GEOPHYSICAL RESEARCH LETTERS, VOL. 30, NO. 2, 1098, doi:10.

1029/2002GL015649, 2003

The wet Nicaraguan slab

Geoffrey A. Abers and Terry Plank
Department of Earth Sciences, Boston University, Boston, MA, USA

Bradley R. Hacker
Department of Geological Sciences, University of California, Santa Barbara, CA, USA
Received 11 June 2002; revised 2 September 2002; accepted 12 September 2002; published 31 January 2003.

[1] Nicaraguan volcanoes show globally high concentra- used to infer a high contribution from altered oceanic crust
tions of geochemical tracers from dehydration of subducting [Patino et al., 2000]. Here, we test for ‘‘wet’’ crust by
crust, which may reflect a slab with unusually high amounts measuring its seismic velocity at sub-arc depths, and find
of H2O. To test this possibility, we measure seismic that the Nicaragua slab segment may be the most hydrated
velocities at the top of the subducted plate and compare yet observed beneath a volcanic arc.
them with predictions for hydrated mafic rocks. Regional
seismic P waves for intraslab events at 100 –150 km depth
show a high-frequency late arrival, apparently trapped in a 2. Data and Methods
low-velocity waveguide 2.5 – 6 km thick at the top of the [4] We examine signals from earthquakes >100 km depth
downgoing plate, 14.5 ± 2.2% slower than surrounding that travel along the strike of the Cocos slab beneath
mantle. The velocities can be explained by 5 wt % H2O in Nicaragua. Such signals travel subhorizontally near the
the subducted crust, 2 – 3 times the hydration inferred for top of the slab for long distances and show considerable
other slabs by similar methods. This interpretation implies distortion as a consequence [Gubbins et al., 1994]. Such
extensive hydration of the Cocos Plate off Nicaragua, signals show early low-frequency arrivals followed by later
perhaps enhanced by up-dip fluid flow within the slab at high-frequency and high-amplitude signals, a behavior best
>100 km depth. INDEX TERMS: 3660 Mineralogy and understood as dispersion caused by interaction with a low-
Petrology: Metamorphic petrology; 7203 Seismology: Body velocity waveguide [Abers, 2000]. Comparison of the
wave propagation. Citation: Abers, G. A., T. Plank, and B. R. arrival times between the high- and low-frequency waves
Hacker, The wet Nicaraguan slab, Geophys. Res. Lett., 30(2), allows the velocity of the waveguide to be estimated, and
1098, doi:10.1029/2002GL015649, 2003. the signal’s frequency content constrains layer thickness. In
the five Pacific arcs previously examined in this way,
Alaska, the Aleutians, Kuriles, Honshu, and the Marianas,
1. Introduction
the waveguides are 5 – 8% slower than the surrounding
[2] Arc lavas feature many geochemical tracers that medium for both P and S waves, and the waveguides are
require input from subducted oceanic igneous crust and estimated to be 2 – 8 km thick [Abers, 2000].
sediment [Miller et al., 1994; Tera et al., 1986]. These [5] The Global Seismic Network station JTS, in central
geochemical signals vary significantly from arc to arc. Some Costa Rica, records Nicaraguan events that have propagated
variations are explained by differing sediment input [Plank along the strike of the downgoing plate (Figure 1). We
and Langmuir, 1993], but for others the cause remains analyze JTS broadband records from 20 earthquakes for the
unclear. Are the variations primarily caused by differences years 1995 –1999 at depths of 100 – 250 km, distances of
in the hydration of the downgoing plate, or by the processes 1– 5
, and magnitudes > 4.0. These records include many P
that transmit fluids from slab to arc? Part of the uncertainty waveforms that show secondary phases (Figure 2); S wave
lies in the lack of information on the state of the subducted records are too noisy and complex to interpret.
material at depths where devolatilization occurs. Recent [6] To understand the path taken by these signals, raypaths
seismological evidence for a thin, low-velocity layer atop are traced in a 3D velocity model of a dipping slab, with a top
many slabs at subarc depths [e.g., Abers, 2000; Helffrich, defined by contours of seismicity and a velocity 5% faster
1996] provides some evidence that the slab remains at least than the overlying mantle. Rays exit the slab at points 50
partially hydrated to depths beyond the volcanic front. km horizontally from the station and 75 km deep. Small
[3] Globally, the Nicaraguan volcanic arc shows some of perturbations to slab geometry (lateral shifts <20 km) or
the highest levels of geochemical tracers for oceanic crustal changes to velocity (to 8%) do not appreciably affect
fluid, including the global maximum concentration of 10Be estimates of ray path lengths, although variations in event
[Tera et al., 1986] and among the highest B/La ratios [Noll location of 25 km (characteristic of teleseismic mislocation)
et al., 1996; Patino et al., 2000]. One possible explanation lead to 5 – 20 km scatter in ray exit depths. Errors in path
is that the flux of slab-derived fluids is overall high here, length produce scatter in single-event velocity measurements
and that Be and B are efficiently removed from the slab by that controls the error bars for velocity quoted below.
these fluids [Turner et al., 1998]. In Nicaragua, high B
concentrations in lavas [Leeman et al., 1994] have been
3. Results
Copyright 2003 by the American Geophysical Union. [7] The secondary phase is clearest for events with depths
0094-8276/03/2002GL015649 of 100– 150 km, and shows linear moveout relative to direct

