PL-TR-91-2192 - "

AD-A246 894* ODELING REGIONAL SEISMIC WAVES III//IlIIII/I/l/iIIil

Donald V. HelubergerDavid G. Harkrider

California Institute of TechnologySeismological Laboratory DTGPasadena, CA 91125

31 July 1991

Final Report24 March 1989-31 July 1991

APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED

92-01224

PHILLIPS LABORATORYAIR FORCE SYSTEMS COMMANDHANSCOM AIR FORCE BASE, NASSACHUSETTS01731-5000

OR 1 o4

SPONSORED BYDefense Advanced Research Projects Agency

Nuclear Monitoring Research OfficeARPA ORDER NO. 5307

MONITORED BYPhillips Laboratory

Contract FY19628-89-K-0028

The views and conclusions contained in this document are those ofthe authors and should not be interpreted as representing theofficial policies, either expressed or implied, of the DefenseAdvanced Research Projects Agency or the U.S. Government.

This technical report has been reviewed and is approved forpublication.

J16E,8F.LETKOWCES F. LEWK(DWICZ /

opt ract Manager B as chIChiefid Earth Geophysics Branch .,'olid Earth Geophysics Branch

Earth Sciences Division Earth Sciences Division

DONALD H. ECKHARDT, DirectorEarth Sciences Division

This report has been reviewed by the ESD Public Affairs Office(PA) and is releasable to the National Technical InformationService (NTIS).

Qualified requestors may obtain additional copies from the DefenseTechnical Information Center. All others should apply to theNational Technical Information Service.

If your address has changed, or if you wish to be removed from themailing list, or if the addressee is no longer employed by yourorganization, please notify PL/IMA, Fanscom AFB, MA 01731-5000.This will assist us in maintaining a current mailing list.

Do not return copies of this report unless contractual obligationsor notices on a specific document requires that it be returned.

UNCLASSI FlED

SECURITY CLASSIFICATION OF THIS PAGE

REPORT DOCUMENTATION PAGEIs REPORT SECURITY CLASSIFICATION lb. RESTRICTIVE MARKINGS

UNCLASSIFIED2e. SECURITY CLASSIFICATION AUTHORITY 3. DISTRIBUTION/AVAILABILITY OF REPORT

Approved for public release: distribution2b. DECLASSIFICATION/DOWNGRADING SCHEDULE tinl imi ted

4. PERFORMING ORGANIZATION REPORT NUMBEAtS) 5. MONITORING ORGANIZATION REPORT NUMBER(S)

PL-TR-91-2192

6a. NAME OF PERFORMING ORGANIZATION b. OFFICE SYMBOL 7&. NAME OF MONITORING ORGANIZATION

California Insitute of Tech. If apptiiablej

Seismoloicical Laboratory Phillips Laboratory

6c. ADDRESS (City. State and ZIP Code 7b. ADDRESS (City. Slate and ZIP Code)

P,!;:Idena, C:il iforni;i c1 12l,('om Air Itor(- Base, M;:S:clhit.,.tt.;01731-5000

as NAME OF FUNDING/SPONSORING Bb. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBERORGANIZATION Defense (if applicable)Advanced Research Projects Ag ncy NMRO F]9628-89-K-0028

Sc ADDRESS IC#tI,. State and ZIP Code) 10. SOURCE OF FUNDING NOS.

1400 1W!ilson Boulevard PROGRAM PROJECT TASK WORK UNITArlington, Virginia 22209-2308 ELEMENTNO. NO. NO.

NO

11 TITLE 'Include Security C"ification, Modeling Regional 61101E qA10 DA ADSeismic W ve._

12. PERSONAL AUTHOR(SI

Donald V. Helmberger and David G. flarkrider13. TYPE OF REPORT 13b. TIME COVERED 14. DATE OF REPORT (Yr., Mo., Day) 15. PAGE COUNT

Final FROM 3./. 9 TO 331 'I July 1991 0816. SUPPLEMENTARY NOTATION

17 COSATI CODES 18. SUBJECT TERMS (Continue on muere if neceuary and identify by block number)FIELD GROUP SUB CR. M,; regional surface wave magnitudes; broad-band modeling regional

seismograms; earthquake relocation Tibet.

19. ABSTRACTThe research performed under the contract, during the period 1 August 1989 through 31 July 1991. can be divided

into three main topics; modeling regional broad-band seismograms from the Imperial Valley to Pasadena. determiningsurface wave magnitudes for NTS events using regional data. and waveform modeling to determine the depth of earthquakesin Tibet leading to new origin Limes and a better estimation of Pn and Sn velocities.

In section 1. we address broad-band wave propagation along a corridor from Imperial Valley to Pasadena.California. The path consists of 50 km of slow basin structure with a shallow moho followed by 250 km of relativelynormal Southern California structure approaching Pasadena. Events occurring in the valley produce extended body wavecodas which carry a dispersive waveform imprint along with longer period Rayleigh waves. Many features of theseseismograms at period greater than a few seconds can be modeled by applying a finite-difference technique to a 2D structure.Shorter periods display complex behaviors not easily modeled by present techniques, especially for shallow events.Events occurring in the normal structure produce broad-band SH waveforms that can be modeled, analytically, if we includesome extra empirically derived parameters determined by calibration events. These parameters are determined by adjustingtravel time differentials between the decomposed wavefield S, sS, etc. For example, delays in sS relative to S correspond tolocal corrections for sediment cover for a particular source region. Asperities and directivity become increasinglyimportant at shorter periods, as demonstrated by numerical experiments and observation. Numerical experiments conductedon 2D models with embedded scatterers appear to explain coda development and some of the properties of the Lg phase.

20 DISTRI BUTION, AVAILABILITY OF ABSTRACT 21. ABSTRACT SECURITY CLASSIFICATION

,,r_ASSIFIED'UNLIMITED Z SAME AS RPT. 0 DTIC USERS 0 UNCLASS[FIED

2, .,,E OF qESF0',.t'BLE NDI DUAL 22b TELEPHONE NUMBER 22c OFFICE SYMBOL(include .4 rea Code)

S.' .'r " -" (617) 377- 02( ",./l,'h

DD FORM 1473, 83 APR EDITION OF 1 JAN 73 IS OBSOLETE UNCLASSIFIEDI SECURITY CLASSIFICATION OF THIS PAGE

UNCLASSIFIED

SECURITY CLASSIFICATION OF THIS PAGE

In section 2, we calculate surface wave magnitudes for 112 Nevada Test Site (NTS) explosions from a data set ofregional long-period scismograms from North American stations. In order to utilize the nearer regional stations (A

FINAL TECHNICAL REPORT24 may 1989 -31 July 1991

ARPA Order No.:

Name of Contractor: California Institute of Technology

Effective Date of Contract: 24 March 1989

Contract Expiration Date: 31 July 1991

Contract Number: F19628-89-K-0028

Principal Inveestigators: Donald V. Ilelmberger(818356-6998

David G. Harkrider(818)356-6910

Program Manager: James F. Lewkowicz(617)861-3028

Short Title of Work: Modeling Regional Seismic Waves

The views and conclusions contained in this document are those of the authorsand should not be interpreted as necessarily representing the official policies,either expressed or implied, of the Defense Advanced Research Projects Agencyor the U.S. Government

Sponsored byDefense Advanced Research Projects Agency (DoD)

Nuclear Monitoring Research OfficeARPA Order No.

Issued by the Air Force Geophysics Laboratory underContract#FI 9628-89-K-(X)28

Seismological Laboratory OOSSlo NiDivision of Geological and Planetary Sciences UTIS- R A&I

California Institute of Technology DTIC TAR 0Pasadena, California 91125 Unanounce4 0

Justif ieat lot .

By - _

PI st r 1but I,al

-- ' eb/i

TABLE OF CONTENTS

PageNumbers

Summary_ V

1. Broad-band modeling of regional seismogramsImperial Valley to Pasadena ......

2. Determining Ms Magnitudes from Regional NTS Data 49

3. A Note on the Relocation of Tibetan earthquakes 81

Iv

Summary

The research performed under the contract, during the period I August 1989through 31 July 1991, can be divided into three main topics; modeling regionalbroad-band seismograms from the Imperial Valley to Pasadena, determining surfacewave magnitudes for NTS events using regional data, and waveform modeling todetermine the depth of earthquakes in Tibet leading to new origin times and a betterestimation of P. and Sn velocities.

In section 1, we address broad-band wave propagation along a corridorfrom Imperial Valley to Pasadena, California. The path consists of 50 km of slowbasin structure with a shallow moho followed by 250 km of relatively normalSouthern California structure approaching Pasadena. Events occurring in the valleyproduce extended body wave codas which carry a dispersive wavefonu imprintalong with longer period Rayleigh waves. Many features of these seismograms atperiod greater than a few seconds can be modeled by applying a finite-differencetechnique to a 2D structure. Shorter periods display complex behaviors not easilymodeled by present techniques, especially for shallow events. Events occurring inthe normal structure produce broad-band SH waveforms that can be modeled,analytically, if we include some extra empirically derived parameters determined bycalibration events. These parameters are determined by adjusting travel timedifferentials between the decomposed wavefield S, sS, etc. For example, delays insS relative to S correspond to local corrections for sediment cover for a particularsource region. Asperities and directivity become increasingly important at shorterperiods, as demonstrated by numerical experiments and observation. Numericalexperiments conducted on 2D models with embedded scatterers appear to explaincoda development and some of the properties of the Lg phase.

In section 2, we calculate surface wave magnitudes for 112 Nevada TestSite (NTS) explosions from a data set of regional long-period seismograms fromNorth American stations. In order to utilize the nearer regional stations (A

from a relatively cool crust which would allow the seismo-genic zone to extend tosuch depths. A detailed investigation of the Tibetan earthquakes, with magnitudegreater than 5.5 from 1964 to 1986, yields a distinctly different picture. Waveformmodeling of depth phases such aspP indicates that only three or four events fromthis population is actually deeper than 25 kn. These few events occur near theedges of the Plateau where active subduction is occurring as suggested by thethrust-like nature of their mechanisms. The events, averaging the entire population,occurred earlier than indicated by the ISC by about 3 seconds which leads to abouta 1.5% and 0.5% over estimation of Pn and Sn velocities respectively applying ISCtables and standard flat-layered models. A more serious error occurs if the Pn andSn velocities are determined by correcting for source depth but assuming the ISCorigin times.

VI

SECTION 1

Broad-band modeling of regional seismograms

Imperial Valley to Pasadena

Broad-band modeling of regional seismograms;Imperial Valley to Pasadena

D. Helmberger, R. Stead, Phyllis Ho-Liu and D. Dreger

Abstract

This paper addresses broad-band wave propagation along a

corridor from Imperial Valley to Pasadena, California. The path

consists of 50 km of slow basin structure with a shallow moho

followed by 250 km of relatively normal Southern California

structure approaching Pasadena. Events occurring in the valley

produce extended body wave codas which carry a dispersive

waveform imprint along with longer period Rayleigh waves. Many

features of these seismograms at periods greater than a few seconds

can be modeled by applying a finite-difference technique to a 2D

structure. Shorter periods display complex behaviors not easily

modeled by present techniques, especially for shallow events.

