G 3. AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES Published by AGU and the Geochemical Society. Why bury ocean bottom seismometers?

Transcription

1 Geosystems G 3 AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES Published by AGU and the Geochemical Society Data Brief Volume 8, Number 2 22 February 2007 Q02010, doi: /2006gc ISSN: Why bury ocean bottom seismometers? Fred K. Duennebier Department of Geology and, S.O.E.S.T., University of Hawaii, 2525 Correa Road, Honolulu, Hawaii 96822, USA (fred@soest.hawaii.edu) George H. Sutton Deceased 26 January 2004 [1] Theory, testing, and use of ocean bottom seismic sensors show that the fidelity of data, particularly measurements of horizontal motion, can be severely distorted and noisy because of mechanical interactions between the sensor package, the ocean bottom, and the ocean water. This lack of fidelity is common, particularly for instruments with large cross sections in the water. These problems should be minimized by burial of the seismic package below the ocean-sediment interface. To test this hypothesis, two identical seismic packages were operated on the shallow ocean floor within meters of each other with a real-time connection to land. The packages were configured in a variety of ways, with both packages buried in the sediment, one buried and one on the bottom, and one on a circular 1-m-diameter plate and the other buried or on the ocean floor. The resulting data are compared with data from both packages resting on a stable cement slab in the frequency range from 0.2 Hz to above 20 Hz. Data show convincingly that burial of seismic sensors in soft sediment greatly increases data fidelity over sensors placed on the bottom or on a plate on the bottom. While vertical sensors provide reasonable fidelity in most situations, the horizontal data are consistently reliable only when the sensor packages are buried. Components: 6507 words, 16 figures. Keywords: ocean bottom seismometer; seismic data quality. Index Terms: 3099 Marine Geology and : General or miscellaneous; 3025 Marine Geology and : Marine seismics (0935, 7294); 3050 Marine Geology and : Ocean observatories and experiments; 3094 Marine Geology and : Instruments and techniques; 7294 Seismology: Seismic instruments and networks (0935, 3025). Received 24 July 2006; Revised 31 October 2006; Accepted 29 November 2006; Published 22 February Duennebier, F. K., and G. H. Sutton (2007), Why bury ocean bottom seismometers?, Geochem. Geophys. Geosyst., 8, Q02010, doi: /2006gc Introduction [2] To be able to interpret seismograms within the context of modern seismology, we often assume that the transfer function applied to arriving signals by our seismic system is well known and invariant. In the case of ocean bottom seismology, where seismic sensors are usually dropped from a ship to record signals on soft sediments, this assumption is unjustified, as the transfer function of the sensors is often heavily modified by interactions of the sensor package with the ocean floor and water [Sutton and Duennebier, 1987; Duennebier and Sutton, 1995; Sutton et al., 1981; Sutherland et al., 2004; Crawford et al., 2006]. The package motion will often be better coupled to the water than to the sediment, and sensitivity to package tilt can Copyright 2007 by the American Geophysical Union 1 of 13

2 dominate the response, particularly for horizontal sensors. In this paper we describe the results of a test in which two identical seismic sensor packages were deployed in a variety of configurations to demonstrate these changes in response. 2. Theory [3] Good coupling is characterized by high-fidelity recording of seismic ground motion. Rather than belabor the theory of coupling and distortions and noise on ocean bottom seismometers, the reader is referred to Duennebier and Sutton [1995] for a detailed investigation of the theory. Other recent papers have also addressed these issues [Crawford et al., 2006; Crawford and Webb, 2000; Sutherland et al., 2004]. [4] In this paper, two statistics, coherence and spectral ratio, are used to quantify the fidelity of the data, together with the response to an internal shaker. Two identical collinear sensors placed adjacent to each other are expected to record identical signals, with a coherence of 1.0, implying stationary phase and amplitude characteristics for spectral estimates of the recorded signals. In the presence of random noise, coherence will decrease as the distance between the sensors increases, and coherence will be low if one sensor observes a different signal than the other. High coherence does not necessarily imply high fidelity for ground motion; high coherence will result if both sensors are responding to the same tilt. Spectral ratios between data from collinear sensors within the same package or placed next to each other should reflect only differences in sensor response, which can be removed from the data by application of appropriate transfer functions. The spectral ratio should