Introduction
Since the start of the industrial revolution, human activities have caused a
rapid increase in atmospheric carbon dioxide (CO2, a major greenhouse
gas) from ∼280 ppm (pre-industrial revolution) to
∼400 ppm (present day) (http://www.esrl.noaa.gov/,
last access: 6 September 2018), which
has, in turn, led to global warming and ocean acidification, indicating that
there is an urgent need to reduce global greenhouse gas emissions (IPCC,
2013) (Fig. 1). As the Anthropocene climate system has rapidly become more
unpredictable, the scientific consensus is that the negative outcomes are a
globally urgent issue that should be resolved in a timely manner for the
sake of all life on Earth (IPCC, 1990, 1992, 1995, 2001, 2007, 2013). The
various ideas and approaches that have been proposed to relieve and resolve the
problem of global warming (Matthews, 1996; Lenton and Vaughan, 2009; Vaughan
and Lenton, 2011; IPCC, 2014; Leung et al., 2014; Ming et al., 2014) largely
fall into two categories: (1) reduction of atmospheric CO2 by the
enhancement of biological CO2 uptake (including ocean fertilization)
and/or the direct capture or storage of atmospheric CO2 through
chemically engineered processes, and (2) control of solar radiation by
artificial aerosol injection into the atmosphere to augment cloud formation
and cloud brightening to elevate albedo (Fig. 2). One of the most attractive
methods among the proposed approaches is ocean fertilization
(https://web.whoi.edu/ocb-fert/, last access: 6 September 2018), which targets the drawdown of atmospheric
CO2 by nutrient addition (e.g., iron, nitrogen, or phosphorus
compounds) to stimulate phytoplankton growth and, subsequently, carbon
export to the deep ocean or sediments via the ocean biological pump (ACE
CRC, 2015).
Schematic representation of several proposed climate-engineering
methods (modified from Matthews, 1996).
The iron hypothesis, as suggested by Martin (1990). (a) Effectiveness of the biological pump under normal conditions.
(b) Effectiveness of the biological pump following iron enrichment (modified
from Sarmiento and Gruber, 2006). (c) Schematic diagram of the decrease in
the downward flux of organic carbon as a function of depth in the water
column (modified from Lampitt et al., 2008). OM is organic matter and DIC
is dissolved inorganic carbon.
Global annual distribution of surface (a) chlorophyll
concentrations
(mgm-3), (b) nitrate concentrations (µM),
and (c) silicate
concentrations (µM). The chlorophyll a concentration
distribution was
obtained from the Aqua MODIS chlorophyll a composite from July 2002 to
February 2016 (https://oceancolor.gsfc.nasa.gov/cgi/l3, last access:
6 September 2018); nitrate and silicate were obtained from the
World Ocean Atlas 2013 dataset
(https://odv.awi.de/en/data/ocean/world-ocean-atlas-2013,
last access: 6 September 2018) and
plotted using Ocean Data View (Schlitzer, 2017). The white circles indicate
the locations of 13 artificial ocean iron fertilization (aOIF) experiments
and the black triangles indicate the locations of six natural OIF (nOIF)
experiments. Note that the numbers indicate the order of the aOIF
experiments and the roman numerals indicate the order of the nOIF
experiments (see Table 1).
The ocean biological pump is frequently depicted as a single combined
process, whereby organic matter produced by phytoplankton during
photosynthesis in surface waters is quickly transported to intermediate
and/or deep waters (Fig. 3a) (Volk and Hoffert, 1985; De La Rocha, 2007).
Although the effectiveness of the biological pump is primarily controlled by
the supply of macronutrients (i.e., nitrate, phosphate, and silicate) from
the deep ocean into the mixed layer (ML), leading to new production
(Sarmiento and Gruber, 2006), iron acts as an essential micronutrient to
stimulate the uptake of macronutrients for phytoplankton growth (Fig. 3b)
(Martin and Fitzwater, 1988; Martin, 1990; Morel and Price, 2003). In the
subarctic North Pacific (NP), equatorial Pacific (EP), and Southern Ocean
(SO), which are well known as high-nutrient and low-chlorophyll (HNLC)
regions (Fig. 4a and b), phytoplankton cannot completely utilize the
available macronutrients (particularly nitrate) for photosynthesis due to a
lack of iron. As a consequence, primary production (PP) in these HNLC
regions is relatively low, despite the high availability of macronutrients
(in particular nitrate and phosphate) (Fig. 4a and b).
Analyses of trapped air bubbles in Arctic–Antarctic ice cores have revealed
that atmospheric CO2 (∼180 ppm) during the Last Glacial
Maximum (LGM; ∼20000 years ago) was much lower than during
preindustrial times (∼280 ppm) (Neftel et al., 1982; Barnola
et al., 1987; Petit et al., 1999). Over the last 25 years, several
hypotheses have been proposed to explain the lowered atmospheric CO2
level during the LGM (Broecker, 1982; McElroy, 1983; Falkowski, 1997;
Broecker and Henderson, 1998; Sigman and Boyle, 2000). Dust inputs are
generally regarded as a major natural iron source for ocean fertilization,
and Martin (1990) hypothesized that during the LGM increased dust inputs
relieved iron limitation and, thereby, substantially enhanced the biological
pump in HNLC regions, particularly in the SO (Fig. 3b). Since Martin's
hypothesis was first published, there has been an enormous interest in ocean
iron fertilization (OIF) because only a small amount of iron (C:Fe ratios = 100000:1,
Anderson and Morel, 1982) is needed to stimulate a strong
phytoplankton response. Therefore, much of the investigative focus has
centered on the artificial addition of iron to HNLC regions as a means of
enhancing carbon fixation and subsequent export via the biological pump (ACE
CRC, 2008).
Summary of ocean iron fertilization (OIF) experiments: time,
location, research vessel, added iron(II) (values in brackets correspond to
the number of days from the first iron addition, e.g., the first iron
addition becomes (0)), initial iron concentrations, target iron
concentrations (iron concentrations after iron addition), tracer, initial
patch size, experiment duration, and regional characteristics (HNLC:
high-nutrient and low-chlorophyll).
Experiment
Time
Location
Research vessel
Added iron(II)
Initial iron
Target iron
Tracer
Patch size
Duration
Regional characteristics
(kg) (day)
(nM)
(nM)
(km2)
(days)
1
IronEx-1
Oct 1993
Equatorial Pacific 5∘ S, 90∘ W
RV Columbus Iselin
1: 450 (0)
0.06
3.60
SF6
64
10
HNLC
2
IronEx-2
May 1995
Equatorial Pacific
RV Melville
1: 225 (0)
0.02
2.00
SF6
72
17
HNLC
3.5∘ S, 104∘ W
2: 112 (3)
1.00
3: 112 (7)
1.00
3
SOIREE
Feb 1999
Southern Ocean–Australasian-Pacific sector
RV Astrolab
1: 768 (0)
0.08
3.80
SF6
50
13
HNLC
61∘ S, 140∘ E
2: 312 (3)
2.60
3: 312 (5)
2.60
4: 353 (7)
2.50
4
EisenEx
Nov 2000
Southern Ocean–Atlantic sector
RV Polarstern
1: 780 (0)
0.06
2.00
SF6
50
23
HNLC
48∘ S, 21∘ E
2: 780 (7)
3: 780 (16)
5
SOFeX-N
Jan–Feb 2002
Southern Ocean–Pacific sector
RV Revelle
1: 631 (0)
1.20
SF6
225
40
HNLCLSia
56.23∘ S, 172∘ W
RV Melville
2: 631 (4)
1.20
3: 450 (29)
1.50
6
SOFeX-S
Jan–Feb 2002
Southern Ocean–Pacific sector
RV Revelle
1: 315 (0)
0.70
SF6
225
28
HNLC
66.45∘ S, 171.8∘ W
RV Melville
2: 315 (5)
0.70
RV Polar star
3: 315 (8)
0.70
4: 315 (12)
0.70
7
EIFEX
Feb–Mar 2004
Southern Ocean–Atlantic sector
RV Polarstern
1: 1410 (0)
0.08–0.20
1.50
167
39
HNLC
50∘ S, 2∘ E
2: 1410 (13)
0.34
8
SAGE
Mar–Apr 2004
Southern Ocean–southeast of New Zealand
RV Tangaroa
1: 265 (0)
0.09
3.03
SF6
36
15
HNLCLSia
46.5∘ S, 172.5∘ E
2: 265 (6)
1.59
3: 265 (9)
0.55
4: 265 (12)
1.01
9
LOHAFEX
Jan–Mar 2009
Southern Ocean–Atlantic sector
RV Polarstern
1: 2000 (0)
2.00
SF6
300
40
HNLCLSia
48∘ S, 15∘ W
2: 2000 (18)
10
SEEDS-1
Jul–Aug 2001
Subarctic North Pacific–western basin
RV Kaiyo-Maru
1: 350 (0)
0.05
2.90
SF6
80
13
HNLC
48.5∘ N, 165∘ E
11
SERIES
Jul–Aug 2002
Subarctic North Pacific–eastern basin
RV John P. Tully
1: 245 (0)
<0.10
>1.00
SF6
77
25
HNLC
50.14∘ N, 144.75∘ W
RV El Puma
2: 245 (6)
0.60
RV Kaiyo-Maru
12
SEEDS-2
Jul–Aug 2004
Subarctic North Pacific–western basin
RV Hakuho-Maru
1: 332 (0)
0.17
1.38
SF6
64
26
HNLC
48∘ N, 166∘ E
RV Kilo-Moana
2: 159 (6)
13
FeeP
Apr–May 2004
Subtropical North Atlantic–northeast Atlantic
RV Charles Darwin
1: 1840 (0)
0.20–0.40
3.00
SF6
25
21
LNLCb
27.5∘ N, 22.5∘ W
RV Poseidon
Continued.
Experiment
Time
Location
Research vessel
Added iron(II)
Initial iron
Target iron
Tracer
Patch size
Duration
Regional characteristics
(kg) (day)
(nM)
(nM)
(km2)
(days)
I
Polar Frontc
Oct–Nov 1992
Southern Ocean–Atlantic Sector
RV Polarstern
HNLC
48∘ S, 6∘ W
II
PlumExc
Nov 1993
Equatorial Pacific
RV Columbus Iselin
0.05
0.2
HNLC
2∘ S, 89∘ W
III
CROZEXc
Nov 2004
Southern Ocean–Crozet Plateau
RV Discovery
0.55
HNLC
–Jan 2005
44∘ S, 50∘ E
IV
KEOPS-1c
Jan–Feb 2005
Southern Ocean–Kerguelen Plateau
RV Marion Dufresne
0.09
0.35
HNLC
50∘ S, 73∘ E
V
DynaLiFec
Jan–Feb 2009
Southern Ocean–Pacific sector
RV Nathaniel B. Palmer
0.20
0.40
HNLC
74∘ S, 105∘ W
VI
KEOPS-2c
Oct–Nov 2011
Southern Ocean–Kerguelen Plateau
RV Marion Dufresne
HNLC
50.63∘ S, 72.08∘ E
a High-nutrient, low-chlorophyll, and low-silicate (HNLCLSi) region.
b Low-nutrient and low-chlorophyll (LNLC) region. c Natural OIF
experiments (PlumEx: natural iron enrichment experiment near the Galapagos
Islands; CROZEX: CROZet natural iron bloom and EXport experiment;
KEOPS-1 and 2: Kerguelen Ocean and Plateau compared Study 1 and 2; DynaLiFe:
dynamic light on iron limitation program).Sources are Martin et al. (1994), de Baar et al. (1995, 2005), Coale et al. (1996, 1998, 2004), Gordon et al. (1998), Boyd et al. (2000, 2004, 2005, 2007), Boyd
and Law (2001), Gervais et al. (2002), Tsuda et al. (2003, 2007), Bakker et al. (2005), Nishioka et al. (2005), Hoffmann et al. (2006), Law et
al. (2006), Blain et al. (2007, 2015), Rees et al. (2007),
Pollard et al. (2009), Strong et al. (2009), Harvey et
al. (2010), Gerringa et al. (2012), Smetacek et al. (2012),
Martin et al. (2013), and Morris and Charette (2013).
To test Martin's hypothesis, six natural OIF (nOIF) and 13 artificial OIF
(aOIF) experiments have been performed to date in the subtropical North
Atlantic (NA), EP, subarctic NP, and SO (Blain et al., 2007, 2015; Boyd et al.,
2007; Pollard et al., 2009; Strong et al., 2009; Smetacek et al., 2012;
Martin et al., 2013) (Fig. 4a and Table 1). These OIF
experiments demonstrated, particularly for the SO, that PP could be
significantly increased after iron addition (de Baar et al., 2005; Boyd et
al., 2007). However, for aOIF to be considered as a useful geoengineering
approach (IPCC, 2007), in the long run, the most critical issue is the
effectiveness of aOIF: that is, whether a significant portion of the
organic carbon produced by aOIF in the surface waters is exported below the
winter mixed layer depth (MLD) to intermediate–deep layers for long-term
(∼1000 years) storage (Fig. 3c) (Lampitt et al., 2008). A
high carbon export was observed in the nOIF experiments in the SO near the
Kerguelen Plateau and Crozet Islands (Blain et al., 2007; Pollard et al.,
2009). However, no significant increase in carbon exports has been detected
during any aOIF experiments (de Baar et al., 2005; Boyd et al., 2007),
except for the SO European Iron Fertilization Experiment, EIFEX (Smetacek et
al., 2012). The results of these experiments, as well as the potential side
effects (e.g., production of climate-relevant gases and development of hypoxia) (Fuhrman and
Capone, 1991), have been scientifically debated amongst those who support
and oppose aOIF experimentation (Chisholm et al., 2001; Johnson and Karl,
2002; Lawrence, 2002; Buesseler and Boyd, 2003; Smetacek and Naqvi, 2008;
Williamson et al., 2012). A legal framework has been put in place to prevent
venture capitalists from deploying large-scale OIF in any international
waters because of the potential threat of commercialization and large-scale
damage inflicted on the environment by private entities motivated primarily by profit. No other marine scientific institutions are willing to
take up the challenge of carrying out new experiments due to the fear of
negative publicity. Consequently, inaction on the part of scientists might
be an incentive for others to go ahead with illegal experiments as happened
off Canada in 2012 (e.g., the 2012 Haida Gwaii Iron Dump off the west
coast of Canada).
In the context of increasing global (social–political–economic) concerns
associated with rapid climate change, it is necessary to examine the
validity and usefulness of aOIF experimentation as a climate change
mitigation strategy. Furthermore, aOIF experiments have provided insights
into the structure and function of pelagic ecosystems that cannot be
acquired from observational cruises alone. Non-OIF observations provide an
assortment of snapshots from which only an incomplete image of the processes
involved can be rendered, while OIF experiments provide time-ordered focused
frames allowing one to directly follow changes triggered by addition of an
important limiting nutrient (i.e., iron) (Smetacek, 2018). That being said,
it is necessary to plan and carry out the next aOIF experiments within the
framework of international law. Therefore, the purpose of this paper is
to (1) provide a thorough overview of the aOIF experiments conducted over
the last 25 years; (2) discuss aOIF-related important unanswered questions,
including carbon export measurement methods, potential side effects, and
international law; (3) suggest considerations for the design of future aOIF
experiments to maximize the effectiveness of the technique and begin to
answer open questions; and (4) introduce design guidelines for a future
Korean Iron Fertilization Experiment in the Southern Ocean (KIFES).
Summary of artificial ocean iron fertilization (aOIF) experiments:
objectives, significant results, and limitations.
Experiment
Objectives
Significant results
Limitations
1
IronEx-1
To test the hypothesis that artificial iron addition will increase phytoplankton productivity by relieving
the iron limitation of phytoplankton in high-nutrient, low-chlorophyll regions
Small responses in the pCO2 concentrations, Fv/Fm ratio, chlorophyll a
concentration, and primary production (PP)
Insignificant changes in nutrients
Subduction of the fertilized patch
Single iron addition
Insufficient experimental periods to observe the full phases of biogeochemical responses from the onset
to termination after iron additions
Micro- or macronutrient limitations
2
IronEx-2
To test four hypotheses that were advanced to explain the weak biogeochemical response observed during IronEx-1
Dramatic changes in biogeochemical responses; close to support for Martin's hypothesis
Taxonomic shift toward diatom-dominated phytoplankton communities
No export flux measurements in the deep ocean
Insufficient experimental duration
3
SOIREE
To test the iron hypothesis in the Southern Ocean
Diatom-dominated bloom
No measurable change in carbon export
Insufficient experimental duration
4
EisenEx
To test the hypothesis that atmospheric dust inputs might have led to a dramatic increase in ocean productivity during the Last Glacial
Maximum due to the relief of iron-limited conditions for phytoplankton growth
Diatom-dominated bloom
No clear differences in carbon flux between in patch and outside patch
Light limitation by storms
Insufficient experimental duration
5
SOFeX-N
To address the potential for iron and silicate interactions to regulate the diatom bloom
Remarkable increase in diatom biomass
Observation of large export flux event with transmissometers
Entrainment of dissolved silicate into the fertilized patch by physical mixing
No direct measurement of export fluxes with sediment traps
6
SOFeX-S
To address the potential for iron and silicate interactions to regulate the diatom bloom
Significantly enhanced export fluxes out of the mixed layer (ML), but similar to those for natural blooms
Insufficient experimental duration
7
EIFEX
To confirm that aOIF experiments can increase export production
Observation of all the phases of the phytoplankton bloom from onset to termination
Significant carbon export to deeper layers (down to 3000 m) due to the formation of aggregates with rapid sinking rates
8
SAGE
To determine the response of phytoplankton dynamics to iron addition in
high-nutrient,
low-chlorophyll, and low-silicate (HNLCLSi) regions
To test the assumption that the response of phytoplankton blooms to artificial iron addition can be detected by the
enhanced air–sea exchanges of climate-relevant gases
No shift to a diatom-dominated community
No detection of fertilization-induced export
High dilution rate by small patch size
Continued.
