726‘

'Q-b‘
< ,...

 

«Mid

‘_ 3.; hs‘firlfiud >'
.
,-
O

      

 

14‘.
1“.\ ~V

mun-)4

 

 

 

v v v V . . x —‘
. ‘ '.'.'.l'.‘.""— n ‘. _ . - g. ' y _ ‘ . . . v V. v "- v'v‘rv| l ' .n . ,. .v v

. o ‘ ' ‘. J .

' > . ’ _. . ’

. .

- ' ' . , ~ ' . ' -

- . 3 .

. . ' . . ‘

. - .
v . ’ ' I
p. . _ . . ' V : '

.7 ll

r-I- *

‘ii 37:!

 

J «'r
t ’57-‘35"
». ‘. .vmnm.
213-3..

‘3:&

. .
593‘
j,‘ .53.: :& .W . , .‘4 "'3 “39122.

“era N. y“ . -. 7"

 

'4

how

~' - awn; '
”c."

.-.
. :3. 11;
N A‘M‘n

 

'_ 1-
.3.-l)f_ .. . 1..

.zLAv

'-.‘.l

M

o'dsfivwo

THEZNC

 

L.::.’. "’2 A APY

*—

 

5'15‘1‘7-4: 51:31;
University

 

 

This is to certify that the
thesis entitled
Teleseismic Mislocations of Earthquakes in
Island Arcs - Theoretical Results
presented by

Timothy Lynn Nieman

has been accepted towards fulfillment
of the requirements for

M.S. . Geological Sciences
degree 1n

 

 

ZZZ:

Maj professor

Kazuya Fujita
Date May 2, 1985

 

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

MSU

LIBRARIES
-—_

 

 

RETURNING MATERIALS:
Place in book drop to
remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped below.

 

 

 

 

 

TELESEISMIC MISLOCATION OF EARTHQUAKES
IN ISLAND ARCS ‘THEORETICAL RESULTS

BY

Timothy Lynn Nieman

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER or SCIENCE

Department of Geological Sciences

1 985

ABSTRACT

TELESEISMIC MISLOCATIONS 0F EARTHQUAKES
IN ISLAND ARCS -' THEORETICAL RESULTS

BY

Timothy Lynn Nieman

Three-dimensional seismic ray tracing through thermally derived slab
models is used to investigate the effects of subducting lithosphere on
teleseismic earthquake locations in island arcs. Theoretical results show
teleseismic mislocations are greatest in the thrust zone and become
negligible seaward of the trench. Varying the thermal coefficient and
depth of penetration of the model slab have pronounced effects on
mislocations while variations in assumed slab thickness have only minor
effects. Variations in station distributions used to locate island arc
events can result in 10 km difference in determined epicenters for the
same event.

Comparison with observed mislocations gives a best fit model for the
central Aleutian slab extending to 360 km depth with a thermal coefficient
of -.0009 km/sec-‘C. Theoretical mislocations indicate that intermediate
depth events are well located teleseismically. Spurious slab dips and

thinning of the Beniof f zone can result from mislocations of deeper events.

 

ACKNOWLEDGMENTS

i would like to thank my advisor and committee chairman, Kazuya
Fujita, for this thesis topic, for keeping my computer account loaded, and
for numerous forms of aid, encouragement, and criticism. Thanks also go to
Fate, for not presenting'the opportunity for Kaz to Shove raw fish down my
throat. I would like also to thank the other members of my committee,
Hugh F. Bennett and F. William Cambray, for their help and criticism.
Special thanks to Mom and Dad, whose encouragement, understanding, and
money helped make this possible, and to the Cooper's for much needed
breaks at their desert resort complex. I am grateful to my friends and
colleagues, Don, Dave, Cindy, Bruce, Soo Meen, Bill, and Greg, and, of course,
the entire cast of Geo inc in its various forms, for their friendship and
help in attitude adjustment when needed. Thanks to the faculty and staff
at Michigan State University for their support and cooperation.

This research was supported, in part, by NSF grant EAR 80-25267. The
author was supported by a Chevron Graduate Fellowship and the Department
of Geological Sciences at Michigan State University.

ll

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................ Iv
LIST OF FIGURES ............................................................................................................ V
INTRODUCTION ................................................................................................................ l
METHODOLOGY ................................................................................................................. 4
MISLOCATION RESULTS ............................................................................................. l0
Shallow Events ............................................................................................... 12
Effects of Station Geometry .................................................................... Io
Deeper Events .................................................................................................. 19
Comparisons With Observed Seismicity .............................................. 22
DISCUSSION .................................................................................................................... 41
CONCLUSIONS ................................................................................................................. 47
BIBLIOGRAPHY ............................................................................................................... 5 I

ill

LIST OF TABLES

TABLE I. Teleseismic and Local Network Hypocenters

of Actual Events .................................................................................... 24
TABLE 2. Actual Event Mislocation Data ......................................................... 28
TABLE 3. Theoretical Event Mislocation Data .............................................. 30

LIST OF FIGURES

Figure I. Model Slab Parameters ........................................................................... 6
Figure 2. Velocity Profile of 300 x 80, -.0009 Slab .................................. I 1
Figure 3. Travel-time Residuals ......................................................................... 13

Figure 4. Shallow Event Mislocation Vectors for 300 x 80 km,
-.0009 Slab, Small Station Set ....................................................... i4

Figure 5. Shallow Event Mislocation Vectors for 300 x 80 km,

-.0009 Slab, Large Station Set ........................................................ 15
Figure 6. Epicentral Mislocations for Different Station Sets ............... 17
Figure 7. Station Distributions ........................................................................... I8

Figure 8. Deep Event Mislocation Vectors for 300 x 80 km,

-.0009 Slab ....... L ...................................................................................... 20
Figure 9. Mislocation Vectors for Actual Events ........................................ 23
Figure IO. Velocity Profile of 300 x 80, -.OOl l Slab ............................... 33
Figure l I. Velocity Profile of 300 x 100, -.0009 Slab ............................ 35

Figure I2. Shallow Event Mislocation Vectors for 360 x 80,

-.0009 Slab ............................................................................................ 37
v

Figure 13. Shallow Event Mislocation Vectors for 540 x 80,
-.0009 Slab ............................................................................................ 38
Figure 14. Deep Event Mislocation Vectors for 360 x 80 km,

-.0009 Slab ............................................................................................ 4S

 

vi

INTRODUCTION

Analyses of locations of earthquakes and nuclear explosions in island
arc regions show that velocity inhomogeneities due to subducting
lithospheric slabs can induce location errors of 50 km or more (Herrin and
Taggert,1968; Utsu, I971; Engdahl,l977), for teleseismically located
island arc earthquakes, when standard symmetric earth location procedures
are used. Higher seismic velocities within the slab, as compared with the
adjacent mantle, can result in P-wave travel time advances in excess of 5
seconds (Jacob,l970; Sleep,l973) as well as shadow zones caused by
refraction through the slab (Sleep,l973). Clearly, understanding the
geometry and rheological properties of descending lithosphere is
intimately related to being able to accurately locate earthquakes occurring
there. in recent years, data from local seismic networks has greatly
improved our understanding of subducting slabs and their effects on ray
paths (Engdahl et al.,i977b; Hasegawa et al.,1978; Fujita et al., i981,-

Frohlich et al., i982).

 

2
Numerous studies have been performed to delineate slab structure
by looking at how the high-velocity subducting lithosphere affects travel
times (Sleep, 1973; Hasegawa et al., 1978; Fujita et al., I98i; Frohlich et
al.,1982; Spencer and Engdahl, I983; Mclaren and Frohlich,1985). Ideally,
the most precise approach involves the use of three-dimensional seismic
ray tracing through realistic, detailed velocity structures. However,
computational inefficiency in tracing rays from a source to a large number
of stations generally makes the side by side comparison of numerous
models impractical. Fujita et al. ( i98i) developed a method that achieves
a significant reduction in computation time required. Instead of trying to
home a ray into each individual station, they shoot a variable coverage of
the focal sphere and interpolate to find travel times to individual stations.
They also noted that unless the slab is torn or contorted, its effects on
locations are a smooth function of epicentral distance from the center of
arc curvature for any constant given focal depth. It is therefore only
necessary to determine the slab's effects at various evenly spaced points
and interpolate for effects at intermediate points. In this study, i use this
approach to investigate and clearly demonstrate the mislocation effects of
a fairly large number of models for the subducting slab in the central

Aleutian islands, Alaska.

