Introduction
Anthropogenic activity has increased the fugacity of atmospheric carbon
dioxide (fCO2) from 280 µatm (pre-Industrial Revolution) to
over 400 µatm today (Hartmann et al., 2013). The IPCC AR5 long-term
projections for atmospheric pCO2 and associated changes to the climate
have been established for a variety of scenarios of anthropogenic activity
until the year 2300. As the largest global sink for atmospheric CO2, the
global ocean has absorbed an estimated 30 % of excess CO2 produced
(Canadell et al., 2007). With atmospheric pCO2 projected to possibly
exceed 2000 µatm by the year 2300 (Collins et al., 2013; Cubasch et
al., 2013), the ocean will take up increasing amounts of CO2, with a
potential lowering of surface ocean pH by over 0.8 units (Raven et al.,
2005). The overall effect of acidification on the biogeochemistry of surface
ocean ecosystems is unknown and currently unquantifiable, with a wide range
of potential positive and negative impacts (Doney et al., 2009; Hofmann et
al., 2010; Ross et al., 2011).
A number of volatile organic compounds are produced by marine phytoplankton
(Liss et al., 2014), including the climatically important trace gas
dimethylsulfide (DMS, C2H6S) and a number of halogen-containing
organic compounds (halocarbons), including methyl iodide (CH3I) and
bromoform (CHBr3). These trace gases are a source of sulfate particles
and halide radicals when oxidised in the atmosphere and have important roles
as ozone catalysts in the troposphere and stratosphere (O'Dowd et al., 2002;
Solomon et al., 1994) and as cloud condensation nuclei (CCNs; Charlson et
al., 1987).
DMS is found globally in surface waters originating from the algal-produced
precursor dimethylsulfoniopropionate (DMSP; C5H10O2S). Both
DMS and DMSP provide the basis for major routes of sulfur and carbon flux
through the marine microbial food web and can provide up to 100 % of the
bacterial and phytoplanktonic sulfur demand (Simó et al., 2009;
Vila-Costa et al., 2006a). DMS is also a volatile compound which readily
passes through the marine boundary layer to the troposphere, where oxidation
results in a number of sulfur-containing particles important for atmospheric
climate feedbacks (Charlson et al., 1987; Quinn and Bates, 2011); for this
reason, any change in the production of DMS may have significant implications
for climate regulation. Several previous acidification experiments have shown
differing responses of both compounds (e.g. Avgoustidi et al., 2012; Hopkins
et al., 2010; Webb et al., 2015), while others have shown delayed or more
rapid responses as a direct effect of CO2 (e.g. Archer et al., 2013;
Vogt et al., 2008). Further, some laboratory incubations of coastal microbial
communities showed increased DMS production with increased fCO2
(Hopkins and Archer, 2014) but lower DMSP production. The combined picture
arising from existing studies is that the response of communities to
fCO2 perturbation is not predictable and requires further study.
Previous studies measuring DMS in the Baltic Sea measured concentrations up
to 100 nmol L-1 during the summer bloom, making the Baltic Sea a
significant source of DMS (Orlikowska and Schulz-Bull, 2009).
In surface waters, halocarbons such as methyl iodide (CH3I),
chloroiodomethane (CH2ClI), and bromoform (CHBr3) are produced by
biological and photochemical processes: many marine microbes (for example
cyanobacteria: Hughes et al., 2011; diatoms: Manley and De La Cuesta, 1997;
and haptophytes: Scarratt and Moore, 1998) and macroalgae (e.g. brown-algal
Fucus species: Chance et al., 2009; red algae: Leedham et al.,
2013) utilise halides from seawater and emit a range of organic and inorganic
halogenated compounds. This production can lead to significant annual flux to
the marine boundary layer in the order of 10 Tg iodine-containing compounds
(“iodocarbons”; O'Dowd et al., 2002) and 1 Tg bromine-containing compounds
(“bromocarbons”; Goodwin et al., 1997) into the atmosphere. The effect of
acidification on halocarbon concentrations has received limited attention,
but two acidification experiments measured lower concentrations of several
iodocarbons, while bromocarbons were unaffected by fCO2 up to
3000 µatm (Hopkins et al., 2010; Webb, 2015), whereas an additional
mesocosm study did not elicit significant differences from any compound up to
1400 µatm fCO2 (Hopkins et al., 2013).
Measurements of the trace gases within the Baltic Sea are limited, with no
prior study of DMSP concentrations in the region. The Baltic Sea is the
largest body of brackish water in the world, and salinity ranges from 1 to
15. Furthermore, seasonal temperature variations of over 20 ∘C are
common. A permanent halocline at 50–80 m separates CO2-rich, bottom
waters from fresher, lower-CO2 surface waters, and a summer thermocline
at 20 m separates warmer surface waters from those below 4 ∘C
(Janssen et al., 1999). Upwelling of bottom waters from below the summer
thermocline is a common summer occurrence, replenishing the surface nutrients
while simultaneously lowering surface temperature and pH (Brutemark et al.,
2011). Baltic organisms are required to adapt to significant variations in
environmental conditions. The species assemblage in the Baltic Sea is
different to those studied during previous mesocosm experiments in the
Arctic, North Sea, and Korea (Brussaard et al., 2013; Engel et al., 2008; Kim
et al., 2010) and is largely unstudied in terms of its community trace gas
production during the summer bloom. Following the spring bloom
(July–August), a low dissolved inorganic nitrogen (DIN) to dissolved
inorganic phosphorous (DIP) ratio combines with high temperatures and light
intensities to encourage the growth of heterocystous cyanobacteria (Niemisto
et al., 1989; Raateoja et al., 2011), in preference to nitrate-dependent
groups.
Here we report the concentrations of DMS, DMSP, and halocarbons from the 2012
summer post-bloom season mesocosm experiment aimed to assess the impact of
elevated fCO2 on the microbial community and trace gas production in the
Baltic Sea. Our objective was to assess how changes in the microbial
community driven by changes in fCO2 impacted DMS and halocarbon
concentrations. It is anticipated that any effect of CO2 on the growth
of different groups within the phytoplankton assemblage will result in an
associated change in trace gas concentrations measured in the mesocosms as
fCO2 increases, which can potentially be used to predict future
halocarbon and sulfur emissions from the Baltic Sea region.
Methods
Mesocosm design and deployment
Nine mesocosms were deployed on the 10 June 2012 (day t-10; days are
numbered negative prior to CO2 addition and positive afterward) and
moored near Tvärminne Zoological Station (59∘51.5′ N,
23∘15.5′ E) in Tvärminne Storfjärden in the Baltic Sea.
Each mesocosm comprised a thermoplastic polyurethane (TPU) enclosure of 17 m
depth, containing approximately 54 000 L of seawater, supported by an 8 m
tall floating frame capped with a polyvinyl hood. For full technical details
of the mesocosms, see Czerny et al. (2013) and Riebesell et al. (2013). The
mesocosm bags were filled by lowering through the stratified water column
until fully submerged, with the opening at both ends covered by 3 mm mesh to
exclude organisms larger than 3 mm such as fish and large zooplankton. The
mesocosms were then left for 3 days (t-10 to t-7) with the mesh in
position to allow exchange with the external water masses and ensure the
mesocosm contents were representative of the phytoplankton community in the
Storfjärden. On t-7, the bottom of the mesocosm was sealed with a
sediment trap and the upper opening was raised to approximately 1.5 m above
the water surface. Stratification within the mesocosm bags was broken up on
t-5 by the use of compressed air for 3.5 min to homogenise the water
column and ensure an even distribution of inorganic nutrients at all depths.