Experiment
Objectives
Significant results
Limitations
9
LOHAFEX
To trace the fate of iron-stimulated phytoplankton blooms and deep carbon export in HNLCLSi regions
Observation of all the phases of the phytoplankton bloom from onset to termination
No shift to a diatom-dominated community
No detection of fertilization-induced export
High grazing pressure
10
SEEDS-1
To investigate the relationship between phytoplankton biomass/community and dust deposition in the
subarctic North Pacific
To investigate changes in phytoplankton composition and vertical carbon flux
A shift from oceanic diatoms to fast-growing neritic ones
The largest changes in biogeochemical parameters of all aOIF experiments
No detection of large carbon export flux
Single iron addition
Insufficient experimental duration
11
SERIES
To compare the response of phytoplankton in the eastern subarctic with that in the western subarctic ecosystem
To investigate the most significant factor that controls the beginning to the ending of the phytoplankton bloom induced by iron addition
Observation of all phases of the phytoplankton bloom from onset to termination
No significant increases in export fluxes below the ML
High bacterial remineralization and mesozooplankton grazing pressure
12
SEEDS-2
To investigate the most significant factor that controls the beginning to the ending of the phytoplankton bloom induced by iron addition
Observation of all phases of the phytoplankton bloom from onset to termination
No shift to a diatom-dominated community
No significant increases in export fluxes
Extensive copepod grazing
13
FeeP
To investigate the impact of iron and phosphate co-limitation on PP
Increases in picophytoplankton abundances
Sources are Martin et al. (1994), Coale et al. (1996, 1998, 2004),
Bidigare et al. (1999), Boyd et al. (2000, 2004, 2005, 2007), Charette and Buesseler (2000),
Gervais et al. (2002), Tsuda et al. (2003, 2005, 2007), Bakker et al. (2005), de Baar et
al. (2005),
Hiscock and Millero (2005), Nishioka et al. (2005),
Tsumune et al. (2005), Rees et al. (2007), Harvey et al. (2010),
Law et al. (2011), Smetacek et al. (2012), and Martin et al. (2013).
Past: overview of previous aOIF experiments
A total of 13 aOIF experiments have been conducted in the following areas:
12 experiments were conducted in the three main HNLC (i.e., nitrate
>∼10 µM) regions: two in the EP, three in
the subarctic NP, and seven in the SO (Table 1, Fig. 4a and b). One
experiment was conducted in the subtropical NA, known to be a low-nutrient
and low-chlorophyll (LNLC) (i.e., nitrate <1 µM) region. These
aOIF experiments have been conducted with various objectives and multiple
hypotheses to investigate the biogeochemical responses of ocean
environments to artificial iron additions (Table 2). This overview of past
aOIF experimentation begins in Sect. 2.1, with a presentation of the
reasons why each experiment was performed and the main hypotheses (Table 2).
The unique ocean conditions for the various experiments are described in
Sect. 2.2. Iron addition and tracing methods are described in Sect. 2.3.
The biogeochemical responses to the aOIF experiments are presented in
Sect. 2.4, and finally the significant findings from these experiments are
summarized in Sect. 2.5.
Objectives and hypotheses of previous aOIF experiments
Equatorial Pacific
Initially, Martin's hypothesis was supported by the results of laboratory
and shipboard iron-enrichment bottle experiments (Hudson and Morel, 1990;
Brand, 1991; Sunda et al., 1991; DiTullio et al., 1993; Hutchins et al.,
1993). However, the extrapolation of these results based on bottle
incubations that exclude higher trophic levels has been strongly criticized
due to possible underestimates in grazing rates and other bottle effects. To
deal with these issues, in situ iron fertilization experiments at the
whole-ecosystem level are required. Under the hypothesis that aOIF would
increase phytoplankton productivity by relieving iron limitations on
phytoplankton in HNLC regions, the first aOIF experiment, the iron enrichment
experiment (IronEx-1), was conducted over 10 days in October 1993 in the EP
where high light intensity and temperatures would promote rapid
phytoplankton growth (Table 1 and Fig. 4a) (Martin et al., 1994; Coale et
al., 1998).
However, the magnitude of the biogeochemical responses in IronEx-1 was not
as large as expected (Martin et al., 1994). Four hypotheses were advanced to
explain the weak responses observed: (1) the possibility of unforeseen
micronutrient (e.g., zinc, cadmium, and manganese) or macronutrient (e.g.,
silicate) limitations, (2) the short residence time of bioavailable iron in
the surface patch due to colloidal aggregation and/or sinking of larger
particles containing iron, (3) insufficient light brought about by
subduction of the patch, and (4) high grazing pressure by zooplankton
(Martin et al., 1994; Cullen, 1995; Coale et al., 1996; Gordon et al.,
1998). To test the four hypotheses, a second aOIF experiment, IronEx-2, was
conducted in May 1995 (Coale et al., 1996). The IronEx-2 research cruise
investigated the same area for a longer period (17 days), providing more
time to collect information about the biogeochemical, physiological, and
ecological responses to the aOIF.
Southern Ocean
The SO plays an important role in intermediate and deep-water formation and
has the greatest potential of any of the major ocean basins for carbon
sequestration associated with artificial iron addition (Martin, 1990;
Sarmiento and Orr, 1991; Cooper et al., 1996; Marshall and Speer, 2012). It
is known as the largest HNLC region in the world ocean and models simulating
aOIF have predicted that, among all HNLC regions, the effect of OIF on carbon
sequestration is greatest in the SO (Sarmiento and Orr, 1991; Aumont and
Bopp, 2006). However, a simple extrapolation of the IronEx-2 results to the
SO was not deemed appropriate because of the vastly different environmental
conditions (Coale et al., 1996); therefore, based on the lessons from the EP
experiments, several aOIF experiments were carried out in the SO (Frost,
1996; Boyd et al., 2000; Smetacek, 2001; Coale et al., 2004; Harvey et al.,
2010; Smetacek et al., 2012; Martin et al., 2013). To test the roles of iron
and light availability as key factors controlling phytoplankton dynamics,
community structure, and grazing in the SO, the Southern Ocean Iron Release
Experiment (SOIREE) (Table 1 and Fig. 4a), the first in situ aOIF experiment
performed in the SO, took place in February 1999 (13 days) in the
Australasian-Pacific sector (Boyd et al., 2000).
The following year, a second aOIF experiment in the SO, EisenEx
(Eisen means iron in German), was performed in November within an Antarctic
Circumpolar Current eddy in the Atlantic sector (Smetacek, 2001; Gervais et
al., 2002). This region is considered to have a relatively high iron supply,
which is supported by dust inputs and possibly icebergs (de Baar et al.,
1995; Quéguiner et al., 1997; Smetacek et al., 2002). EisenEx was
designed to test how atmospheric dust, an important source of iron in ocean
environments, might have led to a dramatic increase in ocean productivity
during the LGM due to the relief of iron-limiting conditions for
phytoplankton growth (Smetacek, 2001; Abelmann et al., 2006).
In addition to iron availability, the supply of silicate is also considered
to be an important factor controlling PP in the SO. Silicate-requiring
diatoms, which are large-sized phytoplankton, play an important role in the
biological pump and are responsible for ∼75 % of the annual
PP in the SO (Tréguer et al., 1995). The silicate concentrations in the
SO show a decreasing northward gradient, in particular, on either side of
the Antarctic Polar Front (PF), with low silicate concentrations (<5 µM) in the sub-Antarctic waters north of the PF (<61∘ S) and high silicate concentrations (>60 µM)
to the south of the PF (Fig. 4c). Therefore, to address the impact of iron
and silicate on phytoplankton communities and export, two aOIF experiments
were conducted during January–February 2002 in two distinct regions: the
Southern Ocean iron experiment north (SOFeX-N) and south (SOFeX-S) of the
PF (Table 1) (Coale et al., 2004; Hiscock and Millero, 2005). Two years
later, the Surface Ocean–Lower Atmosphere Study (SOLAS) Air–Sea Gas
Exchange (SAGE) experiment was conducted during March–April 2004 (15 days)
in sub-Antarctic waters, which are typically HNLC with low silicate
concentrations (HNLCLSi). The aim was to determine the response of
phytoplankton dynamics to iron addition in an HNLCLSi region (Fig. 4c) (Law
et al., 2011). SAGE was designed with the assumption that the response of
phytoplankton blooms to aOIF could be detected by enhanced air–sea exchanges
of climate-relevant gases (e.g., CO2 and dimethyl sulfide, DMS)
(Harvey et al., 2010; Law et al., 2011).
These early aOIF experiments resulted in clear increases in phytoplankton
biomass and PP, but the impact on export production (i.e., carbon export
from the surface waters to below the winter MLD) was not evident (Fig. 3c)
(de Baar et al., 2005; Boyd et al., 2007). To determine if aOIF could
increase export production, EIFEX was carried out in the closed core of a
cyclonic eddy near the PF during the austral summer of 2004 (Fig. 5).
Because it was designed to investigate the termination of a bloom and
resulting export production, EIFEX was much longer (39 days) than earlier
experiments (mean±SD=22±10 days; SD represents standard
deviation) (Smetacek et al., 2012).
Of similar duration, the Indo-German iron fertilization experiment (LOHAFEX;
Loha means iron in Hindi) was conducted during January–March 2009 (40 days),
also in a PF cyclonic eddy in HNLCLSi waters (Smetacek and Naqvi, 2010;
Martin et al., 2013).
Photographs of the iron addition procedure (a–f) taken during
the European Iron Fertilization Experiment (EIFEX), Surface Ocean–Lower
Atmosphere Study (SOLAS) Air–Sea Gas Exchange (SAGE), and Indo-German iron
fertilization experiment (LOHAFEX). (a) Iron(II) sulfate bags. (b) The
funnel used to pour iron and hydrochloric acid. (c) Tank system used for
mixing iron(II) sulfate, hydrochloric acid, and seawater (Smetacek, 2015).
(d) Preparation for release: the deck of RV Tangaroa with the iron tanks on the left
and the SF6 tracer tanks on the right (photo: Matt Walkington)
(https://www.niwa.co.nz/coasts-and-oceans/research-projects/sage,
last access: 6 September 2018).
(e) Outlet pipe connected to the tank system. (f) Pumping iron into the prop
wash during EIFEX (Smetacek, 2015).
Subarctic North Pacific
The subarctic NP shows a strong longitudinal gradient in aeolian dust
deposition (i.e., high dust deposition in the west, but low in the east)
(Duce and Tindale, 1991; Tsuda et al., 2003; Takeda and Tsuda, 2005), which
is different from the other two HNLC regions (i.e., EP and SO). To
investigate the relationship between the phytoplankton biomass/community and
dust deposition, the Subarctic Pacific iron Experiment for Ecosystem
Dynamics Study 1 (SEEDS-1) was conducted in July–August 2001 (13 days) in the
western subarctic gyre (Tsuda et al., 2003, 2005). In 2004, the experiment
was repeated (SEEDS-2) in almost the same location and season. In the
intervening year, the Subarctic Ecosystem Response to Iron Enrichment Study
(SERIES) was performed in July–August 2002 (25 days) in the Gulf of Alaska
(representing the eastern subarctic gyre ecosystem) to compare the response
of phytoplankton in this area with that in the western subarctic (Boyd et
al., 2004, 2005). The SEEDS-1 and 2 experiments focused on changes in
phytoplankton composition, vertical carbon flux, and climate-relevant gas
production stimulated by artificial iron addition (Tsuda et al., 2005,
2007). The main objective of SEEDS-2 and SERIES was to determine the most
significant factor (i.e., nutrient supply and/or grazing) controlling the
iron-induced phytoplankton bloom from its beginning to its end (Tsuda et
al., 2007; Boyd et al., 2004).
Initial conditions and changes (Δ values) in chemical
parameters during the artificial ocean iron fertilization (aOIF) experiments.
Experiment
Initial NO3-
ΔNO3-
Initial PO43-
ΔPO43-
Initial Si
ΔSi
Initial pCO2
ΔpCO2
Initial DIC
ΔDIC
(µM)
(µM)
(µM)
(µM)
(µM)
(µM)
(µatm)
(µatm)
(µM)
(µM)
1
IronEx-1
10.8
-0.70
0.92
-0.02
3.90
-0.02
471
-13.0
2044a
-6.00a
2
IronEx-2
10.4
-4.00
0.80
-0.25
5.10
-4.00
538
-73.0
2051a
-27.0a
3
SOIREE
25.0
-2.90
1.50
-0.24
10.0
-2.90
349
-(38.0–32.0)
2137
-(18.0–15.0)
4
EisenEx
23.5
-1.60
1.60
-0.16
14.2
∼0
∼360
-(20.0–18.0)
2131
-(15.0–12.0)
5
SOFeX-N
21.9
-1.40
1.40
-0.09
2.50
-1.10
367
-26.0
2109
-14.0
6
SOFeX-S
26.3
-3.50
1.87
-0.21
62.8
-4.00
365
-36.0
2176
-21.0
7
EIFEX
25.0
-1.60
1.80
∼-0.30b
19.0
-11.0
360
-30.0
2135
-13.5
8
SAGE
7.90–10.5
1.30–3.90
0.62–0.85
0.83–0.97
330
8.00
2057
25.0
9
LOHAFEX
20.0
-2.50
1.20–1.30
∼-0.15c
0.60–1.60
∼358d
-(15.0–7.00)
10
SEEDS-1
18.5
-15.8
31.8
-26.8
390
-130
-58.0
11
SERIES
10.0–12.0
∼-(9.00–7.00)
1.00
-0.50
14.0–16.0
∼-(14.0–12.0)
350
-85.0
2030
-37.0
12
SEEDS-2
18.4
-5.70
36.1
370
∼-6.00
13
FeeP
<0.01
∼0.01
∼-1.00
a Dissolved inorganic carbon (DIC) values in IronEx-1 and 2 indicate
normalized DIC (normalized DIC = DIC × 35 / Salinity).
b ΔPO43- in EIFEX was digitized from Fig. 3
of Smetacek et al. (2012). c ΔPO43- in LOHAFEX
was digitized from Fig. 5.1 of Smetacek and Naqvi (2010). d ΔpCO2 in LOHAFEX was digitized from Fig. 6.1 of Smetacek and Naqvi
(2010).Sources are Martin et al. (1994), Steinberg et al. (1998), Boyd et al. (2000, 2005, 2007), Bakker et al. (2001, 2005), Frew et al. (2001),
Bozec et al. (2005), Hiscock and Millero (2005), Smetacek et
al. (2005, 2012), Takeda and Tsuda (2005), Tsuda et al. (2005, 2007), Marchetti et al. (2006a), Wong et al. (2006),
Tsumune et al. (2009), Harvey et al. (2010), Smetacek and Naqvi (2010), Berg
et al. (2011), Currie et al. (2011), Law et al. (2011), Assmy et al. (2013), Ebersbach et al. (2014), and Latasa et
al. (2014).
Initial values of biological parameters and the values after
fertilization. Note that maximum values were attained after fertilization.
