View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Kyoto University Research Information Repository

Attenuation of Coda Waves in the Source Area of the 1990 Title July 16 Earthquake,

Author(s) KANAO, Masaki; ITO, Kiyoshi

Bulletin of the Disaster Prevention Research Institute (1992), Citation 42(2): 31-51

Issue Date 1992-06

URL http://hdl.handle.net/2433/124987

Right

Type Departmental Bulletin Paper

Textversion publisher

Kyoto University Bull. Disas. Prey. Res. Inst. , Kyoto Univ. , Vol. 42, Part 2, No. 365, June, 1992 31

Attenuation of Coda Waves in the Source Area of the 1990 July 16 Luzon Earthquake, Philippines

By Masaki KANAO and Kiyoshi ITO

(Manuscript received on December 15, 1991)

Abstract

Coda Q values are determined by using small aftershocks of the Luzon earthquake of July 16, 1990 in Philippines (M = 7.8) recorded by temporary observations. The observations were conducted from August to December in 1990 as a cooperative project between the Disaster Prevention Research Institute, Kyoto University and the Philippine Institute of Volcanology and Seismology. The temporary stations were set at 21 sites, in an area of more than 150km along the Philippine fault and around the city of . 16-88 events, having S-P time less than 5.0s were analyzed at each station. The single scattering model of coda generation was applied to the band-pass filtered RMS coda amplitudes in five frequency bands from 2 to 32Hz. The time window was taken for all the events from more than twice the arrival time of S wave to 25s from origin time, which indicates that the scattering area contributing to the coda excitation is within a radius of approximately 45km. By fitting the coda Q to the power law, Q,=Qef, the averaged values of coda Q and frequency exponent n were obtained as Q,,= 68, and n=1.06 around the city of Baguio. On the other hand, Qe= 165, and n=0.84 in the southern part of the fault where the rupture of the main shock initiated. The lower value of coda Q and the higher value of n in the northern part suggest the existence of large heterogeneities, such as small cracks in the crust made by active aftershock activity around the city of Baguio. As for the southern part of the area, fault movement is considered to be smooth enough not to develop more cracks than in the northern area. The frequency dependence of coda source factors was determined to examine the difference in site amplifications according to respective geological aspects at each observation station. The temporal change in coda Q and coda source factors during four months after the occurrence of the main shock were not obvious when considering their errors.

I. Introduction

Coda parts of S waves are considered to consist of S waves scattered by many heter- ogeneitiesdistributed randomly in the lithosphere7' 8' 25' 26' 33) The decay rate of coda amplitude with time shows the attenuation property of the upper lithosphere. The quality factor determinedfrom the decay of coda amplitude with time is called coda Q (Q). Coda Q values have been determined in various regions in the world' S,6' 9' 10, 17, 20, 23, 27). Regionaldifferences in Q, have also been revealed by the studies of n in the formula of Q= Qofn, where f is frequency. The relationship among tectonic features, seismicity and coda Q has also been pointed out by analyzing coda waves. For example, Jin and Aid 14)searched the regional variations in coda Q for the oceanic lithosphere. They found that younger oceanic lithosphere has a higher frequency exponent n value than older one, which is similar to the continents where the active regions have higher n values. Thus coda Q and n values are 32 M. KANAOand K. ITO effective in determining the attenuation characteristics and their relation to tectonics in studied area. In and around the Philippine Islands, geophysical studies of crustal velocity structure for the Manila Trench and its fore arc basin were carried out by means of multichanned seismic reflection prospecting11,18). Attenuation of earthquake—shaking intensity with epicentral distance in the Philippines was determined from the data of 83 earthquakes 34). The result was quite similar to that in the San Andreas fault region. On the other hand, Acharya 2.'3) suggested that the attenuation of ground motion in the Philippines is less severe than in the San Andreas area. Since then, no detail analyses of seismic waves have been performed to reveal the structures of velocity and attenuation in Luzon Island. In particular, Q values have never been determined in the Philippines by using small local earthquakes. On July 16, 1990, a large earthquake with a magnitude of 7.8 occurred in the central part of Luzon Island. The event was a large shallow intra—plate earthquake common along the Philippine Fault, which, in fact, has caused many large historical earthquakes in Luzon Island. The ground ruptures by the earthquake extended about 150km from Gabaldon (GBD) to Digdig (DIG) as shown in Fig. 1. The seismic strong motions caused extensive damage along the Philippine Fault and around the city of Baguio. In order to study the distribution of after- shocks in detail, the observations of aftershocks were conducted two times as a cooperative project between the Disaster Prevention Research Institute, Kyoto University (DPRI), and the Philippine Institute of Volcanology and Seismology (PHIVOLCS). Temporary observation stations were set along the earthquake fault and in the vicinity of Baguio City. Many aftershocks were recorded and about 200 hypocenters were determined accurately by Ohkura et cd.21). Many records of the aftershocks have sufficient excitation of coda waves to determine coda Q. Seismograms of these temporary observations were used for the determination of coda Q in the crust and upper mantle. The records at 13 observation stations were used for the analyses of coda waves. The regional difference in coda Q values in the aftershock area of the Philippine earthquake is discussed in relation to the aftershock activities. Coda source factors were also determined from the data. Under the assumption that the averaged value of coda source factors of many events at a station indicates the site effect near the station, the frequency dependence of coda source factors for all the observation stations are studied to reveal the relationship between the frequency dependence of coda source factor and the surface geology. Furthermore, site amplicication is discussed on the basis of the coda source factor. Temporal change in Q,, is also discussed.

2. Data and Analyses

2. 1 Data Seismic wave forms obtained by the two observations of the aftershocks of the 1990 Philippine earthquake of July 16 were used for the analyses of coda Q in this paper. The first observation was conducted from Aug. 30 to Sep. 4, 1990. Twelve temporary stations were set along the Philippine Fault and in the vicinity of Baguio City. The Locations of the observation stations are listed in Table 1(a) and shown in Fig. 1 by solid triangles and Attenuation of Coda Waves in the Source Area of the 1990 July 16 Luzon Earthquake, Philippines 33 . , 1 t I 11– N i 17.0 N la- - TAG I I 50kin SAB/BUR _ A I\ ATK/AT2 PUG A-1 BAY A BAG * M0.0 1 CP1AM