3

I also compare modeled mislocations with mislocations of actual
earthquakes in an effort to constrain several physical properties of the
subducting slab. Estimates of errors in ISC (International Seismological
Centre) and PDE (Preliminary Determination of Epicenters) reported
teleseismic hypocentrai solutions can be made by comparison with
locations determined from only local network data. This study concerns
the area near Adak Island, where a local seismic network has been in
operation since late July, I974. Local epicentral solutions for shallow
thrust zone events near Adak are believed to be accurate to within a few

kilometers (Frohlich et al.,I982).

METHODOLOGY

The method of model generation, ray tracing, and relocation is basically
the same as that of Fujita et al. (I98i) with several minor changes.
Essentially the procedure is as follows. Thermal modeling was used to
determine velocity structure for various combinations of three parameters,-
slab thickness, depth of slab penetration, and thermal coefficient of
seismic velocity. Seismic ray tracing was performed through the models to
obtain P-wave travel time anomalies from hypothetical earthquakes to a
network of stations. Computed travel times were then used to relocate the
theoretical events using a standard symmetric earth location routine. The
original hypocenter was compared with the relocated hypocenter in order to
determine the mislocation vector for a given theoretical earthquake and
slab model. Mislocation vectors for actual earthquakes could then be
compared with model mislocation vectors to evaluate the feasibility of the
specific model.

The finite difference modeling technique of Toksoz et al. (I97I,I973)
was used to produce temperature profiles for each model. Using the Herrin
et al. (I968) velocities as the base model, velocity profiles were

4

S

constructed from the thermal profiles using a linear relation between
velocity and temperature anomaly. Laboratory experiments give values for
this thermal coefficient (av/6T in eq. l: Sleep, 1973) around -5 x 10‘4
km/sec-‘C (Anderson et al.,1968), while many authors have reasoned that
realistic earth values are greater in magnitude (Sleep,l973,Engdahl et
al.,I977b; Fujita et al., l981; also Jacob, 1972; Hasegawa et al., I978 based
on velocity contrast considerations). I allowed the thermal coefficient to
vary from -S x ID"4 to an assumed maximum of -9 x 10'4 km/sec-‘C in an
attempt to constrain its value.

Three dimensional velocity representations were constructed by
rotating the two dimensional model about the center of curvature of the
island arc located at 63.3i4'N and I78.0S9'W (Engdahl, i977; Fujita et
al.,I981). Positions on a cross section of the subduction zone can then be
described simply in terms of depth and distance (a) from the pole. This
places the volcanic are at I I.4' a (distance from center of arc curvature)
and the trench axis at 13.0' a for the Adak region. A diagram of the model
slab is shown in Figure I.

I followed Fujita et al.‘s (l98I) scheme for ray tracing, plotting, and

digitizing exactly except that I digitized every 5' of azimuth instead of

 

.ousuo>u:o cum mo umucou oz» Eouu mooummc :fl mocoumflo Owuucoooom ecu mg n
.xaaoc0wmcoeflciooucu ocucomoumou ma gnaw #0608 30: mcfi3osm Emummflo
.muquEmumm nodm Hone: .H ousmflm

 

 

3.5.2.:
as.»
a. 5:33.85
cf :2-
.0533.
g
\ .. ..
. .
Q — .
Y...
\ X a. Q3008.Ob..202n.nw.
0. “i bu \\ 02°>50
e \ 9:. .o
\\ . o 8.50
es s N
‘. “‘ ‘
........ .. ix .g o. a o
killliicnlll. N. 93 I! Z

 

a. 5...: 252.5

7
every 10', and I used l3 Fourier coefficients instead of 9. These
improvements in accuracy were needed since I examined models having
more pronounced azimuthal variations of residuals.

Relocations of the hypothetical earthquakes were performed using two
different station sets, each consisting of stations between 30' and 90'
geocentric distance, with the addition of local station ADK for depth and
origin time control. Travel times were determined by applying interpolated
residuals to a modified Herrin travel time table. In an effort to obtain
more realistic travel times to ADK, Herrin travel times for distances of
less than 5' were adjusted to reflect the local ADK shallow velocity
structure (Toth and Kisslinger, l984), assuming that this structure is
representative of both source and receiver velocities. Standard Herrin
travel times were used for teleseismic distances.

The existence of a strong azimuthal bias in travel time residuals
necessitated the consideration of station distributions used to locate
different earthquakes. For the central Aleutians, smaller magnitude events
are generally recorded best in western North America and moderately well
in northern Europe while larger events are very well recorded in Europe and
southern Asia as well as North America. Relocations of hypothetical

events indicate epicentral differences of up to lo km can occur for the

8

same event when located using twodifferent station sets, one set typical
of smaller central Aleutian events (“4.7 Mb) and one typical of larger
' events ("5.4 Mb). Ideally, for each actual earthquake used to compare with
model earthquakes, I would like to have performed theoretical event
relocations using the same station set reporting the actual event, but time
and cost limitations made this impractical. Therefore, station geometries
for approximately 50 thrust zone events were examined to determine an
“average“ station set. It was noted that the distribution of stations
reporting arrivals is more dependent on the number of stations reporting
than on listed body wave magnitudes, with a general change in geometric
distribution occurring at about 175 reporting stations. I thus compiled
two station sets using the most commonly reporting stations for the
events examined. One station set consisting of 220 stations was used to
represent an ”average“ station distribution for larger earthquakes (events
reported by more than 175 total stations) while the other, consisting of
i 10 stations, was used to represent smaller events (reported by 50 to 175
total stations).

Earthquakes to be used as the data base with which to compare model
mislocations had to be well recorded both teleseismically and locally. For

the time span of August, 1974 through February, 1979 and January, 1982

9
through October, 1983, I compiled an initial list of 40 shallow (<50 km
depth) events between 175'w and 179'w recorded by the local network
(Engdahl et al., I977a; Engdahl et al., 1982; Kisslinger et al., 1982;
Kisslinger et al., 1984) and by at least 50 total stations In the ISC or PDE
bulletins. In order to be consistent with the hypothetical event location
method, the actual events were relocated using Herrin travel times and
only stations at distances of 30' to 90' ,plus ADK Stations with spurious
travel times (residual > 3 sec) were discarded. Relocations were done both
with and without station corrections (Dziewonski and Anderson, 1983)
with an average location difference of 2.7 km in epicenter and 2.9 km in
depth between the two methods for the 40 events used in this study.
Generally, smaller events (50 to 100 stations) showed the greatest
location changes when station corrections were used, with a maximum
epicentral change of 10.4 km in the Feb 18,1976 event. The average RMS
residual for the locations was reduced from .823 to .778 seconds when
corrections were used. The hypocenters determined with station

corrections were used as the data set.

MI SLOCATION RESULTS

Most previous studies employing thermal modeling and ray tracing for
the central Aleutians have assumed slabs which penetrate no deeper than
300 km. Sleep (1973) used a slab extending to 180 km depth with av/aT =
-9 x 10"4 kmlsec-‘C for the region near Amchitka island, approximately
200 km west of Adak Island. One would expect the slab to penetrate to a
shallower depth at Amchitka than at Adak since subduction becomes more
oblique as one moves west along the Aleutian arc. A number of recent
studies of the Adak region have used a slab which is 80 km thick,
penetrates to a depth of 300 km, and also with a thermal coefficient of -9
x 10‘4 (Fujita et al., 1981; Engdahl et al., 1982; Rogers, 1982). Figure 2
shows the velocity profile for this model, which gives an average velocity
contrast of "6% and a maximum velocity contrast of “10% compared with
the ambient mantle. I chose this as the first model for study. Assuming
this model is essentially correct, I performed theoretical relocations to

determine the mislocations that should be observed for events in this area.

