erosional base, a fine sandy basal layer and
an upper layer of silt and clay that fines upwards. Becausethis sediment was derived from the biologically-rich
outer shelf and upper slope, it contains shallow-water shell
fragments and is high in carbon from partially decayed

organicdebris [Griggset al., 1969, p. 168]. Once triggered, deposition from the turbidity current takes a few
hours to days. Even the clay fraction must be deposited
quite rapidly becausedeep-water currents would move any
remaining suspendedclays away from the newly-deposited
turbidite. After all deposition has ceased, pelagic or
hemipelagic sedimentation resumes, which is partly biological- foraminifera and radiolaria tests - together with
small amounts of clay carried in suspensionfrom the continent. In contrast to the turbidity currents which deposit
300 - 3000 mm in a matter of days, the pelagic sediment

may accumulateat lessthan 0.1 mm yr-1, slowerfarther

away from the continent. Between turbidity currents there
is colonisation of the newly-formed sea-bottom by burrow-

Adams: Paleoseismicityof the CascadiaSubductionZone

571

Scale

50

122 ,=,

124W
I

126

t00

150kin

IN

53-18

MOUNT
ST. HELENS
6705-5

6705

6508-KI
6705-9

6502-PC

IORTLAND

6609-27

6509-15

509 27

24 j
/

//

.2 --

MT.MAZAMA

:APE

BLANCO

/609-1
__..
124

Fig.

CRATER LAKE/

ORE

%/////
.___-

12.2

1. Map of the Oregon-Washingtonmargin showing the extent of the air-fall Mazama

tephra [Fryxell,1965],the pattern of sedimentdispersalnorth from the ColumbiaRiver (with

amountsof sedimentas percentageof the river input, after Sternberg[1986]), canyonson the
Washingtoncontinentalslope (J, Juan de Fuca, Q, Quinault, G, Grays, W, Willapa, and A,
Astoria), submarinechannelsleadingto the deep-seafloor, the locationof coresmentioned(all
from OregonState Universityexceptcore53-18 whichis Universityof Washington),and (within
large circles)the numberof post-Mazamaturbidiresas discussed
in the text. The baseof the
continentalslope(dashedline) and the 200-misobath,markingthe edgeof the shelf,are also
shown.

572

Adams: Paleoseismicity of the Cascadia Subduction Zone

CORE
6609-24

Fig. 2. Perspectiveview showingschematicallythe genesis

of turbidity currents in the Cascadia Channel. Sediment

is carrieddownthe ColumbiaRiver (1), drifts north along

the shelf (2), and accumulatesat the top of the continental slopein the headsof deep-seacanyons(3). According
to the hypothesis presented in this paper, a great thrust

earthquakeon the subductionzone(4) causesstrongshaking of the shelf and slope. The shaking causessediment
liquefaction and slumping simultaneously at many places
along the margin. The resultant massiveunder-sea debris
flows mix with the water to become a series of turbidity

currentstravelling synchronously
down the channels(5).
At junctions, the tributary turbidity currents coalesceto
travel down the Cascadia Channel as one large turbidity

current (6).

ing organisms(thriving on the high organiccontentof the

turbidite silts) which causebioturbationof the upper part
of the turbidite and of the slowly accumulating pelagic
sediment[Griggset al., 1969].
The turbidites off Oregon and Washington have been

Fig. 3. Sketch of core 6609-24, made from the core log,

showing the even thicknessesof the turbidite and pelagic

sampledas 2 - 8 m longpistoncores(e.g., Figure3) taken

(shaded)layers. Only the top 6.4 m of the coreis shown.

in 2.0 - 2.5 km of water. For the purposesof this study

The x's in the 13th turbidite show the position of the earliest Mazama glass.

the detailed core logs (which describepiston and associated pilot coresseparately)compiledby studentsand staff
at Oregon State University have been used to determine
the turbidite and pelagic thicknessesshown on Table 1.
Most of the coresstill exist in storageat the university, but
the logs have the advantagethat they were written when
the cores were freshly recovered, and were described by
people with considerableexperience. In almost all cases,
the original colorsand other evidencerecordedin the logs
are sufficient to determine the base and top of each turbidite, the thicknessof the overlyingpelagicsediment, and
the amount of bioturbation; key parameters used in the
present analysis.
CORRELATION
BY

TtIE

MAZAMA

AND

DATING

OF THE

TURBIDITES

TEPtIRA

Becauselarge turbidity currents travel down the length

of the Cascadia Channel, each can leave a single deposit,
and turbidite layers along the length of the channelshould

correlate(thoughsomemay be missinglocally in parts of

the channelwhereerosiondominates).A uniqueeventin

the last 10,000 years was the eruption of Mount Mazama

at 6845 4- 50 radiocarbonyrBP [Bacon,1983].Duringthe

brief eruptionabout34 kma of tephra("volcanicash")

was depositedon Oregon and Washington. Sites on land
show that prevailing westerly winds allowed little if any
of the airfall tephra to fall as far west as the coastal

ranges(Figure 1). Much, however,fell in the catchment

of the Columbia River and this loose material was easily
washed into the sea and deposited on the shelf. Experience with the much smaller Mount St. Helens eruption in
1980 showed that only about 18 months were needed for

the tephra to travel to the headof Quinault Canyon[e.g.,

Sternberg,1986],and there is no reasonto think that the
Mazama tephra would have behaved differently.
When the shelf-edge deposits slumped, the turbidity
current carried the tephra-rich sediment down into channels. Cores taken along and near the channelscontain the

Adams: Paleoseismicityof the CascadiaSubductionZone

TABLE 1'

Pelagic

and Turbidire

573

Thicknesses

in Selected

Oregon-Washington

Cores

Turbidire

sequence

6705-2

6705-6

6705-5

6508-K1

6509-15

6509-27

25

15

6609-24

6509-26

6502-PC1

6604-12

number #

20

55
50

>60
80

90

10
3

60

285

35

140
50

170
50

110
40

75

10

40

180
25

11

280
45

100
25

12

60

13

240

50
80

14

110
40

75
70

30
30

430*

120
10

15

75

16

125'

170

350
60

150
210

>690

620
50

170
150

730
130

220

370

170

340

50

15
90*

190
240

180

730

685

70

640

890

10

10

20

120

50

345

160
130

80

85
130

430

120

30

75

15

10

550

290
30

130

720

480

75

130

25

30

15
125

130

>1630

500
40

110

515

620

660
60

180

55

40

95

900

30

30

80

10

35
110

1530

20
130

720

455

240

160

130
640

60

20

40

20

30

60

60

75
880

130

230

85

620

565

1280

70

40

65

230
150

120

55

55

25

30

30

50

30

160

180
120
35

60

60

300

1070

180
630

445

50

250

30

55

25

15

25

40

80

50

55

290

760

10

10

40

45

40

100
110

180

170

110

65

>420

780

70

470
25

70

20
50

20

45

30

50

90

540

730

40

15

>230
60

25

15
120

35

25

25

570

750
50

10

50
680

180
80

>240

530

10

> 20
15

25

10
55

170
85

660

40

120

>160
20

40

40
110

15

35

20
4

>670

140
385*

550
10

100'

590*

750*

70
5

690*

75

Pelagic
(left-hand)
and turbidire
(right-hand
column of each pair)
# "1" represents
the topmost turbidire
in each core,
*
Lowest post-Mazama turbidire
(see text).

