You are on page 1of 50

3 Review

Geochemistry
G
Volume 4, Number 5
Geophysics 3 May 2003
8905, doi:10.1029/2002GC000392
Geosystems
AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES ISSN: 1525-2027
Published by AGU and the Geochemical Society

Subduction fluxes of water, carbon dioxide, chlorine, and


potassium
Richard D. Jarrard
Department of Geology and Geophysics, University of Utah, Salt Lake City, Utah 84112, USA
( jarrard@mines.utah.edu)

[1] The alteration of upper oceanic crust entails growth of hydrous minerals and loss of macroporosity,
with associated large-scale fluxes of H2O, CO2, Cl, and K2O between seawater and crust. This age-
dependent alteration can be quantified by combining a conceptual alteration model with observed age-
dependent changes in crustal geophysical properties at DSDP/ODP sites, permitting estimation of crustal
concentrations of H2O, CO2, Cl, and K2O, given crustal age. Surprisingly, low-temperature alteration
causes no net change in total water; pore water loss is nearly identical to bound water gain. Net change in
total crustal K2O is also smaller than expected; the obvious low-temperature enrichment is partly offset by
earlier high-temperature depletion, and most crustal K2O is primary rather than secondary. I calculate
crustal concentrations of H2O, CO2, Cl, and K2O for 41 modern subduction zones, thereby determining
their modern mass fluxes both for individual subduction zones and globally. This data set is complemented
by published flux determinations for subducting sediments at 26 of these subduction zones. Global mass
fluxes among oceans, oceanic crust, continental crust, and mantle are calculated for H2O, Cl, and K2O.
Except for the present major imbalance between sedimentation and sediment subduction, most fluxes
appear to be at or near steady state. I estimate that half to two thirds of subducted crustal water is later
refluxed at the prism toe; most of the remaining water escapes at subarc depths, triggering partial melting.
The flux of subducted volatiles, however, does not appear to correlate with either rate of arc magma
generation or magnitude of interplate earthquakes.

Components: 28,530 words, 11 figures, 2 tables, 1 Auxiliary material.


Keywords: Subduction; flux; carbon dioxide; potassium; oceanic crust; Ocean Drilling Program.
Index Terms: 1030 Geochemistry: Geochemical cycles (0330); 1020 Geochemistry: Composition of the crust.
Received 18 June 2002; Revised 31 December 2002; Accepted 10 January 2003; Published 3 May 2003.

Jarrard, R. D., Subduction fluxes of water, carbon dioxide, chlorine, and potassium, Geochem. Geophys. Geosyst., 4(5),
8905, doi:10.1029/2002GC000392, 2003.

————————————
Theme: Oceanic Inputs to the Subduction Factory Guest Editors: Terry Plank and John Ludden

1. Introduction elements. Oceanic crustal alteration, associated


with axial and ridge-flank hydrothermal circula-
[2] Long-term average production of oceanic tion, modifies crustal abundances of several ele-
crust at spreading centers is balanced by subduc- ments. For upper crust, the largest changes occur
tion. The same is not true, however, for individual for structural water, carbon dioxide, and potas-

Copyright 2003 by the American Geophysical Union 1 of 50


Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

sium [Staudigel et al., 1996], all of which are off-axis hydrothermal circulation may be largely
enriched by alteration. low-temperature (<30C) with seafloor communi-
cation or higher-temperature (30–60C) if sedi-
[3] Subduction recycling of H2O, CO2, Cl, and
mentation isolates the circulation from seawater
K2O affects the geochemical evolution of the
[Fehn and Cathles, 1986; Davis et al., 1992;
mantle, oceans, and atmosphere, impacting our
Humphris, 1995]. Hydrothermal circulation con-
understanding of a wide suite of global geochem-
tinues within ridge flanks for at least tens of
ical cycles [Peacock, 1990]. Volatile outgassing
millions of years (e.g., Anderson et al. [1977],
from MORB and arc magmas is probably the main
Fisher et al. [1990], and many others). Off-axis
source of the water and chlorine for early growth of
waning of hydrothermal circulation is caused by
the oceans [Rubey, 1951], but these fluxes are
decrease in crustal heat, accumulation of a rela-
currently nearly balanced by loss to subduction.
tively low-permeability sediment cover, and
For no-growth continents, long-term sediment sub-
decrease in crustal permeability caused by crustal
duction should approximately equal arc magma-
alteration [Anderson et al., 1977; Stein and Stein,
tism [Armstrong, 1968, 1991]. Slab dehydration
1994]. Of these three mechanisms, crustal alter-
generates overpressures and reflux at shallow
ation has received the most attention, perhaps
depths [Moore and Vrolijk, 1992] and partial melt-
because only it can account for the age-dependence
ing at 100 km responsible for arc magmatism
of upper crustal seismic velocities [Schreiber and
[Gill, 1981]. Subduction permanently removes
Fox, 1976, 1977]. My focus is on this crustal
CO 2 that was temporarily isolated from the
alteration.
ocean/atmosphere system by deposition of carbo-
nates and organic matter. Subduction recycling of [6] According to this conceptual model, alteration
volatiles and potassium are also likely to generate of oceanic crust begins as soon as crust is created.
mantle heterogeneity and may affect compositions The combination of axial heat and highly fractured,
of arc magmas [Plank and Langmuir, 1993; Stau- permeable basalt promotes vigorous hydrothermal
digel et al., 1996], though these effects are yet to circulation, which causes crustal alteration. Alter-
be explored comprehensively. ation minerals eventually fill macroporosity (cracks
and interpillow voids), thereby reducing porosity
[4] The subduction-related fluxes of this study
and permeability. Nevertheless, circulation on
contribute to the Geochemical Earth Reference
flanks is responsible for 70% of the advective
Model (GERM), an interdisciplinary, long-term
heat loss from oceanic crust [Stein and Stein, 1994]
initiative with the much wider aims of characteriz-
and substantial geochemical exchange with sea-
ing the Earth’s major chemical reservoirs and the
water [Mottl and Wheat, 1994; Elderfield and
fluxes between them [Staudigel et al., 1998a].
Schultz, 1996].
Subduction fluxes for all major elements have been
estimated for both upper crustal extrusives [Stau- [7] Low-temperature alteration patterns of upper
digel et al., 1996] and overlying sediments [Plank oceanic crust have been determined from petro-
and Langmuir, 1998]. This study incorporates geo- graphic studies, core descriptions, and geochemis-
physical models for age-dependent crustal altera- try; Honnorez [1981], Thompson [1991], Alt [1995,
tion into calculations of fluxes for all subduction 1999], and Bach et al. [2003] provide comprehen-
zones. sive reviews. Early formation of celadonite and
smectites (particularly saponite), often as inter-
2. Age Dependence of growths, is followed by precipitation of carbonates
Upper Crustal Properties (mainly calcite). These three minerals are the
dominant alteration minerals in upper oceanic
[5] The evolution of hydrothermal circulation is crust; others include Fe-oxyhydroxides and zeo-
well established qualitatively. At the ridge axis, lites. Early low-temperature alteration is oxidative,
magmatic heat drives vigorous and deep axial with high water-rock ratios and open seawater
circulation [Humphris, 1995; Fisher, 1998]. Early circulation. Reducing conditions generally follow,
2 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Figure 1. Locations of DSDP and ODP sites with significant penetrations into normal oceanic crust and with either
downhole logging or core physical properties. Also shown are locations of Holes 504B (Figure 3), Site 801 (Figure 4),
and gabbro Hole 735B.

associated with decreases in both flow rates and These analyses of crustal alteration patterns
exchange between seawater and pores. exclude data from ophiolites, for two reasons:
(1) most are formed in supra-subduction settings
[8] Basalt coring and logging by DSDP and ODP
that are nonrepresentative of normal, open-ocean
provide abundant information on physical proper-
crust, and (2) ophiolitic alteration is generally
ties of upper oceanic crust, despite generally low
more intense, with much higher total fluid flux,
basement core recoveries and rare basement log-
than normal crustal alteration [Alt and Teagle,
ging [Goldberg, 1997]. Most of the analyses of
2000]. Sedimented ridges and back-arc spreading
this section draw heavily from a synthesis of
are also excluded, because their hydrothermal
DSDP/ODP petrophysical data on aging of upper
circulation histories probably differ from that of
oceanic crust by Jarrard et al. [2003]. Of 1200
open-ocean crust.
sites drilled by DSDP and ODP, 13 fulfill the
following two criteria: (1) basement composed of [9] Basalt physical properties at intergranular core-
normal oceanic crust formed in an open-ocean plug scale may differ from those at the meter scale
environment, and (2) velocity logging of at least measured by logs, because of differences in type of
40 m of basement (Figure 1). Jarrard et al. porosity and in alteration history. At plug scale,
[2003] confined their analyses to the top 300 m porosity consists mainly of vesicles and micro-
of basement, because most sites have only 90– cracks, whereas log-scale porosity also includes
200 m of velocity log and only four holes (395A, fractures and interpillow voids. At plug scale,
418A, 504B, and 801C) penetrated more than alteration converts glass, olivine, and plagioclase
300 m of basement. A second search of the DSDP mainly to clay minerals, thereby decreasing matrix
and ODP databases and publications (Initial density and matrix velocity. Porosity may increase
Reports and Scientific Results) focused on index due to this hydration, or decrease due to precip-
properties of basalts. Only sites on normal oceanic itation. At log scale and larger, porosity is thought
crust were considered, and data from multiple to decrease due to filling of cracks and interpillow
holes at the same site were combined. Useful voids by alteration minerals. Consequently, a com-
index data are available for 25 sites (Figure 1). parison of plug-scale and log-scale basalt physical
3 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

properties may yield insights concerning the scales ity is well correlated both with hydration intensity
and varieties of alteration. determined by light-absorption spectroscopy
(LAS) (significant at 99% confidence level) and
2.1. Permeability, Porosity, and Alteration with matrix density (significant at 99% c.l.).
Because hydration substantially reduces matrix
[10] Permeability is the architect of the crust’s
density, matrix density is correlated with LAS-
hydrothermal systems. Nearly all measurements
based hydration (significant at 99% c.l.) [Jarrard
of the permeability of upper oceanic crust are
et al., 2003].
averages for portions of a borehole tens to hun-
dreds of meters in extent. Fisher [1998] provided
an excellent review and synthesis of these bulk 2.2. Timing of Waning of Hydrothermal
permeability measurements and their implications. Circulation: Previous Evidence
This large-scale crustal permeability is a critical
[12] Geophysical techniques disagree on the tem-
control on hydrothermal circulation patterns and
poral evolution of hydrothermal circulation and
associated heat flux, because the volumetrically
resulting alteration in oceanic crust. Cementation
dominant flux is channelized within the most
associated with ridge-flank hydrothermal circula-
permeable zones, particularly large open fractures
tion appears to terminate as early as 5–10 Ma,
[Fisher, 1998; Fisher and Becker, 2000] and brec-
based on global data syntheses of upper crustal
cias [Bach et al., 2003]. Bulk permeabilities for the
seismic velocities [Grevemeyer and Weigel, 1996;
top 1200 m of oceanic crust suggest two main
Carlson, 1998]. Similarly, crustal permeabilities
layers: the top 500 m has permeabilities of
decrease exponentially from 1 Ma to 8 Ma; sub-
1014 –1013 m2, and the next 700 m has per-
sequent changes (if any) are undetermined [Fisher
meabilities of 1017 m2 [Fisher, 1998]. The
and Becker, 2000].
intense intergranular alteration of pillows and flow
margins, despite intergranular permeabilities that [13] In contrast, the global heat flow synthesis of
are more than four orders of magnitude lower than Stein and Stein [1994] indicated that the heat flow
the fracture and interpillow permeabilities meas- deficit (theoretical minus observed) gradually
ured by packers, occurs via both microbial activity decreases with age, disappearing at about 65 Ma.
[e.g., Furnes and Staudigel, 1999; Furnes et al., The deficit indicates advection via open circulation
2001; Staudigel et al., 1998b] and diffusion. The between seawater and the crust; beyond 65 Ma,
lower limit for sufficient diffusion to produce heat flow is dominantly conductive, with an age-
significant alteration may be 1019 m2, based dependent decrease in heat flow variance that
on the observation that lower intergranular perme- suggests continued fluid flow [Stein and Stein,
abilities are encountered mostly in fresh, massive 1994]. Von Herzen [2003] reanalyzed 58 detailed
basalts [Karato, 1983a, 1983b; Johnson, 1979a; heat flow surveys from old crust and concluded
Hamano, 1979; Jarrard et al., 2003]. that ongoing hydrothermal circulation is common
in Mesozoic crust; it is likely in half of the surveys
[11] ODP Hole 801C, which obtained the world’s
on crust 65–95 Ma, and in a third of those on >95-
oldest section of in situ, normal oceanic crust,
Ma crust. Heat flow data cannot unambiguously
provides the opportunity to examine relationships
determine whether hydrothermal circulation in old
among hydrologic properties (porosity, permeabil-
crust is closed-cell (no interchange with seawater)
ity, fluid flow), crustal alteration, and geophysical
or slow open-cell [Fisher, 1998], yet the two differ
properties, at both core-plug and downhole-log
crucially in geochemical mass balance. For closed-
scales [Busch et al., 1992; Jarrard et al., 1995,
cell fluid flow, precipitation equals solution, and
2003]. Within these upper crustal basalts, higher
crustal geophysical properties may be unchanged
porosity is responsible for higher permeability and
by ongoing alteration.
therefore higher fluid-flow rates. High fluid flux, in
turn, fosters alteration, particularly the hydration [14] These conflicting indications of when off-axis
reactions that generate clays. Consequently, poros- hydrothermal circulation wanes have been partly
4 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

reconciled. Fisher and Becker [2000] suggested


that the rapid initial increase in velocity and
decrease in permeability are not necessarily incom-
patible with heat flow evidence for persistence of
open-cell convection. They hypothesized that
hydrothermal circulation continues to 65 Ma, in
localized channels within the crust that are too
small a percentage of total upper crust for signifi-
cant effects on average crustal velocity. Crustal
alteration beyond 5–10 Ma is too subtle for most
geophysical methods to resolve, but the next three
sections examine continuing geophysical changes
that are detectable.

2.3. Core Alteration and


Matrix Density Versus Age
[15] Johnson and Semyan [1994] examined the
possibility of age-dependent changes in velocity,
porosity, density, oxidation state, and hydration,
based on a compilation of core-plug measurements
from DSDP and ODP basalts. The increase of
crustal alteration with increasing age is most evi-
dent among parameters sensitive to hydration.
Age-dependent oxidation is clearest for 0.1–5 Ma
and is not resolved beyond 10–30 Ma [Johnson
and Semyan, 1994; Zhou et al., 2001b]. Analyses
of bound water, or H2O+, provide a semiquantita-
tive measure of the extent of alteration-induced
hydration [Alt et al., 1992]. The H2O+ compilation
of Johnson and Semyan [1994] confirmed earlier
indications [Hart, 1970, 1973; Donnelly et al.,
1979a; Muehlenbachs, 1979] that hydration
increases with increasing age (R = 0.35; significant
Figure 2. Age-dependent geophysical properties of
at 95% confidence level), but dispersion is too high upper oceanic crust. Each point represents one DSDP or
to identify when most of this hydration occurs. ODP site. Regression lines are used to predict crustal
properties at subduction zones, based on crustal age. A–
[16] Matrix density is the average density of the C are modified from Jarrard et al. [2003].
minerals forming the solid part of the rock, includ-
ing any alteration minerals. Matrix densities provide
the strongest demonstration of systematic increase density continues to decrease, at a rate proportional
in alteration versus time (Figure 2b) (R = 0.70; to logarithm of age, throughout the period 0–167
significant at 99.9% confidence level), indicating Ma. However, data dispersion in the time period
that approximately half of all intergranular-scale 10–70 Ma is high, and the hypothesis of no change
crustal alteration occurs after the first 10–15 Ma. beyond about 40–60 Ma cannot be rejected.
Site 597, where sampling concentrated on a massive
flow [Shipboard Scientific Party, 1985] and matrix 2.4. Velocity Versus Age
density is anomalously high, is excluded. The [17] Compilations of published seismic experi-
simplest interpretation of this pattern is that matrix ments from throughout the world ocean demon-
5 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

strated that seismic velocities of the upper oceanic [19] Before comparing atmospheric-pressure core-
crust increase rapidly for the first 5–8 Ma, from an plug velocities to in situ log velocities, pressure
average of 2.3 km/s at the spreading center to 4.3– effects must be considered. Rebound, the core
4.4 km/s at 5–10 Ma (Figure 2a) [Grevemeyer and expansion that accompanies change from in situ
Weigel, 1996; Carlson, 1998]. The amount of to laboratory pressure, can reduce plug velocities
associated crustal alteration is undetermined. [e.g., Nur, 1971; Bourbié et al., 1987]. Rebound is
Velocity modeling suggests that the large near- generally minor in oceanic extrusive basalts,
ridge velocity increases may be accomplished by because burial depths are much too small to induce
changing the shape of pore spaces via secondary microcracks. The extensive suite of core velocity
mineralization with relatively small overall poros- measurements for Hole 504B [Wilkens et al., 1983;
ity reduction [Wilkens et al., 1991; Shaw, 1994; Christensen and Salisbury, 1985; Christensen et
Moos and Marion, 1994]. Beyond 10 Ma, no al., 1989; Iturrino et al., 1995; Salisbury et al.,
significant change in upper crustal velocity is 1996] shows that the average pressure-induced
resolvable with seismic velocity data [Carlson, percentage difference between atmospheric and in
1998]; data dispersion is too high to resolve situ velocities for the upper 600 m of basement is
systematic changes of <1 km/s. DSDP and ODP only 0.3% based on differential pressures,
velocity logs, with an age range of 6–167 Ma that 0.6% based on lithostatic pressures, and 2.0%
is beyond any seismically resolvable age depend- based on confining pressures. In contrast, the two
ence, indicate that crustal velocity increase con- diverge at greater depths: at about 1100 m subbase-
tinues (Figure 2a, significant at 95% c.l.), rising ment, 300 m into the dikes, the unmistakable
from 4.2–4.7 km/s at 6 Ma to 5.0–5.1 km/s for signature of microcracks appears and atmospheric-
>100 Ma. This gradual increase in large-scale pressure measurements are no longer a useful
velocity occurs despite systematic velocity indicator of in situ velocities (Figure 3). Conse-
decrease at the intergranular scale, as indicated quently, I assume that atmospheric-pressure veloc-
by core-plug velocities for 0–100 Ma [Christensen ities can be used for macroporosity calculations
and Salisbury, 1973; Johnson and Semyan, 1994; within this and other extrusive sections, whereas
Jarrard et al., 2003]. Precipitation in cracks and high-pressure measurements would be required for
interpillow voids must be dominant over intergra- dikes and gabbros.
nular-scale alteration in controlling the evolution
of large-scale crustal velocity. [20] Figure 2c compares core-plug and log veloc-
ities for the 13 sites with significant basement
penetration. Because a systematic difference
2.5. Macroporosity Versus Age between the two is thought to result primarily from
[18] Plugs reveal intergranular patterns, but they the influence of macroporosity on logs but not on
tend to miss cracks and crack-filling that poten- cores, I used the Wyllie et al. [1956] time-average
tially are major controls on large-scale velocity equation to express the difference between each
[Hyndman and Drury, 1976; Anderson and core/log velocity pair as an apparent macroporos-
Zoback, 1982]. Evaluation of crack porosity based ity. Matrix velocity may also decrease with age
on incomplete core recovery is difficult [Johnson, [Jarrard et al., 2003] but no macroporosity bias
1979b]. If much of basalt porosity is in the form of results, because both core-plug and log velocities
large-scale voids and fractures that are not are similarly affected. Total macroporosities exhibit
sampled by 10-cm3 core plugs, then comparison a decrease with increasing age (R = 0.87) that is
of core and log measurements of porosity has the significant at the 99.9% confidence level, and this
potential of detecting this macroporosity. I use decrease appears to continue throughout the avail-
core and log velocity for macroporosity determi- able age range (Figure 2c). Initial macroporosities
nation, because velocity logs provide the most of 8–10% eventually are reduced to near zero. This
robust, unbiased measure of in situ porosities of systematic macroporosity decrease contrasts with
oceanic basalts. the absence of significant age-dependent porosity
6 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Figure 3. Downhole K2O variations, seismic layers, lithologic units, and P wave velocity rebound at Hole 504B.
Sources in text. K2O variations are shown with red dots for core XRF analyses, black lines for well log, and vertical
green line for assumed pre-alteration K2O concentration. Velocity rebound is shown as percentage ratio of velocity at
atmospheric pressure compared to that at in situ (lithostatic) pressure.

change among core plugs [Johnson and Semyan, This carbon sink has been demonstrated to be an
1994; Jarrard et al., 2003]. In contrast to the important term in global carbon budgets [Staudi-
precipitation in cracks and interpillow voids that gel et al., 1989; Peacock, 1990; Alt and Teagle,
causes macroporosity decrease, intergranular alter- 1999].
ation can either fill small cracks and vesicles or
generate porosity by breakdown of individual min- [22] Progressive CO2 enrichment of oceanic crust
erals such as olivine. can be quantified by analysis of DSDP and ODP
sites, but avoidance of sampling bias is non-trivial.
2.6. CO2 Versus Age Most whole-rock CO2 analyses target the least
weathered samples and avoid carbonate-filled
[21] Ridge volcanism releases CO2 at axial hydro-
cracks. Core recovery is also biased, losing some
thermal vents, mainly via degassing at magma
of the larger-scale crack filling. Two studies of
chambers rather than via crustal alteration [Ger-
crustal CO2 have succeeded in minimizing these
lach, 1989]. Residual CO2 concentrations within
biases: Staudigel et al. [1996] and Alt and Teagle
the basalts are minor, only 0.045 wt.% [Gerlach,
[1999].
1989]. Subsequent low-temperature alteration adds
CO2 to the upper crust, primarily in the form of [23] Adjacent Holes 417A, 417D, and 418A in the
CaCO3, using CO2 from seawater and CaO mostly west-central Atlantic provide an unusually good
from the basalts [Staudigel et al., 1989, 1996]. record of the alteration of old oceanic crust, due to
7 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

penetration of almost the entire extrusive section, an increase in CO2 between 6 Ma and 110 Ma but
relatively high core recovery (70%), and sam- not the specific timing of this increase. Given the
pling of both hydrothermal downflow and upflow evidence of continuing upper-crustal alteration
zones. Staudigel et al. [1996] took several steps to based on velocity (Figure 2a), macroporosity (Fig-
obtain representative samples of downhole geo- ure 2c), matrix density (Figure 2b), and variance of
chemical variations within these holes: they global heat flow [Stein and Stein, 1994], however,
sampled a range of alteration states for all extrusive the simplest explanation of age-dependent CO2 is
styles (flows, pillows, and volcaniclastics), includ- that CO2 enrichment is ongoing but waning, with a
ing vein materials, and then they blended the logarithm of time dependence (Figure 2d; R =
results using best estimates of the in situ propor- 0.88).
tions of alteration states and extrusive styles. These
[26] Cretaceous atmospheric CO2 levels were an
composite CO2 concentrations provide an excellent
order of magnitude higher than Neogene levels
demonstration of the difference between average
[Ekart et al., 1999; Pearson and Palmer, 2000;
geochemical analyses and representative samples;
Berner and Kothavala, 2001], possibly explaining
the former are almost all 0–2 wt.% CO2, whereas
the pattern of Figure 2d without requiring contin-
the latter exhibit a systematic decrease downhole:
ued off-axis carbonate precipitation. Within-site
5.6% for 0–100 m subbasement, 5.0% for 130–
correlations of highest whole-rock and vein CO2
290 m, and 1.0% for 455–540 m. The resulting
with highest present-day porosity and presumably
weighted average for the upper extrusive section
permeability are consistent with – but cannot
(0–300 m subbasement) is 5.2%, similar to an
prove – ongoing, progressive fluid flow and asso-
earlier 5.0% value by Staudigel et al. [1989], and
ciated carbonate enrichment. Petrographic studies
the overall average CO2 concentration of the
indicate that carbonate precipitation generally post-
extrusives at these three holes is 2.95% [Staudigel
dates smectite alteration [Alt and Teagle, 1999].
et al., 1996].
This observation is consistent with the regressions
[24] Alt and Teagle [1999] combined best estimates of Figure 2, which suggest that half of all matrix
of bulk-rock and vein CO2 to obtain representative density lowering (smectite/celadonite alteration)
samples of the average CO2 contents of upper occurs after the first 15 m.y., whereas half of all
oceanic crust at Sites 504, 896, 843, and 801. CO2 addition occurs after the first 25–30 m.y. The
Bulk-rock CO2 measurements were averaged for calculated modern CO2 subduction fluxes of this
different alteration types, then these results were study depend only on the pattern of Figure 2D, not
weighted based on observed proportions of the its mechanism, whereas modern precipitation flux
alteration types, producing an average bulk-rock is sensitive to both.
CO2 for each crustal section. The amount of vein
carbonate was calculated from detailed core 2.7. Potassium Versus Age
descriptions of carbonate vein abundances and
[27] Low-temperature crustal alteration processes
thicknesses plus breccia carbonate abundance.
include oxidation, hydration, and alkali fixation.
These results demonstrated that carbonate precip-
The latter two generate clay minerals, the most
itation in veins continues beyond the 6 Ma age of
abundant alteration products [e.g., Donnelly et al.,
Sites 504 and 896, resulting in enrichment of total
1979a; Andrews, 1980; Gillis and Robinson, 1988;
CO2 content of the upper extrusive section.
Alt and Honnorez, 1984; Alt et al., 1986]. Saponite
[25] Figure 2d combines the upper crustal CO2 (a smectite clay) and celadonite (a mica) are
determinations of Alt and Teagle [1999] with that commonly formed during low-temperature diagen-
of Staudigel et al. [1996]. The pattern of progres- esis [e.g., Donnelly et al., 1979b; Alt and Hon-
sive CO2 enrichment extends far beyond 10 Ma, norez, 1984], usually incorporating potassium
once hypothesized to be the conclusion of calcite extracted from seawater [Hart, 1969; Andrews,
precipitation [Staudigel et al., 1981]. Alt and 1980]; an exception is saponite formed by alter-
Teagle [1999] emphasized that these data establish ation of interstitial glass [Zhou et al., 2001a].
8 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