Events occurring in the normal structure produce broad-band SIt

waveforms that can be modeled, analytically, if we include some

extra empirically derived parameters determined by calibration

events. These parameters are determined by adjusting travel time

differentials between the decomposed wavefield S, sS, etc. For

example, delays in sS relative to S correspond to local corrections for

sediment cover for a particular source region. Asperities and

directivity become increasingly important at shorter periods, as

demonstrated by numerical experiments and observation. Numerical

experiments conducted on 2D models with embedded scatterers

3

appear to explain coda development and some of the properties of

the Lg phase.

Introduction

One of the fundamental reasons why quantitative seismology

and waveform modeling is contributing to source retrieval and to the

monitoring of nuclear weapons is its prediction capability. That is,

given a set of seismograms we can correct for propagational effects

and predict the nature of the source. This is relatively easy when

dealing with teleseismic body waves and well dispersed long period

surface waves. But since small events cannot be seen teleseismically,

we must address the more difficult problem of source retrieval from

regional phases.

An intermediate stage in this development is to examine broad

band records, BB, at regional distances for events large enough to be

well-recorded teleseismically. This feature allows the source

parameters to be determined by modeling teleseismic bodywave

phases, direct P and reflected phases (e.g. pP, etc.). Starting with a

known source allows detailed studies of regional seismograms in

terms of path effects and the development of BB Green's functions.

Hopefully, these Green's functions can be used to estimate smaller

events in the general vicinity of the larger event or master event.

Such a procedure has been used earlier to calibrate surface wave

paths, Romanowicz (1982). Unfortunately, there are not many BB

recordings of significant events presently available. Some results

from detailed studies of these existing records are encouraging while

others are not.

4

An example of the former is given in a paper by Zhao and

Helmberger (1991), see figure 1, for a path from Quebec to Harvard.

Their strategy for modeling these records was to break the

seismograms into segments Pnl, containing Pn, pPn, sPn, PmP, P,

coupled PL waves; Snl, containing Sn, sSn, SmS, (etc.); and the

fundamental Rayleigh waves. Synthetics are generated for the

various segments with generalized rays, normal modes and

reflectivity methods where the advantages of each technique are

exploited. Information about the upper crust is obtained from the

fundamental modes and crustal thickness and velocity gradients in

the mantle from Pnl and SnI waves.

It is rather remarkable that a flat layered model (see Table 1)

can explain so many features. In fact, a simple layer over a half

space does a good job for the first 100 seconds of long period

motions.

These results suggest that in stable regions where Sn propagates

to large distances, the possibility of using sSn to help fix source

depths could prove useful. Note that Snl and sSn are considerably

larger than Pn and PPn and can probably be seen for smaller events.

The usefulness of the interference produce by Pn and PPn has

already been demonstrated by Burdick, Saikia and Smith (1991) for

explosions.

Explaining the jitter in the BB observations displayed in figure 1,

(assuming a flat-layered model), has proven difficult. Thus, it

appears appropriate at this stage to introduce scattering into the

synthetics as proposed by Flatte and Wu (1988), Kennett (1989) and

others. Note that this method supposes that a first-order model

5

already exists. We will show that establishing these models is no

easy task, especially in tectonic regions, as discussed for a BB dataset

for a corridor from Imperial Valley to Pasadena. Complexity

abounds along this profile and the short period modeling results are

not so encouraging.

Observations

The observations addressed in this study consist of a collection

of low-gain short period records from moderate sized events along

the San Jacinto fault system which have known source mechanisms,

and BB observations of recent smaller events. Figure 2 gives the

locations of these events which range (roughly) from 170 to 270 km.

The more distance events are beneath the Imperial Valley. Table 2

list some of the event information, source depth, etc, These events

are all strike-slip in nature and have been studied in detail by

several authors: for example see Bent and Helmberger (1991).

One of the reasons for working with the Pasadena data is the

presence of a recently installed Streckeisen instrument. Example

seismograms from this system are displayed in figures 3, 4, and 5.

The bottom three rows of simulations correspond to three older

types of instruments operated at Pasadena and bracket the range of

frequencies of interest. These three events are located along the

extended San Jacinto fault system and can be used to compare with

the older well-studied large events such as the events located in

figure 2.

Since the azimuth towards Pasadena is near the (P-SV) node

assuming a strike-slip orientation, we would expect considerable

6

variation on the vertical component which is apparently observed,

(see figures 4 and 5), where the P-waves appear to be of opposite

polarity. The tangential motions are similar as expected along this

azimuth. Recordings from the aftershock of the Superstition Hills

event, see (c) in figure 2, display distinctly different properties from

the first two sets. Note that the BB tangential motions are longer

period and the ratio of wa.lp-to-wa.sp period torsion responses

increase from (3) to (5). Stead (1989) compares these ratios for 20

events in the region and finds this same change on average. Another

difference is the enhanced strengths of surface waves for basin

events and their relatively slow velocity relative to to the

bodywaves.

Note that as the Rayleigh waves become stronger, the SV waves

become weaker. These same general features can be seen in the

older records as displayed in figure 6. The wa.sp records are from

the large events and the wa.lp is from a smaller aftershock as

indicated in Table 2. Comparing the wa.sp waveforms for a large

event with a wa.lp waveforms of a aftershock or smaller event

generally shows good agreement as displayed in figure 7. This

similarity is produced largely by the differences in faulting area

where the distributed nature of the main event tends to produce an

integrated effect. This subject becomes interesting because of

directivity and the sensitivity of the wa.sp waveforms to fault

asperities. We will address these issues in this paper, along with the

development of depth phases and the effects of random scatterers in

the crustal waveguide.

7

Modeling events in the Imperial Valley

We have not had much success in modeling Imperial Valley

events with uniform layers, see Ho-Liu and HeImberger (1989).

Finite-difference modeling, however, appears to explain some of the

observed long period features, see Ho-Liu and Helmberger (1989).

Their model is displayed in figure 8 and is based on studying a

profile of long-period (30-90) observations along this same corridor.

This model allows events occurring in the basin to have prolonged

wavetrains at the Pasadena station, PAS while events north of the

boundary to have rather sharp pulses, see figure 2.

Synthetics for basin structures can be generated by finite-

difference methods following the approach discussed by Helmberger

and Vidale (1988). The basic technique is based on the expansion of

the complete three-dimensional solution of the wave equation in

cylindrical coordinates in an asymptotic form which provides for the

separation of the motions into SH and P-SV systems. Closed form

expressions appropriate for finite-difference source excitation are

obtained from two-dimensional Cagniard-de Hoop theory with rather

elaborate near-field terms, see Vidale and Helmberger (1987). Note

that the synthetics generated by this mapping preserve the (P-SV)

and (SH) systems or two dimension scattering but fail to treat the

SV-SH scattering problem which becomes increasingly important at

the higher frequencies, see figures 3, 4 and 5."

Profiles of synthetics for the geometry displayed in figure 8 are

given in figure 9 for the three pass-bands. The first fifty km of

propagation in the slow Imperial Valley structure causes the rapid

development of dispersion. After the shallow Lovewave leaves the

basin, it speeds up and disperses more slowly. Three component

synthetics generated for a source depth of 12 km appropriate for the

1/28/88 event are given in figure 10. The overall timing looks good

but the shorter-periods are not matched well, especially on the

tangential component.

Comparing these synthetics with the records of event I of the

1987 sequence, trace 4 of figure 6, displays about the same degree of

fits as discussed earlier, see Bent et al.(1989) where this event was

studied in detail.

Results for the Westmoreland event are given in figure 11.

These results are not bad, but the short period complexity becomes

very severe for shallower events, as displayed in figure 12. Some

progress in explaining the long period motion for very shallow

events is given in Ho-Liu and Helmberger (1989) where very slow

surface waves in the sediments can be scattered into the coda by

valley edge effects. However, generating BB synthetics for these

types of models is very expensive relative to analytical techniques

and therefore, we will address events occurring to the North of the

Imperial Valley in the remaining sections assuming more

conventional layered models.

Waveform modeling north of the Imperial Valley.

In this section, we will discuss the analytical decomposition of

the wavefield in terms of generalized rays, see Helmberger (1983)

for a discussion of the method employed. This method allows

separation of the wavefield into energy that starts downward

relative to upward and isolation of depth phases such as sS. The

9

latter phase (sS) has potential for source discrimination since

explosions are obviously shallow.

First, we examine the behavior of a smooth model followed by

that of a coarsely layered model with fewer parameters and discuss

their properties when dipping layers are allowed. A particularly

effective wavefield decomposition is presented in terms of travel

paths: namely,

1. Direct arrival plus surface layer multiples (shallow Love

waves)

2. Diving energy paths (lower crustal triplications)

3. Surface reflected paths which turn below the source (sS)

These latter two wave-packets are particularly affected by vertical

directivity, as demonstrated by broadband experiments, and behave

as if they had different time histories. In short, combining the three

contributions with various time shifts which correspond to depth

changes or lateral variation allows some extra freedom. Modeling

observations with these added parameters becomes much easier as

will be demonstrated later.

A velocity model composed of approximately 50 layers and

obtained by smoothing a southern California model proposed by

Hadley and Kanamoii (1979), is given in figure 13. Figure 13 also

displays two generalized ray sets used in constructing the wavefield,

namely the down-going ray set and up-going ray set (excluding the

direct arrival). The upper portion of figure 14 displays the various

contributions of these three raysets to the total potential field. These

responses were produced by applying the Cagniard-de Hoop

10

technique to the generalized rays, see Helmberger and Malone

(1975). The upper row shows the loss of short-period energy with

range as the signal becomes diffracted. The down-going rays or

diving rays contribute significantly to the high-frequency content.

The moho reflection and head wave complicate the picture, especially

due to contributions from sS. The interplay between S and sS

changes dramatically when we change mechanism to a dip-slip

because of the polarity shift across the horizontal axis.

This feature is easily seen in some preliminary dipping models

used in modeling waveform data from the 1969 and 1968 events,

see figures 15 and 16. In this situation, we decomposed the rayset

into two groups, those containing the direct ray plus multiples in the

surface layer (Lovewave) and those diving below the source either

directly or following a bounce off the free surface, see figure 15. The

direct arrival and diving rays have the same polarity for strike-slip

events but opposite in the dip-slip case. For this reason the dip-slip

case produces much different looking synthetics as displayed in

figure 15.

A comparison of synthetics generated from this dipping model

are presented in figure 16 where the depth effects are emphasized

by the ray decomposition. Note that in the Coyote Mtn. example, the

source depth is 14 km and the shallow Lovewaves become less

important. For the shallower Borrego Mtn. event we see a stronger

Lovewave contribution. The sSmS phase is especially obvious in the

Coyote Mtn. record.

