Download Hydroacoustic monitoring of seismicity at the slow

GEOPHYSICAL RESEARCH LETTERS, VOL. 29, NO. 11, 10.1029/2001GL013912, 2002
Hydroacoustic monitoring of seismicity at the slow-spreading
Mid-Atlantic Ridge
Deborah K. Smith,1 Maya Tolstoy,2 Christopher G. Fox,3 DelWayne R. Bohnenstiehl,2
Haru Matsumoto,4 and Matthew J. Fowler4
Received 13 August 2001; accepted 20 September 2001; published 8 June 2002.
[1] In February 1999, long-term hydroacoustic monitoring of the
northern Mid-Atlantic Ridge (MAR) was initiated. Six
autonomous hydrophones were moored between 15°N and
35°N on the flanks of the MAR. Results from the first year of
data reveal that there is significant variability in along-axis event
rate. Groups of neighboring segments behave similarly, producing
an along-axis pattern with high and low levels of seismic activity
at a wavelength of 500 km. This broad scale pattern is likely
influenced by the axial thermal regime. Several earthquake
sequences with variable temporal characteristics were detected,
suggesting fundamental differences in the cause of their seismicity.
Off-axis, most seismic faulting occurs within a zone <15 km from
the axis center.
INDEX TERMS: 3035 Marine Geology and
Geophysics: Midocean ridge processes
1. Introduction
[2] Seismic studies of mid-ocean ridges have suffered from
two primary limitations: land-based networks cover broad areas
over long time periods, but detection is limited to large magnitude earthquakes; ocean bottom seismometer experiments detect
the microseismicity associated with spreading-center processes,
but only monitor small sections of the ridge for short durations.
The problem of recording small-magnitude earthquakes over
large regions of a mid-ocean ridge for long periods of time
has been addressed by NOAA’s Pacific Marine Environmental
Laboratory (PMEL). Since 1990, PMEL’s work, using the Navy’s
SOSUS hydrophone arrays in the NE Pacific and an autonomous
hydrophone array in the equatorial Pacific, has demonstrated that
small magnitude earthquakes can be detected at long range using
the acoustic propagation properties of the oceanic water column
[Fox et al., 2001].
[3] In February 1999, six of NOAA/PMEL’s autonomous
hydrophones were moored in the North Atlantic. Each hydrophone is capable of recording 1 to 50 Hz of acoustic signal
continuously for up to 14 months. The configuration of the
hydrophone array (Figure 1) allows earthquakes within the array
to be located with sufficient accuracy to be associated with largescale seafloor structures. The region of the MAR within the array
encompasses most of the French American Ridge Atlantic
(FARA) study area that extends from 15°-40°N. The wealth of
information available for the 15°-40°N area makes it ideal to
monitor, as existing studies provide a framework within which
the data can be interpreted.
1
Department of Geology and Geophysics, Woods Hole Oceanographic
Institution, Woods Hole, Massachusetts, USA.
2
Lamont-Doherty Earth Observatory, Palisades, New York, USA.
3
NOAA/PMEL, Newport, Oregon, USA.
4
OSU/NOAA, Cooperative Institute for Marine Resources Studies,
Newport, Oregon, USA.
Copyright 2002 by the American Geophysical Union.
0094-8276/02/2001GL013912$05.00
2. Results from the First Year of Data
[4] The data from the first year of deployment yielded 3,240
earthquakes within a geographical range from 58°N to 40°S.
Within the hydrophone array, where detection levels and earthquake location accuracy are the best (location errors < 2 km) [Fox
et al., 2001], a total of 2,015 earthquakes were detected, 1,495 of
which were detected on four or more hydrophones (Figure 1). This
event detection rate suggests a magnitude of completeness (Mc) of
3.0 Mw (assuming the Gutenburg-Richter relationship observed
teleseimically can be extrapolated to lower magnitudes), with a
number of smaller events (< 2.5 Mw) also being recorded
[Bohnenstiehl et al., 2000]. This represents an improvement of
more than an order of magnitude relative to the teleseismic
detection level, which is 4.7 Mw for earthquakes in this area.