16 .0 N —ss''-DIG CAR d

A APB1/PB2

11111 i Phillipine 1\47.8 II,,\iiki.. LUP`-) "Dfault • • ...Zi: rur; II TAL A pity f Io,G. . FMSGBD !111\11 \ I I l% 120.0 E 121.0 E Fig. 1. Map of Luzon Island showing the locations of temporary stations. Solid triangles show the stations set for the first observation. The open triangles show the stations set for the second observation. The half solid triangles show the stations set for both obvervations. Stations SAB, ATK, and PB1 were moved to more adequate neighboring places BUR, AT2 and PB2, respectively,during the observations. The large star shows the epicenter of the 16 July 1990 Luzon earthquake (USGS), The small star shows the epicenter of the largest aftershock. The thick lines show the faults. half–solid triangles. The seismograms analyzed were recorded with high–gain vertical–component short period (2Hz) velocity–type seismometers at all stations except the station CAR. A horizontal–com- ponent was also set at CAR as the vertical one. Three types of records were used for the analyses. One is digital seismograms recorded in floppy discs at nine stations, another is digital seismograms recorded on audio magnetic tapes at stations GBD and CP1, and the other is FM records at the station CAR. The seismograms at CAR were digitized on a personal computer for the following analyses. Details of the recording systems are described by Ohkura et aL21). The second observation was conducted from Nov. 17 to Dec. 7 by staff members both of

S ' \ 34 M. KANAOand K. ITO

Table 1. Locations of observation stations ; (a) stations of the first observation, (b) stations of DPRI in the second observation, and (c) stations of PHIVOLCS in the second observation. The abbreviation codes corres- pond to those in Fig. 1.

station code lat. (N) long (E) Height (km)

(a) Palayan PLY 15.5865 121.1194 0.090 Gabaldon GBD 15.4501 121.3354 0.125 Talavera TAL 15.5999 120.9713 0.045 Fort Magsaysay FMS 15.4363 121.0758 0.100 Pantabangan 1 PB1 15.8143 121.1058 0.150 Pantabangan 2 PB2 15.8110 121.1070 0.220 Lupao LUP 15.8473 120.9499 0.160 Digdig DIG 15.9489 121.0007 0.300 CAR 15.9547 121.1055 0.265 San Manuel SAM 16.1327 120.6862 0.100 Camp One CP1 16.2353 120.5265 0.120 Baguio BAG 16.4162 120.6210 1.520 Atok ATK 16.5058 120.6806 1.824 (b) Sablan SAB 16.4965 120.4867 0.460 Burgos BUR 16.5183 120.4630 0.460 Atok ATK 16.5058 120.6806 1.824 Atok2 AT2 16.4983 120.6733 1.750 Pugo PUG 16.3321 120.4765 0.100 Camp One CP1 16.2353 120.5265 0.120 (c) Tagudin TAG 16.9180 120.4590 0.100 Bayombong BAY 16.4830 121.1300 0.320 Baguio BAG 16.4162 120.6210 1.520

DPRI and PHIVOLCS. In this observation period, four stations (SAB/BUR, ATK/AT2, PUG, BAG) were installed as listed in Table 1(b). In addition, PHIVOLCS operated stations BAG, TAG and BAY from Sep. 30 to Dec. 18 (Table 1(e)). In total, seven stations were operated simultaneously for about three weeks. The stations for the second observation were concentrated in the vicinity of Baguio City, in order to reveal the precise distribution of aftershocks in the northwestern end of the aftershock area. The locations of the stations are shown in Fig. 1 by open and half-solid triangles. Seismometers and recording systems were the same as those used in the first observation. Details of the observation and the method of hypocenter determination are given by Shibutani et a!. 31).

2.2 Coda Q Analysis The procedure of waveform analysis is as follows. First, digital waveforms were band- pass-filtered at five central frequencies of 2, 4, 8, 16, 32Hz by using Chebychev's band-pass- filter 24)with a fall-off rate of about 48db/oct. An example of a waveform of raw data and its band-pass filtered traces at station ATK are shown in Fig. 2. The maximum and the mini- mum amplitudes are shown at the top-left of each trace. Next, root-mean-square (RMS) Attenuation of Coda Waves in the Source Area of the 1990 July 16 Luzon Earthquake, Philippines 35

Raw data '' -14701414." " " ' '" " " ' '" ' ' ' '

. . I i I I I 1 I I I I I I 1 I I 1 I i i I 1 I I I ,,,i 52 03 04 as 06 07 00 09 to 11 12 Ip 11 15 1,6 17 18 tp 20 21 22 23 Band-Pass-Filtered -818 769mv fa =32Hz • 1 . 10 ,I % i I j -1'5')lc"' Inv --101044104041011414\4011104**44+1P0f0 =16Hz I1 -398 354my 1 ' ' ' ' ''''' ' . . . -441ANHANAVV4iMIMPAIMVAINWIVA"Iti""\ f0 =8Hz

I1\ foL4Hz 437", i j f 1fi, =2Hz

DI 02 ixs 04 05 06 0, 08 09 10 11 12 13 14 !5 16 17 te 19 20 2,1 a 23 1

Fig.2. An exampleof waveformand time-codesignal of JJY at the stationATK/AT2 (upper traces). Band-pass filtered outputs using the digital-filterwith centralfrequencies of 2, 4, 8, 16,32 Hz (lowertraces). Theamplitudes are normalizedfor eachtrace.

amplitudes were generated for the filtered outputs in each frequency band, and the amplitudes were moving averaged over a few periods of the waves (1-3s) for each frequency band (Fig. 3). Single scattering model for the generation of coda waves was applied to determine coda Q. When coda waves are assumed to be scattered S body waves, the RMS amplitude of coda wave (A) for the central frequency f at the time t after twice the travel time of S waves (t,) is approximated as,

Ac= Ct-1 exp (—a-fot/Q), (t>2t), (1)

where C is the coda source factor, Q. is the quality factor of coda wave and t is the lapse time measured from the origin time of an earthquake. Coda Q is determined by the least square analyses,applying the formula (1) to the RMS waveform. Figure 3 shows an example of the coda part of a seismogramto which the formula (1) is fitted. The time window used to determine coda Q is shown in the figure by a pair of arrow heads. The beginning point of the time window is measured from twice the travel time of S

- i1 ]J " 36 M. KANAOand K. ITO

RMS Amplitude Time Window -1.- 552p, fa =32Hz

1-111111723m~ /11;'.11i1111!1 fa =16Hz

264Tev f e =8Hz

fo =4Hz

1 54 1MY 1 1 1 1 1 1 : ; 1 : ; f 1 o =2Hz

oz op C% op op o7 op op io 11 12 ip IA 15 16 I? 15 ip zo 21 22 23

Fig. 3. An example of moving—averagedroot—mean—square (RMS) amplitudes of the event shown in Fig. 2 at five frequency bands, 2, 4, 8, 16 and 32 Hz. The time window to which the single scattering model is applied is also indicated by a pair of arrow heads.