One of the advantages in working with large events in

developing a model is the constraints on source parameters obtained

11.

from the teleseismic modeling of body waves, The Coyote Mtn. event

appears very simple teleseismically and occurred at the relatively

deep depth of 14 kin, see Bent and HeImberger (1991a). A

disadvantage of large events is seen in the Borrego Mtn. case where

the source contains at least two spikes. This complexity is clear in

the teleseismic results as well, see Ebel and Helmberger (1982).

The dipping model discussed above was derived in conjunction

with a cross-section running from Imperial Valley to Pasadena as

displayed in figure 8. The model predicts synthetic that fit these

waveforms quite well, although it is poorly constrained. In fact,

there are only three principal pulses controlling these waveforms at

this range. They are the shallow Lovewave, SInS and sSmS. Since a

flat-layered model has a broader applicability, we will address such

a model next.

Unfortunately, we do not have a profile of shear wave

observations from a calibrated event and, thus, we must constrain

the model from other sources and keep the parameters to a

minimum. Our preferred model is given in Table 3 and has just five

layers, similar to Hadley and Kanamori (1979). Here the gradient at

the top of the model given in figure 13 was replaced by a single

layer with a thickness of 4 km. This thickness was adopted from a

study of local events recorded at PAS where a single-layered model

fits the three component data very well, see Dreger and HelImberger

(1990). The smooth positive gradient at mid-depths displayed in

figure 13 was replaced by a single layer with velocity 3.6. This is

consistent with studies from NTS to Tucson; see for example,

Langston and Helmberger (1974). The transition zone at the base of

12

the crust was added to reduce the strength of the moho triplication

as suggested by the P-wave data, (Hadley and Kanamori (1979).)

The problem discovered in modeling the waveforms is that of

preventing the moho triplication from completely overwhelming the

relative weak arrival from the nearly constant velocity crust.

Applying the above constraints and varying the thickness to

match the timing and waveshapes produced the model given in Table

3. Synthetics appropriate for this model are displayed in figure 17

along with a decomposition which proves useful in modeling near-

regional data. The first-column shows the development of the Love

wave associated with the surface layer. Note that it also contains the

direct arrival. The second column gives the results contributed by

the down-going arrivals and the third displays the build up of sS. It

is clear that the direct wave plus the surface multiples dominate the

motions at epicentral distances less than 100 km. This feature was

observed and modeled earlier by Heimberger and Malone (1975) for

the Morgan Hill earthquake in central California.

Most of the San Jacinto events occurred near distances at 200

km where all three raysets contribute, sometime destructively and

sometime constructively. Adding random velocity anomalies to the

various crustal layers produces shifts in timing between these

groups, which in turn yields dramatic high frequency effects as

discussed in the next section.

Changing the source depth also has a strong effect on the

excitation of Lovewaves and on the separation in arrival times

between S and sS. Figure 18 displays the depth sensitivity

appropriate for a distance of 205 km, where we show BB, long-period

13

and short period Wood-Anderson synthetics. The phase sS is the

most obvious for deep events where the Lovewaves are the weakest

and less interference occurs.

Application of the decomposed wavefield with "time shifts" is

presented in figure 19. The fits to the data were obtained by shifting

the phase sS back in time by .3 seconds for the 7/02/88 event and .5

seconds for the 5/17/88 event. Although these shifts were applied

as rather arbitrary corrections, they can be pictured as upper crustal

adjustments for these particular locations. For instance, Hamilton

(1970), reports about .4 km of sediments in the Borrego Spring

Valley near the epicenter of the 5/17/88 event. Thus, sS would be

delayed about .5 seconds relative to the diving energy in accordance

with the .5 sec. lag applied. These two events were sufficiently small

to treat as point sources and apply a simple trapezoid time history,

namely btl, 8t2 and 8t3 yielding .2, .4, and .2 seconds for both

events. Their moment estimates are 4.5x10 21 ergs for the 7/02/88

event and sightly smaller for the 5/17/88 earthquake, 4.5x10 21 ergs.

Since we have, presently, only one broad-band instrument we must

treat these moment results as preliminary estimates.

For larger events, we expect considerably more complications

involving both directivity and a slip distribution containing

asperities. Figure 20 displays predicted synthetics for a number of

rupture geometries. A point source synthetic is included for

comparison. The case of forward rupturing event is nearly the same

as the point source if the rupture velocity is near the body velocity

as usually assumed. Rupturing away from the receiver produces a

strong reduction in short period signals, as expected. Downward

14

rupture tends to enhance the moho reflection while rupture upward

strengthens sS. Since both the waveshapes and relative short period

to long period amplitude ratios are so sensitive to rupture direction

it would appear that broad-band data will prove particularly useful

for the purpose of studying rupture direction.

If we scale the motions given in figure 20 to those appropriate

for the 69 event discussed earlier with a Mo=4.8x1024 from Bent

and Helmberger (1991a) we obtain an estimate of 168 cm for the

case of upward rupture, The observed peak motions were 53 cm and

the waveform fit is quite good, see figure 21. A .30% strength of

short period energy release to long period level is about the average

behavior of many California earthquakes, (Bent and Helmberger

(1991b)). However, to avoid biasing this estimate by directivity, we

clearly require multiple samples in different locations, as

demonstrated in figure 20. This will be possible in a few years with

modern instrumentation, at least in California with the installation of

Terrascope.

Numerical Scattering Experiments

The introduction of laterally varying structure causes

considerable interference between the three ray-groups discussed

earlier. This is especially true at the larger ranges where the

summation has about the same amplitude as each individual group.

In this section, we discuss the theoretical results from models

containing random scatterers embedded in the various layers. We

are particularly concerned with the development of high frequency

coda, and with amplitude decay with distance.

15

The model and geometrical setup is displayed in figure 22. We

investigated four situations with variation allowed in particular

layers; namely RMOHO (thin layer at the top of the mantle), RSURF

(surface layer only), RBOTH (bottom transition layer) and RALL

(variations in all the layers). The scattering model follows the

scheme discussed by Frankel and Clayton (1984) where a Gaussian

correlation function with correlation distances (X) in the vertical and

(Y) in the horizontal dimensions is applied. Variations in velocity are

up to 20%. Locations of variation are random and their positions are

indicated in the above figure. Variations of this size have been

suggested by Frankel and Clayton (1984) and others and are

probably on the high side.

The complete wave-field was generated for these models

ranging in distance from 50 to 275 km at intervals of 5 km with

frequencies up to about 2 hz. We will concentrate primarily on the

cross-over ranges as displayed in figure 23 for the SH field. The

synthetics displayed on the left are difficult to distinguish from the

synthetics discussed earlier from the homogeneous model, see figure

17. The next three columns of the above figure shows the effects of

adding random scatterers to the various crustal layers. Scatterers in

the top layer show a much stronger effect than do those in the

bottom crustal layer (3.75 km/sec) since the second and third

columns are nearly the same. Adding in scatterers in the main

crustal layer (3.6 km/sec) causes the strongest distortions. The

relative amplitude variation across the various rows reflects

primarily the velocity at the receiver which is the same for the three

columns on the right.

16

It becomes difficult to interpret these complex waveforms since

we can no longer decompose the waveform into subgroups, though

we can still identify some of the more important phases such as SmS

and sSmS. A brief review of figure 17 indicates that sSmS becomes

strong near 200 km and SmS and sSmS are particularly obvious at

the ranges 170 and 185 in the first three columns. Direct SmS

becomes less strong in the most severe case given on the right.

Apparently as the ray paths flatten, they become more sensitive to

lateral variation as one would expect and the multiples which travel

more nearly vertical become important.

The introduction of random scatterers into the surface layer

produced significant amplitude anomalies in the tangential case but

did not greatly increase the coda which is so commonly observed in

regional data. The effects on the (P-SV) system are more dramatic as

displayed in figure 24. Synthetics for the RMOHO case are displayed

on the left for comparison. Note the simplicity of the P arrivals,

essentially Prp and pPmP, which are not affected much by the

shallow velocity anomalies, nor are the corresponding SV phases.

Significant short period scattering occurs following SV which is

apparently caused by the interaction of the Rayleigh wave with the

irregular fine structure near the surface. The latter contributes to

the mysterious Lg phase which is prevalent on short period regional

records. Convolving the broad-band SH responses of figure 23 with a

(wa.sp) instrument yields similar looking records, but showing

considerable modulation in amplitudes, as displayed in figure 25.

The amplitude fall-off or attentiation is determined by the

strong-motion community reflecte the choipo of crustal model The

17

thick mid-crustal laver of velocity (3.6 km/sec) is probablv ci"

common in western United States and is responsible for the rapid

decay between 50 to 90 km. The increase at larger ranges is

influenced by the lower crustal transition and the sharpness of the

moho. Note that these amplitudes are also influenced by radiation

pattern. These plots are appropriate for the strike-slip case which is

the most common type observed along the San Jacinto fault system.

The strong arrival at the range of 105 km in the full-scattering

model is caused by constructive interference between the moho

reflection and first multiple in the top layer (Lovewave). In general,

the strong scattering introduced in this exercise was sufficient to

obscure the moho reflection, but the possibility of large motions near

the moho-cross over is hard to avoid.

The increased amplitudes near 100 km, as discussed here, are

not apparent in most strong-motion datasets. However, this may

reflect the prevalent processing methods normally applied; where

one truncates the range of interest at the first strong-motion station

that does not trigger. In light of the strong motions observed in San

Francisco for the Loma Prieta (1989) earthquake, (Somerville and

Yoshimura, 1990) and observations from the Saguenay earthquake

(1988), Somerville et.al (1990), this subject should probably be

revisited.

Conclusions

This paper investigates BB wave propagation along a corridor in

Southern California from the Imperial Valley to Pasadena.

Seismograms from Imperial Valley events are characterized by SH

18

waves which appear dispersed, and possess well-developed codas.

Rayleigh waves are dispersed and relatively strong compared to

body phases. These features can be modeled with a finite-difference

code at periods greater than a few seconds by assuming the following

two-dimensional model; a slow basin with a shallow moho followed

by a thick crust. Propagating 50 km in the basin is sufficient to

disperse the wave-field. Shallow sources apparently excite the

surface waveguide where the field travels slowly to the edge of the

valley. There it is re-radiated very strong coda signatures. It

appears difficult to propagate the wave-field across boundaries of

this type by present analytical techniques.

Modeling events at smaller ranges in the normal crust proved

possible with analytical codes. In these codes, a decomposition of the

wavefield into the following individual phases was useful: direct S

and Love, diving S and sS in the tangential case. Shifts in SmS

relative to sSmS of a few tenths of a second allow excellent fits in BB

records in many situations. Directivity plays an important role in the

sharpness of these phases which adds to the richness in modeling BB

regional phases. At these ranges the Moho critical angle reflections

contribute the strongest peak amplitudes, at least in flat-layered

models. Numerical experiments conducted with the inclusion of

scatterers into the various layers did not change this result.

