Earth and Planetary Science Letters 398 (2014) 66–76
Contents lists available at ScienceDirect
Earth and Planetary Science Letters
www.elsevier.com/locate/epsl
Deconvolving the controls on the deep ocean’s silicon stable isotope
distribution
Gregory F. de Souza a,∗ , Richard D. Slater a , John P. Dunne b , Jorge L. Sarmiento a
a
b
Program in Atmospheric and Oceanic Sciences, Princeton University, 300 Forrestal Road, Princeton, NJ 08544, USA
National Oceanic and Atmospheric Administration/Geophysical Fluid Dynamics Laboratory, Princeton, NJ 08540, USA
a r t i c l e
i n f o
Article history:
Received 12 February 2014
Received in revised form 22 April 2014
Accepted 23 April 2014
Available online 16 May 2014
Editor: G.M. Henderson
Keywords:
silicon isotopes
marine silicon cycle
Southern Ocean
ocean biogeochemical cycles
general circulation model
a b s t r a c t
We trace the marine biogeochemical silicon (Si) cycle using the stable isotope composition of Si dissolved
in seawater (expressed as δ 30 Si). Open ocean δ 30 Si observations indicate a surprisingly strong influence
of the physical circulation on the large-scale marine Si distribution. Here, we present an ocean general
circulation model simulation that deconvolves the physical and biogeochemical controls on the δ 30 Si
distribution in the deep oceanic interior. By parsing dissolved Si into its preformed and regenerated
components, we separate the influence of deep water formation and circulation from the effects of
biogeochemical cycling related to opal dissolution at depth. We show that the systematic meridional
δ 30 Si gradient observed in the deep Atlantic Ocean is primarily determined by the preformed component
of Si, whose distribution in the interior is controlled solely by the circulation. We also demonstrate that
the δ 30 Si value of the regenerated component of Si in the global deep ocean is dominantly set by oceanic
regions where opal export fluxes to the deep ocean are large, i.e. primarily in the Southern Ocean’s
opal belt. The global importance of this regionally dynamic Si cycling helps explain the observed strong
physical control on the oceanic δ 30 Si distribution, since most of the regenerated Si present within the
deep Atlantic and Indo-Pacific Oceans is in fact transported into these basins by deep waters flowing
northward from the Southern Ocean. Our results thus provide a mechanistic explanation for the observed
δ 30 Si distribution that emphasizes the dominant importance of the Southern Ocean in the marine Si cycle.
© 2014 Elsevier B.V. All rights reserved.
1. Introduction
Diatoms are siliceous phytoplankton with an obligate requirement for silicon (Si) to form their opaline cell walls, or frustules. As a result of their large cell size and boom–bust ecological strategy, diatoms are important contributors to the ocean’s
biological carbon pump (Buesseler, 1998; Henson et al., 2012),
and are the dominant driver of the oceanic Si cycle (Tréguer and
De La Rocha, 2013). The resulting link between the oceanic cycles of silicon and carbon (e.g. Dugdale and Wilkerson, 2001;
Ragueneau et al., 2006) has motivated much research into the controls on the biogeochemical Si cycle, both in the modern ocean and
during past glacial–interglacial cycles (e.g. Dugdale et al., 2002;
Brzezinski et al., 2003; Griffiths et al., 2013; Meckler et al., 2013).
A promising tool for such research is the stable isotope composition of silicon, expressed as δ 30 Si (defined in permil units as
*
Corresponding author. Tel: +1 609 258 0846.
E-mail address: [email protected] (G.F. de Souza).
http://dx.doi.org/10.1016/j.epsl.2014.04.040
0012-821X/© 2014 Elsevier B.V. All rights reserved.
(R sample /R std − 1) × 1000, where R sample is the 30 Si/28 Si isotope
ratio in a sample and R std is this ratio in the standard NBS28).
Diatoms preferentially incorporate the lighter isotopes of Si
into their frustules during silicification (De La Rocha et al., 1997;
Milligan et al., 2004; Sutton et al., 2013). As a result, biological utilization of Si in surface waters enriches dissolved Si in its
heavier isotopes, imparting an elevated δ 30 Si signature to surface
waters that have experienced diatom Si uptake (Varela et al., 2004;
Cavagna et al., 2011). Within the ocean interior, the dissolution of
sinking diatom opal that has been exported from the surface also
influences the δ 30 Si value of dissolved Si, an effect that might be
enhanced by Si isotope fractionation during dissolution (Demarest
et al., 2009). Since these two processes, diatom Si uptake and
opal export, drive the largest Si fluxes in the ocean (Tréguer and
De La Rocha, 2013), the oceanic δ 30 Si distribution bears information on the dominant pathways and processes by which Si is cycled
within the ocean (e.g. Cardinal et al., 2005; Fripiat et al., 2011a;
de Souza et al., 2012a).
Due to isotope fractionation by diatom Si uptake at the surface, δ 30 Si values of dissolved Si reach their maximum at the
surface, and exhibit lower values in the subsurface water column.
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
In addition to this intuitive vertical δ 30 Si gradient, δ 30 Si values
also vary in deep waters at the global scale. Values of δ 30 Si exhibit a systematic inter-basin gradient along the deep limb of
the ocean’s meridional overturning circulation (MOC), decreasing from high values in the deep Atlantic, through intermediate values in the Southern Ocean, to low values in the deep
North Pacific (De La Rocha et al., 2000; Cardinal et al., 2005;
Reynolds et al., 2006). The exact controls on this large-scale deep
water δ 30 Si gradient – and thus the nature of oceanic Si cycling
reflected by this feature – remain incompletely understood. Early
observational and modeling studies invoked a variety of processes
to explain its origin. For instance, De La Rocha et al. (2000) attributed the difference between Atlantic and Pacific deep water
δ 30 Si values to the cumulative effect of the dissolution of low-δ 30 Si
opal along the deep limb of the MOC. A later study by Beucher
et al. (2008), however, used δ 30 Si observations in the deep Pacific
and Southern Oceans to argue that opal dissolving in the deep Pacific bears a higher δ 30 Si value than deep waters. Wischmeyer et
al.’s (2003) early general circulation model simulation, using the
Large-Scale Geostrophic model of Maier-Reimer et al. (1993), was
unable to shed light on this issue, since the model did not produce
the inter-basin deep water δ 30 Si gradient. The box models implemented by Reynolds (2009) did simulate systematic δ 30 Si variation
along the deep limb of the MOC, which he attributed to a combination of physical Si transport by the MOC and high Southern
Ocean opal export fluxes.
Recently, a number of observational studies have documented
strong regional- to basin-scale coherence in δ 30 Si values, implying
a strong physical control on the δ 30 Si distribution at a range of
depths in the water column. Fripiat et al. (2011b) and de Souza
et al. (2012b) demonstrated the importance of physical processes
in communicating the elevated δ 30 Si signal produced at the surface into the ocean interior. These studies revealed that high δ 30 Si
values produced during summer are retained in the deep winter mixed layers of the Southern Ocean, and that this utilization
signal is transported into the subtropical interior by Subantarctic
Mode Water (SAMW) and Antarctic Intermediate Water (AAIW),
consistent with the Si∗ distribution (Si∗ =Si–NO3 ; Sarmiento et al.,
2004). Similarly, equatorial Pacific studies suggest that the Equatorial Undercurrent’s δ 30 Si signature may be traced across 60 degrees of longitude (Beucher et al., 2008, 2011; Ehlert et al., 2012;
Grasse et al., 2013). Furthermore, a dominant control of the circulation has been observed on deep water δ 30 Si variations at the
basin scale: de Souza et al. (2012a) demonstrated that the meridional δ 30 Si gradient within the deep Atlantic Ocean is related not
to opal dissolution, but rather to the quasi-conservative mixing of
Si between North Atlantic Deep Water (NADW) and Antarctic Bottom Water (AABW).
The recent increase in the volume and spatial resolution of
oceanic δ 30 Si observations has thus provided powerful evidence for
the largely conservative behavior of δ 30 Si values in the ocean interior. The implied dominance of the physical circulation on the
large-scale oceanic distribution of δ 30 Si – and hence that of Si –
is somewhat surprising for a biogeochemically-cycled nutrient. In
this study, we aim to identify the processes that determine this
behavior, with a focus on the global deep water δ 30 Si distribution.