IO

ll

DISTANCE FROM POLE (deg)

 

 

 

 

 

 

 

 

 

 

 

IO II £7 l2 l3 l4
ooo;,gtf?rooooo o o Q73 0 o J
I00 -d
825‘—————'
200 '—
350‘_—_“T
300 -‘
875
900
4OOF- '4
925-——-——
9.50 km/sec—
SOC

 

 

 

Figure 2. Velocity Profile of 300 x 80, -.0009 Slab.
Triangle at a distance 11.4° at the surface indicates
location of volcanic front. Solid dots indicate locations
of theoretical events for this model. Slab is outlined

by dotted line. (from Fujita et al., 1981)

Shallow Events

Ray tracing and relocations were computed for theoretical earthquakes
located every 0.1' from 11.8' to 13.0’ a and every 0.2' from 13.2' to 14.6'
a at a depth of 25 km. The left sides of the plots in Figure 3 show two of
the contoured residual maps from ray tracing for events located at 12.0’ a
and 12.4' a. For these events, the greatest travel time advances occur at
an azimuth of ”60' and “55', respectively. In general, rays traveling
laterally down dip, as opposed to directly down dip, travel the greatest
distance within the slab and thus result in the greatest residuals. As the
geometry of the situation would predict, the region of greatest residuals
rotates toward 0' azimuth with increasing a. The right sides of the plots
show emergence points for equal takeoff azimuths (solid lines) and equal
takeoff angles (dotted lines or solid dots).

Figures 4 and 5 show plots of mislocation vectors for shallow events in
this model done with the two station sets, open circles indicating original
locations and solid circles showing hypocenters upon relocation. Figures 4
and 5 verify the assumption that mislocations are a smooth function of a
(Fujita et al., 1981). Therefore, in order to save time and expense, all

subsequent models were done with a wider spacing of theoretical events

13

Aamma ..~m um mafiflsm EOHM. .ooooo: cums: czozm one monco can mnusefino wmoioxou
HoucOEOHQmsm .mam>umucfi com uo cog um mwcfia UAHOm an :3ocm who mausefluo
mwoioxmu Hmfiuficd “moon CAHOm we go m~o>HOucfi av um mocfia couuoc ha c3ozm

mum mooumoc CH awamcm mmoimxmu Hoduwcfi “mam: unmwu mcu co coauOAQ who mucfloa
mocomuoso new muoumEouoa any HmHuACM couoasuaou .Auzmfiu. ov.-uc 0cm Aumev
oc.mdum um muoucoooa>z nummc HMOOW Ex mm usono cofiuoofloum Honussfiuo ucoumflc
1wsvm co mo mam: uuoa on» :0 coquAQ can mam>uoucfi ccooom mm.o um nou30ucoo
cum Uoioom\Ex mooo.1 mo acofiofiuwooo HoEHozu m LUHS gnaw xowcu Ex om >3 moon
Ex com o nmsounu mcfiomuu any OwEmwmm >9 couoasoHoo maozcflmou oEHUIHo>cuB
.mamscfimom osfiuiao>mue .m ousmam

onnw co. co.

 

 

 

14

maoacflmou mzm
.nocouu can
.cofluoooHou com:

.noam mo ooowusm momma mOCOmouQmu mafia Gunman

.musmflm «0 when EOuuon cw czozm who mcofluoooHou HOm
.>Ho>quommmou
accum OacmoHo> ucomoumou cameras» ammo can camcofiuu caaom
muoucmoomxn ucomouamu moon cflaom .muco>o Hmuauonuomxn wo

 

 

 

 

 

mCOMUMOOH answmwuo acommummu mmaouwo Como .uom :Ofluoum HHoEm .noam
mooo.1 .Ex cm x com now moouoo> :oHumoonHz oco>m soHHmnm .v oosmsm

- q - q a u a . q q 4 q q q a q u 4‘. u 4 q - LOO
Goa
. m
.3)
. a
l. Nomi

l .
"lillulu"nulullllfllluxlTR
\
.3 32.2- .3.5... . \
8.. .5. 8.52 \

\ G
- \ 183
E \\ Id.
®\m.\\ \ KAN H
o a / A/zffil. .. E...

p p — b . . L _ . lullp.1i.il—\ - . . . _ C . k . _ G

on. a». 3. 3. a:

:53 moon. :2: mozfima

lS

mooo.i

.v musmfim 2% mm osmm maonaxm .uom :ofiumum,mmumq .Qoam

 

 

 

 

 

 

.Ex cm X COM HOW mHOu00> COHHMOOHmHZ HC®>H 3OHHM£m .m QHDOHM
q 4 d d d d d d q q d d d 1 d 114 q d u 1 u q d no
. _ HG
. . o . 09W
1111. a
p . — p p L p — . . n . _ . p . . _ . p p . _ ON
a a d a a . < q q q a u q q a q 4 q 1 q q \4 a d
.3 3:2. .93.. \
92o .5. oo - 8m \x G
\\ l 83
am I; M
H
3 A” Q 1. mm M.
. . L F . p . — _ O
I . .
on. on. a. N. o N. o :

so... Boa :9: 52455

16
with interpolation used between events. For this model (and, as will be
shown, for all models in this study), mislocations are greatest in the
thrust zone near 122' a. Also seen is that epicentral mislocations become
negligible, given the limits of location procedures, seaward of the trench.
The mislocation vectors suggest that events may actually be located too

shallowly near to and seaward of the trench.

Effects of Station Geometry

Figure 6 shows only epicentral mislocations as a function of a for
relocations done with the two different station sets. Mislocations are
greater for locations done with the large station set on the volcanic arc
side of the thrust zone. Mislocations are about equal at 12.1’ a; seaward of
this point, mislocations are greater when the smaller station set is used.
One reason for this pattern can be seen by considering the station sets
(Figure 7) and the residual patterns (Figure 3). The major difference in
coverage between the two station sets occurs in Europe and southern Asia
between about 76' and 90' geocentric distance from 320' and 360'
azimuth, with the large station set showing a significantly better coverage
of this area. Travel time residuals are up to one second greater in this

area for a a of 12.0“ than for 12.4', hence the greater mislocations. This

l7

Mislocation (hull

 

 

 

oIrrrIIrIrrrFIIfr'rrirrrIrT
l LS IZO IL! l3.0 l3.5 “.0

Distance from pole (dog)

Figure 6. Epicentral Mislocations for Different Station
Sets. Comparison of epicentral mislocations using

station sets representative of "small" and "large" events
near Adak for 300 x 80 km model with a thermal coefficient

of -.0009 km/sec-°C.

18

 

.Azommo.mma .zovfia.~m. musum>uno ouc mo Hmucoo
onu EOHu o~.NH macs umauzu us» :« couoooH ma muoHQ mo umucoo .mu:m>o
Hmofiuosuomxc mumooHou Cu com: mcofluoum mo muOHQ HmSuDEHNo ucoumflcflsvm

.mcofiusnfluumflo :owumum .5 ousmflm

uom mound uom HHoEm
00. co.

 

19
general pattern (Figure 6) holds for all models used in this study; as the
length of the model slab increases, though, the point where the
mislocations are equal for both station sets shifts slightly towards

smaller as.

Deeper Events

For this model (and for one other model which will be discussed later) I
also performed ray tracing and relocations for a series of theoretical
events located at depth along the upper surface of the descending slab. The
degree of mislocation for deeper events proved to be relatively insensitive
to which station set was used. The probable reason for this is that rays
traveling from deep events to European stations, where the major
difference in the station set coverages exists, are largely unaffected by
the slab. All deep event teleseismic relocations shown in this study are
done with the large station set. The results are basically the same if the
relocations are done with the small station set.

Figure 8 shows the mislocation vectors, which predict that teleseismic
mislocations for events deeper than about 100 km are small (always less
than 10 km) supporting previous findings (Barazangi and Isacks, 1979;

Engdahl et al., 1982) which suggest that teleseismic locations of deep

20

DISTANCE FROM POLE (deg)
II L

r If T

 

 

 

I00

(km)

DEPTH

200

 

SOC

 

 

93
a)

 

p
(n

RMS(sec)
qrI l l at?! $T
I
i

Figure 8. Deep Event Mislocation Vectors for 300 x 80 km,
-.0009 Slab. Symbols same as in Figure 4. Relocations are
done with large station set. Original hypothetical event
locations (open circles) indicate the top surface of
subducted slab.

' 21
events in subduction zones are probably more accurate than local network
solutions even though they often show greater RMS residuals than the local
solutions (Mclaren and Frohlich, 1985).

