Вы находитесь на странице: 1из 19

Bulletinof the SeismologicalSocietyof America,Vol. 75, No. 4, pp.

1105-1123,August1985

CHANGES IN V J V , WITH DEPTH: IMPLICATIONS FOR APPROPRIATE VELOCITY MODELS, IMPROVED EARTHQUAKE LOCATIONS, AND MATERIAL PROPERTIES OF THE UPPER CRUST
BY CRAIG NICHOLSON* AND DAVID W . SIMPSON ABSTRACT

Properties associated with the arrival times of shallow microearthquakes are determined for three tectonically different sites: Soviet Central Asia; the Central United States; and southern California. Simple graphical presentations of the earthquake data, including Wadati and Riznichenko diagrams, are used to resolve the trade-off in origin time and focal depth. Estimates of the average half-space velocities for P and S waves, and of the travel-time ratio (t./tp) are also obtained that are relatively free of any model constraints used to initially locate the earthquake. In all cases, P and S arrivals exhibit a systematic decrease in t,/tp with depth in the upper few kilometers. This decrease is significant and matches similar variations observed in situ for the velocity ratio (Vp/V.). Based on laboratory and theoretical studies, such observations are consistent with the closing:of saturated microcracks with increasing confining pressure. Independent determinations of velocity structure and rock properties from both refraction studies and borehole experiments show analogous results. By specifying separate velocity models for both P and S waves that reflect this variation, rather than assuming a constant value of Vp/V., we find improved stability in hypocenter determination and increased resolution of shallow seismicity patterns.
INTRODUCTION

The use of microearthquakes to determine t h e orientation of active faults, focal mechanism solutions, and details of local crustal structure strongly depends on the velocity model used in the earthquake location process and, as a consequence, is subject to systematic errors introduced by inappropriate model parameters. With the advent of digital data and the increasing use of three-component stations, large numbers of good quality S-wave arrivals are being recorded by microearthquake networks. These S-wave arrivals require proper modeling if full advantage is to be taken of the increased resolution such secondary arrivals can provide (Buland, 1976). For these reasons, we began to investigate the characteristics of microearthquake travel times as a means of identifying: (1) model-induced errors; (2) average variations in local velocity structure; and (3) general material properties o f the upper crust.
TECHNIQUES

The principal techniques involve the use of Wadati (1933) and Riznichenko (1958) diagrams as shown in Figure 1. Both methods have had a long history of use as primary tools in the earthquake location process. Wadati diagrams helped to verify the existence of deep focus earthquakes under the islands of Japan (Wadati, 1933) and were subsequently adopted as a common procedure for determining origin times of local and regional earthquakes (e.g., Bune et al., 1960). James et aL (1969) comment on the improvement in earthquake locations if the standard four parameter hypocentral solution is reduced to three spatial parameters by using the Wadati
* Also at the Department of Geological Sciences of Columbia University. 1105

1106

CRAIG N I C H O L S O N

A N D DAVID W. S I M P S O N

method. The simple geometries involved in the Riznichenko diagram were first discussed by Green (1938), although the semi-log form of presentation for use in determining velocities and earthquake focal depths was developed by Riznichenko (1958). More recently, Hales et al. (1981) and Toth and Kisslinger (1984) have shown how both techniques may be used for determining origin times, focal depths, and velocities in subduction zones, provided terms to account for the sphericity of the earth are used and higher order coefficients are included in the Wadati diagram. The Wadat! diagram (Figure 1A) is based on a linear relationship between the arrival time of P (Tp) and the time difference between P and S (T, - Tp)
T, T,~ =

(Tp - To)(t,/t,,

1).-}-

This produces an intercept that is an estimate of the origin time (To) and a slope that is a function of the average trave-time ratio (t~/tp).
A 4=

INPUT: Ts, Tp m .E OUTPUT: To, ts/tp, &

ASSUMPTIONS: Consfant
Poisson's

2-

Ratio
Ts-Tp = T p ( f s / t p - 1 ) - ToOa/tp - 1 ) 0

sfddev ...... ....... ; ........ i ........ , .....i n,,j ........ li~ll,,,,


29 30 31 Tp 10.52 33 34 35

in sec.

INPUI::

Unear Ray Paths


d ._c
Tz ~ Tz = H / o

Tp, To, A
OUTPUT:

.~k I ~ - I

Tp-To = ~ A 2 +

H2

Vav Vav

H, ray, Tz,
&

stcldev
1.0 0 5 10 15 20 25

(epicentral disfance) in kin.

FIG. 1. Equations and figure,~ for (A) Wadati diagrams and (B) Riznichenko diagrams. To the left are input and output parameters, in the middle are the graphical forms, and to the right are the equations.

The Riznichenko diagram (Figure 1B) is based on a linear relationship between the square of travel time (t 2) and the square of epicentral distance (A 2)
t2 _ A 2 -I- H 2
V2
ao

'

appropriate for close distances and shallow focal depths. This produces a slope that is a function of the average half-space velocity (Vav) and an intercept that is a function of the Vertical travel time (tz = H/Vav, where H is the earthquake focal
Arrival times are upper case (e.g., Tp); travel times are lower case (e.g., Tp - To = tp).