We deconvolve the physical and biogeochemical influences on the
Si and δ 30 Si distributions in an ocean general circulation model
(OGCM) simulation that explicitly traces the preformed and regenerated components of Si.
Our conceptual distinction between preformed and regenerated
Si is motivated by the differing controls on their distributions in
the ocean interior. At the ocean’s surface, the balance between Si
supply by the circulation on the one hand, and Si uptake & export
by biology on the other, determines Si concentrations in the mixed
layer. The properties of the winter mixed layer, including its Si
67
concentration and δ 30 Si value, are communicated to the ocean interior during subduction or deep water formation (Stommel, 1979;
de Souza et al., 2012b), giving rise to the preformed component
of the interior ocean Si inventory. Thus, whilst the concentration of preformed Si and its δ 30 Si value reflect the physical–
biogeochemical Si balance in the mixed layer at the sites of deep
water formation, this component enters the ocean interior in dissolved form, without being cycled by biology. Once within the
ocean interior, preformed Si behaves conservatively, such that its
interior concentration and isotopic distributions are influenced
only by the physical circulation. Differences in the δ 30 Si value
of preformed Si sourced from different deep water formation regions thus leads to variations in deep ocean δ 30 Si values that
are purely related to the physical transport of these conservative
source signals. The regenerated component, on the other hand, is
added to the ocean interior by the dissolution of sinking opal that
has been exported from the euphotic zone, representing the nonconservative component of the interior Si inventory that has been
cycled biogeochemically. The distribution of total dissolved Si in
the ocean interior is governed by the combination of these two
components. Our approach of separating preformed and regenerated Si in the context of a model simulation, and tracing their
isotopic compositions, thus allows assessment of the relative importance of physical versus biogeochemical controls on the distribution of Si and its isotopes in the deep sea.
2. Methods
2.1. Model description
The physical ocean model used in this study is the Modular
Ocean Model 3.0 (MOM3; Pacanowski and Griffies, 1999). MOM3
is a z-level primitive-equation ocean general circulation model,
run here with 3.75◦ × 4.5◦ (longitude–latitude) horizontal resolution and 24 vertical levels. Our study uses the P2A configuration
of MOM3 (Gnanadesikan et al., 2004). This configuration conforms
to the observational constraints of low diapycnal diffusivity in the
low-latitude thermocline (e.g. Ledwell et al., 1993, 1998) and enhanced diapycnal mixing in the Southern Ocean (e.g. Naveira Garabato et al., 2004), and forces deep water formation in the model’s Southern Ocean via strong winter salinity-restoring in four
grid cells. It reproduces observed distributions of ocean ventilation
tracers such as radiocarbon and chlorofluorocarbons (CFC-11), as
well as biogeochemical variables such as particle export, dissolved
phosphate and dissolved oxygen with good fidelity (Gnanadesikan
et al., 2004; Matsumoto et al., 2004).
The physical model is coupled to a slightly modified version of
the biogeochemical model of Jin et al. (2006). The representation
of Si cycling follows the OCMIP-2 protocol (Najjar and Orr, 1998)
of restoring nutrient concentrations to observations in the surface
ocean (≤85 m), with an e-folding timescale of 3 months. The surface nutrient-restoring drives a diagnosed Si uptake in the surface
ocean, which is exported as a sinking opal flux that dissolves with
an e-folding lengthscale of 1000 m. The oceanic Si cycle in this
model is closed, with no external inputs from riverine or æolian
sources and no losses to sediment, such that sinking opal reaching
the sea floor is converted to dissolved Si. Unlike the simulations of
Jin et al. (2006), biogeochemical tracers are fully prognostic in the
ocean interior in our simulation.
We have expanded Jin et al.’s (2006) model to also (a) simulate Si isotope fractionation and trace its stable isotope composition, and (b) explicitly trace the preformed component of Si
and its stable isotope composition. Silicon stable isotope composition is traced by including an additional tracer of the isotope
30
Si, scaled to a 30 Si:Si ratio of 1 in order to minimize potential numerical errors associated with the model’s advection scheme
68
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
(cf. Somes et al., 2010). Isotope fractionation during diatom Si uptake in the surface ocean is simulated by scaling the Si uptake flux
diagnosed from nutrient-restoring by the fractionation factor (e.g.
Maier-Reimer, 1993; Schmittner et al., 2013):
30
Si
J prod
(t ) = α · J Si
prod (t ) ·
30
Si(t )/Si(t )
x
where J prod
is the uptake flux of the tracer x,
(1)
α is the fractiona-
tion factor associated with diatom Si uptake, and 30 Si (t ) and Si(t )
are the concentrations of 30 Si and Si respectively. This general formulation represents kinetic Si isotope fractionation without relying
on assumptions of isotope systematics (e.g. open- or closed-system
behavior), which instead emerge from interaction of the biogeochemical model with the circulation. The fractionation factor α is
set to a constant value of 0.9989 (i.e. an isotope effect of −1.1h),
consistent with the isotope effect of diatom Si uptake documented
by the culture studies of De La Rocha et al. (1997) and inferred
from in situ Southern Ocean observations by Fripiat et al. (2011a),
although the recent culture study of Sutton et al. (2013) has shown
that individual diatom species’ isotope effects vary from −0.5h to
−2.1h.
Our model does not consider Si isotope fractionation during
opal dissolution (Demarest et al., 2009). Unlike Si isotope fractionation during uptake, which directly imposes an isotopic signal upon
seawater Si, fractionation during dissolution occurs at the level of
the individual diatom frustule. Since the downward decrease in
opal flux is primarily due to the complete dissolution of more easily dissolved members of the diatom assemblage (Shemesh et al.,
1989; Demarest et al., 2009, and references therein), the isotopic
effect of fractionation during dissolution is likely strongly muted
(Demarest et al., 2009). This inference is supported by observations
that the δ 30 Si value of sinking opal remains constant throughout
the water column (Fripiat et al., 2012), and by preliminary OGCM
results (Gao et al., 2013), which indicate that simulating fractionation during dissolution results in seawater δ 30 Si depth profiles
inconsistent with observations. Based on these considerations, our
model treats 30 Si and Si identically during regeneration.
Simulated silicon stable isotope composition is calculated offline using the simplified formula (30 Si/Si − 1) × 1000 in order to
derive δ 30 Si values. Due to the low abundance of 28 Si, this simplification slightly (∼8%) underestimates δ 30 Si relative to the standard delta notation, well within the uncertainty associated with
the isotope effect of diatom Si uptake (De La Rocha et al., 1997;
Milligan et al., 2004; Sutton et al., 2013).
The preformed component of Si is traced by defining an additional prognostic tracer. Following the definition of preformed
nutrients, preformed Si is set equal to total Si in the surface ocean
(the uppermost 85 m, in which Si uptake takes place), and behaves conservatively in the ocean interior. Its isotopic composition
is traced by including an analogous tracer of preformed 30 Si. Regenerated Si (and 30 Si) are calculated ex post facto by subtracting
preformed Si (and 30 Si) from total Si (and 30 Si). This calculation
ensures mass balance, such that the combination of preformed Si
and regenerated Si always equals total Si in terms of both concentration and isotope composition. In the following, we use italics for
clarity when referring to the components derived from our model’s
tracer deconvolution.
2.2. Simulation: initialization, forcing, and equilibrium diagnostics
As described in Gnanadesikan et al. (2004), the physical model
is forced by heat and salt fluxes from da Silva et al. (1994) including flux corrections diagnosed from restoring surface temperatures and salinities towards observational climatologies (Levitus
and Boyer, 1994; Levitus et al., 1994), with additional salinity
forcing as described in Section 2.1. Wind stress forcing is derived from the ECMWF reanalysis of Trenberth et al. (1989). Initial physical ocean fields are taken from a multi-millennial spinup simulation, whilst biogeochemical tracers are initialized to the
objectively-analyzed annual mean climatologies of World Ocean
Atlas 2009 (WOA09; Garcia et al., 2010a, 2010b). This initialization
yields an oceanic Si inventory of 1.19 × 105 Tmol, similar to the
value of 0.97 × 105 Tmol estimated by Tréguer and De La Rocha
(2013). Oceanic δ 30 Si is initialized to a globally constant value of
+1.2h, the mean δ 30 Si value of the Si-rich deep Southern and
Pacific Oceans (≥2000 m) that dominate the ocean’s Si mass balance (Cardinal et al., 2005; Beucher et al., 2008, 2011; Fripiat et
al., 2011b; de Souza et al., 2012a, 2012b; Grasse et al., 2013).