Unlike in previous experiments, there was no addition of inorganic nutrients
to the mesocosms at any time during the experiment; mean inorganic nitrate,
inorganic phosphate, and ammonium concentrations measured across all
mesocosms at the start of the experiment were 37.2 (±18.8 SD), 323.9
(±19.4 SD), and 413.8 (±319.5 SD) nmol L-1 respectively.
To obtain mesocosms with different fCO2, the carbonate chemistry of
the mesocosms was altered by the addition of different volumes of
50 µm filtered, CO2-enriched Baltic Sea water (sourced from
outside the mesocosms), to each mesocosm over a 4-day period, with the
first day of addition being defined as day t0. The addition of the enriched
CO2 water was by the use of a bespoke dispersal apparatus (“Spider”)
lowered through the bags to ensure even distribution throughout the water
column (further details are in Riebesell et al., 2013). Measurements of
salinity in the mesocosms throughout the experiment determined that three of
the mesocosms were not fully sealed and had undergone unquantifiable water
exchange with the surrounding waters. These three mesocosms (M2, M4, and M9)
were excluded from the analysis. Two mesocosms were designated as controls
(M1 and M5) and received only filtered seawater via the Spider; four
mesocosms received the addition of CO2-enriched waters, with the range of
target fCO2 levels between 600 and 1650 µatm (M7, 600; M6,
950; M3, 1300; M8 1650 µatm). Mesocosms were randomly allocated a
target fCO2; a noticeable decrease in fCO2 was identified in
the three highest-fCO2 mesocosms (M6, M3, and M8) over the first half
of the experiment, which required the addition of more CO2-enriched
water on t15 to bring the fCO2 back up to maximum concentrations
(Fig. 1a; Paul et al., 2015). A summary of the fCO2 in the mesocosms
can be seen in Table 1. At the same time as this further CO2 addition on
t15, the walls of the mesocosms were cleaned using a bespoke wiper
apparatus (see Riebesell et al., 2013, for more information), followed by
weekly cleaning to remove aggregations on the film which would block incoming
light. Light measurements showed that over 95 % of the photosynthetically
active radiation (PAR) was transmitted by the clean TPU and PVC materials
with 100 % absorbance of UV light (Riebesell et al., 2013). Samples for
most parameters were collected from the mesocosms at the same time every
morning from t-3 and analysed daily or every other day.
Daily measurements of (a) fCO2, (b) mean
temperature, and (c) total chlorophyll a in the mesocosms and
surrounding Baltic Sea waters. Dashed lines represent the three phases of the
experiment, based on the Chl a data.
Summary of fCO2 and pHT (total scale) during phases
0, I, and II of the mesocosm experiment.
Whole experiment
Phase 0
Phase I
Phase II
(t-3 to t31)
(t-3 to t0)
(t1 to t16)
(t16 to t31)
Mesocosm*
Target
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
fCO2
fCO2
pHT
fCO2
pHT
fCO2
pHT
fCO2
pHT
(µatm)
(µatm)
(µatm)
(µatm)
(µatm)
M1
Control
331
7.91
231
8.00
328
7.95
399
7.86
M5
Control
334
7.91
244
7.98
329
7.94
399
7.52
M7
390
458
7.80
239
7.99
494
7.81
532
7.76
M6
840
773
7.63
236
7.99
932
7.59
855
7.59
M3
1120
950
7.56
243
7.98
1176
7.51
1027
7.52
M8
1400
1166
7.49
232
8.00
1481
7.43
1243
7.45
Baltic Sea
380
350
7.91
298
7.91
277
7.98
436
7.86
* Listed in order of increasing fCO2.
Trace gas extraction and analysis
DMS and halocarbons
A depth-integrated water sampler (IWS, HYDRO-BIOS, Kiel, Germany) was used to
sample the entire 17 m water column daily or every other day. As analysis
of chlorophyll a (Chl a) showed it to be predominantly produced in the
first 10 m of the water column, trace gas analysis was conducted only on integrated samples collected from the surface 10 m, with all corresponding
community parameter analyses with the exception of pigment analysis performed
also to this depth. Water samples for trace gas analysis were taken from the
first IWS from each mesocosm to minimise the disturbance and bubble
entrainment from taking multiple samples in the surface waters. As in Hughes
et al. (2009), samples were collected in 250 mL amber glass bottles in a
laminar flow with minimal disturbance to the water sample, using Tygon tubing
from the outlet of the IWS. Bottles were rinsed twice before being carefully
filled from the bottom with minimal stirring and allowed to overflow the
volume of the bottle approximately three times before sealing with a glass
stopper to prevent bubble formation and atmospheric contact. Samples were
stored below 10 ∘C in the dark for 2 h prior to analysis. Each
day, a single sample was taken from each mesocosm, with two additional
samples taken from one randomly selected mesocosm to evaluate the precision
of the analysis (< 4 %, no further data shown).
On return to the laboratory, 40 mL of water was injected into a purge and
cryotrap system (Chuck et al., 2005), filtered through a 25 mm Whatman glass
fibre filter (GF/F; GE Healthcare Life Sciences, Little Chalfont, England)
and purged with oxygen-free nitrogen (OFN) at 80 mL min-1 for 10 min.
Each gas sample passed through a glass wool trap to remove particles and
aerosols, before a dual Nafion counterflow drier (180 mL min-1 OFN)
removed water vapour from the gas stream. The gas sample was trapped in a
stainless steel loop held at -150 ∘C in the headspace of a liquid-nitrogen-filled dewar. The sample was injected by immersion of the sample
loop in boiling water into an Agilent 6890 gas chromatograph equipped with a
60 m DB-VRX capillary column (0.32 mm ID, 1.8 µm film thickness,
Agilent J&W Ltd) according to the programme outlined by Hopkins et
al. (2010). Analysis was performed by an Agilent 5973 quadrupole mass
spectrometer operated in electron ionisation, single-ion mode. Liquid
standards of CH3I, diiodomethane (CH2I2), CH2ClI,
iodoethane (C2H5I), iodopropane (C3H7I), CHBr3,
dibromoethane (CH2Br2), dibromochloromethane (CHBr2Cl),
bromoiodomethane (CH2BrI), and DMS (standards supplied by Sigma Aldrich
Ltd, UK) were gravimetrically prepared by dilution in high-performance liquid chromatography (HPLC) grade methanol
(Table 2) and used for calibration. The relative standard error was expressed
as a percentage of the mean for the sample analysis, calculated for each
compound using triplicate analysis each day from a single mesocosm, and was
< 7 % for all compounds. Gas chromatography–mass spectrometry instrument drift was corrected by
the use of a surrogate analyte standard in every sample, comprising
deuterated DMS (D6-DMS), deuterated methyl iodide (CD3I), and
13C dibromoethane (13C2H4Br2) via the method
described in Hughes et al. (2006) and Martino et al. (2005). Five-point
calibrations were performed weekly for each compound with the addition of the
surrogate analyte, with a single standard analysed daily to check for
instrument drift; linear regression from calibrations typically produced
r2 > 0.98. All samples measured within the mesocosms were
within the concentration ranges of the calibrations (Table 2).
Calibration ranges and calculated percentage mean relative standard
error for the trace gases measured in the mesocosms.