Scatterers in the surface layer produce signals which resemble

typical Lg phases, at least as observed on the vertical component

waveform.

In conclusion, it appears possible to model depth phases such as

%S in tectonic regions with analytical methods if major 2eological

19

boundaries are not crossed. Secondly, the ratio of S to Love waves

can be used to estimate source depths if paths are calibrated. Future

efforts will include the application of source inversion techniques to

these type of records.

Acknowledgments

We would like to thank Laura Jones, Larry Burdick and Hiroo

Kanamori for their reviews. This research was supported by the

Advanced Research Projects Agency of the Department of Defense

and was monitored by the Air Force Geophysical Laboratory under

the contract F19628-89-K-0028. Contribution No. 4987, Division of

Geological and Planetary Sciences, California Institute of Technology,

Pasadena, California.

20

References

Bent, A. L., D. V. Heimberger, R. J. Stead, and P. Ho-Liu, 1989. Waveform modeling

of the November 1987 Superstition Hills earthquakes, Bull. Seism. Soc. Am., 79, 500-514.

Bent, A. L. and D. V. HeImberger (1991a). A re-examination of historic earthquakes in

the San Jacinto fault zone, California, submitted Bull. Seism. Soc. Am.Bent, A. L. and D. V. HeImberger (1991b). Seismic characteristics of earthquakes along

the offshore extension of the western transverse ranges, California, in press, Bull.

Seism. Soc. Am.

Burdick, L. J., C. K. Saikia, and N. F. Smith, 1991. Pn for the Nevada Test Site, AGUMonograph on Explosion Source Phenomenology.

Dreger, D. and D. V. Helmberger, 1990. Broad-band modeling of local earthquakes. B:'!Seism. Soc. Am., 80, 1162-1179.

Ebel, J. E. and D. V. Helmberger, 1982. P-wave complexity and fault asperities: TheBorrego Mountain, California, earthquake of 1968, Bull. Seism. Soc. Am., 72, 413-437.

Flatte, S. M. and R. S. Wu, 1988. Small-scale structure in the lithosphere and

asthenosphere deduced from arrival times and amplitude fluctuations at NORSAR, J.

Geophys. Res., 93, 6601-6614.Frankel, A. and R. W. Clayton, 1984. A finite difference simulation of wave propagation

in two-dimensional random media, Bull. Seism. Soc. Am., 74, 2167-2186.Hadley, D. and H. Kanamori, 1979. Regional S-wave structure for southern California

from the analysis of teleseismic Rayleigh waves, Geophys. J. R. astr. Soc., 58, 655-

666.Hamilton, R. M., 1970. Time-term analysis of explosion data from the vicinity of the

Borrego Mountain, California earthquake of 9 April, 1968, Bull. Seism. Soc. Am., 60,

367-381.Helmberger, D. V. and S. D. Malone, 1975. Modeling local earthquakes as shear

dislocations in a layered half space, J. Geophys. Res., 80, 4881-4888.Helmberger, D. V. 1983. Theory and application of synthetic seismograms, In

Earthquakes: Observation, Theory and Interpretation, Proc. Int. Sch. Phys. "Enrico

Fermi" Course LXXXV, (eds. Kanamori, H. and Boschi, E.)(North-Holland Publ.,

Amsterdam) pp. 174-221.

Ho-Liu, P. and D. Helmberger, 1989. Modeling regional love waves: Imperial Valley toPasadena, Bull, Seism Soc. Am., 79, 1194-1209.

21

Kennett B.L.N., 1989. Lg-wave propagation in heterogeneous media, Bull Seism. Soc.

Am., 79, 860-872.

Kennett B. L. N., 1989. On the nature of regional seismic phases I - phase representations

for Pn,PgLg, and Sn, Geophysical Journal, 98, 447-456.

Langston, Charles A. and Donald V. HeImberger 1974. Interpretation of body andrayleigh waves from NTS to Tucson, Bull Seism. Soc. Am., 64, No. 6, 1919-1929.

Romanowicz, B., 1982. Lateral heterogeneity in continents; moment-tensor inversion oflong-period surface waves and depth resolution of crustal events; body-wave modeling

and phase-velocity calibration, Phys. of the Earth and Plan. Int., 30, 269-27 1.

Somerville, P. G. and J. Yoshimura, 1990. The influence of critical Moho reflections on

strong ground motions recorded in San Francisco and Oakland during the 1989 Loma

Prieta earthquake, Geophys. Res. Lett., 17, 1203-1206.

Somerville, P. G., J. P. McLaren, Saikia, C. K. and D. V. Helmberger, 1990. TheNovember 25, 1988 Saguenay, Quebec earthquake: Source parameters and the

attenuation of strong ground motion, Bull. Seism. Soc. Am., 80, 1118-1143.Stead, R. J., (1989). Finite Differences and a Coupled Analytic Technique with

applications to Explosions and Earthquakes, Phd Thesis, Caltech, Pasadena, California.Vidale, J. E., and D. V. Helmberger, 1987. Path effects in strong motion seismology.

(chapter in 1986 volume of Methods of Computational Physics, Bruce Bolt, ed.).

Zhao, L. S., D. Helmberger, 1991. Broadband modeling along a regional shield path,

Harvard recording of the Saguenay Earthquake, Geophysical Journal. (in press).

22

Table 1. Crustal model MPM

Layer 1 2 3 4 5 6 7 8 9 10

a km/sec 6.04 6.24 630 6.52 6.58 7.90 8.10 8.20 8.27 8.13

0 kn/sec 3.49 3.61 3.70 3.77 3.80 4.60 4.70 4.55 4.723 4.74

Th km 8.0 8.0 8.0 8.0 3.0 10.0 10.0 90.0 35.0 45.0

Table 2. Earthquakes along profile

date origin time latitude longitude depth magnitude distanceGCT degrees N degrees E km from Pas.

03/23/54 04.14 33.28 -116.18 16.0 5.1 230 km

04/09/68 02.28 33.18 -116.12 11.0 6.4 224

04/28/69 23.20 33.33 -116.33 20.0 6.1 191

04/26/81 12.09 33.13 -115.65 6.0 5.3 260

11/24/87 01.54 33.08 -115.78 5.0 5.7 251

11/24/87 13.15 33.01 -115.84 2.0 6.0 252

01128/88 02.54 32.91 -115.68 6.0 4.6 270

05/17/88 19.38 33.24 -116.25 8.0 3.8 205

07/02/88 00.26 33.49 -116.44 12.0 4.1 176

03/06/89 22.16 33.17 -115.59 1.0 4.7 262

Table 3. Normal Southern Cal Crust

Layer 1 2 3 4 5

a 5.4 6.2 6.6 7.5 7.8 km/sec

3.2 3.6 3.75 4.1 4.25 km/sec

p 2.7 2.85 3.2 3.42 3.45 gr/cm 3

Th 4 16 8 m3 k

23

EPO withManyLayem ( M S)C11

Figue I Thetoppairof 'ace conpars a Syntcthtfis 0 o el ftevria

Len g-Period Vertical Layer Over Half-Spane

Figurese. The tofir o tra0cees compsed sdyticiffatdtnryhln the fistop e of the vria

mantle, Pnl. The latter 30 secs is composed mostly of SV diffracted energy along the same mantle

path, Snl.

46.3

000

4--.~

Al -

toJ

A'7/

4*=

c? /~ ~ a ~ p cb- / / ~i 5a

Z~ *J

o -qC.)

00

CD 0 nc 0

o0

-E

ocoo

o E14

W 0 cl 0-

0 Uo oECIOC

00.

.0

co 0 Lo

26

C\2 C\2 CD0 0

o 0 0

t*0 000

L.o

0 - 0

r oc 000

LoLao2

m co 00 0

C) >

00

oo cr

W. CV)O1

a) I)

00-mvj

28o

(U 0

cy W- -

o- o 0CDco ( 0 'E CVv

V *

U 10~~II UE

CD m

II VVE0u

-Wo)

29

J..

CL

J, iz. -~ c

- -

- -0

C "~:.- 30

3 B=jS = 4.18 19=3.78 3.38

p- 2.65 p =3.2 p =2.7 p=2. 6

\ . 50 km

B= .0 B=2.3 B-3.18p = 1.4 p= 2.3 p=2.5

Figure 8. Seismic parameters for a 2D model connecting Imperial Valley to Pasadena, after Ho-Liu and Helmberger (1988).

31

LP 30-90 LP-torsion SP-torsionA, km 1.88 2.90 1.26

262

2.00 3.23 14

212

2.64 3.82 1.67

162

3.27 4.37 1.73112

626.42 11.71 6.34

56.09 106.62 45.50

12

60 sec

Figure 9. Synthetics appropriate for a suike-slip source beneath Imperial valley (d = 7 kin)assuming the model defined in figure 8.

32

m =4.9 1/28/88 AFTERSHOCK

20 sec

Pn Pg SinS sSmS Rayleigh

M - 1. 1 )( 10 23ergs

Uilgurc 10. CVOnparisci -,r synthetics and obse,-va-,ions for the Superstition Hills aftershockconstructed from the model displayed in figure 8 assuming a=1.733.

33

1981 Westmorland Earthquake

Data

Synthetic

data

I-wc. filtered

synthetic

data

2-sec. filtered

synthetic

C.0 SO-.Oec.

Figure 1I. WA (100x) recording from the Westmorland earthquake. Two sources were used ingenerating the synthetics, one with a MO of 2.5 x 1024 dyne-cm at a depth of 7 km and the secondone with an Mo of 4.5 x 1024 dyne-cm at a depth of 10.5 km, delayed by about 2.5 secs. This

2.5 sec. delay is also observed in local strong motion accelerograms. The lower comparisons are

made after filtering the synthetics and data by one and two second triangles.

34

02 ooCOCV) 0 mColc 0(D 0

Qc

5.ca

(7) .)

CVV0 I)

CD2

o o

10

CVu

C-

to 0

C0 0 0MO

0.

35

00

cr) >

C4)

(LUM)4160

364

0 U0itt

- U) U)

(.1) U)-

IC a U) U

W 0

C Li

0'0

0C

E

0L O) C U

FILgUre 14. Wave-field decomposition displaying the response of the direct arrival on the top

followed by the contribution from down-going paths (S) and up-going paths (sS). The bottom

ro%% displays the synthetic assuming a (.2.. 2,.2) trapezoidal source.

37

- - - - - - -- - - -

I I I

CI N 0 I 0.in cIC (4

Pe W) (/)- a.