be unity (0 db) if the sensors are identical and close to each other, or smoothly varying depending on differences in the response if the sensors are not identical. In practice, the spectral ratio will be noisy if the signals recorded by the two sensors are not coherent, as when they see different signals or when signal-tonoise levels are low. [5] Horizontal sensors respond both to horizontal acceleration and to tilt. Equation (23) of Duennebier and Sutton [1995] is S 2 ¼ 1 þ u þ q zþ g = S 2 ; y s y s where I h is the equivalent horizontal displacement measured at the location of the sensor, s is the I h Laplace variable, y s is the particle motion of the ocean floor under the package, u is the displacement of the center of mass relative to y s, q is the package tilt, z is the height of the sensor above the center of mass, and g is the acceleration of gravity. This transfer function will equal 1 when fidelity is perfect, with no relative motion between the true particle motion and the center of mass (u = 0) and no tilt (q = 0). The second part of the tilt term represents static tilt sensitivity important at low frequencies [Crawford et al., 2006], while the first term (dynamic tilt) dominates at high frequencies. The vertical sensor will respond to dynamic tilt if the sensor is displaced horizontally from the axis of tilt. The tilt signal recorded on the vertical sensor may be coherent with the tilt signal recorded on horizontal sensors. Since horizontal motions are more likely to be generated by tilt than vertical ground motion, it is more likely to observe horizontal cross talk on vertical sensors than to observe vertical motion on horizontal sensors [Crawford et al., 2006]. [6] We hypothesize that seismic sensors (particularly those measuring horizontal motion) placed on the soft ocean floor will show effects of package coupling problems, both relative motion of the center of mass with respect to the motion of the ground, and tilt. These effects should be characterized by distorted signals and increased noise levels. We expect that the buried sensors and those resting firmly on competent material will be much less affected by these problems. We also expect that bottom currents will cause tilt and increase noise levels on unburied sensor packages [Duennebier and Sutton, 1995, Figure 9]. [7] Part of the transfer function of a seismic package can be obtained by shaking the package with a calibrated internal source. The packages used in this test contain a centrally located variable-frequency mechanical shaker (offset cam connected to a DC motor) which can be commanded to run at frequencies from about 8 to 70 Hz, generating an internal force in a direction that should yield equal amplitudes on the two horizontal components and twice the amplitude on the vertical. The force on the package will be proportional to frequency squared, and the velocity response to the shaker should have a slope of 18 db/octave. At very low frequencies the offset of the shaker mass should generate a static tilt with a velocity response of 6 db/octave. At very high frequencies, package inertia dominates and the shaker will generate a signal with a 6 db/octave 2of13

3 Figure 1. Configuration of sensor packages for testing. (1) Slab: both packages on a cement slab in the laboratory. (2) Buried-Buried: both packages buried in the bottom. (3) Bottom-buried: one package on the ocean bottom and the other buried. (4) Plate-Buried: one package on a 1-m-diameter plate and the other buried. (5) Plate-Bottom: one package on the plate and the other on the ocean floor. (6) Bottom-Bottom: both packages resting directly on the ocean floor. slope [Duennebier and Sutton, 1995]. A wellcoupled seismic sensor will operate in the band where the shaker response is 18 db/octave from about 8 Hz (the low-frequency limit of our shaker) to more than 50 Hz. Poor response to a shaker does not necessarily imply poor coupling to the ocean floor; a package with the same density as its surroundings will move with its surroundings in response to seismic signals (perfect coupling), but it can still have resonant frequencies in response to a shaker. 3. Seismic Packages [8] Two identical broad-band seismic systems containing orthogonal broad-band seismometers (Guralp CMG-3) and collinear geophones (4.5 Hz Geospace HS-1), tilt sensors, shaker, compass, and separate hydrophone, pressure gauge, differential pressure gauge, thermal sensor, and current meter, were built at the University of Hawaii for the U.S. Navy and tested off the coast of Oregon for twelve days in 1991 [Duennebier et al., 1991]. The seismic packages are right circular cylinders m in diameter and 0.53 m high, with a density of approximately 1,800 kg/m 3, roughly matching the density of the sediment. The seismic sensors are mounted on a table in each package that is leveled to within of vertical by command from shore. In the 1991 Navy experiment, the packages were buried below the ocean floor using an ROV to sink a caisson