Also, note in Figure 8 the appearance of a spurious increase in slab dip
at about 100 km depth resulting from the errant locations. This suggests
that reported increases in the dip of actual Beniof f zones at depth based on
teleseismically determined locations may at least be partially the result
of earthquake location errors. The mislocation vectors also show a
subsequent decrease in dip at a depth of about 200 km, but this would
probably not be seen since very little seismicity is recorded at this depth
in the central Aleutians. It should be noted that a similar increase in dip at
about 100 km depth has also been shown to appear as a result of slab
induced location errors in local network solutions (Spencer and Engdahl,
1983; McLaren and Frohlich, 1985). Furthermore, the spurious increase in
slab dip demonstrated for local network solutions is significantly greater

than the effect determined here for teleseismic solutions.

22

Comparisons With Observed Seismicity

At this point I wish to determine if the model i have been using is a
good match to the actual velocity structure in our study area. Before
discussing the degree of fit to the model, evaluation of the actual
mislocation data must be made to ensure that they are consistent. A major
assumption in the modeling procedure is slab symmetry about a pole of
rotation. If this assumption is a good approximation for the actual slab,
then earthquakes along strike of the arc having the same a value should
show approximately the same degree of mislocation. Figure 9 shows
mislocation vectors for the 40 events used in this study while Table 1 lists
local and teleseismic locations for the events. As can be seen from Figure
9, mislocations for events between 175' and 176.2'W longitude (shown by
an asterisk in Table I) show significant variations for similar values of a.
Topper (1978) suggested that a bend or tear exists in the slab at about
175.5‘W with the slab to the east having a steeper and more northerly dip.
The possible existence of considerable complexity in the slab in this area
makes comparison with models assuming symmetry tenuous. Events
between 176.2'w and 179‘w (hereafter referred to as the ”non-eastern"
events) do, however, show a good deal of consistency with a few

exceptions. Event ’ 19 (Table 1) did not converge upon relocation and was

23

* O 1» 4- " 53N

100 km

9

Figure 9. Mislocation Vectors for Actual Events.
Epicentral mislocation vectors for the 40 events used in
this study. Heads of arrows represent teleseismically
determined locations, tails represent epicenters from the
Adak local network. Bathymetric contours for the trench

are given in meters. Adak and nearby islands are also
shown.

24

Table 1. Teleseismic and Local Network Hypocenters Of Actual Events.

For each event, the first line gives the local network solution, the second
line gives the teleseismic solution done with Herrin et al. (1968) travel
times and Dziewonski and Anderson (1983) station corrections. *‘s
indicate eastern events; "5 indicate small non-eastern events, f's indicate
depths fixed due to convergence above the surface.

_Daia anMWMmim

 

I. 74AU013 034619.78 51.147 177.880
21.4 51.486 178.115

2. 74AUGI4 053453.31 51.118 177.904
55.2 51.484 178.188

3. 74110016 094131.03 51.113 177.555
33.0 51.433 177.875

4.*74NOY28 0528 46.18 51.420 175.189
48.5 51.778 175.293

5.’ 75JAN10 20 40 36.55 51.101 178.398
39.4 51.536 178.432

6. 750MB 192912.69 50.870 178.108
15.3 51.098 178.224

7. 75HAR12 10 43 31.90 51.143 177.527
33.7 51.459 177.778

8.*751‘1AR17 17 39 29.17 51.421 175.318
31.4 51.789 175.384

937514141220 03 23 32.21 50.341 175.926
33.0 50.233 176.095

10.'75APR02 1443 20.26 51.279 178.296
21.8 51.543 178.276
11.7SJUL0820572121 51.147 178.213
22.6 51.491 178.332

1237611818 08 00 58.8 51.466 178.381
56.9 51.481 178.665

13.'76FE822 07224.52 51.315 176.629
26.2 51.588 176.844

14.'761JUL22 1430 14.49 51.085 177.870
16.2 51.366 178.020

15. 76AU016051137.00 51.140 178.233
38.8 51.502 178.409

16.'76AU028 17 29 27.80 50.923 177.753
28.5 51.139 177.880

l7.*765EP22 02 30 25.80 51.401 175.857
27.3 51.652 175.962

18.*77JAN0616 02 06.72 51.255 175.556
08.4 51.444 175.515

19. 77APR20 00 19 15.99 51.035 178.642
(non-oonv.) 16.1 51.214 178.971
20.'77APR28 I3 37 36.41 50.354 177.438
38.1 50.516 177.713

14.7 5.7
45.6 5.7 241 .831
12.9 5.6
46.7 5.6 221 .862
1501' 5.6
43.4 5.6 235 .823
11.4 5.1
46.1 5.1 94 .911
20.0f 4.8
42.3 4.8 48 .720
10.0f 4.8
24.0 4.8 62 .687
18.6 5.2
39.8 5.2 133.733
11.9 5.0
50.5 5.0 100.706
501' 4.9
10.0i 4.9 91 .918
2001' 4.8
40.6 4.8 43 .647
16.9 4.8
40.0 4.8 64 .866
13.8 5.0
10.0f 5.0 41 .746
25.0f 5.0
45.4 5.0 631.100
16.3 4.9
31.8 4.9 48 .745
17.5 5.2
45.2 5.2 126.689
15.0f 5.0
11.1 5.0 45 .478
20.6 4.8
42.8 4.9 102 .816
11.2 5.3
27.8 5.3 162 .883
5.01' 4.8
32.2 4.8 44 .576
5.0f4.7
10.0f 4.7 47 .985

Table 1 (cont'd.).

2S

21.*77JUN29 08 47 14.86 51.418 176.143
16.7 51.744 176.256
22.*77AU01716 48 30.94 51.469 175.282
33.0 51.817 175.362
23.775EP21 1035 27.91 51.068 178.194
30.0 51.461 178.367
2417700111 05 03 09.75 50.896 176.712
12.8 51.112 176.883
25.*77NOV04 09 52 57.32 51.380 175.879
59.7 51.642 176.024
26.*77NOY04 10 02 05.78 51.409 175.624
06.6 51.607 175.619
27.*77NOV04 18 07 32.34 51.329 175.621
34.2 51.518 175.680
28.*77NOV05 14 44 03.79 51.369 175.476
06.6 51.584 175.664
29. 77DEC19 10 52 29.54 50.838 176.262
38.6 51.152 176.529
30. 780AN02 20 57 39.55 50.964 177.854
39.9 51.200 178.209
31.*78\JAN06 07 08 43.75 51.423 175.977

44.6 51.710 176.016

32.*78APR24 04 28 47.37 51.354
48.5 51.650

176.057
176.124

33.’78JUL21 20 50 29.72 51.038 177.904
31.0 51.463 178.311
34.'78(X3T17 20 50 48.22 51.358 176.723
49.0 51.595 176.927
35.*79UAN31 03 07 32.22 51.513 175.793
32.5 51.732 175.859
36. 79FE812 05 1 1 06.88 51.058 178.943
08.2 51.225 179.071
37.'82JAN04 23 37 35.05 51.249 177.869
30.5 51.409 178.409
38. 82JUN04 03 01 03.88 51.282 177.094
05.1 51.635 177.375
39. 820UN15 1957 36.91 50.976 178.156
39.1 51.339 178.506
40.'831‘1AR22 01 32 29.79 50.952 178.278
30.8 51.317 178.570

22.7 4.9
51.6 4.9
16.7 5.4
55.3 5.4
13.6 4.9
38.9 4.9
15.1 4.6
27.8 4.6
18.3 5.6
50.6 5.6
22.4 5.2
38.9 5.2
18.9 5.4
40.8 5.4
20.7 5.3
46.5 5.3
501' 5.1
34.8 5.1
5.0T 5.0
28.6 5.0
20.8 5.3
52.7 5.3
1001' 5.2
48.2 5.2
5.0f 4.9
37.0 4.9
2501' 4.9
43.0 4.9
2501' 5.1
46.0 5.1
5.0T 4.7
36.8 4.7
11.3 4.8
42.1 4.9
21.4 5.7
50.3 5.8
1.0f 5.0
35.5 5.0
6.8 4.9
33.5 4.9

108.780
162.830
98.889
82.659
2681.080
86.692
204.985
230.990
143.707
134.684
1441.030

172 .742

72 .606
44 .784
44 .799
51 .639
48 .749
173 .717
73 .585

53 .754

26

rejected. Event 4' 36 , at the far western edge of the study area, occurred
at a time when the three westernmost stations of the Adak network were
not operating and thus does not have a well constrained local solution.