wave. Since events near observation stations are necessary to study the regional difference in coda a local events with S—Ptime less than 5s are selected and the beginning of the time window is more than 12s in the present analyses. The end point of the window (lapse time) was fixed at 25s. The S/N ratio of all events was more than two by the end point of coda analyses. According to the single scattering model, the area contributing to the coda generation is approximated as,

r= VSt/2, (2)

where r is the radius of the scattering area, Vs is the velocity of S wave, and t is the lapse time. In this case, we assume that stations were located quite close to the epicenter, so we selected events with S—Ptimes within 5s in the actual analyses. When the lapse time is taken to be 25s and the velocity of S wave is assumed to be 3.5km/s, the scattering area contributing to coda excitation is approximately 45km in radius from formula (2). Frequency dependence of coda Q in a frequency range higher than 1Hz is roughly given as,

Q= Qof, (3)

where Qo is the coda Q value at 1Hz. By studying the difference in the parameter n in the various regions, the characteristics of seismic—waveattenuation in the crust and upper mantle Attenuationof CodaWaves in the SourceArea of the 1990July 16 LuzonEarthquake, Philippines 37 were revealed in relation to tectonic activity 14). The value n is determined by a least squares method to study the regional difference in it

3. Regional Difference in Coda Q

The resultant values of coda Q and their standard deviations at five frequency bands for each station in the first aftershock observation are listed in Table 2. These values are the averaged ones for all the events at each station. The exponent of frequency n, the averaged value of S—Ptimes and the total number of events used for the analyses at each station are also listed in Table 2. The stations are divided into three sets for convenience in Table 2 ; (a) stations in the northern part, (b) stations in the central part and (c) stations in the southern part of the aftershock area. The obtained values of coda Q and their standard deviations in the second observation are listed in Table 3. The stations are also divided into two sets for convenience, (a) stations installed by DPRI, and (b) stations installed by PHIVOLCS. Since station ATK was moved to AT2, which had quieter surroundings, on Nov. 27 in the course of observation, the values at the two stations and the totally averaged value are also listed in Table 3. The averaged coda Q values over the entire observation period at station BAY are listed as well as those divided into three observation periods, (1) from Sep. 30 to Oct. 22, (2) from Oct. 22 to Nov. 13 and (3) from Nov. 13 to Dec. 18. The number of data at BAY is about 300 and enough to examine temporal variations in Q„ so the averaged Q. values during the three successive observation periods are calculated as well as that for the entire period. The temporal variations in Q, are discussed in section 5. Stations are divided into five sets to show all data clearly and to consider regional variations in coda Q. Figures 4-1---4-5 show Q, at (1) stations in the northern part in the

Table2. The values of Q, and their standard deviationsat five frequency bands for each station in the first observation. The exponent of frequency dependencen, the averaged value of S—P time (s) and the total number of events used in the analyses are also listed for (a) stations in the northern part, (b) stations in the central part and (c) stations in the southern part.

station code 2Hz 4Hz 8Hz 16Hz 32Hz n s—p(s) N

(a) NORTH Baguio BAG 209±533 761±1839 1284±23091932±2720 2764±3777 0.88 3.32 34 SanManuel SAM 43±65 224±413 776±1261 1170±2243 1605±26451.28 3.77 23 Atok ATK 69±83 451±1286 892±1228 1665±24102692±3795 1.25 3.27 25 (b) CENTRAL Carranglan CAR 198±55 344±93 651±256 1471±575 4318±1761 1.10 ---- 16 Digdig DIG 74±90 521±1679 431±547 1213±1752 933±1152 0.85 3.06 16 Lupao LUP 218±621 296±412 1265±16691423±1622 2103±2461 0.88 2.62 30 Pantabangan PB1/2 176±322 166±223 686±1104 1363±19892167±3255 1.03 2.89 28 (c) SOUTH Palayan PLY 322±776 777±2032 683±811 1851±28941928±2297 0.64 2.88 26 Talavera TAL 1066±2023 543±684 1404±17504087±4811 4598±5513 0.71 3.49 22 Fort Magsaysay FMS 171±247 508±742 1651±23312786±3527 3449±4709 1.11 3.54 25 38 M. KANAOand K. ITO

Table 3. The values of Q, and their standard deviations at five frequency bands for each station in the second observation. The exponent of frequency dependence n, the averaged value of S- P time (s) and the total number of events used in the analyses are also listed for (a) stations of DPRI and (b) stations of PHIVOLCS. The station ATK/AT2 was moved to a quieter neighboring place on Nov. 27, so the values at the two observation stations and those averaged over the entire data are also listed. For station BAY, averaged Q, values for the entire data and those divided into three successiveperiods are listed.

station code 2Hz 4Hz 8Hz 16Hz 32Hz n s-p (s) N

(a) Burgos BUR 90±130 228±367 487±626 838±1070 1586±20961.02 3.33 33 CampOne CP1 71±94 249±371 507±972 780±912 1730±22091.09 3.58 34 Pugo PUG 77±152 268±395 316±393 1348±19952221±3201 1.20 3.15 34 Atok ATK (11/17-11/27) 127±396 275±359 640±924 1514±19521929±2694 1.03 2.72 31 Atok2 AT2 (11/27-12/9) 78±87 310±396 898±1471 1459±18101950±2630 1.15 2.01 32 Atok/Atok2(Total) (1st.+2nd. Obs.) 93±234 341±759 808±1198 1537±19972146±2949 1.23 2.61 88 (b) Tagudin TAG 76±132 228±360 483±748 972±1658 2072±3288 1.16 4.07 18 Bayombong BAY (9/30-10/22) 53±44 381±694 1083±17202219±3044 1895±2891 1.29 2.58 99 (10/22-11/13) 120±220 506±1208 1151±15822569±3593 3123±4294 1.17 2.84 97 (11/13-12/18) 68±81 366±591 1288±19091921±2741 2505±3724 1.28 3.15 98 Bayombong(total) 80±54 421±844 1172±16882199±3035 2485±3591 1.23 2.86 294 first observation (Table 2a), (2) stations in the northern part in the second observation (Table 3a), (3) stations in the northern part in the second observation (Table 3a, b), (4) stations in the central part in the first observation (Table 2b), and (5) stations in the southern part in the first observation (Table 2c). Comparing the results of Figs. 4-1, 4-2 and 4-3, differences in coda Q are not seen in the northern part of the studied area. Therefore, the regional variation in coda Q is discussed by adding the results of the three datasets. Figures 4-4 and 4-5 are similar to each other, reflecting the fact that no regional difference in Q was detected between the central and southern parts of the area. The Q values in the two regions were, therefore, added and the difference is discussed in the following. Resultant frequency dependence of Q.-1 values are shown for the northern and southern parts of studied area in Fig. 5. Thick lines in two figures indicate the averaged coda Q values at all stations for each region. Solid circles show the locations of the stations in the northern part, and solid squares and triangles show those in the southern part. Figure 5 shows that a-1 values in the northern part are larger than those in the southern part, particularly in the low frequency range. The averaged values of coda Q and frequency exponent n from the power law in the formula (3) are Q0=68, n=1.06 in the northern part, and Q0=165, n=0.84 in the southern part, respectively. Epicentral distribution determined by the first observation 21) is Attenuation of Coda Waves in the Source Area of the 1990 July 16 Luzon Earthquake, Philippines 39

iiiir 11111 11111 -1 C1 NORTH,„7,.,- t" .u e IQc 2 NORTH . Camp!Burg"c 3 NORTH payombongTagmlin oRtok 2nd-Obs. 0 pup 2nd-Obs. 0 Rtnk2 o Rtok 10 10 — 10—2 —