70 - 1
70 - 2 ABERS ET AL.: THE WET NICARAGUAN SLAB

represent signals caused by interaction with upper-plate

crust except within 10 km of JTS. Such a scatterer would
not produce the linear moveout and dispersion. Second,
velocity estimates from phase moveout (Figure 2) and from
3D ray tracing are identical, suggesting that ray bending
caused by the subducting slab does not bias velocity
estimates. Third, velocity anisotropy probably does not
produce the variation of observed signals between Nicar-
agua and other slabs [Abers, 2000]. Ray paths are along-
strike in all cases, and at both Nicaragua and the Aleutians
the subducting crust was generated at ridges subparallel to
the trench. Hence, effects of anisotropy should similar
between Nicaragua and the other slabs. In conclusion, a
Figure 1. Central America and raypaths. Thin dotted lines low-velocity waveguide along the top of the subducting slab
contour top of seismicity. Hypocenters (circles) from PDE best explains the observations.
catalog. Thick black lines show portions of raypaths within
the fast slab, and thick gray lines show portions above it. 4. Discussion
[11] The layer thickness determined here for Nicaragua is
comparable to other northern Pacific arcs (Alaska, the
P (Figure 2). Its relative velocity is estimated in two ways. Aleutians, Kuriles, Honshu, and the Marianas), but the
First, linear regression of picked secondary arrivals (Figure wave speed anomaly is roughly double the 5 – 8% seen
2) provides an apparent velocity that is 14.5% ± 2.2% elsewhere [Abers, 2000]. The 2.5– 6.0 km layer thickness is
slower than the direct arrival. Their moveout reaches zero at comparable to part or all of the subducted oceanic crust, so
range of 50 km, suggesting that the delay is produced >50 we infer that the signal originates from propagation in slow
km distant from the station. At this distance signals depart subducted crust. Other possibilities may exist (discussed
the slab on their way to the surface (Figure 1), confirming below), but the layer thickness (> 2.5 km) precludes a
the supposition that the secondary arrival is slab-derived. narrow shear zone or metasediments (which should be 400
Events deeper than 160 km show secondary arrivals at m thick prior to compaction and dehydration [Kelly et al.,
inconsistent times, or do not show them. Either the deeper 2001]), as has been suggested by previously [Kerrick and
earthquakes are not located in a portion of the slab capable
of generating them (deep in the subducting mantle), or the
relevant structure is not present at these depths (e.g.,
eclogite has formed).
[8] Second, we fit measured dispersion of the P wave
train to a velocity model, as described in Abers [2000]. In an
example (Figure 3), the dominant arrival at frequencies
above 1.5 Hz arrives 3 seconds later than the initial,
low frequency signal. The measured dispersion curve is
inverted for 2 parameters describing the waveguide needed
to produce that behavior, a fractional velocity perturbation
of the layer relative to surroundings, dlnVp, and the layer
thickness H, using a simplex nonlinear minimization
scheme. This one record gives dlnVp = 14% and H = a
3.9 km. Sensitivity tests (white lines) show that 50%
changes to layer thickness (to 2.1 km or 6.9 km) result in
predicted dispersion that significantly violates the data.
[9] Dispersion curves for the 7 events shallower than 151
km are jointly inverted using this scheme, giving dlnVp = –
0.145 +0.031/ –0.038 and H = 2.9 + 0.5/ – 0.4 km (90%
confidence limits from bootstrap tests). In other words,
velocities within the waveguide appear to be 14.5% slower
than surroundings, the same value inferred from Figure 2.
These results assume dispersion by fundamental-mode
acoustic excitation of the waveguide. Numerical experi-
ments (not shown) suggest that the fundamental-mode
approximation may underestimate layer thickness by up to
a factor of two depending upon details of source location Figure 2. Record section aligned by first P arrival. Vertical
and focal mechanism, so the upper bound on layer thickness component broadband record is bandpass filtered at 0.75–
should be twice the inversion estimate, or H = 2.5 – 6.0 km. 8.0 Hz. Depths labeled on left. Gray lines: event depths >
[10] The ray geometry and travel-time systematics place 150 km; circles: picked high-frequency secondary arrivals,
constraints on the origin of these signals. First, the secon- gray for event depth >150 km. Dashed line shows linear
dary phases arrive as little as 1 s after the first P, too early to regression on secondary arrivals for events <150 km deep.
 ABERS ET AL.: THE WET NICARAGUAN SLAB 70 - 3