Experiment
Initial
After
Initial chlorophyll a
After chlorophyll a
Initial PP
After PP
Initial mesozooplankton
After mesozooplankton
Initial heterotrophic bacteria
After heterotrophic bacteria
Fv/Fm
Fv/Fm
(mgm-3)
(mgm-3)
(mgCm-2day-1)
(mgCm-2day-1)
biomass (mgCm-3)
biomass (mgCm-3)
abundance (×105 cellsmL-1)
abundance (×105 cellsmL-1)
1
IronEx-1
∼0.30
0.63
0.24
0.65
300–450a
805–1330a
2
IronEx-2
0.25
∼0.57b
0.15–0.20
4.00
∼630c
∼2430c
∼5.00d (0–55 m)
∼14.0d (0–55 m)
6.50
10.8
3
SOIREE
0.22
0.65
0.25
2.00
∼120e
∼1300e
22.8f (0–65 m)
30.1f (0–65 m)
1.70
3.90
4
EisenEx
0.30
0.56
0.50
2.50
130–220
790
2.00
6.20
5
SOFeX-N
0.20
0.50
∼0.15g
∼2.60g
∼144h
∼1500h
10.9
6
SOFeX-S
0.25
0.65
∼0.30g
∼3.80g
∼216h
∼972h
3.28
5.25
7
EIFEX
∼0.28i
∼0.6i
0.70
3.16
∼750
1500
8
SAGE
0.27
0.61
0.63
1.33
540
900
9
LOHAFEX
∼0.33
0.50
0.50
1.25
<960
1560
10
SEEDS-1
∼0.19j
∼0.42j
0.80–0.90
21.8
420
1670
6.80f (0–20 m)
7.50f (0–20 m)
∼3.20
8.07
11
SERIES
0.24
∼0.55
0.35
∼5.00
300
>2000
7.30f (0–30 m)
5.50
12.0
12
SEEDS-2
0.29
∼0.43k
0.80
3.00
390
>1000
18.9f (0–20 m)
38.0f (0–20 m)
13
FeeP
0.06
0.07
a Primary productivity (PP) in IronEx-1 was estimated by multiplying PP
(mgCm-3day-1) with the mixed layer depth (initial: 30 m and
after: 35 m). b Fv/Fm in IronEx-2 was digitized from the Fig. 3 of
Behrenfeld et al. (1996). c PP in IronEx-2 was digitized from the Fig. 2 of Boyd (2002).
d Mesozooplankton biomass in IronEx-2 was digitized
from the Fig. 1 of Rollwagen Bollens and Landry (2000); values in brackets
correspond to the sampling layer. e PP in SOIREE was digitized from the
Fig. 3 of Gall et al. (2001b). f Mesozooplankton biomass indicates
copepod biomass; values in brackets correspond to the sampling layer; after
mesozooplankton biomass is the mean value averaged for the experimental
period after iron addition. g Chlorophyll a concentrations in SOFeX-N and SOFeX-S
were digitized from the Supplement Fig. 5 of Coale et al. (2004).
h PP values in SOFeX-N and SOFeX-S were digitized from the Fig. 4 of Coale et
al. (2004). i Fv/Fm in EIFEX was digitized from the Fig. 2 of Berg et
al. (2011). j Fv/Fm in SEEDS-1 was digitized from the Fig. 2 of Tsuda
et al. (2003). k Fv/Fm in SEEDS-2 was digitized from the Fig. 6 of
Tsuda et al. (2007).Sources are Kolber et al. (1994), Behrenfeld et al. (1996), Coale et al. (1996, 2004), Steinberg et al. (1998), Boyd et al. (2000, 2004, 2005, 2007), Rollwagen Bollens and
Landry (2000), Boyd and Law (2001), Cochlan (2001), Gall et al. (2001b),
Hall and Safi (2001), Zeldis (2001), Boyd (2002), Gervais et al. (2002),
Tsuda et al. (2003), Arrieta et al. (2004), Oliver et al. (2004), de Baar et
al. (2005),
Suzuki et al. (2005), Takeda and Tsuda (2005), Tsuda et al. (2005, 2007, 2009),
Levasseur et al. (2006), Kudo et
al. (2009), Harvey et al. (2010), Berg et al. (2011),
Currie et al. (2011), Peloquin et al. (2011b), Smetacek et al. (2012),
Thiele et al. (2012), Martin et al. (2013), and Latasa et al. (2014).
Subtropical North Atlantic
Unlike HNLC regions, PP in LNLC regions, which are predominantly occupied by
N2 fixers, is generally co-limited by phosphate and iron (Mills et al.,
2004). To investigate the impact of iron and phosphate co-limitation on PP,
the in situ phosphate and iron addition experiment (FeeP) was conducted by adding
both phosphate and iron in a LNLC region of the subtropical NA during
April–May 2004 (21 days) (Rees et al., 2007). The location of the subtropical
NA experiment corresponded to a typical LNLC region (Fig. 4a and b, Tables 3 and 4)
with low nutrients (nitrate: <0.01 µM, phosphate:
∼0.01 µM, and iron: <0.4 nM) and chlorophyll a
(<0.1 mgm-3) conditions much lower than other experimental
sites. The FeeP experiment reported that picoplankton (0.2–2.0 µm)
abundances increased after iron and phosphate additions (Rees et al., 2007);
however, no other details on the biogeochemical response to aOIF in FeeP
have been reported. This experiment will, therefore, not be discussed
further.
(a) Maximum (bar with dotted line) and initial (bar with solid line)
patch size (km2) during artificial ocean iron fertilization (aOIF)
experiments. (b) First target iron concentrations (nM). (c) Maximum (bar
with dotted line) and minimum (bar with solid line) mixed layer depth (MLD,
m) during aOIF experiments. (d) Initial sea surface temperature (SST,
∘C). (e) Initial nitrate concentrations (µM). (f) Initial
silicate concentrations (µM). (g) Initial Fv/Fm ratios. (h) Initial
chlorophyll a concentrations (mgm-3). Note that the numbers on the
x axis indicate the order of aOIF experiments as given in Fig. 4 and Table 1
and are grouped according to ocean basins: equatorial Pacific (EP) (yellow
bar), Southern Ocean (SO) (blue bar), subarctic North Pacific (NP) (red
bar), and subtropical North Atlantic (NA) (green bar). Sources are Kolber et
al. (1994), Martin et al. (1994), Behrenfeld et al. (1996), Coale et al. (1996, 1998, 2004), Steinberg et al. (1998), Boyd et
al. (2000, 2005, 2007),
Boyd and Law (2001), Gall et al. (2001a), Gervais et al. (2002), Law et al. (2003, 2006, 2011), Tsuda et al. (2003, 2005, 2007), Turner et
al. (2004),
Bakker et al. (2005), Bozec et al. (2005), de Baar et
al. (2005), Hiscock and Millero (2005), Takeda and Tsuda (2005), Tsumune et al. (2005, 2009), Marchetti et al. (2006a), Rees et al. (2007), Suzuki
et al. (2009), Harvey et al. (2010), Smetacek and
Naqvi (2010), Berg et al. (2011), Hadfield (2011),
Peloquin et al. (2011b), Smetacek et al. (2012), Thiele et al. (2012),
Martin et al. (2013), Ebersbach et al. (2014), and Latasa et al. (2014).
Environmental conditions prior to iron addition
The initial environment (∼1–7 days before iron addition) can
affect the outcome of an aOIF experiment, and the experiments described
above were conducted under a wide range of physical and biogeochemical
conditions. Below we consider the similarities and differences in these
environments according to the physical and biogeochemical properties of the
sites (Coale et al., 1998, 2004; Steinberg et al., 1998; Bakker et al., 2001; Boyd
and Law, 2001; Gervais et al., 2002; Boyd et al., 2005;
Takeda and Tsuda, 2005; Tsuda et al., 2007; Cisewski et al., 2008; Harvey et
al., 2010; Cavagna et al., 2011) (Fig. 6, Tables 3 and 4).
Equatorial Pacific
The first two aOIF experiments, IronEx-1 and IronEx-2, which were both
conducted in the EP, were performed in different seasons (i.e., IronEx-1:
October, IronEx-2: May). However, the initial surface physical conditions
were similar, with warm temperatures (24.1±1.2 ∘C), high
surface photosynthetic available radiation values (∼51.7±2.1 molm-2day-1), and shallow MLDs (27.5±2.5 m)
(Fig. 6c and d) (Coale et al., 1996, 1998; Steinberg et al.,
1998; de Baar et al., 2005).
The initial surface biogeochemical conditions were high nutrients (i.e.,
nitrate = 10.6±0.2 µM, phosphate = 0.86±0.06 µM,
and silicate = 4.5±0.6 µM) and low chlorophyll a
concentrations (0.2±0.05 mgm-3) (Tables 3 and 4). The
picophytoplankton community, including Synechococcus and Prochlorococcus, was dominant (Martin et al.,
1994; Coale et al., 1996; Cavender-Bares et al., 1999). Initial surface
nutrient concentrations were relatively low compared with other ocean basin
aOIF sites (Table 3 and Fig. 6e). Initial photosynthetic quantum efficiency
(i.e., Fv/Fm ratio, where Fm is the maximum chlorophyll fluorescence yield
and Fv is the difference between Fm and the minimum chlorophyll fluorescence
yield) (Butler, 1978), which is widely used to determine the degree to which
iron is the limiting nutrient for phytoplankton growth (the Fv/Fm ratio
ranges from 0.2 to 0.65 where conditions are less iron limited as Fv/Fm
approaches 0.65), was less than ∼0.3 (Fig. 6g and Table 4),
suggesting severe iron limitation (Behrenfeld et al., 1996; Barber and
Hiscock, 2006; Aiken et al., 2008). In the EP, initial surface partial
pressure of CO2 (pCO2) values were 504.5±33.5 µatm,
which were much higher than those observed in the SO (355.6±11.7 µatm)
or the subarctic NP (370.0±16.3 µatm) (Table 3)
(Steinberg et al., 1998).
Southern Ocean
The initial physical conditions for the aOIF experiments in the SO (SOIREE,
EisenEx, SOFeX-N, SOFeX-S, EIFEX, SAGE, and LOHAFEX) were very different from
those found in the EP; MLDs were much deeper (57.9±19.2 m) (Fig. 6c)
and sea surface temperature (SST) was much lower (4.7±3.4 ∘C)
(Fig. 6d). During SOFeX-N and SOFeX-S, which were conducted along the
same line of longitude, on either side of the PF, there were distinct
differences in SST: 5.0 ∘C in SOFeX-N and -0.5 ∘C in
SOFeX-S (Coale et al., 2004). SAGE was the northernmost of the aOIF
experiments in the SO (Table 1) and, therefore, had the highest SST
(11.5 ∘C) (Fig. 6d) (Harvey et al., 2010).
The locations for the aOIF experiments were selected following preliminary
surveys to confirm the HNLC conditions, i.e., based on satellite imagery,
nutrient concentrations, and Fv/Fm. Initial nitrate concentrations ranged
from 7.9 µM (SAGE) to 26.3 µM (SOFeX-S) (Fig. 6e and Table 3).
Among the various aOIF HNLC experiment sites, the SO had the highest initial
nitrate concentrations (21.4±5.8 µM), while the EP had the
lowest (10.6±0.2 µM). Initial nitrate and phosphate
concentrations at aOIF sites in the SO followed a latitudinal gradient, with
higher values to the south of 50∘ S (nitrate: 24.6±1.6 µM
and phosphate: 1.6±0.2 µM) and lower values to the
north (nitrate: 17.1±6.7 µM and phosphate: 1.1±0.4 µM)
(Table 3, Figs. 4b and 6e). The full range of initial silicate
concentrations has been covered by the various SO aOIF experiments, with
values ranging from ∼1.0 µM (SAGE) in the most
northernmost site to ∼60 µM (SOFeX-S) in the most
southernmost (Table 3, Figs. 4c and 6f). With the specific intent of
investigating the co-limitation of iron and silicate, SOFeX-N, SAGE, and
LOHAFEX were all conducted in HNLCLSi regions, with initial silicate
concentrations less than 2.5 µM (Figs. 4c and 6f) (Coale et al., 2004;
Harvey et al., 2010; Martin et al., 2013; Ebersbach et al., 2014). Initial
pCO2 values were low in the SO (355.6±11.7 µatm), ranging
from 330 µatm (SAGE) to 367 µatm (SOFeX-N) (Table 3).
As in the EP, initial Fv/Fm values were below ∼0.33 (Table 4
and Fig. 6g), indicating severe iron limitation. Prior to iron addition,
initial chlorophyll a concentrations ranged from ∼0.15 to
0.70 mgm-3. The maximum initial chlorophyll concentrations occurred in
EIFEX, which started with a community dominated by diatoms (Hoffmann et al.,
2006; Assmy et al., 2013), while the lowest initial chlorophyll
concentrations occurred in SOFeX-N, with a community dominated by a
nanoplankton (2.0–20 µm), such as prymnesiophytes, pelagophytes, and
dinoflagellates (Coale et al., 2004).
Subarctic North Pacific
The subarctic NP aOIF experiments (i.e., SEEDS-1, SEEDS-2, and SERIES) were
performed in regions with high nitrate (15.6±4.0 µM) and low
chlorophyll a concentrations (0.7±0.2 mgm-3) (Tables 3 and 4,
Fig. 6e and h). Compared with the other aOIF experiments, these subarctic
experiments had much higher initial silicate concentrations (27.3±9.6 µM) (Table 3 and Fig. 6f) and shallower MLDs (Fig. 6c). Although
SEEDS-1 and SEEDS-2 were conducted in almost the same location and season in
the western basin (Tsuda et al., 2007), the MLD in SEEDS-1 (8.5 m) was
shallower than in SEEDS-2 (28 m) (Fig. 6c).
Unlike the latitudinal gradients seen in the aOIF experiments in the SO,
there were longitudinal gradients in physical and biogeochemical properties
in the subarctic NP experiments (Tables 3, 4, Figs. 4 and 6d–h). Initial SSTs
in the subarctic NP were lower in the western region (7.5 ∘C in
SEEDS-1 and 8.4 ∘C in SEEDS-2) than in the eastern region
(12.5 ∘C in SERIES) (Fig. 6d). Initial nutrient concentrations
were much higher in the west (nitrate: 18.5±0.1 µM and
silicate: 34.0±2.2 µM) compared to the east (nitrate: 10 µM
and silicate: 14 µM) (Table 3, Figs. 4b, c and 6e, f). There was also a
longitudinal gradient in chlorophyll a concentrations, with relatively high
values in the west (SEEDS-1: 0.8 mgm-3 and SEEDS-2: 0.8 mgm-3)
and low value in the east (SERIES: 0.35 mgm-3) (Fig. 6h). Before the
first SEEDS-1 iron infusion, microphytoplanktons (20–200 µm), such as
the pennate diatom Pseudo-nitzschia turgidula, were dominant, whereas the areas for SERIES and SEEDS-2
were exclusively occupied by pico- and nanophytoplankton, such as
Synechococcus and haptophytes (Boyd et al., 2005; Tsuda et al., 2005, 2007;
Sato et al., 2009). Initial Fv/Fm ratios in the subarctic NP aOIF
experiments were <0.3, indicating a severe iron limitation (Fig. 6g).
Iron addition and tracing methods
Iron addition
Iron(II) and sulfate aerosols are ubiquitous in the atmosphere and,
therefore, iron sulfate (FeSO4⚫H2O), a common form of
combined iron that enters the ocean environment via dust deposition, has
been frequently regarded as a bioavailable iron source during glacial
periods (Zhuang et al., 1992; Zhuang and Duce, 1993; Spolaor et al., 2013).
Iron sulfate is a common inexpensive agricultural fertilizer that is
relatively soluble in acidified seawater (Coale et al., 1998). Therefore,
all aOIF experiments have been conducted by releasing commercial
iron sulfate dissolved in acidified seawater into the propeller wash of a
moving ship (Fig. 5), to ensure mixing with surface waters during iron
additions.
In general, background dissolved iron concentrations in HNLC regions are
<0.2 nM (Table 1). Iron-enrichment bottle incubation experiments
performed in deck incubators using in situ seawater have indicated the maximum
phytoplankton growth rates in response to iron additions of
1.0–2.0 nM
(Fitzwater et al., 1996). In aOIF experiments performed in the ocean,
targeted iron concentrations within the ML have ranged between
∼1.0 and 4.0 nM, depending on the site (Table 1 and Fig. 6b)
(Martin et al., 1994; Coale et al., 1996, 2004; Boyd et al., 2000; Bowie et al.,
2001; Tsuda et al., 2003; Nishioka et al., 2005; Law et
al., 2006; Smetacek et al., 2012; Martin et al., 2013). If injected iron is
well dispersed throughout the ML within 24 h by convective mixing
(Martin and Chisholm, 1992), the amount of added iron required to raise the
background iron concentration to the target level can be calculated using a
volume estimate (i.e., iron-fertilized water patch area × MLD)
(Watson et al., 1991). To minimize uncertainty between the first iron
addition and phytoplankton response, aOIF experiments have involved
multiple small iron injections to the surface waters in the study area at
∼0.4 to ∼1.5 km intervals over a 1–2-day
period (Coale et al., 1998). The patch size fertilized by the first iron
addition varied from 25 km2 (e.g., FeeP; iron(II) addition of 1840 kg)
to 300 km2 (e.g., LOHAFEX; iron(II) addition of 2000 kg) and by the
end of these experiments had spread to a maximum of ∼2400 km2 (Coale et al., 2004; Boyd et al., 2007; Strong et al., 2009; Martin
et al., 2013) (Table 1 and Fig. 6a).
During the experiments, dissolved iron concentrations increased to the
target ∼1.0–4.0 nM (Table 1 and Fig. 6b), but decreased to
background concentrations within days. The fast decrease in dissolved iron
concentrations indicates that iron was horizontally dispersed and/or rapidly
incorporated into particles. These processes occur more rapidly in warmer
waters (ACE CRC, 2015). For example, the first aOIF experiment, IronEx-1,
showed that the dissolved iron concentration rapidly decreased from 3.6 to
0.25 nM ∼4 days after iron addition in the center of the
fertilized patch, suggesting a limit to the level required for phytoplankton
growth (Coale et al., 1998; Gordon et al., 1998). As a result, except for
the single iron addition experiments of IronEx-1, SEEDS-1, and FeeP (Martin
et al., 1994; Tsuda et al., 2003; Rees et al., 2007), most aOIF experiments
have involved multiple iron additions at the patch center, to continuously
derive the stimulation of phytoplankton during the experiments. These
experiments included (two additions) EIFEX, SERIES, SEEDS-2, and LOHAFEX (Boyd et
al., 2005; Tsuda et al., 2007; Smetacek et al., 2012; Martin et al., 2013);
(three additions) IronEx-2, EisenEx, and SOFeX-N (Coale et al., 1996, 2004; Gervais et
al., 2002; Nishioka et al., 2005); and (four additions)
SOIREE, SOFeX-S, and SAGE (Boyd et al., 2000; Coale et al., 2004; Bakker et al.,
2005; Harvey et al., 2010) (Table 1).