38 Q

Synthetics Synthetics

•7 X 10 4 CM I.

up-going

1.4 .8

diving

1.1 sum

191 km 1-224 km

Coyote Mtn. (h= 14km) Borrego Mtn. (h= 9km)

Figure 16. Synthetics constructed from the step-responses (strike-slip) displayed in figure

15 for a range of 191km along with a similar construction for A=224 km. Comparisons

with thc dtt;a is given in the bottom row. The time function for the Coyote Mtn. event

coi ited 4,fa trapezoid (.2,2,.2). The time function used in the Borrego Mtn. event

, ,is ed of two trapezoids with the same timing where the second is three times the first

,nd liged .7 sec,. 3q

co

T)

An"

Cto

5 I0E %n %n C% v .

cli~- 0Y C

0. -O-

CY E 0

C- I

oE cct,0

!U,

wo 0V

CD-a>C

40~

a, "

E CI

cn -1

a.a

CLo W

CD- cj

E4

obs = .54 obs = .36

obs= 1.8 obs = 96syn= 1,5 S =.99

-- li~P) I'-

A jf .27 Sum .20

!.1f .16- Iv -

32 dn13

I I V.21up2

lOsec

Figure 19. Comparison of observations (solid) and synthetics (dashed) for event (a) and

(b), see figure 2, the lower panels display the wave-field decomposition assuming a

(.2,4,.2) trapezoidal source.

42

E E EECYJ CD

00

Ch)E 0 o~

E E E E-t- LO. 0g

C c

00

ECu C u .0

*- rn 0

cc~

m~ CLoO

ET U) !e- w

En E EOD CD(P 09I CY

0

Y

43

LsynobsVA0.9M

I--10 sec--

Figure 21. Comparison of the upward propagation case with the observed Coyote Mtn

event. The peak amplitude of the data is .19mm corrected for instrument gain (low gain

100x), predicted to have a peak amplitude of 53 cm if recorded on the standard Wood-

Anderson instrument (2800x).

44

r- 30 l)LL 0C i C j C~ I- li c

11 0CL 2

LO LOC\I (D I'-- V--

c'J 0

o 0

U) C

1~E

E

o 0112

0(I)

45

FD with Random Media Reducing V=3.90

RUOHO RSURF RBOTH RALLS0km ~3.62e-05 cm 4 41e-0b cm 4 41e-05 cmj f 4 38e-05 err

65km 2 26e-05 cm 2.25e-05 cm 2 25e-05 cm 2 68e-0 c

80km 2 12e-05 Cm 197e-05 cm 1A. I 97e-05 cm A e 3e-05 eni

95km 52e-05 cm 2.29le-05 cm 2 07e-O5 cm I 94e-05 err,

110kn S6e-05 cm_ A 2 5Oe-05 cm 4e-05 cm 3 02e-05 cm

125km , 2.23e-05 cm 1.7 e-05 cm 1.92e-05 cm I 49e-05 cm

140km 1.4 6e-05 cm 1.6 e-05 cm-A1. 47e-05 cm__, , I 50e-05 cr,

155km 1 e-05 cm e-05 cm . I 64e-05 cm i 36e-Ob cm

170kmjj, 5e 05 cm 1.57e-05 cm I 62e-05 cm 6"4e-05 c.m

1 3km 1.33e-05 cm i 1.,28e-05 cm I 30e-05 cm 3be-05 c 1

200km .l1ee-05cm 1 4e-05 cm ~ I 9e05 cm b- ,c

V1km _ 4 / 372 9 .km A A 9 8 3 e -0 6 c m 7 5 3 e -0 6 c m 8 8 9 e -0 6 c m e I]3 7

230km7__ ?e-O6 cm p 8 8le-06 cm 8 78e-06 cm , 7 .e-o.,"~~~ (:,..: ,.

245km I 13e 05 cm I OBe-O5 cm I 19e Pr j M f 9 ,t

280kmI 1 e-O0 cn, I 07e US cm I I e ), ", 06 E

16 00 sec

Figure 23. Profiles of BB tangential synthetics through the cross-over distances of 5(0 to

2(A) km. The column on the left (RMOHO) is nearly identical to the homogenences case,

s c figure 17. The next three columns include increasingly more layered anomalies.

4 6

4 i Ifl C 'i N~~ 4aO c.~U, Ln r M ci Cl c c.) N~ N

W) V m M m O ) c.) Cl2 m* c.)

.0 ~ 0

.r 0

cfl (n (0c

cco EU

47

II 1

00

I ~000

ogo

04 -

44.3 E00.

0 0 :

0 0r_

0 0 C0o 0000t

4104:0

4:0

a)) LO 0

IEOBJ 410

48

SECTION 2

Dezernfining M., Magnitudes firm Regionial Ni'S Data

49

Determining M, Magnmitudesfrom Regional NTS Data

Brad Woods, David Harkrider

August 2, 1991

Abstract

We lta'v ca~lcu~lated surface wave mnagnitudes for 102 Nevada TestSite ( NTS) underground nuclea~r explosions froin a dat a set of regionalIlong- period svismlogralus froin Nor th A mericani stattion s. [it order to

IIt -lhe iierer re'ginllstiimis(A < 25'),ia new miod I'M11fo deter-milling N1, has hen developeo Which enliploys slit lid i' "eil.-iiop r; illsito (stah~jlii at lonlshtil l)etwevii the anilflitiidc oif thle re'gioniit .%iryphase, or Rayleigh pulse, of the data and( alt associated surface wave

Ilili~tluillilc. ImaseI oill conaventional NI. dletermiinationis, calculatevd fromt

aI s ,itntet ic svisliltgrain propalgated to 1 0 '. This, Itiethiod easil dsitself to inmplementting pah corrections. 'I'lle imicitsiomi of path mi(rrec-Ilimis decreases tiw NI, varianlce fly a factor of t wo)ai al( 51)s (I cre sesOle average value hY 0.08 imagnitude uniits. his Lat t''r effect I, Mtril(Itted to thev ptdicilark stamitioiinetwork iised. T[he it' lngie

si ihIv NM, valuevs t hat correlate well with oilier ntiagitit itli 5(if dtiesover aI rnige of thIree orders of umagntitudle in source Yld. 0111 liow'4refinied M,~ values 'ield tOwi reltiotli~p nit, t)0.'-2 , I, -, It. \. howe Itis dIep~end~ent import ou rueIT regil oi a shot iltedi uml. Tlti, r wia imi lphtoldls for vvetts of ;ill sizes. W\livii events are groitpcil with Iisct~o source regiont, sigiiiin1tY l better fits to t hie' ilmdlvioii~l uiI ill-eair r4'gressioii ('1ii 111 ;vsaic i tIII I lilt o ipa red to 114 t 11 (1011ie4 lit is i . igaI siiige, aII ll icliiv iiiinlel. 'This olisvrvat iOnit iplifs thfit 4l iloti

;itiieteIr.liim smitrce t rlittirle 1,11,V4t sIir1;tc' ve iInagniltid nica 11(;siirie'its atII h(ioItgli E~ Ii Vi4.I s Ite ( dIstI r-ili to) I k Ia II II; hi 1 pInres po IIsi 1)14.

Wi I'll'( eIt 1 V(ill t(&V;IId5 miakisig I hie dal ;I sel comtprel1.iusive anid(livvise ill 1,rtus of YivIdl, smiriPc lot and 51101 iiiediit n ini orderto det erii e he' portability of Seismic nteasu rinig scales. Ini pa rticuilarWe Oxl i ii Pa linte Mesa, RIainfier Mesa an 11( j't ta Flat explosionstle ii ated abhove and t below the water table.

Si fOe' onur ntagn it de valutes are based ott a theoretical continuenitalsi ritturv. we regresseo oiur %,allies with the more st andardl vales ofMarial! I al (19 79). U sing 1.1 cofumon0f N'JS events we foun d t hatour values were greater by bY 0.53+- 0.03 rttagttit(- ilnts.

1 Introduction

We re-eXa tmi te th li s( of Surface- waves for itdergrtou id iclear Cxp;losioi IrI Iag InIt It(lu' (etItof-1 1iatm lo IIs, p~art ieti Iarly for sitlle I r leItl ( Y < 20l\*1t even t s.

'Ilt' ttr~tu WaE' n~gliIt tl--vIel saliiig law for siich hm-~ % ceviei'~'it. mit 11flow, was fiot. kriowti wt-ll. 'I'' dIat a used ame Ilitg periodu 'Noul It .\uti'rjo0tsA at ion vert icaI records for 10'. spec'iflied Nevada. T'est Silte ( NTIS) v~vitls. I Icst ations used are front several net~works. rteir respective itust rimetit s a.1 la&

pass5 bandls thtat Hei withi n the 10 t~o 60 second range. Surface waves axte veryulsefuil for vivid estimnationt purposes, for ( M.,) is det erminned front relativelYlonlg-period seistmic wvaves which are itisensitive to high frequency utear-sourceeffect's. wh ill itlong wit hL several othter possible mtechtanitsms, may be' cauised(b)y asvnimet ries in thie shot. cavity, see Zha) and Ilarkrider (19)91 ). Thesehigh frequency sotirce effects may ca use appreciable bias in uiagnilt ides 111,atare based ott higher frequency waves, suich as the mlt, and~ Ig scales.

F.or t lie lower i ielol ev~ents it, becomies necessary to inc(lud(e the lta frontre'gionial stationts (A < 230). for teleseisinic recordintgs hav oo0 low a signtalto noise ratio. whliich ma kes themun uusa ble. At, regional distances surfacewa VS art, Hot wel dlisp~ersedl, ha vinrg a. prottintat Air 'y phatse5 pulse with Itperiod betw~eent G and( 2(0 secondls ( Alewite, 1972). so that It Is tiot possihuIClii tuteasitfe M, (ulfivefit tontally ( I hial Is rIleastiring I ie( aitllt tfilC of tile *21SC'u. wave). h'or Noth Attertica in gen .eral thevre is tii i ll itthe gtuttjvelocitv cntrye tieax 12 secondos for the fti ndamnttalII Ray liglt- wav fI Ma rshllI

fI (it 19719).TlO miiua site Mwe emtiploy a tech niqtue whereb)'y t heoret ica I seism togramu

ill Conjuti(tion with Ithlesurface wave data are erployedl to ittllitectlk~calo'tlate

51

it.. fin using this procedure several propagation path modlels were testedl todletermnhre the effect, of attenuation and~ seismic velocity stiietutre upoti 1 lieM., values. These calcullated NM. values remain stable. have reas(,IlaIblv Smuallerrors and correlate well with associated 11b maginitudes and~ log yield forthe event data set. The M, - Mb relationships are (leternhiri('( byv a weiglitedleast-squares linear regression.