Tracer fields used as targets for nutrient-restoring are derived from
WOA09’s objectively-analyzed monthly nutrient climatologies.
The model is integrated forward for 5000 model years to
steady-state. Over the last 100 years of the simulation, δ 30 Si values
in the deep Pacific Ocean, the basin slowest to approach equilibrium, vary by <0.0001h. Similarly, the oceanic inventories of
preformed and regenerated Si are near equilibrium, the preformed
inventory increasing by only ∼0.07 Tmol century−1 (a negligible
fraction of the oceanic preformed Si inventory of 6.23 × 104 Tmol),
whilst its isotopic composition (+1.42h) remains constant to
within 0.0003h. The model’s average opal export flux in the last
century of the simulation is 189 Tmol Si yr−1 (with a δ 30 Si value
of +0.98h), slightly different from the value of 178 Tmol yr−1
reported by Jin et al. (2006), likely due to Si-restoring below the
euphotic zone in their model. The value we obtain is at the higher
end of the range of estimates in the literature (69–185 Tmol
yr−1 ; Dunne et al., 2007, and references therein) and consistent
with a recent optimized data-assimilating model of the Si cycle
(171 ± 31 Tmol yr−1 ; Holzer et al., 2014).
3. Results
3.1. Model validation: comparison to δ 30 Si data
The oceanic Si isotope systematics of open-ocean observations
and our model simulation are compared in Fig. 1, which plots
δ 30 Si values against Si concentrations. The model reproduces the
observed non-linear relationship between Si concentration and its
isotopic composition in the global ocean (Fig. 1a). Furthermore, the
model captures a key feature of the regional variability in oceanic
δ 30 Si systematics (Fig. 1b): the simulated isotope systematics of
the Southern Ocean south of 50◦ S faithfully reproduces the observed enhanced trend towards elevated δ 30 Si values at intermediate Si concentrations (20–60 μM). Note, however, that at very low
Si concentrations (<5 μM), the model δ 30 Si extends to higher values than observed. This mismatch may partially reflect analytical
constraints on the observations (i.e. inability to analyze extremely
Si-depleted samples), but also results from the model’s overprediction of subtropical δ 30 Si values (see below).
Fig. 2 shows a latitudinal section of δ 30 Si values in the model’s southeast Pacific at ∼103◦ W, overlain by data from 85◦ W to
110◦ W. A more detailed comparison between the model’s δ 30 Si
field and data from 103◦ W is presented in the Supplementary
Information (Fig. S1). The model exhibits considerable fidelity in
reproducing the observed δ 30 Si distribution in the deep ocean,
as well as in the intermediate/mesopelagic regions of the high
southern latitudes and the tropics, both in terms of the range of
δ 30 Si values and their distribution in the water column. However,
in the subtropical thermocline, the model strongly overestimates
δ 30 Si values compared to the sparse observations in these severely
nutrient-depleted waters (Fig. 2). This discrepancy, which results
from the subduction of preformed Si with greatly elevated δ 30 Si
values at the sites of thermocline ventilation in the model, may
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
69
Fig. 1. Comparison of the δ 30 Si systematics in the open-ocean observational dataset (open blue circles) and the model (red dots). Isotope systematics are shown for (a) the
global ocean and (b) the ocean south of 50◦ S. The model reproduces the observed δ 30 Si systematics in terms of degree of Si and δ 30 Si variability as well as the shape of
their non-linear relationship (panels a, b), and captures the elevated δ 30 Si values at intermediate Si concentrations of 20–60 μM at high southern latitudes of the Southern
Ocean (panel b). Typical 2σ SD analytical reproducibility of the data is ±0.15h, whilst data from different laboratories agree to within ±0.2h (Reynolds et al., 2007). Data
are from De La Rocha et al. (2000, 2011), Varela et al. (2004), Cardinal et al. (2005), Reynolds et al. (2006), Beucher et al. (2008, 2011), Cavagna et al. (2011), Fripiat et al.
(2011a, 2011b), de Souza et al. (2012a, 2012b), Ehlert et al. (2012) and Grasse et al. (2013). (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)
Fig. 2. Model-data comparison of the δ 30 Si distribution in the eastern equatorial and southeastern Pacific Ocean. Good agreement is seen between the observations (colored
dots) and the simulation (color map) in the deep ocean and over the entire water column at tropical and high southern latitudes, in terms of gradients as well as absolute
values (see also Fig. S1). However, the model strongly overestimates δ 30 Si values in the subtropical thermocline and in the shallow tropical ocean (40◦ S–10◦ S; see Section 3.1
for discussion). The model section is from 103W◦ ; data are from Beucher et al. (2008), de Souza et al. (2012b) and Grasse et al. (2013).
result from our nutrient-restoring approach, which leads to nutrient uptake that is temporally smoother than the episodic diatom
blooming in these Si-poor regions (e.g. Brown et al., 2008). Such
episodic Si uptake and export may limit the expression of isotope fractionation due to near-complete mass transfer to the opal
pool. However, it is beyond the scope of this work to include a
discussion of this effect and its potential influence on the δ 30 Si
distribution.
The model’s positive δ 30 Si bias in the subtropical thermocline
does not affect our deep-ocean analysis here. A comparison of
the model’s δ 30 Si field at a depth of ∼2000 m with a compilation of observations shows that the model captures the main
characteristics of the deep ocean δ 30 Si distribution well (Fig. 3).
Note the high δ 30 Si values in the northern North Atlantic associated with the downward penetration of NADW; this Si-poor water
mass bears an elevated δ 30 Si signature of +1.79h, consistent with
de Souza et al. (2012a) observations in the North Atlantic Ocean
(+1.58h to +1.78h). Values of δ 30 Si in the deep Arctic are also
uniformly high at +1.63h, consistent with the elevated δ 30 Si values (+1.69h to +1.86h) observed in Nordic Sea overflows by
de Souza et al. (2012a). The high-δ 30 Si North Atlantic signal propagates into the South Atlantic and even south of Africa, extending
in a tongue of very slightly elevated δ 30 Si values (Fig. 4b) associated with a tongue of low [Si] (Fig. 4a). Southward of this
feature, the Si-rich deep Southern Ocean exhibits uniformly low
δ 30 Si values, with the model’s average value of +1.14h being
70
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
Fig. 3. Comparison of the δ 30 Si distribution at ∼2000 m water depth in observations (colored dots) and model simulation (color map). The simulated distribution reproduces
the elevated δ 30 Si values observed in the northern North Atlantic, low δ 30 Si values in the Southern Ocean and the meridional δ 30 Si gradient in the deep Atlantic Ocean. Note
that the model does not reproduce the low δ 30 Si values (below +1h) seen in the observational dataset in the high-latitude North Pacific. Observations are from De La Rocha
et al. (2000), Cardinal et al. (2005), Reynolds et al. (2006), Beucher et al. (2008; 2011), Fripiat et al. (2011a, 2011b), de Souza et al. (2012a, 2012b) and Grasse et al. (2013).
highly consistent with observations (+1.2h; Cardinal et al., 2005;
Fripiat et al., 2011b; de Souza et al., 2012b). In contrast to observations, model δ 30 Si values do not decrease between the deep
Southern Ocean and the North Pacific. The model’s North Pacific
exhibits δ 30 Si values largely similar to those in the Southern Ocean,
and higher than observations by ∼0.3h (De La Rocha et al., 2000;
Reynolds et al., 2006). The model is also insufficiently Si-rich in
the North Pacific, with concentrations ranging up to 130 μM at
2000 m, contrasting with observed concentrations of >170 μM
(Garcia et al., 2010b).
3.2. Distribution of preformed and regenerated components
Having documented the model’s skill at reproducing both
oceanic δ 30 Si systematics and the main features of the deep δ 30 Si
distribution, we now turn to the distributions of the preformed and
regenerated components of Si. Here and in the following discussion
of the deep ocean, we present model results from a depth level
of ∼2000 m, which best illustrates the main features of the deep
ocean Si and δ 30 Si distributions (additional figures are in the Supplementary Information). The maps at ∼2000 m depth in Fig. 4
illustrate that the distributions of both preformed and regenerated
Si differ from that of total Si, and bear a number of interesting
features central to the deconvolution of the Si cycle. Preformed Si
constitutes almost 100% of total Si at 2000 m at the primary sites
of deep water formation in the model, the North Atlantic and the
Weddell Sea, where deep waters are freshly ventilated from the
surface (Figs. 4c, e). Its contribution decreases away from these
regions to about half of total Si in the deep Pacific (the global preformed Si inventory is 52% of the total Si inventory). The two sites
of deep water formation are associated with preformed Si bearing
strongly differing isotope compositions (Fig. 4d): in the Weddell
Sea, preformed Si has a low δ 30 Si signature of +1.29h, whereas the
high-latitude North Atlantic has a preformed δ 30 Si value of +2.1h.