Compound
Calibration range
% Mean relative
(pmol L-1)
standard error
DMS
600–29 300*
6.33
DMSP
2030–405 900*
CH3I
0.11–11.2
4.62
CH2I2
5.61–561.0
4.98
C2H5I
0.10–4.91
5.61
CH2ClI
1.98–99.0
3.64
CHBr3
8.61–816.0
4.03
CH2Br2
0.21–20.9
5.30
CHBr2Cl
0.07–7.00
7.20
* Throughout the rest of this paper, these measurements are
given in nmol L-1.
DMSP
Samples for total DMSP (DMSPT) were collected and stored for later
analysis by the acidification method of Curran et al. (1998). A 7 mL
subsample was collected from the amber glass bottle into an 8 mL glass
sample vial (Labhut, Churcham, UK), into which 0.35 µL of 50 %
H2SO4 was added, before storage at ambient temperature. Particulate
DMSP (DMSPP) samples were prepared by the gravity filtration of
20 mL of sample through a 47 mm GF/F in a glass filter unit, before careful
removal and folding of the GF/F into a 7 mL sample vial filled with 7 mL of
Milli-Q water and 0.35 µL of H2SO4 before storage at
ambient temperature. Samples were stored for approximately 8 weeks prior to
analysis. DMSP samples (total and particulate) were analysed on a PTFE purge
and cryotrap system using 2 mL of the sample purged with 1 mL of 10 M NaOH
for 5 min at 80 mL min-1. The sample gas stream passed through a
glass wool trap and Nafion counterflow (Permapure) drier before being trapped
in a PTFE sample loop kept at -150 ∘C by suspension in the
headspace of a liquid nitrogen-filled dewar and controlled by feedback from a
thermocouple. Immersion in boiling water rapidly re-volatilised the sample
for injection into a Shimadzu GC2010 gas chromatograph with a Varian
Chrompack CP-Sil-5CB column (30 m, 0.53 mm ID) and flame photometric
detector (FPD). The gas chromatography (GC) oven was operated isothermally at 60 ∘C, which
resulted in DMS eluting at 2.1 min. Liquid DMSP standards were prepared and
purged in the same manner as the sample to provide weekly calibrations of the
entire analytical system. Involvement in the 2013 AQA 12-23 international DMS
analysis proficiency test (National Measurement Institute of Australia, 2013)
in February 2013 demonstrated excellent agreement between our method of DMSP
analysis and the mean from 13 laboratories measuring DMS using
different methods, with a measurement error of 5 %.
DMSP was not detected in any of the samples (total or particulate) collected
and stored during the experiment, and it was considered likely that this was
due to an unresolved issue regarding acidifying Baltic Sea samples for later
DMSP analysis. This method had been used during a previous mesocosm
experiment (SOPRAN II, Bergen, Norway), and the results correlated well with
those measured immediately on a similar GC-FPD system (Webb et al., 2015). It
was considered unlikely that rates of bacterial DMSP turnover through
demethylation rather than through cleavage to produce DMS (Curson et al.,
2011) were sufficiently high in the Baltic Sea to remove all detectable DMSP
yet still produce measurable DMS concentrations. Also, rapid turnover of dissolved DMSP in surface waters
being the cause of low DMSPT concentrations does not explain the
lack of intracellular particulate-phase DMSP. Although production of DMS is
possible from alternate sources, it is highly unlikely that there was a total
absence of DMSP-producing phytoplankton within the mesocosms or Baltic Sea
surface waters around Tvärminne; DMSP was measured in surface waters of
the southern Baltic Sea at 22.2 nmol L-1 in 2012, indicating that
DMSP-producing species are present within the Baltic Sea (C. Zindler,
personal communication, 2014).
A previous study by del Valle et al. (2011) highlighted up to 94 % loss
of DMSPT from acidified samples of colonial Phaeocystis globosa culture and field samples dominated by colonial Phaeocystis antarctica. Despite filamentous, colonial cyanobacteria in the samples from
Tvärminne mesocosms potentially undergoing the same process, these
species did not dominate the community, at only 6.6 % of the total
Chl a, implying that the acidification method for DMSP fixation also failed
for unicellular phytoplankton species. The findings of this mesocosm study
suggest that the acidification method is unreliable in the Baltic Sea and
should be considered inadequate as the sole method of DMSP fixation in future
experiments in the region. The DMSP acidification method is used worldwide as
a simple and effective method of DMSP storage; the findings here, alongside
those of del Valle et al. (2011), question the applicability of this method
in other marine environments and suggest significant testing prior to
reliance on this method as a sole means of DMSP storage.
Measurement of carbonate chemistry and community dynamics
Water samples were collected from the 10 and 17 m IWS on a daily basis and
analysed for carbonate chemistry, fluorometric Chl a, phytoplankton
pigments (17 m IWS only), and cell abundance to analyse the community
structure and dynamics during the experiment. The carbonate system was
analysed through a suite of measurements (Paul et al., 2015), including
potentiometric titration for total alkalinity (TA), infrared absorption for
dissolved inorganic carbon (DIC), and spectrophotometric determination for pH.
For Chl a analysis and pigment determination, 500 mL subsamples were
filtered through a GF/F and stored frozen (-20 ∘C for 2 h for Chl a and -80 ∘C for up to 6 months for pigments), before
homogenisation in 90 % acetone with glass beads. After centrifuging (10 min at 800 g at 4 ∘C) the
Chl a concentrations were determined using a Turner AU-10 fluorometer by
the methods of Welschmeyer (1994), and the phytoplankton pigment
concentrations were determined by reverse phase high-performance liquid chromatography
(WATERS HPLC with a Varian Microsorb-MV 100-3 C8 column) as described by
Barlow et al. (1997). Phytoplankton community composition was determined by
the use of the CHEMTAX algorithm to convert the concentrations of marker
pigments to Chl a equivalents (Mackey et al., 1996; Schulz et al., 2013).
Microbes were enumerated using a Becton Dickinson FACSCalibur flow cytometer
(FCM) equipped with a 488 nm argon laser (Crawfurd et al., 2016), and counts
of phytoplankton cells > 20 µm were made on
concentrated (50 mL) sample water, fixed with acidic Lugol's iodine solution
with an inverted microscope. Filamentous cyanobacteria were counted in
50 µm length units.
Statistical analysis
All statistical analysis was performed using Minitab V16. In analysis of the
measurements between mesocosms, one-way ANOVA was used with Tukey's post hoc
analysis test to determine the effect of different fCO2 on
concentrations measured in the mesocosms and the Baltic Sea (H0 assumes
no significant difference in the mean concentrations of trace gases measured
through the duration of the experiment). Spearman's rank correlation
coefficients were calculated to compare the relationships between trace gas
concentrations, fCO2, and a number of biological parameters, and the
resulting ρ values for each correlation are given in Supplement
Table S1 for the mesocosms and Table S2 for the Baltic Sea data.
Results and discussion
Biogeochemical changes within the mesocosms
The mesocosm experiment was split into three phases based on the temporal
variation in Chl a (Fig. 2; Paul et al., 2015) evaluated after the
experiment was completed:
Phase 0 (days t-5 to t0) – pre-CO2 addition;
Phase I (days t1 to t16) – “productive phase”;
Phase II (days t17 to t30) – temperature-induced autotrophic decline.
(a) Mean DMS concentrations measured daily in the mesocosms and
Baltic Sea from an integrated water sample of the surface 10 m. Dashed lines
show the phases of the experiment as given in Fig. 1; fCO2 shown in the
legend is mean fCO2 across the duration of the experiment. (b) Mean DMS
concentrations from each mesocosm during Phase I (crosses) and Phase II
(diamonds), for ambient (blue), medium (grey), and high fCO2 (red), with
error bars showing the range of both the DMS and fCO2.