[29] Only seven of the thirteen logged basement


sites in normal oceanic crust have potassium logs.
These logs show no consistent pattern of potassium
increase or decrease with depth. Among these
seven sites, average log-based potassium content
is strongly correlated with average core-based
matrix density (R = 0.86; significant at 99%
confidence level) (Figure 5). This correlation, like
the Hole 801C correlation between potassium and
velocity logs (Figure 4), demonstrates that alter-
ation entails potassium enrichment. Also shown on
Figure 4. Comparison of velocity and potassium logs Figure 5 are results for Site 768, which may be the
for Hole 801C tholeiites. The strong inverse correlation most altered thick sequence of oceanic crust yet
between these logs indicates that precipitation of sampled by DSDP or ODP [Shipboard Scientific
potassium-rich clays and celadonite occurs mainly in Party, 1990a]. This crust was formed by back-arc
the most porous (lowest-velocity) zones. Modified from
Jarrard et al. [2003]. spreading and is consequently excluded from these
analyses of normal crustal alteration, but it illus-
trates extreme alteration and associated potassium
enrichment.
Consequently, altered basalts are generally higher
in potassium than unaltered basalts [Hart, 1969; [30] Despite the excellent correlation of Figure 5,
Hart et al., 1974]. Deep penetration of old oceanic the temporal evolution of potasium enrichment is
crust at Holes 417A/417D/418A provided the first an outstanding problem. Potassium-rich celadonite
compelling evidence of widespread alteration- is generally one of the earliest-formed alteration
induced K2O enrichment in the upper crust [Don- minerals [Alt, 1999], and the few sites with potas-
nelly et al., 1979a, 1979b].
[28] Potassium enrichment is greatest in the most
porous, permeable zones. Figure 4, which overlays
K2O and velocity logs versus depth for the Hole
801C tholeiites, demonstrates that potassium var-
iations are closely linked to porosity changes
(velocity is used here as a porosity indicator).
The spectral gamma-ray log of potassium is here
converted to K2O weight percent by multiplying by
1.2. The lowest observed K2O contents, 0.07–
0.12%, are comparable to values typical of unal-
tered MORB tholeiites, suggesting that the most
massive, low-porosity units are minimally altered.
K2O maxima of 1% result from a high proportion
of potassium-bearing alteration minerals and occa-
sional sediment interlayers. This strong inverse
Figure 5. Black dots: average core-based matrix
correlation between velocity and potassium-bear- density versus average log-based K2O concentration,
ing alteration minerals is not inconsistent with the for upper oceanic crust at all DSDP/ODP sites with both
earlier observation that log velocity increases with data types; regression line is also shown. Dashed lines
time and crustal alteration (Figure 2a): the highest- and associated labels identify 3-component (fresh basalt,
saponite, celadonite) mixing models. For prediction of
porosity zones experience the most alteration and K2O and structural water at subduction zones, the best
associated porosity reduction, yet they remain fit alteration mineral assemblage of 80% saponite and
more porous than fresher, more massive intervals. 20% celadonite is assumed.
9 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

sium logs show only a fair correlation (R = 0.47) of [Ransom and Helgeson, 1995]. Furthermore, sam-
potassium with logarithm of crustal age [Jarrard et ple drying, required for either matrix-density meas-
al., 2003]. urement or chemical analysis, drives off not only
pore water but also some of the loosely held
2.8. Alteration Mineralogy Model structural interlayer water. Consequently, measured
matrix densities are biased upward [Brown and
[31] Geochemical changes associated with the pro-
Ransom, 1996] and H2O+ analyses are too low.
gressive alteration of oceanic crust can be modeled
Measured structural-water content of saponite aver-
as an age-dependent mix of fresh basalt and alter-
ages 12 wt.%, based on 35 analyses by Honnorez
ation minerals. I assume that the dominant alter-
et al. [1983], but I assume that the true value is
ation minerals in the upper extrusives are saponite
20 wt.%. Brown and Ransom [1996] estimated that
and celadonite, though this is only a first-order
the true water content of largely authigenic smec-
approximation. Given the matrix densities and K2O
tites offshore Barbados is 20 wt.%, and Schlum-
contents of smectite, celadonite, and fresh basalt,
berger used neutron and density log analysis to
one can compute the proportions of these three
propose that montmorillonites have a water content
components from the relationship between matrix
of 41% by volume [Serra, 1986], or 20 wt.%.
density and K2O (Figure 5).
Applying this estimate to basalt saponites, how-
[32] The average matrix density of fresh basalt is ever, could be inaccurate by as much as 5 wt.%.
3.01 g/cc, based on applying the regression of The assumed bound-water content of 5% for cela-
matrix density on age (Figure 2b) to ages <1 Ma. donite is based on only one measurement by
The local value depends on concentration of dense Buckley et al. [1978], because most analyzed
phenocrysts such as olivine and pyroxene. I oceanic crustal celadonites are a mix of celadonite
assume a K2O content of 0.085 weight percent, and smectite [Andrews, 1980] (e.g., water contents
based on analyses of fresh basalts from Holes of 9.5–11% for mixed celadonite/smectite analyses
417A/417D/418A [Spivack and Staudigel, 1994; of Buckley et al. [1978]).
Staudigel et al., 1996] and 801C [Shipboard Sci-
[35] For calculation of alteration mineral percen-
entific Party, 2000a]. Average unaltered N-MORB
tages using Figure 5, it is appropriate to use
[Hart et al., 1999] and fresh glasses from dredged
measured rather than true saponite matrix density,
basalts [Jochum et al., 1983] are slightly higher:
based on measurements with similar sample drying
0.106% and 0.12%, respectively. The freshest
procedure to those applied to basalts: 24 hours
basalts from Holes 504B and 896A have an aver-
heating at 110C. Smectite sediments from Site
age K2O of 0.04% [Shipboard Scientific Party,
1222 have an average measured matrix density of
1992], much lower than typical MORB. This
2.51 g/cc (std. dev. 0.05, N = 17) [Shipboard
depletion may slightly affect the lowest two points
Scientific Party, 2002]. I assume that smectites
on Figure 5, but the overall influence of magmatic
within basalts are similar in water content to these
K2O variations is clearly minor in comparison to
and have an apparent matrix density of 2.50 g/cc.
alteration-induced variations.
[36] The dashed lines on Figure 5 show expected
[33] The matrix density of celadonite is assumed to
positions on a K2O versus matrix density plot of
be 2.56 g/cc, the same as that of glauconite [Serra,
mixtures of fresh basalt, saponite, and celadonite.
1986], based on compositional similarity of the two
Observed data points, including Site 768, lie along
minerals [Buckley et al., 1978]. Celadonite has an
a line for a mixture of fresh basalt and an alter-
average K2O content of about 9.07%, based on 13
ation-mineral assemblage of 80% saponite and
analyses by Buckley et al. [1978]. Saponite has an
20% celadonite; the data regression line lies vir-
average K2O content of about 0.21%, based on 35
tually on top of that mixture line. Consequently, I
analyses of Honnorez et al. [1983].
assume that the alteration minerals in extrusive
[34] The structural interlayer water of smectites can basalts are – to a first approximation 80%
vary from 10% to 25% by weight or higher saponite and 20% celadonite. Other alteration min-
10 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

erals are present, but I assume that they either are subduction zones are those tabulated and analyzed
similar in properties to these minerals or have a by Jarrard [1986] and von Huene and Scholl
minor effect on total K2O and matrix density. For [1991], though some names have been updated.
example, smectite is dominant in both early oxi- The total length of active subduction zones in this
dizing and later reducing alteration, but early study is 44,450 km, close to the 43,250 km total of
celadonite can be followed by zeolites with similar von Huene and Scholl [1991]. Total ridge length is
K2O and matrix density. This model composition 49,000 km [Kominz, 1984]. The total length of
for alteration minerals, along with their K2O con- convergent plate boundaries, including 10% con-
tents and matrix densities, permits estimation of the tinent-continent collisions, is also 49,000 km, so
K2O content of an extrusive sequence, based on the global average convergence and divergence rates
age-dependent function for matrix density (Figure are nearly equal.
2b). The direct regression of K2O on matrix density [39] Convergence rates for the major plates are
(Figure 5) yields an almost identical result for based on the NUVEL-1A global motion model
predicted K2O and total global K2O flux. I use [DeMets et al., 1990, 1994], supplemented by
the mineralogic model instead because of its geo- GPS data [Bevis et al., 1995; Tregoning et al.,
logic grounding and because it predicts both K2O 1998] and the Philippine plate motion of Seno et
and structural water from matrix density (Table 1). al. [1993]; some Philippine plate GPS data are too
[37] Both matrix density functions – versus age recent to be included [Miyazaki and Heki, 2001;
and versus K2O – are based on core-plug matrix Kato and Kotake, 2002]. Only the component of
density rather than log matrix density. However, convergence perpendicular to each trench is consid-
large-scale matrix density may differ from inter- ered. When applicable, back-arc spreading rates are
granular matrix density because the amount or kind added to the major-plate motions, using back-arc
of fracture-fill minerals differs from intergranular rates summarized by Jarrard [1986]. A few con-
alteration minerals. This difference does not bias vergence rates for short trenches in Southeast Asia
the K2O value, which is log-based and therefore (Manila, Cotabato, Negros, Sulu, and Trobriand) are
representative, but it does mean that computation poorly known but slow, and crustal age at San
of structural water is accurate only for intergranular Cristobal is ill-determined; subsequent calculations
alteration minerals. When macroporosity of old of global fluxes are insensitive to these local uncer-
basalts has been filled by alteration minerals, tainties. For most of the longest subduction zones,
large-scale matrix density is modified slightly from McCaffrey [1997] has computed perpendicular-con-
intergranular matrix density. Fortunately, the per- vergence rates and average subducting-plate ages
centage of filled macroporosity is so small that its pointwise along each trench, obtaining a more
influence on total matrix density and structural- accurate average for each than is obtainable by
water estimates is minor; nevertheless, I employ a subjectively selecting a representative crustal age,
correction for this effect (Table 1). trench azimuth, and convergence rate for an entire
subduction zone. In general, I use McCaffrey’s
[1997] averages when available (Table 2), but I
3. Data Set: Global Subduction have not undertaken a similar pointwise approach
3.1. Subduction Zones for the other subduction zones.

[38] Nearly all of the world’s currently active [40] Total consumption of oceanic crust on the
subduction zones are included in this analysis subduction zones of Table 2 is 2.42 km2/yr, much
(Table 2), regardless of whether the subducting less than the 3 km2/yr (3.3 km2/yr including back-
oceanic crust formed in an open-ocean or back- arc spreading) of crustal generation calculated from
arc environment. Continent/continent collisions early global motion solutions [Minster and Jordan,
(e.g., India/Asia) and portions of a subduction zone 1978; Parsons, 1981] and usually used for estima-
undergoing continental collision (e.g., Aegean, E. tion of subduction fluxes [Ito et al., 1983; Peacock,
Sunda) are excluded. With few exceptions, the 1990; Bebout, 1995; Staudigel et al., 1996]. Part of
11 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Table 1. Parameters for Flux Calculations


Variable Units Symbol Zone Value

Trench length km L meas.


Converg. rate, perp. mm/yr V? meas.
Crustal age Ma A meas.
Fluid density g/cm3 rf 1.02
Zone thickness m Ts sed. meas.
Tu upper ext. 300
Tl lower ext. 300
Td dikes 1400
Tg gabbros 5000
Microporosity % fm,u upper ext. 7.8
fm,l lower ext. 5.1
fm,d dikes 2.2
fm,g gabbros 0.7
Macroporosity % fm,u upper ext. fm,u = 13.01  5.625*log(A)
fm,l lower ext. fm,l = fm,u/2
fm,d dikes 0.84
fm,g gabbros 0
Total porosity % ftot,u upper ext. ftot,u = fm,u + fm,u
ftot,l lower ext. ftot,l = fm,l + fm,l
ftot,d dikes ftot,d = fm,d + fm,d
ftot,g gabbros ftot,g = fm,g + fm,g
Matrix density g/cm3 rma,u upper ext. rma,u = 3.01  0.0631*log(A)
rma,l lower ext. rma,l = 3.01  0.5*0.0631*log(A)
rma,d dikes 2.98
rma,g gabbros 2.99
Bulk density g/cm3 rb,s sed. meas.
rb,u upper ext. rb,u = (0.01*ftot,u)*rf + (1  0.01*ftot,u)*rma,u
rb,l lower ext. rb,l = (0.01*ftot,l)*rf + (1  0.01*ftot,l)*rma,l
rb,d dikes rb,d = (0.01*ftot,d)*rf + (1  0.01*ftot,d)*rma,d
rb,g gabbros rb,g = (0.01*ftot,g)*rf + (1  0.01*ftot,g)*rma,g
K2O wt. % K2Os sed. meas.
K2Ou upper ext. K2Ou = 11.58  3.823*rma,u
K2Ol lower ext. K2Ol = 11.58  3.823*rma,l
K2Od dikes 0.014
K2Og gabbros 0.06
CO2 wt. % CO2s sed. meas.
CO2u upper ext. CO2u = 1.55 + 2.493*log(A)
CO2l lower ext. CO2l = CO2u/2
CO2d dikes 0.14
CO2g gabbros 0.02
Structural H2O wt. % H2OSs sed. meas.
H2OSu upper ext. H2OSu = (103.1  34.27*rma,u) + (0.17*(13.01  fm,u))
H2OSl lower ext. H2OSl = (103.1  34.27*rma,l) + (0.17*(13.01  fm,l))
H2OSd dikes 1.76
H2OSg gabbros 0.79
pore H2O wt. % H2OPs sed. meas.
H2OPu upper ext. H2OPu = ftot,u*rf/rb,u
H2OPl lower ext. H2OPl = ftot,l*rf/rb,l
H2OPd dikes H2OPd = ftot,d*rf/rb,d
H2OPg gabbros H2OPg = ftot,g*rf/rb,g
chloride wt. % Cl each Cl = 0.019*H2OP
K2O flux g/yr FK2O each FK2O = K2O*(1  0.01*H2OP)*T*rb*L*V?*104
CO2 flux g/yr FCO2 each FCO2 = CO2*(1  0.01*H2OP) *T*rb*L*V?*104
structural H2O flux g/yr FH2OS each FH2OS = H2OS*(1  0.01*H2OP)*T*rb*L*V?*104
pore H2O flux g/yr FH2OP each FH2OP = H2OP*T*rb*L*V?*104
chloride flux g/yr FCl each FCl = 0.019*FH2OP

12 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

this difference is due to substantial reductions (10– trench-fill sediments and subduction erosion. They
25% in the Pacific) of many seafloor spreading estimated that the global volume of subduction
rates by NUVEL-1, 2% of which is attributable to erosion is comparable to that of deeply subducted
time-scale revisions [DeMets et al., 1990], with a incoming sediment, and that the combined total is
further 4.4% time-scale-induced reduction by 1.3–1.8 km3/yr.
NUVEL-1A [DeMets et al., 1994]. I am not aware
of a recalculation of global crustal generation based [43] Rea and Ruff [1996] used a quite different set
on NUVEL-1 or NUVEL-1A. In addition, total of assumptions to assess global sediment subduc-
consumption of oceanic crust is less than total tion. Rather than consider trench-fill, their sedi-
generation because of continent-continent conver- ment thicknesses were based on DSDP and ODP
gence, such as the 0.2 km2/yr convergence reference sites seaward of trenches. This approach
between India and Asia. allowed them to calculate the amount of incoming
sediment for three lithologies (terrigenous, calcite,
and opal) and to accurately separate mass fluxes of
3.2. Sediment Subduction
minerals and pore water. Uneven distribution of
[41] Some previous analyses of global subduction reference sites necessitated grouping of many
fluxes have excluded sediment subduction [Stau- adjacent subduction zones into ‘‘trench systems’’.
digel et al., 1996; Alt and Teagle, 1999], and some Accretion was assumed to be zero for the incom-
have included it by assuming a ‘‘typical’’ package ing sedimentary blanket sampled at the reference
of incoming sediments [Ito et al., 1983; Peacock, sites and 100% for the unconsidered trench fill.
1990; Bebout, 1995]. The partitioning of incoming Rea and Ruff [1996] estimated that the global
sediments into accreted and subducted components sedimentary flux of solids is 1.43  1015 g/yr
has presented a challenge to quantitative analyses. and of pore fluids is 9.1  1014 g/yr. This fluid
Hilde [1983] emphasized the role of flexural gra- flux is virtually identical to that of von Huene and
bens as sediment traps fostering sediment subduc- Scholl [1991], whereas the solid flux is only 35%
tion, and he documented the circum-Pacific of theirs. The analysis of Rea and Ruff [1996] did
occurrences of these grabens. More recent multi- not consider fluxes of structural water, K2O, or
channel seismic profiling across accretionary mar- CO2, but their calcite flux of 2.2  1014 g/yr
gins has permitted identification of the décollement suggests a sediment-hosted CO2 flux of 1.0 
separating accreting from underthrusting sedi- 1014 g/yr.
ments. Three analyses dealt with the extreme
regional heterogeneity of sediment subduction by [44] The most comprehensive analysis of sediment
undertaking global syntheses. subduction is that of Plank and Langmuir [1998],
who calculated subduction fluxes for major and
[42] Von Huene and Scholl [1991] compiled trench trace elements at 26 subduction zones. Like Rea
length, pre-NUVEL-1 orthogonal convergence and Ruff [1996], they computed sedimentary com-
rate, and trench sediment thickness for virtually ponents based on analysis of DSDP and ODP
all modern subduction zones. They used these data, reference sites on the incoming plate, but their
along with reference porosity/depth curves and more detailed lithologic analysis was calibrated
identification of accretionary versus nonaccretion- with geochemical measurements, permitting calcu-
ary margins, to calculate total fluxes of both sub- lation of bulk compositions and therefore elemental
ducting sediment (1.5 km3/yr, or 4.0  1015 g/yr) fluxes. By comparing sediment-hosted elemental
and pore water (0.9 km3/yr or 9  1014 g/yr). fluxes for individual subduction zones to the com-
Incoming sediment was partitioned between sub- position of arc volcanics, they were able to eval-
duction and accretion by assuming 100% subduc- uate the effects of sediment subduction on arc
tion for subduction zones lacking prisms, 80% for volcanism.
those with small to medium-sized prisms, and 70%
for those with large prisms. von Huene and Scholl’s [45] The analysis of Plank and Langmuir [1998]
[1991] analysis is the only one that has evaluated quantitatively assessed the sediment-hosted sub-
13 of 50
Table 2. Subduction Zones Analyzed and Their Calculated Fluxesa
Geophysics
Geosystems
Geochemistry

Subducting Sediment Fluxes Igneous Crust Fluxes


G

Length, Conv. Rate, Age, Mw Chloride, Chloride,


3

K2O, CO2, H2O+, Pore H2O, K2O, CO2, H2O+, Pore H2O,
Trench km Ref mm/yr Ref Ma Ref Thrust g/yr g/yr g/yr g/yr g/yr g/yr g/yr g/yr g/yr g/yr

Aegean 850 l 5 b 90 l 7.00E + 10 2.05E + 11 1.11E + 12 5.43E + 11 1.03E + 10


Alaska 1490 a 56 a 49 a 9.2 1.46E + 12 0 5.00E + 12 3.80E + 13 7.21E + 11 1.31E + 12 3.35E + 12 2.11E + 13 1.12E + 13 2.13E + 11
Aleutian, E. 1246 a 59 a 56 a 9.1 5.21E + 11 0 1.98E + 12 1.75E + 13 3.33E + 11 1.16E + 12 3.09E + 12 1.87E + 13 9.79E + 12 1.86E + 11
Aleutian, W. 1102 a 18 a 72 a 7.1 3.21E + 11 8.99E + 11 5.13E + 12 2.59E + 12 4.91E + 10
Andaman 1089 a 23 a,d 83 a 5.7 2.93E + 12 7.17E + 10 1.24E + 13 3.62E + 13 6.89E + 11 4.10E + 11 1.18E + 12 6.54E + 12 3.22E + 12 6.13E + 10
Antilles, N. 400 e 24 e 87 a 6.1 4.25E + 10 1.39E + 10 2.98E + 11 1.40E + 12 2.65E + 10 1.58E + 11 4.60E + 11 2.51E + 12 1.23E + 12 2.34E + 10
Antilles, S. 400 e 24 e 87 a 6.1 6.19E + 11 1.77E + 10 2.05E + 12 9.48E + 12 1.80E + 11 1.58E + 11 4.60E + 11 2.51E + 12 1.23E + 12 2.34E + 10
C. America 1506 a 72 a 16 a 7.8 2.60E + 11 1.02E + 13 1.76E + 12 3.63E + 13 6.91E + 11 1.54E + 12 2.77E + 12 2.54E + 13 1.60E + 13 3.03E + 11
Cascadia 990 a 36 a 5 a 7.2 1.69E + 12 0 6.46E + 12 3.02E + 13 5.73E + 11 4.53E + 11 3.86E + 11 7.73E + 12 5.71E + 12 1.09E + 11
Chile, C. 1306 a 71 a 23 a 9.5 1.36E + 12 2.80E + 12 2.23E + 13 1.33E + 13 2.52E + 11
Chile, N. 1579 a 74 a 41 a 8.5 1.80E + 12 4.42E + 12 2.92E + 13 1.60E + 13 3.03E + 11
Chile, S. 1218 a 16 a 16 a 5.7 2.76E + 11 4.98E + 11 4.57E + 12 2.87E + 12 5.45E + 10
Colombia 1355 a 65 a 21 a 8.8 2.96E + 10 5.08E + 12 1.58E + 12 1.62E + 13 3.08E + 11 1.28E + 12 2.56E + 12 2.10E + 13 1.27E + 13 2.41E + 11
Cotabato 500 k 20 f 42 n,i 1.54E + 11 3.81E + 11 2.50E + 12 1.36E + 12 2.59E + 10
Hikurangi 794 a 21 a 100 a 7.7 2.77E + 11 8.30E + 11 4.40E + 12 2.11E + 12 4.01E + 10
Izu-Bonin 1167 a 50 g 145 a 6.6 4.06E + 11 1.89E + 12 2.45E + 12 1.50E + 13 2.85E + 11 9.99E + 11 3.20E + 12 1.58E + 13 7.15E + 12 1.36E + 11
Japan, NE 1061 a 76 a 132 a 8.2 4.79E + 11 0 2.84E + 12 1.81E + 13 3.44E + 11 1.37E + 12 4.32E + 12 2.17E + 13 9.96E + 12 1.89E + 11
Java 1200 f 71 b 100 f 7.8 6.29E + 11 2.69E + 11 2.06E + 12 1.70E + 13 3.23E + 11 1.42E + 12 4.24E + 12 2.25E + 13 1.08E + 13 2.05E + 11
jarrard: subduction fluxes of water

Kamchatka 900 a 74 a 115 a 9.0 1.84E + 11 0 2.90E + 12 1.56E + 13 2.96E + 11 1.12E + 12 3.44E + 12 1.77E + 13 8.33E + 12 1.58E + 11
Kermadec 1422 a 52 a 100 a 8.0 2.00E + 11 0 7.74E + 11 1.20E + 13 2.28E + 11 1.23E + 12 3.68E + 12 1.95E + 13 9.37E + 12 1.78E + 11
Kurile 1243 a 75 a 128 a 8.5 5.54E + 11 0 3.29E + 12 2.09E + 13 3.98E + 11 1.58E + 12 4.95E + 12 2.50E + 13 1.16E + 13 2.19E + 11
Makran 950 c 37 b 97 d 6.79E + 12 1.23E + 13 1.28E + 13 6.10E + 13 1.16E + 12 5.82E + 11 1.73E + 12 9.26E + 12 4.46E + 12 8.48E + 10
Manila 1050 c 10 c 30 h 1.58E + 11 3.54E + 11 2.57E + 12 1.47E + 12 2.80E + 10
Mariana 1812 a 47 e 134 a 7.8 7.90E + 11 1.52E + 12 3.94E + 12 1.73E + 13 3.30E + 11 1.45E + 12 4.58E + 12 2.29E + 13 1.05E + 13 2.00E + 11
10.1029/2002GC000392

14 of 50
Geophysics
Geosystems
Geochemistry

Table 2. (continued)
G
3

Subducting Sediment Fluxes Igneous Crust Fluxes


+
Length, Conv. Rate, Age, Mw K2O, CO2, H2O , Pore H2O, Chloride, K2O, CO2, H2O+, Pore H2O, Chloride,
Trench km Ref mm/yr Ref Ma Ref Thrust g/yr g/yr g/yr g/yr g/yr g/yr g/yr g/yr g/yr g/yr