CHANGES IN

VjV, WITH DEPTH

1107

depth). Following the Soviet practice (Bune et al., 1960), we plot the log of the travel time against distance to accentuate travel times to the nearer, more important stations. These methods have the advantage of being simple and robust. The Wadati diagram depends only on the observed arrival times of P and S waves and is completely independent of any parameters used in the hypocentral location procedure. The Riznichenko diagram normally requires estimates of the origin time (To) and epicentral distance (A). In practice, we use the origin time as determined from the Wadati diagram and distances as calculated from the trial computer location. If stations are well distributed in azimuth around the epicenter, these distances are generally well controlled. Half-space velocities for both P and S waves can then be easily determined, as well as several independent estimates of focal depth. For example, if values for T~ - Tp are used in the Riznichenko diagram, average effective velocities for the separation time between S and P are calculated, and focal depths independent of origin time are produced. The different estimates of focal depth can then be used to identify cases where compensating errors in depth and origin time have constrained the focal depth to remain near the starting depth used in the iterative computer location program. The velocity information these methods provide is also useful in a number of other ways. Wadati diagrams for a suite of earthquakes at various focal depths gives tJtp as a function of depth. This in turn can be used to relate the S-wave velocity structure to that of the P wave. The average half-space velocities determined using the Riznichenko method represent averages over all rays to all stations. They are neither interval velocities, nor strictly simple average vertical velocities from the focus to the surface. They can, however, provide sufficient information to resolve a number of first-order velocity variations independent of the original velocity model used for location. Conversion of the average velocity estimates into nominal interval layer velocities is accomplished by either differentiating the vertical travel-time versus distance curve (Bune et al., 1960), or by backstripping from the surface the average velocity-depth profile. Layer boundaries are identified by significant changes in the travel-time or average velocity plots. Variations of velocity within the network can be identified by a slightly different technique. In this method, average velocities are calculated using the travel time and the hypocentral distance to individual stations. These estimates are more dependent on local geometry and so are more useful in identifying gross lateral changes within a small region. As these data are ilot, however, realistic estimates of regional averages, they are not subsequently converted to interval values. In this paper, we show that for shallow earthquakes recorded at close distances, tJtp initially decreases with depth. This decrease is significant and must be considered if accurate locations for shallow earthquakes are to be achieved. Furthermore, if ts/t~ initially decreases, so too must Vp/V~, the ratio of the compressional to shear wave velocities. Since Vp/Vs eventually begins to increase with depth into the upper mantle (Hales et al., 1981), Vp/Vs attains a broad minimum that approximately coincides with the depth range of most common crustal earthquakes and reflects certain depth-dependent fracture properties of the upper crust.
RESULTS FROM SOVIET CENTRAL ASIA

Our original intent in applying these methods was to provide a convenient means of checking computer located earthquake hypocenters for sources of error and proper convergence. Figure 2 shows an epicentral map and vertical cross section

1108
C3 0 0

CRAIG NICHOLSON AND DAVID W. SIMPSON


CD

C
c3 ~)

C3

ED
.

Ib~ C~

u~oo

F
Z LD

C%1 ( r )

(uuH)

H.Lc130 o

0,1:

O
o o D

++

~+

xx o

o o

o~

[[*;i',t XI+
* "4"
E -I~

~o
,

r.~

x x +

,(
x

+.,.

o
I i i i i i i i i i I , , , , ,

~=

EEl
i ~ i i

fx
i i [x i I

0 (~I 0

R~

, x ~

F-

0 +

C3 ~

+ iV +
+
m ,

.++

+e

+ +

bJDO

,+ +

*+~

"+

+i x

~++ + ++

+ X

)+~

,:r

.1-

CHANGES

IN

VJV,

WITH DEPTH

1109

near Toktogul Reservoir located along the Naryn River in the Soviet republic of Kirgizia. The network was installed to monitor induced earthquake activity associated with the filling of the reservoir and consists of 10 high-gain stations, each connected by radio telemetry to a central recording site (Simpson et aL, 1981). Three of the stations are three-component, with station spacing on the order of 10 km or less. In addition, network data are supplemented by data supplied by several regional Soviet stations at greater distances. The location program used for determining the hypocenters shown in Figure 2 was HYPOINVERSE (Klein, 1978). The velocity model used by the program assumed homogeneous layers and a constant value for VJVs of 1.78. The cluster of events directly below the dam is interpreted as the seismicity induced by the reservoir. Other well-located events that extend to depths of 40 km form part of the natural background seismicity. As seen from the figure, a large number of earthquake hypocenters remain constrained at the starting depth of 3 km, a depth chosen as appropriate for the shallow induced events. For many of these earthquakes, sufficient data were available to determine that their actual focal depths were much greater. Using the methods of Wadati and Riznichenko, we were able to determine that one of the primary causes for this result was the incorrect assumption used in the computer location program that Vp/V~ was constant and did not vary with focal depth. Changes in t~/tp with depth. Figure 3 shows Wadati and Riznichenko diagrams for six events near the center of the Toktogul array. Focal depths range from 1 to 38 kin. The Riznichenko diagrams clearly show the change in shape of the traveltime versus distance curve with increasing focal depth, indicative of both a geometrical effect and an increasing P-wave velocity. The Wadati diagrams show a systematic decreases in ts/tp, suggesting that the velocity for shear waves increases faster than that for the compressional waves in the upper kilometers of the crust. An interesting feature, however, is the linearity of the Wadati diagrams. If t~/tp (and therefore VJV~) indeed changes with depth, the linear relation assumed in the Wadati diagram is no longer valid (Kisslinger and Engdahl, 1973). Yet it is apparent that, at close distances and shallow focal depths, the expected curvature of the Wadati diagram is not sufficiently resolvable with the available data. Average velocity structures for P and S. A summary of average half-space velocities for P and S at Toktogul calculated using the Riznichenko technique and values for tJtp from the Wadati diagram are shown in Figure 4. Values shown are those data reported during a 1-month interval and are plotted as a function of focal depth for each individual earthquake. The internal consistency of the data suggests that these values are useful as estimates of the first-order variation in velocity structure and can be used to improve the velocity model used for location. Consequently, the data were smoothed by a polynomial fit (dashed lines), and then backstripped to yield interval velocities (solid lines). Backstripping simply starts with the surface value and then determines the incremental velocity that, when averaged with all the p~eceding values, maintains the shape of the best-fitting curve. Depth intervals over which the backstripped velocities remained relatively uniform were chosen for layer boundaries. Interval values for VJVs were calculated using the ratio of the interval P and S velocities. At depths greater than 20 km, where the earthquakes were few and velocities less well resolved, the velocity for the S wave was restricted to a value corresponding to a VJV~ of 1.68. Earthquake relocations using P and S velocity models. Because of the systematic variation in tJtp with depth, the location program was modified to accept separate

1110

CRAIG NICHOLSON AND DAVID W. SIMPSON

velocity models for P and S waves. The values used were those shown in Figure 4 as interval velocities. Figure 5 shows a comparison of this modification on the depth distribution of earthquake hypocenters at Toktogul. The data chosen for comparison include only those earthquakes for which depth control and a good azimuthal

I 0 50 A

I _J 20

&
4
elm i-I I.-

/..~.,

g~/ / / s" ,0/" /J" ~f"

2
911" / /

6 8 Tp in secs.