Tracing iron-fertilized patch
To trace the iron-fertilized patch, aOIF experiments have used a combination
of physical and biogeochemical approaches. All the aOIF experiments except
EIFEX have used sulfur hexafluoride (SF6) as a chemical tracer (Table 1)
(Martin et al., 1994; de Baar et al., 2005; Smetacek et al., 2012). The
SF6, which is not naturally found in oceanic waters, is a useful tracer
for investigating physical mixing and advection–diffusion processes in the
ocean environment due to its nontoxicity, biogeochemically inert
characteristics, and low detection limit (Law et al., 1998). The injected
SF6 is continuously monitored using gas chromatography with an electron
capture detector system (Law et al., 1998; Tsumune et al., 2005). Usually
only one SF6 injection is necessary because background levels are
generally extremely low in the ocean (<1.2 fM; f: femto-,
10-15) (Law et al., 1998, 2003; Boyd et al., 2004);
however, in the SAGE experiment, with its higher mixing and lateral
dilution, there were three injections (Harvey et al., 2010). Although these
earlier experiments demonstrated that the injection of artificial SF6
is a useful technique for following iron-fertilized patches, SF6 can
only be used for a limited period (∼2 weeks) due to the loss at
the surface through air–sea gas exchange (Law et al., 2006; Tsumune et al.,
2009; Martin et al., 2013). Furthermore, caution is required because
artificially high levels of SF6 injection may negatively impact the
interpretation of low-level SF6 signals dissolved in seawater via
air–sea exchange to estimate tracer-based water mass ages for understanding
physical circulation (Fine, 2011). These techniques have been widely used to
estimate anthropogenic carbon invasion as well as to understand ocean
circulation in various ocean environments, with SF6 being an
important time-dependent tracer that has a well-recorded atmospheric
history. Thus, continuous sampling systems, measuring biogeochemical
parameters such as Fv/Fm, pCO2, and chlorophyll fluorescence, have also
been used as an alternative means of following iron-fertilized patches
(Gervais et al., 2002; Boyd et al., 2005; Tsuda et al., 2007; Harvey et al.,
2010; Smetacek et al., 2012). The Fv/Fm ratio displays a particularly rapid
increase (within 24 h) in response to a first iron addition (Kolber et
al., 1994; Behrenfeld et al., 1996; Smetacek et al., 2012), suggesting that
it is an easy and convenient tracer for following a fertilized patch.
In addition, surface-drifting buoys equipped with Argos or GPS
systems have been successfully used to track the movement of fertilized
patches along with biogeochemical tracers (Coale et al., 1998; Boyd and Law,
2001; Law et al., 2006; Martin et al., 2013). However, floats tend to drift
out of the fertilized patches under strong wind forcing (Watson et al.,
1991; Law et al., 1998; Stanton et al., 1998). NASA airborne oceanographic
lidar and ocean-color satellites have also been employed to assess the
large-scale effects of iron addition on surface chlorophyll in fertilized
patches, as compared to surrounding regions (Martin et al., 1994; Westberry
et al., 2013).
Initial values of the export flux and the values after
fertilization (mgCm-2day-1), the corresponding depth inside and
outside the fertilized patch for artificial ocean iron fertilization (aOIF)
experiments, and measurement method. Values in brackets correspond to the
day of measurement after fertilization.
Experiment
In patch
In patch
Outside patch
Outside patch
Depth
Method
initial (day)
after (day)
initial (day)
after (day)
(m)
1
IronEx-1
2
IronEx-2
84 (0)
600 (7–14)
25
Water-column 234Th
3
SOIREE
∼87
100
Water-column 234Th
185 (11–13)
146 (0–2)
78 (11–13)
110
Drifting trap
74 (11–13)
73 (0–2)
38 (11–13)
310
Drifting trap
4
EisenEx
5
SOFeX-N
6
SOFeX-S
36 (5)
112 (27)
48 (6)
49 (26)
50
Water-column 234Th
19 (5)
142 (27)
38 (6)
56 (26)
100
Water-column 234Th
7
EIFEX
∼340 (0)a
∼1692 (32)a
∼396 (0)a
∼516 (32)a
100
Water-column 234Th
8
SAGE
9
LOHAFEX
∼62 (0)b
∼94 (25)b
∼77 (4)b
∼54 (34)b
100
Water-column 234Th
∼6 (0–2)c
∼5 (13–15)c
∼29 (26–27)c
200
Neutrally buoyant sediment trap
∼12 (28–37)c
∼11 (24–29)c
450
Neutrally buoyant sediment trap
10
SEEDS-1
234 (1–3)
141 (12–14)
148 (1–6)
154 (10–14)
40
Drifting trap
100 (0–2)
423 (9–13)
50
Water-column 234Thd
68 (1–3)
85 (12–14)
61 (1–6)
91 (10–14)
100
Drifting trap
121 (0–2)
460 (2–9)
200
Water-column 234Th
11
SERIES
∼138 (3)e
480 (24)
192 (3)
139 (15)
50
Drifting trap
∼48 (3)e
∼192 (24)e
100
Drifting trap
12
SEEDS-2
290 (1–4)
580 (19–22)
300 (1–8)
509 (18–31)
40
Drifting trap
316 (1–4)
337 (19–22)
213 (1–8)
204 (18–31)
100
Drifting trap
13
FeeP
a Export flux in EIFEX was digitized from the Supplement Fig. 5.1
of Smetacek et al. (2012). b Export flux in LOHAFEX was digitized from
the Fig. 4 of Martin et al. (2013). c Export flux in LOHAFEX was
digitized from the Fig. 6 of Martin et al. (2013). d Export flux in
SEEDS-1 was determined from the suspended particles. e Export flux in
SERIES was digitized from the Fig. 2 of Boyd et al. (2004).Sources are Bidigare et al. (1999), Charette and Buesseler (2000), Nodder
and Waite (2001), Boyd et al. (2004), Aono et al. (2005), Buesseler et al. (2005), Aramaki et al. (2009), Smetacek et al. (2012), and Martin et al. (2013).
Biogeochemical responses
Biogeochemical responses to artificial iron addition, in particular, Fv/Fm
ratio, chlorophyll a, PP, nutrients, CO2 variables, and carbon export
fluxes, are given in Tables 3–5 and Figs. 7–8. The results are important,
as they have been used as a basis to determine whether the aOIF is
effective. Here we address the biogeochemical response in each of the ocean
basins to the aOIF experiments to date.
(a) Maximum (bar with dotted line) and initial
(bar with solid line)
Fv/Fm ratios during artificial ocean
iron fertilization (aOIF) experiments.
(b) Changes in nitrate concentrations (ΔNO3-=[NO3-]post-fertilization(postf)-[NO3-]pre-fertilization(pref);
µM). (c) Maximum (bar
with dotted line) and initial (bar with solid line) chlorophyll a
concentrations (mgm-3). (d) Distributions
of chlorophyll a
concentrations (mgm-3) on day 24 after iron addition
in the Southern
Ocean iron experiment north (SOFeX-N) from MODIS Terra Level-2 daily image
and on day 20 in the Southern
Ocean iron experiment south (SOFeX-S) from SeaWiFS Level-2 daily image
(white dotted box indicates phytoplankton bloom during aOIF experiments).
(e) Changes in primary productivity (PP)
(ΔPP=[PP]postf-[PP]pref; mgCm-2day-1).
(f) Changes in partial pressure of
CO2 (pCO2) (ΔpCO2=[pCO2]postf-[pCO2]pref; µatm). The color bar indicates changes in
dissolved inorganic carbon (DIC) (ΔDIC=[DIC]postf-[DIC]pref; µM). Note that the PP
(mgCm-2day-1) of aOIF
experiment number 1 (IronEx-1) was estimated by multiplying the PP
(mgCm-3day-1)
with the mixed layer depth (initial: 30 m and after: 35 m).
The numbers on the x axis indicate the order of aOIF experiments as
given in Fig. 4 and Table 1 and are grouped according to ocean basins:
equatorial Pacific (EP) (yellow bar), Southern Ocean (SO) (blue bar),
subarctic North Pacific (NP) (red bar), and subtropical North Atlantic (NA)
(green bar). Sources are Kolber et al. (1994), Martin et al. (1994),
Behrenfeld et al. (1996), Coale et al. (1996, 2004), Steinberg et al. (1998), Boyd
et al. (2000, 2004, 2005, 2007), Boyd and Law (2001), Frew et al. (2001), Gall et al. (2001b),
Boyd (2002), Gervais et al. (2002), Tsuda et al. (2003), Bakker et al. (2005), Bozec
et al. (2005), de Baar et al. (2005), Hiscock and Millero (2005), Smetacek
et al. (2005, 2012), Takeda and Tsuda (2005), Tsuda et al. (2005, 2007), Wong et al. (2006), Kudo et al. (2009), Tsumune
et al. (2009), Harvey et al. (2010), Smetacek and Naqvi (2010), Berg et al. (2011), Currie et al. (2011), Law et al. (2011), Peloquin et
al. (2011b),
Thiele et al. (2012), Assmy et al. (2013), Martin et al. (2013),
Ebersbach et al. (2014), and Latasa et al. (2014).
(a) Time series of particulate organic carbon (POC) fluxes estimated
from the water-column-based 234Th method (mgCm-2day-1) of
the upper 100 m layer inside (red bar) and outside the fertilized patch
(blue bar) during the European Iron Fertilization Experiment (EIFEX)
(modified from Smetacek et al., 2012). (b) Time series of vertically
integrated 234Th (dpmL-1) inside (red circles) and outside the
fertilized patch (blue diamonds) relative to the parent uranium-238 (238U; dpmL-1; dotted black line) during the
Southern Ocean Iron Release Experiment (SOIREE) (modified from Nodder et al., 2001).
Equatorial Pacific
The IronEx-1 and 2 experiments, which were conducted in similar initial
conditions (refer to Sect. 2.2.1), presented quite different
biogeochemical responses (Tables 3–4 and Fig. 7). In IronEx-1, there were
small responses in the Fv/Fm ratio, chlorophyll a concentration, PP, and
pCO2 concentrations, but no significant changes in nutrients (Martin et
al., 1994). On the other hand, IronEx-2 found dramatic changes in
biogeochemical responses, providing support for Martin's hypothesis (Coale
et al., 1996). Unexpected small responses during IronEx-1 were due to
subduction of the fertilized surface layer by adjacent water (Coale et al.,
1998). The contrasting results from the two experiments are also likely to
be associated with whether or not there were additional iron injections
(IronEx-1: no extra addition; IronEx-2: two additional injections) and
different experiment durations (IronEx-1: 10 days; IronEx-2: 17 days).
The Fv/Fm ratios provided further detail. In IronEx-1 and IronEx-2, Fv/Fm
rapidly increased within ∼24 h of iron addition and
reached a maximum of ∼0.60 on the second day (Table 4 and
Fig. 7a) (Barber and Hiscock, 2006; Aiken et al., 2008). While the elevated
IronEx-1 Fv/Fm ratios promptly disappeared, suggesting rapid iron loss due
to the subduction of the fertilized patch and/or adsorption onto colloidal
particles (perhaps indicative of insufficient iron supply), increased
IronEx-2 Fv/Fm ratios were maintained for 8 days through multiple iron
additions, suggesting that additional iron enrichments are likely to be a
determining factor in successfully artificially increasing PP through OIF
(Kolber et al., 1994; Behrenfeld et al., 1996).
During IronEx-1, chlorophyll a concentrations increased significantly
(3-fold), reaching a maximum value of 0.65 mgm-3 in the first 4 days
following iron addition (Martin et al., 1994). In IronEx-2, surface
chlorophyll a increased nearly 27-fold, with a maximum of 4 mgm-3
after day 7 (Table 4 and Fig. 7c) (Coale et al., 1996). To quantify the
changes in carbon fixation following iron addition, the depth-integrated PP
(from the surface to the critical depth, euphotic depth, or MLD) was
estimated in the iron-fertilized patches. The depth-integrated PP values
increased significantly compared to the initial values. The IronEx-2
ΔPP (where ΔPP = [PP]post-fertilization(postf) – [PP]pre-fertilization(pref))
was the highest (∼1800 mgCm-2day-1)
of all the aOIF experiments discussed here (Table 4 and
Fig. 7e).
Changes in pCO2 during IronEx-1 were less than expected (ΔpCO2=[pCO2]postf-[pCO2]pref=-13 µatm)
(Martin et al., 1994). However, substantial drawdowns of pCO2
(ΔpCO2=-73 µatm) and dissolved inorganic
carbon (ΔDIC=[DIC]postf-[DIC]pref=-27 µM) during
IronEx-2 were derived through the increased PP (Table 3 and Fig. 7e–f)
(Steinberg et al., 1998). As the bloom developed, a significant nitrate
uptake (e.g., ΔNO3-=[NO3-]postf-[NO3-]pref=-4.0 µM) was observed (Table 3 and Fig. 7b)
and silicate concentrations also gradually decreased from 5.1 to 1.1 µM
(i.e., limiting diatom growth) over 8 days (Coale et al., 1996;
Boyd, 2002). The depletion of macronutrients in fertilized patches provides
indirect evidence that phytoplankton growth in surface waters was driven by
aOIF (Boyd and Law, 2001).
Although no phytoplankton community change was observed in IronEx-1, after
iron addition in IronEx-2 a shift from a picophytoplankton-dominated
community to a microphytoplankton-dominated community was observed,
resulting in a diatom-dominated bloom (Behrenfeld et al., 1996; Coale et
al., 1996; Cavender-Bares et al., 1999). Diatom biomass increased nearly 70-fold over 8 days early in the experiment, compared to a less than a
2-fold increase for the picophytoplankton (Landry et al., 2000). The
biomass of mesozooplankton (200–2000 µm), such as copepods, grew
simultaneously, substantially increasing the community grazing effect of
larger animals on phytoplankton standing stocks from 7.8 %day-1
outside the patch to 11.4 %day-1 in the patch (Coale et al., 1996).
However, grazing did not prevent the development of a diatom bloom over
8 days early in the IronEx-2 experiment (Table 4) (Coale et al., 1996;
Rollwagen Bollens and Landry, 2000). The iron-induced diatom bloom began to
decline after day ∼8 of the experiment (Landry et al., 2000).
The decline was probably associated with the combined effects of both the
elevated grazing pressure and the onset of nutrient depletion (i.e.,
limitation in silicate and/or iron) (Cavender-Bares et al., 1999; Boyd,
2002).
To determine whether the biological pump (i.e., export production) is
enhanced after iron addition, the export flux of particulate organic carbon
(POC) was estimated using a chemical tracer, the natural radiotracer
thorium-234 (234Th; half-life = 24.1 days) (Table 5) (Bidigare et
al., 1999). The 234Th radionuclide has a strong affinity for particles,
and the extent of 234Th removal in the water column is indicative of
the export of POC associated with surface PP out of the ML (Buesseler,
1998). IronEx-2 was the first aOIF experiment in which the POC flux from the
surface to 25 m was measured (Table 5). However, no 234Th measurements
were made in the unfertilized patch for comparison, and no measurements in
the deep ocean were undertaken to demonstrate deep carbon export (Bidigare
et al., 1999).
Southern Ocean
As in the EP IronEx-1 and 2 experiments, there were initial rapid increases in
the Fv/Fm ratio within 24 h
of iron addition in the SO experiments indicating that
phytoplankton growth was mainly limited by iron availability. Maximum values
of the Fv/Fm ratio ranged from 0.50 (SOFeX-N and LOHAFEX) to 0.65 (SOIREE
and SOFeX-S) (Table 4 and Fig. 7a). However, the time taken to reach the
maximum Fv/Fm ratio was usually longer than ∼10 days, i.e.,
much slower than in IronEx-1 and 2 (∼2 days) (Boyd and Abraham,
2001; Gervais et al., 2002; Coale et al., 2004; Smetacek et al., 2005;
Peloquin et al., 2011b; Martin et al., 2013). The slower response time in
the SO compared to the EP might be attributed to the colder temperatures
(∼5 vs. ∼24 ∘C) and/or
the deeper MLDs (∼60 vs. ∼30 m) (Fig. 6c
and d) (Boyd and Abraham, 2001; Boyd, 2002).
The aOIF experiments in the SO recorded >2-fold increases in
chlorophyll a concentrations compared to initial levels (<0.7 mgm-3),
and maximum values between 1.25 mgm-3 (LOHAFEX) and
∼3.8 mgm-3 (SOFeX-S) were obtained after artificial
iron additions (Table 4 and Fig. 7c). Satellite observations were used to
investigate the changing spatial and temporal distribution of chlorophyll a
concentration in response to iron fertilization in the fertilized patches
compared to the surrounding waters; for example, SOFeX-N and SOFeX-S found elevated
chlorophyll a concentrations in fertilized patches after iron addition
through satellite images (Fig. 7d) (Boyd et al., 2000; Coale et al., 2004;
Westberry et al., 2013).