Besides comparing the N1, results with several dJifferenit iideperident mnag-niitud~e scales, the dat-a have also been separated wit It respect to source regionlmtid shot, mraterial . M a vaties at Yucca Flat tenl to be larIger tiahtrthosc atltaiiier. Mes ' 0.t)9 ta-gtilitide tiijtIIs lor a giver) III,,. 'lcreg a ppear"I~ii 14)

vSol IIt d It~een FV i I('(,rit Ie Iet evIt ISCO- l I VVlwsSe1'1 tS IWO Soil ue rgil )I S.lPahlit,e Mesa events axe 0.22 umagniitude uniits larger I ball I lose at Yucca flatfor explosions set-oif lielow the water tale and( Wit Ii I Ie( salli II,.

WVe (to not account for tectonic release effects uipon I ie( magii t ude Inca-surmnuts. Such effects are best accounted for withi miomnrt tensor inversOinsof sources which involves more sophlist icatedl (atita aialysis, St alidlard M,mueasuirement I echniqtuies ignore t his fact or as Wvell.

2 Data

'lThe (lata are long- I)(rio(l vertical seismograms recordled at Nort Ii Ainericastations for 102 exp~losiorns at NTS arid consist of digitized W\orld Wide SennicNetwork (WWSN) and Caiiadian Seismiographic Net work (CSN) recordls.D igitl W \otld W~ide Seisitiic Net work (DIWWSN ), Lawrenice Livceriiore Vve-glotia I Seismuic Net work (I.N N) and IRegioiial 'lest Seism ic Nr Iw. 4 k (161 S N)digital dataa for eemts occiririlg later Ot 1 981. '[he analog \\\\.'N ariIdC an~adianr stat ion (data were dligit ized by l'NS( '. Fifl~v- ighit stattns corn-prise the network, alt hough fewer thatn 60 pe(rcenIt of thle stat tomir hiad dat aavailable for al ' y si tigl(evn - Fig. I shows at miap of Illie stat im t nietwi rkEphic('ttIral (list aIteCs ranIige fronm 220 k ill for N'lS to C SC ( ;olditin. C allforiria). to 5200) kill for Ni'S to ST.] (Saint ioliis. N (wl'Oilaiii). F'or IlicsinaIler events, particriIa rly R~airier Mest ('xplosiotis. mii Ili t~erca cr stationls(distance < I1000 kill) hiad eit her dlata available or teasotiable sigtial t.o n1oiseratios. Station coverage varies widlely bet wevti (,\(its. Lieof r he sriallhrevents only had oite, viable st ationi scismnogramr each, while Somre events hadover :30. 'F'lie a verage rtniumber of stations rej)ort iilg jper event is It).

52

Silrface wave', I Ihat propagatec across occaiiic/cmit Iiietal m iarginis undergosigi fialliin)l oi ficat iol in I hiir wa veformys beca use of' IeI( great, lateral vari-at ionl III t alld ufltpi4'r ale sitructitre at such bounfdaries. T'Ivscpr' pagati'o 'Iieis at-e not straight forward to molII, hence(( ap)propriiat('( ree' ' fu nils, or Irldlsfi-fr flniils. are 'Ii Iicitdt to ob alit. With1oult ro-Iiti'st ( '.r4-4lif 's fit it 0 t n11 I s It ail I4 o iItIl.'r acc I Irat v S4oIIrct. Il iformta IlollI fnr)II IIII,' datit. Smialler cvvils also ilre( 114) likely Io) bc4 4ili4'l''4'4 ait 11'(- dist aitistations, which oftenl iltid ('( oceanlic si ruuc0t1 re alonig I heir projiagal ion path,atid make Ilie se loniger pI, ts evenI less attractive to iichiude in thleioutitorilignetw~ork. I hence, we chose to conlfinie ou r simliy to1 siirface waves t ra vellinigsOlv'ly ahat , cni iettal paths s. . withbin North Am\ lerica.

Of I ht- 102 even4'ts. 27 are f'ront lPa l ite Mesa. H; arte fromt tla i ier Mesa. 58ff ii Yucca Flat anid on e, Pijled river, is loca ted at ('lii ax Stock. W\e considerlitse to 1)4' I (distinc Soft[.(.(,rc regions. For sonme specific stations, wa veforn is

va rl('( somiewha t bet ween events, depend~linig 111)011 soilrc'e location.The I'iledriver (la fromt a given station look appreciabily different, fromt

t hat of allY ot her evelit s re'cord(edl at that sa ftte stat ioli. llThis wa~s rite foru'evstation rec(ordhing P i led riv~erlaitd prlobably' is caiised by dIifferenices ill lCne

s(Jli-c('eliott fot I his e'xplosiont. Ililedlriver was deli(tatell ilt a. gl'amlt.i(' souirct

re'gion, nolrthI of I lie fit r sitems'. Tlhe souirce' to r('ce' vet geottuettnics for this('vo-lt ate applloxiltial ('lv I ie( santle as I hose as I1114' ot her NIS (ens, so t Ite,411 hl-creic' I t wa vefortits , loiesnit appearii I) b e alt I nili able I4o 4Ii5 1 t('rsive evi'ect'sCalis(ld l),\ (ifel'tii4' Ill propagat ion pail l len1gil It. lielri vet wa~s thle onlly( 1 ia nix Stocke('cliii withl readi lY avallabled(at a, so no0 ftirt 14re(xal)mat ion of1bis sotircI' was carried out.

At Scitlif' Of tithe nearer regn lal stat ionts ((Iistarice < W~), there are alsoSbt ledd l'c'iccs bl)'wee'4'fi t, nYucca F~lat arnd R~ai ner Mesa evenlt waveformis.

At 1) ; '( i )tgwaY. Ut al), (or ('xamph4, tie( lI aiier e'venit wa veforitis 1l(ok asif ille Ati i phase, has be en II iibert t ransfortned (equivalent t~o a 900 phlase'Shift* ) relative o, Ite Yuicca Flat waveforms. The DUGI Ilailier wav('fortitsalso contIaI tin i more Itiglt freq tiency codla energy than th Iose from Yucca, Flat.l'alI t 4'ev4nt-4 a rc simniila r ill wa veforrti to Rlalinier (' itsalId hlave( le ss highi

hIle S( .islIo 1gra Ills well' banld-passed filt~eredl bet-weeti 6 anid 10 stte'ondcs

l I d114 Ibf the lg pocriod altic shlort. pertiod lnoise. whichl woiildlu "Ilerwisaffect 1the jnctk Io) peak iiu'asitt'elliill of till' Ravigli IttIi.sc. Then albsollilhaixiplit tile's of* the dat a were ve'rifiedl by choosing seve'ral diffrenr(lt stat iolis

53

mid comnparing the p~eak to peak aniplitutde of thie (ligit ized recordl to t hatiieasiired (hilPetll fromi I lIe( respective analog W\N'SSN film chip. Friomi thisstelb We iiscertiiIICil haitc stat iou gaill factor hadi' lbee ll orrt 11Y factoredl(lilt of Che- seistiogramns.

3 M, Calculation Technique

We have developed it lillo to measuire surface miaguiitudles Ind~irect ly. Be-cuse a large portion of thed(at a for low viv(1( events is from stat ions rec-ord i rig

at, regionial (list aices (A < 250) , it, is not possibIl~e to ca lou it e Mcurliyel-tional ly, for the Hay leigh -wave is puilse-Ii ke whiich precl udes invtasmIni g it welldlispersedl 20 sec. phiase (A lwinte, 1972). We adodress thins prullcli withI thle Uts(of synthetic svismnogra-ins of thle fundamental Ra leigh-wa c unsinrg a mod ifiedllaskel-Thounpsoii mat rix method (Ilarkrider, 1 96-1l).

For each souirce to receiver path a theoretical hlavleigh-wave Is gerierat vd.'lhie Earth model iused to createv this syli thietic is filt-alit to relicclt the a vera "eFEarth structutre between NTS and thev givenl stat ioll. '[hle Flart 11Iild. ;141in inhs study were deteriiiied from Iiversionis of (1ispe(r51 Il at ;t elilt Wlul&la tas well a~s forward modeling of the waveform to fill(- t lilt(- t( ie oduels.Tlhe criteria for determ inihig thef( goodness of fit of t ie( spit lietic to I te dat aare dispersion, absolntte travel timie anid waveform fit (relative am plituode ofdi Iferent dispersed phases). I hence the svnhieti' scislnogrann dlisplay, s t Ie(samte spectral anid tiime dloma~Iin waveformi chitract eristics as t ie dat a whichit. simulates. T[hiis wats done for all pat Is. 'I'lue paths t o \\\\S.\ it~ I Cariadia must ations were taken fro i a 1i dY bY St evenls (St evenls, I 986i) . 'I'lie IN.LLN amidl I)W\VS N pat hs wered(et enmii ed previoiuslYy vthlis resc arch grouip.

Jodoeterminie N11, for a part icular sorc-rceve Ieme \%v ( t *u svit het, ic ,are generated. Oile which is propagated the act tial patl Id(ist rice that lmneatit. t~o simulate( IlI(' dfata and( oine whichi Is prupagat- 14I 10u".t At tU' 1 Itsurface wave traiii is wellI dispersed ando stabie, so that at tcolon imal Ml,value call be CaIcil ilt ed. To calculate, M, we is(, I lie, miodified Von Sevt-wrifornituIla ( VOit Seggern . 1 977):

IM, log11 1( /'l -I . t~ 1g1 (A) jis

Wi('e A is I te peak 14) pea~k amlpit iide (it) 1l;llllrec4r1S) (r I tIe wauvclct 114ld

siire,(l from tlie vertical record. T' is thle period of thle wauhI l. mne'a5,r111 i

5

seconds, and A is the propagation distance in degrees. Tlhis scale was choselibeca IIse t I e dlisft rI ce cot.llIicivlt. ( 1 .08) mhore clIoselyv a p proxiwates t Ie effect, ofat tewtial 101 alonig cont inetl pjathus. A vertical collponi('t mieasiireiiet hasIwo adjvatagesl.4 over hiorizonit al comlliellt. wiwtasiwrt'iwltt'w . 'I'le Ili'zoiital

ptorterit awiti gtiit'rally ale 11101V likelY it) Iwc conlIanliaietl 1) 'v Love' wave sigita Iswih mia 'v he generat ed I) tetctoiiic release, source effccts, or scadt i wg dm-ito lateral variations in the E~arth's structure.

Rot 11 t he regional antd teleseisniic synthetics are generatedl withi the samesotaC ruc fiw111111 so t hat t lbe peak to peak amnpuit ide of the Rayleigh pulse oflie regionalI syit hict ic call ke dIirectly related to thle M,% va I ie det ermuined for

a t heoret ical RaYleigh- wave train propagatedl out to 100 .Ihie relationishipbet weeni I lie- (ata peak to peak amhplittile antd its Indtirect M, is:

Nildata) = N\k(svwwt hwI.i()) + ig~[hhA~;1~/ t ;j~i~]

where P hA. is tIe peak to peak anwip1it nih' of the Rayleigh ,)twlse. A pat hicorrectioni mway be includwted oi tihe right side of t his expression.