Physical parameters
fCO2 decreased over Phase I in the three highest-fCO2
mesocosms, mainly through air–sea gas exchange and carbon fixation by
phytoplankton (Fig. 1a). All mesocosms still showed distinct differences in
fCO2 levels throughout the experiment (Table 1), and there was no
overlap of mesocosm fCO2 values on any given day, save for the two
controls (M1 and M5). The control mesocosm fCO2 increased through
Phase I of the experiment, likely as a result of undersaturation of the water
column encouraging dissolution of atmospheric CO2 (Paul et al., 2015).
Salinity in the mesocosms remained constant throughout the experiment at 5.70 ± 0.004 and showed no variation with depth (data not shown but available
in Paul et al., 2015). It remained similar to salinity in the Baltic Sea
surrounding the mesocosms, which was 5.74 ± 0.14. Water temperature
varied from a low of 8.6 ± 0.4 ∘C during Phase 0 to a high of
15.9 ± 2.2 ∘C measured on day t16, before decreasing once
again (Fig. 1b).
Summertime upwelling events are common and well described (Gidhagen, 1987;
Lehmann and Myrberg, 2008) and induce a significant temperature decrease in
surface waters; such an event appears to have commenced around t16, as
indicated by significantly decreasing temperatures inside and out of the
mesocosms (Fig. 1b) and increased salinity in the Baltic Sea from 5.5 to 6.1
over the following 15 days to the end of the experiment. Due to the enclosed
nature of the mesocosms, the upwelling affected only the temperature and not
pH, fCO2, or the microbial community. However, the temperature decrease
after t16 was likely to have had a significant effect on phytoplankton
growth (and biogenic gas production), explaining the lower Chl a in
Phase II.
Community dynamics
Mixing of the mesocosms and redistribution of the nutrients throughout the
water column after closure (prior to t-3) did not trigger a notable
increase in total Chl a in Phase 0 as was identified in previous mesocosm
experiments. During Phase I, light availability, combined with increasing
water temperatures, favoured the growth of phytoplankton in all mesocosms
(Paul et al., 2015) and was unlikely to be a direct result of the CO2
enrichment, as no difference was identified between enriched mesocosms and
controls. Mean Chl a during Phase I was 1.98 (±0.29) µg L-1
from all mesocosms, decreasing to 1.44 (±0.46) µg L-1 in
Phase II; this decrease was attributed to a temperature-induced decreased in
phytoplankton growth rates and higher grazing rates as a result of higher
zooplankton reproduction rates during Phase I (Lischka et al., 2015; Paul et
al., 2015). Mesocosm Chl a decreased until the end of the experiment on
t31.
The largest contributors to Chl a in the mesocosms during the summer of
2012 were the chlorophytes and cryptophytes, with up to 40 and 21 %
contributions to the Chl a respectively (Table 3; Paul et al., 2015).
Significant long-term differences in abundance between mesocosms developed as
a result of elevated fCO2 in only two groups: picoeukaryotes I showed
higher abundance at high fCO2 (F = 8.2, p< 0.01;
Crawfurd et al., 2016, and Supplement Fig. S2), as seen in previous mesocosm
experiments (Brussaard et al., 2013; Newbold et al., 2012), and picoeukaryotes
III showed the opposite trend (F = 19.6, p< 0.01; Crawfurd et al.,
2016). Temporal variation in phytoplankton abundance was similar between all
mesocosms (Figs. S1 and S2).
Abundance and contributions of different phytoplankton groups to the
total phytoplankton community assemblage, showing the range of measurements
from total Chl a (Paul et al., 2015), CHEMTAX analysis of derived Chl a
(Paul et al., 2015), and phytoplankton abundance (Crawfurd et al., 2016).
Data are split into the range of all the mesocosm measurements and those from
the Baltic Sea.
Mesocosm
Baltic Sea
Range
Range
% contribution
Range
Range
% contribution
Integrated 10 m
Integrated 17 m
to Chl a
Integrated 10 m
Integrated 17 m
to Chl a
Chl a
0.9–2.9
0.9–2.6
100
1.3–6.5
1.12–5.5
100
Phytoplankton taxonomy (equivalent chlorophyll µg L-1)
Cyanobacteria
0.01–0.4
8
0.0–0.1
1
Prasinophytes
0.04–0.3
7
0.01–0.3
4
Euglenophytes
0.0–1.6
15
0.0–2.6
21
Dinoflagellates
0.0–0.3
3
0.04–0.6
9
Diatoms
0.1–0.3
7
0.04–0.9
9
Chlorophytes
0.3–2.0
40
0.28–3.1
41
Cryptophytes
0.1–1.4
21
0.1–1.0
15
Small phytoplankton (< 10 µm) abundance (cells mL-1)
Cyanobacteria
55 000–380 000
65 000–470 000
30 000–180 000
30 000–250 000
Picoeukaryotes I
15 000–10 0000
17 000–111 000
5000–70 000
6100–78 000
Picoeukaryotes II
700–4000
600–4000
400–3000
460–3700
Picoeukaryotes III
1000–9000
1100–8500
1000–6000
950–7500
Nanoeukaryotes I
400–1400
270–1500
200–4000
210–4100
Nanoeukaryotes II
0–400
4–400
100–1100
60–1300
Diazotrophic, filamentous cyanobacterial blooms in the Baltic Sea are an
annual event in summer (Finni et al., 2001), and single-celled cyanobacteria
have been found to comprise as much as 80 % of the cyanobacterial biomass
and 50 % of the total primary production during the summer in the Baltic
Sea (Stal et al., 2003). However, CHEMTAX analysis identified cyanobacteria
as contributing less than 10 % of the total Chl a in the mesocosms
(Crawfurd et al., 2016; Paul et al., 2015). These observations were backed up
by satellite observations showing reduced cyanobacterial abundance throughout
the Baltic Sea in 2012 compared to previous and later years (Oberg, 2013). It
was proposed that light availability and surface water temperatures during
the summer of 2012 were suboptimal for triggering a filamentous
cyanobacteria bloom (Wasmund, 1997).
DMS and DMSP
Mesocosm DMS
A significant 34 % reduction in DMS concentrations was detected in the
high-fCO2 treatments during Phase II compared to the ambient-fCO2 mesocosms (F = 31.7, p< 0.01). Mean DMS
concentrations of 5.0 (±0.8; range 3.5–6.8) nmol L-1 in the
ambient treatments were compared to 3.3 (±0.3; range 2.9–3.9) nmol L-1
in the 1333 and 1075 µatm mesocosms (Fig. 2a). The primary
differences identified were apparent from the start of Phase II on t17,
after which maximum concentrations were observed in the ambient mesocosms on
t21. The relationship between DMS and increasing fCO2 during
Phase II was found to be linear (Fig. 2b), a finding also identified in
previous mesocosm experiments (Archer et al., 2013; Webb et al., 2015).
Furthermore, increases in DMS concentrations under high fCO2 were
delayed by 3 days relative to the ambient- and medium-fCO2
treatments, a situation which has been observed in a previous mesocosm
experiment. This was attributed to small-scale shifts in community
composition and succession which could not be identified with only a
once-daily measurement regime (Vogt et al., 2008). DMS measured in all
mesocosms fell within the range 2.7 to 6.8 nmol L-1 across the course
of the experiment. During Phase I, no difference was identified in DMS
concentrations between fCO2 treatments, with the mean of all mesocosms
being 3.1 (±0.2) nmol L-1. Concentrations in all mesocosms gradually
declined from t21 until the end of DMS measurements on t31. DMS
concentrations measured in the mesocosms and Baltic Sea were comparable to
those measured in temperate coastal conditions in the North Sea (Turner et
al., 1988), the Mauritanian upwelling (Franklin et al., 2009; Zindler et al.,
2012), and the South Pacific (Lee et al., 2010).