would produce the requisite seismic velocities, it is difficult

to see why such an anhydrous layer would persist at Nicar-
agua and not elsewhere. A pronounced system of normal
faults penetrate to depths of >20 km at the outer rise [Ranero
et al., 2001], which should provide effective conduits for
hydration of the subducting crust and mantle offshore
Nicaragua [von Huene et al., 2000]. Such features exist at
some many outer rises and could vary in their efficiency of
hydration, but those off Nicaragua appear typical.
Figure 3. Spectrogram for record designated by (*) on [16] A wetter subducting plate at Nicaragua would deliver
Figure 2. Lighter shades correspond to higher power in greater H2O to the mantle than elsewhere, as well as
arrivals; circles show maximum-energy at each frequency. geochemical tracers for slab-derived fluids (e.g. boron).
Thick black line shows best fit for a low-velocity waveguide Indeed, mafic magmas in Nicaragua have H2O concentra-
for White lines show effect of increasing or decreasing tions among the highest globally [Roggensack et al., 1997].
assumed layer thickness 50%. [17] Were the base or top of a serpentinized layer a
dehydration boundary (subparallel to an isotherm), one
would expect at least some cold subduction zones to show
Connolly, 2001]. Likewise subduction may erode the base a waveguide several tens of km thick, the depth to serpen-
of the upper plate beneath the Nicaraguan forearc [Ranero et tine-out boundary [Hacker et al., 2002b]. Also, Vp/Vs
al., 2000], providing an additional source of material, but it constraints at other slabs are inconsistent with the high
seems unlikely that such a process could produce a layer Poisson’s ratio of serpentinite [Abers, 2000]. Hence, of the
>2.5 km thick. possibilities, the waveguide seems best explained as
[12] We compare the Nicaragua waveguide velocities to hydrated mafic crust.
predictions made for metamorphosed mafic and ultramafic [18] Still, this does not explain why the Nicaragua slab is
rocks using the database of Hacker et al. [2002a] (Figure 4). twice as wet as other slabs (Figure 4). The 5 wt% exceeds
All seismic velocities are calculated at a common pressure estimates for oceanic crust from drilling oceanic basalts
(3 GPa) and temperature (400
C) and referenced to a [Staudigel et al., 1996]. Some process must extensively
nominal background composition for surrounding mantle
(pyrolite).
[13] First, we consider the possible equilibrium mineral-
ogies at pressures of 2.5 – 5 GPa, corresponding to the
depths sampled seismically (80 – 150 km). The equilibrium
mineralogy of mafic rocks ranges from lawsonite-rich
blueschist (low temperature, 5.4 wt% H2O) to eclogite (high
temperature, dry). Surrounding mantle harzburgites could
hydrate to serpentinite or chlorite-bearing assemblages. The
H2O content of these rocks correlates with dlnVp (Figure 4),
such that only rocks containing 5– 8 wt% H2O can produce
the observed low velocity layer in Nicaragua. The appro-
priate mafic assemblage is a fully saturated jadeite-law-
sonite blueschist, requiring 5.4 wt % crystallographically
bound H2O throughout the layer. The other North Pacific
slabs can be explained by partial saturation of the same
assemblage or higher-temperature rocks. The absence of
eclogite at Nicaragua indicates a substantial part of the crust
stays relatively cool, consistent with rapid subduction rate,
the presence of earthquakes to 300 km depth (following the
logic of Peacock and Wang [1999]) and thermal modeling Figure 4. Observed and predicted variations in seismic
extrapolated from Costa Rica [Hacker et al., 2002b]. velocities vs. crystallographically bound H2O content for
[14] Second, we consider compositions that may have potential rock types at 3 GPa, 400
C. Gray circles: rocks
been transported metastably to depth (Figure 4, open stable at 3 Gpa; open circles: rocks in equilibrium at lower
circles). Metastability is favored for rapid subduction, large pressures. Abbreviations: aE, amphibole eclogite; zaE,
grain sizes, and low H2O content [Hacker, 1996]. During zoisite-amphibole eclogite; laE, lawsonite-amphibole eclo-
subduction, fine-grained, hydrous assemblages (e.g., gite; jlB, jadeite-lawsonite blueschist; eA, epidote amphi-
hydrated open circles on Figure 4) are unlikely to survive bolite; GS, greenschist; PP, prehnite – pumpellyite; Z, zeolite.
to >2.5 GPa without re-equilibrating, as even trace quantities Values and calculations from Hacker et al. [2002a]. dlnVp:
of H2O can greatly catalyze reaction rates, but anhydrous, fractional P velocity change relative to reference anhydrous
coarse-grained gabbro may persist to such depths. Metasta- mantle (pyrolite-2). Dashed lines: Hashin-Shtrikman bounds
ble gabbro does have the appropriate seismic velocities. on velocities of mixtures between jlB and eclogite or
[15] Thus, there are three possibilities for the composition between harzburgite and hydrated harzburgite (serpentine +
of the waveguide at 100 – 150 km depth: fully saturated chlorite + brucite). Horizontal gray bands: constraints on
blueschists partly serpentinized harzburgite, or anhydrous low-velocity channels from dispersion for N. Pacific Slabs
gabbro. Although anhydrous metastable basalt or gabbro [Abers, 2000] and Nicaragua (this study).
70 - 4 ABERS ET AL.: THE WET NICARAGUAN SLAB