the tephra, but coresdistantfrom channelsdo not [Nelson,

1976], confirmingthat air-fall tephra did not fall offshore
and that deposition was via river, shelf, and channelled
turbidity current.
By examining the coarsefraction at the base of each
turbidite for freshvolcanicglass,it is possibleto determine

thicknesses

(in

mm) are given

to the nearest

5mm.

and pelagic thicknessesand the lowest layer containing

Mazama glass have been determined from the core logs,
in almostall casesusingthe describer's
boundaries(Table
1). In CascadiaChannel cores6509-15and 6609-24the
Mazama glass occurs in the 13th turbidite from the top;
in cores6705-6 and 6508-K1 it is in the 14th; for core 6705-

the first (lowest) turbidite that containsabundantfresh

glass. Deeper layerscontainsparse,weatheredglassfrom

5 there are no marginal notes on the core log indicating

the position of the Mazama glass,but from the recurrence

eruptions older than the Mazama, but the first turbidite

containing Mazama glassis very distinctive, and in some
coresthe sandy fraction is 25% glass(e.g., in core 6508-

intervalsgivenby Griggsand Kulm [1970,Table3] it is in

K1).
The first presenceof the redepositedMazam glass in
the deep-seaturbidites providestime control with two important consequences:it allows the first turbidire with
abundant glass to be traced down the Cascadia Channel,
so confirmingthe nature of the event, and it allowscalculation of the mean period betweenturbidity currents. Griggs

and Kulm [1970]usedthe Mazamaglassto computerecurrenceintervals of 410 - 510 yearsfor the turbidity currents
in the Cascdia

Channel.

the 15th turbidite; and for core 6705-2 it is probably the

16th, but might be the as low as the 19th, as someof the
sedimentary units are poorly defined.
The position of the Mazama glassin the three coresfrom

the lowerCascadiaChannelsupportsGriggs'[1969]assertion that not only must the lowest turbidite containing
Mazama glasscorrelate, but so must each of the 12 over-

lying turbidites(i.e., the 13 turbiditesin the coresrepresent the same 13 turbidity currents). Griggs'assertionis
further supported by correlatable vertical variations in the
degreeof bioturbation in the top eight layersof cores6509-

27, 6609-24,and 6509-26 [Griggset al., 1969]. Although

In this pper I invert Griggs and Kulm's approachin

order to discussfirsfly the number of turbidity currents

the pattern of bioturbation is distinctive in these three

since the Mzam eruption, and then the mean recurrence

interval. A typical core is shown in Figure 3. Turbidire

channel, it is unlikely that similar correlationscould be ex-

cores and allows their

correlation

over 65 km of the lower

tendedto the coresin the tributary channels(discussed

be-

574

Adams: Paleoseismicityof the Cascadia Subduction Zone

low), becausethosecoresare moreuniformlybioturbated.

downstream is probably becauseone small turbidity cur-

Also, until the reasonsfor the degreesof bioturbation are

rent did not flow downstream

fully understood- Griggs et al. [1969]suggestedvariable

6508-K1.

time between events as one cause, but also noted that no

correlation exists between pelagic thicknessand depth of
burrowing- it is not possible to correlate the Cascadia
Channel coreswith those off southern Oregon.
Upstream from the correlated cores, core 6508-K1 con-

The coresin the tributariessuggestthat a few (2 or 3

out of a maximumof 16) smallturbiditycurrentsmayhave

tains 14 post-Mazama turbidites (from the core logs;

thoughTable 3 of Griggsand Kulm [1970]implies13),
and evidently one of the 14 was a small turbidity current
that

did not travel

as far downstream

as core 6509-15.

to the confluence

and to core

been generated that were smaller than the 13 turbidity

currents

recorded

in the lower channel

and did not travel

as far. With this in mind, the remarkable inference can

be made that pairs of turbidity currents were generated
synchronously in the two tributaries. This arises because
there have been 15 post-Mazama turbidites in the Willapa
Channel and 14 turbidites in the northern channel, but

Downstream from core 6509-26, only 4 of the 13 turbidity currents since the Mazama eruption passed through
the gap in the Blanco Fracture zone and onto the Tufts
Abyssal plain.
Besidesthe coresdiscussedabove, there are many others
off Oregon and Washington that contain Mazama glass.
These cores contain varying numbers of turbidites over-

only 14 turbiditesbelowtheir confluence(whereup to 29

mightbe expectedif theflowshad occuredindependently).

lying the earliest occurrenceof Mazama glass, although

none, to my knowledge,has more than 13 overlying tur-

in the lower Cascadia

bidites. In the middleCascadiaChannelsomecores(e.g.,

core6705-13,GriggsandKulm [1970])havefewor nopostMazama turbidites becausethe channelis actively eroding
and little sediment is being deposited. Becausethe turbidity currents must have passedthrough this section to
deposit the 13 turbidites in the lower channel, these cores
are seen to provide an incomplete record. Other coreson
the channel levees are high above the channel floor and

contain 8-10 post-Mazamaturbidites (e.g., cores6705-7

and 6705-9). As noted by Griggsand Kulm [1970],they
represent only the largest turbidity currents, and so may
provide interesting information about the size distribution
of the turbidity currents, but only a minimum estimate of
the total number. Away from the CascadiaChannel, there
are many coresthat contain a few post-Mazama turbidites

or someHoloceneturbidites(e.g., coresshownby Duncan

[1968]and Nelson[1976]). They recordsomeHolocene
turbidity currents(not necessarily
the sameonesin different places),but addlittle informationaboutthe maximum
numberof turbiditesthat may be foundin a givenregion.
SYNCHRONOUS
INDIVIDUAL

Core 6508-K1

TURBIDITY

CURRENTS

FROM

TRIBUTARIES

lies in the Cascadia

Channel

downstream

of the confluence of two main tributaries, the northern from the Juan de Fuca Canyon and nearby canyons,
and the southern from the Willapa, Grays, and Quinault

canyons(Figure1). In the northerntributary,core6705-6

has 14 turbidites since the Mazama eruption, the same as
below the confluence. Upstream from core 6705-6, core

6705-2 contains16 post-Mazamaturbidites(accordingto

Griggsand Kulm [1970]),and by inferencetwo smallturbidity currents did not flow down the channel as far as core
6705-6. In the Quinault Canyon, core 53-18 contains "at

least 14" post-Mazamaturbidites[Barnard,1978,p. 111].

Downstream, in the main Willapa Canyon, core 6705-5 has
15 turbidites sincethe Mazama eruption, and the presence
of only 14 post-Mazama turbidites in core 6508-K1 farther

An alternative, though lessplausible explanation, is that

exactly 15 of the turbidity currents, somefrom the Willapa
Channel and the rest from the northern channel, died out
before the confluence.This would require a higher rate of
attrition than is evident in either the upper tributaries or
Channel.

Therefore turbidity currents were generated synchronously in two independent channels, the headwaters of
which are a minimum

of 50 km and a maximum

of 150 km

apart, and each pair of small turbidity currents merged

to form one large turbidity current. That the number in
the main channel is only 13 implies that every turbidity
current in the two tributaries occurredsynchronously,i.e.
that synchronous occurrence is the rule rather than the
exception. A similar argument can be made for other tributary channels that join the main channel farther downstream.

Several of these tributaries

contain 8- 10 Holocene

turbidites(e.g., cores6609-27and 6705-10),and yet the

number of turbidites

in the main channel does not increase

beyond the 13 already present. Either all these turbidity

currents predate the Mazama or at least somewere synchronouswith the 13 turbidity currents in the main channel.