The high North Atlantic values propagate southward through the
Atlantic basin along its western boundary, and as a tongue of elevated δ 30 Si values (> +1.4h) extending eastward south of Africa.
Over the entire Indo-Pacific, the δ 30 Si value of deep water preformed Si is essentially constant (+1.38h).
Unlike the systematic inverse relationship between the concentration and δ 30 Si value of the preformed component (Figs. 4c,
d), the deep ocean distributions of regenerated Si and its isotope
composition show little relation (Figs. 4f, g). The concentration of
regenerated Si at 2000 m is, as expected, lowest at the sites of deep
water formation (where Si is almost entirely preformed), and increases steadily along the deep limb of the MOC (Figs. 4f, h). In
strong contrast to this steady progression in concentrations of regenerated Si, regenerated δ 30 Si values display remarkable regional
constancy (Fig. 4g). Apart from a small region of elevated δ 30 Si values in the northern North Atlantic, the δ 30 Si value of regenerated
Si in the deep Atlantic, Indian and Pacific Oceans is exceptionally uniform: integrated from 2000 m to the sea floor, the δ 30 Si
value of regenerated Si in these basins ranges between +0.91h
and +0.93h. In the deep Southern Ocean south of 50◦ S, regenerated Si bears a lower δ 30 Si value, averaging +0.82h in a roughly
zonal band around Antarctica. The deep Arctic Ocean, in contrast,
exhibits an elevated regenerated δ 30 Si value of +1.48h.
4. Discussion
The deconvolution of deep ocean Si into its physically-controlled
preformed component and its biogeochemically-cycled regenerated
component enables us to take a detailed look at the manner in
which these two controls interact to determine the oceanic Si and
δ 30 Si distributions. We begin our discussion with a consideration
of the most striking feature of the deep δ 30 Si distribution, the
meridional δ 30 Si gradient in the deep Atlantic Ocean.
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
71
Fig. 4. Simulated distributions of Si concentration (left column), isotopic composition (center column) and component fractions (right column) at a model depth of 2063 m.
(a) concentration of total Si; (b) δ 30 Si value of total Si; (c) concentration of preformed Si; (d) δ 30 Si value of preformed Si; (e) fractional contribution of preformed Si to total Si;
(f) concentration of regenerated Si; (g) δ 30 Si value of regenerated Si; (h) fractional contribution of regenerated Si to total Si. Note the differing color bars between panel a and
panels c and f. δ 30 Si color bar applies to entire center column.
4.1. Influence of preformed Si on the deep Atlantic δ 30 Si distribution
It is apparent from Fig. 4 that the strongest δ 30 Si signals in the
deep Atlantic are found in preformed Si, which exhibits a δ 30 Si difference of ∼0.8h between the high-latitude North Atlantic and
the Weddell Sea. Indeed, the close similarity between the spatial distribution of total and preformed δ 30 Si in the deep Atlantic
Ocean (Figs. 4b,d) implies that the systematic δ 30 Si variability here
is primarily determined by the preformed component. This strong
preformed influence on total δ 30 Si is clearly illustrated by the linear correlation (r 2 = 0.91) with a slope near unity between the
isotopic compositions of preformed Si and total Si (Fig. 5).
Preformed Si is transported into the deep ocean from the surface by deep water formation, a result of the buoyancy loss of
surface waters at high latitudes. Since it is, by definition, conservative within the ocean interior, the preformed Si distribution in the
deep ocean is determined solely by the circulation. The preformed
δ 30 Si gradient in the deep Atlantic Ocean (Fig. 4d) thus reflects
a difference in preformed δ 30 Si values between NADW and AABW,
the two sources of deep water in the Atlantic. This preformed δ 30 Si
difference can be related to the degree of Si utilization in the formation regions of NADW and AABW. North Atlantic Deep Water
forms from Si-depleted surface waters of the high North Atlantic
and the Nordic seas (Dickson and Brown, 1994). Its high preformed
δ 30 Si value reflects the elevated surface δ 30 Si signature of these
waters due to isotope fractionation during Si uptake by diatoms.
de Souza et al. (2012a) suggested that this elevated δ 30 Si value is
ultimately derived from the northward transport of the high-δ 30 Si
signature of SAMW and AAIW in the upper limb of the MOC.
Whilst our physical model has been shown to produce significant
inter-hemispheric nutrient transport via this pathway (Palter et al.,
2010), it is not possible to firmly establish such a causal relationship in the absence of additional model diagnostics.
The low preformed δ 30 Si signature of AABW, on the other hand,
is a result of its formation from nutrient-replete waters of the
high-nutrient, low chlorophyll Southern Ocean that have experienced little Si utilization (Huhn et al., 2008; de Souza et al.,
2012a). The lack of high-latitude Si drawdown imparts AABW a
low preformed δ 30 Si value and a preformed Si concentration almost
an order of magnitude higher than that of NADW (Fig. 4c). This
high preformed Si content gives AABW considerable leverage during mixing, allowing it to strongly affect the deep preformed δ 30 Si
distribution as it flows northward in the abyssal Atlantic and mixes
diapycnally into the overlying water column. Fig. 4d shows that the
72
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
high-δ 30 Si preformed signature of NADW, transported southward in
a deep western boundary current (e.g. Rhein, 1994), is gradually
diluted by the admixture of preformed Si from AABW.
A closer look at the correlation between the δ 30 Si values of preformed and total Si in the deep Atlantic (Fig. 5) reveals that the
slope of this correlation is less than unity (m = 0.83), indicating
that total δ 30 Si values vary less strongly than preformed δ 30 Si. This
reduced gradient is the result of the influence of regenerated Si on
the total δ 30 Si distribution. Despite the fact that regenerated δ 30 Si
values bear a rather complex relationship to total δ 30 Si in the deep
Atlantic (Fig. 5), and exhibit their strongest variability in this basin
(Fig. 4g), the primary influence of regenerated Si is to homogenize
Fig. 5. Relationship between the δ 30 Si values of total Si and its preformed (red open
circles) and regenerated (blue open squares) components in the deep Atlantic Ocean.
Note the strong linear correlation (r 2 = 0.91) between total δ 30 Si and preformed
δ 30 Si; the slope of a least-squares fit to this relationship is close to, but less than
unity (m = 0.83). This figure shows data from 2063 m for clarity; essentially identical systematics are seen over the entire deep Atlantic. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of
this article.)
total δ 30 Si values in the deep Atlantic, slightly damping the strong
preformed δ 30 Si signal. In the rest of the deep ocean, regenerated
δ 30 Si values display remarkable regional homogeneity, and it is to
an explanation of this feature that we now turn.
4.2. Circulation control on the regenerated component
The regenerated component of Si is added to the ocean interior
by dissolution of sinking opal. Fig. 6a shows the latitudinal distribution of the model’s zonally-averaged opal export flux: fluxes are
high in the Southern Ocean south of 50◦ S and low elsewhere, with
the exception of a minor peak at northern subpolar latitudes. This
latitudinal distribution is consistent with Dunne et al.’s (2007) estimate from satellite observations and biogeochemical data products,
although tropical export fluxes are underestimated in the model
(the mean low-latitude opal flux is 3% of the mean Southern Ocean
flux in our simulation, as opposed to 10% in Dunne et al., 2007; see
the Supplementary Information).