alter the oceanic crust such that it is saturated (>5 wt% H2O) Geophysical Monograph 106, edited by W. R. Buck, P. T. Delaney, J.
at the >2.5 GPa conditions sampled here. Vigorous hydro- A. Karson, and Y. Lagabrielle, pp. 177 – 218, Am. Geophys. Union,
Washington, DC, 1998.
thermal circulation and resulting hydration has been inferred Kelly, R. K., K. D. McIntosh, E. A. Silver, J. A. Goff, C. R. Ranero, and R.
from low heat flow (<20 mW m 2) on the Cocos plate V. Huene, A quantitative analysis of flexural faulting in the Cocos Plate at
seaward of the trench [Langseth and Silver, 1996]. How- the Middle America Trench from Nicaragua to Costa Rica, EOS Trans.
AGU, 82, Fall Meet. Suppl., F1148, 2001.
ever, it is unclear whether such cooling penetrates very deep Kerrick, D. M., and J. A. D. Connolly, Metamorphic devolatilization of
into the oceanic crust. Fluids may be introduced by the subducted oceanic metabasalts: implications for seismicity, arc magma-
extensive normal faulting at the outer rise offshore Nicar- tism and volatile recycling, Earth Planet. Sci. Lett., 189, 19 – 29, 2001.
Langseth, M. G., and E. A. Silver, The Nicoya convergent margin; a region
agua [Ranero et al., 2001], although the observed faults are of exceptionally low heat flow, Geophys. Res. Lett., 23, 891 – 894, 1996.
several km apart and so could not fully hydrates all of the Leeman, W. P., M. J. Carr, and J. D. Morris, Boron geochemistry of the
crust. Perhaps, fluid produced by dehydration reactions at Central American volcanic arc - constraints on the genesis of subduction-
>75 depth in the slab migrates up dip and hydrate subducted related magmas, Geochim. Cosmochim. Acta, 58, 149 – 168, 1994.
Miller, D. M., S. L. Goldstein, and C. H. Langmuir, Cerium/lead and lead
crust at shallower levels [Hacker et al., 2002b]. This isotope ratios in arc magmas and the enrichment of lead in the continents,
scenario would require a pathway that allows fluids to Nature, 368, 514 – 520, 1994.
ascend up the slab rather than into the overlying mantle Noll, P. D., H. E. Newson, W. P. Leeman, and J. G. Ryan, The role of
hydrothermal fluids in the production of subduction zone magmas: Evi-
wedge. The steep dip of the Nicaragua slab, relative to the dence from siderophile and chalcophile trace elements and boron, Geo-
other Pacific slabs studied, may facilitate such a pathway; chim. Cosmochim. Acta, 60, 587 – 611, 1996.
the other steep slab studied, in the Marianas, also shows a Patino, L. C., M. J. Carr, and M. D. Feigenson, Local and regional varia-
tions in Central American arc lavas controlled by variations in subducted
waveguide that is slower than average. Possibly all three sediment input, Contrib. Mineral. Petrol., 138, 265 – 283, 2000.
factors, hydrothermal circulation, outer-rise faulting and up Peacock, S. M., and K. Wang, Seismic consequences of warm versus cool
dip fluid flow, contribute to making the slab beneath subduction metamorphism: examples from southwest and northeast Ja-
Nicaragua among the wettest in the world. pan, Nature, 286, 937 – 939, 1999.
Plank, T., and C. H. Langmuir, Tracing trace elements from sediment input
to volcanic output at subduction zones, Nature, 362, 739 – 743, 1993.
[19] Acknowledgments. We thank the organizers of the 2001 MAR- Ranero, C. R., R. von Huene, E. Flueh, M. Duarte, D. Baca, and K.