into the bottom. The packages, placed 200 m apart in 620 m of water, yielded excellent data on all components, with high coherence between identical sensors between the packages. Since both packages were buried, the Oregon test did not address the question of how much improvement in fidelity and noise can be obtained by burial vs. a bottom deployment. The Navy hardware was returned to the University of Hawaii in 1995, for use in the Hawaii-2 Observatory (H2O) [Duennebier et al., 2002]. One of these packages recorded at H2O for four years, and is still at the site. 4. Makai 95 Experiment [9] Prior to modifications of the packages for H2O, a test was conducted to address questions on the value of burial and of placement of sensors on a broad base when installing seismometers on the ocean floor. The test was conducted next to the Makai Pier, Oahu, in 5 meters of water on a sandy bottom. The packages were tested in six configurations shown in Figure 1: (1) both packages on a concrete poured slab equivalent to a seismic ob- 3of13

4 servatory installation, (2) both packages buried in the sediment, (3) one package on the bottom and the other buried, (4) one package on a 1 m diameter plate and the other buried, (5) one package on the plate with the other package placed on the ocean bottom, and (6) both packages on the bottom. Currents measured in the test area were up to 10 cm/sec, and wave action is estimated at less than 12 cm peak-to-peak. The sediments at the test site are denser (1800 kg/m 3 ) and coarser (fine sand) than those on much of the deep-ocean floor. Data were recorded using a 2-channel HP 3562A Hewlett Packard Spectrum Analyzer with spectra, transfer function, and coherence files recorded digitally for each data record using a frequency range providing reliable spectral estimates from 0.2 to 50 Hz for most data. Coherence is defined by g 2 = G xyg xy * where G xx is the power spectrum G xx G yy of the channel 1 data, G xy is the cross spectrum, and G yy is the power spectrum of the channel 2 data. No active sources were used except for the internal shaker in the packages. [10] The packages were placed 2-3 meters apart to minimize possible interactions except in Test #1 on the slab, where the packages were placed as close together as possible without touching. Bottom current and pressure data were recorded for some of the tests for comparison with the seismic data. Coherence and spectral data were recorded between the collinear geophone and the broad-band (Guralp) sensors inside each package, and between identical collinear Guralp sensors in the two separate packages. If the packages move with the ocean floor with rectilinear response, then internal collinear sensors should record identical data after differences in instrument response are accounted for, but if the package is responding to motion by tilting then the sensors may respond differently depending on their location in the package and the location of the tilt axis [Duennebier and Sutton, 1995]. [11] If our hypotheses are correct, we should observe high coherence and unity spectral ratios between signals from collinear sensors where the coupling is good and where the coherence lengths are long. Coherence between sensors in different packages will decrease with increasing frequency and increasing distance between the sensors at a rate that depends on the coherence length. If package tilt is a problem, then sensors internal to a rigid package will yield spectral ratios different from the ratios expected, and possibly low coherence. Since each package contains one geophone and one Guralp sensor in each orthogonal direction, it is possible to check for dynamic tilt, although the poor sensitivity of the 4.5 Hz geophones at low frequencies prevents extending this analysis below about 0.2 Hz. Deviations of the shaker response from the nominal 18 db/ octave slope will also be an indication of coupling problems. 5. Observations 5.1. Case 1: Cement Slab [12] The two packages were placed side by side on a stable poured concrete slab in the Hawaii Institute of building and measurements were taken. This test provides baseline data since no tilts or coupling problems should be present. Each package was placed on three equally spaced steel nuts at the edge of the base to provide solid 3-point coupling to the cement slab. We expect that nearperfect data should be obtained on the concrete slab, as this configuration is similar to that used in seismic observatories. Coherence between collinear sensors is expected to be 1.0, and spectral ratios for collinear sensors should be unity except for differences in response. [13] All observed coherence values between collinear sensors in the same package are generally above 0.95 at all frequencies from 0.1 Hz (the lowest frequency adequately sampled) to above 40 Hz (Figure 2a) except for an unexplained decrease in coherence near 0.2 Hz. Coherence between collinear Guralp sensors in the two different packages was significantly lower but still