In Fig. 6b, we compare the flux-weighted mean δ 30 Si value of
the sinking opal flux with the δ 30 Si value of regenerated Si in the
deep ocean. These isotopic distributions clearly document a fundamental feature of the oceanic Si distribution that is not apparent from concentrations alone. The δ 30 Si value of regenerated Si in
the global deep ocean north of 50◦ S is +0.93h. This δ 30 Si value
is considerably lower than that of the opal flux sinking into the
deep ocean here, which has a flux-weighted average δ 30 Si value
of +1.70h. This large disparity between the isotopic composition
of Si being added to the deep ocean north of 50◦ S by sinking opal
and that of regenerated Si present within this volume demonstrates,
unambiguously, that a dominant fraction of the regenerated Si in
the Atlantic, Pacific and Indian basins (cf. Fig. 4g) is not added to
the deep ocean locally. This low-δ 30 Si regenerated Si must in fact
be transported into these basins from a region where the dissolving opal flux releases regenerated Si with a low δ 30 Si value to the
deep ocean. From Figs. 6a and b, it is clear that this region must
be the Southern Ocean. As we detail in Section 4.3 below, our
δ 30 Si deconvolution thus reveals with striking clarity not only that
the Southern Ocean is dominantly important in the global Si cycle
Fig. 6. (a) Zonally-averaged opal export flux at 85 m over the last century of the simulation. (b) δ 30 Si value of regenerated Si (blue line) integrated over the deep ocean
(2063 m to bottom) in year 5000 of the simulation, compared to the flux-weighted average δ 30 Si value of the opal flux sinking past the upper boundary of this volume over
the last century of the simulation (solid red line). These values are averaged over ∼20◦ of latitude in order to emphasize the characteristic regional δ 30 Si value of the sinking
flux; the dashed red line shows the model distribution of sinking opal δ 30 Si. With the exception of the Arctic Ocean (>70◦ N), the δ 30 Si value of regenerated Si in the deep
ocean north of ∼50◦ S is too low to be explained by the local dissolution of sinking opal. (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
73
Table 1
Values of parameters used in the mixing equations to calculate the Southern Ocean contribution to the global deep water Si inventory, and details of their calculation.
Parameter
Value
Remarks
pre
δ SiAABW
30 pre
δ SiNADW
regen
δ 30 SiSO
regen
δ 30 SinonSO
+1.29h
δ 30 Si value of preformed Si within (80◦ S, 50◦ W), (65◦ S, 15◦ W), from 1500 to 5000 m, year 5000
+2.08h
δ 30 Si value of preformed Si within (70◦ N, 60◦ W), (40◦ N, 30◦ W), from 1500 to 2000 m, year 5000
+0.82h
δ 30 Si value of total opal flux at 2063 m South of 50◦ S, years 4900–5000
+1.70h
δ 30 Si value of total opal flux at 2063 m North of 50◦ S, years 4900–5000
30
(cf. Sarmiento et al., 2007), but also that due to this localized overrepresentation, the distribution of regenerated Si and its isotopic
composition in the global deep ocean is strongly influenced by the
northward spreading of deep waters from the Southern Ocean.
4.3. Quantifying the influence of the Southern Ocean
As discussed above, the δ 30 Si distributions of both the preformed and the regenerated components of Si document a strong
Southern Ocean influence on the deep Si and δ 30 Si distributions.
Here, we use simple isotope mixing relationships in order to quantify this influence in the context of our simulation. The isotope
composition of preformed Si in the deep sea is determined by conservative mixing of preformed Si from NADW and AABW, and can
be represented as:
pre
pre
δ 30 Sipre = δ 30 SiAABW · X AABW + δ 30 SiNADW · (1 − X AABW )
(2)
pre
where δ 30 Sipre is the isotopic composition of preformed Si, δ 30 Six
is the preformed δ 30 Si signature of water mass x, and X AABW is the
proportion of preformed Si contributed by AABW (errors in X AABW
introduced by using δ -values in Eq. (2) are <0.001). Eq. (2) can be
solved for X AABW at each grid point in the deep ocean, giving the
proportion of preformed Si derived from the Southern Ocean. Similarly, we can approximately parse the proportion of regenerated
Si added to the deep sea by the high opal fluxes of the Southern
Ocean by using a similar mixing equation:
opal
opal
δ 30 Siregen = δ 30 SiSO · X SO + δ 30 SinonSO · (1 − X SO )
(3)
where δ 30 Siregen is the isotopic composition of regenerated Si,
opal
δ 30 Six
is the flux-weighted mean δ 30 Si value of sinking opal in
region x, SO stands for the Southern Ocean south of 50◦ S, nonSO
is the global ocean North of 50◦ S, and X SO is the proportion of regenerated Si contributed by Southern Ocean opal fluxes. Note that
the use of average δ 30 Si values for the opal fluxes North and South
of 50◦ S means that this calculation is only approximative. Parameter values used in the formulation of Eqs. (2) and (3) are given in
Table 1.
Figs. 7a, b show the distributions of X AABW and X SO , respectively, in the deep ocean at ∼2000 m. The distribution of X AABW
(Fig. 7a) demonstrates a strong influence of the upward mixing
of abyssal AABW in the Atlantic Ocean, with numerous regions
even in the North Atlantic exhibiting >50% Si contribution from
AABW at this depth level. Whilst this is broadly consistent with
observations (van Aken and de Boer, 1995; van Aken, 2000), our
model overestimates the influence of AABW at this depth in the
Atlantic Ocean, since the physical model produces insufficientlydense NADW that is consequently restricted to shallower levels
than observed, allowing the model’s abyssal Atlantic to fill with
Si-rich AABW. Elsewhere in the deep ocean, the high preformed
Si content of AABW results in this water mass contributing the
dominant fraction (>80%) of deep preformed Si. The sensitivity of
regenerated δ 30 Si to the low-δ 30 Si Southern Ocean opal fluxes is
revealed by the X SO distribution in Fig. 7b: our relatively coarse
unmixing calculation (Eq. (3)) picks up the increased proportion
of locally-regenerated Si under the coastal upwelling regions of
the eastern Atlantic and Pacific, and in the recirculating waters of
the Angola Basin, west of southern Africa. Interestingly, regenerated
δ 30 Si values are so sensitive to the Southern Ocean opal flux that
X S O integrated over the deep ocean traces the inflows of Southern
Ocean water into the abyssal southwestern Atlantic, Indian and Pacific Oceans (Fig. S4).
In a final calculation, we estimate the amount of total Si that
is derived from the Southern Ocean by combining X AABW and X SO
with the distributions of preformed Si and regenerated Si:
SiSO = Sipre · X AABW + Siregen · X SO
(4)
where SiSO is the concentration of Si derived from the Southern
Ocean and Sipre (Siregen ) is concentration of preformed (regenerated) Si. We show the result of this calculation in Fig. 7c, which
illustrates the deep distribution of the proportion of total Si derived from the Southern Ocean (i.e. SiSO :Si). The combination of
the high preformed Si content of AABW (Fig. 7a) and the strong
Southern Ocean control on regenerated Si (Figs. 6b, 7b) results in
an overwhelmingly large contribution of Southern Ocean-sourced
Si to the global deep ocean: integrated from 2000 m to the sea
floor, 89% of Si in the deep ocean north of 50◦ S is of Southern
Ocean origin.
4.4. Implications for the deep ocean’s δ 30 Si distribution
Our deconvolution of Si into its preformed and regenerated
components allows assessment of the mechanisms driving the
global deep water δ 30 Si gradient, and their implications for the
marine biogeochemical Si cycle. To a considerable extent, the influence of the physical circulation on the deep δ 30 Si distribution
results from the important contribution of preformed Si, whose
isotopic composition bears strongly contrasting signals in NADW
and AABW. However, our simulation also illustrates that, as a result of the highly non-uniform distribution of opal export, even
the distribution of the biogeochemically-cycled regenerated component of Si is significantly determined by the circulation. These
two mechanisms combine to produce the observed strong water
mass control on the δ 30 Si distribution.
Furthermore, our analysis in Section 4.3 suggests that this
quasi-conservation of δ 30 Si values within the ocean interior reflects
the dominance of Southern Ocean processes in Si cycling. Firstly,
AABW plays a key role in determining the distribution of the
circulation-controlled preformed component due to its high preformed Si content. Secondly, the elevated opal fluxes in the Southern Ocean contribute the dominant fraction of the deep ocean’s
regenerated Si inventory. As a result, the mass-balance of Si in
deep waters, and hence its stable isotope composition, is strongly
determined by Si – both preformed and regenerated – from the
Southern Ocean. This inference is consistent with the recent modeling study of Holzer et al. (2014), who find that the ocean’s
preformed and regenerated Si inventories are dominantly sourced
from the Southern Ocean south and north of the Antarctic Divergence, respectively (see also the Supplementary Information). Since
the preformed and regenerated Si inventories cannot be calculated
from data, the finding of Holzer et al. (2014) could not be validated
against observations. Our study fills this gap by tying the Southern
74
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
Fig. 7. Quantifying the influence of the Southern Ocean on the simulated Si distribution at 2063 m. (a) Fraction of preformed Si sourced from AABW, calculated from the
preformed δ 30 Si distribution using Eq. (2). (b) Fraction of regenerated Si added south of 50◦ S, calculated from the regenerated δ 30 Si distribution using Eq. (3). (c) Fraction of
total Si sourced from the Southern Ocean, calculated in Eq. (4) by combining the information in panels a and b.