How closely in time did the tributary turbidity currents

occur? If the turbidites xvereseparated by 50 years or

so, a thin pelagicdrape (5 mm, at rates derivedlater in

Figure 5) betweenthe pair of eventsmight be expected.
For separations of a decade or so, there need be no perceptible pelagic layer, but some bioturbation in the lower
turbidite could have occured. For separationsof a day or
greater, each turbidite would still have left an individual
fining upwards sequenceso each event would be a doublet:
coarsebase + fine top, coarsebase -I- fine top, pelagic
layer. Furthermore these doubletswould persistdown the
Cascadia

Channel.

Examination

of core 6508-K1

and the

cores downstream does not reveal any double events, so

I conclude that the tributary turbidity currents occurred
lessthan a few days, and perhapslessthan hours,apart.
My rough analysis of the time needed for turbidity
currents to travel downstream from the canyon heads
mapped on Figure 1 suggeststhat only a few tens of hours
might separate currentsfrom different, but synchronouslytriggered sourcesat the canyon heads, so that the transport and deposition event in the main channel would be almost continuous.Using realistic estimatesof flow velocity
and sediment concentration, some tens of hours would be

Adams' Paleoseismicity of the Cascadia Subduction Zone

575

needed to transport the sediment in an average Cascadia

turbidire, which is consistent with xvith the above. Thus
the coalescingflows combineto form the deposit described
as a singleturbidire. Similar coalescingof channelledturbidity currents to form a single turbidire deposit has been

ally thick hemipelagic sediments,again due to closenessto

the Columbia

River

source.

Both the coreoff Cape Blanco(6604-12)and the oneoff

the RogueRiver (6609-1) are remotefrom the Columbia

Rico Trench [Pilkey,1988],and has alsobeeninterpreted

River mouth, and heavy mineral analysis shows that at

least since the Mazama eruption their sediment has come
from the Klamath Mountains, inland and south of Cape

to representthe effectsof a synchronousregional trigger.

Blanco[Duncan,1968]. Despitetheir remoteness

from the

SPATIAL

Cascadia and Astoria Channel systems they too contain

13 post-Mazama turbidites.

documented

RENTS

from the Mediterranean

EXTENT

SINCE

OF

THE

THE

and from the Puerto

13 TURBIDITY

MAZAMA

CUR-

It

ERUPTION

is at least

50

km

from

the

Juan

de Fuca

to

the

Quinault Canyon, 100 km across the Quinault-Willapa

canyon system, 30 km more to the Astoria Canyon, 350
km farther to Cape Blanco, and then 50 km to the Rogue.
From all five sitesalong the margin there have been 13 turbidity currents since the Mazama. I consider the simplest
explanation for the synchronousevents in neighbouring
tributaries, and the same numbers of events at sites that
span 580 km, is a seriesof great earthquakes.

Figure 1 shows three other cores that lie outside the

Cascadia Channel system but contain the Mazama turbidite, one from Astoria Canyon, and two from the Blanco

Valley physiographicprovince [Duncan, 1968] off Cape

Blanco. In all three, the Mazama occurs in the 13th turbidire from the top.
The Astoria Canyon was a major conduit for sediment

up until 7000 yearsago. While Nelson[1976]showsthat

the lower Astoria

Fan has been almost inactive

since about

REGULAR

RECURRENCE

OF

TIIE

13 TURBIDITY

CURRENTS

the time of the Mazama (presumablydueto the changein

sediment dispersal that accompaniedthe sea level change

at that time), core6502-PC1shovs13 thin (averagethickness130mm) post-Mazamaturbidity currentsoccurredin

Up to this point the paper has discussedonly the numbers of turbidires without regard to their timing. As noted

the upper channel. Apparently there is enoughspill-overof

the Columbia River sedimentcarried north alongthe shelf

by Griggs [1969] and Griggs and Kulm [1970], the tur-

currents that do not travel all the way down the Astoria

bidires in the Cascadiasystemhave generally similar thicknessesand are separatedby pelagic intervals also of generally similar thickness. The left half of Figure 4 plots

Channel.

turbidire and "pelagic"(hemipelagicplus pelagic)thick-

(Figure 1) for small turbidity currentsto be generated,

The thin turbidires

CORE

are interbedded

6609-

with unusu-

24

8C

E 60
mm

Pelogic

loo

>,,40

5o

ro 20

OVERLYING

o 80

mm

Turbidite
60

6O

x
x

'F 4O

x
x

200tBOTTOMTOP
2O

x
x

UNDERLYING

TURBIDITE UNIT

2oo

4.oo

Turbidite

6oo

Thickness(ram)

Fig. 4. Pelagicand turbidite thicknesses

in core6609-24. (Left) Plots of unit thicknessagainst
sequencenumber. (Right) Plots of relationshipbetweenpelagicthicknessand the thicknessof
the overlying(top) and underlying(bottom) turbidite.

800

576

Adams: Paleoseismicityof the CascadiaSubductionZone

nessesfor core 6609-24, in the lower Cascadia Channel.

Both the turbidite (440 q- 130mm each)and pelagic(55

q- 19 mm) layersare surprisinglyregularin thickness(the
standard

deviation

is 29% of the mean for the turbidite

units and 35% of the meanfor the pelagicunits). For the

pelagic
thicknesses,
themeasurement
errorisaboutq-to
q- ofeachvalue.
Some correlation between the thicknesses of adjacent
pelagic and turbidite units might be expected. For example, an unusually long period betweeneventswould give

both a thick pelagicunit (becauseof the longertime available for the pelagic sediment to accumulateon the deep
seafloor) and a thick overlyingturbidite unit (becauseof
the longer time availablefor the slope sedimentsto accumulate in the canyonheads).Howeverthis positivecorre-

glass. Because the St. Helens tephra was carried to the

head of Quinault Canyon in 18 months, it is reasonable
to suppose that Mazama tephra travelled similarly fast
and was available for slumping almost immediately after
eruption. When did the next turbidity current occur? For
13 turbidity currents the mean recurrence interval is about
600 years, so the next one would probably have been 300
q- 300 years later. Therefore, 7360 q- 400 yearsbefore 1989
is adopted as the age of the 13th turbidite.
One might speculatethat the large volume of tephra
must have caused more rapid accumulation on the shelf
edge than normal, thus possibly reducing the time to the

4. For someother cores(e.g., 6604-12) there is an in-

next sedimentfailure (which need not have been seismically induced). This might be the explanationfor the
'14th' turbidite in the upper CascadiaChannel(e.g., cores
6705-6 and 6508-K1). Futhermore,if the Mazama eruption were triggeredby one great earthquake,the likelihood

verse relationship between pelagic and overlying turbidite

thicknesses,perhapssuggestingthe thicker turbidites were
depositedfrom larger turbidity currentsthat erodedmore
of the underlying pelagic sediments.
The variation of pelagicthicknessesfor severalcorescan
be seenconvenientlyby plotting cumulativepelagicthick-

of the next would be reduced for about 600 years. Neither

of these extreme adjustmentsto the age of the first postMazama turbidite can yet be proven.
Since the 13th turbidite there have been 12 turbidites,
12 inter-turbidite intervals and a period of time since the
last turbidite. A priori, the length of the last period is

nessagainstturbidite sequencenumber (Figure 5). For

notknown
soistakenas q- interval
fora totalof 12.5

this analysisto work, the rate of pelagic sediment accumulationshouldbe steady,and the amountsof pelagicsediment eroded by subsequentturbidity currents should have
been negligibleor constant. The cumulative curvesshow
similar linearity and scatter, supportingpreviousqualitative assertionsthat the eventswere fairly regular in time,
and in particular that they did not clusterat the beginning
or end of the sequence.
If turbidites in many coreswerethe result of single,spa-

q- 0.5 intervals.

tial extensive events, there should be a correlation of devi-

whichtotal carbon(mostlyplant fibers)in the eighthtur-

ations betweenthe core plots, reflectingsystematicdiffer-

bidite from the surface was dated at 4645 q- 190 radiocar-

ences in the intervals between the events. Correlations

bon yrBP [Griggset al., 1969],or about 5400 q- 500 cali-

lation is not evident on the upper right diagram of Figure

are

not obvious,suggestingthe signal/noiseratio is low, i.e.,

variability of timing (signal)is not very large relative to
the imprecisionin using the measuredpelagicthicknesses
to estimate

MEAN

the intervals.