Mexico 1383 a 49 a 9 a 8.0 7.96E + 10 0 6.47E + 11 9.31E + 12 1.77E + 11 9.11E + 11 1.23E + 12 1.53E + 13 1.04E + 13 1.98E + 11
Nankai 824 a 38 g 23 a 8.1 6.25E + 11 0 1.21E + 12 4.82E + 12 9.16E + 10 4.59E + 11 9.47E + 11 7.53E + 12 4.48E + 12 8.52E + 10
Negros 400 k 20 c 16 m 1.13E + 11 2.04E + 11 1.88E + 12 1.18E + 12 2.24E + 10
New Britain 600 f 110 c,o 50 d 1.03E + 12 2.67E + 12 1.67E + 13 8.87E + 12 1.69E + 11
Peru 1599 a 65 a 37 a 8.1 7.00E + 10 1.35E + 12 5.09E + 11 1.05E + 13 2.00E + 11 1.59E + 12 3.79E + 12 2.58E + 13 1.43E + 13 2.72E + 11
Philippine 1509 a 64 g 43 a 8.0 8.64E + 10 4.90E + 11 3.72E + 11 9.50E + 12 1.80E + 11 1.49E + 12 3.71E + 12 2.42E + 13 1.31E + 13 2.50E + 11
Ryukyu 1153 a 57 g 46 a 6.3 2.28E + 11 0 6.08E + 11 8.54E + 12 1.62E + 11 1.02E + 12 2.59E + 12 1.65E + 13 8.90E + 12 1.69E + 11
San Cristobal 1050 c 49 p 50 d 8.07E + 11 2.08E + 12 1.30E + 13 6.92E + 12 1.31E + 11
Scotia 1005 a 44 a 57 a 7.0 1.30E + 11 0 4.70E + 11 6.59E + 12 1.25E + 11 7.01E + 11 1.87E + 12 1.13E + 13 5.88E + 12 1.12E + 11
Sulawesi, N. 600 c 30 d 42 n,i 2.78E + 11 6.87E + 11 4.50E + 12 2.46E + 12 4.66E + 10
Sulu 500 k 20 f 16 m 1.42E + 11 2.55E + 11 2.35E + 12 1.47E + 12 2.80E + 10
Sumatra 2462 a 50 a 61 a 7.1 6.22E + 12 3.55E + 11 1.83E + 13 8.23E + 13 1.56E + 12 1.96E + 12 5.31E + 12 3.15E + 13 1.63E + 13 3.09E + 11
Sunda, E. 950 f 71 b 145 j 7.10E + 11 2.64E + 12 3.20E + 12 2.01E + 13 3.82E + 11 1.15E + 12 3.70E + 12 1.82E + 13 8.26E + 12 1.57E + 11
Tonga 1460 a 148 a 100 a 8.5 1.56E + 11 0 6.22E + 11 1.24E + 13 2.36E + 11 3.59E + 12 1.08E + 13 5.70E + 13 2.74E + 13 5.20E + 11
Trobriand 590 f 20 c 50 d 1.85E + 11 4.77E + 11 2.98E + 12 1.59E + 12 3.01E + 10
Vanuatu 1189 a 111 a 40 a 7.5 1.29E + 12 3.33E + 12 4.20E + 12 5.26E + 13 1.00E + 12 2.03E + 12 4.95E + 12 3.29E + 13 1.81E + 13 3.43E + 11
jarrard: subduction fluxes of water

Yap-Palau 550 c 3 g 32 d 2.49E + 10 5.70E + 10 4.05E + 11 2.30E + 11 4.37E + 09


Total (26 zones) 2.72E + 13 3.95E + 13 9.27E + 13 5.79E + 14 1.10E + 13
Global total 44454 3.62E + 13 5.26E + 13 1.24E + 14 7.72E + 14 1.47E + 13 3.81E + 13 1.00E + 14 6.14E + 14 3.23E + 14 6.14E + 12
a
Fluxes for subducting sediments are modified from Plank and Langmuir [1998]. Thrust references in McCaffrey [1997]. References are as follows: (a) McCaffrey [1997]; (b) DeMets et al. [1990, 1994];
(c) von Huene and Scholl [1991]; (d) Jarrard [1986]; (e) Plank and Langmuir [1998]; (f) this study; (g) Seno et al. [1993]; (h) Brias et al. [1993]; (i) Beirsdorf et al. [1997]; (j) Sager [1992]; (k) Rangin and
Silver [1990]; (l) Robertson et al. [1996]; (m) Shipboard Scientific Party [1990a]; (n) Shipboard Scientific Party [1990b]; (o) Honza et al. [1987]; (p) Tregoning et al. [1998].
10.1029/2002GC000392

15 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

duction fluxes of four components focused on in ment less accurate for CO2 than for the other three
this study: structural water (H2O+), pore water, components.
CO2, and K2O. Chloride, the fifth component, is
readily calculated from pore water. Their tabula-
tions confirmed that H2O+ and K2O are ubiquitous 3.3. Oceanic Crustal Layers
in subducting sediments: H2O+ occurs mainly in [48] Oceanic crustal layering was first determined
clays but also in opal, and K2O is present in the based on seismic refraction. The Raitt-Hill layer-
nonbiogenic, terrigenous component. In contrast, ing [Hill, 1957; Raitt, 1963] consists of the fol-
CO2 is often virtually absent from subducting lowing: Layer 2, the ‘‘volcanic layer’’, which is
sediments; it is, of course, highest in calcareous 1.71 ± 0.75 km thick; underlain by Layer 3, the
units, but it is also moderately abundant in clays ‘‘oceanic layer’’, which is 4.86 ± 1.42 km thick;
and volcaniclastics. separated by Moho from Layer 4, the upper
mantle. Many subsequent seismic experiments
[46] Plank and Langmuir [1998] used more accu-
have refined the seismic structure of Layer 2, its
rate, trench-specific measurements of partitioning
sublayers 2a, 2b, and 2c, and Layer 3. The syn-
between sediment subduction and accretion than
thesis of White et al. [1992] concluded that normal
any other study. Many of their rates and trench
Layer 2 is 2.1 ± 0.6 km thick and Layer 3 is 5.0 ±
lengths were from von Huene and Scholl [1991]
0.9 km thick, for a total crustal thickness of 7.1 ±
and are now superceded. Applying the revised
0.8 km.
rates (mostly NUVEL-1A) and trench lengths of
this study (Table 2) to Plank and Langmuir’s [49] Attempts to correlate this layering with petro-
[1998] concentrations of structural water, pore logical variations have met with limited success.
water, CO2, and K2O changes fluxes at most Moho is a distinctive boundary, but seismic Moho
individual subduction zones substantially. Their may not correspond exactly with petrological
global fluxes of these four components are quite Moho. The deepest drilling into in situ oceanic
robust, however, and are revised by only 1.5% to crust is at Hole 504B. The top layer here is a 572
+3%. m extrusive volcanic section, underlain by a 208-
m thick transition zone (572–780 m sub-base-
[47] A limitation of the Plank and Langmuir
ment), underlain by sheeted dikes that continue to
[1998] study is that its reliance on reference drill
the 1836 m sub-basement bottom of the hole. The
sites precludes analysis of 14 of the subduction
top 100 m is low-velocity Layer 2A, but the
zones of this study. The most volumetrically sig-
main petrological change within the extrusives is a
nificant omissions are probably the Aegean and the
change from more open to more restricted low-
three segments of Chilean subduction zones. Their
temperature alteration near the middle of the
analysis includes two thirds of the global trench
extrusives [Alt et al., 1996b]. The layer 2B/2C
length represented in Table 2, but a more useful
boundary, generally considered to mark the
correction factor would be percentage of globally
change from extrusive basalts to dikes, occurs
subducted sediment included in their analysis.
within the transition zone. The Layer 2C/3 boun-
Based on the virtually complete suite of global
dary has been commonly considered to be the
subduction zones analyzed by von Huene and
change from dikes to gabbros [Fox and Stroup,
Scholl [1991], the subset analyzed by Plank and
1981], but at Hole 504B it is probably a meta-
Langmuir [1998] accounts for 75% of total glob-
morphic front within the sheeted dikes [Salisbury
ally subducted sediments. Consequently, I assume
et al., 1996].
that the global sediment-hosted fluxes of structural
water, pore water, CO2, and K2O are one-third [50] I follow the example of Hole 504B in
higher than those calculated by Plank and Lang- assuming that the upper extrusive layer is 600 m
muir [1998] and slightly refined above. It should thick, divided into two 300-m thick layers that
be noted, however, that the heterogeneity of sedi- may have significantly different porosity struc-
ment CO2 contents makes that one-third adjust- tures and associated alteration states (Table 1). I
16 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

include the transition zone with the sheeted dikes, the normal crustal sites tabulated by Jarrard et al.
because its alteration history is generally similar to [2003].
that of the upper dikes [Alt et al., 1996b], but its
velocity and porosity structure are more like the 3.4.2. Macroporosity
lower extrusives than the sheeted dikes. I follow [54] Macroporosity, consisting mainly of cracks
many previous investigators in assuming that the and interpillow voids, decreases systematically
base of the dikes is at 2 km sub-basement, while with increasing age, as described by the linear
recognizing that this assumption is largely based regression of Figure 2c. The secondary minerals
on the outdated interpretation of Layer 2C/3 at 2.1 responsible for this macroporosity reduction are
km as this petrological boundary. The seismic assumed to be an 80:20 mix of saponite and
Moho at 7 km is taken as the base of oceanic crust celadonite (section 2.8), the same as at an inter-
and the lower boundary of these analyses. These granular scale. Calcite is undoubtedly present as
layers are generally much thinner in ophiolites well at a relatively small percentage, which is
[Coleman, 1977]. An exception is the Semail already accounted for in the analysis of Alt and
ophiolite, with crust that is 7 km thick. The seismic Teagle [1999] and section 3.4.5 below.
structure of this ophiolite is generally comparable
to that of normal oceanic crust, but the extrusive 3.4.3. Total Porosity
section is twice as thick as at Hole 504B, the
[55] Total porosity is simply the sum of micro-
sheeted dikes extend to 3.3 km, and some perido-
porosity and macroporosity.
tites occur above seismic Moho [Christensen and
Smewing, 1981]. 3.4.4. Matrix Density
[51] Crustal accretion at slow spreading rates [56] Matrix density is assumed to decrease system-
involves more episodic volcanism, a deeper melt atically with increasing age, as described in the
lens [Purdy et al., 1992] and more abundant fault- linear regression of Figure 2b, due to gradual
ing than that at intermediate-to-high rates, resulting replacement of primary minerals by secondary
in disrupted igneous stratigraphy and more com- minerals.
plex, heterogeneous structure [e.g., Cannat, 1996;
Karson, 1998]. Despite these differences, seismic 3.4.5. Carbon Dioxide
crustal thickness is only weakly correlated with
[57] Carbon dioxide is assumed to increase system-
spreading rate [White et al., 1992; Small, 1998],
atically with increasing age, as described in the
except for rare occurrences of rates <6 mm/yr
linear regression of Figure 2d.
[Dick et al., 2002].
3.4.6. Potassium and Bound Water
3.4. Properties of Upper Extrusives
[58] Potassium and intergranular bound water are
[52] Values for the top 300 m of the extrusive calculated from age-dependent matrix density (Fig-
section (Table 1) are based on the core and log ure 2b), attributing reduction in matrix density to
analyses of section 2. an 80:20 mixture of saponite and celadonite (Fig-
ure 5). Bound water is also present within alter-
3.4.1. Microporosity ation minerals filling fractures and interpillow
voids (section 3.4.2); these alteration minerals are
[53] Microporosity is a routine shipboard core
also assumed to be a saponite/celadonite mix.
measurement. The primary control on microporos-
ity of extrusives is volcanic style: pillows have
microporosities that are several percent higher 3.5. Properties of Lower Extrusives
than more massive flows. I assume an age-inde- [59] Significant sections of the lower extrusives
pendent microporosity of 7.8% for the upper (300–600 m sub-basement) have been penetrated
crust, based on averaging 839 measurements from at only four sites: Holes 395A, 418A, 504B, and
17 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

801C. Consequently, generalizations concerning et al., 1985, 1986]. The tholeiitic section at Hole
the properties of this zone are based primarily on 801C penetrates only 100 m of the lower extru-
comparisons of these four sections to their over- sives, and depth-dependent changes in alteration
lying extrusive zones. Tectonic windows provide are difficult to isolate from flow/pillow-induced
another perspective on this zone [e.g., Karson, heterogeneity (Figure 4).
2002], if they are representative [Francheteau et
[62] Chemical analyses generally underestimate
al., 1975].
alteration because of sampling bias and underre-
[60] The upper and lower extrusives are expected to covery of both altered basalts and macroporosity-
be broadly similar in geophysical characteristics, filling alteration minerals. For example, despite the
volcanically and hydrologically. Volcanic style may observed difference in alteration between upper
exhibit recurrent patterns of flows, pillows, and and lower extrusives at Hole 504B [Alt et al.,
hyaloclastics [Robinson et al., 1979; Schmincke 1996b], H2O+ analyses are similar: 0.86% (std.
and Bednarz, 1990; Pezard, 1990; Pezard et al., dev. 0.32%, N = 51) and 0.83% (std. dev. 0.66%,
2000], but resulting stratigraphic architecture has a N = 77), respectively. The short interval of lower
wavelength of tens of meters, not several hundreds extrusives at Hole 801C continues an overall
of meters. Large-scale permeability measurements pattern of downhole decrease in alteration, partic-
of upper oceanic crust are too sparse for quantitative ularly decrease in CO2 and possibly also H2O+ but
comparison of upper versus lower extrusives, but not K2O [Shipboard Scientific Party, 2000a]. Stau-
the first-order generalization of these studies is that digel et al. [1996] quantitatively distinguished
the entire extrusive interval acts as a hydrologic unit recovered and unrecovered materials for Site 417/
during off-axis hydrothermal circulation [Fisher, 418, by carefully weighting their samples to
1998; Fisher and Becker, 2000]. Seismic studies achieve representative sampling. Like some core/
(e.g., ESP, OBS, OBH) indicate strong velocity log comparisons, their analysis demonstrated that
gradients within the top few hundred meters of total abundance of alteration minerals is much
oceanic crust [Vera and Mutter, 1988; White et higher than most XRF measurements detect. In
al., 1992], whereas no systematic pattern of down- the lower extrusives, average K2O is 14%, struc-
hole velocity increase is detected within the upper tural H2O is 64%, and CO2 is 19% of that in the
300 m of basement for DSDP and ODP sites upper extrusives [Staudigel et al., 1996].
[Jarrard et al., 2003].
3.5.2. Log-Based Alteration
3.5.1. Core-Based Alteration and Total Porosity
[61] Core descriptions of the four deep-crustal sites [63] Potassium data from cores indicate that most
provide ground-truth comparisons of variations in K-rich alteration minerals are in the uppermost
alteration with depth. In all four extrusive sections, crust, and they suggest a near-exponential decrease
smectite is the most abundant alteration mineral. At to 600 m sub-basement [Booij and Staudigel,
Hole 395A, depth-dependent alteration was noted 1997]. Potassium logs provide unbiased, integrated
principally as downhole increases in non-carbonate comparisons of total potassium enrichment for
veins [Shipboard Scientific Party, 1978] and tem- upper versus lower extrusives at Holes 395A,
perature of mineral formation [Lawrence et al., 418A, and 504B. The four potassium logs for Hole
1978], though nearly all diagenesis is low-temper- 395A indicate that the lower extrusives have 110%,
ature. At Hole 418A, alteration is consistently 101%, 120%, and 128% of the K2O content of the
higher in the top 200 m than in the underlying overlying extrusives. In contrast, the potassium log
250 m [e.g., Broglia and Ellis, 1990]. At Hole for the lower extrusives at Hole 418A averages
504B, the upper and lower extrusives are generally only 57% of the K2O in the upper extrusives. At
similar in alteration mineralogy, but the lower Hole 504B, the corresponding proportion is only
extrusives are less altered [Alt et al., 1996b] and 35% based on the potassium log and 46% based on
have more chlorite and mixed chlorite-smectite [Alt 458 XRD measurements (Figure 3).
18 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

[64] All three sites have both multichannel sonic legs at this site have cored and logged through a
and reprocessed Schlumberger sonic logs that yield 572-m thick extrusive layer, a 208-m transition
similar large-scale velocities; in conjunction with zone, and 1056 m of sheeted dikes, thought to
core-based velocities, these provide macroporosity represent virtually the entire dike layer. No other
estimates. In general, the deeper extrusives have DSDP or ODP site has sampled the dike layer
macroporosities that are about half those of the (Layer 2C and upper Layer 3), though isolated
shallower extrusives: 0–3.7% versus 6.0–8.2% at dikes have been encountered within the extrusive
Hole 395A, 1.8–2.1% versus 3.4–4.2% at Hole layer. The only other stratigraphic sections of this
418A, and 2.5–5.0% versus 8.0–8.1% at Hole portion of oceanic crust are from ophiolites [e.g.,
504B. Similarly, core-based microporosities for Coleman, 1977; Christensen and Smewing, 1981],
Holes 395A, 418A, 504B, and 801C are signifi- whose geophysical properties may have been
cantly lower for the lower extrusives than upper affected by emplacement and post-emplacement
ones, averaging 5.1% (std. dev. 3.8%, N = 154) and processes. Consequently, this study bases its char-
7.8% (std. dev. 5.9%, N = 839), respectively acterization of the dike layer on Hole 504B,
(medians 4.0% and 5.8%). Matrix densities are combining the transition zone and sheeted dikes
different as well: 2.97 g/cc for lower extrusives into a nominally uniform layer.
versus 2.93 g/cc for upper ones, corresponding to
[68] Alt et al. [1996b] provided a synthesis of
8% and 16% alteration, respectively, based on the
alteration mineralogy and the multistage alteration
mineral alteration model of section 2.8.
history of the Hole 504B dikes. Olivine alters to
[65] In summary, the bulk of the available evidence chlorite, clay, and talc, plagioclase to albite, zeo-
indicates that the lower half of the extrusive layer has lites, chlorite, or clay, and augite rarely alters to
about half as much macroporosity as the upper half. actinolite [Alt et al., 1996b]. Alteration of the upper
This difference is apparently not alteration induced; dikes and transition zone began with greenschist-
indeed, matrix densities and potassium logs suggest facies metamorphism at 350–380C during axial
about half as much hydration and potassium enrich- hydrothermal alteration, followed by mixing of
ment for the lower extrusives as for the upper ones. upwelling hydrothermal fluids with cooler sea-
The weight of the volcanic pile may cause a few water circulating through the extrusives. Alteration
percent more macroporosity reduction of the lower of the lower dikes began with exposure to very
extrusives compared to upper ones, thereby reduc- high-temperature (>400C) hydrothermal vent flu-
ing permeability and associated fluid flow and gen- ids, followed by 300–400C greenschist metamor-
erating a seismically detectable vertical velocity phism, and concluding with rare off-axis (<250C)
gradient. Seawater that does eventually penetrate veins [Alt et al., 1996b].
the lower extrusives generates reducing conditions, [69] Index measurements on 91 samples from the
lacking the celadonite of the upper extrusives but transition zone and sheeted dikes of Hole 504B
retaining the strong smectite dominance. establish a mean matrix density of 2.98 g/cc (stand-
[66] The flux estimates of this paper assume that ard deviation 0.04 g/cc) and mean microporosity of
the lower extrusives exhibit age-dependent macro- 2.25% (standard deviation 2.10). Macroporosity
porosity and alteration that are half those of the calculation based on comparison of lab and log
overlying extrusives. Age-dependent changes in velocities must use lab measurements at elevated
matrix density, structural water, carbon dioxide, pressure rather than at atmospheric pressure,
and potassium are calculated from this assumption because microcrack opening becomes significant
of proportional alteration (Table 1). at depths >300 m into the dikes. With this
approach, Salisbury et al. [1996] estimated dike
macroporosities as 0–1.4%. Average macroporosity
3.6. Properties of Dikes of the dikes is 0.8%, based on lithostatic-pressure
[67] Hole 504B could be the type section for core measurements. Most of this macroporosity is
Layer 2 of oceanic crust. Seven DSDP and ODP concentrated in the transition zone, where macro-
19 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

porosity is 5.6%, in contrast to near-zero (mean logging of 1075– 1823 mbsf on Leg 148, and
0.1%) in the sheeted dikes. 0.057% for logging of 1349–1871 mbsf on Leg
140. The difference between core and log K2O
[70] Concentrations of structural water are rela- could be caused by either non-representative core
tively uniform within the dikes of Hole 504B, recovery (as observed within the extrusives) or
except for significantly higher values within the measurement bias. I attribute the difference to
transition zone (mean 2.37%, std. dev. 0.78%, N = biased log measurements, resulting from K2O
140) than in the sheeted dikes (mean 1.67%, std. levels that are much lower than those anticipated
dev. 0.67%, N = 274). I assume that the overall during tool design. For example, minimum K2O
average of 1.76% (std. dev. 0.72%, N = 314) is detectability and minimum K2O readings are both
representative of Layer 2C globally. Unlike the sensitive to trace amounts of K2O in the detector
situation for the extrusives, this value is probably crystal. Different detector crystals may account for
not significantly biased by loss of structural water the 65% difference in average K2O for logs
during sample preparation, because pure smectites obtained on different legs. I assume that K2O is
are absent and expandable clays are not abundant sufficiently uniform within the dikes (Figure 3) for
[Alt et al., 1985, 1996b]. Although core recovery in the mean XRF-based K2O of 0.014% to provide a
the dikes is generally low, the core-based determi- representative estimate of the K2O content of
nation of structural water may not be significantly Layer 3. This low value results from leaching of
biased by preferential coring loss of crack-fill an already K2O-depleted magmatic composition
alteration minerals, because H2O+ measurements [Alt et al., 1996a, 1996b]. More typical MORB
suggest relatively uniform hydration. Broglia and compositions may or may not have higher final
Ellis [1990] analyzed a neutron log of upper dikes K2O contents; K2O availability for leaching may
(875 – 1250 m sub-basement) of Hole 504B, determine final K2O content (J. Alt, personal
accounting for borehole effects, porosity, and struc- communication, 2002).
tural water contributions. They concluded that
volume percentage of total water is 7–8% within [73] With only one dike section available, it is
the upper dikes; of this, porosity is <2% and is impossible to examine the possibility of age-
probably entirely microporosity, whereas hydrous dependent dike alteration directly. Because the
mineral content of about 10–30% contributes 6% alteration mineral assemblage within the Hole
H2O. 504B dikes is mainly high-temperature [Alt et al.,
1996b], it is largely attributable to near-ridge
[71] Considering both whole-rock CO2 and vein alteration. Consequently, I assume that dike alter-
carbonate, average CO2 at Hole 504B is 0.58 wt.% ation is age-independent, at least for the 5–145 Ma
for the transition zone and 0.07% for the sheeted average ages of subducting crust in modern sub-
dikes [Alt and Teagle, 1999]. The thickness- duction zones (Table 2). The heat flow pattern at
weighted average is 0.14%. Hole 504B suggests present-day hydrothermal cir-
culation within the dikes [Wang and Davis, 2002],
[72] K2O content of the dikes of Hole 504B has
and late-stage dike alteration may be occurring [Alt
been measured via XRF analyses of cores and
et al., 1996b], so the assumption of age-independ-
downhole logging (Figure 3). K2O leaching [Alt
ent alteration could be only a first-order approx-
et al., 1996b] has so depleted this interval, how-
imation. I apply the Hole 504B values for dike
ever, that it is near or at the resolution of both
microporosity, macroporosity, structural water,
methods. Based on 343 XRF measurements, aver-
K2O, CO2, and matrix density to the dike layers
age K2O is 0.014% (std. dev. 0.013), with values
at all subduction zones.
that decrease gradually downhole (Figure 3) from a
level of 0.023% (std. dev. 0.011) in the transition
zone to 0.011% (std. dev. 0.012) in the underlying 3.7. Properties of Gabbros
sheeted dikes. Spectral gamma ray logging indi- [74] The processes that bring the gabbro layer
cates significantly higher K2O: 0.034% for ACT within sampling reach also assure that the samples
20 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