Following artificial iron enrichment in the SO, ΔPP ranged from 360
(SAGE) to ∼1356 mgCm-2day-1 (SOFeX-N) (Table 4
and Fig. 7e). During SOIREE, EisenEx, SOFeX-N, and SOFeX-S, PP increased
continuously throughout the duration of the experiments (Boyd et al., 2000;
Gall et al., 2001b; Gervais et al., 2002; Coale et al., 2004; Assmy et al.,
2007). However, in EIFEX, SAGE, and LOHAFEX there was a significant increase
in PP for ∼10 (SAGE) to 20 (EIFEX) days in response to the
iron addition, and decreasing trends after day ∼12 (SAGE) to 25 (EIFEX).
The decrease was due to various processes such as export (e.g.,
EIFEX), lateral dilution with surrounding waters (e.g., SAGE), and high
grazing pressure and bacterial respiration (e.g., LOHAFEX) (Boyd, 2002;
Gervais et al., 2002; Buesseler et al., 2004; Coale et al., 2004; Peloquin
et al., 2011b; Smetacek et al., 2012; Thiele et al., 2012; Assmy et al.,
2013; Martin et al., 2013; Latasa et al., 2014).
Using both microscopes and high-performance liquid chromatography pigment
analysis, changes in the phytoplankton community affected by iron addition have
also been investigated. Most SO aOIF experiments have resulted in blooms of
diatoms (Boyd et al., 2007). During SOIREE and EisenEx, the dominant
phytoplankton community shifted from pico- and nanophytoplankton (e.g.,
picoeukaryotes and prymnesiophytes) to microphytoplankton (i.e., diatoms)
(Gall et al., 2001b; Gervais et al., 2002; Assmy et al., 2007). In SOFeX-S
and EIFEX, diatoms were already the most abundant group prior to iron
addition (Coale et al., 2004; Hoffmann et al., 2006; Assmy et al., 2013).
The contribution of large diatoms became especially clear in EIFEX where
∼97 % of the phytoplankton bloom was attributed to this
group (Smetacek et al., 2012; Assmy et al., 2013). However, no taxonomic
shift toward diatom-dominated communities (<5 % of total
phytoplankton community) was observed during SAGE and LOHAFEX, which were
conducted under silicate-limited conditions (Harvey et al., 2010; Peloquin
et al., 2011b; Martin et al., 2013; Ebersbach et al., 2014). Although
SOFeX-N was conducted under low silicate conditions (Fig. 6f), the diatom
biomass increased remarkably, making up ∼44 % of the total
phytoplankton community (Coale et al., 2004). This result was partly
influenced by the temporary relief of silicate limitation through lateral
mixing of the iron-fertilized waters with surrounding waters, with
relatively higher silicate concentrations (Coale et al., 2004).
Iron-mediated increases in PP resulted in a significant uptake in
macronutrients and pCO2 throughout the aOIF experiments in the SO
(except for SAGE) (Table 3, Fig. 7b and f). ΔNO3- ranged
from -3.5 µM (e.g., SOFeX-S) to -1.4 µM (e.g., SOFeX-N) and
ΔpCO2 ranged from -38 µatm (e.g., SOIREE) to -7.0 µatm
(e.g., LOHAFEX). Although both were initially dominated by diatoms, SOFeX-S had a
somewhat greater ΔNO3- (-3.5 µM) and ΔpCO2 (-36 µatm) than EIFEX (ΔNO3-: -1.6 µM
and ΔpCO2: -30 µatm) (Coale et al., 2004; Hoffmann et al.,
2006; Smetacek et al., 2012; Assmy et al., 2013). However, the smaller
silicate uptake (ΔSi=[Si]postf-[Si]pref)
observed during SOFeX-S (-4.0 µM) compared to EIFEX (-11 µM) was
associated with a decrease in silicification (i.e., changes in frustule
thickness of the dominant diatom species, Fragilariopsis sp., Twining et al., 2004). During
EIFEX, the ratio of heavily silicified diatoms (e.g., Thalassiothrix antarctica) to total diatom
biomass increased from 0.24 (day 0) to 0.46 (day 37), leading to the higher
Si uptake (Hoffmann et al., 2006; Assmy et al., 2013). Interestingly, the
biogeochemical responses in SAGE were totally different from those seen in
other experiments as increases in ΔNO3- (+3.9 µM),
ΔpCO2 (+8.0 µatm), and ΔDIC (+25 µM) were
observed (Table 3, Fig. 7b and f). These contrasting results were thought
to be the result of entrainment through vertical and horizontal physical
mixing into the iron-fertilized patch of surrounding waters with higher
nutrient and pCO2 concentrations (Currie et al., 2011; Law et al.,
2011).
SOIREE was the first aOIF experiment in the SO to estimate the downward
carbon flux into deep waters (Fig. 3c). A comprehensive suite of methods was
used: drifting traps, 234Th and the stable carbon isotope of
particulate organic matter (δ13Corg) estimates derived
from high-volume pump sampling, and a beam transmissometer (Nodder and
Waite, 2001). However, no measurable change in carbon export was observed in
response to iron-stimulated PP (Table 5 and Fig. 8b) (Charette and
Buesseler, 2000; Nodder and Waite, 2001; Trull and Armand, 2001; Waite and
Nodder, 2001). During EisenEx, an increased downward carbon flux estimated
from 234Th deficiency was observed in the iron-fertilized patch as the
experiment progressed. However, there were no clear differences between in-
and outside-patch carbon fluxes (Buesseler et al., 2005). During SOFeX-S,
significantly enhanced POC fluxes below the MLD, similar to those observed in
natural blooms, were estimated from 234Th measurements after iron
enrichment (Buesseler et al., 2005). During SOFeX-N autonomous profilers
equipped with transmissometers recorded a downward carbon flux between day
∼27 and ∼45 after the first iron addition
(Bishop et al., 2004; Coale et al., 2004). However, it was unclear whether
surface-fixed carbon was well and truly delivered below the winter MLD.
During SAGE and LOHAFEX, which were conducted under silicate-limited
conditions (Table 3, Figs. 4c and 6f), no significant enhancement of carbon
export was detected (Table 5) (Peloquin et al., 2011b; Martin et al., 2013).
This result was likely due to the dominance of picoplankton and grazing
that led to rapid recycling of organic matter in the ML. In contrast to the
other aOIF experiments, EIFEX, which was conducted within the core of an
eddy, showed clear evidence of carbon export well below 500 m, stimulated by
artificial iron addition (Jacquet et al., 2008; Smetacek et al., 2012).
During EIFEX, the initial export flux, estimated from 234Th in the
upper 100 m of the fertilized patch, was ∼340 mgCm-2day-1
(Table 5 and Fig. 8a) (Smetacek et al., 2012). This value remained
constant for about 24 days after iron addition. Between day 28 and 32 a
massive increase in carbon export flux (maximum of ∼1692 mgCm-2day-1)
was observed in the fertilized patch, while the initial
value remained constant in the unfertilized patch (Table 5 and Fig. 8a). The
profiling transmissometer with high-resolution coverage confirmed this
result, showing an increase in exported POC below 200 m after day 24. At
least half the iron-induced biomass sank (via the formation of aggregates of
diatom species, in particular Chaetoceros dichaeta) to a depth of 1000 m, with a 10-fold higher
sinking rate (500 mday-1) compared to the initial conditions (Smetacek
et al., 2012). Significant changes in export production were not found in
any of the other aOIF experiments and, therefore, the impact of artificial
iron addition on diatom aggregate formation needs focused study in future
aOIF experiments (Boyd et al., 2004; Smetacek et al., 2012; Martin et al.,
2013).
Subarctic North Pacific
The observed increase in the Fv/Fm ratio in response to aOIF in the
subarctic NP suggests that the relief in iron limitation may have assisted
phytoplankton growth (Table 4 and Fig. 7a). SEEDS-1 and 2, which were conducted
in the western basin, showed continuous increases in the Fv/Fm ratio, with a
maximum value of ∼0.4 approximately 10 days after the first
iron addition (Tsuda et al., 2003, 2007). During SERIES, which was conducted
in the eastern basin, the Fv/Fm ratio rapidly increased within 24 h of
the first iron addition and reached a maximum value of ∼0.55
on day 4 (Boyd et al., 2005). However, the Fv/Fm ratio returned toward the
initial value of <0.3 as the dissolved iron concentrations
decreased to background levels (<0.2 nM) after about day 10 (Tsuda
et al., 2003, 2007; Boyd et al., 2005).
Increases in chlorophyll a concentrations were detected in the subarctic NP
aOIF experiments in both basins after about the fifth day (Tsuda et al.,
2003; Boyd et al., 2004; Suzuki et al., 2009). These increases were
especially apparent in SEEDS-1, where they reached a maximum value of 21.8 mgm-3
(27 times the initial value of 0.8 mgm-3) (Table 4 and
Fig. 7c). This augmentation was the largest among all the aOIF experiments
(Tsuda et al., 2003). The dramatic surface chlorophyll a increase observed
during SEEDS-1 was partly attributed to the particular range of seawater
temperature in the region, which was conducive to diatom growth (i.e.,
8–13 ∘C) as well as to the shallower MLD (∼10 m),
which provided a relatively longer surface water residence time for the
additional iron (Fig. 6c and d) (Noiri et al., 2005; Takeda and Tsuda,
2005; Tsuda et al., 2005; Tsumune et al., 2005). During SERIES,
chlorophyll a concentrations increased substantially from the initial value
of 0.35 to ∼5 mgm-3 over 17 days, the second highest
concentration recorded in all aOIF experiments (Table 4 and Fig. 7c) (Boyd
et al., 2004). However, on the 18th day there was a downturn in
chlorophyll a as silicate concentrations decreased to <2 µM
(Boyd et al., 2005). Although SEEDS-2 was conducted under similar initial
conditions to SEEDS-1 (refer to Sect. 2.2.3), there was a minimal increase
in chlorophyll a (i.e., maximum value of 3 mgm-3) (Fig. 7c). This
smaller increase was thought to be the result of strong copepod grazing
(SEEDS-2 had almost 5 times more copepod biomass than SEEDS-1) (Table 4)
(Tsuda et al., 2007). A similar range was seen in depth-integrated PP, which
increased 3-fold or more after iron addition in the subarctic NP aOIF
experiments (e.g., from 300–420 to 1000–2000 mgCm-2day-1)
(Table 4 and Fig. 7e).
Changes in the composition of phytoplankton groups were investigated in the
subarctic NP aOIF experiments. In SEEDS-1 there was a shift from oceanic
diatoms (e.g., Pseudo-nitzschia turgidula), with growth rates of 0.5–0.9 day-1, to faster-growing
neritic diatoms (e.g., Chaetoceros debilis, with a growth rate of 1.8 day-1) (Tsuda et al., 2005). The effect on
the biological pump can be quite different depending on the species of
diatom stimulated by the aOIF. Chaetoceros debilis, known to be widespread in coastal
environments, intensifies the biological pump by forming resting spores in
contrast to grazer-protected, thickly silicified oceanic species (e.g.,
Fragilariopsis sp. and Thalassiothrix sp.) that contribute silica but little carbon to the sediments.
The shift in the dominant phytoplankton species during SEEDS-1 was an
important contributor to the recorded increase in phytoplankton biomass.
During SERIES, the phytoplankton community changed from Synechococcus and haptophytes to
diatoms, and the highest SERIES chlorophyll a concentration (day 17) was
associated with a peak in diatom abundance (Boyd et al., 2005). However,
during SEEDS-2, no significant iron-induced diatom bloom was observed.
Instead, pico- and nanophytoplankton (e.g., Synechococcus, picoeukaryotes, and
cryptophytes) (>70 % of the total community) dominated
throughout the duration of the experiment due to the heavy grazing pressure
on diatoms (Table 4) (Tsuda et al., 2007; Sato et al., 2009).
In the subarctic NP experiments, significant changes in macronutrient
uptake (i.e., ΔNO3- and ΔSi), ΔDIC, and
ΔpCO2 in response to aOIF were observed (Table 3 and Fig. 7b
and f). SEEDS-1, which exhibited the largest increases in chlorophyll a
concentrations, also had the largest ΔpCO2 (-130 µatm) and
ΔDIC (-58 µM) (Table 3 and Fig. 7f). These changes led, in
turn, to the largest ΔNO3- (-15.8 µM) (Fig. 7b) and
ΔSi (-26.8 µM) (Table 3) (Tsuda et al., 2003). The second
largest increase in the chlorophyll a concentration was observed in SERIES,
where drawdowns of pCO2 (-85 µatm), DIC (-37 µM), nitrate
(∼-9.0 µM), and silicate (∼-14.0 µM)
were recorded. During SEEDS-2, the nitrate concentration decreased
remarkably from 18.4 to 12.7 µM after day 5; however, there was
no significant change in silicate concentrations, which would have been
expected as a signal of an iron-induced diatom bloom (Tsuda et al., 2007;
Suzuki et al., 2009).
Despite the formation of a massive iron-induced phytoplankton bloom during
SEEDS-1, there was no large POC export flux during the observation period
(Table 5) (Tsuda et al., 2003; Aono et al., 2005; Aramaki et al., 2009).
During SERIES and SEEDS-2, which allowed comprehensive time-series
measurements of the development and decline of the iron-stimulated bloom,
POC fluxes estimated by the drifting traps in the fertilized patch displayed
temporal variations (Boyd et al., 2004; Aramaki et al., 2009). The results
suggested that, subsequently, the drifting trap captured only a small part of
the decrease in ML POC and POC flux losses were mainly governed by bacterial
remineralization and mesozooplankton grazing (Boyd et al., 2004; Tsuda et
al., 2007).
Summary of the significant results from aOIF experiments
Each aOIF experiment has provided new results on basic processes pertaining
to the relationship between pelagic ecology and biogeochemistry, such as
selection of the dominant phytoplankton group or species; the effects of
grazing; interactions within the plankton community; and effects of nutrient
concentrations on the growth of phytoplankton. The aOIF experiments have
generally led to changes in the size of the phytoplankton community from
pico- and nanophytoplankton to microphytoplankton. This effect was
particularly noticeable as diatoms became the dominant species during
IronEx-2, SOIREE, EisenEx, SEEDS-1, SOFeX-S, EIFEX, and SERIES.
Diatom-dominated blooms induced >4.5-fold increases in
chlorophyll a concentrations and accounted for >43 % of the chlorophyll a increase (Cavender-Bares et al., 1999; Boyd et
al., 2000; Gall et al., 2001b; Gervais et al., 2002; Coale et al., 2004;
Tsuda et al., 2005; Marchetti et al., 2006b; Assmy et al., 2007; Smetacek et
al., 2012). The shift to a diatom-dominated community appears to be related
to initial availability of silicate (i.e., initial silicate was >5 µM in all the experiments listed above). However, as silicate
concentrations decreased to <2 µM due to removal by
phytoplankton, diatom blooms rapidly declined. SAGE and LOHAFEX had low
initial levels of silicate (<2 µM). As a consequence, pico-
and nanophytoplankton dominated their communities and diatom growth was
limited by the lack of available silicate. However, during SOFeX-N, initial
silicate limitation (<3 µM) in the iron-fertilized waters was
temporarily relieved through lateral mixing with the surrounding waters that
had relatively higher silicate concentrations (Coale et al., 2004), which
contributed to a taxonomic shift toward diatom-dominated communities (from
16 % to 44 % of total phytoplankton community). These results suggest
that, to develop large-phytoplankton blooms, changeover to a diatom-dominated
community after iron addition is needed. A necessary, but not sufficient,
condition for such a change to occur is the availability of silicate.
Silicate alone is not expected to be sufficient because diatom-dominated
blooms were not observed in all experiments with high initial silicate
concentrations. IronEx-1 and SEEDS-2 had high initial silicate levels
(>3.9 µM) considered conducive to the
development of a diatom-dominated bloom, but blooms were suppressed due to
high grazing pressure. Taken together, the aOIF results suggest that both
mesozooplankton grazing rates and initial silicate concentrations play a
role in limiting the stimulation of diatom-dominated blooms after artificial
iron enrichment.