INTERVAL

VARIABILITY

BETWEEN

OF THE

THE

EVENTS

AND

MEAN

The above argument suggeststhat 13 large eventshave

shakenthe margin. In addition to providinga coprelatable
horizon, the Mazama eruption is well-dated by multiple

4C datesat 6845q- 50 radiocarbon

yrBP (Bacon[1983],
thoughBrownet al. [1989]suggestthat 6480q- 60 mightbe
a better estimate),and providesa reliabledate to compute
the mean recurrenceinterval. The Bacon radiocarbonage

is equivalent
to 7620+159
calibrated
yearsbefore1950(at
-51

The mean interval

between

turbidires

is

then 7360 q- 400 years/ 12.5 q- 0.5 intervalsor 590 q- 50

years(for comparison,the Broxvnet al. [1989]agefor the
Mazama eruptiongives570 q- 60 years). Not usedin this
analysisis the information that if the last turbidity current
was earthquake-triggered,it must have occuredmore than
150 years ago, for there has been no great earthquake on
the Cascadia

subduction

zone in historical

times.

An independent check comes from core 6509-27 from

brated years before 1989. The carbon had accumulated in

the canyon heads in the 600 years prior to the dated tur-

bidity current,soapplyinga correctionof-300 years(half

the inferredaccumulationtime) givesan intervalof about
680 q- 150 years and by extrapolation beyond the base of
the core, about 11 q- 2 turbidites since the Mazama. This

is in good agreement with sedimentologicalcorrelations

made by Griggset al. [1969]on the basisof the degree
of bioturbation, and the presenceof 13 post-Mazama turbidites in the channel upstream and downstream of the
core.

It is important to realize that the above error on the

mean

recurrence

interval

does not measure

the variabil-

ity of turbidity current timing. Such an estimate might

be made directly by examining the thickness variations
of the pelagic intervals. The thicknessesare not simple

the 2rr confidencelevel using version 2.0 of the computer

to determinebecause(1) the grainsizeand colorof the

programof Stuiver and Reimer [1986]). The difference

uppermost turbidite layer and of the pelagic sediment is

between the upper and lower errors is small relative to

the stochasticerror discussedbelow, so adjusting for the
39 years since the 1950 reference date gives 7660 q- 100
calibrated years before 1989.
There is an unknown delay between the eruption and
the deposition of the first turbidite containing Mazama

similar, and (2) bioturbationof the uppermostturbidite

layer and the pelagic sediment confusesthe lower boundary of the pelagicsediment. In addition, the presentthick-

nessmay not recordthe originalthickness

because(3) the
pelagic sedimentmay have been partially eroded during
depositionof the overlyingturbidite, and (4) bottom cur-

Adams: Paleoseismicityof the CascadiaSubductionZone

577

..'."/'"":"':""'"
I/////"'"'""::"'

E-

....'.'..'?'"":ss:'

.......,'.:'..::/.""'

""

....:?.::.'..".5:

..[O.::i....

,:..',%'"""'"'
.. "':.";"'"'":'
......'?',::'"
....

olo

ANNEL

2X

..:i?::'

...:..'.."-"-":"

/ /

....
:.'::'

..:
./,"/'"

'U.BD'rESEeU.

6/o

PELAGIC
ACCUMULATION
RATE

mm/EVENT
(MEAN
+_
'1
(7')

Fig. 5. Cumulative pelagic sedimentdepositionversusturbidite number for sevencoresfrom the

Oregon-Washingtonmargin. The plots are lined-up on the turbidite containingthe first Mazama

glass(star), and penultimateand top pelagiclayersare labelledby P and T respectively.If the

pelagicsedimenthas accumulatedsteadily,and if the turbidity currentsoccurredregularly,each
plot wouldbe a straightline. The data (dotsjoinedby heavylines)showsa goodapproximation
to the expectedrelationship(lightlines).Note that the verticalscalehasbeenadjustedby factors
of 2 for three coresso that all the curveshave a similar slope. The slopeof the fitted line represents
the mean pelagicsedimentaccumulationrate, and the deviationsof the data about the line gives
information on the variability of turbidity current timing, here expressedas a percentageof the
mean pelagic accumulationper event.

578

Adams: Paleoseismicityof the Cascadia Subduction Zone

rents may have eroded or winnowed the pelagic layer as

For a similar reason, events local to one canyon head

it accumulated[Harlett and Kulm, 1972]. All of these

(e.g., a magnitude6.5 earthquakewithin 50 km) do not

factors would tend to reduce the thicknessof the pelagic

layer, but are difficult to quantify unlessgood estimatesof
the pelagic accumulationrate can be applied. The rather
constant thickness of the pelagic layers suggeststhat the
losses,if any, are rather constant. In any event, the vari-

provide a sufficient explanation for synchronousturbidity

currents in separated tributaries and the same numbers
of events along the. margin. Therefore the cause must
be exogenic to the canyon heads and affect the OregonWashington margin as a whole. Three such external triggers with the required spatial extent are: tsunamis, waveinduced slumping during large storms, and great earthquakes on the Cascadia subduction zone.
Tsunami. The most recent damaging tsunami on
the Oregon-Washingtoncoast was generated by the 1964
Alaska earthquake, the secondlargest earthquakeof this
century. It did not trigger one of the 13 great turbidites in
the CascadiaChannel, becauseboth the amount of pelagic

ance from

the four

sources above must

be added

to the

variance due to irregular event occurrence, so that the

events must have been more regular than is implied by
the variability of the pelagic thicknesses.
For example, the numbers against the plot for each core
on Figure 5 representthe best-estimate mean pelagic accumulation per event and the standard deviation expressed
as a percentageof the mean. Both accumulationrates and
their variability are generally largest near shore,while the
deepercoresare the most regular, 4-26%for 6508-K1 and

sedimenton top of the cores(seebelow)and the amount

of bottom life on the channel floor during the 1965-1967

4-36%for 6509-15(not shownon Figure 5). If the cores

cruises[e.g., Griggset al., 1969]are too large for sucha

are consideredto represent the same 13 events, those with

the more variable pelagic thicknessesmust include a larger

recent turbidity current. Tsunami-generatingearthquakes

are relatively commonaround the Pacific, and if the large
tsunami in 1964 did not trigger a turbidity current, it
seems unlikely that a coupling of multiple independent
tsunami sourceswith the time-dependentstability changes
in the canyon heads would give both long recurrenceintervals and the samenumber of eventsalong the margin.
Wave-induced slumping. The case for wave-induced
slumping on such a large scaleis poorly documented,and
computations suggestthat for all but the steepestslopes
wave loading does not influence slope stability in water

amount

of variance

due to measurement

errors

and ero-

sion, as discussedabove, and so provide poor estimates

of the variability of event timing. Therefore, the average

variability for the three lowestcores(let = 28%) is chosen

as representative. For a 590-year recurrence interval, this

translatesto a variability(let) of about170years,and an

implied standard error on the mean of 50 years. Therefore, if the intervals are normally distributed there is one
chance in two that

one of the 12 intervals

would lie outside

it 4- 1.7rr, or outside the range 300 to 900 years.

depthsgreaterthan 120 m [Moranand Hurlbut, 1986],

DO THE

i.e., shallower than most of the canyon heads. Physical

parameters constrain the maximum size of storms and
storm-generatedwavesso that the rate of sizeincreasefalls

TURBIDITES

REPRESENT

GREAT

EARTH-

QUAKES, OR NOT?