The majority of DMS production is presumed to be from DMSP. However, an
alternative production route for DMS is available through the methylation of
methanethiol (Drotar et al., 1987; Kiene and Hines, 1995; Stets et al.,
2004), predominantly identified in anaerobic environments such as freshwater lake
sediments (Lomans et al., 1997), salt marsh sediments (Kiene and Visscher,
1987), and microbial mats (Visscher et al., 2003; Zinder et al., 1977). Recent
studies have also identified this pathway of DMS production from
Pseudomonas deceptionensis in an aerobic environment (Carrión et
al., 2015), where P. deceptionensis was unable to synthesise or
catabolise DMSP but was able to enzymatically mediate DMS production from
methanethiol (MeSH). The same enzyme has also been identified in a wide range
of other bacterial taxa, including the cyanobacterial Pseudanabaena,
which was identified in the Baltic Sea during this and previous
investigations (A. Stuhr, personal communication, 2015; Kangro et al., 2007; Nausch et al.,
2009). Correlations between DMS and the cyanobacterial equivalent Chl a
(ρ= 0.42, p< 0.01; Fig. S1g) and DMS and single-celled
cyanobacteria (ρ= 0.58, p< 0.01; Fig. S2a) suggest that the
methylation pathway may be a potential source of DMS within the Baltic Sea
community. In addition to the methylation pathway, DMS production has been
identified from S-methylmethionine (Bentley and Chasteen, 2004), as well as
from the reduction of dimethylsulfoxide (DMSO), in both surface and deep
waters by bacterial metabolism (Hatton et al., 2004). As these compounds were
not measured in the mesocosms, it is impossible to determine whether they were
significant sources of DMS.
Concentration ranges of trace gases measured in the mesocosms
compared to other open-water ocean acidification experiments, showing the
range of concentrations for each gas and the percentage change between the
control and the highest-fCO2 treatment. SOPRAN: Surface
Ocean Processes in the Anthropocene; NERC: Natural Environment Research Council; EPOCA: European Project on OCean Acidification; UKOA: UK Ocean Acidification Research Programme.
Range fCO2
DMS
CH3I
CH2I2
CH2ClI
CHBr3
CH2Br2
CH2Br2Cl
(µatm)
(nmol L-1)
(pmol L-1)
SOPRAN Tvärminne
346–1333
Range
2.7–6.8
2.9–6.4
57–202
3.8–8.0
69–148
4.0–7.7
1.7–3.1
mesocosm (this study)
% change
-34
-0.3
1.3
-11
-9
-3
-4
SOPRAN Bergen 2011
280–3000
Range
0.1–4.9
4.9–32
5.8–321
9.0–123
64–306
6.3–30.8
3.9–14
Webb et al. (2015)
% change
-60
-37
-48
-27
-2
-4
-6
NERC microbial
300–750
Range
ND–50
2.0–25
ND–750
ND–700
5.0–80
ND–5.5
0.2–1.2
metagenomics experiment,
% change
-57
-41
-33
-28
13
8
22
Bergen (2006),
Hopkins et al. (2010)
EPOCA Svalbard 2010
180–1420
Range
ND–14
0.04–10
0.01–2.5
0.3–1.6
35–151
6.3–33.3
1.6–4.7
Archer et al. (2013),
% change
-60
NS
NS
NS
NS
NS
Hopkins et al. (2013)
UKOA European shelf 2011
340–1000
Range
0.5–12
Hopkins and Archer (2014)
% change
225
Korean mesocosm experiment
160–830
Range
1.0–100
2012 Park et al. (2014)
% change
-82
ND – not detected.NC – no change.
DMS and community interactions
Throughout Phase I, DMS showed no correlation with any measured variables of
biological activity or cell abundance and was unaffected by elevated
fCO2, indicating that measured DMS concentrations were not directly related
to the perturbation of the system and associated cellular stress (Sunda et
al., 2002). Of the studied phytoplankton groupings, neither the cryptophytes
nor chlorophytes as the largest contributors of Chl a were identified as
significant producers of DMSP. During Phase II, DMS was negatively correlated
with Chl a in the ambient- and medium-fCO2 mesocosms (ρ= -0.60, p< 0.01). During Phase II, a significant correlation
was seen between DMS and single-celled cyanobacteria identified predominantly
as Synechococcus (ρ= 0.53, p< 0.01; Crawfurd et
al., 2016, and Table S1) and picoeukaryotes III (ρ= 0.75,
p< 0.01). The peak in DMS concentrations on t21 is unlikely to
be a delayed response to the increased Chl a on t16 due to the time lag
of 7 days. These higher DMS concentrations were likely connected to a peak in
dissolved organic carbon (DOC) on t15, as well as increasing bacterial
abundance during Phase II (Hornick et al., 2016). It is also likely that DMS
concentrations increased as a response to the mesocosm wall cleaning which
took place on t16. The variation in inorganic nutrient concentrations
between mesocosms at the start of the experiment did not have an effect on
DMS concentrations during Phase I, and by the start of Phase II the variation
between mesocosms had decreased.
In previous mesocosm experiments (Archer et al., 2013; Hopkins et al., 2010;
Webb et al., 2015), DMS has shown poor correlations with many of the
indicators of primary production and phytoplankton abundance, as well as
showing the same trend of decreased concentrations in high-fCO2
mesocosms compared to ambient ones. DMS production is often uncoupled from
measurements of primary production in open waters (Lana et al., 2012) and
also often from the production of its precursor DMSP (Archer et al., 2009). DMS
and DMSP are important sources of sulfur and carbon in the microbial food
web for both bacteria and algae (Simó et al., 2002, 2009), and since
microbial turnover of DMSP and DMS play a significant role in net DMS
production, it is unsurprising that DMS concentrations have shown poor
correlation with DMSP-producing phytoplankton groups in past experiments and
open waters.
DMS concentrations have been reported to be lower under conditions of elevated
fCO2 compared to ambient controls, in both mesocosm experiments
(Table 4) and phytoplankton monocultures (Arnold et al., 2013; Avgoustidi et
al., 2012). However, the varying response of the community within each
experiment limits our ability to generalise the response of algal production
of DMS and DMSP in all situations due to the characteristic community
dynamics of each experiment in specific geographical areas and temporal
periods. Previous experiments in the temperate Raunefjorden of Bergen, Norway,
showed lower abundance of DMSP-producing algal species, and subsequently
of DMSP-dependent DMS concentrations (Avgoustidi et al., 2012; Hopkins et al.,
2010; Vogt et al., 2008; Webb et al., 2015). In contrast mesocosm experiments
in the Arctic and Korea have shown increased abundance of DMSP producers
(Archer et al., 2013; Kim et al., 2010) but lower DMS concentrations, while
incubation experiments by Hopkins and Archer (2014) showed lower DMSP
production but higher DMS concentrations at high fCO2. However, in all
previous experiments with DMSP as the primary precursor of DMS, elevated
fCO2 had a less marked effect on measured DMSP concentrations than on
measured DMS concentrations. Hopkins et al. (2010) suggested that “the
perturbation of the system has a greater effect on the processes that control
the conversion of DMSP to DMS rather than the initial production of DMSP
itself”.