[5] Several first-order observations can be made. Earthquakes
are scattered across and along the axis of the MAR. We note that
although earthquake locations derived from the T-wave analysis
may not correspond directly to epicenter locations, our preliminary
analysis indicates that they are not broadly biased with regard to
topographic features, such as the crest of the rift valley walls. The
distribution of seismicity within individual segments varies. In
some spreading segments earthquakes are located throughout the
segment, while in others earthquakes are restricted to limited
regions.
[6] Three well-studied hydrothermal sites are located within the
array: TAG [Rona et al., 1976] (Segment 21), Broken Spur
[Murton et al., 1999] (Segment 28) and Snakepit [Brown and
Karson, 1988] (Segment 15). The TAG segment had 8 earthquakes
located on the valley floor within 10 km of the TAG hydrothermal
mound and an additional 8 earthquakes located within the eastern
bounding wall. The Broken Spur and Snakepit sites were less
active. Four earthquakes are located within 10 km of the Broken
Spur vent and 5 within 10 km of the axial volcanic ridge associated
with the Snakepit site.
[7] Two well studied inside corners (MARK at 23°N [Karson
et al., 1987] and Atlantis at 30°N [Blackman et al., 1998]) have
earthquakes associated with them, suggesting active faulting. The
major transforms and associated fracture zones (Oceanographer,
Hayes, Kane, Atlantis, Fifteen-Twenty) are quiet relative to the
spreading segments during the year’s deployment. This suggests
that the seismic moment deficiency observed teleseismically along
transform faults [Abercrombie and Ekström, 2001] may not be
accounted for by an abundance of smaller magnitude earthquakes.
[8] Figure 2 displays a histogram of the normalized number of
earthquakes vs. spreading segment. Figure 2a reveals an along-axis
pattern with regions exhibiting high and low levels of inner valley
floor activity. The wavelength of this pattern is on the order of
450 – 550 km. Groups of spreading segments appear to behave
similarly. Differences in event rate between groups of segments
may reflect variability in the thermal state of the ridge (inferred
from bathymetry and gravity data [Thibaud et al., 1998]), with the
presence of a thinner brittle layer within ‘‘hotter’’ segments
resulting in fewer detectable earthquakes. Therefore, the longwavelength pattern may reflect the deep crustal/upper mantle
processes that are influencing regional, rather than segment-scale,
magma supply.
13 - 1
13 - 2
SMITH ET AL.: HYDROACOUSTIC MONITORING OF THE MAR
earthquakes in these segments is persistently reduced due to
differences in the geologic setting.
[13] Several large sequences of earthquakes have been identified within the first year of data. The largest sequence occurred
within the western bounding walls of the rift valley at 24.5°N
Figure 1. Earthquake locations. Insert shows hydrophone
locations as stars. Two areas are outlined: (a) Oceanographer to
Atlantis fz; (b) Atlantis to Fifteen-Twenty fz. Circles: Earthquake
locations identified using 4 or more hydrophones. Lines along the
axis: Individual spreading segments identified from published
interpretations [Detrick et al., 1995; Smith and Cann, 1992;
Thibaud et al., 1998] and our interpretation of the bathymetry.
Segment numbers start at the Fifteen-Twenty FZ.
[9] Within the ridge section between the Fifteen-Twenty and
Kane transforms activity decreases toward the center where there is
a broad swell in the topography (consistent with a hotter thermal
regime) centered near 19.5°N. The shallowest axial depths between
the Fifteen-Twenty and Kane transforms occur near 20.5°N (Segment 11) and 21.5°N (Segment 13), where there are also relatively
large gaps in seismicity (Figure 1b).