will be nonrepresentative of normal in situ con- the layered plutonics [Lippard et al., 1986] and by
ditions. DSDP and ODP have encountered short actinolite and Fe-Ti oxides plus cavity fillings in
gabbro intervals at various sites, often where only more isotropic plutonics [Ernewein et al., 1988].
basalts were anticipated. Significant penetrations of
Layer 3 gabbros were obtained at two localities: [77] Measurements of structural water at Hole
Hole 735B on Atlantis Bank of Southwest Indian 735B are most abundant for the interval drilled
Ocean Ridge [Shipboard Scientific Party, 1989, during Leg 118: 0.79% average of 94 samples
1999] and Site 894 at Hess Deep [Shipboard [Shipboard Scientific Party, 1989]. I assume that
Scientific Party, 1993]. At both localities, faulting this value is representative of gabbros in general,
has unroofed the gabbroic section and exposed it at but high-temperature alteration of gabbros may
the seafloor, resulting in stratigraphic sections that generally be limited to the upper few hundred
are enhanced in fracturing, microcracks, and low- meters by available fracture permeability [Alt,
temperature alteration, compared to what is likely 1995]. At Hole 735B, high-temperature alteration
for normal in situ Layer 3 [Shipboard Scientific extends to about 600 mbsf, and unroofing-induced
Party, 1989, 1993, 1999; Bach et al., 2001]. Never- low-temperature alteration is locally common
theless, I assume that these two sites provide the below 500 mbsf [Bach et al., 2001]. Consequently,
most representative sample of Layer 3 currently the 1.97% average structural water concentration
available. for 20 altered samples from below 500 mbsf is less
representative than the 0.79% average for 18 fresh
[75] Hole 735B was cored to 501 mbsf on Leg 118 samples [Bach et al., 2001]. The gabbros cored at
and deepened to 1508 mbsf on Leg 176. Core Hess Deep Site 894 average 1.26% for 22 samples
recovery was exceptionally good, but loss of the (1.39% including one extremely hydrated sample).
bottomhole assembly during Leg 176 prevented Carlson [2001] cited an average bound water
downhole logging of much of the newly drilled content of only 0.04% for the gabbros of Holes
interval. The entire drilled interval consists of 735B, 894G, and 923A.
Layer 3 gabbros. Index measurements on 346
samples establish a mean matrix density of 2.99 [78] CO2 content of gabbros is assumed to be 0.02
g/cc (standard deviation 0.12 g/cc) and mean wt.%, the average of whole-rock analyses from
microporosity of 0.74% (standard deviation 2.76). upper Hole 735B, Hess Deep, and MARK sites
Ildefonse and Pezard [2001] measured 63 addi- [Alt and Teagle, 1999]. Fresh gabbros from the
tional samples, which have a mean matrix density deeper portion of Hole 735B, however, have a
of 2.98 g/cc and microporosity of 1.11%. The much higher median CO2 of 0.09% [Bach et al.,
microporosity distribution is, however, non-normal, 2001].
skewed positively by a few high-porosity samples;
[79] K2O content of the gabbros of Hole 735B has
median microporosity of the 346 samples is only
been measured via XRF core analyses and down-
0.2%. This microporosity is partly microcracks,
hole logging. Based on 280 XRF measurements
based on resistivity measurements [Ildefonse and
from Legs 118 and 176 [Shipboard Scientific
Pezard, 2001] and on velocity measurements that
Party, 1989, 1999], average K 2 O is 0.065%
average 5.6% higher at 200 MPa than at atmos-
(0.061% excluding a single extreme value of
pheric pressure [Stephen, 2001]. Microcracks may
3.97%). Leg 176 samples affected by low-temper-
not be present in unexhumed gabbros of normal
ature alteration average twice as much K2O as
oceanic crust.
fresh samples [Bach et al., 2001]. Three spectral
[76] Amphibolite-facies alteration of upper Hole gamma ray logs of the upper portion of this interval
735B is extensive, averaging 20–30% and with give reasonably consistent average K2O contents
veins comprising 2.4% of total volume [Dick et al., (0.124%, 0.138%, and 0.146%) that are twice as
1991; Robinson et al., 1991]. By comparison, high as XRF results. The good log replicability and
alteration of the Semail ophiolite is characterized mean K2O readings that are apparently well above
by rare conversion of clinopyroxene to actinolite in measurement threshold argue against log measure-
21 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

ment bias as an explanation. The logs may provide range is only 5.5–5.7%, a variation of only 3%
a more representative sample than cores of the K2O among incoming crust at modern subduction
contents of gabbros at Hole 735B, yet be less zones. This similarity arises from two factors. First,
appropriate for evaluating the K2O of gabbros in all age-dependent changes in macroporosity and
general. The logs are more sensitive to occasional alteration within dikes and gabbros are assumed to
K2O spikes caused by low-temperature alteration occur near-ridge, so that the structural and pore
minerals, present at Hole 735B because of its water contents of these portions of the crust are the
unroofing history but not expected to be present same throughout the 5–145 Ma average ages at
in deep gabbros. modern subduction zones. Age-dependent changes
in macroporosity and alteration are confined to the
[80] I assume that Layer 3 gabbros have an average
extrusives, which account for only 16–31% of
K2O of 0.061%, based on the Hole 735B XRF
total crustal structural water and 35–51% of total
results; this value is close to the weighted average
crustal pore water. Second and more important,
of 0.05% determined by Dick et al. [2002]. This
age-dependent decreases in upper crustal macro-
value is significantly less than the 0.11% average
porosity are nearly balanced by increases in struc-
for the upper third of this hole based on strip
tural water. Most of this structural water is within
samples [Hart et al., 1999]. It is consistent with
the mineral matrix, mainly replacement of feldspar
the XRF-based average of 0.08% for the much
and olivine by saponite; only 17% of macroporos-
shorter interval of gabbros sampled at Hess Deep
ity fill is structural water. Although the average
Site 894. It should be noted, however, that uncer-
structural water content of the extrusive section
tainty in gabbro K2O has a huge impact on sub-
varies by a factor of 2.3 (2.2–5.1%) among sub-
sequent calculations of total K 2 O flux into
duction zones, and average pore water content
subduction zones: using the log-based value,
varies by a factor of 1.9 (2.6%–5.0%), leading to
instead, would increase computed total K2O flux
an 18% regional variation in total bound water and
by 71%.
24% variation in total pore water, total water varies
by only 3%.
4. Subduction Fluxes [83] The mass balance for water addition and
removal within oceanic crust appears to permit
4.1. Water Flux off-axis crustal alteration to be confined to the
[81] The annual global flux of subducted water is oceanic crust, without significant interchange with
1.83  1015 g/yr. This flux includes four major seawater, such as might be expected when a
sources (Figure 6): sediment pore water (42%), relatively impermeable sediment blanket inter-
sediment structural water (7%), igneous crustal venes. Similarly, much of late-stage intergranular
pore water (18%), and igneous crustal structural alteration is accomplished by local redistribution of
water (33%). The global contributions from sub- elements without substantial bulk-rock geochemi-
ducting sediments and crust are nearly equal. cal change [Alt, 1999; Zhou et al., 2001a]. How-
Regionally, in contrast, combined structural and ever, interchange with seawater to at least 65 Ma
pore water of incoming igneous crust is nearly is indicated by both heat flow data and by the other
uniform, whereas the quantity of sediment pore elemental mass fluxes of this study. The structural
water subduction per meter of trench length varies water enrichment at the expense of pore water
by more than an order of magnitude. Variations in implies either chlorine outflow or increase in pore
total-water fluxes among individual subduction water chlorinity, and ongoing additions of both
zones (Figure 6) are most sensitive to convergence potassium and CO2 have been documented earlier.
rate, trench length, and thickness of the subducting
sedimentary section. 4.1.1. Comparison to Prior Flux Estimates
[82] The computed water contents for crust of [84] My calculated global fluxes of 1.24  1014
various ages are remarkably uniform: their total g/yr for structural water and 7.72  1014 g/yr for
22 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Figure 6. Water fluxes at global subduction zones, from Table 2. H2O+ is bound water in minerals; for crustal
H2O+, loosely bound water is distinguished from water in stabler, more temperature-resistant minerals.

pore water in subducting sediments are based on my 7.7  1014 g/yr: 1  1015 g/yr [COSOD-II,
minor revisions of the comprehensive evaluation 1987], 1  1015 g/yr [von Huene and Scholl,
of Plank and Langmuir [1998]. An earlier value 1991], 1.8  1015 g/yr [Moore and Vrolijk, 1992],
of 0.7  1014 g/yr for structural water by Peacock and 9  1014 g/yr [Rea and Ruff, 1996]. The
[1990] was based on a single hypothetical sub- good agreement of Rea and Ruff’s [1996] value
ducting sediment column, in contrast to the eval- with Plank and Langmuir’s [1998] result is partly
uations for individual subduction zones by Plank attributable to methodological similarities in the
and Langmuir [1998]. Most prior estimates for two analyses, including use of many of the same
sediment pore water flux are slightly higher than reference sites.
23 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

[85] Most previous analyses of water budgets for oceanic crust (versus my 3.7  1014 g/yr for the
oceanic crust do not quantify pore water; they same portion).
focus instead on structural water. Estimates for the
[87] Nearly all of the water that enters subduction
latter are higher than the 6.14  1014 g/yr of this
zones comes from seawater or – in the case of
study: 8.8 ± 2.9  1014 [Ito et al., 1983] and 8 
bound water in sediments – terrestrial hydrous
1014 [Peacock, 1990]. Ito et al. [1983] used
minerals. Seven studies of fresh MORB glasses
assumed abundances and associated water con-
indicate about 0.15–0.25% H2O+ [Ito et al., 1983].
tents for hydrous alteration minerals in an oceanic
Sobolev and Chaussidon [1996] argued that fluid
crust consisting of alteration-based layers: halmyr-
inclusions in olivine phenocrysts provide a more
olysis, greenstone, amphibolite, and unaltered gab-
reliable measure of initial MORB H2O+ (0.12
bro. Their structural-H2O values are similar to
wt.%) than glass does. This value, in conjunction
ours for the extrusives, higher for the dikes, and
with a crustal consumption rate of 2.4 km2/yr,
much lower for the gabbros. Peacock [1990]
basaltic layer thickness of 2 km, and weighted
assumed 2% structural H2O for basalts and 1%
average density of 2.76 g/cc, indicates that 1.6 
for gabbros, based on consideration of whole-rock
1013 g/yr of subducted water, or 0.9% of the total
analyses from three upper-crustal sites and the
subducted water, is primary water from crustal
gabbros of Hole 735B and Oman. Peacock
generation.
[1996] noted that dredged rocks indicate much
higher structural water: 1.7–4.9% (avg. 3.5%) for
basalts and 1.6–5.8% (avg. 2.5%) for gabbros, 4.1.2. Slab Fluid Expulsion Pattern
based on the data of Anderson et al. [1976]. [88] Subducting fluids control first-order structural
Bebout [1995, 1996] suggested that total water and petrologic problems: structural style and evo-
flux may be 18  1014 g/yr, about double the lution of accretionary prisms [Davis et al., 1983]
structural-H2O values of Ito et al. [1983] and and generation of arc magmas [Gill, 1981] and
Peacock [1990], to account for the high H2O+ consequently long-term growth of continents [e.g.,
contents of outcropping metamorphosed subduc- Reymer and Schubert, 1984]. Unfortunately, even
tion-zone rocks such as the Catalina Schist. He rough fluid budgets demonstrate that our knowl-
noted, however, that this additional water may edge of slab fluid expulsion patterns still faces
come from sedimentary or basaltic pore fluids. first-order uncertainties: present-day fluid expul-
My structural-H2O flux for the extrusives, 1.58  sion rates from some subduction zones are an
1014 g/yr, is significantly higher than the 1.01  order of magnitude higher than fluid sources [Le
1014 of Staudigel et al. [1996], due to a combi- Pichon et al., 1991; Kastner et al., 1991], and an
nation of higher structural H2O, thicker extrusive order of magnitude more fluid enters subduction
section (600 m versus 500 m), and lower porosity zones than is eventually released within arc mag-
(7–13% versus 18.5%). mas [Ito et al., 1983]. Yet, the 1.83  1014 g/yr
loss of water to subduction zones must be nearly
[86] Moore and Vrolijk [1992] calculated global balanced by gains elsewhere, because even a
water fluxes for both subducting sediments and sustained 20% imbalance implies a long-term sea
oceanic crust, using trench lengths and conver- level change of 1 m/Ma, which is unlikely on the
gence rates from von Huene and Scholl [1991], time scale of >100 Ma.
upper crustal porosity and hydrous minerals for
Hole 504B [Becker et al., 1990], and hydrous [89] Figure 7 summarizes slab fluid expulsion
minerals in sediments of Kastner et al. [1991]. patterns, based on the incoming global water
They estimated fluxes (typo-corrected) of 1.8  budgets of Table 2 and dehydration evidence
1015 g/yr for sediment pores (versus my 7.7  1014 discussed in this section. A useful starting point
g/yr), 4.3  1014 g/yr for H2O+ in sediments is to stipulate that there are five types of sub-
(versus my 1.2  1014 g/yr), and 3.5  1014 g/yr ducting water, each of which might be released
for pore and structural water in the top 1 km of at a different point in the subduction process:
24 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Figure 7. Hypothesized water expulsion pattern in a generalized subduction zone. Bottom: cross-section of a
subduction zone, with solid arrows showing escape paths of subducting water, and dashed lines showing mantle flow
paths. Dehydration mechanisms are shown above the cross-section, with solid bars above the portion of slab
undergoing each dehydration reaction. The plot of total global subducted water semiquantitatively illustrates
expulsion magnitudes; starting values are from Table 2, but expulsion rates are only approximate, as discussed in text.

sediment pore water (7.7  1014 g/yr), sediment smectite interlayer water is released at temper-
structural water (1.2  1014 g/yr), igneous-crust atures below 150C.
pore water (3.2  1014 g/yr), igneous-crust
loosely bound structural water (1.6  1014 [90] Subduction at margins containing accretionary
g/yr), and igneous-crust firmly bound structural prisms expels water due to compaction within the
water (4.6  1014 g/yr). A likely sixth source, accreting and underthrust sediments [Bray and
structural water in serpentinites, is ignored here Karig, 1985] and transformation of smectite to illite.
because its magnitude is unknown. Structural Kastner et al. [1991] and Le Pichon et al. [1991]
water within the extrusive portion of oceanic simultaneously demonstrated the discrepancy
crust is mostly in smectites, whereas that in between present rates of water expulsion at Barba-
the dikes and gabbros is in more temperature- dos, Nankai, and Peru prisms and the amount of
resistant minerals such as actinolite and horn- water that can be accounted for by generation within
blende, so release of the former may be more the prism. Kastner et al. [1991] calculated average
closely related to smectite breakdown within water expulsion of 7 m3/yr per m of trench (global
sediments than it is to breakdown of lower mass of 3  1014 g/yr) from internal fluid sources
crustal hydrous minerals. Structural water within (compaction and dehydration), much less than the
the sediments is primarily in clays (particularly 100 m3/yr per m of trench presently being vented at
smectite) and secondarily in opal, and most the three margins. The difference was initially
25 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

thought to be provided mainly by meteoric water loss at depths of 15–45 km is similarly updip,
[Kastner et al., 1991] and shallow seawater con- toward the toe of the prism [Bebout, 1991]. At
vection [Le Pichon et al., 1991]. these depths, fluid loss from subducted sediments
is nearly completed and slab devolatilization
[91] Isotopic studies and identification of low- begins, driven by metamorphism rather than com-
chlorinity and high-methane anomalies demon- paction. Bebout [1995] documented this process
strated the significant contribution from deeper for the metasedimentary and metamafic rocks of
diagenetic processes, particularly smectite dehy- the Catalina Schist, including loss of both water
dration and hydrocarbon generation [Kastner et and CO2 and associated homogenization of oxygen
al., 1991, 1993]. In addition, the assumption of and hydrogen isotopic signatures. Bebout [1996]
steady state compaction may be invalid [Moore extended these observations to other eclogite-facies
and Vrolijk, 1992; Le Pichon et al., 1993]. Recent subduction complexes. For both metasedimentary
modeling of flow and solute transport at Nankai and metamafic rocks of the Catalina Schist, volatile
accretionary prism incorporated transient flow, loss progresses with increasing metamorphic grade,
caused by a hydrofracture-induced temporary from H2O+ concentrations of 5–6% for lawsonite-
increase in permeability along the décollement albite to 1–2% for amphibolite-facies metamor-
[Saffer and Bekins, 1998]. phism [Bebout, 1995]. Bebout [1995] noted that
[92] Subducted sediments are initially overpres- H2O+ contents of up to 10% for the low-grade
sured and underconsolidated [Davis et al., 1983; metabasalts are higher and more homogeneous
Moore and Vrolijk, 1992]; rate of fluid expulsion than those of DSDP/ODP basalts and ophiolites.
depends on a combination of loading, convergence He suggested that low-grade metamorphism trans-
rate, and permeability [Saffer and Bekins, 2002]. forms both original hydrous minerals and pore
Normal consolidation is achieved and most smec- water into new hydrous minerals.
tite dehydrates after perhaps 30–40 km of sub-
duction (5 km burial) [Saffer and Bekins, 1998; [94] Although the Catalina Schist is unlikely to be
Moore and Saffer, 2001]; near-complete loss of representative of all subducted sediments and
pore and structural waters is presumed to occur basalts, its pattern of progressive water loss pro-
within a few additional kilometers of burial. It vides a first-order indication of devolatilization at
appears doubtful, however, that any significant moderate depths. By 15 km depth, a water
net escape of fluids from oceanic crust occurs at content of 5% by weight for sediments, for a global
these shallow depths, because the framework sediment subduction of 1.7  1015 g/yr, implies
strength of basalts prevents compaction. This does 8.4  1013 g/yr of water subduction, only 9% of
not preclude fluid flow within the basalts, possibly the pore and structural water within initially sub-
including flushing of water whose chlorinity has ducted sediments. By 40 km depth, reduction of
been lowered by saponite breakdown, driven by this water content to 1.3% expels all but 2% of
the buoyancy force of warmer waters flowing the initial water. Assuming that the Catalina Schist
updip from deeper sources. Vein and fabric studies metabasalts are representative of the extrusive
of outcropping forearc sedimentary rocks indicate portion of oceanic crust, 5% water content at
an evolution of fluid flow style during subduction: 15 km depth implies 2.1  1014 g/yr of water,
initial mud-filled veins, then cracks with calcite fill, or 72% of the originally subducted upper crustal
then scaly fabric [Fisher, 1996]. Updip migration waters, and 1.3% at 40 km depth implies a
of fluids may occur mainly within the high-perme- residual water content there of 5.6  1013 g/yr,
ability scaly fabric of the upper portion of under- or 19% of original subduction volumes (Figure 7).
thrust sediments, at least for burial depths of By 15 km depth, 50% of the water that entered
<15 km [Fisher, 1996]. the subduction zone has been expelled, mainly
updip toward the prism toe. By 40 km, 60%
[93] Studies of the Catalina Schist showed a slab- has been expelled. Inclusion of water expulsion by
parallel melange fabric, suggesting that most fluid metamorphic reactions within the dikes and gab-
26 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

bros presumably would raise this total substan- [98] Based on a variety of H2O+ studies of arc
tially. For example, similar amphibolite-grade magmas, Ito et al. [1983] calculated that the water
metamorphism of the dikes, reducing their water flux at arcs is 1  1014 g/yr; major uncertainties
content from 2.8 to 1.3%, would expel 1.5  1014 in magma volumes result in a possible range of
g/yr, or an additional 8% of originally subducted 0.95–2.0  1014 g/yr. Peacock [1990] provided a
water. similar estimate, 1.4  1014 g/yr, without mention-
ing data sources. A more appropriate measure of
[95] Magnetotelluric surveys have detected high- the volatile content of magmas may be fluid
conductivity zones attributed to free water in or inclusions in olivine phenocrysts, which average
near the top of the subducting slab at several 2.5% [Sobolev and Chaussidon, 1996]; in conjunc-
subduction zones: at 25–50 km depth in Juan de tion with an arc magmatism rate of 2.9–8.6 km3/yr
Fuca [Wannamaker et al., 1989], at roughly 20– [Crisp, 1984], resulting water flux is 2–6 times
60 km beneath the Okinawa Trough forearc and that of Ito et al. [1983] or Peacock [1990]: 2–6 
somewhere within the interval 60–130 km depth 1014 g/yr. Of the 6  1014 g/yr of subducting
beneath the arc [Shimakawa and Honkura, 1991], water that appears to reach 45 km depth, at least a
extending to >60 km depth in Izu-Bonin [Toh, third and perhaps all is later released to the mantle
1993], and from 13 km to 30 km depth in Mexico wedge and reaches the arc. The remainder may be
[Arzate et al., 1995]. The excellent resolution of released into other parts of the mantle wedge, enter
the Juan de Fuca resistivity model detected an arc magmas but degas during volcanism, or be
order-of-magnitude conductivity decrease from retained into the deep mantle. I emphasize, how-
the trench to 60 km inland, due to pore water ever, that the cumulative errors in this mass balance
loss, followed by a sudden increase in conductiv- are huge and impossible to quantify. For example,
ity at 25 km depth, interpreted as dehydration of Reymer and Schubert [1984] calculated an arc
greenschist-facies minerals [Wannamaker et al., magmatism rate of 1.1 km3/yr (1.3 km3/yr with a
1989]. more accurate total trench length) from seismic
[96] Updip reflux persists at least to 15 km and profiles across arcs that is only 13–38% of Crisp’s
possibly to 45 km, based on the Catalina Schist [1984] rate from active arc volcanism; I use the
melange fabric [Bebout, 1991, 1995]. The Juan de latter. The unknown amount of water loss to the
Fuca conductivity increase at 25 km [Wannamaker deeper mantle makes estimation of the mantle
et al., 1989], however, does not suggest reflux. water budget problematic [Bell and Rossman,
Whether subducting crust below 45 km depth 1992; Thompson, 1992; Williams and Henley,
retains sufficient fracture permeability for volatile 2001].
escape is unknown. Dehydration may generate its
own permeability, by inducing hydrofracture [Fyfe, [99] Many authors have interpreted phase diagrams
1997]. A magnetotelluric survey of the Vancouver based on experimental petrology to infer the cause
Island portion of Cascadia subduction zone of the fluid release responsible for arc magmatism
detected porosities of 0.5–4.0% (assuming sea- [e.g., Peacock, 1990, 1991, 1993, 1996; Pawley
water salinity) well above the slab, attributed to and Holloway, 1993; Pawley, 1994; Poli and
fluid rise from the slab to an impermeable barrier Schmidt, 1995; Liu et al., 1996; Iwamori, 1998,
[Hyndman, 1988]. 2001; Ono, 1998; Ernst, 1999]. A single volatile-
expulsion event is not expected. A suite of hydrous
[97] When the slab reaches 80 –120 km depth, minerals (lawsonite, chlorite, amphibole, epidote/
large-scale release of volatiles into the overlying zoisite, chloritoid, and others) is expected to break
mantle wedge generates the partial melting that down in a series of overlapping depth zones
results in arc volcanism [Gill, 1981]. Except per- [Schmidt and Poli, 1998; Kerrick and Connolly,
haps for very young slabs, the crust undergoes little 2001]. Systematic lateral variations in water-solu-
partial melting [Defant and Drummond, 1990; ble elements in magmas from forearc to arc suggest
Peacock, 1990; Peacock et al., 1994]. a progressive decrease in water input from the slab
27 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

[Ryan et al., 1996], possibly a change from Sub-arc slab depths are readily computed from slab
amphibole breakdown near the trench to phlogo- intermediate dip (0–100 km) and arc-trench gap,
pite decomposition in the backarc [Tatsumi and both tabulated by Jarrard [1986] for 26 of the
Kogiso, 1997]. subduction zones of Table 2. All but two sub-arc
depths are between 80 and 120 km. In modern
[100] Even if a single volatile-release event were subduction zones, subduction rates vary by a factor
likely within each slab, it should not occur at of 30, and average ages of the slab vary from 5 Ma
similar depths in subduction zones with widely to 145 Ma, yet no correlation is found between
varying slab ages and subduction rates, based on sub-arc depth and subduction rate, slab age, or their
pressure-temperature phase diagrams. Slab temper- product, strongly implying that melting within the
ature at any depth is very sensitive to both sub- mantle wedge is insensitive to temperature of the
duction rate and slab age, as confirmed by thermal adjacent oceanic crust. This observation is consis-
models. For example, for a constant subduction tent with thermal models incorporating convection
rate, the temperature of the top of upper crust at of the mantle wedge, whereas models lacking
100 km depth varies from 540C for 145-Ma crust convection exhibit strong slab-induced cooling of
to 1180C for 5-Ma crust [Peacock, 1990]. the mantle wedge [Anderson et al., 1980]. Nor
Accordingly, the same dehydration reaction that does fluid availability affect the depth of melting:
is expected to release water beneath the arc for sub-arc depth is not correlated with CO2 or any of
intermediate geotherms releases that water beneath the four types of water flux, nor with these fluxes
the forearc for high-temperature geotherms (slow per meter of trench length.
subduction and/or very young crust) and retains
that water to the deep mantle for low-temperature [103] Predictions of multiple volatile expulsions and
geotherms (rapid subduction and/or old crust) thermally dependent expulsion imply large-scale
[Kerrick and Connolly, 2001]. Early release of transport prior to release beneath the arc. Even with
volatiles from young, hot slabs may affect compo- available water to lower melting temperature, par-
sitions of later, deeper arc magmas, based on a tial melting of the mantle wedge cannot occur until
possible association between slab age and magma a temperature of at least 1000C is reached. Water
composition in only two subduction zones [Green released between 50 km and 100 km may remain
and Harry, 1999; Harry and Green, 1999]. In within the slab or migrate into the adjacent mantle
contrast, ultrafast subduction, such as occurs today that is dragged downward (Figure 7). In either case,
in Tonga, favors retention of some water to the the upper part of the slab and mantle wedge have
deep mantle, possibly generating chemical anoma- insufficient permeability to permit large-scale
lies in the mantle [Staudigel and King, 1992]. escape of volatiles, until partial melting of the
mantle wedge occurs, with associated volatile
[101] Volatile release within the subducted slab advection. Of the volatiles that are released after
may occur earlier within the extrusives than in much deeper subduction, some may trigger slab or
the dikes and gabbros, not only because of more adjacent mantle melting and reflux along the slab
loosely held water in less stable minerals, but also top to the arc-magma generation zone, and some
because of larger-scale metamorphic phenomena. may be lost to the deep mantle.
Hacker [1996] concluded that conversion of the
extrusives to eclogite is complete by 250C, [104] Because partial melting in the mantle wedge
whereas eclogite conversion of the gabbros usually is triggered by volatile release, a correlation
is complete by 550C and may continue to between rate of arc volcanism and amount of
>800C. deeply subducted water may exist. Early explora-
tions of this hypothesis, using convergence rate as
[102] In contrast to the wide variability of depths a proxy for subducted water, found no correlation
for volatile release, both within and among slabs, between convergence rate and either arc eruption
volatile escape culminating in generation of arc rate [Gill, 1981] or seismic-based arc growth rate
magma is generally confined to 100 km depth. [Reymer and Schubert, 1984]. Subducted water per
28 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