10

12

10.0

.E 1,0
O I-I 13_ I--

/ ),, /

z~ 0 ~" 4,

2 3 4 5 6

4.42 10.78 14.99 25.71 37.91

5.60 5.86 5.94 6.12 6.27

1.85 1.72 1.67 1.63 1.61

0.08 0.04 0.10 0.12 0.10

0,1

20 4o 60 A (epicentral distance) in kms.

80

FIG. 3. Variation in Wadati (top) and Riznichenko (bottom) diagrams with earthquakes of different focal depths at Toktogul. Inset shows a map of the earthquake locations (numbered) relative to the array, tJtp from the Wadati diagrams show a systematic decrease with focal depth as determined by the Riznichenko diagrams, stddev is the standard deviation (in seconds) between the data and the linear least-squares fit on the Wadati plot. c o v e r a g e exist. I n t h e f i r s t c r o s s s e c t i o n ( F i g u r e 5A), a c o n s t a n t v a l u e for Vp/Vs w a s used. T h e s h a l l o w i n d u c e d e v e n t s d e f i n e a t i g h t c l u s t e r b e n e a t h t h e d a m , h o w e v e r , as n o t e d e a r l i e r , few of~ t h e e a r t h q u a k e l o c a t i o n s i t e r a t e d t o d e e p e r d e p t h s a n d a large n u m b e r w e r e a r t i f i c i a l l y h e l d a t t h e s t a r t i n g d e p t h o f 3 k m . C h a n g i n g t h e

CHANGES

IN

VJVs W I T H

DEPTH

1111

starting depth to 10 km (Figure 5B) allowed more hypocenters to relocate deeper, but many were still constrained at the Starting depth. Moreover, the induced events exhibit a pattern of migration towards the surface, implying several did not properly converge. If separate velocity models for P and S are used,-incorporating the observed variation in tJt~ with depth, then both the shallow induced events and the deeper earthquakes are well located and the effect of the starting depth is not as prominent (Figure 5C). Starting depth for the earthquake hypocenters shown in Figure 5C was 10 km.
RESULTS FROM SOUTHERN CALIFORNIA

Another variation of these simple techniques is to determine, for a given station, the average velocity between that station and each earthquake hypocenter. This
Vp ( k r n / s e c ) Vs ( k m / s e c ) t s / f p (Wodafi)

5.0

5.5

6.0

8.5

7.0

5.5

3.0

3.5

4.0

4.5

1.5

1.7

1.9

2.1

Io 1~o 0" , Ii~e ",4.,.--Icl,~0 ] 43D

,Xoo
i,~_ I ~

Iooo
/

,0

o~t 0 ,
I

,o

0:,~o ~>

2:

:!f
o

~o

-'ok O 0

,,o
' GD

o o
20

20

~o
I

~o
,

/ oi:
30 3o

,
~o 30
o f

FIG. 4. Averagehalf-space values of Vp, V,, and tJtp for each earthquake as a function of focal depth at Toktogul. Focal depths are those derived from the Riznichenko diagram. Dotted lines are polynomial fits to the data; solid lines are the backstripped interval velocitiesused to relocate the earthquakes shown in Figure 5C.

velocity is calculated using the travel-time based on the Wadati origin time and the linear hypocentral distance from the Riznichenko diagram. An example is taken from the southern California network, SCARLET, operated by the U.S. Geological Survey and the California Institute of Technology. Network geometry is shown in Figure 6, along with the locations of earthquakes used in the study. Model-induced errors. Figure 7 compares the velocities from this method with similar results using the hypocentral parameters of origin time and focal depth from the H Y P O I N V E R S E computer location program. Average Vp, Vs, and ts/tp to station MLL are plotted for each earthquake as a function of focal depth. The

1112
0

CRAIG NICHOLSON AND DAVID W. SIMPSON


0 0 0

. . . .

w 0 ~

. . . .

.o o , o

Ii"
+

(m~l)

Hd30 o

....
x

i
xxx

o ....... o
=

o .o
oo o

-,-~;;

%~==.
x = x X

o0 0o
~ o
o o

Oo

"
x

"=x ~ ~ o

o o

o oO

=~
o =

~o

x o

o o

~-)

~,

,. . . . . . . . .

~ '~ . . . . . . . .
o o

o~o . . . . .

.~+++ L +* ~ ~ x

x~ x S ~ <x

xx X~l~

o D.^

o oo o = o

:.,3~-+,.,,Jt.

+ ~+ ,

,+.,,,~ro'=~o .+ = o o ~,~ = o=o o o %

,,:~

@o o o

x + ~

Xx = X' Xx

R= Io R

8 ~

oO
o ~

e o o o

o
oo ~"

.-~ "0

"C~

o x T~ : ---L ~ E L ~ I I , I I I , , o , I = I ~ I

o o

oou

~.

>~x

;> OJ ; >

F* .,,+

*~* "kt .,F-e

"x +xx=

o ~o

o 0

x XX

XXx "X ='0

= 0

xx (]~ X X " i L I I ~ 1 I 1 I I I t I I I I I

o (U~t) H/d3(]

CHANGES IN

Vp/V, WITH

DEPTH

1113

overall range in hypocentral depths is similar in the two sets of diagrams, but three differences stand out. The depths of the HYPOINVERSE hypocenters (Figure 7, top) are distinctly clustered above discontinuities in the velocity model used for location (long-dashed lines), whereas the model-independent hypocenters (Figure 7, bottom) are more uniformly scattered with depth. This suggests that the clustering is model dependent, as is the tendency for the velocities to align along the average of the input velocity model (short-dashed lines). Second, the velocity profiles based

116.50

33- 65 i1750
O ~ 10 KM

~io.uwm

FIG. 6. Earthquake epicenters detected by the southern California network, SCARLET from 1977 to 1980, and used in the velocity study. Heavy lines are major faults; shaded areas represent elevation contours above 6,000 ft. Circled stations exhibit velocity minima; stations with squares do not. Hexagonal stations exhibit velocity minima but only for azimuths that generally correspond to propagation along or across the San Jacinto fault. Double-bordered stations are discussed in the text and shown in Figure 8.

on the Wadati origin times and Riznichenko focal depths show a well defined minimum in Vp, Vs, and ts/tp at about 8 to 10 km, suggesting a low-velocity layer above this depth. No evidence for such a velocity minimum is available from the velocity profiles based on the computer-located hypocenters. Third, there is an overall decrease in ts/tp with depth that is in sharp contrast to the assumption of a constant value for VJV~ used in the computer location program (Figure 7, top right diagram).