above 0.9 at all frequencies (Figure 2b). Spectral ratios between collinear sensors internal to a single package (Guralp/geophone) are displayed in Figure 2c. In all cases the internal package ratios have been corrected for the same generic Guralp/geophone response, since the variations between instruments are not significant for this study. Spectral ratios from collinear Guralp sensors in the two different packages are shown in Figure 2d. Note that the ratios are not accurate below about 0.2 Hz, and that above 10 Hz the ratios are dominated by resonances. [14] The shaker data for all sensors on the concrete slab (Figure 3) show the expected 18 db/octave slope over the frequency band from 10 Hz (the minimum frequency measurable above noise level) to above 30 Hz, where the horizontal sensors show signs of resonance. The vertical sensors continue to have good coupling to over 70 Hz. As expected, 4of13

5 Figure 2. The same format is used for Figures 2, 4, 6, and 8, with coherence between collinear geophone and Guralp sensors internal to one package displayed in Figure 2a, coherence between the collinear Guralp sensors in adjacent package displayed in Figure 2b, spectral ratios of collinear Guralp/geophone sensors in the same package corrected for a generic response curve in Figure 2c, and spectral ratios between collinear Guralp sensors (not corrected for response) in different packages displayed in Figure 2d. Red and blue curves are horizontal sensor data; black curves are vertical sensor data. In Figure 2 (Slab) the packages were placed on a cement slab in the Hawaii Institute of, with the expectation of perfect response. Deviations of the coherence from 1.0 and the spectral ratios from 0 db (green lines) show how well this expectation was met. the packages were well coupled to the cement slab, and recorded data with good fidelity Case 2: Buried-Buried [15] Both packages were buried in the coral sand bottom at Makai Pier about 3 m apart by setting a caisson on the ocean floor and sucking the sediment out of the caisson with a pump while the caisson sank into the bottom. Bottom mud was poured into the caisson after the instrument was lowered in so that only the cabling was visible above the mud. The tops of the package were less than 10 cm below the ocean floor. As can be seen in Figure 4a, the coherence between collinear sensors in the same package is high at all frequencies up to above 30 Hz, almost as good as the coherence on the concrete slab, although the vertical sensors display coherence minima near 0.2 Figure 3. Shaker response on cement slab. Red and blue curves are horizontal geophone data; black curves are vertical geophone data. The expected 18 db/octave slope for all curves is shown by the thick green line. All curves approximate this slope up to at least 30 Hz, implying good coupling for all sensors. 5of13

6 Figure 4. Buried-Buried. Comparisons with both packages buried in the ocean floor. (a) Internal coherence is somewhat degraded from the concrete slab case, and (c) internal spectral ratios are similar to those for the packages resting on the concrete slab (Figure 2). (b) The inter-package coherence is considerably less than on the concrete slab, indicating short coherence lengths. (d) The inter-package spectral ratios show similar amplitudes between the buried packages. See Figure 2 for the figure format. and 1 Hz (Figure 2a). The coherence between collinear sensors between the two buried packages (Figure 4b) is considerably lower than expected, and far lower than the equivalent coherences observed on the slab (Figure 2b), implying that the coherence length in the test area is very short and variable with frequency. In hindsight this is not surprising given the geometry of the test site (within a corner of the pier), with possible noisegeneration sources within several meters in several directions. [16] Spectral ratios between buried collinear sensors in the same package (Figure 4c) are similar to those observed on the cement slab. Spectral ratios between collinear sensors in different buried packages average less than 6 db between 0.3 Hz and 30 Hz (Figure 4d). The noise in the spectral ratios between packages is also an indication of short coherence lengths. [17] The shaker data for all buried sensors (Figure 5) show the expected 18 db/octave slope up to above 50 Hz, implying good coupling in all directions. Surprisingly, comparison with Figure 3 shows that the response to the shaker is better for the buried sensors than for the sensors on the Figure 5. Buried shaker response. Good coupling is reflected in the observed 18 db slope (green line) of these data. Red and blue curves are horizontal sensor data, and black curves are vertical data. 6of13