Ocean’s controlling role on these two components directly to the
observed δ 30 Si distribution.
The high Si content of the Southern Ocean and its low δ 30 Si
value are mechanistically linked to each other, and to the MOC.
Strong southern westerly winds drive upwelling in the Southern
Ocean (Marshall and Speer, 2012), producing a dynamic, nutrientrich system that favors diatom-dominated ecosystems (Ragueneau
et al., 2000). Heavily-silicified diatoms strip Si from these waters
as they are transported northward by Ekman drift (Brzezinski et
al., 2003; Sarmiento et al., 2004), efficiently exporting it as opal
to depth, trapping Si within the Southern Ocean (Sarmiento et
al., 2007; Holzer et al., 2014). Since this opal bears a low δ 30 Si
value (Cardinal et al., 2007; Egan et al., 2012), it lowers the deep
Southern Ocean’s δ 30 Si signature, whilst leading to the transport
of a residual high-δ 30 Si signal into the regions where SAMW and
AAIW are formed (Fripiat et al., 2011b; de Souza et al., 2012b).
Importantly, the shoaling of deep isopycnals in the Southern Ocean
favors the regeneration of the low-δ 30 Si signal in dense waters that
will exit the Southern Ocean as AABW (cf. Lumpkin and Speer,
2007; Talley, 2013), even if a large proportion of the exported
opal dissolves within the upper water column, as suggested by
Holzer et al. (2014). The low-δ 30 Si, high-Si signatures of AABW’s
preformed and regenerated components thus reflect the Southern
Ocean’s unique physical-biogeochemical regime.
Our results thus reconcile the intuitive expectation that the
oceanic δ 30 Si distribution should reflect biological uptake and opal
dissolution with the observed strong hydrographic control (e.g.
Fripiat et al., 2011b; de Souza et al., 2012a, 2012b). Returning to
the published hypotheses put forward for the origin of the interbasin deep water δ 30 Si gradient, we see that explanations invoking the influence of opal dissolution (De La Rocha et al., 2000;
Beucher et al., 2008) only capture part of the picture. Dissolution of opal does indeed add regenerated Si with a low δ 30 Si
value to the deep ocean (our model’s global opal flux has a δ 30 Si
value of +0.98h); however, since most regenerated Si is contained in waters of Southern Ocean origin, the isotopic composition of this component is rather homogeneous globally (Fig. 4g),
and contributes little to the large-scale δ 30 Si gradient. In this regard, our simulation agrees with Reynolds’ (2009) box model analysis in highlighting the importance of the physical transport of a
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
high-δ 30 Si signal in NADW, whilst also emphasizing the role of
the contrasting low-δ 30 Si signal in AABW, one that our model indicates is reflective of the strong trapping of Si in the Southern
Ocean.
However, whilst our model performs considerably better at reproducing the oceanic δ 30 Si distribution than the OGCM simulation
of Wischmeyer et al. (2003) (see the Supplementary Information),
it does not reproduce the high Si concentrations and low δ 30 Si values observed in the deep North Pacific (De La Rocha et al., 2000;
Reynolds et al., 2006). This isotopic feature is poorly constrained,
and its origin remains both enigmatic and debated in the literature (Beucher et al., 2008; de Souza et al., 2012b; Grasse et al.,
2013). Our results suggest one possible origin for such a low-δ 30 Si
signature. Higher nutrient supply to the surface subarctic Pacific
than simulated by our model would lead to an increased inventory of locally-regenerated Si in the deep North Pacific, as a result of larger opal fluxes (cf. Gnanadesikan and Toggweiler, 1999;
Gnanadesikan et al., 2002). Since Si utilization in the subarctic Pacific is incomplete (e.g. Whitney, 2011), the regenerated Si released
by dissolution of this opal should possess a low δ 30 Si value, such
that it would lower δ 30 Si values in the mid-depth North Pacific.
Such an influence would mean that the contribution of regenerated
Si to the global deep δ 30 Si gradient is greater than our simulation
indicates. Alternatively, however, deep North Pacific δ 30 Si values
may be influenced by Si derived from hydrothermal sources (e.g.
Johnson et al., 2006). The resolution of this question must await
a clearer observational picture of the North Pacific δ 30 Si distribution.
5. Conclusions
By parsing dissolved Si into its preformed and regenerated components in an OGCM simulation, we have separated the influence
of physical and biogeochemical processes on the deep ocean’s δ 30 Si
distribution. The results indicate a strong control of the preformed
component of Si, whose interior distribution is determined solely
by the circulation, on the systematic meridional δ 30 Si gradient in
the deep Atlantic Ocean. Furthermore, we have shown that, due to
the pronounced regional differences in opal export flux, the δ 30 Si
value of regenerated Si acts as a sensitive tracer of oceanic regions
with large opal export fluxes, which in the model are almost completely restricted to the Southern Ocean. The importance of the
Southern Ocean in cycling Si helps explain the strong physical control on the oceanic Si distribution indicated by δ 30 Si observations:
the small opal fluxes north of the Southern Ocean’s opal belt lead
to the surprising result that most of the regenerated Si in the deep
Atlantic and Indo-Pacific Oceans is in fact transported into these
basins in dissolved form, by deep waters flowing northwards from
the Southern Ocean. Our model’s mechanistic explanation of the
observed water mass control on the oceanic δ 30 Si distribution thus
provides further support for the dominance of the Southern Ocean
in the oceanic Si cycle.
Acknowledgements
Fruitful discussions with Irina Marinov, Mark Brzezinski, Ben
Reynolds and Robbie Toggweiler are gratefully acknowledged.
Feedback from Florian Wetzel helped considerably improve an
earlier version of this manuscript. The authors thank the observational δ 30 Si community for generously sharing data, Damien
Cardinal and an anonymous reviewer for their constructive reviews, and Gideon Henderson for editorial handling. This work
was supported by Swiss National Science Foundation postdoctoral fellowship PBEZP2-140169 granted to GFDS, and NOAA grant
NA11OAR4310066 to JLS.
75
Appendix A. Supplementary material
Supplementary material related to this article can be found online at http://dx.doi.org/10.1016/j.epsl.2014.04.040.
References
Beucher, C.P., Brzezinski, M.A., Jones, J.L., 2008. Sources and biological fractionation of silicon isotopes in the Eastern Equatorial Pacific. Geochim. Cosmochim.
Acta 72, 3063–3073.
Beucher, C.P., Brzezinski, M.A., Jones, J.L., 2011. Mechanisms controlling silicon
isotope distribution in the Eastern Equatorial Pacific. Geochim. Cosmochim.
Acta 75, 4286–4294.
Brown, S.L., Landry, M.R., Selph, K.E., Jin Yang, E., Rii, Y.M., Bidigare, R.R., 2008.
Diatoms in the desert: plankton community response to a mesoscale eddy in
the subtropical North Pacific. Deep-Sea Res., Part 2, Top. Stud. Oceanogr. 55,
1321–1333.
Brzezinski, M.A., Dickson, M.L., Nelson, D.M., Sambrotto, R., 2003. Ratios of Si, C and
N uptake by microplankton in the Southern Ocean. Deep-Sea Res., Part 2, Top.
Stud. Oceanogr. 50, 619–633.
Buesseler, K.O., 1998. The decoupling of production and particulate export in the
surface ocean. Glob. Biogeochem. Cycles 12, 297–310.
Cardinal, D., Alleman, L.Y., Dehairs, F., Savoye, N., Trull, T.W., André, L., 2005. Relevance of silicon isotopes to Si-nutrient utilization and Si-source assessment
in Antarctic waters. Glob. Biogeochem. Cycles 19. http://dx.doi.org/10.1029/
2004GB002364.
