<<

Ocean-bottom seismograph observations on the Mid-Atlantic Ridge near lat 37°N

T.J.G. FRANCIS* Cooperative Institute for Research in Environmental Sciences, University of ColoradolNOAA, Boulder, Col- orado 80309 I. T. PORTER Institute of Oceanographic Sciences, Blacknest, Brimpton, Reading, RG7 4TJ England J. R. McGRATH U.S. Naval Research Laboratory, Washington, D.C. 20390

This article is one of a series appearing in the April and May issues of the Geological Society of America Bulletin on the scientific results of Project FAMOUS. These studies were undertaken in the axial area of the Mid-Atlantic Ridge between approximately 36°30' and 37°N latitudes.

ABSTRACT 200 Hz. The tape speed of the recording system used by the ocean-bottom seismographs set an upper limit of about 40 Hz to Ocean-bottom seismographs were deployed twice in the FA- the frequencies recorded. It was decided to retain the original hyd- MOUS area in 1973. Few earthquakes were observed on the first rophones of the seismographs as well to be able to compare the per- occasion, but 515 were recorded on the second with magnitudes formance of the two systems. Since only four data channels can be ranging from 0 to 2.5. These occurred both in the median valley recorded in each ocean-bottom seismograph, the H-58 hydrophone and in the fracture zone adjacent to the instruments. The latter was substituted for one of the horizontal seismometers. Thus dur- events have been interpreted as activity along a single Riedel frac- ing the October 1973 deployment, each instrument recorded the ture cutting obliquely across the trend of the fracture zone. Only a outputs of the following transducers: vertical seismometer, hori- crude velocity structure could be determined on the basis of the zontal seismometer (of unknown orientation), wideband H-58 earthquake observations. A superficial surface layer with a com- hydrophone, narrowband hydrophone. Instrument problems pre- pressional velocity of 2.7 km/s overlies a main layer with a P veloc- vented the seismometer from being recorded at as wide a band- ity of 5.0 km/s and a S velocity of 2.4 km/s. The low-velocity sur- width as the H-58 hydrophone. The advantages of wideband re- face layer and the high ratio of P to S velocities for the main layer cording in seismology are now well understood, and the contrast (2.08) reflect the highly fissured, water-saturated nature of the between the seismometer channels and the H-58 hydrophone crust. The earthquake foci are thought to lie in the low-velocity sur- channel is analogous to that between the conventional long-period face layer. Finally, in spite of numerous small earthquakes, an seismometer and the broadband Kirnos instrument (Marshall and acoustical analysis of the hydrophone data indicates that distant others, 1972). The broadband instrument produces a more faithful shipping traffic is the dominant source of ambient noise on this part record of the parameter being measured. Thus, the H-58 hyd- of the Mid-Atlantic Ridge in the frequency band 5 to 32 Hz. rophone records are more faithful reproductions of the pressure fluctuations in the water than are the seismograms of the ground INTRODUCTION motion of the sea bed.

Ocean-bottom seismographs (OBS) were deployed twice on the DESCRIPTION OF OBSERVATIONS Mid-Atlantic Ridge in 1973 as part of the British contribution to the FAMOUS project. In May, three ocean-bottom seismographs Although the ocean-bottom seismographs operated for similar were launched and recovered by RRS Shackleton just south of the durations and in similar situations in relation to the rift and small fracture zone offsetting the ridge at lat 36°57'N, designated fracture-zone topography, the May and October operations fracture zone A. In October, the same three instruments were showed a gross difference in seismicity. In May, OBS-2 detected 5 launched from USNS Hayes and recovered by USNS Mizar; the events, 4 of which were also detected by OBS-3. All appeared to array on this occasion straddled the next fracture zone to the north. arrive earlier at OBS-3, but only 2 were large enough for mea- The positions of the seismographs in relation to the bathymetry of surements of arrival time to be made for both instruments. In Oc- the area are shown in Figure 1; their geographic coordinates, water tober, the level of activity observed was two orders of magnitude depths, and operating times are given in Table 1. Unfortunately, greater: 515 events were detected by OBS-2, 184 by OBS-3. Of the tape recorder of OBS-1 failed to operate, for different reasons, these, 106 were detected by both instruments, and depending on on both deployments so that on each occasion only two records the velocity model employed, it was possible to locate between 50 were available for interpretation. and 60. The use of the wider-band H-58 hydrophone improved the The instruments themselves have already been described in the detection threshold for the October operation but did not alter the literature (Francis and others, 1975). In May 1973, the specifica- overall picture. Of the 515 events detected by OBS-2 in October, tions for the ocean-bottom seismographs were essentially un- 342 would have been detected by the seismometer channels alone, modified from those during their deployment on the Mid-Atlantic operating at the same gain settings as in May. A plot of the cumula- Ridge at lat 45°N the previous year (Francis and Porter, 1973). In tive number of seismic events detected by each of the two instru- October 1973, however, the seismographs were improved by the ments as a function of time is shown in Figure 2. OBS-2 was a more addition of U.S. Naval Research Laboratory type H-58 hyd- efficient detector than OBS-3 because the former was both instru- rophones that have a flat response to pressure in the range 0.5 to mentally quieter and closer to most of the activity. The detection rate at OBS-2 varied between about 50 and 200 events per day, but over periods of a few days at a time the rate appeared relatively * Present address: Institute of Oceanographic Sciences, Blacknest, Brimpton, Read- ing, RG7 4RS England. constant.

Geological Society of America Bulletin, v. 88, p. 664-677, 18 figs., May 1977, Doc. no. 70505.

664

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 32°30'W

37° 20^- - 37°20'N

36° ¿tfN - 36°40'N

32 30 W Figure 1. Bathymétrie sketch map of the northeastern part of the FAMOUS area showing positions of ocean-bottom seismographs. Bathymetry from unpublished chart by J. D. Phillips, Woods Hole Oceanographic Institution. Depths greater than 1,500 fm shown by vertical lines; those less than 1,000 fm have cross pattern. Positions of three seismographs in May 1973 shown by Roman numerals; the positions in October 1973 shown by Arabic numerals. Solid circles indicate teleseismically determined earthquake epicenters for the period January 1963 through October 1974.