unit arc length (Table 2) is well correlated (R = however, that a more diagnostic test would con-
0.72) with arc growth rate if arc ages of Jarrard sider individual subduction zones rather than
[1986] are used to calculate the latter, but inversely trench groups. For 23 subduction zones, incoming
correlated (R = 0.57) if arc ages from Reymer contents of bound and pore water of Table 2 can
and Schubert [1984] are used; in both cases, 1–2 be compared to maximum earthquake magnitudes
extreme points dominate the correlation. A com- tabulated by McCaffrey [1997]. Water content is
prehensive analysis of these hypotheses, using used rather than water flux, to remove the effects
water fluxes from Table 2, may be warranted but of trench length and convergence rate. This com-
is beyond the scope of this paper. parison confirms Rea and Ruff’s [1996] conclu-
sion that pore water content does not affect
4.1.3. Slab Devolatilization earthquake magnitude, and it extends this conclu-
and Benioff Zone Seismicity sion to include sediment bound water and total
[105] Devolatilization within the slab has been water from sediments and basalts. A weak corre-
suggested as a mechanism for several aspects of lation (R2 = 0.28  0.30) is found, but it is almost
Benioff zone seismicity. Metamorphic reactions entirely attributable to a single subduction zone,
and associated fluid release could control the updip Andaman, which has the highest content of
and downdip limits of the large earthquakes gen- incoming water and lowest earthquake magnitude.
erated at the contact between underthrusting and Convergence rate, rather than fluids, is the dom-
overriding plates [Hyndman et al., 1997; Kasahara inant control on earthquake magnitude, either
et al., 2001]. Alternatively, the updip edge of because of coupling [Ruff and Kanamori, 1980;
interplate seismicity may be attributable to a suite Jarrard, 1986] or recurrence time [McCaffrey,
of diagenetic changes that increase rock strength 1997]. Slab age may [Ruff and Kanamori, 1980;
[Moore and Saffer, 2001], and the downdip edge Jarrard, 1986] or may not [McCaffrey, 1997] be a
may be compositionally determined, occurring at secondary influence. Subducting fluids are not
the overriding plate’s Moho [Ruff and Tichelaar, even a second-order influence, based on compar-
1996]. Dehydration is likely to elevate pore pres- isons to earthquake magnitude residuals (observed
sures and lower effective stress, thereby inducing minus predicted based on regression on conver-
earthquakes via hydrofracture. This dehydration gence rate).
embrittlement may be responsible for intermedi-
ate-depth earthquakes in subducting crust during 4.2. Chlorine Flux
transformation to eclogite [Kirby et al., 1996] and
in the lower plane of double seismic zones during [107] Assuming a typical seawater chlorinity of
breakdown of serpentine to forsterite and enstatite 19% (34.3% salinity) for subducted pore waters,
[Peacock, 2001; Omori et al., 2001]. Serpentine 2.08  1013 g/yr of chlorine enters the world’s
dehydration may also be the cause of deep-focus subduction zones via pores in sediments and crust.
earthquakes [Raleigh and Paterson, 1965; Meade Amounts of chlorine in subducting minerals are
and Jeanloz, 1991]. In all three cases, the wide less reliably known but certainly much smaller. Ito
depth distribution without a localized seismicity et al. [1983] calculated that 2.9 ± 1.5  1012 g/yr
maximum is surprising. of chlorine is subducted as oceanic crustal miner-
als, based on measured chlorine contents of the
[106] Slab devolatilization might also affect mag- hydrous minerals and approximate mineral abun-
nitude of interplate earthquakes: increased fluid dances; updated subduction rates would change
pressure might decrease coupling between the this value to 2.5 ± 1.3  1012 g/yr. For chlorine
plates, leading to smaller earthquakes. Rea and within subducted sedimentary minerals, the geo-
Ruff [1996] tested this hypothesis by comparing synclinal average of 1200 ppm chlorine [Ronov
maximum earthquake magnitude to incoming and Yaroshevskiy, 1976] for a sedimentary subduc-
sedimentary pore water for their 12 trench groups, tion rate of 1.68  1015 g/yr provides a very rough
and they found no correlation. They suggested, estimate of 2  1012 g/yr. Total chlorine entering
29 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Figure 8. Flowchart of global chloride fluxes. See text for sources. All arc volcanism is shown as occurring on
continents, but some occurs on oceanic crust that may later become continental crust.

global subduction zones is therefore 2.5  1013 Direct measurement of the cyclic salt flux is not
g/yr. possible. Between 28% [Meybeck, 1983] and 82%
[Berner and Berner, 1987] of the pollution-cor-
[108] Figure 8 presents a flowchart of global chlor-
rected chloride flux in rivers is attributable to
ide fluxes. Although mantle degassing has trans-
dissolution of evaporites, and the rest is from cyclic
ferred about 40% of Earth’s chlorine to oceans and
salt. Consequently, cyclic sea salt transport of
crust [Magenheim et al., 1995], the global chloride
chloride from sea to land is 4–16  1013 g/yr.
budget is probably near long-term mass balance,
because salinity of the ocean has been relatively [110] The mantle loses chloride to MORB and arc
stable throughout at least the last 1.5 b.y. The volcanism [Schilling et al., 1978] and to reflux, and
oceanic chloride reservoir is so huge (2.7  it gains chloride from subduction. Using an aver-
1022 g), however, that even massive temporary age chlorine content of 48 ppm for MORB glasses,
imbalances have a negligible impact. The ocean Ito et al. [1983] calculated a chloride supply rate of
loses chloride to cyclic sea salt and pores, and this 2.7  1012 g/yr at spreading centers. Their estima-
loss is likely to be balanced by dissolved river influx tion of chloride flux at arc volcanoes is subject to a
and water reflux from the shallow subducted slab. factor of 20 uncertainty associated with magma
generation rates [Ito et al., 1983]. Using an average
[109] Chloride fluxes between the ocean and con-
chlorine content for arc magmas of 900 ± 500 ppm
tinents are 1–2 orders of magnitude larger than
and the upper range of magma generation estimates
other chloride fluxes. Livingstone [1963] con-
(5–10  1015 g/yr), Ito et al. [1983] computed a
cluded that river influx of dissolved materials to
chloride supply rate of 4.3–9.5  1012 g/yr. The
the sea is 2.5–4  1015 g/yr, of which 6.5% is
MORB Cl flux is not changed by newer geo-
chloride, so the chloride influx is 1.6–2.6  1014
chemical data [Jambon et al., 1995], nor by revised
g/yr. Meybeck [1979] noted that pollution is often a
seafloor spreading rates and crustal thicknesses of
major source of chloride in rivers, and he evaluated
this study. Revised arc magma generation (2.9–8.6
the natural river flux of chloride as 2.2  1014 g/yr.
km3/yr) [Crisp, 1984] changes the arc Cl flux to
Much of this chloride, however, is ‘‘cyclic salt’’,
7–22  1012 g/yr.
sea salt captured by the atmosphere and carried
inland, where it is either deposited or washed out [111] Reflux at and near the toes of accretionary
of the atmosphere by rain or snow [Clarke, 1924]. prisms expels half of total subducted water prior to
30 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

slab subduction to 15 km, and reflux may expel the ated with both crustal alteration and subduction
two thirds of total subducted water released by 45 recycling [Magenheim et al., 1995].
km (section 4.1.2). Associated chloride flux for the
latter consists of 1.4  1013 g/yr from sediment
4.3. CO2 Flux
pores and may include up to 0.6  1013 g/yr from
basalts. The average chlorinity of reflux water is [114] Global CO2 subduction totals 1.53  1014 g/
therefore 11–16%, much less than the normal yr. Two thirds of this amount is within the oceanic
chlorinity of 19% for seawater and pore fluids. crust, particularly the upper portion of the extru-
Chlorinity anomalies appear to be a common sives, and the remainder is in accreted sediments
feature of décollements [Kastner et al., 1991]. (Figure 9). Nearly all of this CO2 is in the form of
For example, chlorinity of the Nankai décollement carbonate, primarily calcium carbonate. CO2 con-
is 15%, less than can be accounted for by tent of most subducting igneous crust varies within
smectite dehydration of subducting sediments and about a factor of two. In contrast, CO2 contents of
therefore implying an additional, deeper source of subducting sediments are much more variable, with
low-chlorinity water [Kastner et al., 1993; Under- negligible carbonate and CO2 in half of the sub-
wood et al., 1993]. The Nankai chlorinity anomaly ducting sediment sections (Table 2). Variations in
has been reproduced with a coupled flow and CO2 subduction fluxes among individual subduc-
solute-transport model that incorporates transient tion zones (Figure 9) are most sensitive to con-
flow within the sediments [Saffer and Bekins, vergence rate, trench length, and amount of
1998]. The water fluxes of Table 2 suggest that subducting sedimentary carbonate.
inclusion of smectite breakdown in the oceanic
crust may be appropriate for reproducing the Nan- 4.3.1. Comparison to Prior Flux Estimates
kai chlorinity anomaly; structural water in the [115] My 5.26  1013 g/yr for CO2 in subducting
Nankai upper crust is 50% higher than that in the sediments is based on very minor revisions of the
subducted sediments. comprehensive evaluation of Plank and Langmuir
[1998]. These results supercede that of Peacock
[112] If the interpreted patterns of slab dewatering [1990], whose 1.6  1014 g/cc was based on a
and associated reflux (Figure 7) are valid, then single hypothetical subducting sediment column
fluid inclusion studies of eclogites are predicted to that lacked hemipelagics and turbidites and was
exhibit the following chlorinities: 19% (sea- 72% carbonate; in contrast, Plank and Langmuir
water) for metasediments with possible smectite/ [1998] found that average subducting sediment is
illite transformation followed by compaction flush- only 7% carbonate.
ing, 6–13% (depending on crustal age) for meta-
[116] Crustal CO2 subduction is calculated to be
basalts with partial expulsion of pore and structural
1.00  1014 g/yr, within the wide range of previous
water, and near-zero for metabasalts with extensive
estimates: 1.63  1014 [Staudigel et al., 1989],
flushing from downdip portions of the slab. Meta-
0.6  1014 [Peacock, 1990], 1.8  1014 [Bebout,
morphic rocks from subduction zones (but not
1995, 1996], 1.18  1014 [Staudigel et al., 1996],
continental collisions) are consistently low-chlor-
and 1.48  1014 [Alt and Teagle, 1999]. These
inity [Bebout, 1996], generally 9–12% [Philippot
previous studies are based on estimating the CO2
et al., 1998].
content of typical old subducting crust. In contrast,
flux calculations for the extrusives based on very
[113] The mantle chloride budget may be close to
young 6-Ma crust [Alt et al., 1996a] and 3.5 Ma
mass balance currently (Figure 8), but uncertainties
hydrothermal fluids [Sansone et al., 1998] were an
are large: total loss to arc and MORB volcanism
order of magnitude lower – 0.7–1.6  1013 and
and reflux is 2.4–4.5  1013 g/yr, and total gain
0.4–1.1  1013, respectively.
from subduction is 2.5  1013 g/yr. In the near
future, chlorine isotopic studies may be able to [117] Peacock [1990] assumed 0.1% CO2 for
refine our understanding of chloride fluxes associ- both basalts and gabbros, based on consideration
31 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Figure 9. Carbon dioxide fluxes at global subduction zones, from Table 2.

of whole-rock analyses from three upper-crustal value for the extrusive section and near-zero at
sites and the gabbros of Hole 735B and Oman, 1100 m subbasement, and Staudigel et al. [1996]
whereas Bebout [1995, 1996] used 0.3% for the confined their flux assessment to the extrusive
entire oceanic crust, based on data of Staudigel section.
et al. [1989]. Staudigel et al. [1989, 1996]
analyzed Holes 417A/417D/418A, and Alt and [118] My CO2 flux calculation for oceanic crust is
Teagle [1999] analyzed Sites 843 and 801. Both based largely on the excellent and comprehensive
studies were carefully weighted to obtain a rep- analysis of Alt and Teagle [1999]. Indeed, I adopt
resentative sample (see section 2.6 above) for the their CO2 values for the dikes and gabbros and for
extrusives. Alt and Teagle [1999] extended this 4 of the 5 upper extrusive sections of Figure 2d.
approach to encompass the dikes of Hole 504B Rather than assume a uniform CO2 content for all
and gabbros at three ODP localities. In contrast, altered crust, however, I use the tentative age-
Staudigel et al. [1989] interpolated between their dependent CO2 enrichment pattern of Figure 2d
32 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

to compute upper crustal CO2 for each subduction comparing these total CO2 atmosphere/ocean sour-
zone. The resulting total flux is only 70% of that of ces (3.5 ± 1.2  1014 g/yr) to sinks (5.6 ± 1.2 
Alt and Teagle [1999], mainly because I use a total 1014 g/yr). Additional possible sources, such as
subduction rate that is only 71% of theirs (see abiogenic methane and non-volcanic CO2 degass-
section 3.1 discussion of revised subduction rates). ing, are discussed by Kerrick [2001]. The close
coupling between CO2, climate, and silicate weath-
4.3.2. CO2 Flux Budget ering may prevent major long-term imbalance
between atmospheric CO2 sources and sinks
[119] Consideration of the complete global CO2 [Berner and Caldeira, 2002].
flux budget (including mantle, continental crust,
oceanic crust, oceans, and atmosphere) is beyond [122] Subduction of sediments and oceanic crust
the scope of this paper. Berner et al. [1983] and removes 1.53  1014 g/yr of CO2 from the
François and Walker [1992], among others, pre- ocean/atmosphere system. In contrast to the initial
sented CO2 flux budgets and models. My focus, in subduction fluxes of water and chloride, most of
contrast, is on mantle and oceanic crust CO2 which are lost to reflux, a negligible portion of
budgets. subducted CO2 is lost to reflux. The CO2 content
of reflux waters is likely to be close to saturation,
[120] The dominant CO2 additions to the ocean/ or 0.02% for seawater at 0C, so a water reflux of
atmosphere system from crust and mantle sources two-thirds initial subduction volume entrains 2.9
are magma degassing at spreading centers and arc  1011 g/yr of dissolved CO2, or only about 0.2%
volcanoes, organic carbon breakdown, and possi- of subducting CO2.
bly metamorphic breakdown of calcite to silicate.
The dominant removal mechanisms are subaerial [123] From the perspective of CO2 mass balance
silicate weathering, which generates bicarbonate for the mantle, the subducting flux of CO2 exceeds
and culminates in calcite precipitation, carbonate the individual supplies from ridge volcanism and
uptake by oceanic crust, and organic carbon arc volcanism, but it is much less than the total
burial. degassing rate of 2.6  1014 g/yr (range 1.8–4.4
 1014) [Marty and Tolstikhin, 1998]. Some
[121] Degassing at spreading centers contributes crustal CO2 reaches the deep mantle, based on
2.8–5.5  1013 g/yr [Gerlach, 1989] or 5.7– the excess of CO2 subduction over arc magma-
13.6  1013 g/yr [Marty and Tolstikhin, 1998] of tism, the observation that CO2 appears to be
CO2; these values should probably be reduced mostly retained during metamorphism at depths
10% to account for lower spreading rates for of 15–45 km [Bebout, 1995], and the relatively
NUVEL-1A than for earlier plate motion models. large amount of CO2 still present in the mantle
Degassing of arc volcanoes contributes another [Zhang and Zindler, 1993]. Phase equilibria sug-
11 ± 3  1013 g/yr [Zhang and Zindler, 1993; gest only minor decarbonation at and above subarc
Marty and Tolstikhin, 1998] or 8–11  1013 g/yr depths [Kerrick and Connolly, 2001], but dehy-
[Kerrick, 2001], and plumes may add a compara- dration enhances decarbonation and vice versa
ble amount [Marty and Tolstikhin, 1998]. Subaerial [Peacock, 1990; Kerrick and Connolly, 2001].
weathering removes 2  1014 g/yr of CO2 from Kerrick and Connolly [1998] suggested that the
the atmosphere [Berner, 1990]. Burial of organic CO2 in arc magmas is derived mainly from sub-
carbon removes 2.2 ± 0.9  1014 g/yr, but 9 ± 4  ducted sediments and extrusives, whereas CO2
10 13 g/yr is released by thermal breakdown within subducting mantle serpentinites (not con-
[Berner, 1990]. Carbonate uptake by oceanic crust sidered in my flux calculations) is released in the
may approximately equal its current subduction deep mantle.
rate, 1.0  1014 g/yr, or it may be much less (see
sections 2.6 and 4.5). The ocean/atmosphere sys- [124] The modern rate of crustal CO2 subduction is
tem must be close to steady state balance with more significant than that of sedimentary CO2 for
crust and mantle, but that is not apparent from several reasons. First, total crustal CO2 subduction
33 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Figure 10. Potassium fluxes at global subduction zones, from Table 2.

is double that from sediments. Second, all subduct- carbon deposition [Dymond and Lyle, 1994] are
ing crust contains significant CO2 whose ultimate on continental shelves and are therefore isolated
release affects melting in the mantle wedge, from mantle recycling for tens to hundreds of
whereas most sedimentary CO2 is in a few sub- millions of years.
duction zones. Third, the modern rate of crustal
CO2 consumption is a robust indicator of long-term
removal from the ocean/atmosphere system. In
4.4. K2O Flux
contrast, present-day sediment subduction is less [125] Sediments and igneous crust contribute about
relevant to ocean/atmosphere CO2 change than is equal amounts to total subduction flux of K2O
carbon removal as deposition of calcium carbonate (Figure 10): 3.62  1013 g/yr and 3.81  1013
and organic matter. A significant portion of carbo- g/yr, respectively. Variations in K2O fluxes among
nate deposition [Hay, 1994] and most organic individual subduction zones (Figure 10) are most
34 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

K2O fluxes (x 10 12 g/yr)


air transfer
1-4
dissolved
continents in rivers 60 oceans
suspended
in rivers
sed.
370 diagenesis
arc low-T 5-7
volcanism 52 alteration high-T
400-1200 12 alteration
sediments
36 subduction

MORB 37 oceanic
mantle volcanism
subduction crust
38

Figure 11. Flowchart of global potassium fluxes. All arc volcanism is shown as occurring on continents, but some
occurs on oceanic crust that may later become continental crust. Most suspended sediments transported to oceans are
deposited on oceanic crust as shown, but some are deposited on continental margins.

sensitive to convergence rate, trench length, and and K2O loss to oceanic crust during low-temper-
thickness of subducting terrigenous sediments. ature alteration. The mantle potassium budget,
which is related to mantle thermal regime and
4.4.1. Comparison to Prior Flux Estimates 40
Ar [Sleep, 1979; Sleep and Wolery, 1979; Joc-
[126] Subduction recycling of K2O was first quan- hum et al., 1983; Azbel and Tolstikhin, 1990],
tified roughly by Sleep and Wolery [1979], who consists of loss to magmas during MORB, arc,
suggested that crustal alteration absorbs 1–2  and hot spot volcanism and plutonism, and gain
1013 g/yr of K2O and that subducting red clays during subduction of sediments and igneous oce-
raise the total to 3  1013 g/yr. These values are anic crust. Neither budget need be in short-term
less than half of mine. The estimated K2O flux of or long-term balance. In particular, erosion
3.62  1013 g/yr for subducting sediments is roughly balances sediment subduction only on a
based on minor revisions to the comprehensive time scale of tens to hundreds of millions of
evaluation of Plank and Langmuir [1998]. Based years.
on geochemical comparison of Holes 417A/417D/ [128] The oceanic K2O budget (Figure 11) is domi-
418A to unaltered crust, global K2O flux into the nated by eroded K2O that is transported by rivers to
extrusives was estimated as 2.2  1013 g/yr the oceans as suspended sediment load and bed-
[Spivack and Staudigel, 1994] and 1.95  1013 load, which total 3.7  1014 g/yr of K2O [Milliman
g/yr [Staudigel et al., 1996], similar to my 1.58  and Meade, 1983]. Nearly all of this sediment is
1013 g/yr for this zone. K2O fluxes have also delivered to oceanic crust; only a few percent of
been calculated for ophiolites [Lecuyer et al., Quaternary terrigenous marine sediments are
1990]. deposited on continental shelves [Hay, 1994].
Annual river influx of dissolved K+ is also signifi-
4.4.2. K2O Flux Budget cant: based on a total dissolved load of 2.5–4 
[127] The oceanic potassium budget (Figure 11) 1015 g/yr, of which 1.9% is K+ [Livingstone, 1963],
consists of supply from rivers (both dissolved and potassium transport (expressed as K2O) is 6–9 
suspended load), sediment loss by deposition, 1013 g/yr; Meybeck [1979] estimated this influx as
diagenesis of seafloor sediments, K2O gain from 6  1013 g/yr. Residence time of K2O in seawater
oceanic crust during high-temperature alteration, (1.4  1024 g @ 0.38% K+) is therefore 8–13 m.y.
35 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

A small portion of this dissolved potassium comes Party, 1992], is representative of dikes in general,
from airborne cyclic sea salt: 1–4  1012 g/yr but leaching is even less if initial K2O is less. The
(expressed as K2O) based on air transfer of chlor- 2.15  1013 g/yr K2O in subducting gabbros
ide and the relative proportions of chloride and appears to be almost entirely primary; average
potassium in seawater. Potassium is absorbed from K2O at Hole 735B is 0.05% for fresh rocks [Bach
seawater by early diagenesis of aluminosilicates et al., 2001] and 0.06% for all rocks (section 3.7).
(zeolites and clays) at and very near the seafloor. Thus, only 13% of subducted crustal K2O is
Pore water K+ profiles indicate a global K+ flux of attributable to addition from seawater; the rest
1.1  1018 mEq/yr (5.2  1013 g K2O/yr) [Martin (3.31  1013 g/yr) is inherited from MORB mag-
and Sayles, 1994]. Compared to most other oce- mas. Because more than half of the total crustal
anic potassium fluxes, the net removal of 5–7  K2O subducted is in the gabbros, any alteration-
1012 g/yr from high- and low-temperature crustal induced change in gabbro K2O could easily surpass
alteration is comparatively small. For example, I those from the extrusives and dikes.
conclude that only 20% of the river input of
potassium is absorbed during low-temperature [131] The K2O budget for igneous oceanic crust
alteration of oceanic crust (Figure 11), in contrast (excluding sediments) almost balances, with 4.9 
to an earlier 50% estimate [Spivack and Staudigel, 1013 g/yr input and 4.3–4.5  1013 g/yr output
1994]. Calculated total input of dissolved K2O into (Figure 11). The slight imbalance is because the
the ocean (6.7  1013 g/yr; Figure 11) is surpris- rate of crustal generation is 10% higher than that
ingly close to total output (6.6  1013 g/yr), of subduction: 10% of convergence is continent/
considering the 10% errors in both. continent collision, so total K2O flux from mantle
to oceanic crust at spreading centers is 3.64  1013
[129] The subduction flux of crustal K2O includes g/yr (= 1.1  3.31  1013 g/yr).
both inherited magmatic K2 O and alteration-
induced changes. Earlier discussions of crustal [132] The largest uncertainty in global K2O fluxes
K2O fluxes concentrated on addition of K2O to is in the dominant component, K2O flux from the
the extrusive section during off-axis low-temper- mantle to arc magmatism, mainly because of
ature alteration. This process was recognized long uncertainties in the ratio of observed arc volcanism
ago to be one of the dominant geochemical sig- to hidden arc plutonism. Based on 0.4–0.6 km3/yr
natures of low-temperature alteration [Hart et al., of arc volcanism and 2.5–8 km3/yr of arc pluton-
1974; Donnelly et al., 1979b]. In contrast, high- ism [Crisp, 1984], a density of 2.7 g/cm3, and
temperature alteration of the dikes at and near the average K2O of 3.8% for silicic igneous rock and
ridge axis removes K2O and is more deeply pen- 5.5% for calc-alkalic granite [Nockolds, 1954], the
etrating [Hart and Staudigel, 1982; Elderfield and total K2O flux due to arc magmatism is 4–12 
Schultz, 1996]. 1014 g/yr. This flux to continents is comparable to
the total K2O flux from continents to oceans of
[130] If the initial K2O content of basalts is 0.085% 4.3–4.7  1014 g/yr, but it is more than an order of
(section 2.8 and Figure 3), then only 22% of the magnitude higher than the MORB crustal K2O flux
1.58  1013 g/yr extrusive K2O flux to subduction and about an order of magnitude higher than the
zones is primary. Both primary and alteration K2O combined subduction flux of 7.4  1013 g/yr from
fluxes for dikes are poorly known. Near-ridge sediments and oceanic crust.
high-temperature leaching apparently reduces the
primary K2O dike flux (8.3  1012 g/yr) by 65– [133] K2O fluxes from subduction and arc magma-
80%, to 1.4–2.9  1012 g/yr. The higher estimate tism need not balance: most of the latter is thought
assumes reduction from a typical MORB magmatic to come from low-degree partial melting of the
content of 0.085% K2O to the present Hole 504B mantle wedge rather than the slab [e.g., Ionov et
content of 0.014%. The lower estimate assumes al., 1997], as required by the flux imbalance of
that the 65% K2O leaching of Hole 504B dikes, Figure 11. Arc K2O is accompanied by long-term
from an initial K2O of 0.04% [Shipboard Scientific mantle 40Ar, and helium similarity of arc and
36 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