Mean concentrations (pmol L-1) of (a) CH3I,
(b) C2H5I, (c) CH2I2, and (d)
CH2ClI taken from a water sample integrated from the surface 10 m.
Dashed lines indicate the phases of the experiment, as given in Fig. 2.
fCO2 shown in the legend is mean fCO2 across the duration of
the experiment.
Previous mesocosm experiments have suggested significant links between
increased bacterial production through greater availability of organic
substrates at high fCO2 (Engel et al., 2013; Piontek et al., 2013).
Further, Endres et al. (2014) identified significant enhanced enzymatic
hydrolysis of organic matter with increasing fCO2, with higher
bacterial abundance. Higher bacterial abundance will likely result in greater
bacterial demand for sulfur and therefore greater consumption of DMS and
conversion to DMSO. This was suggested as a significant sink for DMS in a
previous experiment (Webb et al., 2015), but during the present experiment,
both bacterial abundance and bacterial production were lower at high
fCO2 (Hornick et al., 2016). However, as it has been proposed that
only specialist bacterial groups are DMS consumers (Vila-Costa et al.,
2006b) and there is no determination of the DMS consumption characteristics
of the bacterial community in the Baltic Sea, it is not known if this loss
pathway is stimulated at high fCO2. As microbial DMS yields can vary
between 5 and 40 % depending on the sulfur and carbon demand (Kiene and
Linn, 2000), a change in the bacterial sulfur requirements could change DMS
turnover despite lower abundance.
Iodocarbons in the mesocosms and relationships with community
composition
Elevated fCO2 did not affect the concentration of iodocarbons in the
mesocosms significantly at any time during the experiment, which is in
agreement with the findings of Hopkins et al. (2013) in the Arctic but in
contrast to Hopkins et al. (2010) and Webb (2015), where iodocarbons were
measured to be significantly lower under elevated fCO2 (Table 4).
Concentrations of all iodocarbons measured in the mesocosms and the Baltic
Sea fall within the range of those measured previously in the region
(Table 5). Mesocosm concentrations of CH3I (Fig. 3a) and C2H5I
(Fig. 3b) showed concentration ranges of 2.91 to 6.25 and 0.23 to
0.76 pmol L-1 respectively. CH3I showed a slight increase in all
mesocosms during Phase I, peaking on t16, which corresponded to higher
Chl a concentrations and correlated throughout the entire experiment with
picoeukaryote groups II (ρ= 0.59, p< 0.01) and III (ρ= 0.23, p< 0.01; Crawfurd et al., 2016) and nanoeukaryotes I
(ρ= 0.37, p< 0.01). Significant differences identified
between mesocosms for CH3I were unrelated to elevated fCO2
(F = 3.1, p< 0.05), but concentrations were on average
15 % higher in Phase II than Phase I. C2H5I decreased slightly
during Phases I and II, although concentrations of this halocarbon were close
to its detection limit (0.2 pmol L-1), remaining below
1 pmol L-1 at all times. As this compound showed no significant effect
of elevated fCO2 and was identified by Orlikowska and
Schulz-Bull (2009) as having extremely low concentrations in the Baltic Sea
(Table 5), it will not be discussed further.
Concentration ranges of trace gases measured in the Baltic Sea
compared to concentrations measured in the literature.
Halocarbon concentration range (pmol L-1)
CH3I
CH2I2
C2H5I
CH3ClI
CHBr3
CH2Br2
CH2Br2Cl
Study
DMS concentration
range (nmol L-1)
SOPRAN Tvärminne Baltic Sea (this study)
1.9–11
4.3–8.6
66.9–374
0.6–1.0
7.0–18
93–192
7.1-10
3.3–5.0
Orlikowska and Schulz-Bull (2009)
0.3–120
1–16
0–85
0.4–1.2
5–50
5.0–40
2.0–10
0.8–2.5
Karlsson et al. (2008)
3.0–7.5
35–60
4.0–7.0
2.0–6.5
Klick and Abrahamsson (1992)
15–709
11–74
14–585
Klick (1992)
ND–243
ND–57
40–790
ND–86
ND–29
Leck and Rodhe (1991)
0.4–2.8
Leck et al. (1990)
ND–3.2
ND – not detected.
No correlation was found between CH3I and Chl a at any phase, and the
only correlation of any phytoplankton grouping was with nanoeukaryotes II
(ρ= 0.88, p< 0.01; Crawfurd et al., 2016). These CH3I
concentrations compare well to the 7.5 pmol L-1 measured by Karlsson
et al. (2008) during a cyanobacterial bloom in the Baltic Sea (Table 5) and
the summer maximum of 16 pmol L-1 identified by Orlikowska and
Schulz-Bull (2009).
Karlsson et al. (2008) showed Baltic Sea halocarbon production occurring
predominately during daylight hours, with concentrations at night decreasing
by 70 % compared to late afternoon. Light-dependent production of
CH3I has been shown to take place through abiotic processes, including
radical recombination of CH3 and I (Moore and Zafiriou, 1994). However,
since samples were integrated over the surface 10 m of the water column, it
was impossible to determine whether photochemistry was affecting iodocarbon
concentrations near the surface where some UV light was able to pass between
the top of the mesocosm film material and the cover. For the same reason,
photodegradation of halocarbons (Zika et al., 1984) within the mesocosms was
also likely to have been significantly restricted. Thus, as photochemical
production was expected to be minimal, biogenic production was likely to have
been the dominant source of these compounds. Karlsson et al. (2008)
identified Pseudanabaena as a key producer of CH3I in the
Baltic Sea. However, the abundance of Pseudanabaena was highest
during Phase I of the experiment (A. Stuhr, personal
communication, 2015) when CH3I concentrations
were lower, and as discussed previously, the abundance of these species
constituted only a very small proportion of the community. Previous
investigations in the laboratory have identified diatoms as significant
producers of CH3I (Hughes et al., 2013; Manley and De La Cuesta, 1997),
and the low, steady-state abundance of the diatom populations in the
mesocosms could have produced the same relatively steady-state trends in the
iodocarbon concentrations.
Measured in the range 57.2–202.2 pmol L-1 in the mesocosms,
CH2I2 (Fig. 3c) showed the clearest increase in concentration
during Phase II, when it peaked on t21 in all mesocosms, with a maximum of
202.2 pmol L-1 in M5 (348 µatm). During Phase II,
concentrations of CH2I2 were 57 % higher than Phase I and were
therefore negatively correlated with Chl a. The peak on t21 corresponds
to the peak identified in DMS on t21, and concentrations through all
three phases correlate with picoeukaryotes II (ρ= 0.62,
p< 0.01) and III (ρ= 0.47, p< 0.01) and
nanoeukaryotes I (ρ= 0.88, p< 0.01; Crawfurd et al., 2015).
CH2ClI (Fig. 3d) showed no peaks during either Phase I or Phase II,
remaining within the range of 3.81 to 8.03 pmol L-1 and again correlated
with picoeukaryotes groups II (ρ= 0.34, p< 0.01) and III
(ρ= 0.38, p< 0.01). These results may suggest that these
groups possessed halo-peroxidase enzymes able to oxidise I-, most likely
as an antioxidant mechanism within the cell to remove H2O2 (Butler
and Carter-Franklin, 2004; Pedersen et al., 1996; Theiler et al., 1978).