TABLE 1. PARAMETERS OF OCEAN-BOTTOM SEISMOGRAPH STATIONS

Instru- Lat Long Water Recording on Recording off ment (N) (W) depth (Greenwich (Greenwich (m) mean time, mean time, date) date)

OBS-I* 36°54.0' 33-13.2' 2,896 (Tape recorder failed to 25 EVENTS / DAY unclamp) in 2 obs-ii* 36-49.3 ' 33-16.0' 2,640 16:00, 18:00, LU > May 11 May 19 iLl li. OBS-III* 36°48.7' 33-10.1' 2,022 16:00 18:00 o May 13 May 20 Iloci CD OBS-lf 37°04.2' 32°53.8' 3,079 (Tape recorder jammed 12 min z 3 after start) Z OBS-2f 37-00.1' 32-53.25' 2,480 21:00 05:09, LU Oct. 13 Oct. 21 < OBS-3f 37-08.3' 32°51.8' 2,424 16:00, 06:00 Z Oct. 14 Oct. 22 <_>

* From RRS Shackleton deployment, May 1973. fFrom USNS Hayes and Mizar deployment, October 1973.

Figure 2. Cumulative number of earthquakes detected by OBS-2 and OBS-3 in October 1973 as a function of time. The number is plotted at 2-h 286 287 288 289 290 291 292 293 294 295 intervals. JULIAN DAY

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 666 FRANCIS AND OTHERS

r +5 Pa | H . i ill

- - 5 Pa tj +0.5 Aim r P

t ' ^ III!. * ^ '

nnnrTTnminrTirT^ B I is Figure 3. Set of four earthquakes detected by OBS-2 in October 1973. In each case H is the wideband hydrophone trace filtered from 2 to 40 Hz. Z and Y are the vertical and horizontal seismometer traces, respectively, filtered from 1 to 12.5 Hz. Equivalent ground displacement on the seismometer channels is shown for the peak magnification at about 8 Hz. The response of the hydrophone channel to pressure is effectively flat within the filtered band. Each event is identified as follows by the time to the nearest minute and the Julian day: (A) 03:18/290, (S - P) = 1.42 s; (B) 18:30/293, (S - P) = 2.49 s; (C) 18:08/293, (5 - P) = 2.68 s; (D) 15:42/292, (5 - P) = 3.03 s.

A small selection of the records obtained by OBS-2 and OBS-3 events [(S — P) > 2.5 s], T phases occur that have propagated are shown in Figures 3 and 4. The same nomenclature is used as largely through the water. As at lat 45°N, the S waves were consid- that for the ocean-bottom seismograph observations along the erably more energetic than the P waves, so that for the weaker Mid-Atlantic Ridge at lat 45°N (Francis and Porter, 1973). In addi- events, only the S waves emerged above the noise on the seismome- tion to clear P and S phases, compressional waves reflected from ter channels. A few records showed good evidence of the conver- the sea surface (R„ R2, . . .) are present, and for the most distant sion of compressional to shear waves and vice versa. This is par-

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 OCEAN-BOTTOM SEISMOGRAPH OBSERVATIONS, MID-ATLANTIC RIDGE 667

+1 Pa H

-L-1 Pa ì t tp tp ' T+0.25//m K, r: f -L - 0.25 /um -r +0.25 Mm ill

---0.25Mm rrvTrrxJuinrrTTTTT^ C

-r+25 Pa H

25 Pa f t i

-1-+2 Mm P P R, I 1

-L -2 Mm

• +2 Mm

Y *

-L -2 Mm nnrrrrunnrrrr^ D ifs Figure 3. (Continued).

ticularly marked on the record shown in Figure 4 where P —* S, and from teleseismic observations (Sykes, 1970; Francis and Porter, possibly 5 —» P conversion as well, has taken place relatively close 1971). Surprisingly, perhaps, the same characteristic appears to be to the seismographs. present on the much smaller scale of microearthquake observa- In addition to those mentioned above, phases were identified on tions. As at lat 45°N in 1972, a number of identical sets of records the October 1973 records that were not apparent on the records were obtained in October 1973. The largest of these was generated obtained in 1972 from lat 45°N. These were apparent only on the by a swarm of 18 separate events occurring between 18:28 and hydrophone and vertical seismometer traces and have been iden- 23:50 on Julian day 293. This particular burst of activity is tified as later-arriving compressional waves. They are labeled P', reflected in the steepening of the slope of the OBS-2 plot at this time P", . . . when impinging on the sea bed from beneath, R/, R,", . . . in Figure 2. The swarm occurred much closer to OBS-2 than to when reflected from the sea surface. These later P arrivals differ OBS-3, and only a few of the events were large enough to be de- markedly in character from the first arriving P waves, and their de- tected by the latter. At OBS-2, however, their amplitudes ranged tailed interpretation is discussed below. over almost the whole dynamic range of the recorder. The The tendency of Mid-Atlantic Ridge earthquakes to occur in amplitude of the largest event was 59 times greater than that of the swarms, grouped closely both in time and space, has been noted smallest. The identical nature of the records, even as recorded on

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 1 Mlinfllifii11 if'iMii'PV: M! i ' I

1 -J- -4 jum spsi'l' ., pft T +4 Am Il I I I IT Y 1 ' ii 'i'1 (hill

-- -4 Aim

11 i

rnrrrxjirrrvrinm

Figure 4. Event 01:38/293 detected by OBS-3. (S - P) = 4.18 s.