MORB volcanism suggests similar source regions Wilson cycles. Alternatively, long-term growth of
[Sleep and Wolery, 1979]. In contrast, the system- continents may result; Reymer and Schubert [1984]
atic increase in arc magma K2O with increasing used volumes and ages of arcs to estimate that this
slab depth [Dickinson and Hatherton, 1967; Dick- growth averages 4  1015 g/yr.
inson, 1975] suggests a slab contribution. K2O is
scoured from the slab and released through solu- 4.5. Are Present Fluxes Representative
tion in fluids, and amount of K2O release may of Prior Ones?
depend more on dehydration history than on initial
[136] Modern subduction fluxes of H2O, CO2,
concentrations of K2O or hydrous minerals [Lee-
K2O, and Cl are probably quite different from
man, 1996]. Metabasalts from <45 km burial depth
those that characterized early earth history. Differ-
retain their pre-subduction K2O concentrations
ent modeling approaches agree that most volatile
[Bebout, 1995]. Most of the hydrous phases that
release from the mantle occurred quite early
break down within the slab at depths of <110 km
[McGovern and Schubert, 1989; Franck and Bou-
are potassium-free. In contrast, phengite (a K-rich
nama, 1997; Harrison, 1999]. Continental growth
mica) could be the dominant K-rich hydrous phase
was probably also mainly Archean [e.g., Harrison,
to break down at depths of >100 km and enrich arc
1999], but rate of Phanerozoic growth is debated
magmas [Schmidt, 1996].
[Reymer and Schubert, 1984]. Thermal modeling
[134] Subducted water has been suggested to affect suggests that Precambrian spreading rates were no
arc magma chemistry, based on comparison of arc more than twice modern ones [Sleep, 1979]. Rapid
K2O contents to convergence rates [Sugisaki, Archean spreading implies generally younger
1976], but confounding variables such as continen- slabs, which may have dehydrated at a shallower
tal contamination and degree of partial melting depth than modern ones [McCulloch, 1993], but
were not considered. Plank and Langmuir [1993] rapid subduction also implies retarded slab heating,
minimized these confounding effects by consider- which leads to deeper devolatilization [Staudigel
ing only primitive arc basalts and normalizing arc and King, 1992; Kerrick and Connolly, 2001]; the
potassium contents to sodium. For eight analyzed latter is likely to dominate. Archean ophiolites of
subduction zones, they found a very good correla- South Africa and Australia appear to have under-
tion (R2 = 0.739) between arc potassium content gone hydrothermal histories quite similar to mod-
and the amount of potassium subducted in sedi- ern oceanic crust [DeWit and Hart, 1993].
ments. This correlation is just as good with the Phanerozoic variations in seafloor spreading rate
parameters of Table 2 (R2 = 0.763), but it is [Gaffin, 1987] have associated changes in MORB
significantly degraded if crustal K2O sources are degassing, subduction fluxes, and possibly arc
included: R2 = 0.624 including upper extrusives, degassing [François and Walker, 1992].
and R2 = 0.044 for total subducted K2O. This
[137] Modern subduction fluxes of H2O, CO2,
pattern suggests that K2O recycling to arc magmas
K2O, and Cl are sensitive to the volume of
is largely confined to subducted sediments.
subducting sediments, which changes dramatically
[135] The present-day imbalance between K2O as plate geometries evolve. Rea and Ruff [1996]
subduction (7.5  1013 g/yr) and arc magmatism discussed this problem in the context of the order-
(4–12  1014 g/yr) is expected, based on the of-magnitude discrepancy between current terrige-
overall imbalance between 1.60  1015 g/yr of nous sedimentation and subduction. von Huene
terrigenous sediment subduction [Plank and Lang- and Scholl [1991] appraised the long-term rate of
muir, 1998] and 10–20  1015 g/yr of terrigenous sediment subduction as only about two thirds of the
sediment deposition [Holland, 1981; Milliman and current rate, because sedimentation rates have been
Syvitski, 1994]. Rea and Ruff [1996] explored the unusually rapid during the last 15–30 m.y. [Rea,
implications of the terrigenous imbalance and con- 1993]. These rates have been even more rapid
cluded that mass balance is likely to prevail only during the Quaternary, associated with increased
over the hundreds-of-million-years time scale of glacial activity [Hay, 1994]; this factor affects
37 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

sediment subduction calculations incorporating [140] The ‘‘modern’’ subduction fluxes of this
trench fill [von Huene and Scholl, 1991] much study are actually time-averaged fluxes, represen-
more than those based on distal reference sites [Rea tative of several million years. As mentioned
and Ruff, 1996; Plank and Langmuir, 1998; this previously (section 4.1.2), present-day fluid fluxes
study]. from accretionary prisms, observed at the seafloor
via submersibles or inferred from heat flow, are
[138] Comparisons of subduction fluxes to the
much higher than long-term averages [e.g., Le
dissolved and suspended loads of rivers (e.g.,
Pichon et al., 1991; Kastner et al., 1991]. This
Figures 8 and 11) are severely limited by tempo-
discrepancy is significantly reduced by considering
ral changes in the latter: river fluxes for the
not just shallow sediment sources but also deeper
Recent are nonrepresentative of earlier fluxes
reflux water from both crust and sediments. For
[Hay, 1994; Meybeck, 1994], and both dissolved
Nankai, Peru, and Barbados, whose long-term fluid
and suspended loads have varied by one order of
fluxes were estimated as 7 m3/yr per m along
magnitude in the past [Meybeck, 1994]. For
trench [Kastner et al., 1991], revised fluxes are
example, the suspended-sediment estimate of
17, 24, and 35 m3/yr per m (not including com-
Milliman and Meade [1983], which we used for
paction within each prism), based on Table 2 and
K2O, has been revised upward by 50% [Milliman
the conclusion that two thirds of subducted water is
and Syvitski, 1994], but the flux prior to wide-
refluxed. Transient flow can account for the orig-
spread farming was less than half of this value
inal discrepancies [Moore and Vrolijk, 1992; Le
[Milliman and Syvitski, 1994].
Pichon et al., 1993], with a flux cycle on an
[139] Late Mesozoic and Cenozoic variations in individual trench lasting perhaps 2.5 – 5 Ma
atmospheric CO2 have probably affected the [Saffer and Bekins, 1998]. If most of a several-
observed pattern of age-dependent upper crustal million-year budget of subducted water and CO2 of
CO2 (section 2.6). More rapid seafloor spreading a high-flux subduction zone like Makran were
rates during the Cretaceous [Kominz, 1984; Lar- refluxed within a few-thousand-year period, it
son, 1991] have been causally linked not only to would raise sea level by 1 m and double atmos-
higher sea levels [Hays and Pitman, 1973], but pheric CO2. However, the sea level effect is minor
also to higher atmospheric CO2 levels and con- in comparison to ice-induced sea level change, and
sequently warmer temperatures [Berner et al., CO2 reflux is unlikely to exceed saturation levels.
1983; François and Walker, 1992]. Rate of ridge-
axis CO2 degassing is proportional to spreading
rate, but this effect is counteracted by increased 4.6. Uncertainties
subduction of CO2-rich sediments and crust; the [141] A weakness of the preceding flux analyses is
net effect is flux to the atmosphere/ocean system if the inability to attach confidence limits to com-
generally younger, less altered crust with thinner puted fluxes. Calculation of confidence limits from
sediments is subducted. Faster subduction may regression relations such as Figure 2 is straightfor-
foster increased arc CO2 degassing, but this ward, as is formal uncertainty analysis for means
assumption is hazardous [Kerrick, 2001]. Further- given standard deviation and number of points.
more, calcareous phytoplankton deposition has The problem, however, is that systematic errors are
fluctuated widely [Nakamori, 2001]. Numerical likely to overwhelm random errors in most cases.
modeling of the Phanerozoic carbon cycle, For example, mean microporosity for gabbros of
attempting to consider most of these variables Hole 735B is 0.74%, with 95% confidence limits
and their feedbacks, suggests that the dominant of ±0.29%, but this value is probably strongly
temporal change is either seafloor alteration [Fran- biased by unroofing-induced fracturing; in situ
çois and Walker, 1992] or–more likely–terrestrial microporosities of most gabbros might be much
silicate weathering [Caldeira, 1995]. Alternatively, lower. This section focuses on some of the domi-
individual CO2 enrichment events may have differ- nant uncertainties - particularly biases - that affect
ent sources [Kerrick, 2001]. these flux estimates. An Excel spreadsheet com-
38 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

bining the calculations of Tables 1 and 2 is important than sedimentation in determining the
included as an auxiliary material file, to encourage transition from convective to conductive heat
readers to supercede my calculations by revising loss [Stein and Stein, 1994], but sedimentation
parameters or assumptions. can locally control hydrothermal circulation and
resulting alteration. These analyses exclude areas
4.6.1. Unquantified Variables Affecting such as the Juan de Fuca sedimented ridge
Upper Crustal Alteration where sedimentation undoubtedly dominates, yet
they include Hole 504B which has sufficient
[142] The assumed alteration mineralogy for upper
sediment blanket for more conductive than con-
oceanic crust–saponite, celadonite, and calcite (sec-
vective heat flow [Fisher et al., 1990] despite its
tion 2.8)–is only a first-order approximation. Actual
young age.
alteration mineralogy is invariably more complex,
but this alteration model probably provides a rea- [148] These four sources of local variation in
sonable indication of the age-dependent enrichment alteration are probably the main causes of disper-
of K2O and H2O+ within the upper extrusives. sion among sites in Figure 2, but they do not
[143] The time-dependent alteration of upper oce- introduce systematic bias into these patterns. For-
anic crust, as well as associated changes in geo- tunately, statistically significant age-dependent
physical properties, can be locally overwhelmed by crustal alteration patterns are evident despite this
four other sources of variance. variance: macroporosity reduction and consequent
chloride expulsion (99.9% confidence level), CO2
[144] 1. Initial permeability is sensitive to porosity enrichment (97.5% c.l.), matrix density reduction
and therefore also to volcanic style. Flows are and consequent hydration (99.9% c.l.), and K2O
generally much more massive and impermeable enrichment (based on age versus matrix density
than pillows, so they are less subject to alteration. @99.9% c.l. and matrix density versus K2 O
Consequently, systematic age-dependent changes @99% c.l.).
in log-scale velocity, core-plug velocity, and mac-
roporosity are much stronger for pillows (signifi- [149] Calculated global fluxes are relatively insen-
cant at 95%, 95%, and 99% c.l., respectively) than sitive to the assumed patterns of age-dependent
for flows (none is significant) [Jarrard et al., upper crustal alteration. If age-dependence is
2003]. Volcanic style is the dominant variable ignored and the upper crustal alteration state of
controlling within-site alteration variations, but all subduction zones is assumed to be that of 77-
the ubiquitous preponderance of pillows over flows Ma crust, the global average age [Sprague and
prevents volcanic style from obscuring the age- Pollack, 1980], global fluxes change by 12% for
dependent relationships. CO2, 3% for K2O, 2% for bound water, and 3%
for pore water and chloride. The impact on calcu-
[145] 2. Local basement topography often controls lated fluxes for individual subduction zones is
the geometry of hydrothermal circulation, concen- much higher, ranging from 19% to +28% for
trating outflow at topographic highs [Fisher et al., most fluxes and >80% for CO2 in young subduct-
1990, 1994]. This pattern is evident for paired ing crust.
holes on adjacent topographic highs and saddles,
at 417A/417D/418A and at 504B/896A.
4.6.2. Rejuvenation of Hydrothermal
[146] 3. Spreading rate affects crustal structure (see Circulation
section 3.3) and therefore 3-D permeability struc-
[150] The geochemical signature of age-dependent
ture. A relationship between spreading rate and
crustal alteration, though only roughly known, can
crustal alteration is likely, but none has been
provide an essential starting point for geochemical
detected yet.
mass balances associated with subduction. Implicit
[147] 4. Sedimentation restricts access of seawater to such calculations, however, is the assumption
to crustal porosity. Crustal age is more globally that normal oceanic crust is not modified by the
39 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

subduction process itself. This assumption may be [2001] noted that typical upper mantle velocities of
invalid. 8.2 km/s are consistent with anhydrous peridotite.
Dehydration of serpentinites, which contain 13%
[151] The flexure of lithosphere prior to subduction
H2O+, could be a major source of the H2O in arc
generates an outer rise, with associated normal
magmas [Ulmer and Trommsdorff, 1995; Kerrick
faulting that might rejuvenate hydrothermal circu-
and Connolly, 1998], and subducting ophicarbon-
lation. The two lowest matrix densities on Figure 2b
ates (calcite-rich serpentinites) may transport CO2
are Holes 1149B and 1149D, the only holes that
to the deep mantle [Kerrick and Connolly, 1998].
have sampled basaltic basement from an outer rise
Even 5% serpentinization of the upper mantle
[Shipboard Scientific Party, 2000b]. If this correla-
would constitute about one third of the subducting
tion is not coincidental, analyses of geochemical
water flux [Peacock and Hyndman, 2001].
fluxes at subduction zones should incorporate the
effects of hydrothermal rejuvenation within oceanic 4.6.5. Subduction Erosion
crust immediately seaward of the trench axis. The and Underplating
effects of extensional cracking extend to depths of
at least 25 km [Christensen and Ruff, 1988], much [154] The comparisons of subduction fluxes to arc
deeper than the crust. These cracks are likely to fluxes in this paper neglect the possibility that the
permit fluid percolation into the mantle and induce slab may exchange material with the overriding
serpentinization [Peacock, 2001]. plate through subduction erosion or underplating.
The inverse correlation between shallow slab dip
4.6.3. Properties of Gabbros and arc age (duration of subduction) [Jarrard,
1986] implies substantial subduction erosion, and
[152] The properties assumed for gabbros (section
von Huene and Scholl [1991] estimated that this
3.7 and Table 1) are based on Hole 735B, with
erosional flux may be comparable to the flux of
supporting evidence from Site 894, but both sites
sediment subduction. Both subduction erosion and
have been affected by the unroofing that made
underplating are poorly known [Dewey and Wind-
gabbro drilling possible. For example, I introduced
ley, 1981]; neglecting them, however, biases com-
section 4.6 with the example of gabbro micro-
parisons of subduction fluxes to arc fluxes.
porosity bias. Furthermore, Hole 735B appears to
be nonrepresentative of the geochemically hetero-
geneous lower crust [Dick et al., 2000]. Gabbro
5. Conclusions
uncertainties have a disproportionate impact on [155] High-temperature alteration of the oceanic
mass flux calculations, because 70% of oceanic crust occurs at and very near the spreading center;
crust is gabbro. The impact is greatest for K2O reference Holes 504B and 735B offer perspectives
fluxes: section 3.7 noted that determining gabbro on this alteration for the dikes and gabbros, respec-
K2O based on the K2O log rather than XRF would tively. In contrast, a dozen logged DSDP and ODP
raise the global subduction flux for crustal K2O by sites document the progressive, age-dependent
71%. The high-temperature K2O fluxes between low-temperature alteration of extrusives. Two main
gabbros and seawater, if any, are undetermined. variables control alteration of the extrusives: (1)
extrusive type (flows, pillows, hyaloclastites),
4.6.4. Peridotite Subduction Fluxes which affects permeability and therefore water/
[153] This study considers only those subduction rock ratio, and (2) time. Intersite variations in
fluxes associated with oceanic crust and its over- proportions of flows versus pillows are a principal
lying sediments. Mantle peridotites are excluded, source of local variance: pillows are much more
although they comprise the majority of subducting porous, permeable, and altered than more massive
lithosphere, because their in situ alteration state is flows. Regionally, however, this variance is aver-
virtually unknown [e.g., Peacock, 1996]; obtain- aged, and the waning impact of off-axis hydro-
able samples are unlikely to be representative of thermal circulation on geophysical properties is
normal lower lithosphere. Peacock and Hyndman evident. Seismic velocity and permeability detect
40 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

only the large changes during the first 10 Ma after tions for the same components within subducting
crustal formation [Carlson, 1998; Fisher and sediments at 26 of these subduction zones by Plank
Becker, 2000], but some properties have the resolv- and Langmuir [1998].
ing power to detect subtler off-axis changes. I
identified the following systematic changes in [ 158] Subducted crustal water is released and
properties of the upper extrusives, each of which expelled at various slab depths by several mecha-
appears to wane with a log(age) dependence, nisms, depending on how the water is held–in
throughout the lifetime of oceanic crust: (1) macro- sediment pores, in sedimentary smectites, in basalt
porosity reduction, (2) intergranular alteration to pores, in loosely bound hydrous minerals in the
smectite, celadonite, and calcite, with the calcite extrusives, and in relatively stable hydrous miner-
cementation occurring generally later than the als within the dikes and gabbros (Figure 7). At
others, and (3) increase in large-scale velocity burial depths of <10 km, sediment compaction
due to macroporosity reduction, mainly in pillows. dominates, with additional water coming from
These changes involve progressive uptake of smectite/illite conversion in both sediments and
H2O+, CO2, and K2O, with attendant expulsion extrusive basalts. At greater depths, metamorphic
of pore water and its Cl. Given crustal age, one reactions convert pore water to structural and
can calculate the average concentrations of pore expelled waters. I estimate that half to two thirds
H2O, H2O+, CO2, Cl, and K2O in the extrusives. of subducted water is later refluxed at or near the
toe of the prism, rising along the slab from at least
[ 156 ] Surprisingly, low-temperature alteration 15 km and possibly as deep as 45 km. This long-
causes no net change in total water. Macroporosity distance transport is inferred from patterns of
is replaced with alteration minerals containing dehydration along slab-parallel permeability at
some structural water, and intergranular alteration 15–45 km in the Catalina Schist [Bebout, 1995,
forms hydrous minerals, but pore water loss is 1996] and from magnetotelluric evidence of free
nearly identical to H2O+ gain. Chloride is presum- water in slabs at 15–60 km depth. Crustal dehy-
ably expelled by these transformations, but reliable dration may be as important as transformation of
pore water samples have not been obtained from sedimentary smectite to illite in generating the low-
the extrusives. The overall alteration effect on total chlorinity anomalies seen in some décollements.
crustal content of K 2 O is also smaller than
expected. Enrichment by low-temperature altera- [159] The remaining water is more deeply sub-
tion of the upper extrusives is obvious at abundant ducted as stable hydrous minerals. Much or most
DSDP and ODP crustal sites, but substantial K2O of this water is released by a series of dehydration
depletion occurs within the dikes during high- reactions at depths of 45–100 km. It is carried along
temperature alteration. The majority of crustal with the slab until it escapes by triggering the partial
K2O for all crustal ages is primary rather than melting that permits it to reach Earth’s surface at the
secondary, and most is in the gabbros. The largest arc. Surprisingly, there is little evidence that amount
net geochemical flux from crustal alteration, by an of subducted volatiles affects either amount of
order of magnitude, is the CO2 enrichment that magma generation or earthquake magnitude on
occurs at both intergranular scale and via fracture the interplate boundary, despite indications that
filling. Its pattern of age-dependence is yet to be reflux water traverses the entire boundary.
documented in detail, but existing data are com-
[160] Global flux balances for H2O, CO2, Cl, and
patible with the log(age) dependence seen for
K2O are constrained partly by the ocean/crust
crustal geophysical properties (Figure 2).
interchanges documented above, partly by the
[157] By applying these predictions to 41 modern subduction fluxes above, and additionally by other
subduction zones, one can determine modern mass published data. In general, fluxes among oceans,
fluxes of pore H2O, H2O+, CO2, Cl, and K2O, oceanic crust, continental crust, and mantle bal-
both for individual subduction zones and globally. ance, within the uncertainties. A major exception is
This data set is complemented by flux determina- the present order-of-magnitude excess of marine
41 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

sedimentation over sediment subduction, most Alt, J. C., Very low-grade hydrothermal metamorphism of
obvious in the K2O and CO2 budgets but also basic igneous rocks, in Low-Grade Metamorphism, edited
by M. Frey and D. Robinson, pp. 85 – 114, Blackwell Sci.,
affecting porosity and therefore H2O and Cl. Malden, Mass., 1999.
Presumably, a closer balance is achieved on a time Alt, J. C., and J. Honnorez, Alteration of the upper oceanic
scale of hundreds of millions of years, though crust, Deep Sea Drilling Project Site 417: Mineralogy and
some imbalance may persist in the form of long- chemistry, Contrib. Mineral. Petrol., 87, 149 – 169, 1984.
Alt, J. C., and D. A. H. Teagle, The uptake of carbon during
term growth of continents. Temporal variation in
alteration of ocean crust, Geochim. Cosmochim. Acta, 63,
dissolution and precipitation of evaporites can 1527 – 1535, 1999.
unbalance the global chloride budget dramatically, Alt, J. C., and D. A. H. Teagle, Hydrothermal alteration and
without significantly affecting ocean salinity. fluid fluxes in ophiolites and ocean crust, in Ophiolites and
Oceanic Crust: New Insights from Field Studies and the
[161] A brief overview of CO2 mass fluxes con- Ocean Drilling Program, edited by Y. Dilek et al., Geol.
firms that carbonate precipitation in the oceanic Soc. Am. Spec. Pap., 349, 273 – 282, 2000.
Alt, J. C., C. Laverne, and K. Muehlenbachs, Alteration of the
crust is a major CO2 sink, but it provides an
upper oceanic crust: Mineralogy and processes in DSDP
unsatisfactory balance of mantle and crust input Hole 504B, Leg 83, Initial Rep. Deep Sea Drill. Proj., 83,
versus output. Subducted CO2 reaches subarc 217 – 247, 1985.
depths with minimal loss to reflux, supplying the Alt, J. C., J. Honnorez, C. Laverne, and R. Emmermann, Hy-
CO2 erupted in arc magmas and replenishing drothermal alteration of a 1-km section through the upper
oceanic crust, Deep Sea Drilling Project Hole 504B: Miner-
deeper-mantle CO2. alogy, chemistry and evolution of seawater-basalt interac-
tions, J. Geophys. Res., 91, 10,309 – 10,335, 1986.
[162] The oceanic potassium budget is well bal-
Alt, J. C., C. France-Lanord, P. A. Floyd, P. Castillo, and
anced, with almost as much loss to seafloor dia- A. Galy, Low-temperature hydrothermal alteration of Juras-
genesis as gain from dissolved river input. Some sic ocean crust, Site 801, Proc. Ocean Drill. Program Sci.
signature of subducted sediment K2O on arc Results, 129, 415 – 427, 1992.
magma K2O has been demonstrated [Plank and Alt, J. C., D. A. H. Teagle, C. Laverne, D. Vanko, W. Bach,
J. Honnorez, K. Becker, M. Ayadi, and P. A. Pezard, Ridge
Langmuir, 1993]. The major excess of arc magma flank alteration of upper ocean crust in the eastern Pacific:
K2O over slab K2O, however, requires that most of A synthesis of results for volcanic rocks of Holes 504B and
the former comes from the mantle wedge and 896A, Proc. Ocean Drill. Program Sci. Results, 148, 435 –
crustal enrichment. 452, 1996a.
Alt, J. C., et al., Hydrothermal alteration of a section of upper
oceanic crust in the eastern equatorial Pacific: a synthesis of
Acknowledgments results from Site 504 (DSDP Legs 69, 70, and 83, and ODP
Legs 111, 137, 140, and 148), Proc. Ocean Drill. Program
[163] I am grateful to J. C. Alt and an anonymous reviewer Sci. Results, 148, 417 – 434, 1996b.
for their constructive reviews. This project was made possible Anderson, R. N., and M. D. Zoback, Permeability, underpres-
by the data acquisition efforts of shipboard scientific parties, sures and convection in the oceanic crust near the Costa Rica
operations managers, and particularly the logging scientists of Rift, eastern equatorial Pacific, J. Geophys. Res., 87, 2860 –
two dozen DSDP and ODP legs. Just as important to the 2868, 1982.
success of this project was the availability of log and core data, Anderson, R. N., S. Uyeda, and A. Miyashiro, Geophysical
on CD-ROM and at the Web sites of ODP Logging Services and geochemical constraints at converging plate boundaries;
and JANUS. This research used samples and data provided by Part I. dehydration in the downgoing slab, Geophys. J. R.
DSDP and ODP. ODP is sponsored by the U.S. National Astron. Soc., 44, 333 – 357, 1976.
Science Foundation and participating countries under manage- Anderson, R. N., M. G. Langseth, and J. G. Sclater, The me-
ment of Joint Oceanographic Institutions, Inc. chanisms of heat transfer through the floor of the Indian
Ocean, J. Geophys. Res., 82, 3391 – 3409, 1977.
Anderson, R. N., S. E. DeLong, and W. M. Schwarz, Dehy-
References dration, asthenospheric convection, and seismicity in sub-
duction zones, J. Geol., 88, 445 – 451, 1980.
Alt, J. C., Subseafloor processes in mid-ocean ridge hydrother- Andrews, A. J., Saponite and celadonite in layer 2 basalts,
mal systems, in Seafloor Hydrothermal Systems, Physical, DSDP Leg 37, Contrib. Mineral. Petrol., 73, 323 – 340, 1980.
Chemical, Biological, and Geological Interactions, Geo- Armstrong, R. L., A model for the evolution of strontium and
phys. Monogr. Ser., vol. 91, edited by S. E. Humphris et lead isotopes in a dynamic Earth, Rev. Geophys., 6, 175 –
al., pp. 85 – 114, AGU, Washington, D.C., 1995. 199, 1968.