Eleven of the smallest earthquakes (events S,10,12,13,14,l6,20,33,34,
37,40 in Table 1) were recorded by a distribution of teleseismic stations
that is significantly different from the average station distribution for the
rest of the events. Generally, mislocations for these events showed
considerable scatter when compared to larger events having station
geometries similar to those used to relocate the theoretical earthquakes. I
suggest three possible reasons for this scatter.
1. Heterogeneities in the slab. This undoubtedly causes some scatter in
mislocation effects, though it seems unlikely that this is the primary
cause, since larger events do show a fair amount of consistency.
2. Errors in both local and teleseismic hypocentrai determinations. The
relatively small size of these events makes this a probable cause of a
significant portion of the variation in mislocations.
3. Variations in station geometry used to locate these small events.
Theoretical relocations indicate that station geometry variations will
contribute somewhat to the scatter and make less reliable the comparisons

with theoretical events located with the 'average" station sets. Attempts

27
to make use of these smaller events will be discussed later.

After the foregoing analysis, 1 am left with 13 well located events
recorded by a distribution of stations similar to the theoretical station
sets. These are marked ' in Table 1.

Mislocations for the “non-eastern“ events (excluding those events with
anomalous station geometries) were averaged for increments of 0.1 ' a with
a separation into large and small events depending on whether more or less
than 175 stations reported the event. These averages, which are shown in
Table 2, form the primary data set with which to compare model
mislocations. It is obvious that this data set is rather limited both in
quantity and spatial coverage with all of the events occurring in the thrust
zone. The 6 events for a a of 12.2' are quite consistent and constitute the
statistically best set of data.

Since depths are much harder to constrain than epicenters, the primary
emphasis in modeling will be on obtaining good agreement between
epicentral mislocations. The evaluation procedure then, is to first evaluate
the degree of fit of epicentral mislocations near 12.2“ a in the seismically
active thrust zone, while looking secondarily at other thrust zone

mislocations and at depth mislocations.

28

Table 2. Actual Event Mislocation Data.
Mislocations determined by comparing Adak local network solutions with
teleseismic solutions computed with Herrin travel times and Dziewonski
and Anderson (1983) station corrections (see text). Positive epicentral
values indicate teleseismic mislocation to the north; positive depth values
indicate teleseismic mislocation too deep. Right hand columns give

averages of “large“ and 'small" events for each a; for 122' the standard
deviation is also given.

Ave. misloc. for that a (km)

Event 1' Mislocation (km) “large” ”small”
’1 1 Illl‘l E' Ell E‘EIIE'EII
12.1 38 317 39.1 28.9 39.1 28.9
12.2 1 311 40.6 33.8 37.9 31.0 37.8 24.0
2 365 37.6 30.9 :26 :27 :26 :3.3
3 330 35.5 28.4
7 171 35.0 21.2
11 96 38.1 23.1
15 174 40.2 27.7
12.3 23 136 43.7 25.3 43.7 25.3
12.4 30 175 26.2 23.6 33.3 29.1
39 117 40.3 34.5
12.5 6 81 25.4 14.0 34.8 29.8
24 105 23.9 12.7 24.7 13.4
29 185 34.8 29.8

29

Comparison with the actual data set (see Tables 2 and 3) indicates that
mislocations for the 300 x 80 km model with av/aT of -9 x 10'4
km/sec-‘C are too small. Actual epicentral mislocations at 122' a range
from 35 to 40 km while the model predicts only 29 to 30 km of
mislocation.

I suggest three possible reasons for the poor fit of this model. First,
the poor agreement is due to incorrect position of the slab with respect to
surface topography, which a study of amplitude anomalies on this identical
model (Rogers, 1982) has indicated is plausible. This seems an unlikely
explanation since the actual seismicity we have at 122' a occurs where
the model predicts the greatest mislocations. if the slab itself was
significantly mislocated in the theoretical calculations, then the actual
events now located at 122‘ would be located at a different a where the
model would predict even smaller mislocations. A second possibility for
the poor fit is that the modeling technique used does not adequately
represent the true velocity structure near Adak. It is, of course, difficult
to assess the absolute reliability of the modeling routine given the
inherent uncertainty of the nature of the slab/mantle interaction at depth.
It Should be noted, however, that thermally derived models have been used

in numerous ray tracing studies (eg. Jacob, 1972; Engdahl et al., 1977) to

 

5.351 3.33 5.3 5.55 3.53 5.55 5.55 5.55 5.53 5.55 5.53 5.53 3
5.531 5.53 5.53 3.55 5.53 5.55 5.55 5.55 5.55 5.55 5.53 5.5 55.553x555
5.53 5.55 3
5.33 5.55 55.553x555
5.55 3.53 3
5.55 5.33 533.555555
5.531 5.5 5.51 5.55 5.33 5.55 3.53 5.55 5.35 5.55 5.35 3.55 5.53 5.35 3.53 3.53 3
5.351 5.53 5.3 5.55 5.53 5.55 5.35 5.55 5.55 5.55 5.35 5.55 5.35 5.55 3.53 5.53 5 5.555555
5.5 5.55 3
5.53 5.55 m 5.55x555
5.531 3.5 5.51 5.53 5.33 3.55 5.53 5.55 5.33 3.55 5.5 5.5 3
3.531 5.53 5.5 5.55 5.33 5.55 5.55 5.55 3.53 5.53 5.33 5.51 55.553555
5.531 3.5 5.51 5.55 3.33 5.55 3.35 5.35 5.53 5.55 5.53 5.53 3
5.551 5.53 5.5 5.55 5.53 5.55 5.35 5.55 3.55 5.55 5.53 5.5 533.55x555
5.551 5.3 5.51 5.33 5.53 5.35 5.33 3.55 5.53 5.55 3.53 5.55 5.33 5.35 3.5 5.5 3
5.551 5.5 5.51 5.53 5.53 5.55 5.53 3.35 5.53 5.55 5.53 3.55 5.53 5.53 5.53 5.3 5 5.555555
.55 .55 355 .55, .55 .55 .55 .55 .55 .55. .55 .55 .55 355 355 .55. 3555:
5.53 5.53 5.53 5.53 5.53 3.53 5.33 5.33 15.

.85 583558353:

dump 00... 3303500353333 3335035 0535535550355... 9335505933 583.5 5335035 o>3ufimom

35:35 3350: 5:113 05 3303350035335 053335355533 933550333333 55355 3555:5035 53535035 :33 :32 :3
swim 01.335 533035003533: swamp 6:35 335553350353
com: uom 33030535 6:5 501053552 ...131 x .533 . :6: 550530333 an .36: 330305qu33535 50 swamp

6535353 .305 3 .3395 505 m 6350333350on55 on» 3053

3 3533.5 55135 5353502 6013035553300 5303593 :5 new 5.5 3503300355 505 5330335003553: 3350553513035.

.35qu 3303535003532 ”3:025 350350130532. .m 23335.3.

Table 3 (cont'd.).

Mislocations (km)

13.2

E‘p.

12.1 12.2 12.3 12.5 12.8

11.9
Ep.
28.9 18.4

11.6

fi—r

Dp_.

1513- DP. EP- 99- EP- 013- 513- DP- 519- Up.
37.7 19.9 34.3 14.6

09.

Ep. 09.

 

Model

420x80,7 s

33.9 18.4 27.8 12.2

30.5 16.5

1

40.0 18.5

32.5 21.5 42.2 23.6

420x80,8 s

38.4 21.5 32.9 15.9

34.3 18.9

17.6 -ll.6

44.8 21.9 34.9 12.5

46.4 27.4
42.8 24.8

16.5 18.6 35.4 24.5
23.6 13.9 38.0 21.0

420x80,9 s

12.5 -24.2

37.2 18.8 26.9 5.2

l

31

43.3 22.7 40.0 18.0

38.5 21.6

34.5 20.4

480x80,7 s

32.5 15.9

35.1 18.8

1

540380,5 s

1

13.4 -15.3

42.3 20.9 41.3 19.5 37.7 16.8 28.1 7.2

34.8 19.0

540x80,6 S

9.5 -15.0

1.3

37.9 20.7 35.1 19.0 30.2 15.1 21.2

34.4 18.5

1

39.4 22.3

540x80,7 s

39.5 21.1

1

32

successfully demonstrate various slab effects. One area where the models
used in this study are likely in error is the leading edge of the downgoing
slab, which I have represented as having a rectangulér shape. More
realistic models assume the leading edge to be tapered or rounded in shape.
However, since so few rays actually exit at the bottom of the slab, this
discrepency will have negligible effect on the results. Assuming, then,
that the modeling technique is essentially reliable, I suggest that the poor
mislocation agreement is due to the inadequacy of this particular model.
Therefore, I need to adjust the model parameters in order to increase
mislocations.