PEAK TO PEAK AMPLITUDE, cm Figure 5. Cumulative plot of earthquakes recorded by the horizontal seismometer of OBS-2, October 1973. Maximum peak-to-peak amplitudes Figure 6. Cumulative plot of earthquakes recorded by the H-58 hyd- were measured in the 5-wave train and refer to a particular gain and filter rophone of OBS-2, October 1973. The maximum peak-to-peak amplitude setting when playing back the magnetic tape onto paper. At the magnifica- was measured regardless of phase. Generally one of the sea-surface- tion used, 3 cm (on abscissa) is approximately equivalent to 1 fim of reflected arrivals gave the largest signal. At the playback setting used, 1 cm ground motion; b = slope of line. (on abscissa) is equivalent to about 3.5 Pa.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 OCEAN-BOTTOM SEISMOGRAPH OBSERVATIONS, MID-ATLANTIC RIDGE 669

the wideband hydrophone channel, indicates that the whole swarm sufficient to regard the mh and M, scales as numerically equivalent. was generated within a volume no more than about 100 m across. Thus, given the limitations of the magnitude scales themselves and Cumulative plots of the distribution of the amplitudes of the the fact that with one or two stations it is not possible to eliminate earthquakes detected by OBS-2 are shown in Figures 5 and 6. In the effect of the radiation pattern of an earthquake, a reasonably Figure 5 the amplitude is the maximum observed in the S—wave consistent picture of the magnitudes of the earthquakes observed train on the single horizontal seismometer (of unknown orienta- has emerged. It can be concluded that the magnitude of the largest tion). In Figure 6 the amplitude is the maximum observed on the earthquake observed, on either the mb or ML scales, was about 2.5. complete hydrophone trace of the earthquake. In each case, the The magnitude of the smallest earthquakes observed was about 0. logarithm of the number of events greater than or equal to a given amplitude is plotted against the logarithm of that amplitude. The ACOUSTIC SPECTRA slopes generated by the two plots are consistent with each other, but significantly greater than that obtained at lat 45°N in 1972 The entire H-58 hydrophone record of OBS-2 was digitally pro- (Francis and Porter, 1973). cessed to determine the ambient sea-noise background in the fre- The magnitudes of the earthquakes observed can be estimated in quency band 5 to 32 Hz. Levels that exceeded the dynamic range of a number of ways: the OBS-2 recorder were excluded. The data were processed 1. The local magnitude ML (Richter, 1958) can be determined through a spectrum analyzer to provide 1-min averages, resulting from the maximum amplitude observed in the S—wave train by the in more than 10,000 spectral amplitudes with a 0.25-Hz resolution single horizontal seismometer. On this basis, the range of mag- over the acoustic band. Unacceptably high spectral amplitudes, nitudes observed by OBS-2 lay in the range 0.5 =£ ML < 2.5. The amounting to less than 3 percent of the total data, were excluded magnitude of the largest event detected by OBS-3 (see Fig. 4) was from further processing. These overloads were caused primarily by 3.0. The fact that a single seismometer is used rather than two at intense earthquakes signals and, occasionally, by nearby ship pas- right angles to measure the maximum horizontal deflection means sages. A summary of the acoustic analysis is shown in Figure 7, that any individual measurement of magnitude must be treated where sound-pressure spectrum levels are plotted against fre- with caution, but this should not affect too seriously the estimate of quency. A more complete description of this work has been given the range of magnitudes observed. elsewhere (McGrath, 1975). 2. Reid and others (1973) derived a formula for determining the Two conclusions are drawn from the data shown in Figure 7. local magnitude of an earthquake from the amplitude of the pres- First, mean sea-noise levels in the infrasonic band (frequencies sure disturbance that it creates in the water: below 15 to 20 Hz) are about 80 dB relative to 1 /nPa (80 dB//l /¿Pa). This result is surprisingly low in comparison to the levels M = logP - 1.85 - log A + log A2o- l 20 0 predicted by Wenz (1962) and Urick (1974). However, it should be The last two terms in this formula are distance correction factors borne in mind that the magnitudes of the events observed were low, given by Richter (1958) and Brune and Allen (1967), respectively. not exceeding about 2.5, and that Wenz's curve is really an esti- P2o's the amplitude (zero to peak) of the P wave (measured in pas- cals), whose spectrum is assumed to peak in the region of 20 Hz. This is roughly true for the wideband hydrophone channels of OBS-2 and OBS-3. The amplitudes of the P waves observed on these channels lay in the range 0.1 to 17.5 Pa, corresponding, when • their ranges are taken into account, to the magnitude range 0.2 £ 100 • • ML < 2.6 O- • • 3. Navarro and Brockman (1970) developed a formula for de- 4. riving mb, normally determined from teleseismic observations, from close-range observations of P: 90 - UJ mb = log V + 2.3 log r - 2.0, > • • O where V is the maximum (zero to peak) vertical velocity (in mic- • O rometres per second) in the P—wave train and r is the range (in • a o 80 • • kilometres). V can be obtained directly from the vertical seismome- • s O •° • ter record. In the case of the largest event recorded, 01:38/293 on o o Ui • OBS-3 (Fig. 4), mb = 2.3. Alternatively, V may be obtained from CL • the hydrophone record, since vertical (particle) velocity and pres- CO — — " UJ 70 - ~ ^ sure for a plane wave propagating upward through the water are CE 3 related by the expression OT to PIV = p (a/cos 8), UJ a: 60 where p and a' are the density and compressional wave velocity of CL the water and 6 is the angle of the ray with the vertical (Schneider => and Backus, 1964). Since the velocity contrast between the sea O • MAXIMUM AND MINIMUM SEA NOISE water and the sea bed is large, P waves transmitted into the water to 50 ELECTRONIC NOISE LEVEL from the sea bed have small values of 9. Hence, cos 6—1, and it is a • MEAN VALUE good approximation to write PIV = 1.5 (where P is measured in O MEDIAN VALUE pascals and V is measured in micrometres per second). Using this expression to obtain V from the hydrophone record of 01:38/293 I I I I 1 ll ill 1 1 1 1 (Fig. 4), the Navarro-Brockman formula gives mb = 1.9. 4 5 6 7 8 9 10 12 14 16 18 20 25 30 40 The relationship between mb and ML is not well established and FREQUENCY, Hz varies with locality. The best available is probably that determined Figure 7. Summary of the acoustic measurements made from the H-58 for shallow New Zealand earthquakes by Gibowicz (1972). This hydrophone record of OBS-2, October 1973. Note that 1 /¿bar is equiva- suggests that in the range of magnitudes considered here, it is lent to 100,000 /u.Pa; thus 100 dB//l /xPa is equivalent to 0 dB//l /¿bar.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 670 FRANCIS AND OTHERS