1114

CRAIG

NICHOLSON

AND

DAVID

W.

SIMPSON

The relocated hypocenters, based on the methods of Wadati and Riznichenko, were then used to define the orientation of active faults and the pattern of deformation as defined by the resulting focal mechanism solutions (Nicholson et al., 1984a). The results were sufficient to resolve a number of small-scale secondary structures that had not been previously identified. Regional variations in velocity. In dealing with a complicated geologic environment, such as the San Andreas fault through the Eastern Transverse Ranges, lateral

Vp HYP ( k m / s e c ) 4.0 5.0 L I 6.0 I I


i

Vs HYP ( k i n / s e e ) 2.0
I

V p / V s HYP 4.0 1.6


I

7.0 I i

3.0

1.7
I t I I

| .8
i J

1.9
-'3

I I
12

I I I

I E

10

10

I I I

20

20

Vp Wad ( k m / s e c )

Vs Wad ( k m / s e c )

20L
o

vp/vs (Wadati)
1.7 I
1,8 1.9

4.0
L

5.0
I I

8.0
i I

7.0 I~

2.0
I

3.0
I I

4.0 I

.6 I

~'<>

<>

10

o
o~. ~

10

10 o\
o

\o /o
o

Io I o

o 20

~oL

o%

20

FIG. 7. Average half-space velocities calculated to station MLL. Top three figures show values based on origin times and focal depths determined by the computer location program; those at the bottom are based on Wadati origin times and Riznichenko focal depths. Long-dashed lines in top diagrams are interval velocities used in the location model; short-dashed lines are average vertical velocities. Dashed lines in bottom diagrams are polynomial least-squares approximations. Top data show pronounced clustering above discontinuities in the location velocity model. Bottom data show less clustering and suggest a low-velocity layer at or above a depth of 8 to 9 km. Values of t,/tp also show a significant decrease in the first few kilometers.

CHANGES IN V~,/V, WITH DEPTH

1115

variations in velocity are to be expected. A simple procedure to detect these variations is to compare the average velocity profiles to various stations within the network. Significant differences will t h e n identify large-scale heterogeneities. Figure 8 shows average velocity profiles for six stations in southern California. T h r e e of
Vpw (kin/s) BlL 4.0 5.0 6.0 7.0
I

Vpw ( k i n / s ) CKC 4.0 5.0 6.0 7.0 4.0


I

Vpw (kin/s) MLL


5.0
I P

6.0
I I

7.0
I

o /o

E
10

~lo
D

~d

oC o

I0
o

o S o ~
o

lo

8 0 iD

o~o~

\%
%
,

\o

~o
\

i
o m o N~ D []

2c
4.0
I

g
Vpw ( k i n / s ) SM0 5.0
I b

20

~
Vpw ( k m / s ) PSP
5.0 6.0 7.0

20
Vpw (krn//s) RAY
4.0
I

6.0
I ~

7.0

4.0

5.0
I [

6.0
I I

7.0
E

o 00

od
I
111

ol
r~

o 10

o
10
o o o o

%
i o o

0 0 ol:l ~ ~00

1
oEt

~
20

20

oI

20

FIG. 8. Model-independent half-space velocities calculated for six stations of the California network. At top are three stations located north of the northern branch of the San Andreas fault zone, in the vicinity of the San Bernardino Mountains; those at bottom are located south of this fault, in the San Jacinto Mountains (see map, Figure 6). the stations (including M L L ) located n o r t h of the n o r t h e r n branch of San Andreas fault exhibit velocity minima, while the three stations south of this fault do not. This suggests a mid-crustal low-velocity layer is present under the San Bernardino M o u n t a i n s (where the three stations t h a t show the velocity minima are located) but not under the San Jacinto M o u n t a i n s (Figure 6). Other stations exhibited

1116

CRAIG NICHOLSON AND DAVID W. SIMPSON

velocity minima, but only for earthquakes recorded at certain azimuths. Most of these stations are located adjacent to the San Jacinto fault, and the azimuths for which velocity minima were most prominent corresponded to propagation along or across the fault zone. Evidence for such localized low velocities associated with either fault zones or regional mid-crustal layers have been previously identified on the basis of seismic refraction surveys (Mooney and Colburn, 1985), seismic reflection profiles (Feng and McEvilly, 1983), sonic well-log studies (Moos et al., 1983), microearthquake travel-time delays (Healy and Peake, 1975), and by the presence of regional phase arrivals Py or P (Gutenberg, 1951). If the data presented here are taken at face value, then the possibility exists that the overthrust San Bernardino Mountains are rootless, and that the dividing line between this structure and that of the San Jacinto Mountains to the south is indeed
89.66 89 .S1 8936 36.38

.LV /

.WBTN
-. \V / MORT.. o / L

A"

TN

" ~
~ \ "

MADT o-

PUSH / /-

[
I

36.29

%o
.e,.
~

o.PUMP /
, SGT~
o

.
o

~ P ~ / ~ . j / ~ V ~
,-._1/ ~

_a\]

/nf])

e CWTN

go

fGRT (:"'
/

_~.~ .~ .~
~q

oHRCTA oARGOo ~ ~ o
.