One way to increase travel time residuals, and hence mislocations, is
to simply increase av/aT, thereby increasing the velocity contrast for the
same size model. Analysis of a 300 x 80 km slab model with a av/aT of
-l l x 10'4 km/sec-‘C shows that this does indeed increase mislocations,
though the mislocations are still slightly too small (see Tables 2 and 3). A
300 x 80 km slab with a av/aT slightly larger than -l l x lO’4 would likely
result in a good fit with the data. However, the model with a av/aT of -l l
x 10'4 gives an average velocity contrast of "8% and a maximum velocity
contrast of "12% (Figure l0), values which are probably unreasonably high

(Utsu, l97l ; Hasegawa et al., 1978; Suyehiro and Sacks, 1979). Such

33

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

DISTANCE FROM POLE (deg)
IO U A l2 I3 I4
I r —J
' “i
sofa:
100 ' 3
8.25-————1
3200 _
:5 o
z 3.50
p.
O.
IUSOO 3
° * 3.75——
300
400- 3
925——-———
N 9.50 fun/:53
L L L L 1 l L 1
50° 30018014100"
Figure 10. Velocity Profile of 300 x 80, -.0011 Slab.

Symbols same as in Figure 2.

_ 34
contrasts would result in very pronounced shadow zones, the degree of
which have not been observed (Sleep, l973; Rogers, l982). Therefore, a
300 by 80 km slab with such a high velocity contrast does not appear to be
a reasonable model. Such a model also makes ray tracing into these shadow
zones very difficult, with increments in takeoff angle as small as .OOOl'
resulting in differences of greater than 10' geocentric distance upon
emergence at the surface.

Another possible way to increase travel time advances is to increase
the size of the slab. First, I attempt to improve the fit by increasing the
thickness from 80 to lOO km, while keeping the depth at 300 km and the
thermal coefficient at -9 x 10"4 km/sec-‘C. Examination of the resulting
velocity profile (Figure l 1) shows that this increases the overall velocity
contrast with the adjacent mantle very little; it merely has the effect of
widening the zone of anomalous velocity. Also seen is that increasing the
thickness affects the degree of mislocation significantly only in the
forearc and outer rise regions, with little difference in the thrust zone
(see Table 3). Similar results are obtained by comparing 360 x 80 km and
360 x 100 km models. Mislocations at as of l2.8' and 13.2' are affected

by 3 to 4 km in epicenter and up to 10 km in depth. Since 1 have little or no

actual data outside the thrust zone, I cannot confidently constrain the

35

D|ISTANCE |S'ROM POITSEMeq)
l

 

 

 

 

 

 

 

 

10 I4

0 + I J

9 cl

locum‘

Ioo .. .+....—:

J

825-—-—

€200 -
:‘.‘.

I 8150

p...
1.5

,33005- 3

8.75 —-—

d

 

 

 

 

 

 

 

900
400- -
925-——-
9.50 Ian/sec-
L L L L l L L 1 L
500 BOOXIOOI-ODOOS

Figure 11. Velocity Profile of 300 x 100, -.0009 Slab.
Symbols same as in Figure 2.

36
thickness of the slab in this study. Therefore, I assume a thickness of 80
km in all subsequent models since this agrees roughly with the predicted
values based on f lexural studies (Watts and Talwani, I974) and age
considerations (Yoshii et al., I976).

The other alternative is to increase the length of the downgoing slab.
Examination of model mislocations for a 360 km deep by 80 km thick slab
with a av/aT of -9 x 10'4 shows good agreement with the actual data set.
At l2.2’ a, the model predicts epicentral mislocations of 39.2 and 36.5 km
for the small and large station sets, respectively, while the data set gives
average epicentral mislocations of 37.8 and 37.9 km for ”small" and 'Iarge“
events, respectively. Subsequent increases in depth with corresponding
decreases in the thermal coefficient give models with similar mislocations
in the thrust zone. (refer to Figures l2 and I3 and Table 3). For a depth of
420 km, the best fit model would have a thermal coefficient between -7 x
W“ and -8 x 10‘4 ’C-km/sec. For a depth of 480 km, the thermal
coefficient‘would be somewhat less than -7 x IO 4; for 540 km, av/aT
would be between -S x IO'4 and -6 x 10-4.

The rest of the thrust zone data shows more scatter. The one "large"
earthquake at I2. I ' a has an epicentral mislocation of 39.1 km which would

require a slightly longer slab and/or higher thermal coefficient than the

37

.300 :03umum
05303 5333 0:00 030 030:0 00033000305 .3 033535 :3 mm 0500 0300555
.0535 5555.1 .55 x 55m 305 030300> :0330003532 uc0>m 303305m .53 035535

 

I
L ‘9
O
SW8

h b
3 3 d 3

‘-
IO
IN

 

I
a

db

.5. 33:25 33.3
50801.52. .5. on 3. own

0
1n

 

1‘:
(1131‘) HidBCl

 

 

. — 3 LI 1 a
5.5. om. 3.5. 5.5. 3.:
.533 5355 .255... 55255.5

38

.300 0033030
05303 £333 0:00 030 :zocm 00033000305 .3 030535 :3 00 0500 0300E>m
.0030 5555.1 .55 x 533 305 030300> 50330003032 uc0>m 30330nm .m3 030535

 

 

 

 

     

 

 

 

Do
1 3 3 q 3 3 3 3 1 3 3 3 3 3 3 3 3 3 3 3 3 3 3 d
-55 M
9 0 0 D
.IIIIIIIIIII vomw
. a
“mom.
“5333"..333333i3 335
3: 33:25 .53..
0000.01.52. .5. 05 3. 0.6 G
- - 53 3
310i. 5
H
/ 1 ON W.
p . — . . 3 3 h . |Iw|l31l . — . 3 . b 3 L 3 5 3 _ G
On. On. ON. ON. n...

3532 51.05 2555 5523555

39
above results indicate. The same is true of the one ”small“ event at l2.3' a
(epicentral mislocation of 43.7), the "smali' event at l2.4' a (40.3 km), and
the “large“ event at l2.S' a (34.8 km). The "small” event at l2.4’ a
(epicentral mislocation of 26.2 km) and the two "small" events at l2.5‘ a
(25.4 and 23.9 km) argue for a shorter slab and/or a smaller av/aT.

Depth mislocations for actual events are generally greater than for
model events. Model depth mislocations can be increased by up to lo km if
the travel time to ADK is determined using a standard Herrin travel-time
table rather than the table modified to include shallow structure under
ADK Using a J-B travel-time table (Jeffreys and Bullen, l940) results in a
further increase in model depth mislocations of l or 2 km. Hence, improved
agreement with actual depth mislocations is obtainable by simply changing
the travel-time table used. Therefore, due to the variability in depth
locations, I do not attempt any conclusions based on depths.

An attempt was made to utilize several smaller events having
anomalous station geometries. Theoretical event relocations corresponding
to the location of the actual event were made for several of the “favored“
models using the identical station set used to locate the actual earthquake.
In general, this did not decrease the scatter or improve the fit to the

“favored” models for these events. Thus, the variation in mislocations for

40
these smaller events does not appear to be primarily caused by station
geometry differences. Likely, the scatter is predominantly due to
uncertainties in the locations for events of this small size.

There appears to be, in general, an approximately linear relationship
between depth and thermal coefficient in the model mislocations. The
maximum reasonable depth of the slab, assuming experimental values for
av/bT of -5 x l0"4 km/sec-‘C, would likely occur at a depth slightly less
than 600 km. Comparison with the data, then, gives us a range of feasible
models from 360 km depth and a ma of -9 x 10‘4 to about 600 km with a

mm of -s x 0‘“ km/sec-‘C.

DISCUSSION

Given the limited set of earthquakes with which to compare the
modeling data, i cannot easily differentiate at this time between the range
of best fit models given above based on the observed seismicity for the
central Aleutian slab. However, theoretical results indicate that, if a more
complete set of data were available, better constraints are possible. One
possible discriminator between models, especially for thickness, is the
variation of mislocations with varying a. Some models show similar
mislocations in the thrust zone while having somewhat different
mislocations outside the thrust zone; for example compare models
300x80,9 to 300xioo,9,and 360x80,9 to 360x l 00,9, at a = l3.2' in Table 3.
At this location, better constraints on local network solutions would have
to be obtained to make reliable comparisons. The additional deployment of
ocean bottom seismographs would be very helpful (Frohlich et al., l982).
Better depth constraints, perhaps through the use of depth phases, could be
very useful given the relatively large differences in depth mislocations.