50r 0 BS-2 OCTOBER 1973

8-0 9.0 0 1.0 2.0 3.0 4.0 5-0 6.0 7.0 8.0 9.0

(S-P), S A (S-P). 5 B

Figure 8. Histograms of (S - P) observations made by OBS-2 (A) and OBS-3 (B).

mate of the uppermost levels expected from an underwater earth- perpendicular to the ridge axis, Whitmarsh (1975) detected a nar- quake. row zone of low velocity associated with the median valley. Such a Second, distant shipping appears to be the dominant source of zone could introduce considerable scatter into the earthquake infrasonic sea noise at the Mid-Atlantic Ridge. This conclusion is epicenters determined. drawn from the trend of the spectral slope (+3 to +4 dB/octave) in An indication of the ranges of the earthquakes observed is given the 5- to 32-Hz band. The spectral slope of the observed mean by the (S — P) times. Histograms of these time differences at OBS-2 acoustic levels is similar to Wenz's predicted slope for sea noise due and OBS-3 are shown in Figures 8A and 8B, respectively. They in- to heavy distant shipping traffic. No more than 10 percent of the dicate that most of the events are taking place closer to OBS-2 than total recording time could be dominated by seaquake noise, based OBS-3. If the P and S velocities were known, measurement of (S — on an average of observed event rates of about 3 h-1 and on event P) times at both OBS-2 and OBS-3 would be sufficient to locate an durations of no more than 2 min per event. Thus, distant shipping event. But there would be an ambiguity: on which side of the two- traffic is the dominant source of sea noise for at least 90 percent of station array did the event lie. This ambiguity can be resolved by an the time at the Mid-Atlantic Ridge. examination of the (Rj — P) time differences. OBS-2 was located on the eastern flank of the median valley so that earthquakes occur- LOCATION OF EVENTS ring to the west of the array were on the deep-water side of the in- strument; those to the east of the array were on the shallow-water In general it is not possible to locate earthquakes recorded at just side. Whether the P waves have arrived from the deep- or two seismic stations unambiguously. But the presence of a water shallow-water side of an ocean-bottom seismograph on a sloping layer, well defined in velocity and thickness, above the ocean- sea bed is apparent from the (Rt — P) time differences. In Figure 9, bottom seismographs makes determination of the locations feasible (R! — P) times computed for OBS-2 are shown as a function of the for the October 1973 observations. It is, however, necessary to slope of the sea bed and the angle (8) of the rays with the vertical. make a number of assumptions about the earthquakes themselves In the case of a flat bottom (fault scarps facing inward toward the median valley are not approximation that is necessary to be able to locate the events, but unlikely — would allow (Rt — P) to exceed the sounding times. that in practice cannot be a very good one. Almost certainly a Histograms of the (R, — P) observations made at OBS-2 and strong positive velocity gradient exists in the top few kilometres of OBS-3 are shown in Figures 10A and 10B, respectively. In neither the Earth's crust at the axis of the ridge (Francis and Porter, 1973). situation is the sounding time an upper limit to the distribution. In- The presence of large open fissures at the sea floor (Heirtzler, deed, in the case of OBS-2, all but a few percent of the (R t — P) 1975), which must close with depth, also implies an initial positive times exceed the sounding time. It is reasonable to conclude, there- gradient of both compressional and shear velocity with depth. fore, that most of the events observed by OBS-2 were occurring on There is evidence, too, that lateral velocity gradients are present the deep-water (that is, median-valley) side. near the axis of the ridge. This is indicated by most of the seismic Before proceeding to locate the events, it is necessary to select P refraction stations in the FAMOUS area (Whitmarsh, 1973; Fowler and S velocities for our simple model. Examination of the arrival and Matthews, 1974; Poehls, 1974). From seismic refraction lines times at the two seismograph stations allows good estimates to be

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 (.0

Figure 9. (ft, - P) as a function of the angle of the ray with the vertical computed for various values of sea-bed slope i\>. a, = 1.501 km/s and

a2 = 4.5 km/s are the compressional wave s o velocities of the water and rock layers, respec- tively; d = 2.480 km is the depth of OBS-2. A „ negative ray angle indicates that the rays came t from the deep-water side of OBS-2. The source is - far enough away for differences between the

angles of incidence ofiZt and P on the sea bed to be negligible.

2.0

-90 - 80 -70 - 60 - 50 - 40 .-30 - 20 -10 0 10 20 30 ¿ 0 50 60 70 80 90

ANGLE Of RAY WITH VERTICAL DEGREES

3.4 3.6 (R, - P), S ( R, - P ), S Figure 10B. Histograms of (ft, - P) observations at OBS-3. Figure 10A. Histogram of (ft, — P) observations at OBS-2. The area with diagonal lines corresponds to events plotted as dots in Figure 14, and the area with cross-hatching corresponds to events plotted as crosses in Fig- ure 14.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 672 FRANCIS AND OTHERS

o 1.0

(Pm-pn). s

Figure 11. Histogram of differences in the P arrival times of earth- quakes recorded by OBS-2 and OBS-3, October 1973. The maximum time difference sets an upper limit to the P velocity.

made of these. The estimates depend, however, on the distance be- tween the stations. Two methods were employed for determining this: 1. The distance was calculated from the individual drop points of the ocean-bottom seismographs on the basis of satellite naviga- tion (Table 1). The distance obtained thus is 15.3 km. Each drop

ocity for locating the earthquakes. It implies that no earthquakes AP = Pm - P„ = h. for which P waves were recorded by both OBS-2 and OBS-3 lay on «3 a2 the extension of the line joining the two instruments. Earthquakes dì d2 occurring in between are more likely to be well recorded by both aAS -— •>!! instruments since both ranges can then be short. The sum of the (S - P 3 P 2 P) times at OBS-2 and OBS-3 is therefore more likely to yield useful Hence, information. The minimum value of this sum sets a lower limit to AS = k3AP + -*-(k3- k2)

where

k k 3 03 ' 2 0. '