42 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Armstrong, R. L., The persistent myth of crustal growth, Aust. bon dioxide over the past 100 million years, Am. J. Sci., 283,
J. Earth Sci., 38, 613 – 630, 1991. 641 – 683, 1983.
Arzate, A. J., M. Mareschal, and D. Livelybrooks, Electrical Bevis, M., et al., Geodetic observations of very rapid conver-
image of the subducting Cocos Plate from magnetotelluric gence and back-arc extension at the Tonga arc, Nature, 374,
observations, Geology, 23, 703 – 706, 1995. 249 – 251, 1995.
Azbel, I. Y., and I. N. Tolstikhin, Geodynamics, magmatism, Booij, E., and H. Staudigel, Controls and diversity of altered
and degassing of the Earth, Geochim. Cosmochim. Acta, 54, oceanic crust composition, Eos Trans. AGU, 78(46), AGU
139 – 154, 1990. Fall Meet. Suppl., F806, 1997.
Bach, W., J. C. Alt, Y. Niu, S. E. Humphris, J. Erzinger, and Bourbié, T., O. Coussy, and B. Zinszner, Acoustics of Porous
H. J. B. Dick, The geochemical consequences of late-stage Media, 334 pp., Educ. Technol., Paris, 1987.
low-grade alteration of lower ocean crust at the SW Indian Bray, C. J., and D. E. Karig, Porosity of sediments in accre-
Ridge: Results from ODP Hole 735B (Leg 176), Geochim. tionary prisms and some implications for dewatering pro-
Cosmochim. Acta, 65, 3267 – 3287, 2001. cesses, J. Geophys. Res., 90, 768 – 778, 1985.
Bach, W., S. E. Humphris, and A. T. Fisher, Fluid flow and Brias, A., P. Patriat, and P. Tapponnier, Updated interpretation
fluid-rock interaction within ocean crust: reconciling geo- of magnetic anomalies and seafloor spreading stages in the
chemical, geological, and geophysical observations, in Sub- South China Sea: Implications for the Tertiary tectonics of
surface Biosphere at Mid-Ocean Ridges, Geophys. Monogr. Southeast Asia, J. Geophys. Res., 98, 6299 – 6328, 1993.
Ser., edited by W. Wilcock et al., AGU, Washington, D.C., in Broglia, C., and D. Ellis, Effect of alteration, formation absorp-
press, 2003. tion, and standoff on the response of the thermal neutron
Bebout, G. E., Geometry and mechanisms of fluid flow at 15 porosity log in gabbros and basalts: Examples from Deep
to 45 kilometer depths in an Early Cretaceous accretionary Sea Drilling Project-Ocean Drilling Program sites, J. Geo-
complex, Geophys. Res. Lett., 18, 923 – 926, 1991. phys. Res., 95, 9171 – 9188, 1990.
Bebout, G. E., The impact of subduction-zone metamorphism Brown, K. M., and B. Ransom, Porosity corrections for smec-
on mantle-ocean chemical cycling, Chem. Geol., 126, 191 – tite-rich sediments: Impact on studies of compaction, fluid
218, 1995. generation, and tectonic history, Geology, 24, 843 – 846,
Bebout, G. E., Volatile transfer and recycling at convergent 1996.
margins: Mass-balance and insights from high-P/T meta- Buckley, H. A., J. C. Bevan, K. M. Brown, J. R. Johnson, and
morphic rocks, in Subduction Top to Bottom, Geophys. V. C. Farmer, Glauconite and celadonite: two separate miner-
Mon., vol. 96, edited by G. E. Bebout et al., pp. 179 – 193, al species, Mineral. Mag., 42, 373 – 382, 1978.
AGU, Washington, D.C., 1996. Busch, W. H., P. R. Castillo, P. A. Floyd, and G. Cameron,
Becker, K., et al., Drilling deep into young oceanic crust, Hole Effects of alteration on physical properties of basalts from
504B, Costa Rica Rift, Rev. Geophys., 27, 79 – 102, 1990. the Pigafetta and East Mariana Basins, Proc. Ocean Drill.
Beiersdorf, H., et al., Age and possible modes of formation of Program Sci. Results, 129, 485 – 499, 1992.
the Celebes Sea basement, and thermal regimes within the Caldeira, K., Long-term control of atmospheric carbon diox-
accretionary complexes off SW Mindanao and N Sulawesi, ide: Low-temperature seafloor alteration or terrestrial sili-
paper presented at International Conference on Stratigraphy cate-rock weathering?, Am. J. Sci., 295, 1077 – 1114, 1995.
and Tectonic Evolution of Southeast Asia and the South Cannat, M., How thick is the magmatic crust at slow-spreading
Pacific, GEOTHAI’97 conference, Int. Union of Geol. Sci., oceanic ridges?, J. Geophys. Res., 101, 2847 – 2857, 1996.
Bangkok, Thailand, 1997. Carlson, R. L., Seismic velocities in the uppermost oceanic
Bell, D. R., and G. R. Rossman, Water in Earth’s mantle: The crust: Age dependence and the fate of layer 2A, J. Geophys.
role of nominally anhydrous minerals, Science, 255, 1391 – Res., 103, 7069 – 7077, 1998.
1397, 1992. Carlson, R. L., Lower crustal water contents, P-wave velo-
Berner, E. K., and R. A. Berner, The Global Water Cycle: cities, and modal mineralogy of oceanic diabase and gab-
Geochemistry and Environment, 397 pp., Prentice-Hall, bro, Eos Trans. AGU, 82(47), Fall. Meet. Suppl., F1154,
Old Tappan, N.J., 1987. 2001.
Berner, R. A., Global CO2 degassing and the carbon cycle: Christensen, D. H., and L. J. Ruff, Seismic coupling and outer
Comment on ‘‘Cretaceous ocean crust at DSDP sites 417 and rise earthquakes, J. Geophys. Res., 93, 13,421 – 13,444,
418: Carbon uptake from weathering versus loss by mag- 1988.
matic outgassing,’’ Geochim. Cosmochim. Acta, 54, 2889 – Christensen, N. I., and M. H. Salisbury, Velocities, elastic
2890, 1990. moduli and weathering-age relations for Pacific Layer 2 ba-
Berner, R. A., and K. Caldeira, The geologic carbon cycle and salts, Earth Planet. Sci. Lett., 19, 461 – 470, 1973.
the evolution of atmospheric carbon dioxide, Eos Trans. Christensen, N. I., and M. H. Salisbury, Seismic velocities,
AGU, 83(47), Fall Meet. Suppl., F383, 2002. densities and porosities of Layer 2B and Layer 2C basalts
Berner, R. A., and Z. Kothavala, GEOCARB III: A revised from Hole 504B, Initial Rep. Deep Sea Drill. Proj., 83, 367 –
model of atmospheric CO2 over Phanerozoic time, Am. J. 370, 1985.
Sci., 301, 182 – 204, 2001. Christensen, N. I., and J. D. Smewing, Geology and seismic
Berner, R. A., A. C. Lasaga, and R. M. Garrels, The carbonate- structure of the northern section of the Oman Ophiolite,
silicate geochemical cycle and its effect on atmospheric car- J. Geophys. Res., 86, 2545 – 2555, 1981.

43 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Christensen, N. I., W. W. Wepfer, and R. D. Baud, Seismic ice Ewing Ser., vol. 2, edited by M. Talwani, C. G. A. Har-
properties of sheeted dikes from Hole 504B, ODP Leg 111, rison, and D. E. Hayes, pp. 369 – 382, AGU, Washington,
Proc. Ocean Drill. Program Sci. Results, 111, 171 – 176, D.C., 1979b.
1989. Dymond, J., and M. Lyle, Particle fluxes in the ocean and
Clarke, W. M., The Data of Geochemistry, U. S. Geol. Surv. implications for sources and preservation of ocean sedi-
Bull., 770, 841 pp., 1924. ments, in Material Fluxes on the Surface of the Earth, edited
Coleman, R. C., Ophiolites: Ancient Oceanic Lithosphere?, by W. W. Hay et al., pp. 125 – 142, Natl. Acad. Press, Wa-
229 pp., Springer-Verlag, New York, 1977. shington, D.C., 1994.
COSOD-II, Report of the Second Conference on Scientific Ekart, D., T. E. Cerling, I. P. Montanez, and N. J. Tabor, A 400
Ocean Drilling, Strasbourg, 6 – 8 July, 1987, 142 pp., million year carbon isotope record of pedogenic carbonate:
JOIDES, Washington, DC, 1987. Implications for paleoatmospheric carbon dioxide, Am. J.
Crisp, J. A., Rates of magma emplacement and volcanic out- Sci., 299, 805 – 827, 1999.
put, J. Volcanol. Geotherm. Res., 20, 177 – 211, 1984. Elderfield, H., and A. Schultz, Mid-ocean ridge hydrothermal
Davis, D. M., J. Suppe, and F. A. Dahlen, Mechanics of fold- fluxes and the chemical composition of the ocean, Annu.
and-thrust belts and accretionary wedges, J. Geophys. Res., Rev. Earth Planet. Sci., 24, 191 – 224, 1996.
88, 1153 – 1172, 1983. Ernewein, M., C. Pflumino, and H. Whitechurch, The death of
Davis, E. E., et al., FlankFlux: An experiment to study the an accretion zone as evidenced by the magmatic history of
nature of hydrothermal circulation in young oceanic crust, the Samail ophiolite (Oman), Tectonophysics, 151, 247 – 274,
Can. J. Earth Sci., 29, 925 – 952, 1992. 1988.
Defant, M. J., and M. S. Drummond, Derivation of some mod- Ernst, W. G., Hornblende, the continent maker: Evolution of
ern arc magmas by melting of young subducted lithosphere, H2O during circum-Pacific subduction versus continental
Nature, 347, 662 – 665, 1990. collision, Geology, 27, 675 – 678, 1999.
DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein, Current Fehn, U., and L. M. Cathles, The influence of plate movement
plate motions, Geophys. J. Int., 101, 425 – 478, 1990. on the evolution of hydrothermal convection cells in the
DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein, Effect of oceanic crust, Tectonophysics, 125, 289 – 312, 1986.
recent revisions to the geomagnetic reversal time scale on Fisher, A. T., Permeability within basaltic oceanic crust, Rev.
estimates of current plate motions, Geophys. Res. Lett., 21, Geophys., 36, 143 – 182, 1998.
2191 – 2194, 1994. Fisher, A. T., and K. Becker, Channelized fluid flow in oceanic
Dewey, J. F., and B. F. Windley, Growth and differentiation of crust reconciles heat-flow and permeability data, Nature,
the continental crust, Philos. Trans. R. Soc. London Ser. A, 403, 71 – 74, 2000.
301, 189 – 206, 1981. Fisher, A., K. Becker, T. N. Narasimhan, M. Langseth, and
DeWit, M. J., and R. A. Hart, Earth’s earliest continental litho- M. Mottl, Passive, off-axis convection on the southern flank
sphere, hydrothermal flux and crustal recycling, Lithos, 30, of the Costa Rica Rift, J. Geophys. Res., 95, 9343 – 9370, 1990.
309 – 335, 1993. Fisher, A. T., K. Becker, and T. N. Narasimhan, Off-axis hy-
Dick, H. J. B., P. S. Meyer, S. H. Bloomer, S. H. Kirby, D. S. drothermal circulation: Parametric tests of a refined model of
Stakes, and C. Mawer, Lithostratigraphic evolution of an in- processes at Deep Sea Drilling Project/Ocean Drilling Pro-
situ section of oceanic layer 3, Proc. Ocean Drill. Program gram Site 504, J. Geophys. Res., 99, 3097 – 3121, 1994.
Sci. Results, 118, 439 – 538, 1991. Fisher, D. M., Fabrics and veins in the forearc: A record of
Dick, H. J. B., et al., A long in situ section of the lower ocean cyclic fluid flow at depths of <15 km, in Subduction Top to
crust: Results of ODP Leg 176 drilling at the Southwest Bottom, Geophys. Monogr. Ser., vol. 96, edited by G. E.
Indian Ridge, Earth Planet. Sci. Lett., 179, 31 – 51, 2000. Bebout et al., pp. 75 – 89, AGU, Washington, D.C., 1996.
Dick, H. J., J. Lin, P. J. Michael, H. Schouten, and J. E. Snow, Fox, P. J., and J. Stroup, The plutonic foundation of the ocea-
Ultra-slow spreading - a new class of ocean ridge, Eos Trans. nic crust, in The Sea, vol. 7, The Oceanic Lithosphere, edited
AGU, 83(47), Fall Meet. Suppl., F1267, 2002. by C. Emiliani, pp. 119 – 218, John Wiley, New York, 1981.
Dickinson, W. R., Potash-depth (K-h) relations in continental Francheteau, J., P. Choukroune, R. Hekinian, X. Le Pichon,
margin and intra-oceanic magmatic arcs, Geology, 3, 53 – 56, and H. D. Needham, Oceanic fracture zones do not provide
1975. deep sections in the crust, Can. J. Earth Sci., 13, 1223 –
Dickinson, W. R., and T. Hatherton, Andesitic volcanism and 1235, 1975.
seismicity around the Pacific, Science, 157, 801 – 803, 1967. Franck, S., and C. Bounama, Continental growth and volatile
Donnelly, T. W., R. A. Pritchard, R. Emmermann, and exchange during Earth’s evolution, Phys. Earth Planet. Int.,
H. Puchelt, The aging of oceanic crust: synthesis of the 100, 189 – 196, 1997.
mineralogical and chemical results of Deep Sea Drilling François, L. M., and J. C. G. Walker, Modelling the Phaner-
Project, Legs 51 through 53, Initial Rep. Deep Sea Drill. ozoic carbon cycle and climate: Constraints from the
87
Proj., 51 – 53, 1563 – 1577, 1979a. Sr/86Sr isotopic ratio of seawater, Am. J. Sci., 292, 81 –
Donnelly, T. W., G. Thompson, and P. T. Robinson, Very-low- 135, 1992.
temperature hydrothermal alteration of the oceanic crust and Furnes, H., and H. Staudigel, Biological mediation in ocean
the problem of fluxes of potassium and magnesium, in Deep crust alteration: How deep is the deep biosphere?, Earth
Drilling Results in the Atlantic Ocean: Ocean Crust, Maur- Planet. Sci. Lett., 166, 97 – 103, 1999.

44 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Furnes, H., H. Staudigel, I. H. Thorseth, T. Torsvik, K. Mueh- by W. W. Hay et al., pp. 15 – 27, Natl. Acad. Press, Washing-
lenbachs, and O. Tumyr, Bioalteration of basaltic glass in the ton, D.C., 1994.
oceanic crust, Geochem. Geophys. Geosys., 2, Paper number Hays, J. D., and W. C. Pitman, III, Lithospheric plate motion,
2000GC000150, 2001. sea-level changes, and climatic and ecological consequences,
Fyfe, W. S., Deep fluids and volatile recycling; crust to mantle, Nature, 246, 18 – 22, 1973.
Tectonophysics, 275, 243 – 251, 1997. Hilde, T. W. C., Sediment subduction versus accretion around
Gaffin, S., Ridge volume dependence on seafloor generation the Pacific, Tectonophysics, 99, 381 – 397, 1983.
rate and inversion using long term sealevel change, Am. J. Hill, M. N., Recent geophysical exploration of the ocean floor,
Sci., 287, 596 – 611, 1987. Phys. Chem. Earth, 2, 129 – 163, 1957.
Gerlach, T. M., Degassing of carbon dioxide from basaltic Holland, H. D., River transport to the oceans, in The Sea, vol.
magma at spreading centers: 2. mid-oceanic ridge basalts, 7, The Oceanic Lithosphere, edited by C. Emiliani, pp. 763 –
J. Volcanol. Geotherm. Res., 39, 221 – 232, 1989. 800, John Wiley, New York, 1981.
Gill, J., Orogenic Andesites and Plate Tectonics, 390 pp., Honnorez, J., The aging of the oceanic crust at low tempera-
Springer-Verlag, New York, 1981. ture, in The Sea, vol. 7, The Oceanic Lithosphere edited by
Gillis, K., and P. T. Robinson, Distribution of alteration zones C. Emiliani, pp. 525 – 587, John Wiley, New York, 1981.
in the upper oceanic crust, Geology, 16, 262 – 266, 1988. Honnorez, J., C. Laverne, H.-W. Hubberten, R. Emmermann,
Goldberg, D., The role of downhole measurements in marine and K. Muehlenbachs, Alteration processes in layer 2 basalts
geology and geophysics, Rev. Geophys., 35, 315 – 342, 1997. from Deep Sea Drilling Project Hole 504B, Costa Rica Rift,
Green, N. L., and D. L. Harry, On the relationship between Initial Rep. Deep Sea Drill. Proj., 69, 509 – 546, 1983.
subducted slab age and arc basalt petrogenesis, Cascadia Honza, E., H. L. Davies, J. B. Keene, and D. L. Tiffin, Plate
subduction system, North America, Earth Planet. Sci. Lett., boundaries and evolution of the Solomon Sea region, Geo.
171, 367 – 381, 1999. Mar. Lett., 7, 161 – 168, 1987.
Grevemeyer, I., and W. Weigel, Seismic velocities of the Humphris, S. E., Hydrothermal processes at mid-ocean ridges,
uppermost igneous crust versus age, Geophys. J. Int., 124, U. S. Natl. Rep. Int. Union Geod. Geophys., 1991 – 1994,
631 – 635, 1996. Rev. Geophys., 33, 71 – 80, 1995.
Hacker, B. R., Eclogite formation and the rheology, buoyancy, Hyndman, R. D., Dipping seismic reflectors, electrically con-
seismicity, and H2O content of oceanic crust, in Subduction ductive zones and trapped water in the crust over a subduct-
Top to Bottom, Geophys. Monogr. Ser., vol. 96, edited by ing plate, J. Geophys. Res., 93, 13,391 – 13,405, 1988.
G. E. Bebout et al., pp. 337 – 346, AGU, Washington, D.C., Hyndman, R. D., and M. J. Drury, The physical properties of
1996. oceanic basement rocks from deep drilling on the Mid-Atlan-
Hamano, Y., Physical properties of basalts from Holes 417D tic Ridge, J. Geophys. Res., 81, 4042 – 4052, 1976.
and 418A, Initial Rep. Deep Sea Drill. Proj., 51 – 53, 1457 – Hyndman, R. D., M. Yamano, and D. A. Oleskevich, The
1466, 1979. seismogenic zone of subduction thrust faults, Island Arc, 6,
Harrison, C. G. A., Constraints on ocean volume change since 244 – 260, 1997.
the Archean, Geophys. Res. Lett., 26, 1913 – 1916, 1999. Ildefonse, B., and P. Pezard, Electrical properties of slow-
Harry, D. L., and N. L. Green, Slab dehydration and basalt spreading ridge gabbros from ODP Site 735, Southwest In-
petrogenesis in subduction systems involving very young dian Ridge, Tectonophysics, 330, 69 – 92, 2001.
oceanic lithosphere, Chem. Geol., 160, 309 – 333, 1999. Ionov, D. A., W. L. Griffin, and S. Y. Oreilly, Volatile-bearing
Hart, R. A., Chemical exchange between sea water and deep minerals and lithophile trace elements in the upper mantle,
ocean basalts, Earth Planet. Sci. Lett., 9, 269 – 279, 1970. Chem. Geol., 141, 153 – 184, 1997.
Hart, R. A., A model for chemical exchange in the basalt- Ito, E. D., M. Harris, and A. T. Anderson Jr., Alteration of
seawater system of oceanic Layer II, Can. J. Earth Sci., oceanic crust and geologic cycling of chlorine and water,
10, 799 – 816, 1973. Geochim. Cosmochim. Acta, 47, 1613 – 1624, 1983.
Hart, S. R., K, Rb, Cs contents, and K/Rb, K/Cs ratios of fresh Itturrino, G. J., N. I. Christensen, K. Becker, L. O. Boldreel,
and altered submarine basalts, Earth Planet. Sci. Lett., 6, P. K. H. Harvey, and P. Pezard, Physical properties and
295 – 303, 1969. elastic constants of upper crustal rocks from core-log mea-
Hart, S. R., and H. Staudigel, The control of alkalis and ura- surements in Hole 504B, Proc. Ocean Drill. Program Sci.
nium in seawater by ocean crust alteration, Earth Planet. Sci. Results, 137/140, 273 – 292, 1995.
Lett., 58, 202 – 212, 1982. Iwamori, H., Transportation of H2O and melting in subduction
Hart, S. R., A. J. Erlank, and E. J. D. Kable, Sea floor basalt zones, Earth Planet. Sci. Lett., 160, 65 – 80, 1998.
alteration: some chemical and Sr isotopic effects, Contrib. Iwamori, H., Transportation of H2O and melting beneath the
Mineral. Petrol., 44, 219 – 230, 1974. Japan arcs, Bull. Earthquake Res. Inst. Univ. Tokyo, 76,
Hart, S. R., J. Blusztajn, H. J. B. Dick, P. S. Meyer, and 377 – 389, 2001.
K. Muehlenbachs, The fingerprint of seawater circulation Jambon, A., B. Deruelle, G. Dreibus, and F. Pineau, Chlorine
in a 500-meter section of ocean crust gabbros, Geochim. and bromine abundance in MORB; The contrasting beha-
Cosmochim. Acta, 63, 4059 – 4080, 1999. viour of the Mid-Atlantic Ridge and East Pacific Rise and
Hay, W. W., Pleistocene-Holocene fluxes are not the Earth’s implications for chlorine geodynamic cycle, Chem. Geol.,
norm, in Material Fluxes on the Surface of the Earth, edited 126, 101 – 117, 1995.

45 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Jarrard, R. D., Relations among subduction parameters, Rev. Kerrick, D. M., and J. A. D. Connolly, Metamorphic devola-
Geophys., 24, 217 – 284, 1986. tilization of subducted oceanic metabasalts: Implications for
Jarrard, R. D., R. L. Larson, A. T. Fisher, and L. J. Abrams, seismicity, arc magmatism and volatile recycling, Earth Pla-
Geophysical aging of oceanic crust: Evidence from Hole net. Sci. Lett., 189, 19 – 29, 2001.
801C, Proc. Ocean Drill. Program Sci. Results, 144, 649 – Kirby, S. H., E. R. Engdahl, and R. Denlinger, Intraslab earth-
663, 1995. quakes and arc volcanism: dual physical expressions of crus-
Jarrard, R. D., L. J. Abrams, R. Pockalny, R. L. Larson, and tal and uppermost mantle metamorphism in subducting slabs,
T. Hirono, Physical properties of upper oceanic crust: ODP in Subduction Top to Bottom, Geophys. Monogr. Ser., 96,
Hole 801C and the waning of hydrothermal circulation, J. Geo- edited by G. E. Bebout et al., pp. 195 – 214, AGU,
phys. Res, 108(D7), 2188, doi:10.1029/2001JB001727, Washington, D.C., 1996.
2003. Kominz, M. A., Oceanic ridge volumes and sea-level change:
Jochum, K. P., A. W. Hofmann, E. Ito, H. M. Seufert, and An error analysis, in Interregional Unconformities and Hy-
W. M. White, K, U and Th in mid-ocean ridge basalt glasses drocarbon Accumulation, edited by J. S. Schlee, Mem. Am.
and heat production, K/U and K/Rb in the mantle, Nature, Assoc. Petrol. Geol., 36, 109 – 127, 1984.
306, 431 – 436, 1983. Larson, R. L., Geological consequences of superplumes, Geol-
Johnson, D. M., Fluid permeability of oceanic basalts, Initial ogy, 19, 547 – 550, 1991.
Rep. Deep Sea Drill. Proj., 51 – 53, 1473 – 1477, 1979a. Lawrence, J. R., J. J. Drever, and M. Kastner, Low temperature
Johnson, D. M., Crack distribution in the upper oceanic crust alteration of basalts predominates at DSDP Site 395, Initial
and its effects upon seismic velocity, seismic structure, for- Rep. Deep Sea Drill. Proj., 45, 609 – 612, 1978.
mation permeability, and fluid circulation, Initial Rep. Deep Lecuyer, C., M. Brouxel, and F. Albarede, Elemental fluxes
Sea Drill. Proj., 51 – 53, 1479 – 1490, 1979b. during hydrothermal alteration of the Trinity Ophiolite (Ca-
Johnson, H. P., and S. W. Semyan, Age variation in the physical lifornia, USA) by seawater, Chem. Geol., 89, 87 – 115,
properties of oceanic basalts: implications for crustal forma- 1990.
tion and evolution, J. Geophys. Res., 99, 3123 – 3134, 1994. Leeman, W. P., Boron and other fluid-mobile elements in vol-
Karato, S., Physical properties of basalts from Deep Sea Dril- canic arc lavas: Implications for subduction processes, in
ling Project Hole 504B, Initial Rep. Deep Sea Drill. Proj., Subduction Top to Bottom, Geophys. Monogr., vol. 96, edited
69, 687 – 695, 1983a. by G. E. Bebout et al., pp. 269 – 276, AGU, Washington,
Karato, S., Physical properties of basalts from the Galapagos, D.C., 1996.
Leg 70, Initial Rep. Deep Sea Drill. Proj., 70, 423 – 428, Le Pichon, X., P. Henry, and the Kaiko-Nankai Scientific
1983b. Crew, Water budgets in accretionary wedges: A compari-
Karson, J. A., Internal structure of oceanic lithosphere: a per- son, Philos. Trans. R. Soc. London Ser. A, 335, 315 – 330,
spective from tectonic windows, in Faulting and Magmatism 1991.
at Mid-Ocean Ridges, Geophys. Monogr. Ser., vol. 106, edi- Le Pichon, X., P. Henry, and S. Lallemant, Accretion and
ted by W. R. Buck et al., 177 – 218, AGU, Washington, D.C., erosion in subduction zones: The role of fluids, Annu. Rev.
1998. Earth Planet. Sci., 21, 307 – 331, 1993.
Karson, J. A., Geologic structure of the uppermost oceanic Lippard, S. J., A. W. Shelton, and I. G. Gass, The Ophiolite of
crust created at fast- to intermediate-rate spreading centers, Northern Oman, 178 pp., Blackwell Sci., Malden, Mass.,
Annu. Rev. Earth Planet. Sci., 30, 347 – 384, 2002. 1986.
Kasahara, J., A. Kamimura, G. Fujie, and R. Hino, Influence of Liu, J., S. R. Bohlen, and W. G. Ernst, Stability of hydrous
water on earthquake generation along subduction zones, phases in subducting oceanic crust, Earth Planet. Sci. Lett.,
Bull. Earthquake. Res. Inst. Univ. Tokyo, 76, 291 – 303, 2001. 143, 161 – 171, 1996.
Kastner, M., H. Elderfield, and J. B. Martin, Fluids in conver- Livingstone, D. A., Chemical composition of rivers and lakes,
gent margins: What do we know about their composition, U. S. Geol. Survey Prof. Paper, 440-G, G1 – G64, 1963.
origin, role in diagenesis and importance for oceanic chemi- Magenheim, A. J., A. J. Spivack, P. J. Michael, and J. M.
cal fluxes, Philos. Trans. R. Soc. London, Ser. A, 335, 243 – Gieskes, Chlorine stable isotope composition of the oceanic
259, 1991. crust: Implication for Earth’s distribution of chlorine, Earth
Kastner, M., et al., Geochemical and isotopic evidence for fluid Planet. Sci. Lett., 131, 427 – 432, 1995.
flow in the western Nankai subduction zone, Japan, Proc. Martin, W. R., and F. L. Sayles, Seafloor diagenetic fluxes, in
Ocean Drill. Program Sci. Results, 131, 397 – 413, 1993. Material Fluxes on the Surface of the Earth, edited by W. W.
Kato, T., and Y. Kotake, Tectonics of the Philippine Sea plate Hay, et al., pp. 143 – 163, Natl. Acad. Press, Washington,
as seen from GPS observations, Eos Trans. AGU, 83(47), D.C., 1994.
Fall Meet. Suppl., F1321, 2002. Marty, B., and I. N. Tolstikhin, CO2 fluxes from midocean
Kerrick, D. M., Present and past nonanthropogenic CO2 degas- ridges, arcs and plumes, Chem. Geol., 145, 233 – 248, 1998.
sing from the solid Earth, Rev. Geophys., 39, 565 – 585, McCaffrey, R., Influences of recurrence times and fault zone
2001. temperatures on the age-rate dependence of subduction zone
Kerrick, D. M., and J. A. D. Connolly, Subduction of ophicar- seismicity, J. Geophys. Res., 102, 22,839 – 22,854, 1997.
bonates and recycling of CO2 and H2O, Geology, 26, 375 – McCulloch, M. T., The role of subducted slabs in an evolving
378, 1998. earth, Earth Planet. Sci. Lett., 115, 89 – 100, 1993.