7 Figure 6. Bottom-Buried. (a) The internal coherences between collinear sensors on the bottom shown here are high for the horizontal sensors (red and blue) but relatively low for the vertical sensors (black). (b) The coherence between the bottom and buried packages is generally low except for the vertical sensors at frequencies below 0.5 Hz. (c) Internal spectral ratios for the bottom package are close to zero db except for one case. (d) The spectral ratios between the bottom and buried packages show that the amplitudes of horizontal motion are as much as 8 times larger on the bottom than in the buried package, while vertical amplitudes are nearly equal. concrete slab, implying that burial of sensors even on land could improve seismic fidelity of horizontal data at high frequencies. This improvement is likely the result of restricting the horizontal motion of the sides of the package Case 3: Bottom-Buried [18] Leaving one of the packages buried, the other package was removed from its caisson and placed directly on the ocean bottom 2 m away from the buried package, taking care to orient the horizontal sensors to collinear directions with those in the buried package. [19] Internal coherence between the horizontal geophones and the Guralp sensors for the package on the bottom are high up to about 20 Hz, except for a dip in coherence below 0.2 Hz (Figure 6a), while the coherence between the vertical geophone and vertical Guralp on the bottom is significantly less than the coherence of the buried vertical sensors (Figure 4). The high coherence between horizontal sensors in the package on the bottom implies that the signals recorded were the same on each pair of collinear components, but it does not imply that the signal was recording rectilinear seismic ground motion. Strong package tilt will be coherent on horizontal sensors at different locations in the same package, while response to tilt need not be coherent for two vertical sensors in the same package. The relatively poor coherence of the vertical sensors indicates that the package is suffering tilt, despite the high coherence of the internal horizontal sensors. [20] Coherence values between collinear Guralp sensors in the buried and bottom package (Figure 6b) show that the vertical coherence is high below 0.5 Hz and low at higher frequencies, while the horizontal coherences are lower than the coherences between the buried horizontal sensor packages (Figure 4b). 7of13

8 Figure 7. Shaker response for seismic package on the bottom. The vertical sensors (black) respond to the shaker with the expected 18 db/octave slope, but the horizontal sensors (blue and red) show extreme departures from the expected response. [21] The observed spectral ratios for the collinear sensors in the package on the bottom (Figure 6c) yield expected values except for one vertical pair where the Guralp was consistently more sensitive than the geophone. Spectral ratios between collinear buried and bottom sensors in the two packages (Figure 6d) indicate that the horizontal sensors placed on the ocean floor detect signal levels as much as 20 db higher (a factor of 10 in amplitude) than the equivalent levels observed in the buried package below 2 Hz, and more than 6 db higher up to 30 Hz. The vertical noise levels are roughly 5 db higher on the bottom sensors than on the buried sensors over the same frequency range. There is no reason to suspect that particle motion of the ocean floor should be different when measured by sensors above the bottom or below the bottom, so at least one of these packages must be recording distorted particle motion. Since package tilt can generate apparent horizontal accelerations that amplify horizontal motion, we conclude that tilting of the package deforms the motions detected by the sensors on the bottom. [22] Shaker data for the package on the bottom show good vertical coupling up to 40 Hz, but the horizontal sensor response to the shaker is severely distorted at frequencies as low as 12 Hz (Figure 7). Compare these results with those in Figure 5 for the buried sensors. The extreme excursions from the desired 18 db/octave slope imply large distortions and resonance [see Duennebier and Sutton, 1995, Figure 6] Case 4: Plate-Buried [23] One package was placed on a 1 m diameter plate with 6 cm fins extending below it to supply rigidity and improve horizontal coupling. The package was placed at the center of the plate on three nuts at the edge of the package to give a stable tripod base. The rationale for placing the package on the 1 m diameter plate is to increase the area in contact with the soft bottom (optimally by a factor of about 10 over just placing the package on the bottom) and to increase the diameter of the base to lessen tilt. Coherence between collinear sensors within the package on the plate is variable (Figure 8a), but generally above 0.4 except near 1 Hz. Horizontal coherence is lower than that observed in case 3 (Figure 6a), again implying that the high coherences observed in case 3 reflect tiltgenerated signals. Coherences between collinear Guralp sensors in the buried package and the package resting on the plate (Figure 8b) are similar but somewhat less than the coherences between the two buried packages (Figure 4b), and higher