Thus, if k2 = k3 = k, AS = k AP, and the plot of AP against AS is a straight line through the origin. A regression line fitted to the data in Figure 13 (assuming all the error is in AS) yields the equation AS = (2.08 ± 0.10) AP + (0.09 ± 0.12). The slope of this line is significantly greater than V3, the ratio of P to S velocity for a Poisson solid, but the intercept is not sig- nificantly different from 0. We can therefore adopt the ratio 2.08 for our velocity model for locating the earthquakes. Note, how- ever, that this result has been obtained without any assumptions p) + ( s - P ) [

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 OCEAN-BOTTOM SEISMOGRAPH OBSERVATIONS, MID-ATLANTIC RIDGE 673

The compressional wave velocity so obtained is a plausible one to 3?W 50' to' use for the top few kilometres of crust in this area. It represents some sort of spatial average of the structure between the seismog- raph stations and the earthquakes. The significance of the velocity ratio 2.08 is more tangible and is discussed fully below. Any event for which three of the four possible P and S arrival times are available can now be located. Most were better recorded by OBS-2 than by OBS-3 so that location depended on Pn, 5„, and P,„. Events located on this basis are shown in Figure 14 by dots. However, there was a small group of events that generated much better records at OBS-3 than OBS-2; most of these occurred within a 4-hr period on day 292. These were located usingPm, Sul, and Pn. Closer examination of these events revealed that their (Rt — P) times at OBS-2 all grouped at the short end of the distribution (cross-hatched part of histogram in Fig. 10A). It seems plausible, therefore, that these events occurred on the eastern, shallower- water side of OBS-2. The locations so obtained are shown in Figure 14 by crosses. Confidence in this interpretation is increased by comparing the (R1 — P) times with those for events in roughly complementary positions on the other side of the array. The latter almost all exceed 3.5 s and average about 3.7 s. To be able to locate the earthquakes from the two stations, two Figure 14. Bathymétrie chart of the area studied showing the positions geometrical conditions have to be satisfied. Where r„ and rm are the of the ocean-bottom seismographs and of earthquakes recorded. ranges from OBS-2 and OBS-3, respectively, the two conditions are Bathymetry from unpublished chart by J. D. Phillips, Woods Hole Océanographie Institution. Contour interval is 100 fm. Depths greater than (rn + rni) a 15.6 km 1,500 fm have vertical lines. Earthquake epicenters are shown as dots and crosses, according to the method of location (see text). The decision to kii - »m| 15.6 km. place the dots to the west of the array and the crosses to the east was based on (H, - P) time differences at OBS-2. Provided that the ratio al ¡3 is fixed, choosing too low a value for a will result in some events failing the first condition. If too high a value is chosen, then other events are likely to fail the second condi- OBS-3 records, since many more T phases were recorded by this tion. With a = 5.04 km/s, only 1 out of the 56 events that it was instrument than by OBS-2. Unfortunately, it proved to be very thought worth trying to locate failed these geometrical constraints. difficult to recognize the start of the T phase as this was frequently Setting a = 4.0 km/s increased the failure rate to 5. lost in higher order sea-surface—reflected phases. The T phase be- ginnings shown in Figures 3B, 4, and 15 are considerably more ob- T PHASES vious than most. However, the arrival time (Tmax) of the maximum amplitude of the T-phase envelope was easier to identify, and T phases could be identified on most of the longer-range records numerous measurements were made of this. Furthermore, where [(S — P)> 2.5 s], but were not apparent at close range. The main both T and Tmax could be identified, the time difference between reason for this is probably that the amplitude of the T phase rela- them averaged about 6 s. Therefore, a plot of (Tmax - P) against (S tive to that of the P and S phases increases with distance (Reid and — P) might be expected to fall on the line others, 1973). At close range, the T phase is also likely to get lost in (Tmax - P) = m(S - P) + 6. the sea-surface-reflected arrivals R,, R2, .... Since the T phase ar-

rives at the ocean-bottom seismograph through the water, it is — A plot of (Tmax — P) against (S — P) for OBS-3 is shown in Figure like R,, R2, . . . — best detected by the hydrophone. The T phase 16. The values show considerable scatter, greater than would be travels for most of its path through the water at known velocity (co) expected from the errors likely in the measurement of Tmax. The so that comparison of T arrival times with those for? and5 should, line corresponding to a = 5 km/s forms an approximate lower in principle, yield information about the velocity structure of the limit to the distribution of points. It seems unlikely that the average rock layer. P velocity is either so high or so variable as indicated by the plot, so Suppose an earthquake occurs at very shallow depth at a dis- another explanation for the scatter must be found. Two processes tance d from the seismograph. Then the time differences between may be taking place to upset the simple model taken for T-phase the arrival of T, S, and P can be written propagation: 1. Energy may be leaking into the water layer at all distances along the path between the earthquake focus and ocean-bottom 0) a \ a - Q) / seismograph. This would tend to advance the arrival of the T and phase, that is, shorten the (T — P) time at any particular range — which appears to be the opposite of what is happening. Further- more, it seems unlikely in the rugged topography prevailing on the axis of the Mid-Atlantic Ridge that such an extended process could Hence (T - P) = m(S - P) where provide a coherent source of energy in the water. a/3 I aM 2. A coherent energy source co.uld be provided, however, at the m = -—/ ; . • (a — /3)' (<*-«) sea bed immediately above a shallow-focus earthquake. But this energy need not propagate to the seismograph by the direct route. But we already know that ccl/3 = 2.08 and thatw = 1.5 km/s so Indeed, the rugged topography may frequently prevent this from that the constant m is dependent only on a. A plot of (T - P) being a possible path. The energy would then reach the seismog- against (S — P) can therefore yield the prevailing P velocity a. raph only indirectly, after numerous reflections from side to side off An attempt was made to produce a (T — P), (S — P) plot from the the rocky sea floor. This process would tend to delay the arrival of

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 51s |

Figure 15. The same record as that shown in Figure 4 played back on a more compressed scale to demonstrate the T phase.