~ ~

- . , , ,

o
, ,

o o
, ,

~CCTN
,

,,

89.66

89.51

36.20 89 36

FIG. 9. Earthquakes located during the operation of the portable array (10 May to 11 June 1978) in the central Mississippi valley (56 events). The town of Ridgely is in the center; Mississippi river is to the left. Horizontal errors in location are less than 0.5 km.

the northern branch of the San Andreas fault through San Gorgonio Pass. Regional gravity data and the distribution of Pg velocities also support this interpretation (Hearn and Clayton, 1984) Earthquake phase data are, therefore, useful not only to improve local velocity models, but to identify significant lateral inhomogeneities and possible allochthonous terrains.
RESULTS FROM THE NEW MADRID AREA IN THE CENTRAL UNITED STATES

The methods of Wadati and Riznichenko can also be used as a basis for more sophisticated techniques. Figure 9 shows a map of a dense network of portable seismographs installed to monitor a portion of the earthquake activity associated with the New Madrid seismic zone in the Central United States. The collected data were analyzed for earthquake arrival times and subsequently inverted for velocity structure and hypocentral locations (Nicholson et al., 1984b). The main purpose of

CHANGES IN VJV, WITH DEPTH

1117

the study was to investigate the influence of hypocenter relocations using a variable Vp/V+ in resolving details of local seismicity patterns and to compare the velocity structure obtained from the earthquake data with models determined from refraction studies, borehole data and a formal velocity inversion. Determination of the first-order velocity structure. Figure 10 reproduces the principal results of this analysis. In each case, the average half-space velocity from the Riznichenko diagram and 4/tp from the Wadati diagram are plotted against the

vp Ckm/sec)
3.0
I

5.0
I /,{
O

7.0
I I o
o o

1.0 I

Vs (km/se) 2.0 3.0


i L_!__.. I .... I f

Ts/Tp (Wadafl)
4.0
I

1.4
I

1.8 I

2.2 I

2.6 I

3.0 I

"9 o

"-i o

j.

o='

o o

\~o

o ~, f~ o o ~ o

~',,o o o,,

:zlO
i.u

X~ 'Uo

10

~o

10

i i
1 1

i
20

f
o Ts/Tp AVG (Wodofi) 1.8 2.2 2.6 3.0
I 10" I<> L. I I I
I

20
Vp AVG 5.0

20

(km/sec)
7.0
1.0
I Jo -

3.0

Vs AV+ (km/sec) 2.0 3.0


~t_l ~ o I ~--~__~ 0 I I

4.0
I

1.4
|

o Ib 01

"~. 14b--

eJ

+-

~I0
o

10--

lO

20

20--

2O

FI6. 10. Average half-space velocities for P and S waves as a function of focal depth for each individual earthquake shown in Figure 9. Travel-time ratios (t+/tp) are from Wadati diagrams; focal depths are from Riznichenko diagrams. Top are original data; bottom are the same data smoothed by a moving-window average starting from known surface values. Arrows identity significant inflection points, representing discontinuities in velocity structure. The heavy dashed line in Vp AVG plot is from refraction study (Mooney et al., 1982); dotted lines are results of inversion study for both P and S waves. Both have been smoothed by a filter to account for averages over up-going rays.

1118

CRAIG NICHOLSON AND DAVID W. SIMPSON

earthquake focal depth determined from the Riznichenko diagram. At the top are the original data, and at the bottom are the same data interpolated at constant intervals and smoothed by a moving-window average. Like the results from Toktogul and southern California, ts/t~ is not constant, but changes systematically with depth. Both V~ and Vs increase rapidly in the upper 2 km from their known surface values and exhibit a mid-crustal low velocity zone below a depth of about 2.5 to 3 km. Another significant inflection occurs at about 5 to 6 km depth, and marks the return to increasing velocities that asymptotically approach constant values below a depth of 10 km. A velocity model composed of uniform layers would represent such a structure with velocity discontinuities at 2.5 to 3 and 5 to 6 km depth. Mooney et al. (1982) identify just such boundaries at 2.0 and 5.0 km for this area based on a long-baseline refraction experiment (Table 1). They also found unusually high velocities at a depth of 1 to 2 km, and a low-velocity zone. A comparison of their model with the earthquake velocity data is shown in Figure 10 as the heavy dashed line. The data imply that any attempt to accurately model the local velocity structure requires separate studies for both S and P.
TABLE 1
CRUSTAL VELOCITIES AND B U L K PROPERTIES IN THE N E W MADRID AREA Depth* (kin) Case 1 Mooney et al. Vp (km/sec) Case 2 Final Model Vp (km/sec) Case 3 Final Model V~ (km/sec) Interval Vp/V8 2/3-1/3 Mooney et al. Density p (grn/cma) Poisson's Ratio ~ Bulk Modulus K (101Pa) Shear Modulus (101Pa)

0.0 0.5 2.0 5.0 18.0 30.0 40.0

1.80 5.95 4.90 6.20 6.60 7.30 8.00

1.68 6.00 4.62 6.14 6.50 7.35 8.10

1.68t 3.04 2.76 3.65 3.79 4.35 4.70

4.20-4.50~: 1.97-1.95 1.67-1.77 1.69-1.70 1.72-1.74 1.69-1.68 1.73-1.72

2.2 2.65 2.55 2.75 2.80 3.00 3.25

0.47-0.47 0.33-0.32 0.22-0.27 0.23-0.24 0.25-0.25 0.23-0.23 0.25-0.24

0.57-0.67 6.28-6.09 2.85-3.53 5.48-5.69 6.47-6.83 8.64-8.42 11.80-11.20

0.004 2.45 1.94 3.66 4.02 5.68 7.18

* Depth to top of layer. t Held fixed after first iteration; S converts to P at the base of this layer. :~ Assumes a value of 0.4 km/sec for S.