Cardinal, D., Savoye, N., Trull, T.W., Dehairs, F., Kopczynska, E.E., Fripiat, F., et al.,
2007. Silicon isotopes in spring Southern Ocean diatoms: large zonal changes
despite homogeneity among size fractions. Mar. Chem. 106, 46–62.
Cavagna, A.-J., Fripiat, F., Dehairs, F., Wolf-Gladrow, D.A., Cisewski, B., Savoye, N., et
al., 2011. Silicon uptake and supply during a Southern Ocean iron fertilization
experiment (EIFEX) tracked by Si isotopes. Limnol. Oceanogr. 56, 147–160.
da Silva, A., Young, A.C., Levitus, S., 1994. Atlas of Surface Marine Data 1994, vol. 1.
National Oceanic and Atmospheric Administration. Silver, Spring, MD.
De La Rocha, C.L., Brzezinski, M.A., DeNiro, M.J., 1997. Fractionation of silicon isotopes by marine diatoms during biogenic silica formation. Geochim. Cosmochim.
Acta 61, 5051–5056.
De La Rocha, C.L., Brzezinski, M.A., DeNiro, M.J., 2000. A first look at the distribution
of the stable isotopes of silicon in natural waters. Geochim. Cosmochim. Acta 64,
2467–2477.
De La Rocha, C.L., Bescont, P., Croguennoc, A., Ponzevera, E., 2011. The silicon isotopic composition of surface waters in the Atlantic and Indian sectors of the
Southern Ocean. Geochim. Cosmochim. Acta 75, 5283–5295.
de Souza, G.F., Reynolds, B.C., Rickli, J., Frank, M., Saito, M.A., Gerringa, L.J.A., et
al., 2012a. Southern Ocean control of silicon stable isotope distribution in the
deep Atlantic Ocean. Glob. Biogeochem. Cycles 26. http://dx.doi.org/10.1029/
2011gb004141.
de Souza, G.F., Reynolds, B.C., Johnson, G.C., Bullister, J.L., Bourdon, B., 2012b. Silicon
stable isotope distribution traces Southern Ocean export of Si to the eastern
South Pacific thermocline. Biogeosciences 9, 4199–4213.
Demarest, M.S., Brzezinski, M.A., Beucher, C.P., 2009. Fractionation of silicon isotopes
during biogenic silica dissolution. Geochim. Cosmochim. Acta 73, 5572–5583.
Dickson, R.R., Brown, J., 1994. The production of North Atlantic Deep Water: Sources,
rates, and pathways. J. Geophys. Res., Oceans 99, 12319–12341.
Dugdale, R.C., Wilkerson, F.P., 2001. Sources and fates of silicon in the ocean: the role of diatoms in the climate and glacial cycles. Sci. Mar. 65 (Suppl. 2), 141–152.
Dugdale, R.C., Wischmeyer, A.G., Wilkerson, F.P., Barber, R.T., Chai, F., Jiang, M.S., et
al., 2002. Meridional asymmetry of source nutrients to the equatorial Pacific
upwelling ecosystem and its potential impact on ocean-atmosphere CO2 flux;
a data and modeling approach. Deep-Sea Res., Part 2, Top. Stud. Oceanogr. 49,
2513–2531.
Dunne, J.P., Sarmiento, J.L., Gnanadesikan, A., 2007. A synthesis of global particle
export from the surface ocean and cycling through the ocean interior and on the
seafloor. Glob. Biogeochem. Cycles 21. http://dx.doi.org/10.1029/2006GB002907.
Egan, K.E., Rickaby, R.E.M., Leng, M.J., Hendry, K.R., Hermoso, M.l., Sloane, H.J., et al.,
2012. Diatom silicon isotopes as a proxy for silicic acid utilisation: a Southern
Ocean core top calibration. Geochim. Cosmochim. Acta 96, 174–192.
Ehlert, C., Grasse, P., Mollier-Vogel, E., Böschen, T., Franz, J., de Souza, G.F., et al.,
2012. Factors controlling the silicon isotope distribution in waters and surface
sediments of the Peruvian coastal upwelling. Geochim. Cosmochim. Acta 99,
128–145.
Fripiat, F., Cavagna, A.-J., Savoye, N., Dehairs, F., André, L., Cardinal, D., 2011a. Isotopic constraints on the Si-biogeochemical cycle of the Antarctic Zone in the
Kerguelen area (KEOPS). Mar. Chem. 123, 11–22.
Fripiat, F., Cavagna, A.-J., Dehairs, F., Speich, S., André, L., Cardinal, D., 2011b. Silicon
pool dynamics and biogenic silica export in the Southern Ocean, inferred from
Si-isotopes. Ocean Sci. 7, 533–547.
Fripiat, F., Cavagna, A.J., Dehairs, F., de Brauwere, A., André, L., Cardinal, D., 2012.
Processes controlling the Si-isotopic composition in the Southern Ocean and application for paleoceanography. Biogeosciences 9, 2443–2457.
76
G.F. de Souza et al. / Earth and Planetary Science Letters 398 (2014) 66–76
Gao, S., Voelker, C., Wolf-Gladrow, D., 2013. Fractionation during biogenic silicon
dissolution: consequences for global marine silicon isotopes distribution – a
modelling study. in: ASLO 2013, Aquatic Sciences Meeting, Poster #400, New
Orleans, USA.
Garcia, H.E., Locarnini, R.A., Boyer, T.P., Antonov, J.I., 2010a. World Ocean Atlas 2009,
vol. 3: Dissolved Oxygen, Apparent Oxygen Utilization, and Oxygen Saturation.
U.S. Government Printing Office, Washington, DC.
Garcia, H.E., Locarnini, R.A., Boyer, T.P., Antonov, J.I., 2010b. World Ocean Atlas, 2009,
vol. 4: Nutrients (phosphate, nitrate, silicate), U.S. Government Printing Office,
Washington DC..
Gnanadesikan, A., Toggweiler, J.R., 1999. Constraints placed by silicon cycling on vertical exchange in general circulation models. Geophys. Res. Lett. 26, 1865–1868.
Gnanadesikan, A., Slater, R.D., Gruber, N., Sarmiento, J.L., 2002. Oceanic vertical exchange and new production: a comparison between models and observations.
Deep-Sea Res., Part 2, Top. Stud. Oceanogr. 49, 363–401.
Gnanadesikan, A., Dunne, J., Key, R., Matsumoto, K., Sarmiento, J.L., Slater, R., et al.,
2004. Oceanic ventilation and biogeochemical cycling: understanding the physical mechanisms that produce realistic distributions of tracers and productivity.
Glob. Biogeochem. Cycles 18. http://dx.doi.org/10.1029/2003GB002097.
Grasse, P., Ehlert, C., Frank, M., 2013. The influence of water mass mixing on the
dissolved Si isotope composition in the Eastern Equatorial Pacific. Earth Planet.
Sci. Lett. 380, 60–71.
Griffiths, J.D., Barker, S., Hendry, K.R., Thornalley, D.J.R., van de Flierdt, T., Hall, I.R.,
et al., 2013. Evidence of silicic acid leakage to the tropical Atlantic via Antarctic Intermediate Water during Marine Isotope Stage 4. Paleoceanography 28,
307–318.
Henson, S.A., Sanders, R., Madsen, E., 2012. Global patterns in efficiency of particulate organic carbon export and transfer to the deep ocean. Glob. Biogeochem.
Cycles 26, GB1028. http://dx.doi.org/10.1029/2011gb004099.
Holzer, M., Primeau, F.W., DeVries, T., Matear, R., 2014. The Southern Ocean silicon
trap: data-constrained estimates of regenerated silicic acid, trapping efficiencies,
and global transport paths. J. Geophys. Res., Oceans 119. http://dx.doi.org/10.
1002/2013jc009356.
Huhn, O., Hellmer, H.H., Rhein, M., Rodehacke, C., Roether, W., Schodlok, M.P., et al.,
2008. Evidence of deep- and bottom-water formation in the western Weddell
Sea. Deep-Sea Res., Part 2, Top. Stud. Oceanogr. 55, 1098–1116.
Jin, X., Gruber, N., Dunne, J.P., Sarmiento, J.L., Armstrong, R.A., 2006. Diagnosing
the contribution of phytoplankton functional groups to the production and export of particulate organic carbon, CaCO3 , and opal from global nutrient and
alkalinity distributions. Glob. Biogeochem. Cycles 20. http://dx.doi.org/10.1029/
2005GB002532.