(S-P), S

Figure 16. Plot of {Tmax - P) against (S - P) for OBS-3, October 1973. Tmax is the arrival time of the maximum amplitude of the T phase.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 OCEAN-BOTTOM SEISMOGRAPH OBSERVATIONS, MID-ATLANTIC RIDGE 675

the T phase at any particular range, with the earliest T arrival [that value obtained for the from seismic refraction shoot- is, the smallest (T — P) value] corresponding to the direct path. ing is 1.85 (Christensen, 1972). Christensen drew mineralogical This appears to be a good explanation of the scatter observed in conclusions about the nature of the oceanic crust from the ratios of Figure 16. The same process also accounts for the extensive coda compressional to shear velocity observed, and it is interesting to following Tmax (Fig. 15). note that the values of a = 5 km/s and al ¡3 = 2.08 (Poisson's ratio = 0.35) correspond on Christensen's (1975) diagrams to a rock INTERPRETATION containing 90 percent serpentine. This is not thought to be a plaus- ible interpretation of the high ratio al ¡3 observed. A more likely In addition to the observations described in this paper, microear- explanation is that the ratio is reflecting high rock porosity. thquake activity in the FAMOUS area was studied by Reid and Numerous fissures have been observed in the median-valley floor Macdonald (1973) and Spindel and others (1974) using in the FAMOUS area by both manned submersibles and remotely sonobuoys. R. B. Whitmarsh (1975, oral commun.) also detected operated cameras (Ballard, 1975; Heirtzler, 1975). The great earthquakes in the area on bottom receivers deployed for seismic horizontal extent of these fissures makes it likely that they also refraction work. extend to some depth, possibly a few kilometres, though with in- The earthquakes located in this study (Fig. 14) fall into two dis- creasing overburden the cracks are likely to close. It is likely then tinct groups: most of the events occurred in the median valley to that the rocks remain highly porous to a comparable depth. Such the west of OBS-2, and some occurred in the fracture zone to the a picture is consistent with that found in the neovolcanic zone of east. The median-valley events all lie west of the axis of the valley, Iceland, where porosities of about 0.03 have been inferred from mostly under its western wall. The one-sidedness of this distribu- electrical resistivity measurements for the top of layer 3, lying at a tion is similar to that obtained in the median valleys south of frac- depth of some 3 km (Bjornsson, 1975). ture zones A and B by Spindel and others (1974). In their work, The effect of porosity on seismic velocities has been demon- however, all the activity lay on the eastern side, and they correlated strated in the laboratory by Nur and Simmons (1969). They mea- this asymmetry with that postulated for sea-floor spreading in the sured compressional and shear velocities of both dry and water- area. Our result makes any such correlation unlikely. A more likely saturated rock samples as a function of confining pressure (with association may be with the steepness of the valley walls. The active zero pore pressure). As confining pressure increases, the cracks in western slope of the median valley in this study is considerably the rocks close up and the velocities increase. But with saturated steeper than the opposite eastern slope on which OBS-2 lay. The rocks, VP increases less than Vs with increasing pressure. Hence, activity detected by Spindel and others farther south correlates well the ratio VPIVS is high for cracked, saturated rocks at low pressures with steep scarps, although not all steep scarps were found to be and decreases to its intrinsic value as increasing pressure closes up seismic. the cracks. The results obtained by Nur and Simmons for one par- The second group of earthquakes (crosses in Fig. 14) form a ticular rock type are shown in Figure 17. The results for other rock quite different distribution. They lie in the fracture zone to the east types are similar. Comparable measurements on oceanic basaltic of the ocean-bottom seismographs and with one exception fall rocks are not available in the literature. quite closely on a line trending N62°E. The length of this linear pat- Nur and Simmons made their measurements on small samples at tern of earthquakes is about 6 km — much too great to be gener- frequencies of a few megahertz. Thus, their seismic wavelengths ated by some systematic error in the observations — and it inter- were approximately a millimetre, but still much greater than the sects the trend of the fracture zone at an angle of about 31°. Locat- width of the microcracks contained in their samples. Similarly, in ing these events with a lower compressional velocity ( a) of 4 km/s this study the seismic wavelengths of a few hundred metres were moves the line closer to the center of the array and rotates it much greater than the width of the fissures present in the ocean clockwise some 13°. But it still intersects the trend of the fracture floor. It is reasonable to conclude, therefore, that the high ratio of zone at a considerable angle — 18°. One cannot escape the conclu- compressional to shear velocity, 2.08, obtained here reflects the sion, therefore, that an active fault has been observed in the frac- high fissure and crack porosity in the top few kilometres of oceanic ture zone, trending approximately east-northeast and intersecting crust in which the seismic waves were propagating. This conclusion the trend of the fracture zone itself obliquely. is compatible with Whitmarsh's (1975) finding of a low-velocity Near-bottom observations in fracture zone A have revealed the zone in the median-valley floor. The high al ¡3 ratio and Whit- presence of fault scarps having both east and east-northeast trends marsh's low value for a are both manifestations of a highly frac- (Detrick and others, 1973; Detrick, 1974). It is noteworthy that tured and disturbed volume of rock saturated with sea water. If the although an east-trending scarp dominates the western end of frac- zone is as narrow as Whitmarsh claims, 2 to 3 km, then within it, ture zone A, at its eastern end the principal scarp trends east- al [3 may be even higher, for the seismic-wave propagation studied northeast. Both are between 5 and 10 km in extent. The active fault here covered a more extensive area. detected here lies at the eastern end of its fracture zone, terminating close to the axis of the section of to the north. OTHER PHASES More recently, large-scale side-scan observations of frac- ture zone B have revealed the presence there of numerous scarps The existence of later-arriving compressional wave phases (P' that trend northeast and east-northeast, between 2 and 10 km in . . .) has already been mentioned, and such phases are identified in length, cutting obliquely across the trend of the fracture zone Figure 3. They were frequently observed on the OBS-2 records and (Whitmarsh and Laughton, 1975). Whitmarsh and Laughton in- occasionally on OBS-3 records as well. These phases differed terpreted these as an en echelon system of tensional and shear frac- markedly in character from the first-arriving P waves, and any ex- tures of the type first observed in clay models by Riedel (1929). It planation of their nature must account for this. In general, they seems plausible that the earthquakes detected here represent activ- were much richer in lower frequencies than P (compare Figs. 3A ity along one such fracture. and 3C), but the presence of high frequencies in some is indicated Although it has not been possible in this study to derive a de- by their impulsive nature (compare Fig. 3B). For no record is it pos- tailed velocity structure for the axial region of the ridge, an unam- sible to derive P' simply by passing? through a low-Q filter, where biguous determination of the velocity ratio al ¡3 has been made. Q is the specific dissipation constant of the rock filter. This does The value obtained, 2.08, is unusually high. In contrast, the largest not mean that P' has not suffered attenuation in propagating