Although G. Griggsoriginally believedthat earthquakes

were the causeof the turbidity currents(L. D. Kulm,

personalcommunication,1988), the lack of large earthquakesnearby made this position difficult to support in the

middle-to-late1960s.GriggsandKulm [1970]notedthat if
sediment is supplied by the Columbia River at a constant
rate, it would accumulate in the canyon heads until there
is enoughto trigger collapse,whereuponthe cycle would
start again. Such a self-triggeringsystem would tend to
repeat itself, and would generate similar-sized turbidites
in the channel system; to borrow a term from seismology, it would generate the "characteristic"turbidite in the
channel system.
Of course,triggering by external eventsof a cyclic na-

ture (suchas great earthquakes)wouldalsotend to displacesimilar massesof sedimentand producesimilar-sized

turbidites. Perhaps the strongest argument against the
self-triggeringhypothesisfor the Cascadiamargin is that
similar numbers of events are found at widely separated
places along the margin; it would be highly improbable that every canyon had the same temporal response
to the spattally-varying rates of sedimentationthat occur
along the margin. It is simpler to concludethat, while
self-triggering might occur in the absenceof an external
trigger, the triggers occur more frequently than the time
needed to reach the critical mass in most canyons,and so
usually short-circuit the endogenicslumpingprocess.

off rapidly with decreasingprobability(e.g., the 1000-year

wave may be only 30% larger than the 100-yearwave,and
not ten times the size). Thereforethe 100- and 1000-year
storms would have rather similar effects, it is likely that
the sedimentswouldneedto be closeto failure anyway,and
this would happen at different times for different places.
Like tsunamis, each large storm would trigger some turbidity currents when individual canyonheads were ready,
and so the processwould be unlikely to give both long recurrence

intervals

and the same number

of events for the

whole margin.
Earthquakes. Earthquakes are an unusual natural phenomenon in that even for quite rare events the expected
ground shaking increasesdramatically as the probability
level drops. Thus a great earthquake is so overwhelmingly large that it would trigger both marginally stableand
"stable" canyon head deposits. Great earthquakes,should
they occur on the Cascadia subduction zone, would probably have a long return period, and indeedreturn periodsof
several hundred to a thousand years have been estimated

by Adams [1984]and Heaton and Hartzell [1986],based

on the rate of geodeticstrain accumulation. As a first approximation, plate-boundary earthquakesoccur fairly reg-

ularly (a consequence
of the steadybuild-upof strain due
to plate movement and the constant physical parameters

of the fault zone), and so tend to have a "characteristic"

size. Very short intervals are unlikely becauseinsufficient
strain is available to be released,and very long intervals

Adams: Paleoseismicityof the Cascadia Subduction Zone

579

are unlikely becausethe strength of the fault zone limits

the amount

of strain

that

can be stored.

For such char-

acteristic earthquakes there may be a scale-independent

variabilityin timingamountingto aboutq--of the mean

recurrenceinterval[Nishenkoand Buland,1987],120years
for a 590-year recurrenceinterval. This is closeto the 2030% variability in pelagic thicknessesfound in the cores

(whichincludesnot only variabilitydue to the earthquake

---

....

II

o/ ---''I I/

cycle but also that due to variable partial erosionand measurementerrors in determiningthe thicknessof the pelagic

400

layers).
Evidence that the 13 turbidity currents indeed represent earthquakes is circumstantial and proof is unlikely for
several years. Nevertheless,becausea great thrust earthquake is an extremely large and relatively rare event, it
should cause synchronouseffects throughout the coastal
Pacific Northwest. Such confirmatory effects would include: sudden coastal subsidenceor uplift, landslides,sediment liquefaction and sand volcanoes,sedimentslumping
in large lakes, and submarine debris flows. Other less direct evidence would be: abnormal sedimentation events,
deformedtree growth, abandonmentof Indian settlements,
and secondaryfaulting on crustal faults. It should also be
noted that while the turbiditc record may well be complete becauseof the muddy nature of the sediments,the inferred earthquake recordmight be incompletebecausesub-

sequentnear-byearthquakes(a secondmainshockor large

aftershock)soonafter a great earthquakewouldprobably
not generate their own turbidity current; all of the unstable sediment having already slumped.
The first onshoreevidenceinferredto indicate past great

44

44

I I 6 o

4 o

I6 o

t4 o

Fig. 6. Spatial distribution of derivedvaluesfor (a) Recurrence

intervals

for turbidites

and onshore

subsidence

events,and (b) Age of the last turbidite, submarineslumping, or onshoresubsidenceevent. Data taken from sources
discussed

in the text.

While I do not claim that there is yet any completely

convincing proof that the 13 turbidity currents do represent 13 great thrust earthquakes on the Cascadia subduction zone, I believe that it is the best interpretation
available at present.

Cascadiaearthquakeswas reported by Atwater [1987a]

who ascribed

buried

Holocene

marshes in coastal Wash-

ington to sudden coseismicsubsidencesof the coast. His

IMPLICATIONS

FOR

GREAT

QUAKES ON THE CASCADIA

TIIRUST

SUBDUCTION

EARTH-

ZONE

mostrecentwork [Atwater,1988b]suggests
five subsidence
events and one shaking event in the last 3100 years. Correspondingevidencecomesalso from eight buried soils in
about 5000 years from another coastal Washington site

buried marshesin about 5000 yearsfrom southernOregon

The occurrence of 13 great thrust earthquakes, their

mean recurrence interval of 590 years, a variability of 170
years, and the assumed580 km extent of triggered turbidity currents and henceinferred strong ground motion that
are deduced above place some constraints on the slip, the
rupture dimensions,magnitude, and style of rupture of the

[Nelsonet al., 1989].At the southernend of the Cascadia

subductionzone,Carverand Burke [1987]deducedrecur-

subductionzone. Rogers[1988]has divided the Cascadia

subduction zone into plausible segments. The Juan de

rence intervals of about 600 years for thrust faults in the

accretionary prism just north of Cape Mendocino. In each
place the implied recurrence interval is about 500 - 600
years, in good agreementwith the mean recurrenceinterval and spatial extent establishedfor the turbidity currents

Fuca plate extends 900 km from the Nootka Fault Zone

off Vancouver Island to a boundary with the South Gorda
plate off northern California. An alternative segmentation
stops the Juan de Fuca plate at the Blanco Fracture zone.