Johnson, H.P., Hautala, S.L., Bjorklund, T.A., Zarnetske, M.R., 2006. Quantifying the
North Pacific silica plume. Geochem. Geophys. Geosyst. 7. http://dx.doi.org/10.
1029/2005gc001065.
Ledwell, J.R., Watson, A.J., Law, C.S., 1993. Evidence for slow mixing across the pycnocline from an open-ocean tracer-release experiment. Nature 364, 701–703.
Ledwell, J.R., Watson, A.J., Law, C.S., 1998. Mixing of a tracer in the pycnocline.
J. Geophys. Res., Oceans 103, 21,499–21,529.
Levitus, S., Boyer, T., 1994. World Ocean Atlas 1994, Temperature, vol. 4. U.S. Government Printing Office, Washington, DC.
Levitus, S., Burgett, R., Boyer, T., 1994. World Ocean Atlas 1994, Salinity, vol. 3. U.S.
Government Printing Office, Washington, DC.
Lumpkin, R., Speer, K., 2007. Global ocean meridional overturning. J. Phys.
Oceanogr. 37, 2550–2562.
Maier-Reimer, E., 1993. Geochemical cycles in an ocean general circulation model.
Preindustrial tracer distributions. Glob. Biogeochem. Cycles 7, 645–677.
Maier-Reimer, E., Mikolajewicz, U., Hasselmann, K., 1993. Mean circulation of the
Hamburg LSG OGCM and its sensitivity to the thermohaline surface forcing.
J. Phys. Oceanogr. 23, 731–757.
Marshall, J., Speer, K., 2012. Closure of the meridional overturning circulation
through Southern Ocean upwelling. Nat. Geosci. 5, 171–180.
Matsumoto, K., Sarmiento, J.L., Key, R.M., Aumont, O., Bullister, J.L., Caldeira, K., et al.,
2004. Evaluation of ocean carbon cycle models with data-based metrics. Geophys. Res. Lett. 31, L07303. http://dx.doi.org/10.1029/2003gl018970.
Meckler, A.N., Sigman, D.M., Gibson, K.A., Francois, R., Martinez-Garcia, A., Jaccard,
S.L., et al., 2013. Deglacial pulses of deep-ocean silicate into the subtropical
North Atlantic Ocean. Nature 495, 495–498.
Milligan, A.J., Varela, D.E., Brzezinski, M.A., Morel, F.M.M., 2004. Dynamics of silicon
metabolism and silicon isotopic discrimination in a marine diatom as a function
of pCO2 . Limnol. Oceanogr. 49, 322–329.
Najjar, R., Orr, J.C., 1998. Design of OCMIP-2 simulations of chlorofluorocarbons,
the solubility pump and common biogeochemistry. Unpublished technical
document, available at, http://ocmip5.ipsl.jussieu.fr/OCMIP/phase2/simulations/
design.ps/ (accessed 2013/09/24).
Naveira Garabato, A.C., Polzin, K.L., King, B.A., Heywood, K.J., Visbeck, M., 2004.
Widespread intense turbulent mixing in the Southern Ocean. Science 303,
210–213.
Pacanowski, R.C., Griffies, S.M., 1999. The MOM3.0 Manual, NOAA/Geophysical Fluid
Dynamics Laboratory. Princeton, NJ.
Palter, J.B., Sarmiento, J.L., Gnanadesikan, A., Simeon, J., Slater, R.D., 2010. Fueling export production: nutrient return pathways from the deep ocean and
their dependence on the meridional overturning circulation. Biogeosciences 7,
3549–3568.
Ragueneau, O., Tréguer, P., Leynaert, A., Anderson, R.F., Brzezinski, M.A., DeMaster,
D.J., et al., 2000. A review of the Si cycle in the modern ocean: recent progress
and missing gaps in the application of biogenic opal as a paleoproductivity
proxy. Glob. Planet. Change 26, 317–365.
Ragueneau, O., Schultes, S., Bidle, K.D., Claquin, P., La Moriceau, B., 2006. Si and
C interactions in the world ocean: Importance of ecological processes and implications for the role of diatoms in the biological pump. Glob. Biogeochem.
Cycles 20. http://dx.doi.org/10.1029/2006GB002688.
Reynolds, B.C., 2009. Modeling the modern marine δ 30 Si distribution. Glob. Biogeochem. Cycles 23. http://dx.doi.org/10.1029/2008GB003266.
Reynolds, B.C., Frank, M., Halliday, A.N., 2006. Silicon isotope fractionation during
nutrient utilization in the North Pacific. Earth Planet. Sci. Lett. 244, 431–443.
Reynolds, B.C., Aggarwal, J., André, L., Baxter, D., Beucher, C., Brzezinski, M.A., et al.,
2007. An inter-laboratory comparison of Si isotope reference materials. J. Anal.
At. Spectrom. 22, 561–568.
Rhein, M., 1994. The Deep Western Boundary Current: tracers and velocities. DeepSea Res., A, Oceanogr. Res. Pap. 41, 263–281.
Sarmiento, J.L., Gruber, N., Brzezinski, M.A., Dunne, J.P., 2004. High-latitude controls
of thermocline nutrients and low latitude biological productivity. Nature 427,
56–60.
Sarmiento, J.L., Simeon, J., Gnanadesikan, A., Gruber, N., Key, R.M., Schlitzer, R., 2007.
Deep ocean biogeochemistry of silicic acid and nitrate. Glob. Biogeochem. Cycles 21. http://dx.doi.org/10.1029/2006GB002720.
Schmittner, A., Gruber, N., Mix, A.C., Key, R., Tagliabue, A., Westberry, T.K., 2013.
Biology and air-sea gas exchange controls on the distribution of carbon isotope
ratios (δ 13 C) in the ocean. Biogeosciences 10, 5793–5816.
Shemesh, A., Burckle, L.H., Froelich, P.N., 1989. Dissolution and preservation of
Antarctic diatoms and the effect on sediment thanatocoenoses. Quat. Res. 31,
288–308.
Somes, C.J., Schmittner, A., Galbraith, E.D., Lehmann, M.F., Altabet, M.A., Montoya,
J.P., et al., 2010. Simulating the global distribution of nitrogen isotopes in the
ocean. Glob. Biogeochem. Cycles 24. http://dx.doi.org/10.1029/2009gb003767.
Stommel, H., 1979. Determination of water mass properties of water pumped down
from the Ekman layer to the geostrophic flow below. Proc. Natl. Acad. Sci.
USA 76, 3051–3055.
Sutton, J.N., Varela, D.E., Brzezinski, M.A., Beucher, C.P., 2013. Species-dependent silicon isotope fractionation by marine diatoms. Geochim. Cosmochim. Acta 104,
300–309.
Talley, L.D., 2013. Closure of the global overturning circulation through the Indian, Pacific, and Southern Oceans: schematics and transports. Oceanography 26,
80–97.
Tréguer, P.J., De La Rocha, C.L., 2013. The world ocean silica cycle. Ann. Rev. Mar.
Sci. 5, 477–501.
Trenberth, K.E., Olson, J., Large, W., 1989. A global ocean wind stress climatology
based on ECMWF analyses. Tech. Rep. NCAR/TN-338+STR. National Center for
Atmospheric Research, Boulder, CO.
Varela, D.E., Pride, C.J., Brzezinski, M.A., 2004. Biological fractionation of silicon isotopes in Southern Ocean surface waters. Glob. Biogeochem. Cycles 18. http://dx.
doi.org/10.1029/2003GB002140.
van Aken, H.M., de Boer, C.J., 1995. On the synoptic hydrography of intermediate
and deep water masses in the Iceland Basin. Deep-Sea Res., A, Oceanogr. Res.
Pap. 42, 165–189.
van Aken, H.M., 2000. The hydrography of the mid-latitude northeast Atlantic Ocean
I: the deep water masses. Deep-Sea Res., A, Oceanogr. Res. Pap. 47, 757–788.
Whitney, F., 2011. Nutrient variability in the mixed layer of the subarctic Pacific
Ocean, 1987–2010. J. Oceanogr. 67, 481–492.
Wischmeyer, A.G., De La Rocha, C.L., Maier-Reimer, E., Wolf-Gladrow, D.A., 2003.
Control mechanisms for the oceanic distribution of silicon isotopes. Glob. Biogeochem. Cycles 17. http://dx.doi.org/10.1029/2002GB002022.