_3 101 0--1-3--

11111 11111 2 4 8 16 32 2 4 8 16 32 2 4 8 16 32 III!! 1 I I I EREQUENCY[Hz] 1 UC 4 CENTRALgdCarranglan CPalayan Talavera o Lupao 5 SOUTH Ft.flagsaysay Pantabagan -2 10 — i —

".^ -3 -3 I — — 10 —

I I 1 1 I I I I I 2 4 8 16 32 2 4 8 16 32 FREQUENCY[Hz] FREQUENCY[Hz] Fig. 4. Frequency dependence of a-1. Each line shows the averaged values at each station. (1) stations in the northern part in the first observation, (2) stations in the northern part in the second observation, (3) stations in the northern part in the second observation, (4) stations in the central part in the first observation and (5) stations in the southern part in the first observation. also indicated for comparison with the aftershock activity. Figure 6 shows aftershock distribution determined by PHIVOLCS from the occurrence of the main shock as of Nov. 29, 1990. This figure shows that more aftershocks occurred in the northern part of the aftershock area than in the southern part. This is also supported by the aftershock distribution reported by USGS 19). Lower values of coda Q in the northern part suggest that the attenuation is stronger due to many heterogeneities, such as small cracks developed throughout a wide area of the crust during the main shock. This corresponds to the high values of n and to the high seismicity around the city of Baguio even a few months after the main shock. As for the southern part, the movement of the fault was so smooth with the ruptures certainly being limited to such a narrow area along the fault that cracks did not form extensively in either number or area compared to the northern part. Similar results of the fault movements during the main shock were obtained from the analyses of long period seismic waves by Abe °. 40 M KANAO and K ITO

-1 1 1 I I I Q C NORTH

,:110-2—. \_

1 \ ',.. .„..,.: ..,.„.

..;,, 0.-.0.i.• e% 0 • % o 0 :::)'„-, di"'w a °°°°f3• go *o.t,,:iike, agee*MG.0 5ty.tt0 ' .,f!2' .gt,,11111 M .' 4 8 15 32 :f:.9::,i..,1.• 00 FREQUENCY CHz ]

18.0 N 4;::,::;!.i'''''. ..,:,..; 4...... ,,, a \ e ^.`ff.. .i.2.i. — 1 1II I a, Q Cu,..i.i1;.:i:,-'1b. ,Fi:). i N skco...,7 SOUTH4-,,1,47. lirA;.. 8.0 -10- -\ A e3 „ 0!:',.i0 , 0.•I ,0r\ :1.11k50km ". .:,c, 110^••

-3 lat I 1 0 — \ .... - _ 121.0El NIAiih:

, I I t [ i 2 zi 8 16 32 FREQUENCY [Hz ] Fig.5. Frequencydependence of Qc.-I in the northernand southernparts of the aftershock area. Data in Figs.4-1 to 4-3 are addedand shownin the figureof the northern part. Whilethe figurefor the southernpart consistsof data in Figs.4-4 and 4-5. Thick lines showthe averagedvalues for the northernand southernparts. Solidcircles show the locationof the stationsin the northernpart and solidsquares and trianglesshow those in the southernpart. Epicenterswere determined by the two temporaryobservations (after Ohkura et al21) ., and Shibutaniet al.")). Largeand smallstars show the main shockand the largestaftershock (USGS), respectively.

i — 4 Attenuation of Coda Waves in the Source Area of the 1990 July 16 Luzon Earthquake , Philippines 41

19°N, i 1

,..

,...

I, t .4`--- • % 18°-,1• • ' • • -, • • 6 • • \

•.1 • ilitat • • r'''• ,0 • A,,, III i • • A_,e•I %.. 1 •et 4 , t mie lia 't• IP • 1 A s'iV:• I8 A , .•%i , : , ^ 11ss . I 1 iv."' I . • •• 1 ti ,upgre -Itr_II4 I.4vi I i,./ I.1.1 I 16° — Is 1 1. • , • awalke Nit s • le /I r ° — _ a ^ . I 1 . ^ % Nr: ,11 jr, . • , ;^.§ . 1 „in?in., 1), 9 •• • • I• •II1114• I•V II•V •I, • • i— ^II\\ 2% 15°i • , ,•, , ,,\— i.I • • I% ,N.,IZZ;)• I^• • 1 ^ % • • I \ • •

i 14° , , ... I 1 .1\<1.-r-IrP1- 119° 120° 121° 122° 123°E

Fig. 6. Aftershock distribution for the 1990 July 16 Luzon earthquake as of Nov. 29. Epicen- ters were determinedby PHIVOLCS.

4. Source factor of Coda

In this section, we discuss the coda source factor (C) in the formula (1) . The value of C can be calculated simultaneously with the determination of coda Q for an event at a station . The coda source factor indicates the amplitude at the origin time extrapolated from coda part of seismograms. C is considered to include the effects of radiation pattern , propagation path, and site amplification at an observation station. Aid and Chouet 8) showed the differences in

, I,• % P r • , \ 42 M. KANAO and K. ITO

1 I 1 I I

`11111 101—A A _ 4._ 0-„, .,,...... ,e-,:z---::--:--- i a- 4)7/A7kr#\- 11111 Iif-'47190-19%-likikAiry:4 ...';:%litp ,,,, 401Wsk.;;;,,,,ItFar":0---' trii&\,,,*4-17-.4^ ' ,-- ifif.N.V.T.41P-- ''' 1111- 4f ,444 , L upao-2 i 0 — .' tam,-----,Nr— 101— 4 'Sit fr3e: — I I t ----,. ...7.-1:: FREQUENCY2 4 8 16[H 32•,.7.•.,..01)...... 1.i.H,.._...,.-,--...,i P ant abittig an ..•iritAt z 3iV-ifr 4 -44/40"- ...ott,‘.:,. I I I I I 1w_irj„,,.....-- _ ,, .:1i, .2 4 8 16 32 ..r, FREQUENCY El-iz l I 1,.,:z:94,::::.,' ilkI Pal ay an

'' ....'-' .