[10] Between the Kane and Atlantis, and the Atlantis and Hayes
transforms a different trend is observed with activity generally
increasing with distance from the transforms. Both of these areas
are topographically influenced by the Azores hotspot and going
northward there is a general shallowing in the bathymetry, a
decrease in the relief of the valley walls, and an overall broadening
of a topographic swell centered on the ridge. The most inflated
segments in these areas are directly south of the Atlantis (Segments
28, 29) and Hayes (Segment 37) transforms, and are notably quiet
seismically. In the middle of the Kane-Atlantis ridge section the
centers of Segments 23 and 25 had almost no activity during the
year, again perhaps linked to their thermal state.
[11] Earthquake activity is low in Segments 30 – 33 just to the
north of the Atlantis transform. In this case, however, there does
not seem to be a relationship to topography. Further to the north,
Segments 37 – 40 are relatively quiet. Segment 37 is just to the
south of the Hayes transform and Segments 38, 39 and 40
correspond to Segments OH3, OH2 and OH1 [Detrick et al.,
1995].
[12] The one-year pattern of seismicity cannot be predicted
from the 10-yr pattern of teleseismicity (Figure 2b). Segments
with no history of teleseismic activity within the last ten years have
similar levels of normalized numbers of earthquakes as those that
have experienced a lot of teleseismic activity (compare Segments
23 and 24). This suggests that these segments are producing
seismicity with magnitudes consistently below the threshold of
the land-based networks during the time period of observation. It is
entirely possible that given a longer observational period as on the
land-based systems, the same pattern of seismicity would emerge,
but conversely, it is possible that the overall magnitude range of
Figure 1.
(continued)
SMITH ET AL.: HYDROACOUSTIC MONITORING OF THE MAR
13 - 3
and ultra-slow spreading rates [e.g., Embley et al., 1995], it seems
unlikely that these events represent full-scale eruptions. One
possibility is that the events may represent a response to deformation resulting from movement of magma at depth. This is speculative, however, and on site measurements would be required to
identify the source of these sequences definitively.
[16] Many aspects of extensional faulting remain poorly constrained at the MAR, including the timing of fault initiation and
growth. Morphologic indicators (e.g., talus slopes and scarp
height) suggest active faulting continues out to 10 – 40 km from
the axis [Escartı́n et al., 1999]. Results from our data indicate
seismic faulting decreases to undetectable levels at distances >15
km from the axis. Variability in the width of the active plate
boundary zone between and within segments will be assessed in
more detail as monitoring continues.
[17] There is one notable exception to concentration of activity
within 15 km of the axis. Between 13° and 14°N a segment length
band of ridge-parallel seismicity is observed 70 km west of the
ridge axis (Figure 1b). The origin of this seismicity may be due to
unusual plate stresses, or the initiation of a ridge jump within the
area. Clearly, a suite of geological and geophysical data needs to be
collected in this region to understand the processes at work.
[18] A variety of other acoustic signals are found in the data. A
large landslide deposit [Tucholke, 1992] in Segment 22 has
Figure 2. Number of earthquakes (normalized to a 50-km long
segment) versus spreading segment. (a) Earthquakes located within
the valley floor (423 events). The valley floor is defined as the region
between the inner-most bounding faults with scarp heights >200 m.
(b) Earthquakes located within a 30 km swath centered on the axis.
Gray bars: Hydrophone events (1046 events). White bars: Teleseismic events during the same time period (22). Black bars:
Teleseismic earthquakes for the 10-yr period 1990 – 2000 (261).
(Segment 16). Nearly 200 aftershocks followed a magnitude 5.3
normal-faulting event in April 1999 (Figure 3a). Following the
mainshock, the rate of activity decays in a manner consistent with
that commonly observed after large magnitude tectonic events,
following a modified Omori’s Law [Utsu et al., 1995].
[14] An earthquake sequence of a somewhat different nature
was initiated in March 1999 following a magnitude 5.1 event near
22.5°N (Figure 3b). The long-term activity is inconsistent with
Omori’s Law, with activity (0.3 events/day) in an area < 20
km from the mainshock continuing for a period of >1 year. Events
are located primarily on the floor of the rift valley. The character
of these sequences could not be deduced from the teleseismic data
alone, as they are associated with only 1 to 3 teleseismic events.