46 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

McGovern, P. J., and G. Schubert, Thermal evolution of the Ono, S., Stability limits of hydrous minerals in sediment and
Earth: Effects of volatile exchange between atmosphere and midocean ridge basalt compositions: Implications for water
interior, Earth Planet. Sci. Lett., 96, 27 – 37, 1989. transport in subduction zones, J. Geophys. Res., 103,
Meade, C., and R. Jeanloz, Deep-focus earthquakes and recy- 18,253 – 18,267, 1998.
cling of water into the Earth’s mantle, Science, 252, 68 – 72, Parsons, B., The rates of plate creation and consumption, Geo-
1991. phys. J. R. Astron. Soc., 67, 437 – 448, 1981.
Meybeck, M., Concentrations des eaux fluviales en éléments Pawley, A. R., The pressure and temperature stability limits of
majeurs et apports en solution aux océans, Rev. Geol. Dyn. lawsonite: Implications for H2O recycling in subduction
Geogr. Phys., 21, 215 – 246, 1979. zones, Contrib. Mineral. Petrol., 118, 99 – 108, 1994.
Meybeck, M., Atmospheric inputs and river transport of dis- Pawley, A. R., and J. R. Holloway, Water sources for subduc-
solved substances, in Dissolved Load of Rivers and Surface tion zone volcanism: New experimental constraints, Science,
Water Quantity/Quality Relationships, Proceedings of Ham- 260, 664 – 667, 1993.
burg Symposium, IAHS Publ., 141, edited by B. W. Webb, Peacock, S. M., Fluid processes in subduction zones, Science,
pp. 173 – 192, Int. Assoc. of Hydrol. Sci. Press, Wallingford, 248, 329 – 337, 1990.
Oxfordshire, U.K., 1983. Peacock, S. M., Numerical simulation of subduction zone pres-
Meybeck, M., Origin and variable composition of present day sure temperature time paths - constraints on fluid production
riverborne material, in Material Fluxes on the Surface of the and arc magmatism, Philos. Trans. R. Soc. London Ser. A,
Earth, edited by W. W. Hay et al., pp. 61 – 73, Natl. Acad. 335, 341 – 353, 1991.
Press, Washington, D.C., 1994. Peacock, S. M., The importance of blueschist-eclogite dehy-
Milliman, J. D., and R. H. Meade, World-wide delivery of dration reactions in subducting oceanic crust, Geol. Soc.
river sediment to the oceans, J. Geology, 91, 1 – 21, 1983. Amer. Bull., 105, 684 – 694, 1993.
Milliman, J. D., and J. P. M. Syvitski, Geomorphic/tectonic Peacock, S. M., Thermal and petrological structure of subduc-
control of sediment discharge to the ocean: the importance of tion zones, in Subduction Top to Bottom, Geophys. Monogr.
small mountainous rivers, in Material Fluxes on the Surface Ser., vol. 96, edited by G. E. Bebout et al., pp. 119 – 133,
of the Earth, edited by W. W. Hay, et al., pp. 74 – 85, Natl. AGU, Washington, D.C., 1996.
Acad. Press, Washington, D.C., 1994. Peacock, S. M., Are the lower planes of double seismic zones
Minster, J. B., and T. H. Jordan, Present-day plate motions, caused by serpentine dehydration in subducting oceanic
J. Geophys. Res., 83, 5331 – 5354, 1978. mantle?, Geology, 29, 299 – 302, 2001.
Miyazaki, S., and K. Heki, Crustal velocity field of southwest Peacock, S. M., and R. D. Hyndman, On the importance of
Japan: Subduction and arc-arc collision, J. Geophys. Res., serpentine in the subduction factory, Eos Trans. AGU,
106, 4305 – 4326, 2001. 82(47), Fall Meet. Suppl., F1154, 2001.
Moore, J. C., and D. Saffer, Updip limit of the seismogenic Peacock, S. M., T. Rushmer, and A. B. Thompson, Partial
zone beneath the accretionary prism of southwest Japan: An melting of subducting oceanic crust, Earth Planet. Sci. Lett.,
effect of diagenetic to low-grade metamorphic processes and 121, 227 – 244, 1994.
increasing effective stress, Geology, 29, 183 – 186, 2001. Pearson, P. N., and M. R. Palmer, Atmospheric carbon dioxide
Moore, J. C., and P. Vrolijk, Fluids in accretionary prisms, Rev. concentrations over the past 60 million years, Nature, 406,
Geophys., 30, 113 – 135, 1992. 695 – 699, 2000.
Moos, D., and D. Marion, Morphology of extrusive basalts and Pezard, P. A., Electrical properties of mid-ocean ridge basalt
its relationship to seismic velocities in the shallow oceanic and implications for the structure of the upper oceanic
crust, J. Geophys. Res., 99, 2985 – 2994, 1994. crust at Hole 504B, J. Geophys. Res., 95, 9237 – 9264,
Mottl, M. J., and C. G. Wheat, Hydrothermal circulation 1990.
through mid-ocean ridge flanks: Fluxes of heat and magne- Pezard, P. A., F. Einaudi, D. Hermitte, J.-J. Cocheme, C. Cou-
sium, Geochim. Cosmochim. Acta, 58, 2225 – 2237, 1994. lon, and C. Laverne, MORB emplacement and structure;
Muehlenbachs, K., The alteration and aging of the basaltic Insights from the Semail Ophiolite, Oman, Geophys. Res.
layer of the sea floor: oxygen isotope evidence from Lett., 27, 3933 – 3936, 2000.
DSDP/IPOD Legs 51, 52, and 53, Initial Rep. Deep Sea Philippot, P., P. Agrinier, and M. Scambelluri, Chlorine cycling
Drill. Proj., 51 – 53, 1159 – 1167, 1979. during subduction of altered oceanic crust, Earth Planet. Sci.
Nakamori, T., Global carbonate accumulation rates from Cre- Lett., 161, 33 – 44, 1998.
taceous to Present and their implications for the carbon cycle Plank, T., and C. H. Langmuir, Tracing trace elements from
model, Island Arc, 10, 1 – 8, 2001. sediment input to volcanic output at subduction zones, Nat-
Nockolds, S. R., Average chemical compositions of some ure, 362, 739 – 743, 1993.
igneous rocks, Geol. Soc. Am. Bull., 65, 1007 – 1032, 1954. Plank, T., and C. H. Langmuir, The chemical composition of
Nur, A., Effects of stress on velocity anisotropy in rocks with subducting sediment and its consequences for the crust and
cracks, J. Geophys. Res., 76, 2022 – 2034, 1971. mantle, Chem. Geol., 145, 325 – 394, 1998.
Omori, S., S. Kamiya, S. Maruyama, and D. Zhao, Morphol- Poli, S., and M. W. Schmidt, H2O transport and release in
ogy of the intraslab seismic zone and devolatilization phase subduction zones: Experimental constraints on basaltic and
equilibria of the subducting slab peridotite, Bull. Earthquake andesitic systems, J. Geophys. Res., 100, 22,299 – 22,314,
Res. Inst. Univ. Tokyo, 76, 455 – 478, 2001. 1995.

47 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Purdy, G. M., L. S. L. Kong, G. L. Christeson, and S. C. Saffer, D. M., and B. A. Bekins, Hydrologic controls on the
Solomon, Relationship between spreading rate and the seis- morphology and mechanics of accretionary wedges, Geol-
mic structure of mid-ocean ridges, Nature, 355, 815 – 817, ogy, 30, 271 – 274, 2002.
1992. Sager, W., Argo abyssal plain magnetic lineations revisited:
Raitt, R. W., The crustal rocks, in The Sea, vol. 3, edited by Implications for the onset of seafloor spreading and tectonic
M. N. Hill, pp. 85 – 102, Wiley-Interscience, New York, 1963. evolution of the Eastern Indian Ocean, Proc. Ocean Drill.
Raleigh, C. B., and M. S. Paterson, Experimental deformation Program Sci. Results, 123, 659 – 669, 1992.
of serpentinite and its tectonic implications, J. Geophys. Salisbury, M. H., N. I. Christensen, and R. H. Wilkens, Nature
Res., 70, 3965 – 3985, 1965. of the Layer 2/3 transition from a comparison of laboratory
Rangin, C., and E. Silver, Geological setting of the Celebes and logging velocities and petrology at the base of Hole
and Sulu Seas, Proc. Ocean Drill Program Initial Rep., 124, 504B, Proc. Ocean Drill. Program Sci. Results, 148, 409 –
35 – 42, 1990. 414, 1996.
Ransom, B., and H. C. Helgeson, A chemical and thermody- Sansone, F. J., M. J. Mottl, E. J. Olson, C. G. Wheat, and M. D.
namic model of dioctehedral 2:1 layer clay minerals in di- Lilley, CO2-depleted fluids from mid-ocean ridge-flank hy-
agenetic processes: Dehydration of dioctahedral aluminous drothermal springs, Geochim. Cosmochim. Acta, 62, 2247 –
smectite as a function of temperature and depth in sedimen- 2252, 1998.
tary basins, Am. J. Sci., 295, 245 – 281, 1995. Schilling, J.-G., C. K. Unni, and M. L. Bender, Origin of
Rea, D. K., Geologic records in deep-sea muds, GSA Today, 3, chlorine and bromine in the oceans, Nature, 273, 631 –
205 – 210, 1993. 636, 1978.
Rea, D. K., and L. J. Ruff, Composition and mass flux of Schmidt, M. W., Experimental constraints on recycling of po-
sediment entering the world’s subduction zones: Implications tassium from subducted oceanic crust, Science, 272, 1927 –
for global sediment budgets, great earthquakes, and volcan- 1930, 1996.
ism, Earth Planet. Sci. Lett., 140, 1 – 12, 1996. Schmidt, M. W., and S. Poli, Experimentally based water bud-
Reymer, A., and G. Schubert, Phanerozoic addition rates to the gets for dehydrating slabs and consequences for arc magma
continental crust and crustal growth, Tectonics, 3, 63 – 77, generation, Earth Planet. Sci. Lett., 163, 361 – 379, 1998.
1984. Schmincke, H.-U., and U. Bednarz, Pillow-, sheet flow-, and
Robertson, A. H. F., and Shipboard Scientific Party, Tectonic breccia volcano-tectonic-hydrothermal cycles in the Extru-
introduction, Proc. Ocean Drill. Program Init. Rep., 160, 5 – sive Series of the northwestern Troodos Ophiolite (Cyprus),
20, 1996. in Proceedings Symposium Troodos 87 – Ophiolites and
Robinson, P. T., M. F. Flower, H. Staudigel, and D. A. Swan- Oceanic Lithosphere, edited by J. Malpas et al., pp. 185 –
son, Lithology and eruptive stratigraphy of Cretaceous ocea- 207, Geol. Surv. Dept., Nicosia, Cyprus, 1990.
nic crust, western Atlantic, Initial Rep. Deep Sea Drill. Proj., Schreiber, E., and P. J. Fox, Compressional wave velocities and
51 – 53, 1535 – 1556, 1979. mineralogy of fresh basalts from the FAMOUS area and the
Robinson, P. T., H. J. B. Dick, D. Von Herzen, and R. Pierre, Oceanographer Fracture Zone and the texture of Layer 2A of
Metamorphism and alteration in oceanic layer 3: Hole 735B, the oceanic crust, J. Geophys. Res., 81, 4071 – 4076, 1976.
Proc. Ocean Drill. Program Sci. Results, 118, 541 – 562, Schreiber, E., and P. J. Fox, Density and P-wave velocity of
1991. rocks from the FAMOUS region and their implication to the
Ronov, A. B., and A. A. Yaroshevskiy, A new model for the structure of the oceanic crust, Geol. Soc. Am. Bull., 88, 600 –
chemical structure of the Earth’s crust, Geochem. Int., 13, 608, 1977.
89 – 121, 1976. Seno, T., S. Stein, and A. E. Gripp, A model for the motion of
Rubey, W. W., Geologic history of sea water: an attempt to the Philippine Sea plate consistent with NUVEL-1 and geo-
state the problem, Geol. Soc. Am. Bull., 62, 1111 – 1147, logical data, J. Geophys. Res., 98, 17,941 – 17,948, 1993.
1951. Serra, O., Fundamentals of Well Log Interpretation, vol. 2, The
Ruff, L., and H. Kanamori, Seismicity and the subduction Interpretation of Logging Data, 684 pp., Elsevier, New
process, Phys. Earth Planet. Int., 23, 240 – 252, 1980. York, 1986.
Ruff, L. J., and B. W. Tichelaar, What controls the seismogenic Shaw, P., Age variations of oceanic crust Poisson’s ratio: In-
plate interface in subduction zones?, in Subduction Top to version and a porosity inversion model, J. Geophys. Res., 99,
Bottom, Geophys. Monogr. Ser., vol. 96, edited by G. E. 3057 – 3066, 1994.
Bebout et al., pp. 105 – 111, AGU, Washington, D.C., 1996. Shimakawa, Y., and Y. Honkura, Electrical conductivity struc-
Ryan, J., J. Morris, G. Bebout, and B. Leeman, Describing ture beneath the Ryukyu Trench-Arc system and its relation
chemical fluxes in subduction zones: insights from ‘‘depth to subduction of the Philippine Sea plate, J. Geomag. Geoe-
profiling’’ studies of arc and forearc rocks, in Subduction Top lect., 43, 1 – 20, 1991.
to Bottom, Geophys. Monogr. Ser., vol. 96, edited by G. E. Shipboard Scientific Party, Site 395, Initial Rep. Deep Sea
Bebout et al., pp. 263 – 268, AGU, Washington, D.C., 1996. Drill. Proj., 45, 131 – 264, 1978.
Saffer, D. M., and B. A. Bekins, Episodic fluid flow in the Shipboard Scientific Party, Site 562, Initial Rep. Deep Sea
Nankai accretionary complex: Timescale, geochemistry, flow Drill. Proj., 82, 255 – 278, 1985.
rates, and fluid budget, J. Geophys. Res., 103, 30,351 – Shipboard Scientific Party, Site 735, Proc. Ocean Drill. Pro-
30,370, 1998. gram Initial Rep., 118, 89 – 222, 1989.

48 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

Shipboard Scientific Party, Site 768, Proc. Ocean Drill. Pro- Staudigel, H., et al., Geochemical Earth Reference Model
gram Initial Rep., 124, 195 – 297, 1990a. (GERM); Description of the initiative, Chem. Geol., 145,
Shipboard Scientific Party, Site 770, Proc. Ocean Drill. Pro- 153 – 159, 1998a.
gram Initial Rep., 124, 343 – 397, 1990b. Staudigel, H., A. Yayanos, R. Chastain, G. Davies, E. A. T.
Shipboard Scientific Party, Site 504, Proc. Ocean Drill. Pro- Verdurmen, P. Schiffman, R. Bourcier, and H. de Baar, Bio-
gram Initial Rep., 140, 37 – 200, 1992. logically mediated dissolution of volcanic glass in seawater,
Shipboard Scientific Party, Site 894, Proc. Ocean Drill. Pro- Earth Planet. Sci. Lett., 164, 233 – 244, 1998b.
gram Initial Rep., 147, 45 – 108, 1993. Stein, C. A., and S. Stein, Constraints on hydrothermal heat
Shipboard Scientific Party, Site 735 [online], Proc. Ocean flux through the oceanic lithosphere from global heat flow,
Drill. Program Initial Rep., 176, 1 – 314, 1999. (Available at J. Geophys. Res., 99, 3081 – 3095, 1994.
http://www-odp.tamu.edu/publications/176_IR/VOLUME/ Stephen, R. A., Data report: physical properties measurements
CHAPTERS/CHAP_03.PDF). in ODP Hole 735B [online], Proc. Ocean Drill Program Sci.
Shipboard Scientific Party, Site 801 [online], Proc. Ocean Drill. Results, 176, 2001. (Available at http://www-odp.tamu.edu/
Program Initial Rep., 185, 1 – 222, 2000a. (Available at publications/176SR/chap.02/chap.02.htm)
http://www-odp.tamu.edu/publications/185_IR/VOLUME/ Sugisaki, R., Chemical characteristics of volcanic rocks: Rela-
CHAPTERS/IR185_03.PDF). tion to plate movements, Lithos, 9, 17 – 30, 1976.
Shipboard Scientific Party, Site 1149 [online], Proc. Ocean Tatsumi, Y., and T. Kogiso, Trace-element transport during
Drill. Program Initial Rep., 185, 1 – 190, 2000b. (Available at dehydration processes in the subducted oceanic crust: 2.
http://www-odp.tamu.edu/publications/185_IR/VOLUME/ Origin of chemical and physical characteristics in arc mag-
CHAPTERS/IR185_04.PDF). matism, Earth Planet. Sci. Lett., 148, 207 – 221, 1997.
Shipboard Scientific Party, Site 1222 [online], Proc. Ocean Thompson, A. B., Water in the Earth’s upper mantle, Nature,
Drill. Program Initial Rep., 199, 1 – 57, 2002. (Available at 358, 295 – 302, 1992.
http://www-odp.tamu.edu/publications/199_IR/VOLUME/ Thompson, G., Metamorphic and hydrothermal processes: ba-
CHAPTERS/IR199_15.PDF). salt-seawater interaction, in Oceanic Basalts, edited by P. A.
Sleep, N. H., Thermal history and degassing of the Earth; some Floyd, pp. 148 – 173, Van Nostrand Reinhold, New York,
simple calculations, J. Geol., 87, 671 – 686, 1979. 1991.
Sleep, N. H., and T. J. Wolery, Sediment subduction and al- Toh, H., Electrical conductivity structure of the Izu-Bonin Arc
teration of oceanic crust, a chemical constraint (abstract), revealed by seafloor electromagnetic observations, Ph.D.
Eos Trans. AGU, 60, 392, 1979. thesis, Ocean Research Inst., Univ. of Tokyo, Tokyo, 1993.
Small, C., Global systematics of mid-ocean ridge morphology, Tregoning, P., F. Tan, J. Gilliland, H. McQueen, and K. Lam-
in Faulting and Magmatism at Mid-Ocean Ridges, Geophys. beck, Present-day crustal motion in the Solomon Islands
Monogr. Ser., vol. 106, edited by W. R. Buck et al., pp. 1 – from GPS observations, Geophys. Res. Lett., 25, 3627 –
25, AGU, Washington, D.C., 1998. 3630, 1998.
Sobolev, A. V., and M. Chaussidon, H2O concentrations in Ulmer, P., and V. Trommsdorff, Serpentine stability to mantle
primary melts from supra-subduction zones and midocean depths and subduction-related magmatism, Science, 268,
ridges: Implications for H2O storage and recycling in the 858 – 861, 1995.
mantle, Earth Planet. Sci. Lett., 137, 45 – 55, 1996. Underwood, M. B., et al., Sediment geochemistry, clay miner-
Spivack, A. J., and H. Staudigel, Low-temperature alteration of alogy, and diagenesis: A synthesis of data from Leg 131,
the upper oceanic crust and the alkalinity budget of seawater, Nankai Trough, Proc. Ocean Drill. Program Sci. Results,
Chem. Geol., 115, 239 – 247, 1994. 131, 343 – 363, 1993.
Sprague, D., and H. N. Pollack, Heat flow in the Mesozoic and Vera, E. E., and J. C. Mutter, Crustal structure in the ROSE
Cenozoic, Nature, 285, 393 – 395, 1980. area of the East Pacific Rise: One-dimensional travel time
Staudigel, H., and S. D. King, Ultrafast subduction: The key to inversion of sonobuoys and expanded spread profiles,
slab recycling efficiency and mantle differentiation, Earth J. Geophys. Res., 93, 6635 – 6648, 1988.
Planet. Sci. Lett., 109, 517 – 530, 1992. Von Herzen, R., Geothermal evidence for continuing hydro-
Staudigel, H., S. R. Hart, and S. Richardson, Alteration of the thermal circulation in older (>60 M.y.) ocean crust, in Hy-
oceanic crust: Processes and timing, Earth Planet. Sci. Lett., drology of the Oceanic Lithosphere, edited by E. Davis and
52, 311 – 327, 1981. H. Elderfield, Cambridge Univ. Press, New York, in press,
Staudigel, H., S. R. Hart, H.-U. Schmincke, and B. M. Smith, 2003.
Cretaceous ocean crust at DSDP sites 417 and 418; Carbon von Huene, R., and D. W. Scholl, Observations at convergent
uptake from weathering versus loss by magmatic outgassing, margins concerning sediment subduction, subduction ero-
Geochim. Cosmochim. Acta, 53, 3091 – 3094, 1989. sion, and the growth of continental crust, Rev. Geophys.,
Staudigel, H., T. Plank, B. White, and H.-U. Schmincke, 29, 279 – 316, 1991.
Geochemical fluxes during seafloor alteration of the basal- Wang, K., and E. E. Davis, In the pursuit of hydrothermal
tic upper oceanic crust; DSDP sites 417 and 418, in Sub- circulation and hydration depths of oceanic plates, Eos
duction Top to Bottom, Geophys. Monogr. Ser., vol. 96, Trans. AGU, 83(47), Fall Meet. Suppl., F1252, 2002.
edited by G. E. Bebout et al., pp. 19 – 38, AGU, Washing- Wannamaker, P. E., J. R. Booker, A. G. Jones, A. D. Chave,
ton, D.C., 1996. J. H. Filloux, H. S. Waff, and L. K. Law, Resistivity cross

49 of 50
Geochemistry 3
Geophysics
Geosystems G jarrard: subduction fluxes of water 10.1029/2002GC000392

section through the Juan de Fuca subduction system and its Wyllie, M. R. J., A. R. Gregory, and G. H. F. Gardner, Elastic
tectonic implications, J. Geophys. Res., 94, 14,127 – 14,144, wave velocities in heterogeneous and porous media, Geo-
1989. physics, 21, 41 – 70, 1956.
White, R. S., D. McKenzie, and R. K. O’Nions, Oceanic crustal Zhang, Y., and A. Zindler, Distribution and evolution of car-
thickness from seismic measurements and rare earth element bon and nitrogen in Earth, Earth Planet. Sci. Lett., 117, 331 –
inversions, J. Geophys. Res., 97, 19,683 – 19,715, 1992. 345, 1993.
Wilkens, R. H., N. I. Christensen, and L. Slater, High-pressure Zhou, W., D. R. Peacor, J. C. Alt, R. Van der Voo, and L.-S.
seismic studies of Leg 69 and 70 basalts, Initial Rep. Deep Kao, TEM study of the alteration of interstitital glass in
Sea Drill. Proj., 69, 683 – 686, 1983. MORB by inorganic processes, Chem. Geol., 174, 365 –
Wilkens, R. H., G. J. Fryer, and J. Karsten, Evolution of por- 376, 2001a.
osity and seismic structure of upper oceanic crust: impor- Zhou, W., R. Van der Voo, D. R. Peacor, D. Wang, and Y. Zhang,
tance of aspect ratios, J. Geophys. Res., 96, 17,981 – Low-temperature oxidation in MORB of titanomagnetite to
17,995, 1991. titanomaghemite: A gradual process with implications for
Williams, Q., and R. J. Hemley, Hydrogen in the deep earth, marine magnetic anomaly amplitudes, J. Geophys. Res.,
Ann. Rev. Earth Planet. Sci., 29, 365 – 418, 2001. 106, 6409 – 6421, 2001b.

50 of 50

You might also like