...,.. s, \ \ 1$ 1 1 . 16.0 N,'' .11111 ..1,';-.1'. .-.--,..,...,. \ FREQUENCY2 4 8 16 [Hz] 32 PB1/PB2 :C. LUP IP 7.!;.,, . f!-:... t._._ M7.8.::.N- 10 .____--O ...... > TA L PLY /00.T.--.—-----_s•,/ 1-02 -'l/r4r----a''''4%, _'''..., - F MS ::. N \ , i 5 U km • Talavera 11111 1 1 1 1 1 2 4 8 16 32 121.0 E FREQUENCY EHz3 1 10— AA- 'It llre.4b. 11-2I 0.6...... /.-v....z

1 fftt:o-i- agis ays ay , 1 I 1 1 2 4 8 16 32 FREQUENCY [Hz] Fig. 7. Frequency dependence of coda source factors at five stations in the southern part. Solid circles are the station locations and the star shows the main shock (USGS). The results of all events are drawn for each station. Thick lines show the averaged values at each station.

frequency response of the coda source factor between California and Japan. Iwata and Irikura 13) tried to separate site effects from source and path effects of S waves. Phillips and Akin) tried to separate the source and site effects from coda level in California.

1 Attenuation of Coda Waves in the Source Area of the 1990 July 16 Luzon Earthquake, Philippines 43

1111, It 0 0 L-/ 1...- _

IXi°1 2-..., „..- 11111 C.,1-- / /4*..47/1A-‘11‘ = , \ U14J112?—, ^ 7 / Alii.k A_ . = 0 11%1'-44"ti„, — %pa In N _\ 14:1- •./107-;ip,,. .' Ill.' ^, TAG 1± ,.I'-:_' ,'%,,,,-:\''' \, SIN .,14>A,.- 27REGIJE 4 8NCY 15H z32 I 1-62_Nei-..7i4.*g' 411/k% *-411AT _SA.B/BUR1 K/AT 2 BAY tY VillIrIIII Baguio Ifrilu nPUG BAII5131cma -sr _____ ii I . , 1 - _ .,. CP1 ; 2 4 9 15 32 F2RE4QUEBNCY16CHz32]Alk .SAM ,.FREQUENCY [Hz] '...'AV. . ",;. i i I I .7 1\ , I I I 1 I - Z\ 1TaI—-AMer.,-Alk . 4014'Z ,,— 'N- etri_si''' 4^ "".40%^*—'e — :gAN„,...... ,,,: --, blalvo,.. ITaa—v‘‘_-;-_,L='-;,.., _i/ler poiroc---,,,,,:-ort----:.4,,::,,\,e— P ugo•: 162 _,,,,- — St ','\---71 .ManueI 1 1 I _i1 1 - Culp1 i.1^1 2 4 816 32 121.0E2 4 9 15 32 FREQUENCYEFizi 11111 FREQUENCY[Hz ] 2 4 8 IS 32 FREQUENCY1Hz ]

Fig. 8. Frequency dependenceof coda source factors at six stations in the northern part. Solid circles show station locations and the large and small stars show the main shock and the largest aftershock (USGS), respectively. The results of all events are drawn at each station. Thick lines show the average values at each station.

We calculated the frequency dependence of the coda source factors for all the observation stations, then tried to reveal the relationship between the frequency dependence of source factors and surface geology near the stations. In the present anaylyses, the path effect of the medium is assumed to be eliminated by fitting a single scattering model of coda generation. The effect of radiation pattern is assumed to be eliminated by averaging many events from various directions. Figure 7 shows the frequency dependence of coda source factors at five stations in the southern part. The locations of the stations are shown by solid circles in the map. The results at each station for all events are drawn in the figures. Figure 8 shows the frequency dependence of coda source factors at six stations in the northern area. The results at stations ATK/AT2 and BAY are shown in Fig. 9 for three different periods of observation. Since many events were recorded at these stations, any temporal change in coda source factor could be examined as well as that in Q , . However, no clear change was seen in the figures. This is discussed in section 5. Thick solid lines in the figures of the coda source factors (Figs. 7, 8, 9)

B 44 M. KANAO and K ITO

(a) CE 9/30-10/22\ 401‘ ow.- cz ...ia"..,q -11/13 IPA1/13-12/18I u4,3..10::.; Wr •Atill*rr'N._ 1--4"--"gAw.--7, / rr" viN 1"-- -1 L ,Itv::-e.t-7-e4.4.-.,.-.17"y:.c_ ° 10 ?k;-.i4v,,;--;,z,,:^,-''-1 ik,,,4'";,-,- 10 :vp';--;.---,;;,7:-:‘,.:1/,/1'..:"7-7tC16- '1:.."-;,,,--- -,--__- I 0 1- .;, ,-,j,II._-_ ,,,..., 0 0Ar.01:;7:?;.,,Itz,',,1-- CAVA7AiS!'‘_-,------..--,_---.------'---'„..1.-,'‘,..---.--V•;,.:N.., .:_, *fr.-,,-,F.,-.-•-r-----,-''.4-`,•7-.3-,-..,:=, --.,,A, .--:,.., -. ,;-..-- -.. ------. . -, -44',;*-1`.--",-.7-7-,"'' 1fl.,flip.,--°--41ft\j-.,:-t-':1',94,1,,,t--12.:,t-k^,,_--,-t„,-__31--: 1-6 10 -2Virt*1'.9i.'"..:...2,;1-,41411—iti _ CE / At .44V—/ Ii?— "IV4,,-I( Alb, 16-to// 1 \ \ Lnr Bayombong1// Bayombong 2 Bayombong 3

I I r I I 1 1 1

(b) 8/30-9/4 11/17-11/27 11/27-12/9 1 I I 1 1 1 1 1 1 1 ' I 1 1 1

CC 1=i 0 )1,ea **A, 1,..0.0_,/,_i ...1 ..,N_,Ar4_ o—NIP4e,y---, ,,r•— to‘ArAik- Ce\i'C\:4 ,4,f"It"\\di^....„le, l' frdi, * IN, „limy, iatic4 ir_ ,.. .to, *go...4re L'CE102— 4 / \ - id - ,A, ,,.. - le-2 - :,,tott,4„,1W-; 110,.= ,.- - CE 41://RT OK itiiVtir ofit•I*-t.•:- try/10,,Fltok \// 1.1.„Fltok 2 i 1 1 r r 1 1 1 1 1 2 4 8 16 32 2 4 8 16 32 2 4 8 16 32 FREQUENCY[Hz ] FREQUENCY[Hz ] FREQUENCY[Hz 7 Fig. 9. (a) Coda source factor during three successiveperiods at the station BAY, from Sep. 30 to Oct. 22 (left hand side), from Oct. 22 to Nov. 13 (middle figure) and from Nov. 13 to Dec. 18 (right hand side). (b) Same at the station AT/C/AT2, from Aug. 30 to Sep. 4 (left hand side), from Nov. 17 to Nov. 27 (middle figure) and from Nov. 27 to Dec. 9 (right hand side). Thick lines are the averaged values for each period.