Another potential discriminator is the comparison of different size
earthquakes occurring in the same vicinity in the thrust zone. Modeling

4i

42
results show that different size events occurring in the same place can be
teleseismically located differently as a result of station geometry
differences (see Table 3). For model thrust zone events, comparison of
epicentral mislocations for the two model station sets reveals distinct
patterns. For a a of l l.9', shorter slab models (300 x 80 with av/aT of -9
x i0“4 and -l l x 10'4 km/sec-‘C and 360 x80 with av/aT of —9 x 10'4
km/sec-‘C) show mislocation differences between the two station sets of
3.5 to 6 km while reasonable, longer models (480 and 540 km depth of
penetration) have differences of less than i km. For l2.2' a, the opposite
pattern emerges. Shorter slabs show small differences (1 or 2 km) while
longer slabs show greater differences (4 or 5 km). At l2.S' a, the effects
are less dramatic, with shorter slabs showing smaller differences (5 or 6
km) than longer slabs (7.5 km). A large number of events would be required
to make a statistically significant comparison, since differences are on the
order of a few kilometers. I can cautiously apply the above criteria to the
data Note that in the limited set of events at 12.2' a, there doesn't appear
to be a clear relationship between the degree of mislocation and the
number of stations reporting the event. For the six actual events occurring
near l2.2' a, the average epicentral mislocation is almost identical for the

3 ‘large" events and the 3 "small” events. Although this is a rather limited

43
number of events, the lack of variation in mislocation with event size
favors the shorter slab models.

For the determined range of best fit models, my feeling is that the 360
km depth of penetration is most realistic for several reasons. First, the
foregoing discussion of station sets tenuously favors a shorter slab model.
Second, a depth of 360 km agrees roughly with the 386 km best fit depth
determined by Spencer and Engdahl (l 983) using velocity inversion. Third,
depths much greater than 360 km would be difficult to explain based on
assumed plate velocities and the estimated time since subduction
originated (Hays and Ninkovich, l970; Sleep, l973).

The deepest recorded seismicity in the region of this study is at
approximately 260 km depth. This study suggests that the slab penetrates
at least l00 km beyond this point, supporting recent studies that indicate
subducting lithosphere can retain its anomalous character in the mantle far
beyond the point of detectable seismicity (e.g. Jordan, l977; Creager and
Jordan, l984). To my knowledge, the largest previous estimate for the
depth of penetration of the central Aleutian slab is the 386 km estimate of
Spencer and Engdahl (l983). The results presented here indicate that the
slab could conceivably penetrate as much as 200 km beyond this previous

maximum, though I feel this is unlikely.

44

If the 360 x 80 km model (av/3T = -9 x lO‘4 km/sec-'C) is grossly
correct, then an estimate of the temperature at which the deepest observed
seismicity occurs can be made for the slab near Adak. Molnar et al. (l979)
examined the depths of deepest observed seismicity in numerous
subduction zones to derive a relationship between cut-of f temperatures for
seismicity and depth. Their results predict a cut-of f temperature of 630 :
lOO'C for a depth of 260 km. The temperature profile for the preferred
model here, however, gives a coldest temperature of about 800'C at 260 km
depth. Temperatures as cold as those predicted by Molnar et al. (I979)
would not occur at this depth using my modeling procedure unless the slab
was extremely fast and/or thick. it is likely that the close proximity of
the Adak local network allows detection of smaller events occurring at
greater temperatures than those determined by Molnar et al. (l979) for a
given depth.

Using the 360 x 80 km (av/6T - -9 x 10““ km/sec-‘O slab as the
preferred model, i then performed theoretical relocations for two parallel
sets of intermediate depth events intended to represent a Benioff double
seismic zone. One set of events was located along the upper surface of the
descending slab, while the other set was within the slab 31 km from the

upper surface. Figure l4 shows the mislocation vectors and indicates that

DISTANCE FROM POLE (deg)
H

A5

 

 

 

 

 

 

 

    
 

 

 

 

IO L l2
0 i l *r *TT"""
.- h—-———J /
O
l00-'
A P
E
5 b
I D
'3-
a- I-
uJ
= 14
ed
_/ 0.8 9. 8.8 kin/soc
300 i/ —l
<3 .
L L L
Figure 14. Deep Event Mislocation Vectors for 360 x 80
km, -.0009 Slab. Symbols same as in Figure 4.

45
the two major conclusions from intermediate depth mislocations in the
original 300 km deep model are unchanged; events below about 100 km
depth are mislocated by less than l0 km in epicenter and the mislocations
give rise to a spurious increase in slab dip at about lOO km depth. Note
also that below about 200 km depth, the mislocations cause the 3i km
separation of the double seismic zone to appear narrower than it actually
is. The pair of events near 240 km depth appear to be only 22 km apart
upon teleseismic relocation. This implies that estimates of the thickness
of Benioff zones may be too small at certain depths when based on
teleseismically determined hypocenter distributions. The teleseismic
mislocations of these events are not great enough, however, to entirely
account for the eventual merging of the double zone into a single seismic
zone, which has been reported to occur at approximately l7S km depth in

the Japan, Kurile and Kamchatka arcs (F u j ita and Kanamori, 1981).

CONCLUSIONS

In this study, I have attempted to use thermal modeling, along with
seismic ray tracing, to delineate the gross velocity structure of a
subduction zone. The way in which the higher velocity slab affects
compressional wave travel times, and thus, earthquake locations, forms the
basis for the modeling procedure. Published hypocenters for island arc
earthquakes located using teleseismic arrivals and standard location
routines can be in error by more than 50 km (Engdahl, 1977). The degree of
teleseismic mislocation is estimated by comparing ISC and PDE reported
hypocenters with hypocenters determined from local network data only.
Local networks give accurate solutions for events relatively close to the
land based network, primarily thrust zone events (Frohlich et al., l982),
while more distant earthquakes in the forearc and outer rise are less
reliably located by the local network.

Modeling results indicate that teleseismic locations of island arc
earthquakes can be grossly in error when location routines assuming a
spherically symmetric earth are used. The greatest mislocations occur in
the shallow thrust zone, the amount depending on the specific
characteristics of the slab. Changes in the depth of penetration of the slab

47

43

and the assumed thermal coefficient have pronounced effects on model
mislocations, while variations in slab thickness result in only minor
variations to velocity profiles and model mislocations.

From comparison with actual earthquakes occurring in the central
Aleutians, the results suggest that the subducting slab in this region
penetrates to a depth well in excess of 300 km but probably not deeper than
600 km assuming that the thermal coefficient varies between -S x IO'4
and -9 x 10'4 km/sec-‘C. The smaller value is constrained by laboratory
experiments (Anderson, l968), while the larger value is constrained on the
basis of previous studies (eg. Sleep, I973), and velocity contrast and
shadow zone considerations. Within the range of models suggested by
comparison with observed seismicity, the preferred model has a thickness
of 80 km, a depth of penetration of 360 km, and a thermal coefficient of -9
x l0'4 km/sec-‘C based on agreement with the results of Spencer and
Engdahl (l983) and on estimates of the duration of subduction and plate
velocities for the central Aleutians (Hays and Ninkovich, I970).

Theoretical results suggest that models could be further constrained
given a more complete set of seismicity data in the area. Variations in
mislocations with changing a exist for models having similar mislocations

in the seismically active thrust zone. Data in the f orearc and outer rise

49

would be especially helpful in constraining thickness. Also, a larger
number of events in the thrust zone could be used to determine mislocation
differences for various size earthquakes. Theoretical results indicate that
such differences resulting from variations in station geometry could be
used to differentiate between models if a statistically significant number
of events were available. The limited data set used in this study favors the
shorter slab models (i.e. 360 km depth of penetration) based on this
reasoning.

My results, then, suggest a depth of slab penetration of about IOO km,
and perhaps several hundred km, deeper than the deepest recorded
seismicity in the area of study (260 km). This supports the theory that
slabs can retain their anomalous velocity characteristics well beyond the
point where they cease to generate detectable seismicity.

