pro-dnces, an
integrated understanding
of the mantle-crust system is impor-tant.
To achievethe desired degree
of resolution seYeralmeans haYe been
employed. First, core phaseshave
been addedto
thedata
setso that
rays with angles morenearly
vertical than thedirect P arrivals
canbe used
. Theserays help
constrain thestructure,
especially that which isdeep beneath
south- ern California. Also, a tomographic method of inversionhas been used since
it allows for a detailed inversion.The
first part
of thischapter discusses
the data and the reductionpro-
cedures that have been
appliedto bring
thedata to a
set of traveltime
residu- als. This is followedby
a description of the details of the tomographic methodof
inversionneeded in
thisspecific
application. Amore
generaldiscussion
of tomography is thetopic
ofChapter
I. Included area
fe,\· examplesof
recon- structionsperformed
on artificially created "data" computed fromsynthetic
structures.
Finally, the inversion of the actual Pdelay data
isgiven
anddis-
cussed
. Chapter IIIdeals with
the interpretation, modeling, and tectonic significance of the observed features.---- .. '
\ Figure 11.1. Locations of the 163 events used in this study. Map (a) gives the geographical locations, while (b) displays the same events in ray parameter-azimuth space. In b) the inner circle is 5 sec/deg and the outer circle is 10 sec/deg.
I w 0) I
-37-
0 ,
o - r---
o - 0 t.D
o - 0
L{)
0
(/)
0 -
- c
:::J ~0
0 u o -
t<')
•
I! II
0 o -
N \,1
I
I I .
I
0
io - - I
rr TT
r ~~
Tn • ••••••1.5 .5 0 .5 1.5
delay ( in seconds)
Figure ll.2. Histogram of the 9888 travel time residuals used in this study. Each ray has associated with it one residual, and the residuals have been tallied for each 1 /20th second increment.
l'" CAL TECH-USGS SOUTHERN CALIFORNIA
NETWORK May. 1, 1979
SBS~ ---:...
20'
-38-
oa,~,.
"··
•TIN
OL~~<~~•~
•ewe
J
Figure ll.3. Map of the seismic station locations for the Southern California Array. This particular figure shows the station distribution for May 1, 1979, which is representative of the distribution during the time interval that data used in this study were recorded .
-
39-
the observations are determined by using the I\TEIS
locations andthe
Herrin Tables (He
rrin,
1968). The core phases are
represented by the symbols·wi
thepicentral distances greater than 100°, or with slowness less than
4sec/deg.
The
data are seen to cove
rall quadrants and range in ray parameter from 0-10 sec/deg
.The south and northeast directions, however, are much more poorly
represented than the northwestand southeast directions
. Inall, about
160events
recorded by the Southern CaliforniaArray are used
. Thenumber of stat
ionsgiving usable
records forany particular event varies, and the
resulting data setconsists of nearly
10,000 rays.Figure II.2 is a histogram of the data
delayvalues. Most of the data deviate from the mean by no more than half a second.
Astation
locationmap
is shownin
Figure Il.3 fort
heyear
1979.There has
been some change
instation distri but
ionthrough time, but the cov- erage shown
in Figure II.3 istypical for the times
fromwhich t
hedata were
recorded. In total,
158 stations were used inthis study. Each station
recordedan average of 63 events while each event was recorded by an average of 61 sta- tions
.To reduce
the data to a set of t
raveltime delays, seve
ralstandard correc-
tions are appli ed. These corrections are elevation and sediment corrections
(applied
inthe same manner as
Raikes,
1980) and travel timecorrections
. Thetravel time corrections
include reductionsby dT /d..6.. and
d2T / d..6..2(from the
Herrin Tab
les). Also, the averagedelay for each event
is removed to reducethe effect of errors
inthe source parameters. The core phases,
PKPand
PKIKP,a
rereduced somewhat differently since reduc
tion by the HerrinTables
cons
istentlyunderestimates dT /d..6..