'lThis pat h correctioni is t he diff~erence between tb -lid ( iiwual path s ii v-thwetic derived Mi, andI the average t heoretical M, for th i etire network. F~oreach sowi rce- reciever pair, a M, is ca-1lnlalted fronut awit hwetie0. seisiwograwwprop)agat ed to .100 . Each swici synithietic has O le sam sn size source, So idlea llyotwe would want each N1, valuteso measuredl to be equwal in value. Yet this t hiswuot so. for each path's affective at teniwation at. die periodls of interest nwaYb~e ttifferenl . The dlifferenice between the mecan net work NI, alw(l a. Jparticiliarreceiver M, is tie path correct ioni. A nwegativye patti (orwect iow val i(' impliest hat tie t livoret ical .100 st at ion M, is larger thawi t lie wiet work average'. TableI lists thle network patti corrections used.

The quiest ion arises, whether or not it is vatlid to ise thle average Earthst rictuwre for a part ictilar path to propagate a surface wave to 400 when t he[lart h mwodhel is owiY m ieant, t~o reflect the seisi pclrop~ert ies of tIwec Emit In For apat i Ilial imav'N unlv he a small fractijowl of this tlistlawce. 'This is jpartlictitawkrue of Ihei' short est palt is for which thle seismi c wavyes traverse onily west -

ern. Northi Ame nrica, a wi area of relat ively high at t ('iwaltiou comp a red to thleco t i lien tal (rat on and1( shield areas. A siwrface wave propaga t.ecl .10' alonga chiaracirt 'sti je ectonic North Awiericani cruist allth wailate miotle ( N'bS toDUG'(. for examtple) for .100 will be mutch more attewniated thbani a wave prop)-agate u Ite samie (list aice tdhrough ali average st ructuwre front NIS to t lie

55

easterni seab~oardl ( NTS to SCP, for example). lleiice the calcu~latedl M, for-the NTS to DUIG structure would be smnaller than ilthe N'VS to S(T MsI.

Irhere are several m~ethodls to correct for this path dejwrideiit e'Heci. Asexplained above one may implement path corrections which accountlt for t lhethieoretical difference in attenuation bet ween paths. Anlot her ircaits is tomake a mixedl path structure which has the appropriate pat h strilct tire frorilthe source to the actual stat ion distance, with the rest of thle pathI out to40() being a generic sersniic velocityarid attenuat ion m1 odel. *I'l(- miiixedi pathIapplroximation for the funrdamnutal Rayliegh mtodle is easY to i111 lenll(wiThe approximation is equi valent to assuming that thle total hoizon~t al en -ergy flux is constant across the transmission boundary ( llarkrider. 1981).(Levshin, 1985) and (Bache ci al , 1978). Finite element reslts.- show thateven for a continental-oceanic crust transtion zone t his approxirnat ion is rea-sonable (Illarkridcr, 1981). For the cases in this studyv where thle st ructuire'swhich complrise tihe mixedl 1 ath are bo0th continental strutcturcs ( i. c. not toodlissimiilar) the approximation is robust, enough for the syihtc seismiogramcalcinaltioris.

We have i mplemlentedl both p~roceduires i rioividnallY n iil conjunct ionto see what their effects are. Anot her nnret hod would be to iniclidte empiricalstationt corrections ( Yacoub. 1983). (Oiven ando Melli narl . 1986i). T'he findingsconcerning the path corrections are disciussedl ili thle resul ts sect it tr.

4 Data Analysis and Results

Thie sensli ogra nts were h)and1 -pIassedl fiIt ereod het wetl 6i( and I0 I J econ ol" tominninuize cointamninat inrg noise as (lescribedCO pieviou sl '%. Thet ve'rtical recordswere visually inspect ed to inisur'e that they were withl 111c leU rnect t n 1iCWindow andh that t heir signal to nloise rat io was alb(we 9.0. Ni. val ies we-re

heln valenilaletl for the dat a as pel. 111 hi' metod tetrbIatut(eqiat ionl10) withI several variat ions. 'Ihie stiut hnet i( cs 5(5loIga 1is wen e ako hianid-paisse'lfiltecreol bet weel 6 a td 100 setonlds fo oss c.I' he M, are jpl(Awdet againstIseismnic mtagn itutdes of several stales fo t ilie saint' set of ('vt'i is. It shld t henotedl that Comniplete miagnitutde lists were riot available for all 102 evenits.

We chose to conmpare or plot our. dat a Ipriliia-il 'y witl hod * ity wave miag-nituides deterii ned I, Li Iwal I anl( McNeary (1985). Thet Li wal dat a setcontains 75 seven of thle 102 events exam i neo bYi us a rid is I el eyed to 1c

a well deterin red anid sel f-consistant list, of til, valu tes thlat. la-ve smrall er-rors due to, aiiiong other th~inigs, lt( in clusioni of nectwork st~ation correctionis.Fig. 2 show illS Ilh i ,-yield relatitonshi ip for (eent s ill tiis study for which liii,all(l yield in forumat ion were available. It is imitport anlt t~o notice that eventsabove and below thei wat er t ab~le separat I int o t wo (list iiict popiilat loris. Fort his (fatla ,(.t t F1Its sepat Hat toil is oiil I appa rerit near I ie I(1 chi of events wit Itin il,s a roud it]-. 1. A lso riotI We I h very'N Siiiall ercror bars for ItIs (tat a-, for i allyevents Ihe error bars are( smialler than tHie sYlilols deiiiarking a dalti. poinit.

Itie corrl at Iloll bet weeli III I a Ii( yield is goodi, wit III I Ie scaItter I lost IV bet rig(lie( to the above water t able shots.

Several sets of synt lieti t( hayleigh- waves were ca Ictilat ed at t lie 100 (I is-tance. One set was propagated along the single st~riictnire miodIel (hereafter

refe rred to as t lie single p~at ht case) which reflectk I ie( average [art hi st ructureb~etwee'n NTS and a givenl station. T'wo sets of riixedl path svynthet ic seis-niograrns were also genieratedI. For that part of the path beyond the act tialsouirce recrever (istatice, ouit t,o 400. a generic eath structure was uised. '[heNTS-HS F~51 arthI strtct ire was chosen for t-is generic pat Ii section, ais It Isa relativelly sinmIple stiltutre whiich generates st able srface wves and it isrouigh l air lilt 'rilediat c r~ange, sta.tioii (distatice < 1290 kin).so Fat. itlsstintic(ture call be conisidheredl to be anl "average" stritetire for thle network. 'I'liedifferenice between these two mixed path earth structures is ili their sjpectralat tenuiat tol coefficients, with j~ being twice as large, at a giveni frequnecy,for the mixed path 2 case as for the mixed path I case.

Su rface Ilagniit rides were first ca lcuilat ed from ii he *J (Jyn t hiet-i ( genera tedwith a s11ngle sttii lvi' propagation p)ath. Fig. 3 ilisplays sinigle path X1.valuies, calculated as described aboMve, verstis body wave niagii tle (nirI,).These int, 'sare those Lilwahlland MNceary (1985). 1;1 the left fig. the ms'S arecalcirlat ed wit tbouit path corrections, whereas path correctiolis are inlcluidedl InIthe figuire on the right. Fhie solid line Is the best fittinrg weighted least-squaresregression of tie (data, with the weighting factor being inversely proportionalto indlividu tal event. st anidard deviations. 'Ill(! dashed linies represent. tHie twostandard devia tion error of the lit, of the li ne to lte ilata. Solid black circlesare shot s below thle water Itablle, shots above thec water t~able are open circles,aid open squares are shots for which this inifornmat ion is not known. Notelie error baris at-re a pproxiiat.ely .50 p~ercent larger for tOle mncorrected Nl,'s

(Fig. 3Ia t han for the case of pathl-corrected NMs's (Fig. 31)). Thle scalter iii1he data is also slightly less for thle path corrected Nl,'s. It appears that the

57

path correctionis (10 im1prove M., measu rient s.Trhe most significant effect of including path corrections is I te rednctioii 'ii

variance of ind~ividulal rnagiit~des. Without path correct lulls the indi vidullstation magnitudes have a lbi-mo(Ial distribut ion rellect lug the Iwo genericEarth models of North America: the tectonic western anld cratollic easterncrust and1 upper mantle structures. The path corrections lbriiig-ili trile out-lying station niagnihtilcles values tcwards the mean value. Table I lists t hesenetwork path corrections. The first coluiimn list,s the correct ioins for sinlglepath synthetics. A positive value denotes that thle MI for a st at ion is smal ler-than the network theoretical average. '[lie fourtlh column Is thle number ofevents that were recordled at the stat ionl for the dat a (lit Ire set.

We niext, explIoredl thei effect, of mixedl patll t ralsfer filltiorls upJoni tile M,calcualtions. As described ab~ove, we chose the path to 11551 as a genericstructure for the second portion of tile mixedI path synithietic scisIflOgramlcalcllatiols. We generated two sets of these sytathietics. 'Ilh( at teniiat ion ofthe genleric p~athl wa~s doubled for one( ofr tilese sets ( 11551)\2 ). I g. I shlowsthe attenuation fact ors (gammi ia) as a funct ion of p~eriod. lhli Ine Iablc(11551 x2 is Chat, of the increasedl at t elIt natll iic ie It is, refer-red toaus "mixedl path 2" thirougholit tis st uN'v. TlhC' lower, da.Shed (I-Cllr iteaftteiiatioai curve for tie hI 551 striictulre. Sylitlicjs Illiade witIh is 11551generic structure for Ilhe lidtter port ion of t lhe .100 travel pat i will be referredto as "miiixedh path I". T[able I gives the p~atIt correct iolis f*or each st at jolt fortiies(' two cases, also.

F~ig. 5 is aiialogous to F'ig, :. t hie dhifferenice being I hat I lit, MI, IIlagilt Iideswere calculate(] using svaitlletic svisinogramns tisilg Ihe iixed pat It I :nio'el.In Fig. 5a the NI,'s are calcuialt ed withlout pathl correct ionl Ierll s. wil e InIFig. 51b path correct ions ale- iiiclided. As before thle adtlfion of thle pItcorrect ion terns cut thle varianlce by abIoit a fact or of 2. Also thle a vrage MI,vihiie (lecreasi' 1) , a 0.0) 1 mil ts fromi t hlose wit hiout pat Ii (-(,11.4(1 lollus. Thiis4Eifelce laY lbe statl Ist Wallv sigiihicauit . for t lhe stl aii'ard (..r(,.r1 1 t( e llnof thle regressionlin 111,Is (il*v 0J.0J26. What Iimst Ilkt i- bkt %.eii V, S' 3amid 5 is tat slope oft lie regr-essio i lne d lesante 101 t111. twolm"I.d ));I'llI cases, witl)

im1, =0.83 x M, + B.

hForthe sinigle pathi case, pat ha corrected m, vallues give tilie santec nclat iouiliip.but, the slope is appIrecialbley larger ( 0.9) ) for ft( ie ltorrect ed lmagmiitumdes.