[15] The aftershock sequences that follow Omori’s Law are
likely tectonic extension events, which in the case of the event at
24.5°N is consistent with the location on the valley wall. Seismic
activity near 22.5°N may be more volcanically influenced, consistent with the location on the valley floor. The size-frequency
distribution of these events, however, is not significantly different
than that observed elsewhere within the array. This is inconsistent
with the higher b-values sometimes correlated with the presence of
magma in other volcanic regions [Wiemer and McNutt, 1997].
Given what we know about seafloor eruptions at fast, intermediate
Figure 3. Characteristics of two earthquake sequences; locations
are marked on Figure 1b. (a) Source power versus time in days
since 01/01/99 for events within the 24.5°N sequence. Events are
located primarily along the western bounding wall. Three events
were detected by land-based seismic networks. The moment tensor
solution (Harvard CMT) of the largest mainshock event indicates
normal slip along a ridge-parallel fault. The rate of aftershock
activity decays rapidly following the mainshock, as predicted by
Omori’s Law. (b) Source power versus time in days since 01/01/99
for events within the 22.5°N sequence. Events are located
primarily on the valley floor. The largest event in this sequence
also exhibits a normal sense of slip. It occurs near the beginning of
this sequence and a moderate event rate (0.3 earthquakes/day)
continues throughout the year.
13 - 4
SMITH ET AL.: HYDROACOUSTIC MONITORING OF THE MAR
numerous earthquakes scattered on and near it, and it is possible
that these earthquakes represent subsidence or adjustment to
loading due to the deposit. The hydroacoustic method is also
capable of detecting landslide events at long range [CaplanAuerbach et al., 2001], although no such events were observed
in this data set. Many non-geological signals are also evident,
including a wide range of marine mammal vocalizations. Examination of these vocalizations in the Pacific has led to a new
understanding of whale distributions in the generally unobserved
areas of the open ocean [Stafford et al., 1999]. Perhaps the most
dominant signal observed in the data set is not natural at all, but
rather the sound of seismic profilers working offshore Canada,
Brazil, and western Africa. Ships passing through the array are also
clearly recorded. None of these signals mask the signals from the
earthquakes.
References
Atlantic Ridge: The median valley of the MARK area, Mar. Geophys.
Res., 10, 109 – 138, 1988.
Caplan-Auerbach, J., C. G. Fox, and F. K. Duennebier, Hydroacoustic
detection of submarine landslides on Kilauea volcano, Geophy. Res. Lett.,
28, 1811 – 1814, 2001.
Detrick, R. S., H. D. Needham, and V. Renard, Gravity anomalies and
crustal thickness variations along the Mid-Atlantic Ridge between
33°N and 40°N, J. Geophys. Res., 100(B3), 3767 – 3787, 1995.
Embley, R. W., W. W. Chadwick, I. R. Jonasson, D. A. Butterfield, and
E. T. Baker, Initial results of the rapid response to the 1993 CoAxial
event: Relationships between hydrothermal and volcanic processes,
Geophys. Res. Lett., 22, 143 – 146, 1995.
Escartı́n, J., et al., Quantifying tectonic strain and magmatic accretion at a
slow-spreading ridge segment, Mid-Atlantic Ridge, 29°N, J. Geophys.
Res., 104, 10,421 – 10,437, 1999.
Fox, C. G., H. Matsumoto, and T.-K. Lau, Monitoring Pacific Ocean Seismicity from an Autonomous Hydrophone Array, J. Geophys. Res., 106,
4183 – 4206, 2001.
Karson, J. A., et al., Along-axis variations in seafloor spreading in the
MARK area, Nature, 328(681 – 685), 1987.
Murton, B. J., L. J. Redbourn, C. R. German, and E. T. Baker, Sources
and fluxes of hydrothermal heat, chemicals and biology within a segment of the Mid-Atlantic Ridge, Earth Planet. Sci. Lett., 171, 301 – 317,
1999.