than that in Figure 6b, implying that the plate helped in reducing the effects of tilts. [24] Spectral ratios between collinear Guralp and geophone sensors in the package on the plate (Figure 8c) display variations of more than 6 db, implying some distortion. Spectral ratios between sensors on the plate and buried sensors show horizontal levels about twice the amplitude of the vertical levels (Figure 8d). The levels observed on the vertical sensor on the plate are within a few db of those observed on the buried package. [25] Shaker data for the package on the plate (Figure 9) show reasonable vertical coupling at all frequencies, with the response near the expected 18 db/octave. Horizontal sensors are poorly coupled, showing signs of resonance at frequencies as low as 15 Hz. The stiffness supplied by the addition of the plate appears to move the resonances observed in the horizontal shaker response on the bottom to higher frequencies (compare with Figure 7) Case 5: Plate-Bottom [26] Coherence of sensors on the plate with those in the package on the ocean floor is low except at frequencies below 0.5 Hz (Figure 10a). Spectral ratios (bottom sensors/plate sensors) show that levels on the bottom are higher than those on the plate by as much as 12 db (Figure 10b) Case 6: Bottom-Bottom [27] With both packages resting directly on the ocean floor, coherence values between collinear 8of13

9 Figure 8. Plate-Buried. (a) Coherence between the geophones and Guralp sensors within the package on a 1 m diameter plate. (b) Coherence between Guralp sensors on the plate and collinear sensors in the buried package. (c) Spectral ratios for sensors within the package on the plate. (d) Spectral ratios of Guralp sensors on the plate divided by the signals from collinear sensors in the buried package. sensors in the two packages (Figure 11a) are generally lower than any other package configuration (compare with Figures 2b, 4b, 6b, 8b, and 10a). Spectral ratios are highly variable between the two packages, with horizontal amplitudes varying by as much as a factor of 30 between collinear horizontals and typically by a factor of 4. also likely caused by seismic energy generated by the currents. In hindsight, the current data would likely have been more coherent with the seismic data if it had been located closer to the seismic sensors. Collins et al. [2002] attempt to 6. Data From Other Sensors 6.1. Correlation With Ocean Current [28] A current meter was located approximately 5 meters from the seismic packages during the tests, with the sensors oriented in the same directions as the seismic axes. Coherence data recorded (Figure 12) display some correlation with the collinear current direction, but short coherence lengths and the likelihood that currents interact with the pier supports complicate the interpretation. While the current interacting with the package may generate some of the correlating energy, some is Figure 9. Shaker response for package on the plate. Vertical response (black) follows the expected 18 db/ octave slope (green), but the horizontal response is distorted. 9of13

10 6.3. Correlation of Current and Pressure [30] In Figure 14, current appears to be partially coherent with the pressure signal at frequencies below 0.6 Hz, while the two hydrophones, located approximately 1 m apart, are well correlated below 0.5 Hz, with partial correlation at higher frequencies, the same frequencies as those showing significant coherence in Figure Low Frequencies [31] Acquisition of low-frequency (long time series) data was not the general procedure during the test because of limited time available, but a several files were obtained in the frequency band from 0.01 to 2.5 Hz (Figure 15). The hydrophone data appear to be well correlated from below 0.01 Hz to 0.3 Hz, while the Guralp horizontal seismic sensors all show a peak in coherence near 0.13 Hz, decreasing at higher and lower frequencies. Unfortunately, no files were obtained comparing the two buried packages at low frequencies, but Figure 4a suggests that the coherence between buried sensors would be high. 7. Discussion [32] Users of seismic sensors assume that their data record rectilinear motion of the ocean floor although theory predicts that packages will respond Figure 10. Plate-Bottom. (a) Coherence between collinear Guralp sensors in the package on the plate and the package on the bottom. (b) Spectral ratios (Bottom/Plate) between collinear sensors show higher amplitudes observed when the sensors are on the bottom than when they are mounted on the plate. quantify the effects of currents at the OSN-1 Pilot Experiment Correlation With Pressure [29] Two hydrophones were located about 5 m from the seismic packages. In the deep ocean hydrophone signals are often well correlated with the signal recorded by vertical seismic sensors, but that is not the case in this shallow-water test. Coherence values between the vertical seismic sensors and the hydrophones in this test are surprisingly low (Figure 13), again likely because of the short coherence lengths. Figure 11. Bottom-Bottom. (a) Coherence between collinear Guralp sensors with both packages on the bottom. (b) Spectral ratios between collinear sensors with both packages on the bottom. 10 of 13