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 676 FRANCIS AND OTHERS

22

2 0 >to >

1-6

1 -4'

6

Figure 18. Plot of (P' - P) against (5 - P) for OBS-2 observations, Oc- tober 1973. (P" - P) times are included with (P' - P) times. 5 in is a direct wave, and P is refracted. This is confirmation that the £ events observed are shallow, argued earlier from the presence of T § phases. 2. If P' is a direct wave and P refracted, it is not surprising that S 4 the former is frequently larger in amplitude than the latter. But hav- ing propagated through the highly fissured superficial crust, P' is likely to have suffered even more scattering than P. This could ac- count for its lack of high-frequency energy relative to P. 3. There is no evidence for the existence of S', a direct shear 3 wave in the top layer. This may mean that this layer is too highly fissured for shear waves to be easily propagated. 4. Finally, the fact that a value of 2.08 for al fi has been ob- served from the P and S arrivals indicates that the Assuring ob- 2 served at the sea floor penetrates through the superficial low- 0 200 400 600 800 1000 velocity layer well into the higher-velocity layers that propagate S and P. PRESSURE, BARS Figure 17. The effect of confining pressure on the compressional and shear-wave velocities of a rock with porosity in the form of cracks (after CONCLUSIONS Nur and Simmons, 1969). At low confining pressure the cracks remain

open, and with saturated samples a high VPIVS ratio is observed. Ocean-bottom seismographs were deployed twice in 1973 in the FAMOUS area of the Mid-Atlantic Ridge. On the first occasion, through a low-Q medium. But it is likely that P has also been con- few earthquakes were recorded, but on the second, 515 events were siderably modified by its path. The high frequency coda of P, par- detected with magnitudes ranging between 0 and 2.5. Although ticularly noticeable on the hydrophone trace of Figures 3B and 3D, only two instruments operated successfully, the direction from suggests that it has suffered some scattering. This would be ex- which seismic waves were arriving was indicated by the (Rt — P) pected when seismic waves propagate through the highly fissured times. Thus it has been possible to locate the events without am- rocks of the median-valley floor. biguity. Most of them lay beneath the western slope of the median The time differences between later-arriving compressional valley, but a small group originated in a fracture zone along what phases and the first-arriving P phases are shown as a function of appears to be a single fault. The fault cuts obliquely across the range in Figure 18. All these measurements were made on OBS-2 trend of the fracture zone, but has the same trend as some of the records with (R, — P) times indicating propagation from the scarps detected by long-range side-scan sonar and deep-tow obser- median-valley side. IfP' represented head waves or waves reflected vations. or refracted at depth, (P' — P) would be expected to decrease with It has not been possible to derive a precise velocity structure for increasing range. Although the points on the plot are scattered, the the rock through which the seismic waves were propagating, but an opposite is clearly the trend. With the velocities a, /3 already estab- unambiguous determination of the ratio of P to S velocities has lished for P and S, most of the P' waves appear to be traveling at a been made. The value obtained (2.08) is very high and is thought to velocity of about 2.7 km/s. The most plausible interpretation for reflect the highly fractured, water-saturated nature of the rock the P' waves, therefore, is that they are waves traveling in a low- through which the P and S waves were propagating. velocity top layer immediately adjacent to the sea floor. A surface An analysis of the T-phase observations indicated that most of layer (layer 2 A) with low velocities of this order has been observed the T-phase energy radiates into the water from a relatively small on the Reykjanes Ridge (Talwani and others, 1971) and elsewhere area above the source and that much of the propagation to the in the FAMOUS area (Poehls, 1974; Whitmarsh, 1973). It is not ocean-bottom seismograph is by indirect paths. Conversely, it is possible in this study to determine the thickness of this surface plausible to interpret the presence of a T phase as an indication of layer, but the measurements just cited suggest that it may be about shallow focal depth. 1 km. Late-arriving compressional phases were observed that have The interpretation of P' waves has a number of implications: been interpreted as direct waves propagating in a low-velocity sur- 1. Propagation of a compressional wave in a low-velocity sur- face layer (layer 2A). This interpretation implies shallowness of the face layer implies that its source exists in the same layer; that is, P' focal depth of the earthquakes observed.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021 OCEAN-BOTTOM SEISMOGRAPH OBSERVATIONS, MID-ATLANTIC RIDGE 14 677