In experiments with smaller increases (<3.8 times) in plankton
biomass (IronEx-1, SEEDS-2, SAGE, and LOHAFEX) there was little change in
the carbon export flux. Among previous aOIF experiments, the subarctic NP
SEEDS-1 experiment, which was conducted under temperature conditions ideal
for diatom growth (∼8 ∘C) and with shallow MLDs
(∼10 m), produced the greatest changes in surface
phytoplankton biomass. However, influence of iron addition on the
phytoplankton growth extends from surface to euphotic depth as added iron is
mixed within the ML by physical processes (Coale et al., 1998). Although
maximum surface chlorophyll a concentration during SEEDS-1 (∼22 mgm-3) was much higher than EIFEX (∼3.2 mgm-3),
the MLD-integrated chlorophyll a concentrations were similar to
∼250 mgm-2 between the two experiments. Therefore, to
quantify the exact changes in phytoplankton biomass in response to iron
addition, it would be appropriate to consider the MLD-integrated PP for
comparison. During IronEx-2, SOIREE, EisenEx, SEEDS-1, SOFeX-N, SOFeX-S, EIFEX,
and SERIES, a >2-fold increase in PP within the ML, with massive
diatom-dominated blooms, was observed. However, changes in the carbon export
varied substantially and differed from experiment to experiment. In SEEDS-1
and SOIREE there was little increase in export flux. These two experiments
were conducted over only about 2 weeks. The short duration of these
experiments could have prevented the detection of downward carbon export. In
SERIES, there was a distinct increase in the POC export flux within the ML
(MLD = 30 m), but there was no increase in the carbon export flux below
the MLD, and it was reported that the POC produced was rapidly remineralized
due to elevated heterotrophic bacteria respiration within the ML (Boyd et
al., 2004). In SOFeX-S the export flux was enhanced at 100 m, below the MLD
(45 m). However, the changes in export flux, after iron addition, were not
dramatic compared to natural values (Buesseler et al., 2005). It is possible
that the duration of SOFeX-S was also insufficient (∼4 weeks)
(Table 2). EIFEX was the only aOIF experiment that produced significant
carbon export to deeper layers (down to 3000 m). This high flux was due to
aggregate formation with fast sinking rates (Smetacek et al., 2012). EIFEX
observed an entire cycle (i.e., development–decline–fate) of the
iron-induced phytoplankton bloom during the 39 days of the experiment, which
strongly suggests that a sufficient experimental duration is a prerequisite
for diatoms to form aggregates and sink (i.e., carbon export). It should
also be noted that the rates of bacterial remineralization and grazing
pressure on the diatoms were in the same range inside the fertilized patch
as outside, which might have assisted the delivery of iron-induced POC from
the ML to deep layers (Smetacek et al., 2012). These results suggest that to
detect significant carbon exported below the winter MLD following an
increase in PP, at least three conditions are necessary: (1) a shift to a
diatom-dominated community; (2) low bacterial respiration and grazing
pressure rates within the ML; and (3) a sufficient experimental duration,
enabling both immediate and delayed responses to iron addition to be
observed.
Present: unanswered aOIF questions – export flux, possible side effects,
and international law
OIF has been proposed as a potential technique for rapidly and efficiently
reducing atmospheric CO2 levels at a relatively low cost (Buesseler and
Boyd, 2003), but there is still much debate. Over the past 25 years,
controlled aOIF experiments have shown that substantial increases in
phytoplankton biomass can be stimulated in HNLC regions through iron
addition, resulting in the drawdown of DIC and macronutrients (de Baar et
al., 2005; Boyd et al., 2007; Smetacek et al., 2012; Martin et al., 2013).
However, the impact on the net transfer of CO2 from the atmosphere to
below the winter MLD through the biological pump (Fig. 3c) is not yet
fully understood or quantified and appears to vary with environmental
conditions, export flux measurement techniques, and other unknown factors
(Smetacek et al., 2012). There have also been a wide range of the estimates
of atmospheric CO2 drawdown resulting from large-scale and long-term
aOIF based on model simulations (Joos et al., 1991; Peng and Broecker, 1991;
Sarmiento and Orr, 1991; Kurz and Maier-Reimer, 1993; Gnanadesikan et al.,
2003; Aumont and Bopp, 2006; Denman, 2008; Jin et al., 2008; Zahariev et
al., 2008; Strong et al., 2009; Sarmiento et al., 2010). While it is
generally agreed that OIF effectiveness needs to be determined through
quantification of export fluxes, there has been no discussion about which
export flux measurement techniques are the most effective. Meanwhile,
concern has been expressed regarding possible environmental side effects in
response to iron addition (Fuhrman and Capone, 1991). These side effects
include the production of greenhouse gases (e.g.,
nitrous oxide, N2O; and methane, CH4)
(Lawrence, 2002; Jin and Gruber, 2003; Liss et al., 2005; Law, 2008;
Oschlies et al., 2010), the development of hypoxia/anoxia in the water
column (Sarmiento and Orr, 1991; Oschlies et al., 2010; Keller et al.,
2014), and toxic algal blooms (e.g., Pseudo-nitzschia) (Silver et al., 2010; Trick et al.,
2010). These unwanted side effects could lead to negative climate and
ecosystem changes (Fuhrman and Capone, 1991; Sarmiento and Orr, 1991; Jin
and Gruber, 2003; Schiermeier, 2003; Oschlies et al., 2010). Model studies
suggested that the unintended ecological and biogeochemical consequences in
response to large-scale aOIF might cancel out the effectiveness of aOIF. For
example, aOIF-enhanced N2O production may have offset (>∼40 %) the benefits of CO2 sequestration in the EP
(Sarmiento and Orr, 1991; Jin and Gruber, 2003; Oschlies et al., 2010; Hauck
et al., 2016). Core unanswered questions remain concerning the different
carbon export flux results from different measurement techniques (Nodder and
Waite, 2001; Aono et al., 2005), the possible side effects that could
directly influence the aOIF effectiveness, and the legal framework that is
in place to regulate aOIF operations while simultaneously supporting further
studies to increase our understanding of the potential risks and benefits of
aOIF (Williamson et al., 2012). With the design of future aOIF experiments
in mind, the following section discusses these core questions: (1) which of
the methods are optimal for tracking and quantifying carbon export flux, (2) which
of the possible side effects have negative impacts on aOIF
effectiveness, and (3) what are the international aOIF experimentation laws
and can they be ignored?
Export flux measurement methods
A traditional, direct method for estimating POC export fluxes in the water
column is a sediment trap that collects sinking particles (Suess, 1980).
Sediment traps are generally deployed at specific depths for days to years
to produce estimates of total dried mass, POC, particulate inorganic carbon
(PIC), particulate organic nitrogen (PON), particulate biogenic silica,
δ13Corg, and 234Th. A basic assumption for the use of
a sediment trap is that it exclusively collects settling particles,
resulting from the gravitational sinking of organic matter produced in
surface waters. However, although they are designed to ensure the
well-defined collection and conservation of sinking particles, they have
accuracy issues due to (1) interference of the hydrodynamic flow across the
trap (i.e., strong advective flow), (2) inclusion and/or invasion (accounting for
14 %–90 % of the total POC collected) of metazoan zooplankton (e.g.,
copepods, amphipods, and euphausiids) capable of vertical migration (Karl
and Knauer, 1989; Buesseler, 1991; Buesseler et al., 2007), and (3) loss of
trapped particles by bacterial decay and/or dissolution during trap
deployment and storage periods (Gardner et al., 1983; Knauer et al., 1984;
Kähler and Bauerfeind, 2001). The application of sediment traps for the
determination of the carbon export flux is relatively more biased in the ML
where ocean currents are generally faster and zooplankton are much more
active than deep water. These issues suggest that sediment traps alone may
not accurately determine carbon export fluxes within the ML.
Even when used at the same depth, traditional sediment traps, such as the
surface-tethered drifting trap and bottom-moored trap, can greatly over- or
underestimate particulate 234Th fluxes compared to water-column-based
estimates (Buesseler, 1991). The water-column-based total 234Th
deficiency method (the sum of dissolved and particulate activities) is less
sensitive than sediment traps to the issues mentioned above and provides
better spatial and temporal resolution in flux estimates (Buesseler, 1998).
For these reasons, traditional sediment trap POC flux estimates have often
been calibrated using the total 234Th deficiency measured using
rosette bottle or high-volume pump samples (Coale and Bruland, 1985;
Buesseler et al., 2006) as a reference. However, the water-column-based
234Th method is sensitive to the characterization of the POC to
234Th ratio on sinking particles and the choice of 234Th flux
models (Buesseler et al., 2006). Therefore, sampling to estimate the POC to
234Th ratio should be conducted below MLD to accurately detect downward
carbon export flux into intermediate–deep waters.
Several aOIF experiments have used both sediment traps and 234Th
deficiency to estimate the iron-induced POC export flux (Table 5). SOIREE
reported distinct differences in POC fluxes estimated from drifting traps
(185 mgm-2day-1) at a 110 m over day 11–13 of the experiment and
234Th (∼87 mgm-2day-1) at 100 m (Charette and
Buesseler, 2000; Nodder and Waite, 2001). While there was no measurable
change in 234Th-based POC fluxes during the 13 days of the SOIREE
experiment (Fig. 8b), the traps suggested a 27 % increase over the course
of the experiment (from 146 to 185 mgm-2day-1) (Table 5). It was
later discovered that the sediment-trap-based sampling biases caused this
supposed increase (Nodder et al., 2001; Nodder and Waite, 2001). Likewise,
in SEEDS-1, 234Th-based POC fluxes at 50 m over day 9–13 were
estimated to be 423 mgm-2day-1, but the drifting trap only
recorded 141 mgm-2day-1 at 40 m over day 12–14, which is 3 times lower
(Table 5) (Aono et al., 2005; Aramaki et al., 2009). This large discrepancy
between the two methods might be caused by the under-sampling of POC into
the drifting traps (Aono et al., 2005).
To resolve the potential biases in traditional sediment traps, a neutrally
buoyant (and freely drifting) sediment trap (NBST) was developed (Valdes and
Price, 2000; Valdes and Buesseler, 2006). Through preliminary experiments
conducted in June and October 1997 at the Bermuda Atlantic Time-series Study
site, Buesseler et al. (2000) showed that an NBST system could reduce the
horizontal flow and invasion and/or inclusion of zooplankton into the trap
samplers and that NBST-based 234Th fluxes were comparable with
water-column-based estimates. LOHAFEX has been the only aOIF experiment so
far that has measured particle export using PELAGRA (Particle Export
measurement using a Lagrangian trap) sediment traps based on the NBST system
deployed at two depths of 200 and 450 m (below the winter MLD) (Martin et
al., 2013). However, the PELAGRA sediment traps did not detect aOIF-induced
carbon export even though PP did increase within the ML. Water-column-based
234Th measurements estimated the POC flux at 100 m to be
∼94 mgm-2day-1, whereas the PELAGRA sediment traps
estimated the flux at 200 and 450 m to be
<12 mgm-2day-1 (Table 5)
(Martin et al., 2013). It should be noted that
both sediment traps and water-column-based 234Th measurements have a
limited ability to fully scan the vertical profile of POC fluxes and,
therefore, these methods should ideally be complemented with additional
techniques that can measure particle stocks at high depth resolution
throughout the water column.
To resolve the full column more effectively, LOHAFEX employed an underwater
video profiler (UVP), which provided photographic evidence of sinking
particles (particle size ≥100 µm) from the surface down to
∼3000 m, with ∼0.2 m vertical resolution
(Smetacek and Naqvi, 2010; Martin et al., 2013). Through an analysis of
particle size distributions, the UVP also allowed particles to be classified
into fecal pellets, aggregates, and live zooplankton. Total vertical
particle volume profiles obtained from the UVP indicated a maximum
concentration at 75 m (∼0.3 mm3L-1), with a
gradual decrease to 150 m (∼0.15 mm3L-1).
Interestingly, large particles (i.e., zooplankton) were copious between 75
and 100 m, suggesting that there might be high grazing pressure. Heavy
grazing might explain the large discrepancy between the 100 m (water-column-based
234Th method) and 200 and 450 m (PELAGRA sediment trap) POC flux
estimates (i.e., rather than a sampling bias in sediment trap data) (Martin
et al., 2013). To continuously monitor vertical changes in POC stocks
following iron addition, EIFEX used a transmissometer, providing high
vertical resolution (∼24 data points per meter) and tracking
of the iron-induced stocks down to ∼3000 m, even though,
unlike UVPs, transmissometers do not allow classification of particles
(Smetacek et al., 2012). Improving on this method, SOFeX-N applied
autonomous carbon explorers equipped with transmissometers, designed to
float along with the currents. Three autonomous carbon explorers were
deployed, two explored the iron-fertilized patch and one acted as a
control outside the patch. Carbon explorers could continuously monitor
carbon flux in the field for up to 18 months beyond the initial deployment,
which allowed SOFeX-N to observe episodic raining in the iron-fertilized
waters (Bishop et al., 2004), indicating a high carbon export flux after
artificial iron addition. Furthermore, recent studies also reported that use
of optical spike signals in particulate backscattering and fluorescence,
measured from autonomous platforms such as gliders and floats, can provide
high-resolution observations of POC flux (Briggs et al., 2011; Dall'Olmo and
Mork, 2014).
The combination of multiple approaches is essential to the successful
detection of POC produced in response to iron addition and its fate. NBST
systems (e.g., the PELAGRA sediment trap) should be deployed at two depths
(i.e., below both the in situ MLD and the winter MLD) to quantify the aOIF-induced
POC flux. This technique is improved when accompanied by calibration using
water-column-based 234Th. Particle profiling systems (e.g., a
transmissometer and an UVP) can provide continuous quantitative and
qualitative information about sinking particles, with high vertical
resolution and full coverage of the water column (>3000 m).
They are therefore useful for indirectly identifying deep carbon transport.
Autonomous carbon explorers are an excellent alternative, allowing for
continuous observation of POC fluxes during and after an aOIF experiment.
Considering environmental side effects
The purpose of aOIF is to reduce the atmospheric CO2 level by
stimulating the sequestration of oceanic carbon through artificial iron
additions in the HNLC regions, mitigating the global warming threat. Beyond
the benefits of aOIF experimentation, scientists have debated the unintended
secondary consequences of aOIF, such as production of climate-relevant gases
and ocean ecosystem changes. Therefore, it is important to consider the
possible negative consequences of aOIF to evaluate whether the aOIF
experiments are effective (i.e., net profit: positives > negatives).
To investigate changes in climate-relevant gas emissions produced by
biological activities and/or photochemical reactions before and after iron
additions, the production of CH4, N2O, DMS, and halogenated
volatile organic compounds (HVOCs) was measured during aOIF experiments
(Liss et al., 2005) because their emission may lead to unintended
consequences negating the desired effects of aOIF experiments on carbon
sequestration. Among the climate-relevant gases, CH4 has a
∼20 times greater warming potential than CO2 (IPCC,
1990). However, CH4 has been considered to be relatively low risk
because most of the CH4 formed in the ocean is used as an energy source
for microorganisms and is converted to CO2 before reaching the sea
surface (Smetacek and Naqvi, 2008; Williamson et al., 2012). The SO nOIF
experiment conducted in year 2011 (i.e., Kerguelen Ocean and Plateau
compared Study 2: KEOPS-2) (Table 1 and Fig. 4a) showed that CH4
concentrations were 4-fold higher in the naturally iron-fertilized patch
than in the control area (Farías et al., 2015). During the SOFeX-N
experiment, measurements of dissolved CH4 indicated concentrations were
slightly elevated, i.e., by less than 1 % (1.74 ppmv in the fertilized patch
and 1.72 ppmv outside the fertilized patch) (Wingenter et al., 2004). Simulated
SO large-scale aOIF has suggested that a 20 % enhancement of CH4
emissions would offset only <1 % (∼4 TgCyr-1)
of the resulting carbon sequestration (Oschlies et al., 2010).
Hence, additional CH4 production from aOIF experiments is not likely to
be significant.
On the other hand, N2O has a relatively long lifetime in the atmosphere
(∼110 years) and has a global warming potential about 300 times
greater than CO2 (Forster et al., 2007). The ocean is already a
significant source of atmospheric N2O (Nevison et al., 2003; Bange,
2006). Oceanic N2O is mainly produced by bacterial remineralization.
Therefore, increases in N2O production after iron additions are
expected and, in the long run, contribute to an increase rather than a
decrease in the greenhouse effect (Fuhrman and Capone, 1991). During the
SOIREE experiment, a significant increase (∼4 %) in mean
N2O saturation in the pycnocline (65–80 m) of the fertilized patch
(104.4%±2.4 %), as compared to outside the fertilized patch (100.3%±1.7 %), was associated with an increased phytoplankton biomass
(Law and Ling, 2001). Measurements of N2O saturation during SERIES also
showed increases of 8 % at 30–50 m, which were coincident with the
accumulation of ammonium and nitrite attributable to increases in bacterial
remineralization following increased POC levels (Boyd et al., 2004; Law,
2008). SOIREE-based model estimates suggested that potential N2O
production at timescales longer than 6 weeks would subsequently offset
carbon reduction benefits resulting from the bacterial remineralization of
additional carbon fixation by 6 %–12 % (Law and Ling, 2001). This estimate
is in line with the N2O offset of 6 %–18 % suggested by a modeling
study (Jin and Gruber, 2003) and the 5 %–9 % suggested by a more recent
modeling study investigating the effectiveness of long-term and large-scale
SO aOIF (Oschlies et al., 2010). However, the SO nOIF experiment (i.e.,
KEOPS-2) suggested that nOIF acts as both a sink and a source for N2O
(Farías et al., 2015). Excess N2O was not found after iron
addition in EIFEX, where significant vertical export through the formation
of rapidly sinking aggregates was found (Walter et al., 2005; Law, 2008).
One explanation for the absence of N2O accumulation below the EIFEX
patch might be the limited bacterial remineralization due to the rapid
export of organic matter well below 500 m to the seafloor (Law, 2008).
Based on the results of previous studies, no consensus has yet been reached
on the exact extent of additional N2O production after iron additions.
However, because there is the potential for excessive N2O production
that would not only impact the effectiveness of aOIF experiments but also
positively contribute to global warming, further studies are required to
reach a conclusion.