(Figure 6a).
B. Atwater [personalcommunication,1988] has also

the mean recurrenceinterval of 590 yearsgives26 m for the

average slip per earthquake, provided no slip is aseismic.
A 26 m slip for the 750 km long Juan de Fuca plate north

[Hull, 1987], sevenburied marshesin about 3550 years

from northern Oregon[Petersonet al., 1988], and eight

drawn my attention to the coincidencebetweenthe age of

4290 4- 80 radiocarbon yrBP for the 8th peat at Willapa

Subductionof the Juan de Fucaplate at 45 mm/yr and

of the Blanco Fracture

Zone and a fault width

of 100 km

Bay (sampleH1-B of Atwater [1988b])and the ageof 4645

[Rogers,1988]givesa maximummomentmagnitude(Mw)

4- 190 radiocarbon yrBP from the 8th turbidite from core

of 9.1, which is in accordwith the maximum size proposed

6509-27 (discussedpreviously). Applying the correction

by Rogers [1988]. A variability of 28% in event timing

would imply slip displacementsof between 19 and 33 m,
but a variability in magnitude of only 4-0.1 magnitude

of-300 years to the core date brings the two ages into
remarkableagreement.This agreementis all the more significant becauseeach date is on the eighth event - thus
suggestinga one-to-one correspondencebetween the turbidite and subsidenceevents,and reinforcingthe argument
made from the mean recurrence

intervals

above.

unit.

Would a rupture of the Juan de Fuca subduction zone

stop at the Blanco Fracture Zone? Key evidence comes
from core 6604-12, which lies on the extension of the frac-

580

Adams: Paleoseismicityof the Cascadia Subduction Zone

ture zone. This core - as does core 6609-1 to the south, on

the subducting Gorda plate - contains 13 post-Mazama

turbidires, here presumedto representgreat earthquakes

on the Juan de Fuca subduction zone. If the Gorda plate

had a historyof independentsubduction(e.g., in Mw 8.3

earthquakes;Rogers[1988])core6604-12is closeenough
that additional turbidity currentsmight have been generated by earthquakes to both north and south. Therefore
the presenceof only 13 events in these two coressuggests
that either every Juan de Fuca rupture extends past the

BlancoFractureZone(i.e., a Mw 9.2 earthquake),or the

(someshownon Figure5) to computethe time sincethe

last turbidire. From six cores(Table 2) an age of about
300 4- 60 years(before1989) is estimated.This is consistant with the lack of a great earthquake in the 150-year
historical record, but may be a biased underestimate if
some sediment washed out from the top of each core. It
is noteworthy that both the Astoria and the Blanco cores

givesimilar(to within the poorresolution)agesto the four

Cascadiacores(Figure6b), therebyshowingthat theydid
not occur at greatly differenttimes (though this is not
proof that they occurredat the sametime), and perhaps

earthquakeson the Gorda segmentare synchronizedwith

thoseon the Juan de Fuca (i.e., zipper effectin which an
earthquake on one segmenttriggers earthquakeson adja-

suggestinga regional correlationof the last event.

cent segmentsin succession

a few hoursto yearslater).

marine canyons since the sediment last slumped away.

In Quinault Canyon piston core 68-18 only 200 mm of
hemipelagicsediment has been depositedsince the last

In the latter

case insufficient

sediment

would

have accu-

mulated between the paired earthquakes for a turbidity

current to have been generatedby each earthquake.
WHEN

WAS THE

LAST

EVENT?

The age of the last event can be estimated roughly

from the thicknessof the topmost pelagic sediment in the
coresrelative to the mean accumulation rate. As noted by

Griggsand Kulm [1970,p. 1367] "A lack of recognizable

hemipelagicsediment at the surfaceof these coresand the
absenceof extensive burrowing in the surficial layer compared to the extensive reworking in lower layers indicate
that the last flow was recent", although the fact that all
their cores have thinner last turbidires than penultimate
turbidiressuggeststhat the very soft pelagicsedimentand
the top, bioturbated part of the most recent turbidire were
often washed away. In addition, their use of "recent" is in
the context of Holocene time, and might be taken to include the last 1000 years.
The thickness of the topmost pelagic sediment, from
the available core logs which include "Phleger" triggerweight corestaken with the piston cores,has been used
together with the long-term pelagic accumulation rates

A date for the last turbidite could also be derived from

the amount of sediment that has accumulated in the sub-

flushingof the canyon[Barnard, 1978, p. 111], representing400 years(usingBarnard'srates) or 100 yearsof

sedimentaccumulation(usingrates of Thorbjarnarsonet
al. [1986]). Cores53-19 and 63-08 contain200 mm and
360 mm of post-flushingsediment[Barnard, 1978],representing200-400yearsaccumulation(usingratesof Thorbjamarson et al. [1986]). While improvingthe precisionof
these estimates clearly requires a critical assessmentof the
sedimentationrates at the coresitesin question,the above
rates generally support the inference that m300 years or
so have elapsed since the last event.
Tree-ring dates on drowned trees at six sites in coastal
Washington suggest the last synchronous,rapid submer-

genceevent was about 300 years ago [Yamaguchiet al.,

1989]. Comparabledates on the youngestburied marsh
soil suggestthe last subsidenceevent occurred about 300

yearsagoin southwestWashington[Atwateret al., 1987],

lessthan 400 4- 200 yearsagoin northernOregon[Peterson et al, 1988],lessthan 380 4- 60 yearsagoin southern
Oregon [Nelsonet al., 1989], and 270 4- 30 yearsago in
northern California (G. Carver, personalcommunication,
1989);datesthat are consistentwith the preceedinganalysis(Figure6b).

TABLE 2. Age of the Last Turbidite From Thickness

of the Topmost Pelagic Sediment

Core

PelagicThickness
a Accumulation
Rateb
(mm)
(mm/event)

Ageof Last
Turbidite(years)
c

6705-2

23 4- 6

62

220 4- 60

6705-6

25 4- 3

55

270 4- 30

6509-26

17 4- 5

41

245 4- 70

6509-27

16 4- 3

28

335 4- 60

6502-PC1

70 4- 20

148

280 4- 80

6604-12

47 4- 10

105

meanage

260 4- 60

2704- 60a

aThickness(4- possibleerror) in 1965/1967.

bFromTable 1 and Figure5 andsimilarplots.

CRelativeto 1965/1967; add 25 yearsto get yearsbefore 1989.

dOr about300 yearsbefore1989. May be biasedtowardbeingtoo

young becauseof washoutsat top of core;see text.

Adams: Paleoseismicity of the Cascadia Subduction Zone

581

TABLE 3. Age of the Last Five Events in Core 6508-K1

Pelagic Turbidite
(lrtm)
(mm)
?

At a
300+60

Date

160+
224.3

Ageb
(years)
3004-60

A.D.

16904-60

3904.70
240

284-3

6904-130

A.D.

13004-130

5004-70

570
554-12

11904-200

A.D.

8004-200

21704-450

1804-450

B.C.

27804-520

7904-520

B.C.

9804-250

120
344-3

6104-70

65

aTime interval representedat 334-1 mm/590-year event

assuming no loss of sediment.

bCalendaryearsbefore1989.
See Table

2.