However, given the lack of response of these compounds to elevated
fCO2 (F = 1.7, p< 0.01), it is unlikely that
production was increased in relation to elevated fCO2. Production of
all iodocarbons increased during Phase II when total Chl a decreased,
particularly after the walls of the mesocosms were cleaned for the first
time, releasing significant volumes of organic aggregates into the water
column. Aggregates have been suggested as a source of CH3I and
C2H5I (Hughes et al., 2008), likely through the alkylation of
inorganic iodide (Urhahn and Ballschmiter, 1998) or through the breakdown of
organic matter by microbial activity to supply the precursors required for
iodocarbon production (Smith et al., 1992). Hughes et al. (2008) did not
identify this route as a pathway for CH2I2 or CH2ClI
production, but Carpenter et al. (2005) suggested a production pathway for
these compounds through the reaction of HOI with aggregated organic
materials.
Bromocarbons in the mesocosms and the relationships with community
composition
No effect of elevated fCO2 was identified for any of the three
bromocarbons, which compared well with the findings from previous mesocosms where
bromocarbons were studied (Hopkins et al., 2010, 2013; Webb, 2015; Table 4).
Measured concentrations were comparable to those of Orlikowska and
Schulz-Bull (2009) and Karlsson et al. (2008) measured in the southern part
of the Baltic Sea (Table 3). The concentrations of CHBr3,
CH2Br2, and CHBr2Cl showed no major peaks of production in the
mesocosms. CHBr3 (Fig. 4a) decreased rapidly in all mesocosms over Phase
0 from a maximum measured concentration of 147.5 pmol L-1 in M1 (mean
of 138.3 pmol L-1 in all mesocosms) to a mean of 85.7
(±8.2 SD) pmol L-1 in all mesocosms for the period t0 to t31
(Phases I and II). The steady-state CHBr3 concentrations indicated a
production source; however, there was no clear correlation with any measured
algal groups. CH2Br2 concentrations (Fig. 4b) decreased steadily in
all mesocosms from t-3 through to t31, over the range 4.0 to
7.7 pmol L-1, and CHBr2Cl followed a similar trend in the range
1.7 to 4.7 pmol L-1 (Fig. 4c). Of the three bromocarbons, only
CH2Br2 showed correlation with total Chl a (ρ= 0.52,
p< 0.01) and with cryptophyte (ρ= 0.86,
p< 0.01) and dinoflagellate (ρ= 0.65, p< 0.01)-derived Chl a. Concentrations of CH2BrI were below detection limit for
the entire experiment.
Mean concentrations (pmol L-1) of (a) CHBr3, (b) CH2Br2, and (c) CHBr2Cl taken from a water sample integrated
from the surface 10 m. Dashed lines indicate the phases of the experiment as
defined in Fig. 2; fCO2 shown in the legend is mean fCO2 across the
duration of the experiment.
CH2Br2 showed positive correlation with Chl a (ρ= 0.52,
p< 0.01), nanoeukaryotes II (ρ= 0.34, p< 0.01), and cryptophytes (ρ= 0.86, p< 0.01; see Supplement), whereas CHBr3 and CHBr2Cl showed very weak or no
correlation with any indicators of algal biomass. Schall et al. (1997) have
proposed that CHBr2Cl is produced in seawater by the nucleophilic
substitution of bromide by chloride in CHBr3, which given the
steady-state concentrations of CHBr3 would explain the similar
distribution of CHBr2Cl concentrations. Production of all three
bromocarbons was identified from large-size cyanobacteria such as
Aphanizomenon flos-aquae by Karlsson et al. (2008), and in addition,
significant correlations were found in the Arabian Sea between the abundance
of the cyanobacterium Trichodesmium and several bromocarbons (Roy et
al., 2011), and the low abundance of such bacteria in the mesocosms would
explain the low variation in bromocarbon concentrations through the
experiment.
Halocarbon loss processes such as nucleophilic substitution (Moore, 2006),
hydrolysis (Elliott and Rowland, 1995), sea–air exchange, and microbial
degradation are suggested as of greater importance than the production of these
compounds by specific algal groups, particularly given the relatively low
growth rates and low net increase in total Chl a. Hughes et al. (2013)
identified bacterial inhibition of CHBr3 production in laboratory
cultures of Thalassiosira diatoms but that it was not subject to
bacterial breakdown, which could explain the relative steady state of
CHBr3 concentrations in the mesocosms. In contrast, significant
bacterial degradation of CH2Br2 in the same experiments could
explain the steady decrease in CH2Br2 concentrations seen in the
mesocosms. Bacterial oxidation was also identified by Goodwin et al. (1998)
as a significant sink for CH2Br2. As discussed for the iodocarbons,
photolysis was unlikely due to the UV absorption of the mesocosm film and
limited UV exposure of the surface waters within the mesocosm due to the
mesocosm cover. The ratio of CH2Br2 to CHBr3 was also
unaffected by increased fCO2, staying within the range 0.04 to 0.08.
This range in ratios is consistent with that calculated by Hughes et
al. (2009) in the surface waters of an Antarctic depth profile and
attributed to higher sea–air flux of CHBr3 than CH2Br2 due to
a greater concentrations gradient, despite the similar transfer velocities of
the two compounds (Quack et al., 2007). Using cluster analysis in a
time series in the Baltic Sea, Orlikowska and Schulz-Bull (2009) identified
both these compounds as originating from different sources and different
pathways of production.
Macroalgal production would not have influenced the mesocosm concentrations
after the bags were sealed due to the isolation from the coastal environment.
However, macroalgal production into the water column prior to mesocosm
installation (Klick, 1992; Leedham et al., 2013; Moore and Tokarczyk, 1993)
could account for the high initial concentrations with concentrations
decreasing through the duration of the experiment via turnover and transfer
to the atmosphere.
Natural variations in Baltic Sea fCO2 and the effect on biogenic trace
gases
Physical variation and community dynamics
Baltic Sea deep waters have high fCO2 and subsequently lower pH
(Schneider et al., 2002), and the influx to the surface waters surrounding
the mesocosms resulted in fCO2 increasing to 725 µatm on
t31, close to the average fCO2 of the third-highest mesocosm (M6:
868 µatm). The input of upwelled water into the region midway
through the experiment significantly altered the biogeochemical properties of
the waters surrounding the mesocosms, and as a result it is inappropriate to
directly compare the community structure and trace gas production of the
Baltic Sea and the mesocosms. These conditions imply that pelagic communities
in the Baltic Sea are regularly exposed to rapid changes in fCO2 and
the associated pH, as well as having communities associated with the elevated
fCO2 conditions. The changes in biological parameters and trace gas
concentrations are therefore discussed here separately from the
concentrations measured in the mesocosms.
Given the separation of the waters within the mesocosms and the movement of
water masses within the Baltic Sea, it is expected that phytoplankton
population structure could be significantly different inside the mesocosms
compared to the external waters. Chl a followed the pattern of the
mesocosms until t4, after which concentrations were significantly higher
than any mesocosm, peaking at 6.48 µg L-1 on t16,
corresponding to the maximum Chl a peak in the mesocosms and the maximum
peak of temperature. As upwelled water intruded into the surface waters, the
surface Chl a was diluted with low-Chl a deep water: Chl a in the
surface 10 m decreased from around t16 at the start of the upwelling until
t31 when concentrations were once again equivalent to those found in the
mesocosms at 1.30 µg L-1. In addition, there was the potential
introduction of different algal groups to the surface, but chlorophytes and
cryptophytes were the major contributors to the Chl a in the Baltic Sea, as
in the mesocosms. Cyanobacteria contributed less than 2 % of the total
Chl a in the Baltic Sea (Crawfurd et al., 2016; Paul et al., 2015).