58

alt hough the di fferentce li es within the errors lboundis. It would seemn thatboth pathI correctionis antd aim1 I lj xenl pat h ( reeni's fuit iti s i05mprot~ve M,(leteriniit otis for thle' mthod tis(I hiete.

Thu is t.hle I ilusin of path correct ionts ill calcutiniig N1, lowers the itag-nit univ value by) 0.0.1 untits Oil average. Thtis Canl be explaitte'l by the ntetworkcoverage' and 11 he rantge of ' viit'lhetic NI,. valutes. The (list rihiutioti of NI, valutesis skewed. wit hI thIere kviiig a sigificanit. ttttttilwr of stt itill' wit Ii theoreticalM, values significantly larger thati most of thle st ations withlin the network.Events for which such~st at ions reported would yieldl a, larger average M,. thtanfor events t hai ll id01o.

Fig. (;a and 6b are mI, VS. til1, plots for the i xed pathI 2 case withoutanni withI pathI correctionis, resp~ecti vely. For 6a a fi xedl slope regression, wastused, so5 t hat. m tore a miore apt. comiparison coulId be mlade to it~s mi xed pathII coutterpa rt plot. Com11parintg F~ig. 5a t~o Fig. 6a shows t ha t the average NI,value dIrop~ped 0.02 magnitutde units for thme nmixed p~at h 2 case relative to thIemuixedl pathi I case. This follows front the fact, t hat, the att enuiationi for thlesec(ond portijont of the( p)ath is twice as large for thle miixedl pathI 2 case as that.for the i xed pathI I case.

In coimpa rintg lig.s 51) and 61b, however, it is apparent, that, there are 11onlifferneiiccS ill mtagnitudne values atni the regression linie inttercept ainl slopeshav~e thle Sautte value ill hoth casesi ,%weit t hough Ihle ti xed-path ItGreeni's full(--iotts its(-n for M, calcu ilat ion putrposes (Ii ffere I withi respect to a ttenu tat itn

for the two cases. T[he effect of the p~at h correct iont, besides redutcinig the dlatavariance as dlescribedI ab~ove, is to niegate the effect of differences in attentia-tion betweent thte two models. To obtain stab~le, robutst, M, values with thismet hod it is best then to utse mixed path generated svitthet ics in contjunctiontwith pathI correct ionts for thle 400 NI, mteasuremnts. IThe variantce attioitg thlemnixed path Itased Mi, valutes for the network is smnaller thant that, when I, isdleriv~edl frutut sintgle patht synit het ics, so that mtagn itudi~e mteasuiremiets wvil Ibe)mtore coniisstntt, when t hey are dleterinied front ittixetl pat 1t sylit biet ics. 'Thisis particularly niportanit for events with few report ing stat ions. All furtherplots of MI, Ii thlis study use- values obtainted front Ite ixed pathI I case withp~at h correctioits. unless stated othterwise.

f low well tIte fintal NI, %-allies reflect the actual seismhic. tagrtit uiv of these'eents ie('sst tatcns havitg anrot her imeasutre of thiei r size. Ii the evenet t.ofanontaloui'y hi gh or low selinic source coupling, for examtple', hoth body0(waVe's and~ surface waves shoutld be affected sirniilarlv Iby then (oil 'intg effect.

59

A nagiiitiide paramiettr indl)ent of s'ismic of ohs(,rvat1)ns w(uld be us(-fill to plot. he M.,, against, so we have also lifted our resillIts I() e'stiiiat .dlog yields. Fig. 7 shows th(l relationship. Yield values a i, estimi ated to bewithin 10 percent of the actual yield (Springer and Kinnaman. 1971 ). Yieldinformation was available for 97 of the events, thus yields make-up the mostcomprehensive data set, to compare our resuilts to as well. Il( yie ,lds for this

data set. range over three orders of magniitude in size. h'lh(e great(,st scatter.as in the case of MI vs. log yield, is (e to shots above Ite water table. Itshould also be be kept, in mind that the scatter would be furt her reducedif the data were separated into populations based on lheir location at NIS.(i.e. Pahiute Mesa, Rainier Mesa and Yucca F'lat). Becaulse of lie classifiednature of some the yields, it is not possible, here, to closely (,xamile thieseeffects with respect to yield.

Since our magnitude values are based on theoretical conktiniental sIruec-tures, as well as the particular network used, we wanted to coimipare our M,values to those o)taine(d from standard M, met hods. One such sl an(lard data

set, is that of Marshall t al (1979). There is a overlap of I I (,\lits betweiistudies. We preformed a fixed-slope (slope= 1.0), regression of oir M, valuesto theirs. Fig. S shows that the correlation is very good; scatitr is small forevents a)ove an(l below the water table. It's important, to not e t hat witlour met hod we are able to measure M, for events 0.75 units sim aller t liai thesmallest Marshall values. Ve are able to measure M!, for these sinaller ('veilIs.because we arte ail( to make use of near-regional ( < 500 kill) re'ords withtrhei method dlescriled in I his paper. ''lh(- offset in lM, ietwe(,n s'aies is 1.531witrh iea-i standard error of 0.013 magniitu (le units. This olfst is 1ue, ill partto te difference in definit o of ,M.,. At .100 t lhe offs(t ill mhaghmi tide is 1).15.tIis reducing th(' offset to o.38 i units. llowe'v(,r. for the ' ho(,I (1 (hecrilI edin this study, M, is based iipon a theoretical nelwork a erau v .1,. so it willhave a bias attached to it which is (l('pendel iiponl the Ih( ,\VwOk lse(I. I'hisnetwork bias can be assunme(l to be r('sponsible for ia rt of the (oIfs(,t, as well.

Tale 2 lists the, iinal ilxcd- pat i, pat h--correcld (,M i, vaudi(s for the 1(12evvits of this stIIdY. I' , li-e l f o i iiwiii lists 114 the IItIII'I-, of stlll i'( ., o litt1I he4 e'yeiit. Next are given I lie surlface %Wav4' Iiiagint ide aiiui ;1'4(wiate ('Ii('14 Ifor the ,w'et are given. Next is a thr'('e ' htler shot iforina t (, ,(h'. I'lh'first let]er deiot.es its g(,ographiic local iol: Y'wca (Y). Pa ic ( P). lR ai ;ier( o), or Climax Stok (C). 'Tli( s'cond is whethelh(r Its shot ,Iil, was al,,v,'(A) or below (11) the water lable. The last l('te, drscril he, the4 sh,,t sit( rock

6O

as t uf (TI). r Ii *volite (it). granite (G), or alluviumi (A). Art undesr-score meanrsthat I he iniforma tioin is riot known. 'Fle final t.wo r'ol ii n s are tble evenit'sri4ariie( arid J1irlatt (data, respectively. The events are listed in chronologicalorder.

To det eri ne the port abilt iy of this Ml, c:alculationi method the eventsneed to be separatedl into groups based oii their source regions andl~ thencompared. one group to( anrot her, in order to see if t here are systemiatic (hf-fererices lit M,. vatires relative to any otlher mnagnitudle scale. Ilire, mrajingeograph ic Csoiirce regions comipjrise th le event, (tat a set: Pair rie Ni esa , hai ri icrMesa and Ynucca Flat.

Whet her or riot a shot ocurs withinn satunraitcd( rrat erial is aniot her ('ritA'-rion by which to separate events in order to look for syst emratic dIifferencesinl M, values. Other stuidies have found significant seismiic Coupling differ-enices btern(plsoi eor a d bove arld below the waler table (Gupt a,19). so it a reasonable p~arameter t~o study. Reviewirng lFig.s 2 andh 7, it isalso apparent that for shot s fired-off below the water table have a largerseismTic rmagiiit udle than t hose (detoniated above the wvat er talble.

Fig. 9a shows the relationship betweenA M., vs. [ji1wall 1 b1 for all, NTISevents. 'lie surface-wave niaginitrides were all ('alcrrhatel rising mnixed-pathGreen's fnrwt ions and p~at h corrections. Fig.s 91b and~ 9c (divide the datapopuilat ions irnto above and~ below the water table, respectively; shots forwhich water table informuat ion was niot available were left out. Thelire is inoappreciable dliff'erence bet ween the above water table and~ below water tableciirves. T[hiis isTnot. srirprisinrg: referiig to Fig. 2 arid 7it is ajpareit.that, bothIsei srmic in aginit in e-p eld ctirves show t.hat for a. gi vein vieldl air eent. belowthe water tahble has a larger iragni ttidle than a shot ab~ove the water table.It follows t hat thle NI, - n1uh relat ionishiip may niot show the same discrepancybet w(x'ri shots d(etonated above arnd below the water tab~le. for the effect, ofhie water content. in the shot iiediumni should affect surface waves arid body

waves] it tli sarnie miariner. All three regression cnirve(s ar-e sseriitiall \the ,samiewit hin I he error b~ouinds. T'here is considerable scalter' iin all three fi gures, but.that is riot suirprising considIering the diversity of thle sampled populations.-I'veri wit It t his scatter. t lie, best-fitt ing Msint, line is well const rainedl, for the

pop~ulat ion covers a wide range of magnitudes.Fig. 10ta gives thre NIl'-1run, relationship for allI Yucca events. '['li regression

curve is sigriifcantly different, frorm that of Fig. 9a. '[he scatter iin thle (hat ais re(Ired by 25 percent, over that, of the general p~opurlat ion. Separatinug

61

the events with respect their relation to the water tabl yields Iwo (list cICurves, unlike the case where shots from all sites are grouped t oget her. lForthle Yucca event~s below thle water table (Fig. 101)) the regression culrve( fitis, within the errors, is not greatly different from the case of all NTS eventsbelow thle water table (Fig. 9b). [In Fig. 9c, for shots in dlry mediumn, theMe-Mb curve is significantly different from thle NTS above water table curve(Fig. 9c). The significance of this curve is questionable, however, due to thepaucity of data used to establish it.

Fig. I la plots all Pahute event M,,'s vs. their respective ni1 h 5 . I'lie re-altionship is essentially the saiiie as for the aggregate NI'S plo~t ( Fig. 9a).Fig. 111) shows the relationship for Pali ite shots below the water table. 'Ilmeslope of this curve is nearly the same as that for Yucca below water tableshots, although the intercept differs appreciably. Thius result iriplies that fora given mb, surface wave magnitudes for events at hPaliuite Mesa are largerthan those at Yucca Flat. Fig. 1 Ic shows the M8 mi, relationshipl for Plihuteevents above thle water table. Because events in this categor 'y are clusteredaround rill, = 5.5, a well constrained line cannot be obtaiiied. so we app1liedla fixeol-slope regression, using a slope of 0.82 (that of the below water ta-ble case). Comparing Fig.,. I lb and I Ic, an off-set, in MSm)of 0.09 with amean