Working from the pelagic thicknessesin core 6508-K1

(chosenbecauseits pelagicthicknesses
are the most regular), assumingthat no sedimenthas beenerodedby the
overlyingturbidite, and adopting 300 yearsfor the age of
the last event, the followingdates for the last five earth-

PROBABILITY
HAZARD

OF THE

NEXT

EVENT

AND

SEISMIC

IMPLICATIONS

quakesare deduced(Table 3): 1690AD, 1300AD, 800

From the mean interval between events (590 years),

the standard deviation of the mean (170 years), the
time since the last event (300 years), and a normal-

AD, 180 BC, and 790 BC. Although the slow accumulation rates of the pelagic sediment and the disturbance

distribution model for simple recurrent ruptures without

clustering, it is possible to estimate the likelihood of the

by the corer mean these dates are nos very precise(and

next great Cascadiaearthquake(Figure 7). At present

there is about a 5% chance that the next earthquake
should have already happened. For the future, the conditional probabilities are crudely 0.1% in the next year,
5% in the next 50 years, 10% in the next 100 years, and
25% in the next 200 years.
These valuesare probably good to only a factor of 2 and

in particular, the errors compound for the older dates as

shownin Table 3), they shouldproveof interestto other

researchersseeking to match their onshoreevents to the
turbidite

record.

The analysis presented here underlines the need to collect new cores from

the Cascadia

Channnel

and the re-

mainder of the margin. A profile of closely-spacedcores

across the channel, somewhere around core 6508-K1 on
Figure I would reveal much about the history of the tur-

100-

2-10
%

bidity currents. Griggs and Kulm [1970]obtainedseveral cores from the levees of Cascadia

Channel

banks that

contained only 8 post-Mazama turbidites; presumably the

largest ones. At some point between the levees and the
channel floor, there should be a complete section of the
13 turbidites, from which each subsequentturbidity current has eroded little or none of the pelagicsediment. Such
coreswould be invaluable for establishingthe relative sizes
and timing of the 13 turbidity currents. It might also
test whether

the amount

of bioturbation

is an indicator

5O%

60-

40-

590170
yr

a"l

/-/..
:1o"

: 2o-

of

elapsed time or relates to the size of the previous turbidity

current. In addition, radiocarbon dating of detrital carbon or even of individual pelagic foraminifera in the cores
would place independent ageson the events and avoid the
compounding errors intrinsic in an analysis like that in
Table 3. In the upper channel and on the slump scars
of the canyon heads, box cores, or other non-disturbing
sampling methods, should be used to determine the thicknessof the topmost sediment in order to establish the time
since the last event.

80-

o
o

4o

so

6oo'

TIME SINCE PREVIOUS GREAT CASCADIA EVENT(yr)

Fig. 7. Graph showingthe cumulative normal probability

distribution for a mean of 590 years and a standard deviation of 170 years as inferred for great earthquakes on
the Cascadia subduction zone, an estimate of where the

presentlies relativeto the last event (thick bar on the abscissaat 300 4- 60 years), and the range of conditional
probabilities of the next event for the next 50, 100, and
200 years.

582

Adams: Paleoseismicity of the Cascadia Subduction Zone

need to be refined using a revised date for the last earth-

quake and perhapsa lognormaldistributionmodel [followingNishenkoand Buland, 1987];howeversomeof the

uncertainty arisesfrom the possibility that the eventsclus-

ter in time (saythreeeventsin onemilleniumfollowedby a

long quiescence)
and from the still-unknownmodeof failure of the subduction

zone. While the same number of tur-

bidites in the Cascadia Channel, Astoria Channel, and the

sites off Cape Blanco strongly suggestssynchronousturbidity currents,it is not yet possibleto rule out a zipper effect whereby smaller earthquakesrupture the plate boundary in a short-term sequence. Such multiple modes of
rupture -sometimes single great earthquakes,sometimes
sequencesof smaller earthquakes - would complicate the
palcoseismichistory and are probably beyond the resolving powerof the turbiditc record,thoughcouldbe resolved
through onshoreinvestigationsusing dendrochronologyor
other precisedating of earthquake effects.
Even if the rupture mode is complex, the net effect on
hazard estimates may not be great because circumstan-

tial evidencesuggests
that the (hypothetical)rupturesegments would need to 'stay in step' and the time since the
last event is already close to half the mean recurrence interval. In addition, the damage implications for any place
along the margin are not greatly different for the single

M9 or multiple M8 earthquakescenarios;though clearly

should one M8 earthquake happen somewhereon the subduction

zone the likelihood

of others vould be increased

becauseof the temporal clusteringimplied by the alternate

hypothesis.
CONCLUSIONS
Turbidites

in the tributaries

of the Cascadia

Channel

and at other places along the Oregon-Washingtonmargin provide circumstantial evidence for the occurrence of
13 great Cascadia subduction zone earthquakessince the
Mazama eruption. My analysissuggeststhat magnitude
9 earthquakesoccurredevery 590 yearson average.The
pelagicintervalsdepositedbetweenthe turbidites suggest
that the earthquakesoccurredfairly regularly, with a standard deviation of 170 years or lesson the recurrenceinterval, similar to the variability found for great earthquake
cycleselsewhere.
Rhythmic triggering of turbidity currents by great
earthquakes may be a much more common phenomena
than hitherto realised, and might be expected at continental margins such as Alaska, Japan, New Zealand, and
Chile where great thrust earthquakeswith a long return
period are combinedwith a moderate supply of sediment
to the edge of the shelf. If sampled correctly, the turbiditc
record can provide a quick estimate of the palcoseismicity of a margin and so provide evidenceindependent of
onshore palcoseismicitystudies.
The thicknessof the topmostpelagiclayer suggests
the
last earthquakewas 300 4- 60 yearsago, but this number
may be a biasedunderestimatedue to washout at the top
of the core during the collectingprocess.It is, however,
consistent with the youngest subsidence episode on the
southwest Washington coast. The pelagic accumulation

between the turbidites has the potential to determine the

timing of previous eventsby working backwards;however
in the absenceof absolute dates on the turbiditc layers in

the cores(like thoseavailablefrom the submergedpeats

onshore),the errorscompoundand suchdatesshouldbe
used with caution. The current state of great earthquake

expectationfor the PacificNorthwest(Figure7) demonstrates that the near-term hazard of a great earthquake is
appreciable.
Material capable of precisely dating the past earthquakes has already begun to be studied at onshoresites
where

coseismic

deformation

and tsunamis

can be dated

by radiocarbon or dendrochronology.With turbidites to

provide an independent record of shaking, the concordance
of diverse data may ultimately reveal the history of great
thrust earthquakes in the Pacific Northwest.
Acknowledgments.This work would not have been possible without the meticulouswork of G. B. Griggs, his contemporaries, and their supervisor La Verne Kulm. Their
work, which I. have reinterpreted in this paper, is proof
that good sciencehas ramificationswell beyond the aspirations of the individual scientist. I thank Mitch Lyle and
the Collegeof Oceanographyfor their cooperationand for
making the core logs available. I apologizeto my colleagueswho have heard me present this work at the 1984
Chapman Conference on Vertical Crustal Motion and at
subsequentmeetings,who have only now a written version
to examine critically, but who have neverthelesscarefully
cited and attributed my work. D. Piper assistedwith technical advice regarding the behaviour of turbidity currents
and B. Atwater pointed out the importance of the bias in
the uncalibratedradiocarbonages. I thank P. W. Basham,
M. J. Berry, D. J. W. Piper, and G. C. Rogersfor valuable critical commentson early versionsof this manuscript,
and B. Atwater

and L. D. Kulm

for constructive

reviews.

GeologicalSurvey of Canada contribution number 33288.