are the average values at each station. Most of the curves at each station show similar variation with frequency, regardless of their amplitude, so the averaged lines can be compared with each other to discuss their difference. There seem to be clear differences between the frequency responses at the stations. Coda source factor at PB1/PB2 seems to have a peak at 4Hz, while that at PLY is positively proportional to the frequency of 2-32Hz (Fig. 7). Those at TAL, FMS and LUP seem to be flat in the frequency range of 4-32Hz. BAG has a peak at 81-1z,SAB/BUR, SAM and BAY have rather flat response to frequency and PUG, CPI and TAG have high values in the low frequency range (Figs. 8, 9). Phillips and Aki 22) pointed out that the frequency response of site effect is related to surface geology at the stations. They mentioned that 'granite' station has high amplification at high frequency range, 'fault zone' station has high amplification at low frequency range and `sediments' station has the flat response to frequency . The responses at the stations PB1/PB2, PUG, CP1, and TAG seem to be similar to that of 'fault zone' station (Figs. 7, 8). The

- Attenuationof Coda Wavesin the SourceArea of the 1990 July 16 Luzon Earthquake,Philippines 45

Baguio -4-) C.)

181— — —\ $.4

cf) Average / — PICJ 0 C.)

2 4 8 16 32 Frequency (Hz)

Fig. 10. Frequency dependence of coda source factor at the station BAG (thick line with dries), at other four stations in the northern part (dashed lines) and the average of the four stations (thick line with diamonds).

frequency responses at PLY and ATK/AT2, have the same type as 'granite' station (Figs. 7, 9). At the stations TAL, FMS, LUP, SAB/BUR, SAM, BAY the frequency responses are similar to 'sediments'. The geological map of the Philippines (after Bureau of Mines, Manila 1963) in the paper by Hirano et al.12) roughly supports the above mentioned relation between the frequency response of coda source factor and geology near observation stations. However, the frequency response at station BAG seems to be exceptional to this classification. The response at BAG is rather flat with a peak at 8Hz (Fig. 8). When the source effects are assumed to be canceled by averaging many events, site amplicication can be detected from the coda source factors. As described above, there seems to be relationship between frequency response of coda source factor and surface geology, so the coda source factor can be related to the site effect. However, when we compare the amplitudes of coda source factors at different stations, magnitudes of the events should be considered. The magnitudes of events at all the stations are almost in the same range, since the stations are distributed over a small area such as the vicinity of Baguio City. In addition, events of the same size were selected.

— 46 M. KANAOand K ITO

Thus, we can now compare the absolute values of the coda source factors at five stations in the vicinity of the city of Baguio. Figure 10 shows the frequency dependence of coda source factors at the station BAG (thick line with circles), those at other four stations SAB/ BUR, PUG, CPI and ATK/AT2 (dashed lines). The thick line with diamonds represents the averaged value of the four stations. BAG has larger values of coda source factors than the average values of the other four stations, three times at 8Hz, and more than two times in the other frequency ranges. This suggests that larger amplification is caused by site effect at station BAG than at the other stations. This may be one of the possible factors that caused the severe damage of buildings in the city of Baguio.

5. Discussion

The averaged value of Qo and frequency exponent n for all the stations in the areas studied are 115 and 0.92 and the averaged value of coda Q at 8Hz is 841. The coda Q value is smaller than that in the southwestern part of Japan, being about 445 at 8Hz in the same lapse time of 25s '7), while the value of the frequency exponent n is 0.88 in southwestern Japan. Therefore, values n are almost the same in both areas, but the intensity of attenuation seems to be weaker in Luzon Island than in the inner zone of southwestern Japan. The temporal variations in coda duration, or that of coda attenuation involving the occurrences of large earthquakes have been reported by many authors in the various regions 15,16, 28, 29, 32) Coda Q values increased after the main shock in some cases. On the other hand, the values decreased or there were no evident changes in coda Q30)in other cases. Furthermore, if coda Q changes with time, it is very difficult to detect the regional variations in coda Q. In the present analyses, however, unfortunately there are no digital seismic wave data before the occurrence of the main shock, so only the change after the main shock could be examined. The data at BAY could be analyzed to examine temporal changes since many events of about 300 could be recorded with a digital recorder from Sep. 30 to Dec. 18. Figure 11 shows the temporal changes of Qc-I at station BAY at five frequency bands. The solid lines are the running averaged values of a-I over 10 days. The main shock which occurred on July 16 is shown by an arrow. The change in averaged Qc-1 is small, but it is not certain whether it is due to error or not. Thus, the temporal change in averaged coda Q is not clear for the observation period of about 80 days, considering the deviations. Site amplification of S waves is important, because most of the severe damage is caused by direct S waves. Therefore, it is necessary to examine the relation between site amplification of S waves and that of coda waves. Figure 12 shows the frequency dependence of the maximum amplitudes of S waves (upper figures), coda source factors (middle figures), and their ratios (lower figures). The stations are BAG (left hand side) and SAB/BUR (right hand side). The ratios of maximum amplitudes of S wave to source factors of coda are roughly constant at both stations, BAG and SAB/BUR in the frequency range of 4-32Hz. This indicates that the frequency dependence of direct S waves and coda waves has almost the same ground response. Therefore, the site amplification determined from coda waves seems to have characteristics similar to that of S waves. Attenuationof Coda Wavesin the SourceArea of the 1990July 16 Luzon Earthquake,Philippines 47

-1 111114.111.1.."1: 11111111111 QC 74 - 1 ,,'--,.. *. g: - -2 Bayombong...s-'4;t1'1?:••-2 Bayombong 10-°8• t--°•:-••• -10- - %* ° ... ., ..."-- . . • - :- g 10--3.. . -3 ... . M7.8 2Hz.10M7.8 °-- 1GHz%-:4:°°,•,' ' °I. -

".., 9°:C-4- , I 1 1 1 1 1 1 1 1 1 , 1 1 1 1 1 1 1%1 1 el 1 7 8 9 10 11 12 7 8 9 10 11 12 -i 111111111.11 11111111111 QC° °

-2 Bayombong,,,,7; : ° - •°-.. -2 Bayombong 10— ,,°.;.°Cow-..e. -10- - ..: 8 ..% e°°-z- - -3-3 -. ., 10-n°°-10M7.8 4Hz ..0...... M7 .8 32Hzjs•-.:°.°-

.• f`:,'' • °.° .. /•..11111111Th I.,:p.,i',.111111neel 2 elm., °1117 1 7 s 9 10 11 12 7 B 9 10 11 12 - ' 1 1 I i 1 1 I i I 1--1 Month( 1990) Q c -2Bayombong io-• °. •- 1 : •.° .:,, 8, - 3 . M7 0% -4,%— le - .88Hz.:..... -%,„p%a. .°. . ... e . oe 11 1 1 1 1 1 ° *1 °I, ° " 1 " •-1 ° " e 1 1 ? e 9 10 11 12 Month( 1 9 9 0 ) Fig. 11. Temporal change in (2,71 at station BAY at five frequency bands. The observation period was from sep. 30 to Dec. 18. The solid lines are the running averaged values of Qc-1over 10 days. The main shock which occurred on July 16 is shown by an arrow head.

