Introduction
Besides water itself, light absorption in the marine environment is due to
three main biogeochemical constituents: (1) chromophoric dissolved organic
matter (CDOM), also known as gelbstoff, gilvin and yellow substances, and
chromophoric particulate matter, subdivided into (2) phytoplankton
(photoautotrophic microorganisms), composed of both prokaryotic
(cyanobacteria) and eukaryotic species (diatoms, dinoflagellates,
coccolithophores, …), and (3) non algal particles (NAP),
comprising organic and minerogenic detritus, and heterotrophic organisms.
Absorption spectra of CDOM, phytoplankton and NAP have been extensively
studied over the last two decades in various oceanic provinces including
coastal waters and open ocean (Blough and Del Vecchio, 2002; Babin et al.,
2003; Bricaud et al., 2010; Matsuoka et al., 2014). Indeed, in addition to
their key role in the oceanic carbon cycle, these three constituents
strongly influence the underwater light field and the apparent optical
properties of seawater. The knowledge of their absorption spectra is thus
essential for bio-optical modelling and remote sensing applications but can
also be used to investigate biological processes in the ocean.
Absorption coefficients of CDOM [ag(λ)] and NAP
[anap(λ)] typically decrease monotonically (exponentially) from
ultraviolet (UV, 280–400 nm) to visible (400–700 nm) wavelengths (Nelson et
al., 1998; Swan et al., 2009; Tilstone et al., 2012). Even though CDOM
absorption spectra are usually featureless, some “shoulders” have been
observed sporadically in the UV and visible spectral domains and attributed
to the presence of dissolved absorbing pigments released by phytoplankton
cells: mycosporine-like amino acids (MAAs) at 310–320 or at 330–360 nm,
and phaeopigments or non-chlorin metal-free porphyrins at 410–420 nm
(Whitehead and Vernet, 2000; Röttgers and Koch, 2012; Organelli et al.,
2014; Pavlov et al., 2014). In contrast, absorption coefficients of
phytoplankton [aϕ(λ)] determined from natural samples
commonly display two main peaks in the visible range, around 435–450 and 675 nm,
attributable to its content in total chlorophyll a (TChl a= mono Chl a+
divinyl Chl a) (Lutz et al., 1996; Dupouy et al., 1997; Bricaud et al.,
2004), but may also reveal other peaks or shoulders resulting from the
presence of other pigments: MAAs at 325 nm (Bricaud et al., 2010), TChl b,
TChl c and photoprotective carotenoids at 460–470 nm, photosynthetic
carotenoids and photoprotective keto-carotenoids at 490 nm (Carreto, 1985;
Stuart et al., 1998; Wozniak et al., 1999; Lohrenz et al., 2003) as well as
phycoerythrin at 550 nm (Morel, 1997). Hence, while chromophoric detrital
matter (CDM = CDOM + NAP) is the major contributor to total absorption
in the UV domain (∼ 60–95 %), in the blue region (440–490 nm),
the contributions of CDM and phytoplankton tend to be equivalent
(∼ 40–50 %), while CDOM alone is accounting for
∼ 80–95 % of CDM in the UV and blue ranges (Siegel et al.,
2002, 2005; Tedetti et al., 2010; Nelson and Siegel, 2013).
In “Case 1 waters” (Morel and Prieur, 1977), which are generally – but not
necessarily – open ocean clear waters, optical properties are controlled by
phytoplankton and all its derived material, and TChl a concentration may be
utilized as an index of optical properties thanks to its covariation with
aϕ(λ), ag(λ), anap(λ) and
particulate backscattering coefficient [bbp(λ)] (Antoine et
al., 2014). Due to the covariation with aϕ(λ) in Case 1
waters, CDOM is considered as being a by-product of phytoplanktonic
production. Nonetheless, recent studies have highlighted some degree of
de-phasing between the dynamics in phytoplankton and that of CDOM at the
global, regional or seasonal scale (Siegel et al., 2002; Morel et al., 2010;
Xing et al., 2014). Whilst photobleaching is now considered as a major
degradation process of CDOM in surface waters (Del Vecchio and Blough, 2002;
Helms et al., 2008; Bracchini et al., 2010; Swan et al., 2012), the main
source of CDOM in open ocean is still a matter of debate, particularly for
its “humic-like” component, which absorbs light over a broad range of UV
and visible wavelengths and fluoresces in the visible domain (Andrew et al.,
2013). Some works suggest that this humic-like CDOM is in part a remainder
of terrestrial matter that has been diluted and transformed during transit
to and within the ocean (Blough and Del Vecchio, 2002; Hernes and Benner,
2006; Murphy et al., 2008; Andrew et al., 2013). Conversely, other studies
put forward its autochthonous marine source and its production from
phytoplankton, including green algal, diatoms, dinoflagellates (Vernet and
Whitehead, 1996; Romera-Castillo et al., 2010, 2011; Chari et al., 2013),
the diazotrophic (N2-fixing) cyanobacteria Trichodesmium spp. (Subramaniam et al.,
1999; Steinberg et al., 2004) and the non-diazotrophic picocyanobacteria
Synechococcus spp. and Prochlorococcus spp. (Romera-Castillo et al., 2011), from zooplankton (Steinberg
et al., 2004; Ortega-Retuerta et al., 2009), or from the bacterial
degradation (mineralization) of phytoplankton-derived organic matter (Nelson
et al., 1998, 2010; Swan et al., 2009).