REFERENCES

Adams, J., Active deformation of the Pacific Northwest

continental margin, Tectonics,3, 449-472, 1984.
Adams, J., Deformation above the Juan de Fuca subduction zone: some enigmas with bearing on greatearthquake risk, Eos Trans. AGU, 66, 1071, 1985.
Ando, M., and E. I. Balazs, Geodetic evidencefor aseismic

subductionof the Juan de Fuca plate, J. Geophys.Res.,

8J, 3023-3027, 1979.

Atwater, B. F., Evidencefor great Holoceneearthquakes

alongthe outer coastof WashingtonState, Science,236,
942-944, 1987a.
Atwater, B. F., Subduction-earthquake telltales beneath

coastal lowlands, U.S. Geol. Surv. Open File Rep.,

87-673, 157-162, 1987b.

Atvater, B. F., Coastlineuplift in Oregonand Washington

and the nature

of Cascadia

subduction-zone

tectonics -

Comment, Geology,16, 952, 1988a.

Atwater, B. F., Geologic studies for seismiczonation of
the Puget Lowlands, U.S. Geol. Surv. OpenFile Rep.,
88-16, 120-133, 1988b.

Adams: Paleoseismicityof the Cascadia Subduction Zone

Atwater, B. F., A. G. IIull, and K. A. Bevis, Aperiodic Holocene recurrence of widespread, probably coseismic subsidence in southwestern Washington, Eos
Trans. A GU, 68, 1468, 1987.
Bacon, C. R., Eruptive history of Mount Mazama and
Crater Lake Caldera, Cascade Range, U.S.A., J. Volcanol. Geotherm. Res., 18, 57-115, 1983.
Barnard, W. D., The Washington continental slope: Quaternary tectonics and sedimentation, Mar. Geol., 27,
79-114, 1978.
Brown, T. A., D. E. Nelson, R. W. Mathewes, J. S. Vogel, and J R. Southon, Radiocarbon dating of pollen by
accelerator mass spectrometry, Quat. Res., N.Y., 32,
205-212, 1989
Carver, G. A., and R. M. Burke, Late Holocene paleoseismicity of the southern end of the Cascadia Subduction
Zone, Eos Trans. A GU, 68, 1240, 1987.
Duncan, J. R., Late Pleistoceneand Postglacial sedimentation and stratigraphy of deep-sea environments off
Oregon, Ph.D. thesis, Oregon State University, Corvallis, 222 pp., 1968.
Fryxell, R., Mazama and Glacier Peak volcanic ash layers:
Relative ages, Science, 17, 1288-1290, 1965.
Griggs, G. B., Cascadia Channel: the anatomy of a deepseachannel, Ph.D. thesis,Oregon State University, Corvallis, 183 pp, 1969.
Griggs, G. B., and L. D. Kulm, Sedimentation in Cascadia
deep-seachannel, Geol. Soc. Am. Bull. 81, 1361-1384,
1970.

Griggs, G. B., A. G. Carey, and L. D. Kulm, Deep-seasedimentation

and sediment-fauna

interaction

in Cascadia

Channel and on Cascadia Abyssal Plain, Deep-Sea Res.

16, 157-170, 1969.
Harlett, J. C., and L. D. Kulm, Some observationsof nearbottom currents in deep-seachannels. J. Geophys. Res.,
77, 499-504, 1972.
Heaton, T. H., and S. H. Hartzell, Source characteristicsof
hypothetical subduction earthquakes in the northwestern United States, Bull. Seismol. Soc. Am., 76, 675708, 1986.
Heaton, T. H., and S. H. Hartzell, Earthquake hazards on
the Cascadia subduction zone, Science, 236, 162-168,
1987.

Heaton, T. H., and H. Kanamori, Seismicpotential associated with subduction in the northwestern United States,
Bull. Seismol. Soc. Am., 7, 933-941, 1984.
Heezen, B.C., and M. Ewing, Turbidity currents and submarine slumps, and the 1929 Grand Banks Earthquake,
Am. J. Sci., 250, 849-873, 1952.
Hull, A. G., Buried lowland soilsfrom Willapa Bay, southwest Washington: Further evidence for recurrence of
large earthquakesduring the last 5000 years, Eos Trans.
A G U, 68, 1468, 1987.
Moran, K., and S. E. Hurlbut, An analysis of potential

583

Nelson, A. R., S. F. Personius, and S. Rhea, Earthquake

recurrence and Quaternary deformation in the Cascadia
subduction zone, coastalOregon, U.S. GeologicalSurvey
Open File Rep., 89-53, 481-484, 1989.
Nelson, H., Late Pleistocene and Holocene depositional
trends, processes,and history of Astoria Deep-sea Fan,
northeast Pacific, Mar. Geol., 20, 129-173, 1976.
Nishenko, S. P., and R. Buland, A generic recurrenceinterval distribution for earthquake forcasting, Bull. Seismol. Soc. Am., 77, 1382-1399, 1987.
Peterson C. D., M. E. Darienzo, and M. Parker, Coastal
neotectonicfield guide for Netarts Bay, Oregon, Oregon
Geol., 50, 99-106 and 117, 1988.
Pilkey, O. H., Basin Plains; Giant sedimentation events,
Spec. Pap. Geol. Soc. Am., 229, 93-99, 1988.
Reilinger, R., and J. Adams, Geodetic evidence for active
landward tilting of the Oregon and Washington coastal
ranges, Geophys. Res. Lett., 9, 401-403, 1982.
Riddihough, R. P., and R. D. Hyndman, Canada's active
westernmargin- the casefor subduction, Geosci. Can.,
3, 269-278, 1976.
Rogers, G. C., An assessmentof the megathrust potential
of the Cascadia subduction zone, Can, J. Earth Sci.,
25, 844-852, 1988.
Savage,J. C., and M. Lisowski,Geodetic deformationmeasured in Washington and British Columbia, U.S. Geol.
Surv. Open File Rep., 88-37, 230-238, 1988.
Sternberg, R. W., Transport and accumulation of riverderived sediment on the Washington Continental Shelf,
J. Geol. Soc. London, 13, 945-956, 1986.
Stuiver, M. and P. J. Reimer, A computer program for radiocarbon age calibration, Radiocarbon,28, 1022-1030,
1986.

Thorbjarnarson, K. W., C. A. Nittrouer, and D. J. DeMaster, Accumulation of modern sediment in Quinault

Submarine Canyon, Mar. Geol., 71, 107-124, 1986.
West, D. O., and D. R. McCrumb, Coastline uplift in
Oregon and ]ashingtonand the nature of Cascadia
subduction-zonetectonics, Geology,16, 169-172, 1988a.
West, D. O., and D. R. McCrumb, Coastline uplift in
Oregon and Washington and the nature of Cascadia
subduction-zone tectonics- Reply, Geology, 16, 952953, 1988b.

Yamaguchi, D. K., C. A. Woodhouse, and M. S. Reid,

Tree-ring evidence for synchronousrapid submergence
of the southwestern Washington coast 300 years B.P.,
Eos Trans. A GU, 70, 1332, 1989.

J. Adams, Geophysics Division, Geological'Survey of

Canada, Ottawa, K1A 0Y3, Canada.

slopeinstability dueto waveloadingon the NovaScotJan

(ReceivedSeptember30, 1988;

Shelf, Proc. Can. Conf. Mar. Geotech. Eng., 3rd., p.

980-999, 1986.

revised December 12, 1989;

acceptedDecember18, 1989.)