11 Figure 12. Correlations with current. Blue curves are for x direction; red are for y direction. Note that the frequency scale covers 0.1 to 1 Hz. There is no appreciable energy in the current signal above 1 Hz. The dotted curves are for correlation of current with sensors on the bottom, dashed curves are for sensors on the plate, and blue curves are for buried sensors. Figure 14. Coherence between two hydrophones located approximately 1 m apart (black curve), and between the hydrophone signals and current sensors (blue, x direction; red, y direction). to relative motion between the water and ocean floor by tilting and by differential translation [Duennebier and Sutton, 1995]. A seismic package will respond well to low-frequency motions of the ocean floor provided that the ocean water and ocean bottom are moving together. When a bottom current flows past the package it will exert a force on the package resulting in tilt that will be recorded as a horizontal acceleration of the ocean floor. The magnitude of this tilt signal will depend on frequency, the speed of the current, the softness of the sediments, and the configuration of the package. At high frequencies, the top of the package attempts to move with (or faster than) the water while the bottom of the package tends to move with the ocean floor, resulting in dynamic tilt. Vertical motion is much less of a problem since vertical motion is continuous across the ocean floor, and since vertical sensors are much less sensitive to tilt than horizontal sensors. [33] The purpose of this test was to determine whether burial of seismic sensors under the ocean floor or mounting the sensors on a large plate appreciably increases signal fidelity compared to mounting the sensors in a package with a relatively small radius and bottom area in contact with the Figure 13. Coherence of vertical Guralp sensor data with hydrophone data approximately 6 m from the sensors. Black curves are for buried vertical sensors, red curves are for vertical sensors on the bottom, and the blue curve is for a vertical sensor on a 1 m diameter plate. No data correlating hydrophone signals with horizontal sensors were obtained in the test. Figure 15. Low-frequency coherence between collinear current and horizontal Guralp sensors (red and orange), between buried and bottom Guralp horizontal sensors (blue), and between the two hydrophones (green). 11 of 13

12 Figure 16. Coherence of horizontal Guralp sensors with vertical sensors. In this case, low coherence indicates better fidelity. Black curves are for buried sensors, blue are for sensors on the plate, and red sensors are in the packages on the bottom. ocean floor. We find that burial of the sensor package in a material of roughly the same density as the package provides signal fidelity nearly equivalent to the fidelity obtained when the packages are resting next to each other on a concrete floor. This does not imply that signal levels and signal-to-noise values for buried ocean bottom sensors will be as good as those at a good land seismic station, since the noise levels and signal amplification present in ocean sediments can be much larger than observed on land, although buried sensor noise levels can approach those of a good seismic observatory [Collins et al., 2001]. The highest fidelity seismic instrument can still only detect the ambient ground motion. [34] Horizontal signal fidelity can be severely degraded if the package is placed directly on the ocean floor or on a plate on the ocean floor, rather than buried. Strong evidence for distortions are documented in this test by (1) low coherence of collinear sensors between the buried package and the package on the plate and package on the bottom, (2) large variations in spectral ratios of noise between collinear sensors in the same package when on the plate and package on the bottom, (3) large amplification of horizontal amplitudes relative to identical sensors in the buried package, and (4) distortions in shaker response. [35] Quantification of this signal degradation and extrapolation to other conditions is difficult since coupling parameters can and do change each time the package is placed on the ocean floor. The amount of degradation is expected to change with frequency, current speed, source direction (for asymmetric packages), and shear strength and density of the sediments. The location of the package tilt axis (if any) will likely depend on the package shape, bottom conditions, and directions of motion. [36] Vertical sensors are considerably less susceptible to coupling problems than horizontal sensors, although package tilt can cross-couple horizontal ground motion into vertical motion at the sensor location if the sensor is located away from the tilt