Velocity inversions for P and S. Well recorded earthquakes from the New Madrid area were then selected for a velocity inversion study. The technique used was that developed by Crosson (1976) for simultaneous inversion of both velocities and earthquake hypocenters. Unlike previous studies of this type, velocity structures for both P and S were determined with no assumptions made regarding the similarity of the S velocity to that of the P velocity except for position of layer boundaries. Layer boundaries were those chosen by Mooney et al. (1982) for velocity discontinuities in the area of the portable array. Details of the inversion procedure, as well as its tectonic implications are discussed by Nicholson et al. (1984b). The results are summarized in Table 1. Also shown are the interval values of Vp/Vs derived from the two inversion studies and the P-wave velocity structure determined from the seismic refraction experiment (Mooney et al., 1982). Both inversions, whether for P or S, proved to be critically dependent on the initial velocity model used to start the inversion procedure. Several trials were made and, in each case, the attempt that converged rapidly and had the greatest stability during the inversion process was the one that used a starting model that closely matched the first-order velocity structure identified by the half-space velocity

CHANGES IN Vp/V~WITH DEPTH

1119

technique (Figure 10). In the case of the P-wave inversion, inclusion of a lowvelocity layer resulted in an 83 per cent reduction in variance and produced station corrections that were less extreme and showed a much more coherent spatial pattern than any of the other attempts. During the S-wave inversion, the velocity in the uppermost layer tended toward that of the P wave. This was not unreasonable since this layer is primarily composed of saturated unconsolidated sediments, the surface velocity for the S wave is known to be low, and the conversion coefficient from S to P is subsequently'high. More importantly, observations from horizontal seismometers operated within the embayment show that SH arrives later than what would normally be identified as S on the vertical component (Andrews and Mooney, 1983), suggesting that t h e latter phase is Sv converting to P. Consequently, the velocity in the uppermost layer was held fixed at the velocity previously determined from the P-wave inversion. The total reduction in variance between the final S-wave model and the one used to initially locate the earthquakes was 98 per cent, indicating that the high residuals commonly observed for S arrivals are more likely a result if improper modeling rather than inherent scatter in the S-wave data. The fact that the inversion process proved stable, converged rapidly, and achieved a considerable reduction in variance implied that the S wave was indeed converting to P at the near-surface, and that the boundary for conversion was consistently the base of the unconsolidated sediments. Weighted moving-window averages of the final models for P and S accounting for averages over upgoing rays are shown in Figure 10 as the fine dotted lines. The correlation with earlier half-space values is readily apparent. Furthermore, relocation of the microearthquakes using a variable V;/Vs model resulted in a clearer hypocentral distribution that showed greater spatial coherence and was in better agreement with regional geology and fault plane solutions. Estimates of general material properties. Besides improving earthquake locations, analysis of the data for both P and S velocity structures allows for the calculation of related bulk properties and elastic constants. Table 1 shows crustal parameters based on the velocities determined from both the New Madrid inversion study and the refraction experiment, as well as density values derived from a regional gravity profile (Mooney et al., 1982). Although no estimates of the surface shear velocity were available from the earthquake data, a value of about 400 m/sec was determined during the seismic refraction experiment (W. Mooney, personal communication, 1981). This value, combined with the measured P velocity, implies the uppermost layer exhibits low P and S velocities, high Poisson's ratio, and low shear strength consistent with its bulk composition as unconsolidated saturated sand and clay. On the other hand, layer 2 exhibits both a high Vp/Vs and relatively high P and S velocities. A highly fractured carbonate sequence would tend to produce such characteristics. A test well drilled near the portable array confirmed that this layer was indeed composed of carbonates and was both highly fractured and highly saturated, with unusually high pore pressure (Crone and Russ, 1979). The high pore pressures would reduce the effective stress levels and insure that cracks remain open, as required to produce the observed high-velocity ratio. Vp/V~ can, therefore, help to discriminate lithologies (Tatham, 1982). For instance, the low velocity for the P wave found for layer 3 could be produced by a high-velocity saturated material with high crack density. However, the low velocity combined with the observed low-velocity ratio implies that this material is likely to have low crack densities, low intrinsic velocities, and possibly higher effective stress

1120

CRAIG NICHOLSON AND DAVID W. SIMPSON

levels. Both the P-wave velocity and the velocity ratio are thus consistent with this layer being composed of early Paleozoic syn-rift sediments (marine shales, siltstones, and sandstones), as proposed by Mooney et al. (1982). It is interesting to note that the microearthquake activity is generally absent in the upper few kilometers. The upper crust is presumably too weak (e.g., layer 1) and/or at insufficient stress levels (e.g., layer 2) to fail by brittle fracture. The earthquakes only occur well into the layers exhibiting lower Vv/Vs, indicative of higher effective stress levels and higher strength. In fact, the elastic moduli of layer 4, in which most of the earthquakes occur, are fairly typical of those assumed for the properties of rocks at depths involved in the brittle deformation associated with earthquake faulting. DISCUSSION In all of the cases studied, ts/tv initially decreases with depth. This implies a similar variation in V;/V~ and requires separate velocity structures for P and S if accurate hypocenters are to be achieved. Similar effects have been well-documented elsewhere, both in the analysis of near-field strong motion records (Archuleta, 1982; Cramer and Shakal, 1983) and in direct borehole measurements of S and P velocities (e.g., Zoback and Hickman, 1982; Moos and Zoback, 1984). Initial decreases in V J Vs with depth have also been observed in seismic velocity experiments conducted through young ocean crust (Bratt and Solomon, 1984). Figure 11 is a compilation from several sources of in situ VJV, measured in deep boreholes, inferred from borehole data, or derived from long-baseline refraction experiments. In all cases, V;/V~ decreases rapidly in the first few kilometers, matching the observed decrease in t,/tp from microearthquake travel times. The high degree of correlation strongly suggests that these techniques are detecting the same first-order spatial variation in velocity structure; a variation that reflects certain common material properties of the upper crust. Both laboratory and theoretical studies predict a decrease in Vv/V, with the closing of saturated microcracks under increasing confining pressure (Nur and Simmons, 1969; O'Connell and Budiansky, 1974). In situ measurements of P and S velocities show that at low effective stress levels, high Vv/V, ratios correlate with high crack densities (Moos and Zoback, 1983). The microearthquake data and the data from borehole studies thus imply that high Vv/V~ , high crack densities, the presence of pore fluids, and low effective stress levels are a fairly ubiquitous feature of the near-surface regime. A more important result, however, is that many of these properties can and do extend to depths of several kilometers. Inversion of the data from New Madrid for both P- and S-wave velocities confirmed the presence of a low-velocity zone; a high-velocity lid with high Vv/Vs and thus high crack density, and a boundary at the near-surface for S to P conversion. The results from southern California also exhibit a low-velocity zone at mid-crustal depths, only in this case, the layer was confined to a region under the San Bernardino Mountains, or localized to the San Jacinto fault. Conversely, the data from Toktogul showed no strong evidence for any low-velocity mid-crustal layer. These simple methods thus serve as useful exploration tools in areas where mid-crustal information is lacking or surface extrapolations are suspect. It is tempting to infer that values of ts/tv f r o m a Wadati diagram represent values of Vv/V~. Unfortunately, this correspondence is not exact as S and P take different ray paths. Other complications arise when the secondary arrival is not a true S wave, as is often the case with vertical component seismographs most commonly