Rona, P. A., R. N. Harbison, B. G. Bassinger, R. B. Scott, and A. J. Nalwalk, Tectonic fabric and hydrothermal activity of Mid-Atlantic Ridge
crest (Lat. 26°N), Geol. Soc. Am. Bull., 87, 661 – 674, 1976.
Smith, D. K., and J. R. Cann, The role of seamount volcanism in crustal
construction at the Mid-Atlantic Ridge (24° – 30°N), J. Geophys. Res., 97,
1645 – 1658, 1992.
Stafford, K. M., S. L. Nieukirk, and C. G. Fox, Low- frequency whale
sounds recorded on hydrophones moored in the eastern tropical Pacific,
J. Acoust. Soc. Am., 106, 3687 – 3698, 1999.
Thibaud, R., P. Gente, and M. Maia, A systematic analysis of the MidAtlantic Ridge morphology and gravity between 15°N and 40°N: Constraints of the thermal structure, J. Geophys. Res., 103, 24,223 – 24,243,
1998.
Tucholke, B. E., Massive submarine rockslide on the rift valley wall of the
Mid-Atlantic Ridge, Geology, 20, 129 – 132, 1992.
Utsu, T., Y. Ogata, and R. S. Matsu’ura, The centenary of the Omori formula for the decay law of aftershock activity, J. Phys. Earth, 43(1 – 33),
1995.
Wiemer, S., and S. McNutt, Variations in frequency-magnitude distribution
with depth in two volcanic areas: Mount St. Helens, Washington, and Mt.
Spurr, Alaska, Geophys. Res. Letts., 24, 189 – 192, 1997.
Abercrombie, R. E., and G. Ekström, Earthquake slip on oceanic transform
faults, Nature, 410, 74 – 77, 2001.
Blackman, D. K., J. R. Cann, B. Janssen, and D. K. Smith, Origin of
extensional core complexes: Evidence from the Mid-Atlantic Ridge at
Atlantis Fracture Zone, J. Geophys. Res., 103, 21,315 – 21,333, 1998.
Bohnenstiehl, D. R., M. Tolstoy, D. K. Smith, and C. G. Fox, Earthquake
Sequences Detected Using Autonomous Underwater Hydrophone Data
From the Northern Mid-Atlantic Ridge: February 1999 – February 2000,
Eos Trans. AGU, 81, Fall Meet. Suppl., Abstract T51D-13, 2000.
Brown, J. R., and J. A. Karson, Variations in axial processes on the Mid-
D. R. Bohnenstiehl and M. Tolstoy, Lamont-Doherty Earth Observatory,
Palisades, New York 10964, USA.
M. J. Fowler and H. Matsumotu, OSU/NOAA, Cooperative Institute for
Marine Resources Studies, Newport, Oregon 97365, USA.
C. G. Fox, NOAA/PMEL, Newport, Oregon 97365, USA.
D. K. Smith, Department of Geology and Geophysics, Woods Hole
Oceanographic Institution, MS 22, Woods Hole, MA 02543, USA.
3. Conclusions
[19] The data from one year of hydroacoustic monitoring of
the northern MAR have revealed many previously unknown
aspects of the overall pattern of seismicity at this slow-spreading
ridge. These data have provided us with a snapshot of the
activity in one specific year. Differences between this data set
and the longer land-based results are due in part to the time
periods observed, but may also reflect differences in the geological processes being observed. As the hydroacoustic data set
is extended in the coming years, the sources of these differences
will become more apparent and our understanding of the longterm behavior of this slow-spreading ridge improved.
[20] Acknowledgments. This project was funded by NSF grant OCE9811575. We thank the captains and crews of the R/V EWING and R/V
KNORR. The manuscript benefited from a review by K. Macdonald and an
anonymous reviewer. LDEO contribution number 6253. WHOI contribution number 10536.