Modeling of deeper focus events occurring along the upper surface of
the slab suggest that events deeper than about lOO km are probably not
significantly mislocated teleseismically. Teleseismic mislocations may
also be at least partly responsible for reported increases in both the dip of
the Benioff zone at depth and the merging of the double zone into a single
seismic zone.

The modeling routine used here suffers from several limitations. The

SO
thermally derived velocity profile representing the subducting slab is
necessarily simplified. While thermal modeling, in theory, can be made as
complex as one wishes, the added uncertainty involved with very detailed
structures precludes their usefulness given the present understanding of
subduction zones. The thermal modeling technique requires numerous
assumptions while the ray tracing technique is approximate.

Other possible sources of uncertainty include plotting and digitizing
errors, representation of residuals by a finite number of Fourier
coefficients, and minor variations in station sets used. In most cases,
these errors are minor, and improvements are not warranted given the

degree of uncertainty in earthquake location routines.

BIBLIOGRAPHY

BIBLIOGRAPHY

Anderson, O.L., Schreiber, E, and Liebermann, R.C., I968. Some elastic
constant data on minerals relevant to geophysics. Rev. Geophys, 6:
49l-S24.

Barazangi, M., and Isacks, B.L., I979. A comparison of the spatial
distribution of mantle earthquakes determined from data provided by
local and by teleseismic networks for the Japan and Aleutian arcs. Bull.
Seismol. Soc. Am., 69: l763-I770.

Creager, KC, and Jordan, T.H., I984. Slab penetration into the lower
mantle, J. Geophys. Res, 89: 303 I -3049.

Dziewonski, AM, and Anderson, D.L., I983. Travel times and station
corrections for P waves at teleseismic distances. J. Geophys. Res, 88:
3295-33 I 4.

Engdahl, E.R., I977. Seismicity and plate structure in the central Aleutians.
In: Talwani, M., and Pitman, W.C. III (eds), Island Arcs, Deep Sea
Trenches and Back-Arc Basins. Am. Geophys. Union, Washington, 0.8,
259-271.

Engdahl, E.R., Kisslinger, C., LaForge, R.C., Price, 5., Topper, R., and Johnston,
AC, I977a. A field study of earthquake prediction methods in the
central Aleutian Islands. Semi-annual tech. report, I Get, l976- 3i
Mar, I977, US. Geol. Survey.

Engdahl, E.R., Sleep, NH, and Lin, M.-T., I977b. Plate effects in North
Pacific subduction zones. Tectonophysics, 37: 9S-I I6.

Engdahl, E.R., Dewey, J.W., and Fujita, K, I982. Earthquake location in
island arcs. Phys. Earth Planet. Inter, 30: l4S-l56.
SI

52

Frohlich, C., Billington, S., Engdahl, ER, and Malahoff, A., I982. Detection
and location of earthquakes in the central Aleutian subduction zone
using island and ocean bottom seismograph stations. J. Geophys. Res,
87: 6853-6864.

Fujita, K., Engdahl, ER, and Sleep, N.H., I98l. Subduction zone calibration
and teleseismic relocation of thrust zone events in the central Aleutian
Islands. Bull. Seismol. Soc. Am., 7I; 1805- I 828.

Fujita, K, and Kanamori, H., l98I. Double seismic zones and stresses of
intermediate depth earthquakes. Geophys. J.R. Astron. Soc., 66: I3I -
IS6.

Hasegawa, A, Umino, N., and Takagi, A, I978. Double-planed deep seismic
zone and upper-mantle structure in the Northeastern Japan Arc.
Geophys. J.R. Astron. Soc., 54: 28 I -296.

Hays, J., and Ninkovich, 0., I970. North Pacific deep sea ash chronology and
age of present Aleutian underthrusting. Geol. Soc. Am. Mem. I26:
263-290.

Herrin, E., and Taggart, J., I968. Regional variations in P travel times.
Bull. Seismol. Soc. Am., '58: I326- I 337.

Herrin, E., Arnold, E.P., Bolt, BA, Clawson, G.E., Engdahl, E.R., Freedman,
H.W., Gordon, D.W., Hales, AL, Lobdell, J.L., Nuttli, 0., Romney, C.,
Taggart, J., and Tucker, W., I968. I968 Seismological tables for P
phases. Bull. Seismol. Soc. Am., 58: I I96-I24l.

Jacob, K, I970. Three-dimensional seismic ray tracing in a laterally
homogeneous spherical earth. J. Geophys. Res, 75: 6675-6689.

Jacob, K, I972. Global tectonic implications of anomalous P travel-times
from the nuclear explosion Longshot. J. Geophys. Res, 77: 2556-2573.

Jeffreys, H. and Bullen, KE., I940. Seismological Tables, Brit. Assn,
London, I940.

53

Jordan, T.H., I977. Lithospheric slab penetration into the lower mantle
beneath the Sea of Okhotsk. J. GeOphys, 43: 473-496.

Kisslinger, C., Billington, 5., Bowman, R., Cruz, 6., Huang, Z.-X., Pohlman, J.,
Toth, T., and Morrissey, S.-T., I982. A field study of earthquake
prediction methods in the central Aleutian Islands. Semi-annual tech.
report, Sep. 30, I982, U.S. Geol. Survey.

Kisslinger, C., Billington, 5., Silliman, A, McDonald, 8, Toth, T., Perin, B.,
Harjes, M., and Morrissey, S.-T., I984. A field study of earthquake
prediction methods in the central Aleutian Islands. Final Tech. Report,
Jan. 30, I984, U.S. Geol. Survey.

Mclaren, JP, and Frohlich, C., I985. Model calculations of regional network
locations for earthquakes in subduction zones. Bull. Seismol. Soc. Am.,
75:397-413.

Molnar, P., Freedman, 0., and Shih, J.S.F., I979. Lengths of intermediate and
deep seismic zones and temperatures in downgoing slabs of lithosphere.
Geophys. J.R. Astron. Soc., 56: 4I -54.

Rogers, W.J.,Jr., I982. The effects of subducting slabs on seismic
amplitudes and travel-times from f orearc and outer rise earthquakes.
M.S. Thesis, Michigan State University, East Lansing, MI.

Sleep, N.H., I973. Teleseismic P-wave transmission through slabs. Bull.
Seismol. Soc. Am., 63: I349- I 373.

Spencer, CR, and Engdahl, E.R., I983. A joint hypocentre location and
velocity inversion technique applied to the Central Aleutians. Geophys.
J.R. Astron. 5°C., 72: 399-4I S.

Suyehiro, K, and Sacks, I.5., I979. P- and S-wave velocity anomalies
associated with the subducting lithosphere determined from
travel-time residuals in the Japan region. Bull. Seismol. Soc. Am., 69:
97-I I4.

Toksoz, M.N., Minear, J.W., and Julian, B.R., I97I. Temperature field and
geophysical effects of a downgoing slab. J. Geophys. Res, 76: I I I3 -
I I38.

54

Toksoz, M.N., Sleep, NH, and Smith, AT., I973. Evolution of the downgoing
lithosphere and the mechanism of deep focus earthquakes. Geophys. JR.
Astron. Soc, 35: 285-3 I O.

Topper, R.E., I978. Fine structure of the Beniof f zone beneath the central
Aleutian arc. M.S. Thesis, University of Colorado, Boulder, CO.

Toth, T., and Kisslinger, C., I984. Revised focal depths and velocity model
for local earthquakes in the Adak seismic zone. Bull. Seismol. Soc. Am.,
74: 1349-1360.

Utsu, T., I97I. Seismological evidence for anomalous structure of island
arcs with special reference to the Japanese region. Rev. Geophys. Space
Phys, 9: 839-890.

Watts, AB, and Talwani, M., I974. Gravity anomalies seaward of deep-sea
trenches and their tectonic implications. Geophys. J.R. Astron. Soc., 36:
57-90.

Yoshii, T., Kono, Y., and Ito, K, I976. Thickness of the oceanic lithosphere.
In: Sutton, G.N., Manghanani, M.H., Moberly, R., and McAf f e, EV. (eds),
The Geophysics of the Pacific Ocean Basin and its Margins. Geophys.
Monogr., A G. U., l9: 423-430.

    

  

R

M'CIIIIIIIIIIIIIISIIIIIIIIIIIIIIIIIIEIIIIIIJIIIIIIIIIIIES