ACKNOWLEDGMENTS Heirtzler, J. R., 1975, Project FAMOUS. Where the Earth turns inside out: Natl. Geog. Mag., v. 147, p. 586-603. We thank all on board USNS Hayes, USNS Mizar, and RRS Marshall, P. D., Burch, R. F., and Douglas, A., 1972, How and why to re- cord broadband seismic signals: Nature, v. 239, p. 154—155. Shackleton who assisted in the deployment and recovery of the McGrath, J. R., 1975, Infrasonic sea noise at the Mid-Atlantic Ridge near ocean-bottom seismographs. Facilities at Blacknest were kindly 37°N: Jour. Acoust. Soc. America, v. 58, p. S120-S121. provided by H.I.S. Thirlaway and his colleagues. For constructive Navarro, R., and Brockman, F. R., 1970, Seismic activity in September criticism of the manuscript, we are grateful to R. C. Lilwall and 1969 near Rulison test site, USCGS Report CGS-746-5. R. B. Whitmarsh. At the outset of this work the ocean-bottom Nur, A. and Simmons, G., 1969, The effect of saturation on velocity in low seismograph group at Blacknest formed part of the Institute of porosity rocks: Earth and Planetary Sci. Letters, v. 7, p. 183-193. Geological Sciences, but is now attached to the Institute of Poehls, K. A., 1974, Seismic refraction on the Mid-Atlantic Ridge at 37°N: Oceanographic Sciences — both are component bodies of the Jour. Geophys. Research, v.79, p. 3370-3373. Natural Environment Research Council, United Kingdom. The Reid, I., and Macdonald, K. C., 1973, Microearthquake study of the Mid- acoustic part of this work was supported by the U.S. Naval Elec- Atlantic Ridge near 37°N, using sonobuoys: Nature, v. 246, p. 88-90. tronics Systems Command, Code 320. Reid, I., Reichle, M., Brune, J., and Bradner, H., 1973, Microearthquake studies using sonobuoys: Preliminary results from the Gulf of Califor- REFERENCES CITED nia: Royal Astron. Soc. Geophys. Jour., v. 34, p. 365-379. Riedel, W. von, 1929, Zur Mechanik geologischer Brucherscheinungen: Ballard, R. D., 1975, Project FAMOUS: Dive to the great rift: Natl. Geog. Zentralbl. fur Min. Geol. Pal. Abt. B, 354-368. Mag., v. 147, p. 604-615. Schneider, W. A., and Backus, M. M., 1964, Ocean bottom seismic mea- Bjornsson, A., 1975, Electrical resistivity of Layer 3 in the Icelandic crust: surements of the California coast: Jour. Geophys. Research, v. 69, Orkustofnun Research Rept. No. OS JHD 7509. p. 1135-1143. Brune, J. N., and Allen, C. R., 1967, A microearthquake survey of the San Spindel, R. C., Davis, S. B., Macdonald, K. C., Porter, R. P., and Phillips,' Andreas fault system in southern California: Bull. Seismol. Soc. J. D., 1974, Microearthquake survey of the median valley of the America, v. 57, p. 277-296. Mid-Atlantic Ridge at 36°30'N: Nature, v. 248, p. 577-579. Christensen, N. I., 1972, The abundance of serpentinites in the oceanic Sykes, L. R., 1970, Earthquake swarms and sea-floor spreading: Jour. crust: Jour. Geology, v. 80, p. 709-719. Geophys. Research, v. 75, p. 6598-6611. 1975, Structure and constitution of the lower oceanic crust: Rev. Talwani, M., Windisch, C. C., and Langseth, M. G., 1971, Reykjanes Geophysics and Space Physics, v. 13, p. 57-86. Ridge crest: A detailed geophysical study: Jour. Geophys. Research, Detrick, R. S., 1974, Fracture Zone A, Mid-Atlantic Ridge 37°N: A near- v. 76, p. 473-517. bottom geophysical study: SIO Reference 74-26. Urick, R. J., 1974, Sea bed motion as a source of the ambient noise back- Detrick, R. S., Mudie, J. D., Luyendyk, B. P., and Macdonald, K. C., 1973, ground of the sea: Jour. Acoust. Soc. America, v. 56, p. 1010-1011. Near bottom observations of an active (Mid-Atlantic Wenz, G. M., 1962, Acoustic ambient noise in the ocean: Spectra and Ridge at 37°N): Nature Phys. Sci., v. 246, p. 59-61. sources: Jour. Acoust. Soc. America, v. 34, p. 1936-1956. Fowler, C.M.R., and Matthews, D. H., 1974, Seismic refraction experi- Whitmarsh, R. B., 1973, Median valley refraction line, Mid-Atlantic Ridge ment on the Mid-Atlantic Ridge in the FAMOUS area: Nature, v. 249, at 37°N: Nature, v. 246, p. 297-299. p. 752-754. 1975, Axial intrusion zone beneath the median valley of the Mid- Francis, T.J.G., and Porter, I. T., 1971, A statistical study of Mid-Atlantic Atlantic Ridge at 37°N detected by explosion seismology: Royal As- Ridge earthquakes: Royal Astron. Soc. Geophys. Jour., v. 24, p. tron. Soc. Geophys. Jour., v. 42, p. 189-215. 31-50. Whitmarsh, R. B., and Laughton, A. S., 1975, The fault pattern of a slow- 1973, Median valley seismology: The Mid-Atlantic Ridge near 45°N: spreading ridge near a fracture zone: Nature, v. 258, p. 509. Royal Astron. Soc. Geophys. Jour., v. 34, p. 279-311. Francis, T.J.G., Porter, I. T., Lane, R. D., Osborne, P. J., Pooley, J. E., and Tomkins, P. K., 1975, Ocean bottom seismograph: Marine Geophys. Research, v. 2, p. 195-213. Gibowicz, S. J., 1972, The relationship between teleseismic body-wave MANUSCRIPT RECEIVED BY THE SOCIETY DECEMBER 3, 1975

magnitude m and local magnitude ML from New Zealand earth- REVISED MANUSCRIPT RECEIVED JUNE 30, 1976 quakes: Bull. Seismol. Soc. America, v. 62, p. 1 — 11. MANUSCRIPT ACCEPTED SEPTEMBER 2, 1976

Printed in U.S.A.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/88/5/664/3418474/i0016-7606-88-5-664.pdf by guest on 03 October 2021