CHANGES IN Vp/V, WITH DEPTH

1121

used in microearthquake studies (Kanasewich et al., 1973). S can convert to P, giving rise to a faster apparent S-wave velocity and a lower ts/tp. Care must therefore be taken in the interpretation of tdt v values. Linearity of the Wadati diagram does not preclude variations in Vp/Vs, nor does an unusually low value of tJtp indicate the presence of a material at depth with unusual properties. Use of regional determinations o f Vv/Vs (e.g., Smith, 1983) may not then be the most appropriate means o f modeling S waves, since they may only reflect the average Vp/Vs as determined by the mean focal depth of the most abundant earthquakes. But if
Velocity Ratio (Vp/Vs) 1.4
I I

1.8
I

2.2
I P

2.6
I J

3.0
I i

3.4
~ I-

3.8
,

4.2

A
/

/ I1'
/11
r ........

E-,
.

:5:A:

&

I'l/
s
CONCLUSIONS

FIG. 11. Vp/V~ measured in situ from boreholes, inferred from borehole data or derived from refraction studies: (A) Michigan basin (Stewart et al., 1981); (B) Imperial Valley, California (Archuleta, 1982); (C) Tadjikistan, U.S.S.R. (after Kulagin and Nikitina, 1968); (D) and (E) Tokyo, Japan area (Ohta et al., 1980); (F) Gulf Cost, Texas (Lash, 1980); (G) San Andreas, central California (Malin et al., 1981).

secondary arrivals are properly modeled (as in the case of the S to P converted waves), increased resolution and improved earthquake locations can be achieved.

Travel times from microearthquakes exhibit ts/t; ratios that decrease with focal depth. This decrease is most rapid in the upper few kilometers and matches similar variations observed in situ for the velocity ratio. Separate velocity structures for P and S are therefore necessary if accurate and realistic earth models are to be determined. If separate models for P and S are used, relocated earthquakes are

1122

CRAIG NICHOLSON AND DAVID W. SIMPSON

generally deeper, are less likely to exhibit model-dependent errors, and allow greater resolution of shallow seismicity patterns. The techniques of Wadati and Riznichenko serve a number of uses in the routine processing of microearthquakes locations: hypocenters can be checked for proper convergence; simple earth models can be developed; spatial variations in ts/tp can be mapped; focal mechanism solutions improved; and regional variations in velocity can be identified. Although their most common application has been to resolve the trade-off in origin time and focal depth, we find these methods have proven useful in a more fundamental sense of identifying significant earth structure and improving our general understanding of material properties of the upper crust.
ACKNOWLEDGMENTS This paper originated from conversations with I. L. Nersesov. We thank Lynn Sykes, Chris Scholz, Keith Evans, and Dan Moos for their critical and substantial reviews of the manuscript. Support for the Soviet network was conducted under the auspices of the United States-USSR exchange program on earthquake prediction and Contracts USGS 14-08-0001-21278 and 14-08-0001-21869. Data from the Central United States was acquired in cooperation with St. Louis University and the Tennessee Earthquake Information Center under Contract USGS 14-08-0001-16794. We gratefully acknowledge the generosity of the California Institute of Technology and the U.S. Geological Survey for the use of the data from SCARLET.

REFERENCES
Andrews, M. C. and W. D. Mooney (1983). Relocation of microearthquakes from the Mississippi embayment using converted phases (abstract), EOS, Trans. Am. Geophys. Union 64, 767-768. Archuleta, R. J. {1982). Analysis of near source static and dynamic measurements from the 1979 Imperial Valley earthquake, Bull. Seism. Soc. Am. 72, 1927-1956. Bratt, S. R. and S. C. Solomon (1984). Compressional and shear wave structure of the East Pacific Rise at 1120'N: constraints from three-component ocean-bottom seismometer data, J. Geophys. Res. 89, 6095-6110. Buland, R. (1976). The mechanics of locating earthquakes, Bull. Seism. Soc. Am. 66, 173-187. Bune, V. I., M. V. Gzovskiy, et al. (1960). Methods for a detailed study of seismicity, Izv. Acad. Sci. USSR, Trudy, 9, no. 176, 327 pp. (in Russian). Cramer, C. H. and A. F. Shakal (1983). Discrepancies in strong motion and sensitive arrival time data from the ML 6.7 Coalinga earthquake of 2 May 1983, in The 1983 Coalinga, California Earthquakes, Special Publ. 66, J. H. Bennett and R. W. Sherburne, Editors, Calif. Div. Mines Geol., Sacramento, California 307-320. Crone, A. J. and D. P. Russ (1979). Preliminary report on an exploratory drill hole--New Madrid test well 1-X in southeast Missouri, U.S. Geol. Surv., Open-File Rept. 79-1216, 12 pp. Crosson, R. S. (1976). Crustal structure modeling of earthquake data. 1. Simultaneous least squares estimation of hypocenter and velocity parameters, J. Geophys. Res. 81, 3036-3054. Feng, R. and T. V. McEvilly (1983). Interpretation of seismic reflection profiling data for the structure of the San Andreas fault zone, Bull. Seism. Soc. Am. 73, 1701-1720. Green, C. H. (1938). Velocity determinations by means of reflection profiles, Geophysics 3,295-305. Gutenberg, B. (1951). Revised travel times in southern California, Bull. Seism. Soc. Am. 41,143-164. Hales, A. L., K. J. Muirhead, and L. Maki-Lopez (1981). The times of origin and depths of focus of intermediate and deep focus earthquakes: model calculations, Bull. Seism. Soc. Am. 71, 1539-1552. Healy, J. and L. Peake (1975). Seismic velocity structure along a section of the San Andreas fault near Bear Valley, California, Bull. Seism. Soc. Am. 65, 1177-1197. Hearn, T. M. and R. W. Clayton (1984). Crustal structure and tectonics in southern California (abstract), EOS, Trans. Am. Geophys. Union 65,992. James, D., I. S. Sacks, E. Lazo, and P. Aparicio G. (1969). On locating local earthquakes using small networks, Bull. Seism. Soc. Am. 59, 1201-1212. Kanasewich, E. R., T. Alpaslan, and F. Hron (1973). The importance of S wave precursors in shear wave studies, Bull. Seism. Soc. Am. 63, 2167-2176. Kisslinger, C. and E. R. Engdahl (1973). The interpretation of the Wadati diagram with relaxed assumptions, Bull. Seism. Soc. Am. 63, 1723-1736. Klein, F. W. (1978). Hypocenter location program: HYPOINVERSE, U.S. Geol. Surv., Open-File Rept. 78-694, 101 pp.