GINS workshop on Central America for bringing focus to these problems. McIntosh, A cross section of the convergent Pacific margin of Nicaragua,
C. Ranero and an anonymous reviewer provided useful comments. This Tectonics, 19, 335 – 357, 2000.
work supported by National Science Foundation grants EAR-0096027 and Ranero, C. R., J. P. Morgan, K. D. McIntosh, and C. Reichert, Flexural
EAR-0096028. faulting and mantle serpentinization at the Middle America Trench, EOS
Trans. AGU, 82, Fall Meet. Suppl., 2001.
Roggensack, K., R. L. Hervig, S. B. McKnight, and S. N. Williams, Ex-
References plosive basaltic volcanism from Cerro Negro volcano: Influence of vo-
Abers, G. A., Hydrated subducted crust at 100 – 250 km depth, Earth and latiles on eruptive style, Science, 277, 1639 – 1642, 1997.
Planet. Sci. Lett., 176, 323 – 330, 2000. Staudigel, H., T. Plank, B. White, and H. R. Schminke, Geochemical fluxes
Gubbins, D., A. Barnicoat, and J. Cann, Seismological constraints on the during seafloor alteration of the basaltic upper oceanic crust: DSDP Sites
gabbro-eclogite transition in subducted oceanic crust, Earth and Planet. 417 and 418, in Dynamics of Subduction, Geophysical Monograph, 96,
Sci. Lett., 122, 89 – 101, 1994. edited by G. E. Bebout, D. Scholl, S. Kirby, and J. P. Platt, pp. 19 – 37,
Hacker, B. R., Eclogite formation and the rheology, buoyancy, seismicity, American Geophysical Union, Washington, D. C., 1996.
and H2O content of oceanic crust, in Subduction: Top to Bottom, AGU Tera, F., L. Brown, J. Morris, I. S. Sacks, J. Klein, and R. Middleton,
Monogr. Ser., 96, edited by G. E. Bebout, D. Scholl, S. Kirby, and J. P. Sediment incorporation in island-arc magmas: Inferences from 10Be,
Platt, pp. 337 – 346, American Geophysical Union, Washington, D. C., Geochim. Cosmochim. Acta, 50, 535 – 550, 1986.
1996. Turner, S., F. McDermott, C. Hawkesworth, and P. Kepezhinskas, A U-
Hacker, B. R., G. A. Abers, and S. M. Peacock, Subduction factory 1: series study of lavas from Kamchatka and the Aleutians: constraints on
Theoretical mineralogy, density, seismic wavespeeds, and H2O content, source composition and melt processes, Contrib. Mineral. Petrol., 133,
J. Geophys. Res., in press, 2002a. 217 – 234, 1998.
Hacker, B. R., S. M. Peacock, G. A. Abers, and S. D. Holloway, Subduction von Huene, R., C. R. Ranero, W. Weinrebe, and K. Hinz, Quaternary
Factory 2. Are Intermediate-Depth Earthquakes in Subducting Slabs convergent margin tectonics of Costa Rica, segmentation of the Cocos
Linked to Metamorphic Dehydration Reactions?, J. Geophys. Res., in Plate, and Central American volcanism, Tectonics, 19, 314 – 334, 2000.
press, 2002b.
Helffrich, G., Subducted lithospheric slab velocity structure: Observations
and mineralogical inferences, in Subduction: Top to Bottom, AGU
Monogr. Ser., 96, edited by G. E. Bebout, D. Scholl, S. Kirby, and J. P.
Platt, pp. 215 – 222, American Geophysical Union, Washington, D. C., G. A. Abers and T. Plank, Department of Earth Sciences, Boston
1996. University, Boston, MA 02215, USA.
Karson, J. A., Internal structure of oceanic lithosphere: a perspective from B. R. Hacker, Department of Geological Sciences, University of
tectonic windows, in Faulting and Magmatism at Mid-Ocean Ridges, California, Santa Barbara, CA 93106, USA.