6. Conclusion

Coda Q and coda source factor were determined along the Philippine Fault zone in Luzon, using the data of aftershocks of the 1990 July 16 Philippine earthquake recorded at 15 temporary stations. The results of our analyses are as follows : 1. Coda Q were determined for 15 observation stations by applying the single scattering model of coda wave generation. When the frequency dependence of coda Q is represented by the formula Q=--Qof in the frequency range from 2 to 32Hz, the average value of Qo and frequency exponent n is Q0=-115, n=0.92, respectively. The Q value is larger than that in the southwestern part of Japan in the same lapse time of 25s. 2. Coda Q is rather small in the northern part along the Philippine Fault than in the

, 48 M. KANAO and K. ITO

I 1 I I 1 1 I 1 1 1 "cl ci.) ,/.4,.- ..,-Aort.. r4i,/,'N''.‘`, .-",.....:_,,, 1:14-1//if\'\'- — 101—LS__— "14 10 —i I'1,,.-/,,---.1;I /I „,,,,,-.....-z.... E /// \\ //''''.±,-;^^^•.\'''^,.- •r 4 10 — 1 —10 — Illy re

Baguio Burgos CID 1 1 1 1 1 1 1 1 1 2 4 8 16 32 2 4 9 16 32 i 1 ./1I k1 1 1 1 1 1 1 i•I 0 -.PU i 4*” CIA - c) id — /7/74.<4\r,--` —1-a — A.,— ---'• — t-1u -willicA.--kreitill--,,,,i;LNK. ...7^14S. I 11-/`. wel4S1/416 .., O 0 2/1 ., ‘ 11/4-\\Ilec.,... • . 1 ;Nr/ ,//„.„„,„00,10,;4_47.0„ .4rdir,i ...4. 414 -lip,....-_:-**--- CID I fr'llf '1" .•.: qratri0.411*.-8,- ,-cfC t 1 0- 2 — - 1-02 — arterix.r."--- viAtik4Wilit,rAlliiiik- — CO ^'4,mg,;,;70--z...... „ .) Baguio i 4^11Bu u .0 t' III I I' I I I I 2 4 9 16 322 2 4 8 16 32 :-.1 1 0a I 1 1 1 1 10 1 I 1 1 1 -40 - U cc Ccc LI ,/,10' — — 101 — — C.)-, r O ,,, 01111111",..../M17111h,1...--,''41.1.M.?...:^...---- AO , „..-..^^•• , -...---^^:,...-, ../- -.. Cn1 Er-hilf`11515".":---7"'7'- 1 0° -fir.'!401117417...... AP'....0.:"..... al1/1:2Ir,1/Ig4:4-4.7^-•-,k,41,/,`-::..::----,---7-•^'."-a.,'--ille47/4,t4.---.'.-:, E-..1'„,741'..-41f, Ak' ii•'•'...11"."11%^11,6, '''"! -1 10 — 1 cd—101—,;ilitA'1""X r f— 11,41 EBaguio7, Burgos (I) 10 -2 , 1 1 1 1 1 1 -02 !It t i 2 4 8 16 32 2 4 8 16 32 Frequency (Hz) Frequency (HZ)

Fig. 12. Frequency dependence of the maximum amplitude of S wave (upper figures), coda source factor (middle figures) and their ratio (lower figures) at stations BAG (left hand side) and BUR (right hand side).

Attenuationof CodaWaves in the SourceArea of the 1990July 16 LuzonEarthquake, Philippines 49

southern part. The differences in Qc seem to be related to the aftershock activity, which is more active in the northern part around the city of Baguio than in the southern part near the starting point of the main shock. 3. Coda source factors were determined for all the events at every station. These factors include source, path and site effects. The frequency dependence of the coda source factor seems to involve the surface geology at observation stations with the coda source factor seemingly sensitive to site amplicication. 4. The temporal changes in coda Q and coda source factor during the observation periods are not evident in these analyses because of the short period of observations of only about three months.

Acknowledgements

The authors wish to express their sincere thanks to the observation staff members of PHIVOLCS and DPRI ; R. S. Punongbayan, B. C. Bautista, J. R. Puertollano, A. G. Lanuza, A. A. Melosantos, A. Chu, R. Pigtain, E. D. Cruz, Y. Iio, K. Nishigami, T. Shibutani, T. Ohkura, T. Iwata, K. Tasaki, Y. Kakehi, M. Kato and N. Hirano. They also thank those who kindly made available their houses or offices to our observations. The authors are also grateful to Prof. K. Aid of Southern California University, Dr. W. S. Phillips in Los Alamos National Laboratory, Dr. H. Sato of Tohoku University, Dr. M. Hoshiba of Meteorological Research Institute for their valuable advice and encouragement during this work. They also acknowledge Drs. J. Akamatsu, K. Matsunami, M. Ando, H. Watanabe and Y. Umeda and many other staff members of the Disaster Prevention Research Institute, Kyoto University for their invaluable help and discussion.

References 1) Abe, K.: Seismologicalaspects of the Luzon, Philippines Earthquake of July 16, 1990, Bull. Earthq. Res. Inst., Univ. of Tokyo, Vol. 65, 1990,pp. 851-873. 2) Acharya, H. K.: Ground Motion Attenuation in the Philippines, Proc. 2nd. International Conferenceon Microzonation for Safer Construction, Vol. 1, 1978,pp. 379-390. 3) Acharya, H. K.: Comment on "Attenuation of Intensity with Epicentral Distance in the Philippines"by Sergio Su, Bull. Seism.Soc. Am., Vol. 71, 1981,pp. 929-931. 4) Akamatsu, J.: Attenuation Property of Coda Parts of Seismic Waves from Local Earthquakes, Bull. Disas. Prey. Inst., Kyoto Univ., Vol. 30, 1980,pp. 1-16. 5) Akamatsu, J.: Seismic Zoning and Seismic Ground Motion in the Southern Parts of Kyoto, SouthwestJapan, Bull. Disas. Prey. Inst., Kyoto Univ., Vol. 36, 1986,pp. 1-42. 6) Akamatsu, J.: Coda Attenuation in the Liitzow-Holm Bay Region, East Antarctica, Phys. Earth Planet. Inter., Vol. 67, 1991,pp. 65-75. 7) Aki, K.: Analysis of the Seismic Coda of Local Earthquakes as Scattered Waves, J. Geophys. Res., Vol. 74, 1969,pp. 615-631. 8) Aki, K. and B. Chouet : Origin of Coda Waves ; Source, Attenuation and Scattering Effects, J. Geophys.Res., Vol. 80, 1975,pp. 3322-3342. 9) Correig, A. M., B. J. Mitchell, and R. Ortiz : Seismicity and Coda Q Values in the Eastern Pyrenees: First Results from the La Cerdanya SeismicNetwork, Pure and Appl. Geophys., Vol. 132, 1990,pp. 311-329. 10) Eck, V. T.: Attenuation of Coda Waves in the Dead Sea Region, Bull. Seism.Soc. Am., Vol. 78, 1988,pp. 770-779. 50 M. KANAO and K. ITO