CHANGES IN Vp/V~ WITH DEPTH

1123

Kulagin, V. K. and S. V. Nikitina (1968). The variation of the ratio of velocities of body waves in the earth's crust, in Deep Structure and Earthquakes of Tadzhikistan, T. I. Kukhtikova, Editor, Acad. Sci. Tadzhik SSR, Donish, Dushanbe, 5-46 (in Russian). Lash, C. C. (1980). Shear waves, multiple reflections, and converted waves found by a deep vertical wave test (vertical seismic profiling), Geophysics 45, 1373-1411. Malin, P. E., M. H. Gillespie, P. C. Leafy, and T. L. Henyey (1981). Crustal structure near Palmdale, California, from borehole-determined ray parameters, Bull. Seism. Soc. Am. 71, 1783-1804. Mooney, W. D. and R. H. Colburn (1985). A seismic-refraction profile across the San Andreas, Sargent, and Calaveras faults, west-central California, Bull. Seism. Soc. Am. 75, 175-191. Mooney, W. D., M. C. Andrews, A. Ginzburg, D. A. Peters, and R. M. Hamilton (1982). Crustal structure of the northern Mississippi Embayment and a comparison with other continental rift zones, Tectonophysics 94, 327-348. Moos, D. and M. D. Zoback (1983). In situ studies of seismic velocity in fractured crystalline rocks, J. Geophys. Res. 88, 2345-2358. Moos, D. and M. D. Zoback (1984). Comparison of the in situ and laboratory determined pressure dependence of P- and S-velocities (submitted for publication). Moos, D., S. H. Hickman, and M. D. Zoback (1983). Sonic velocity and fracture patterns or distribution in a well drilled through the Cleveland Hills fault, Oroville, California (abstract), EOS, Trans. Am. Geophys. Union 64, 834. Nicholson, C., L. Seeber, P. Williams, and L. R. Sykes (1984a). Seismotectonics of the Eastern Transverse ranges: block rotations and shallow-angle thrusts (abstract), EOS, Trans. Am. Geophys. Union 65, 285. Nicholson, C., D. W. Simpson, S. Singh, and J. E. Zollweg (1984b). Crustal studies, velocity inversions and fault tectonics: results from a microearthquake survey in the New Madrid Seismic Zone, J. Geophys. Res. 89, 4545-4558. Nut, A. and G. Simmons (1969). The effect of saturation on velocity in low porosity rocks, Earth Planet. Sci. Letters 7, 183-193. O'Connell, R. J. and B. Budiansky (1974). Seismic velocities in dry and saturated cracked solids, J. Geophys. Res. 79, 5412-5426. Ohta, Y., N. Goto, F. Yamamizu, and H. Takahashi (1980). S-wave velocity measurements in deep soil deposit and bedrock by means of an elaborated down-hole method, Bull. Seisra. Soc. Am. 70, 363378. Riznichenko, Y. V. (1958). Methods for large-scale determination of focus coordinates of nearby earthquakes and velocities of seismic waves in the focal region, Tr. Inst. Fiz. Zemli. Akad. Nauk SSR. 4. Simpson, D. W., M. W. Hamberger, V. D. Pavlov, and I. L. Nersesov (1981). Tectonics and seismicity of the Toktogul reservoir region, Kirgizia, U.S.S.R., J. Geophys. Res. 86, 345-358. Smith, E. G. C. (1983). Joint determinations of seismic velocity ratios: theory and application to an aftershock sequence, Bull. Seism. Soc. Am. 73,405-418. Stewart, R. R., R. M. Turpening, and M. N. ToksSz (1981). Study of a subsurface fracture zone by vertical seismic profiling, Geophys. Res. Letters 8, 1132-1135. Tatham, R. H. (1982). Vp/Vs and lithology, Geophysics 47, 336-344. Toth, T. and C. Kisslinger (1984). Revised focal depths and velocity models for local earthquakes in the Adak seismic zone, Bull. Seism. Soc. Am. 74, 1349-1360. Wadati, K. (1933). On the travel time of earthquake waves, II, Geophys. Mag. 7, 101-111. Zoback, M. D. and S. Hickman (1982). In situ study of the physical mechanisms controlling induced seismicity at Monticello Reservoir, South Carolina, J. Geophys. Res. 87, 6959-6974. LAMONT-DOHERTYGEOLOGICALOBSERVATORY OF COLUMBIAUNIVERSITY PALISADES,NEW YORK 10964 CONTRIBUTIONNO. 3796 Manuscript received 26 September 1984

Вам также может понравиться