Temporal community dynamics in the Baltic Sea were very different to that in
the mesocosms across the experiment, with euglenophytes, chlorophytes,
diatoms, and prasinophytes all showing distinct peaks at the start of Phase
II, with these same peaks identified in the nanoeukaryotes I and II and
picoeukaryotes II (Crawfurd et al., 2016; Paul et al., 2015; Supplement
Figs. S1 and S2). The decrease in the abundance of many groups during Phase II
was attributed to the decrease in temperature and dilution with low-abundance
deep waters.
DMS in the Baltic Sea
The Baltic Sea samples gave a mean DMS concentration of
4.6 ± 2.6 nmol L-1 but peaked at 11.2 nmol L-1 on t16 and were within the range of previous measurements for the region (Table 5).
Strong correlations were seen between DMS and Chl a (ρ= 0.84,
p< 0.01), with the ratio of DMS : Chl a at 1.6
(±0.3) nmol µg-1. Other strong correlations were seen
with euglenophytes (ρ= 0.89, p< 0.01), dinoflagellates
(ρ= 0.61, p< 0.05), and nanoeukaryotes II (ρ= 0.88,
p< 0.01), but no correlation was found between DMS and
cyanobacterial abundance or with picoeukaryotes III, which were identified in
the mesocosms, suggesting that DMS had a different origin in the Baltic Sea
community than in the mesocosms. In addition, the community demands of
sulfur are likely to be very different in the Baltic Sea compared to the
mesocosms, due to differences in community composition and sulfur
availability, and therefore direct comparisons with mesocosm concentrations
are inappropriate.
As CO2 levels increased after t16, the DMS concentration measured in
the Baltic Sea decreased, from the peak on t16 to the lowest recorded
sample of the entire experiment at 1.85 nmol L-1 on t31. As with
Chl a, DMS concentrations in the surface of the Baltic Sea may have been
diluted with low-DMS deep water
Halocarbon concentrations in the Baltic Sea
Outside the mesocosms in the Baltic Sea, CH3I was measured at a maximum
concentration of 8.65 pmol L-1, during Phase II, and showed a limited
effect of the upwelling event. Both CH2I2 and CH2ClI showed
higher concentrations in the Baltic Sea samples than the mesocosms
(CH2I2: 373.9 pmol L-1; CH2ClI:
18.1 pmol L-1) and were correlated with the euglenophytes
(CH2I2: ρ= 0.63, p< 0.05; CH2ClI: ρ= 0.68, p< 0.01) and nanoeukaryotes II (CH2I2: ρ= 0.53, p< 0.01; CH2ClI: ρ= 0.58,
p< 0.01), but there was no correlation with Chl a. Both polyhalogenated
compounds showed correlation with picoeukaryote groups II and III, indicating
that production was probably not limited to a single source. These
concentrations of CH2I2 and CH2ClI compared well to those
measured over a macroalgal bed in the higher-saline waters of the Kattegat by
Klick and Abrahamsson (1992), suggesting that macroalgae were a significant
iodocarbon source in the Baltic Sea. Macroalgal production in the Baltic Sea
is likely the predominant iodocarbon source, compared to the mesocosms where
macroalgae are excluded.
As with the iodocarbons, the Baltic Sea showed significantly higher
concentrations of CHBr3 (F = 28.1, p< 0.01),
CH2Br2 (F = 208.8, p< 0.01), and CHBr2Cl
(F = 23.5, p< 0.01) than the mesocosms, with maximum
concentrations of 191.6, 10.0, and 5.0 pmol L-1 respectively. In the
Baltic Sea, only CHBr3 was correlated with Chl a (ρ= 0.65,
p< 0.05), cyanobacteria (ρ= 0.61, p< 0.01; Paul
et al., 2015), and nanoeukaryotes II (ρ= 0.56, p< 0.01;
Crawfurd et al., 2016), with the other two bromocarbons showing little to no
correlations with any parameter of community activity. Production of
bromocarbons from macroalgal sources (Laturnus et al., 2000; Leedham et al.,
2013; Manley et al., 1992) was likely a significant contributor to the
concentrations detected in the Baltic Sea; over the macroalgal beds in the
Kattegat, Klick (1992) measured concentrations an order of magnitude higher
than seen in this experiment for CH2Br2 and CHBr2Cl. There was
only a slight increase in bromocarbon concentrations as a result of the
upwelling, indicating that the upwelled water had similar concentrations to
the surface waters. These data from the Baltic Sea are presented as an
important time series of halocarbon measurements during the summer of 2012 and are expected to add to existing Baltic Sea trace gas datasets.
The Baltic Sea as a natural analogue to future ocean
acidification?
Mesocosm experiments are a highly valuable tool in assessing the potential
impacts of elevated CO2 on complex marine communities; however, they are
limited in that the rapid change in fCO2 experienced by the community
may not be representative of changes in the future ocean (Passow and
Riebesell, 2005). This inherent problem with mesocosm experiments can be
overcome through using naturally low-pH–high-CO2 areas such as upwelling
regions or vent sites (Hall-Spencer et al., 2008), which can give an insight
into populations already living and acclimated to high-CO2 regimes by
exposure over timescales measured in years. This mesocosm experiment was
performed at such a location with a relatively high-fCO2 excursion,
which was, however, still low compared to some sites (800 µatm compared to
> 2000 µatm; Hall-Spencer et al., 2008), and it was
clear through the minimal variation in Chl a between all mesocosms that the
community was relatively unaffected by elevated fCO2, although
variation could be identified in some phytoplankton groups and some shifts in
community composition. The upwelling event occurring midway through our
experiment allowed the comparison of the mesocosm findings with a natural
analogue of the system, as well as showing the extent to which the system
perturbation can occur (up to 800 µatm). This event was a
fortuitous occurrence during this mesocosm experiment, but as the scale and
timing of these upwelling events is difficult to determine, these upwelling events are extremely challenging to study as natural
high-CO2 analogues.
In this paper, we described the temporal changes in concentrations of DMS and
halocarbons in natural Baltic phytoplankton communities exposed to elevated-fCO2 treatments. In contrast to the halocarbons, concentrations of DMS
were significantly lower in the highest-fCO2 treatments compared to
the control. Despite very different physicochemical and biological
characteristics of the Baltic Sea (e.g. salinity, community composition, and
nutrient concentrations), this is a very similar outcome to that seen in
several other high-fCO2 experiments. The Baltic Sea trace gas samples
give a good record of trace gas cycling during the injection of high-fCO2 deep water into the surface community during upwelling events.
For the concentrations of halocarbons, the measured concentrations did not
change during the upwelling event in the Baltic Sea, which may indicate that
emissions of organic iodine and bromine are unlikely to change with future
acidification of the Baltic Sea without significant alteration to the
meteorological conditions. Further studies of these compounds are important
to determine rates of production and consumption to include them in prognostic and
predictive models. However, net production of organic sulfur within the
Baltic Sea region is likely to decrease with an acidified future ocean
scenario, despite the possible acclimation of the microbial community to
elevated fCO2. This will potentially impact the flux of DMS to the
atmosphere over northern Europe and could have significant impacts on the
local climate through the reduction of atmospheric sulfur aerosols. Data
from a previous mesocosm experiment has been used to estimate future global
changes in DMS production and predicted that global warming would be
amplified (Six et al., 2013); utilising the data from this experiment
combined with those of other mesocosm, field, and laboratory experiments and
associated modelling provides the basis for a better understanding of the
future changes in global DMS production and their climatic impacts.