11) Hayes, D. E., and S. D. Lewis : A Geophysical Study of the Manila Trench, Luzon, Philippines 1. Crustal Structure, Gravity, and Regional Tectonic Evolution, J. Geophys. Res., Vol. 89, 1984, pp. 9171-9195. 12) Hirano, S., T. Nakata, and A. Sangawa : Fault Topography and Quaternary Faulting along the Philippine Fault Zone, , the Philippines, Geol. Journ., Vol. 95, 1986, pp. 1-23. 13) Iwata, T., and K. Irikura : Separation of Source, Propagation and Site Effects from Observed 5- Waves, Zisin (J. Seism. Soc. Jpn.), Ser. 2, Vol. 39, 1986, pp. 579-593 (in Japanese with English Abstract). 14) Jin, A., T. Cao, and K. Aid : Regional Change of Coda Q in the Oceanic Lithosphere, J. Geophys. Res., Vol. 90, 1985, pp. 8651-8659. 15) Jin, A., and K. Aki : Temporal Change in Coda Q Before the Tangshan Earthquake of 1976 and the Haicheng Earthquake of 1975, J. Geophys. Res., Vol. 91, 1986, pp. 665-673. 16) Jin, A., and K. Aki : Spatial and Temporal Correlation Between Coda Q-1 and Seismicity and its Physical Mechanism, J. Geophys. Res., Vol. 94, 1989, pp. 14041-14059. 17) Kanao, M., and K. Ito : Attenuation Property of Coda Waves in the Middle and Northern Parts of Kinki District, Zisin (J. Seism. Soc. Jpn.), Ser. 2, Vol. 43, 1990, pp. 311-320. (in Japanese with English Abstract). 18) Lewis, S. D., and D. E. Hayes : A Geophysical Study of the Manila Trench, Luzon, Philippines 2. Fore Arc Basin Structural and Stratigraphic Evolution, J. Geophys. Res., Vol. 89, 1984, pp. 9196- 9214. 19) NEIC, USGS : Preliminary determination of epicenters, Monthly listing, 1990. 20) Nishigami, K., Y. Iio, C. Gtirbtiz, A. Pinar, N. Aybey, S. B. User, Y. Honkura, and A. M. Isikara : Microseismic Activity and Spatial Distribution of Coda Q in the Westernmost Part of the North Anatolian Fault Zone, Turkey, Bull. Disas. Prey. Res. Inst., Kyoto Univ., Vol. 40, 1990, pp. 41-56. 21) Ohkura, T., T. Shibutani, Y. Iio, K. Nishigami, M. Kanao, K. Tasaki, T. Iwata, Y. Kakehi, M. Ando, B. C. Bautista, J. R. Puertollano, A. G. Lanuza, A. A. Melosantos, A. Chu, R. Pigtain, E. dela Cruz, and R. S. Punongbayan : Aftershock Observation of the 16 July 1990 Luzon Earth- quake, Philippines, Recconaissanse Report on the 16 July 1990 Luzon Earthquake, Philippines, 1991 (in press). 22) Phillips, W. S., and K. Alci : Site Amplicication of Coda Waves from Local Earthquakes in Central California, Bull. Seism. Soc. Am., Vol. 76, 1986, pp. 627-648. 23) Pulli, J. J.: Attenuation of Coda Waves in New England, Bull. Seism. Soc. Am., Vol. 74, 1984, pp. 1149-1166. 24) Saito, M.: An Automatic Design Algorithm for Band Selective Recursive Digital Filters, Geophys. Explor., Vol. 31, 1978, pp. 240-263. 25) Sato, H. : Energy Propagation Including Scattering Effect : Single Isotropic Scattering Approximation, J. Phys. Earth, Vol. 25, 1977, pp. 27-41. 26) Sato, H.: Scattering and Attenuation of Seismic Waves in the Lithosphere-Single Scattering Theory in a Randomly Inhomogeneous Medium, Rep. Natl. Res. Ctr. Disas. Prey., Vol. 33, 1984, pp. 1-186 (in Japanese with English Abstract). 27) Sato, H.: Regional Study of Coda Q-1 in the Kanto-Tokai District, Japan, Zisin (J. Seism. Soc. Jpn.), Ser. 2, Vol. 39, 1986, pp. 241-249 (in Japanese with English Abstract). 28) Sato, H.: Temporal Change in Spectral Coda Attenuation Q-1 Associated with the K=13.3 Earthquake of 1983 Near Garm, Tadjikistan Region in Soviet Central Asia, Zisin, Ser. 2, Vol. 41, 1988, pp. 39-46. 29) Sato, H.: A Precursor-like Change in Coda Excitation Before the Western Nagano Earthquake (M,=6.8) of 1984 in Central Japan, J. Geophys. Res., Vol. 92, 1987, pp. 1356-1360. 30) Sato, H.: Temporal Change in Scattering and Attenuation Associated with the Earthquake Occurrence-A Review of Recent Study on Coda Waves, Pure and Appl. Geophys., Vol. 126, 1988, pp. 465-497. 31) Shibutani, T., T. Ohkura, Y. Iio, M. Kanao, K. Nishigami, K. Tasaki, T. Iwata, Y. Kakehi, N. Hirano, M. Ando, B. C. Bautista, J. R. Puertollano, A. G. Lanuza, A. A. Melosantos, A. Chu, R. Pigtain, E. dela Cruz, and R. S. Punongbayan : Search for a Buried Subfault(s) of the 16 July 1990 Luzon Earthquake, Philippines by Means of Aftershock Observations, J. Natural Disaster Attenuationof Coda Wavesin the SourceArea of the 1990July 16 LuzonEarthquake, Philippines 51

Science, Vol. 13, 1991, pp. 29-38. 32) Tsukuda, T.: Coda Q Before and After a Medium Scale Earthquake, Prog. Abst. 23rd. General Assembly IASPEI, Tokyo, Vol. 1, 1985, p. 82. 33) Tsujiura, M.: Spectral Analysis of the Coda Waves from Local Earthquakes, Bull. Earthq. Res. Inst., Tokyo Univ., Vol. 53, 1978, pp. 1-48. 34) Vanmarcke, E. H., and S. S. P. Lai : Attenuation of Intensity with Epicentral Distance in the Philippines, Bull. Seism. Soc. Am., Vol. 70, 1980, pp. 1287-1291.