axis in the horizontal direction [Duennebier and Sutton, 1995] or if the package is not correctly leveled [Crawford et al., 2006]. Such cross-coupling is displayed in Figure 16, where the coherences between the vertical Guralp sensors in the packages are compared with the horizontal sensors in the same package. Except at frequencies below 0.2 Hz, where the spectral estimates are questionable, coherence between horizontal and vertical sensors in the buried package is uniformly low, while coherence between horizontal and vertical sensors in the packages on the bottom are consistently much higher. The data for the sensors on the plate have intermediate values. [37] Hydrophones are not sensitive to ground motion, unless that motion, such as vertical motion, also changes the ambient pressure. This is the reason that hydrophones make excellent seismic sensors in the ocean, with uniformly consistent fidelity. High coherence between the hydrophones in this test is likely only limited by coherence length. [38] The tests described here were in shallow water with relatively high currents (as high as 10 cm/s), using a relatively tall sensor package, and thus the tests might be considered a worst-case. However, the sandy coral sediments at the site have a relatively high shear strength and density compared to deep-ocean sediments, and they provide a more rigid base for the package to rest on than the soft deep-ocean sediments, resulting in less distortion for the same current. [39] These results suggest strongly that methods be developed to cheaply and reliably bury ocean bottom seismometer seismic sensor packages when they are deployed in soft sediment. While ROV burial of sensors is suitable for permanent observatories [Duennebier et al., 2002], it is too expensive for temporary deployments. None of the existing broadband ocean bottom seismometers 12 of 13

13 utilize buried sensors, and, as a result, it is likely that much of the ocean bottom seismometer data collected to date are seriously degraded. [40] Suggestions for future work: It would be highly desirable to repeat this test when a cheap burial system is developed. The data from the Makai-95 Test presented here suffer because of the proximity of structures that likely broadcast seismic waves and cause short coherence lengths. A repeat experiment at a site with minimal local inhomogeneities would likely result in clearer results. Extension of the tests presented here to lower frequencies is highly desirable, as would be tests in high and low current and soft and hard bottom situations. [41] A better shaker could be designed that could operate from DC to high frequencies, with the force decreasing with increasing frequency to keep the response above noise level over a wider frequency band. Acknowledgments [42] George Sutton, mentor, advisor, and friend, died on 26 January 2004, before we could finish this paper. He is greatly missed by all who knew him. The support of the S.O.E.S.T. Engineering Facility in design, fabrication, and testing of the seismic packages and preparing for and conducting this test is greatly appreciated. We also thank the Hawaii Undersea Research Laboratory for the use of their facility at Makai Pier, Oahu. References Collins, J. A., F. L. Vernon, J. A. Orcutt, R. A. Stephen, K. R. Peal,F.B.Wooding,F.N.Spiess,andJ.A.Hildebrand (2001), Broadband seismology in the oceans: Lessons from the Ocean Seismic Network Pilot Experiment, Geophys. Res. Lett., 28(1), Collins, J. A., F. L. Vernon, J. A. Orcutt, and R. A. Stephen (2002), Upper mantle structure beneath the Hawaiian swell: Constraints from the ocean seismic network pilot experiment, Geophys. Res. Lett., 29(11), 1522, doi: / 2001GL Crawford, W. C., and S. C. Webb (2000), Removing tilt noise from low frequency (<0.1 Hz) seafloor vertical seismic data, Bull. Seismol. Soc. Am., 90(4), Crawford, W. C., R. A. Stephen, and S. T. Bolmer (2006), A second look at low-frequency marine vertical seismometer data quality at the OSN-1 site off Hawaii for seafloor, buried and borehole emplacements, Bull. Seismol. Soc. Am., 96(5), Duennebier, F. K., and G. H. Sutton (1995), Fidelity of ocean bottom seismic observations, Mar. Geophys. Res., 17, Duennebier, F. K., et al. (1991), Geoacoustic noise from 0. 1 to 100 Hz recorded off Oregon: The ULF/VLF Experiment, in Oceans 91 Proceedings, vol. 1, pp , Inst. of Electr. and Electron. Eng., New York. Duennebier, F. K., D. W. Harris, J. Jolly, J. Babinec, D. Copson, and K. Stiffel (2002), The Hawaii-2 Observatory Seismic System, IEEE J. Oceanic Eng., 27(2), Sutherland, F. H., F. L. Vernon, J. A. Orcutt, J. A. Collins, and R. A. Stephen (2004), Results from OSNPE: Improved teleseismic earthquake detection at the seafloor, Bull. Seismol. Soc. Am., 94, Sutton, G. H., and F. K. Duennebier (1987), Optimum design of ocean bottom seismometers, Mar. Geophys. Res., 9, Sutton, G. H., et al. (1981), An overview and general results of the Lopez Island OBS experiment, Mar. Geophys. Res., 5(1), of 13