.Fortunately, a large event (mb
=6.1 )
-40-
occurred near the antipode to southern California (£::.. ::::::::: 175°). The Herrin corrections, small for this event, were applied to produce a reference delay map. It >vas then a straightforward matter to adjust dT / d£::.. for each of the other events in order to best match the reference P delay map in a least- squares sense. Since the antipodal travel time residuals indicate the delay
accumulated directly beneath each observation site, and also because this e.-ent was exceptionally >veil recorded, this P delay map has been chosen as an example P wave map (Figure II.4). For comparison, the P delay map resulting from an event of similar magnitude (mb = 5.9) that occurred in Korea
(.6. =
70°) is also shown. Arrivals from this e.-ent are from the \V. 0V, and the general pattern is seen to shift towards the ESE.In addition to the standard reductions, a correction for variable crustal thickness has been applied. The indiYidual station corrections are calculated from the station P n time-terms of Hearn (198-la). These time terms may be due to variations in either crustal velocity or in Moho depth, but since the time term method is especially sensitive to variations in the ::.1oho depth, this was assumed to be the cause. Corrections are determined by calculating a travel time slab correction v·:ith a slab thickness given by the de•·iation of the ::.1oho from its aYerage depth, as inferred from Hearn's time terms (19 -la), and by using an assumed .-elocity contrast across the ::.Ioho of 1.2 km/sec. The travel time effect on the P delays resulting from varying crustal thickness is less than that produced by varying crustal velocity, and so this is the more conservative of the two approaches. The corrections made in this way are in the range ±0.37 sec. >rith an average deviation of ±0.06 sec.
0 0
delta=175
0
0
rfo
D 0delta= 82
0
0
',,
0
0
'··· ..
,·,
0
0
. .
' ' '
~
0 " ... '".!
0
.... ···
,-'
" ...
.
I___
,_...:
'
',
·.
,-' ' ' '
Figure ll.4. Maps of travel time residuals for two events. The solid triangles represent early arrivals, and the open squares represent late arrivals. The size of the symbol is pro- portional to the value of the delay. The upper map is the result of a nearly antipodal event and thus shows the integrated delay directly beneath each station, while the lower map is for an event in Korea (.b. = 82°, to the WNW) and the delay pattern is shifted to the ESE.
-42-
2.3 Method of Inversion
The method
of tomography
was chosen to invertthe data.
Themajor
advantageoffered
withthis app
roach isthe ab
ility to handle a
detailed inver-sion.
Thetheory and methodology are the subjects of Chapter
I, and only a few int
roductory comments willbe given
here. The statement of the problem
is identicalto that
mostcommonly used
inthe generalized
inverse problem(see,
for example, Aki et aL (1977)
for a discussion that is particularly relevantto the geometry of
teleseismicarrivals).
This involvesdividing the
region in whichone
is inte
rested intoa
numberof
discrete blocks, and the slowness per-turbations
tothese blocks that best produce the observations a
re sought.Th
e geometry of the near
normal incidence of teleseismic rays
result ina
fewspec
ial propertiesthat can be taken advantage of, eithe
r forthe
purpose of gaining insightor
tosimplify the calculational
formulas. An important obser- vation can easily be made that the average slowness perturbation of each layeris
t
he same.(Since
theave
rage delayhas
been removed fromeach events set of t
ravelt
imes, this
value is zero.) There
is noability, therefore
.to resolve the
average vertical structure. This
is simply a statement that all rayst
raverse the entire inversion domaint
hickness and
therefore lackthe
ability to resolvethe
averagevertical
structure,or equivalently, that the eigenvectors of the
information matrix (LTL,Chapter
I) do notspan this dimension and are
independent of changes in it.Two approximations which simplify the computations have been imple- mented. Since
all
rays are fairly vertical,
it is "·ithout significant
lossof accu-
racy
thatone may assign the
ray to one and only one block per layer. Ifthe
ray happens
to penetratemore than one block, only the block "·ith the longest ray segment is
used, andit is
assumed that theray
traverses the entirelayer within
that singleblock.
Thisgreatly
simplifies the geometrical considerations thathave
tobe made. Once ray segments are associated with an entire block,
allray lengths within
anyblock are
approximately equal and equationI.l
canbe simplified
tosb
=~dr /~trb without
perceptible alteration ofthe
r r