Unlike N2O emissions, which have the potential to offset the
effectiveness of aOIF, DMS, a potential precursor of sulfate aerosols that
cause cloud formation, may contribute to the homeostasis of the Earth's
climate by countering the warming due to increased CO2 emissions
(Charlson et al., 1987). DMS is produced by the enzymatic cleavage of
planktonic dimethylsulfoniopropionate (DMSP). Microzooplankton grazing on
nanophytoplankton (e.g., haptophytes) is a key factor controlling oceanic
DMS production (Dacey and Wakeham, 1986; Gall et al., 2001a; Park et al.,
2014). The production of DMS in response to iron addition was measured
during all aOIF experiments. In the EP and SO, DMS production increased, but,
in the subarctic NP, it remained constant or decreased (Boyd et al., 2007;
Law, 2008). There were significant short-term increases in DMS production in
IronEx-2 (from 2.5 to 4.2 nM), SOIREE (from 0.5 to 3.4 nM), EisenEx (from
1.9 to 3.1 nM), and SOFeX-N (7.7 nM in the fertilized patch and
1.6 nM
outside the fertilized patch) (Turner et al., 1996, 2004;
Wingenter et al., 2004, 2007; Liss et al., 2005). The
maximum DMS production observed was a 6.8-fold increase after iron addition
in SOIREE (Turner et al., 2004). During the SOIREE experiment, the initial
dominant phytoplankton species were haptophytes and they remained dominant
until day 7. Since then, DMS production was increased by microzooplankton
grazing on DMSP-rich haptophyte groups (e.g., Phaeocystis) (Gall et al., 2001a).
Similarly, a 4.8-fold enhancement of DMS production was observed in SOFeX-N.
Estimates derived by the extrapolation of SOFeX-N DMS production results
suggested that fertilizing ∼2 % of the SO area over the
course of a week would result in a 20 % increase in the total SO DMS flux,
which would lead to a 2 ∘C decrease in air temperature over the SO
(Wingenter et al., 2007). On the other hand, the SO nOIF experiment
(KEOPS-1) conducted in year 2005 (Table 1 and Fig. 4a) showed that DMS
production was not markedly higher in the naturally fertilized area compared
to the surrounding waters (Belviso et al., 2008). In addition, a 20-year
aOIF simulation through a three-dimensional ocean biogeochemical model did not
show significant increase in DMS emissions from the SO (Bopp et al., 2008).
Interestingly, there were no significant changes in DMS production after
iron additions in the western subarctic NP SEEDS-1 and 2 experiments, despite
increases in PP (Takeda and Tsuda, 2005; Nagao et al., 2009). Furthermore,
in the eastern subarctic NP, SERIES DMS production increased from 8.5–10.9 nM
on day 1 to a maximum of 41.2 nM on day 10, but decreased to <0.03 nM by the end of the experiment due to an increase in bacterial
abundance (Table 4) (Levasseur et al., 2006). It is therefore difficult to
predict the iron-induced DMS response, because OIF itself is not the only
source of DMS. Based on the results of previous aOIF experiments, DMS
production was sensitive in the EP and SO, but was less sensitive in the
subarctic NP (Law, 2008). These results indicate that further process and
modeling studies for each region are required to determine the production
and degradation of DMS, both following iron fertilization and in the natural
environment.
HVOCs, such as CH3Cl, CH3Br, and CH3I, are well known for
their ability to destroy ozone in the lower stratosphere and marine boundary
layer (Solomon et al., 1994) and were also measured during past aOIF
experiments (Wingenter et al., 2004; Liss et al., 2005). However, no
consistent results have been reported for HVOCs production (Liss et al.,
2005). In SOFeX-N, the impact of iron addition on HVOCs was complicated,
with CH3Cl concentrations remaining unchanged and CH3Br
concentrations increasing by 14 % (6.5 pptv in the fertilized patch and
5.7 pptv outside the fertilized patch), while CH3I concentrations
decreased by 23 % (4.9 pptv in fertilized patch and 6.4 pptv outside the
fertilized patch) (Wingenter et al., 2004). In contrast, CH3I
concentrations increased ∼2-fold during EisenEx (Liss et al.,
2005). Such a complicated response suggests that, as for DMS, further study
is needed to fully understand natural cycling of HVOCs and their responses
to aOIF.
Another important consideration is the extent to which the effectiveness of
aOIF is canceled out by its tendency to lead to ocean ecosystem changes
such as a decrease in dissolved oxygen and an increase in domoic acid (DA)
levels. The decomposition of iron-addition-enhanced biomass may cause
decreased oxygen concentrations in subsurface waters, but midwater oxygen
depletion has not been reported from aOIF experiments to date (Williamson et
al., 2012). Early modeling studies suggest that anoxic conditions may
develop after long-term, large-scale aOIF (Fuhrman and Capone, 1991;
Sarmiento and Orr, 1991), whereas a recent study based on more sophisticated
models showed sustained well-oxygenated conditions (O2≈120 µM) even under simulated aOIF south of 30∘ S on a 100-year
timescale from 2010 to 2110 (Oschlies et al., 2010). Keller et al. (2014)
found that simulated SO large-scale aOIF south of 40∘ S from the
year 2020 to 2100 under a high-CO2-emissions scenario (Meinshausen et
al., 2011) may develop suboxia (O2<10 µM) in the year 2125.
Clearly, the circumstances under which a substantial decline in oxygen
inventory can be caused by large-scale aOIF need further study.
The changes in phytoplankton community composition after iron addition
discussed in Sect. 2.4 may also have unintended consequences. For example,
such changes could lead to potentially toxic species dominating plankton
assemblages (Silver et al., 2010; Trick et al., 2010). Some aOIF experiments
(e.g., IronEx-2, SOIREE, EisenEx, SOFeX-N, SOFeX-S, and SERIES) generated large
blooms dominated by pennate diatoms belonging to the genus
Pseudo-nitzschia (de Baar et al., 2005; Silver et al., 2010; Trick et al., 2010). Some
Pseudo-nitzschia species have the capacity to produce the neurotoxin DA that can
detrimentally affect marine ecosystems. However, no DA was found during
EisenEx and SERIES, even though Pseudo-nitzschia were dominant (Gervais et al., 2002; Assmy
et al., 2007; Marchetti et al., 2008). However, phytoplankton samples used
to estimate DA production have sometimes been stored for a long time before
the analysis, for example, 12 years in IronEx-2 and 4 years in SOFeX-S
(Silver et al., 2010). Trick et al. (2010) argued that storage might have
affected the DA content in the samples, which led to an underestimation in
DA concentrations. Nevertheless, discernable changes in DA production were
found in IronEx-2 and SOFeX-S experiments (Silver et al., 2010). It is
likely that detection was possible because these samples were collected with
net tows (20 to 30 µm mesh phytoplankton nets), which provided
concentrated samples of larger phytoplankton including Pseudo-nitzschia (e.g.,
Pseudo-nitzschia abundance: 1.3×106 cellsL-1 in IronEx-2 and
7.5×104 cellsL-1 in SOFeX-S). During IronEx-2 and SOFeX-S,
high cell abundances of Pseudo-nitzschia (106 and 105 cellsL-1,
respectively) combined with moderate DA cell quotas (0.05 and 1 pgDAcell-1,
respectively) produced toxin levels as high as 45 and 220 ngDAL-1 in the water, respectively, i.e., toxin
levels high enough to damage marine communities in coastal waters (Scholin
et al., 2000; Schnetzer et al., 2007). Trick et al. (2010) suggested that
large-scale OIF may induce DA accumulation with developing toxic
Pseudo-nitzschia blooms. However, large uncertainties remain as Trick et al. (2010) simply
extrapolated DA concentration based on bottle incubation experiments with
HNLC surface waters to the DA production expected from large-scale OIF. As a
result, it is necessary to clarify and quantify DA production in response to
aOIF, with concentrated larger phytoplankton samples collected using net
tows (20 to 30 µm mesh phytoplankton nets). Here again, existing
research indicates that the processes involved need to be better understood
in the natural environment before the ramifications of aOIF can be fully
understood.
Whether aOIF is a viable carbon removal strategy is still under debate (Boyd
et al., 2007; Smetacek and Naqvi, 2008). The production of climate-relevant
gases such as N2O, DMS, and HVOCs, which is influenced by the
remineralization of sinking particles that follows OIF-induced blooms; the
decline in oxygen inventory; and the production of DA are particularly
important to understand. These processes can directly and indirectly modify
the effectiveness of carbon sequestration, with either positive or negative
effects. Therefore, monitoring declines in oxygen content and production of
climate-relevant gases and DA to evaluate the effectiveness of aOIF as a
geoengineering approach is essential. The processes discussed here represent
the current state of knowledge concerning aOIF side effects. The direct and
indirect environmental consequences remain largely unresolved due to the
inconsistent and highly uncertain outcomes of the experiments conducted so
far, as well as our poor understanding of the processes involved under both
nOIF and aOIF conditions (Chisholm et al., 2001; Johnson and Karl, 2002;
Williamson et al., 2012). However, considering the increasing evidence for
the necessity to keep warming at or below 1.5 ∘C (Rogelj et al.,
2015), there continues to be a need to quantitatively determine the
effectiveness of aOIF as a long-term means for reducing atmospheric CO2
through the quantification of aOIF side effects.
Assessment framework for scientific research involving ocean
fertilization (OF) (modified from Resolution LC-LP.2, 2010).
Regulation of aOIF: international law of the sea as it applies to
aOIF
To prevent pollution of the sea from human activities, the international
Convention on the Prevention of Marine Pollution by Dumping of Wastes and
Other Matter (London Convention, 1972) was amended in 1972. In 1996,
contracting parties to the London Convention adopted the Protocol to the
London Convention (London Protocol, 1996). This places legal restrictions on
the dumping of wastes and other matter that may cause hazard, harm, and
damage in the ocean and/or interfere with the marine environment. However,
the London Convention and Protocol (LC/LP) did not establish specific laws
to protect the ocean environment against the side effects of fertilization
activities. In 2007, several commercial companies (e.g., GreenSea Venture,
http://www.greenseaventure.com, last access: 6 September 2018; and Climos,
http://www.climos.com, last access: 6 September 2018)
promoted large-scale (10 000 km2) commercial
aOIF as a climate mitigation strategy and as a means to gain carbon credits
(Chisholm et al., 2001; Buesseler and Boyd, 2003; Freestone and Rayfuse,
2008). Meanwhile, assessments of the effectiveness of aOIF have been limited
to small fertilized patches (25–300 km2) (Table 1 and Fig. 6a) due to
the time and expense of comparing fertilized and unfertilized areas (ACE
CRC, 2008). As discussed earlier, these small-scale experiments have left
many unanswered scientific questions regarding both the effectiveness and
the potential impacts of aOIF (Lawrence, 2002; Buesseler and Boyd, 2003). In
the same year, noting the potential risks and benefits, the LC/LP scientific
group released a statement on large-scale ocean fertilization and
recommended that ocean fertilization activities be evaluated carefully to
ensure that such operations were not contrary to the aims of the LC/LP.
At the 2008 LC/LP meeting, the contracting parties adopted Resolution
LC-LP.1 (2008) on the regulation of ocean fertilization. This resolution
prohibited ocean fertilization activities until such time that specific
guidance could be developed to justify legitimate scientific research. There
was an exception for “small-scale scientific research studies within coastal
waters” to permit the development of proposals that would lead to an
assessment framework for scientific ocean fertilization research (Resolution
LC-LP.1, 2008). In the meantime, there was a call to develop an assessment
framework for ocean fertilization experiments to assess, accurately,
scientific research proposals (Resolution LC-LP.1, 2008). In 2010, LC/LP
parties developed Resolution LC-LP.2 (2010), adopting an Assessment
Framework for Scientific Research Involving Ocean Fertilization to be used
to assess, on a case-by-case basis, whether any proposed ocean fertilization
activity constitutes legitimate scientific research falling within the aims
and scope of Resolution LC-LP.1 (2008) (Fig. 9) (Resolution LC-LP.2, 2010).
This framework demands preliminary scientific research prior to any aOIF
experimentation. There must be a transparent/reasonable scientific
rationale/purpose to the experiment and a risk analysis must be undertaken
using parameters such as problem formulation, site selection, exposure and
effect assessment, and risk characterization and management. Monitoring is
also required as an integral component of all approved (i.e., legitimate)
scientific aOIF research activity to assess ecological impacts and to review
actual vs. intended geoengineering benefits (ACE CRC, 2015). The process of
acquiring permission for an aOIF experiment from the LC/LP is a
multifaceted enterprise involving not only ecology, biogeochemistry, and
climate science (i.e., Martin's iron hypothesis), but also social sciences
(i.e., ethics and efficacy of climate-engineering measures) and ocean
governance (i.e., international law of the sea and its enforcement). In
October 2013, the LC/LP parties adopted amendments that categorize aOIF as
marine geoengineering, thereby prohibiting operational aOIF activities, but
enabling aOIF scientific research that meets the permit conditions through
the environmental assessment framework (Resolution LP.4 (8), 2013). This
means that large-scale (i.e., >300 km2 based on previous
aOIF experiments; exact areal sizes are not determined in the LC/LP) and/or
commercial aOIF (e.g., the 2012 Haida Gwaii Iron Dump off the west coast
of Canada) are currently banned by international regulations. Under LC/LP,
commercial aOIF efforts cannot proceed because of the large uncertainties
related to large-scale aOIF.
Future: designing future aOIF experiments
Scientific aOIF research has focused on improving our understanding of the
effectiveness, capacity, and risks of OIF as an atmospheric CO2 removal
strategy both in the future and the past (in particular glacial periods).
Although the first aOIF experiments took place more than 20 years ago,
the legal and economic aspects of such a strategy in terms of the
international laws of the sea and carbon offset markets are not yet clear
(ACE CRC, 2015). Nonetheless, previous small-scale aOIF experiments have
demonstrated a considerable potential for easily and effectively reducing
atmospheric CO2 levels. Accordingly, physical–biogeochemical–ecological
models and nOIF experiments (long-term) have been conducted in an effort to
overcome some of the limitations of short-term aOIF experiments (e.g.,
spatial and temporal scales) and to predict the effectiveness of long-term
and large-scale fertilization (Aumont and Bopp, 2006; Blain et al., 2007;
Denman, 2008; Pollard et al., 2009; Sarmiento et al., 2010). For example,
earlier global biogeochemical models have indicated that massive
fertilization could draw down atmospheric CO2 by as much as 107 µatm
in 100 years (Joos et al., 1991; Peng and Broecker, 1991; Sarmiento and
Orr, 1991; Kurz and Maier-Reimer, 1993). Recent global models, with more
realistic ecosystem and biogeochemical cycles, predict values closer to a 33 µatm
drawdown in atmospheric CO2 (Aumont and Bopp, 2006). These
results suggest that the amount of carbon sequestration resulting from aOIF
represents only a modest offset, i.e., a contribution of ∼10 % over the range of IPCC future emission scenarios (IPCC, 2000; Aumont
and Bopp, 2006; Denman, 2008; Zahariev et al., 2008). The nOIF experiments
have also produced much higher carbon sequestration rates than the
small-scale aOIF experiments (Morris and Charette, 2013). Furthermore, the
results from nOIF experiments do not support the potential negative impacts
proposed for OIF experiments, even at larger scales (Belviso et al., 2008).
However, these nOIF results do not guarantee that aOIF as a geoengineering
approach is able to achieve the high effectiveness associated with carbon
sequestration and enables a simple scaling up as a prediction tool, because
the nOIF experiments differ from the aOIF experiments in the mode of iron
supply. In particular, nOIF is a continuous and slow process and its iron
source is based on the upwelling of iron-rich subsurface waters to the
surface layer, whereas aOIF is intended to be episodic, with massive
short-term iron additions (Blain et al., 2007). In addition, in nOIF it is
difficult to accurately identify iron sources due to the complexity of the
system, whereas in aOIF there is quantitative and qualitative information
about iron additions and sources (Blain et al., 2008). Contrary to the
results of aOIF experiments in the SO (e.g., SOIREE and SOFeX-N), no
increase in DMS emissions was found in the SO nOIF experiment (i.e., KEOPS-1)
(Belviso et al., 2008), suggesting that it might be difficult to identify
the potential long-term negative effects of aOIF through the study of
naturally fertilized systems, at least in the SO. There is also a broad
swath of hypotheses in the fields of pelagic ecology–biogeochemistry that
can be tested with OIF experiments using the
correlations between temperature, CO2 concentrations, and dust over the
past four glacial–interglacial cycles on the one hand and bottle experiments
showing iron limitation of phytoplankton growth in HNLC regions on the
other. Therefore, it is important to continue undertaking small-scale
studies to obtain a better understanding of natural processes in the SO as
well as to assess the associated risks and so lay the groundwork for
evaluating the potential effectiveness and impacts of large-scale aOIF as a
geoengineering solution to anthropogenic climate change. It is therefore of
paramount importance that future aOIF experiments continue to focus on the
effectiveness and capacity of aOIF as a means of reducing atmospheric
CO2, but they should also carefully consider the location (i.e.,
where), timing (i.e., when), and duration (i.e., how long), as well as
modes of iron addition (i.e., how), tracing methods and measurement parameters
(i.e., what), and side effects on marine/ocean
ecosystems (i.e., what concerns). They should build on the results of
previous aOIF experiments to develop our understanding of the magnitude and
sources of uncertainties and provide confidence in our ability to reproduce
results.
Where. The first consideration for a successful aOIF experiment is the
location. The dominance of diatoms in phytoplankton communities plays a
major role in increasing the biological pump because diatom species can sink
rapidly as aggregates or by forming resting spores to efficiently bypass the
intense grazing pressure of mesozooplankton (e.g., copepods, salps, and
krill) and export carbon out of the winter ML (Tréguer et al., 1995;
Salter et al., 2007; Assmy et al., 2013; Rembauville et al., 2015, 2016). Previous aOIF experiments have shown that
silicate concentration and mesozooplankton stocks (i.e., copepods) are the
crucial factors controlling diatom blooms (Boyd et al., 2000; Gervais et
al., 2002; Coale et al., 2004; Tsuda et al., 2007; Smetacek et al., 2012).
Therefore, to obtain the greatest possible carbon export flux in response to
iron addition, aOIF experiments should be designed in regions with high
silicate concentrations and low grazing pressure. It will be important to
conduct initial surveys to measure the degree of grazing pressure in HNLC
regions with high silicate concentrations such as in the subarctic NP (e.g.,
SEEDS-1 experiment) and the south of the SO PF (e.g., SOFeX-S experiment)
>15 µM (Fig. 4c). In selecting sites for
aOIF, it is also important to distinguish the iron-fertilized patch from the
surrounding unfertilized waters to easily and efficiently observe
iron-induced changes (Coale et al., 1996). Ocean eddies provide an excellent
setting for aOIF experimentation because they tend to naturally isolate
interior waters from the surrounding waters. Mesoscale eddies range from 25
to 250 km in diameter and maintain their characteristics for 10–100 days
after formation (Morrow and Le Traon, 2012; Faghmous et al., 2015). Eddy
centers tend to be subject to relatively slow current speeds, with low shear
and high vertical coherence, providing ideal conditions for tracing the same
water from the surface to below the winter MLD, while simultaneously
minimizing lateral stirring and advection (Smetacek et al., 2012). Finding
an appropriate eddy setting in a study area should be a high priority
consideration when designing an aOIF experiment (Smetacek and Naqvi, 2008).
Mesoscale eddies can be reliably identified and tracked with satellite
sea surface height anomalies (Smetacek et al., 2012).
When. The second consideration for a successful aOIF experiment is timing,
which includes when an experiment starts. PP in ocean environment is
generally limited by nutrient availability and/or by light availability,
often referred to as a single or co-limitation. PP in the SO, a
representative HNLC region, is subject to co-limitation by
micro- or macronutrients (i.e., iron and/or silicate) and light availability
(Mitchell et al., 1991). To the south of the SO PF, phytoplankton blooms
usually occur during early summer (i.e., from late December to early
January) due to an increase in the nutrient flux from subsurface waters
induced by winter mixing, along with the favorable light conditions provided
by a shoaling of the ML (Moore and Abbott, 2002). Weekly and monthly
climatological maps of chlorophyll a concentrations derived from satellite
data could provide the necessary information for determining the timing of
blooms in the SO PF (Westberry et al., 2013). Prior to December,
phytoplankton growth is mainly limited due to light availability (Mitchell
et al., 1991; Veth et al., 1997; Abbott et al., 2000), while after January
(i.e., during late summer and early autumn from February to March) it is
mainly limited due to iron and silicate availability (Abbott et al., 2000;
Mengelt et al., 2001; Nelson et al., 2001). In previous SO aOIF experiments
conducted between spring and early autumn, PP was mainly limited by iron
and/or silicate availability rather than light availability (except when
heavy clouds led to severe light limitation, which only occurred for a few days
during EisenEx) (Gervais et al., 2002; Bakker et al., 2005; Smetacek and
Naqvi, 2008; Peloquin et al., 2011a). The grazing pressure of
mesozooplankton on large diatoms was also a major limiting factor in diatom
production (Schultes et al., 2006; Smetacek and Naqvi, 2010) and was
generally higher during late summer and early autumn (February to March)
(Hunt and Hosie, 2006; Rembauville et al., 2015). Considering the key
factors (i.e., micro- or macronutrient availability, light availability, and
grazing pressure) controlling PP in the SO, the most appropriate timing for
the start of an aOIF experiment to the south of the SO PF is likely to be
the early summertime (i.e., late December to early January).
(a) Time series of mixed-layer-depth-integrated chlorophyll a
concentrations (mgm-2) during the Southern Ocean Iron Release
Experiment (SOIREE) (brown line), Subarctic Pacific iron Experiment for
Ecosystem Dynamics Study 1 (SEEDS-1) (coral line), Subarctic Ecosystem
Response to Iron Enrichment Study (SERIES) (cyan line), SEEDS-2 (blue line),
and European Iron Fertilization Experiment (EIFEX) (teal line). (b) The
distributions of chlorophyll a concentrations (mgm-3) on day 5 and day 42
during SOIREE from SeaWiFS Level-2 daily images. Sources are Gall et al. (2001a), Tsuda et al. (2007), and Assmy et al. (2013).
How long. The third consideration for a successful aOIF experiment is the
duration. The periods in which phytoplankton blooms have been maintained by aOIF
have lasted from ∼10 to ∼40 days (Kolber et
al., 1994; Martin et al., 1994; Coale et al., 1996, 2004; Boyd et al., 2000, 2004; Tsuda et al., 2005; Smetacek et al.,
2012). Although the first 2 weeks have a decisive effect on the development
and demise of the bloom, it has been suggested that most aOIF experiments
did not cover the full response times from onset to termination (Boyd et
al., 2005). For example, SOIREE and SEEDS-1 had relatively short observation
periods (13 days) and saw increasing trends in PP throughout the experiments
(Fig. 10a), suggesting that the observation period should have been
extended. Furthermore, after the end of SOIREE, ocean-color satellite images
showed continued high chlorophyll a concentrations (>1 mgm-3)
in the iron-fertilized patch, which was visible as a long ribbon
shape that extended some 150 km for >40 days (∼6 weeks)
after the first iron addition (Fig. 10b) (Abraham et al., 2000;
Westberry et al., 2013). This indicates that short experimental durations
may not be sufficient for detecting the full influence of aOIF on PP and
the ecosystem (Figs. 8b and 10), although some useful information can be gleaned
over short periods, including dominance of spore-forming Chaetoceros species during
the SEEDS-1 experiment (Boyd et al., 2000; Tsuda et al., 2003, 2005; de Baar et al.,
2005). SOFeX-S also resulted in relatively low export
production despite the high PP due to the experimental duration being
insufficient to cover the termination of the phytoplankton bloom. However,
SERIES, SEEDS-2, EIFEX, and LOHAFEX did fully monitor all phases of the
phytoplankton bloom from onset to termination. EIFEX, the third-longest aOIF
experiment, at 39 days, was the only one that observed iron-induced deep
export production between day 28 and 32 (Table 5 and Fig. 8a) (Smetacek et
al., 2012; Assmy et al., 2013). Furthermore, long-term observations covering
the later stage of bloom development during nOIF experiments resulted in
much higher C:Fe export efficiencies compared to the short-term aOIF (Blain
et al., 2007; Pollard et al., 2009). Based on previous aOIF experiments, it
would, therefore, be important to detect the full phase of a phytoplankton
bloom to determine accurately the amount of iron-induced POC exported out of
the winter ML. The observation period is, therefore, an important
consideration with regard to budget and effectiveness estimates. It is
suggested that the experimental duration should be a minimum of
∼40 days based on the SOIREE experiment, which produced the
longest iron-induced bloom (>40 days). In addition, autonomous
observation platforms are essential to monitor post-assessment of
effectiveness, capacity, and risks of aOIF for at least 12 months after
experiment termination.
How. The fourth consideration for a successful aOIF experiment lies in the
strategy/approach of adding and maintaining dissolved iron within the ML to
produce a phytoplankton bloom. First, the chemical form for iron addition
should be acidified iron sulfate, which is less expensive and more
bioavailable than other iron compounds. The amount of iron sulfate required
is calculated according to the target concentration of the dissolved iron
and volume (MLD × patch size). Based on bottle incubation
experiments, target iron concentrations of ∼2–4 nM are
recommended to stimulate maximum phytoplankton growth due to the rapid
losses of added iron by horizontal advection–diffusion and oxidation to
poorly bioavailable iron(III) (Coale et al., 1996, 1998; Bowie
et al., 2001). For patch size, a biogeochemical model study showed that a
fertilized patch size of 156 km2 maintained an iron concentration above
0.3 nM for 56 days, while a longer period of 194 days required a fertilized
patch size of 160 000 km2 (Xiu and Chai, 2010). As a consequence of
expansion and dilution, previous aOIF experiments also produced similar
results to this model study. The lateral dilution rate (<0.25 day-1)
during SAGE, which had the smallest fertilized patch size (36 km2)
of the SO experiments, was 2 times higher than the rates
(<0.11 day-1) in the SO experiments with a larger fertilized
patch size (e.g., EIFEX fertilized with a patch size of 167 km2 and
SOFeX-S fertilized with a patch size of 225 km2) (Coale et al., 2004;
Harvey et al., 2010; Law et al., 2011; Smetacek et al., 2012). Therefore, it
would be more appropriate to fertilize a large area (e.g., LOHAFEX had the
largest aOIF experiment at 300 km2), which would reduce the dilution
effect with unfertilized waters (Xiu and Chai, 2010). Based on a
∼2 nM iron concentration for a patch size of 300 km2 and
MLD of ∼60 m, it would need ∼2000 kg of
iron(II) to be applied in a fertilization experiment. Iron should be
released into the wake of a ship, with the release track describing an
expanding spiral (or square) in the eddy center, with a regular interval of
∼1 km throughout the patch, because it is easier to locate a
fertilized patch than a point release (Watson et al., 1991). In addition, it
should be completed within ∼24 h because of the
time-dependent phytoplankton response within the iron-fertilized patch.
Previous aOIF experiments have shown that multiple iron additions (≥2 infusions)
are needed to maintain the dissolved iron concentration required
to derive maximum phytoplankton growth within the fertilized patch. For
example, in SOIREE it was found that four additions of iron at intervals of
about 3 days led to persistently high levels of both dissolved and
particulate iron within the ML, with a rapid reduction at the end of the
experiment, combined with an increase in the concentration of iron-binding
ligands (Bowie et al., 2001). In both EIFEX and SOFeX-S, it was also found
that multiple iron(II) infusions (in particular, two infusions with intervals
of 13 days in EIFEX and four infusions with intervals of 4 days in SOFeX-S)
allowed iron to persist in the ML longer than its expected oxidation
kinetics. The relatively low oxidation rates were related to a combination
of photochemical production; slow oxidation; and, possibly, organic
complexation (Croot et al., 2008). Blain et al. (2007) explained that the
higher carbon sequestration effectiveness of nOIF experiments compared to
aOIF experiments partly resulted from the slow and continuous iron addition
that occurs in the natural environment. Large amounts of iron addition at
one time can lead to a substantial loss of artificially added iron.
Therefore, for an experimental duration of >40 days, a minimum of two or three
iron infusions at intervals of ∼10–15 days would be required to prevent the iron limitation on
phytoplankton growth, based on the EIFEX experiment (Smetacek et al., 2012).
What. The fifth consideration for a successful aOIF experiment is the
effective tracing of the fertilized patch, including the detection of carbon
sequestration (Buesseler and Boyd, 2003). The first step in tracing a
fertilized patch is to investigate in advance the development and fate of
natural blooms appearing as chlorophyll patches using satellite data from
pre-experiment investigations. All aOIF experiments used physical tracers to
follow the iron-fertilized patches, in particular GPS- and Argos-equipped
drifting buoys that provide the tracked positions of a fertilized patch as a
passive system moving with local currents. GPS- and Argos-equipped drifting
buoys should be released before fertilization (to provide a baseline), and
ensuing aOIF experiments should be carried out in the region described by
the drifting buoys deployed. Drifting buoys are, however, not perfect
representations of water motion and due to the effects of winds are likely
to escape a fertilized patch within a few days to a week regardless of how
deep their drogues are (Watson et al., 1991; Law et al., 1998; Stanton et al.,
1998). An inert chemical tracer, such as SF6, would also be an
excellent option for following the fertilized patch after iron addition.
Previous aOIF experiments have shown that the SF6 measurements based on
underway sampling systems can be used to accurately determine time-dependent
vertical and lateral transport of iron-fertilized patches. However, tracing
via SF6 allows for only a limited period (∼2 weeks) due
to air–sea gas exchange (Law et al., 2006; Tsumune et al., 2009; Martin et
al., 2013). Thus, many subsequent aOIF experiments have also used tracing
methods based on the observation of biogeochemical parameters (such as the
Fv/Fm ratio, chlorophyll fluorescence, and underway pCO2) before and
after iron addition (Martin et al., 1994; Coale et al., 1996, 2004; Boyd et al.,
2000, 2004; Tsuda et al., 2005; Smetacek et
al., 2012). The Fv/Fm ratio can be easily and promptly used as an indicator
to track the fertilized patch due to its rapid response to iron addition.
Direct measurements of carbon export fluxes to determine the effectiveness
of aOIF should be conducted by deploying an NBST at two depths: (1) just
below the in situ MLD to detect increases in iron-induced POC in the surface layer
along with the calibration of the water-column-based 234Th method and
(2) at the winter MLD to detect iron-induced carbon export fluxes below
winter MLD (Bidigare et al., 1999; Nodder et al., 2001; Boyd et al., 2004;
Buesseler et al., 2004, 2005; Coale et al., 2004; Aono et al., 2005; Tsuda et al., 2007; Smetacek et al., 2012; Martin et al., 2013).
Sinking-particle profiling systems (e.g., transmissometers mounted on
autonomous floats and gliders) that measure sinking particles could provide
a record of the temporal and vertical evolution of iron-induced POC stocks
through successive depth layers down to ∼3000 m for
∼20 months after deployment, once calibrated using POC fluxes
measured from sediment traps and/or the water-column-based 234Th method
(Bishop et al., 2004; Smetacek et al., 2012). Repeat casts with UVPs mounted
on the rosette could also serve a similar purpose providing a photographic
history of the water column (Martin et al., 2013). Future aOIF experiments
would benefit from these technological advances, enabling a more efficient
tracing of the carbon export flux and particle size and composition at
higher vertical and temporal resolution than has been possible in the past.
Hence, the application of an NBST system and water-column-based 234Th
method to direct flux estimates, combined with autonomous sinking-particle
profilers of a transmissometer and an UVP, will enable the quantitative and
qualitative evaluation of the effectiveness of aOIF and direct observation
of iron-induced carbon export fluxes after artificial iron additions.
What concerns. The sixth consideration for a successful aOIF experiment is
the monitoring of possible side effects. The LC/LP parties recently adopted
Resolution LC-LP.2 (2010), which includes the Assessment Framework for
Scientific Research Involving Ocean Fertilization. This considers possible
side effects on marine/ocean ecosystems after artificial iron additions,
such as the production of climate-relevant gases and negative ecosystem
changes, which are vital to assess when proposing an aOIF experiment. The
emissions of climate-relevant gases, such as N2O, DMS, and HVOCs, may
directly contribute to warming or cooling effects, and oxygen decrease and
toxic DA production may have a negative impact on marine/ocean ecosystems
(Law, 2008; Silver et al., 2010; Trick et al., 2010; Williamson et al.,
2012), resulting in significant offsets against the benefits of aOIF
experiments. However, there is little quantitative and qualitative
information regarding possible side effects following the previous aOIF
experiments. Therefore, the future monitoring of these potential side
effects is a prerequisite to evaluate accurately the effectiveness of an
aOIF experiment in the future. The possible side effects after an aOIF
experiment can be continuously monitored from optical-sensor-equipped
autonomous moored profilers and/or autonomous benthic vehicles (e.g., crawler,
which is capable of performing long-term benthic oxygen measurements for
∼12 months) (Dunne et al., 2002; Purser et al., 2013;
Wenzhöfer et al., 2016).
In summary, to maximize the effectiveness of aOIF experiments in the future,
we suggest a design that incorporates several conditions. (1) Experiments
are conducted in the center of an eddy structure when grazing pressure is
low and silicate levels are high (e.g., in the case of SO, at the south of
the PF during the early summer). (2) Shipboard observations are made during a
minimum of ∼40 days, with multiple iron injections (iron
infusions of ∼2000 kg at least two or three times, with an
interval of ∼10–15 days, to fertilize a patch of 300 km2
to obtain a ∼2 nM concentration). (3) The
iron-fertilized patch is traced using both physical (e.g., a drifting buoy)
and biogeochemical (e.g., SF6, the Fv/Fm ratio, and online pCO2
measurements) tracers. (4) The NBST system and water-column-derived 234Th
method are employed at two depths (i.e., just below the in situ MLD and at the
winter MLD), with autonomous profilers equipped with an UVP and a
transmissometer to estimate accurately the carbon export flux. (5) The side
effects on marine/ocean ecosystems, including decline in oxygen
content and
the production of climate-relevant gases (e.g., N2O, DMS, and HVOCs)
and toxic DA, are monitored using optical-sensor-equipped autonomous moored
profilers and/or autonomous benthic vehicles.
Schematic diagram of the Korean Iron Fertilization Experiment in
the Southern Ocean (KIFES) representing the experiment target site (eddy
structure) and survey methods (underway sampling systems, multiple sediment
traps, sub-bottom profilers, sediment coring systems, and satellite
observations).