The New Caledonian coral lagoon, located in the south-west Pacific, is a
tropical, oligotrophic low-nutrient low-chlorophyll (LNLC) ecosystem in
which diazotrophs such as cyanobacteria Trichodesmium spp. (Dupouy et al., 1988, 2008;
Masotti et al., 2007; Rodier and Le Borgne, 2010) and diazotrophic
picocyanobacteria (Biegala and Raimbault, 2008) but also non-diazotrophic
picocyanobacteria such as Synechococcus spp. and Prochlorococcus spp.
(Biegala and Raimbault, 2008; Neveux et al., 2009) play a significant role. Although the biogeochemical
conditions in the New Caledonian coral lagoon are well documented for
several years (see review by Grenz et al., 2010), the dynamics of CDOM
remains poorly known in this environment. In the framework of the
VAHINE (VAriability of vertical and tropHIc transfer of fixed N2 in the south
wEst Pacific) mesocosm experiment, the objectives of the present
study were (1) to assess the spectral characteristics and the variability of
dissolved and particulate chromophoric materials throughout a 23-day
mesocosm experiment and (2) to tentatively identify the main biogeochemical
contributors (diazotrophic and non-diazotrophic primary producers,
heterotrophic bacteria) driving changes in chromophoric material over the
course of the experiment. Chromophoric parameters we examined here were
absorption coefficients of CDOM [ag(λ)] and particulate matter
[ap(λ)=aϕ(λ)+anap(λ)],
determined over the spectral domain 370–720 nm, the spectral slope of CDOM
(Sg), computed over the range 370–500 nm, as well as fluorescent DOM
(FDOM) components, determined from excitation-emission matrices (EEMs)
combined with parallel factor analysis (PARAFAC).
Material and methods
The mesocosm experiment
Study site and mesocosm description
The VAHINE mesocosm experiment
was conducted from 13 January to 4 February 2013 in the south-west
Pacific at a mouth of the New Caledonian coral lagoon, 28 km off the coast of
New Caledonia (22∘29.073 S–166∘26.905 E) (Fig. 1). At
the deployment site the water depth was 25 m and the bottom was sandy. The
site was protected by land from the dominant trade winds (SE sector) and
characterized by high influence of oceanic oligotrophic waters coming from
outside the lagoon through the Boulari passage (Ouillon et al., 2010). Three
large mesocosms (hereafter called M1, M2 and M3), of 50 m3 volume each,
were deployed (Fig. 2). All details concerning the mesocosm design and
deployment are given in Bonnet et al. (2016). In brief, the mesocosms
consisted in large cylindrical bags made of one polyethylene film and one
ethylene vinyl acetate (EVA, 19 %) film, each 500 µm thick, with
nylon meshing in between to allow maximum resistance and light penetration.
They were 2.3 m in diameter, 15 m in depth and were equipped with removable
sediment traps, allowing collection of sinking material. The top of the bags
were maintained 1 m above the surface with floats to prevent inflow of
external water. Their straightness was maintained by weights at the bottom of
the mesocosms. Before starting sampling, the mesocosms were left opened from
the bottom for 24 h to insure a total homogeneity of the water column.
Location of the site of the VAHINE mesocosm experiment at the mouth
of the New Caledonian coral lagoon, 28 km off the coast of New Caledonia, in
the south-west Pacific (Ocean Data View software version 4.6.5, Schlitzer,
R., http://odv.awi.de, 2014, and Google Earth).
Pictures of the VAHINE mesocosms deployed at the mouth of the New
Caledonian coral lagoon.
Nutrient fertilization
To prevent phosphate limitation, the
mesocosms were fertilized in the evening of day 4 with dissolved inorganic
phosphorus (DIP) to a final concentration of 0.8 µM (see details of
the fertilization procedure in Bonnet et al., 2016). This phosphate
fertilization aimed at stimulating the diazotroph activity.
Sampling and in situ measurements
During the 23 days of the experiment, seawater sampling was performed every
morning from a 4 m2 floating platform at three depths (1, 6 and 12 m)
in each mesocosm and in the surrounding waters close to the mesocosms
(“OUT”) using a compressed air-driven, metal-free pump
(AstiPure™) connected to a polyethylene tubing. Samples were
filled into 50 L polypropylene carboys and immediately transported for
subsampling and sample treatments onboard the R/V Alis, moored 1 nautical mile
away from the mesocosm site. Along with discrete sampling, vertical profiles
of temperature, salinity, Chl a fluorescence, turbidity and light intensity
were obtained daily (at 7 a.m. local time) in each mesocosm and in the
surrounding waters using a 911plus conductivity temperature depth (CTD)
profiler (Sea-Bird Electronics, Inc.). For our specific parameters, i.e.
dissolved and particulate chromophoric materials, we only sampled the
mesocosm M1 at 1, 6 and 12 m depth and the surrounding waters at 1 m depth.
Filtration
Onboard R/V Alis, samples for CDOM absorption and fluorescence measurements
were immediately filtered under low vacuum (< 50 mm Hg) through 0.2 µm
polycarbonate filters (25 mm diameter, Nuclepore) using small,
pre-combusted (450 ∘C, 6 h) glass filtration systems. Prior to
sample filtration, the Nuclepore filters were cleaned by first soaking them
for several minutes in 1 M HCl, then in ultrapure water, and processing them
by filtering through and discarding 300 mL of ultrapure water and lastly 50 mL
of the sample. Then, 1 L of the sample was filtered and the 0.2 µm
filtrate transferred into pre-combusted Schott®
glass bottles for analyses. Powder-free disposable gloves were worn during
sampling, filtration and analyses to avoid sample contamination. All
absorption coefficient measurements [ag(λ) and
ag+p(λ)] were performed directly onboard (see Sect. 2.2),
while samples for fluorescence measurements were stored at 4 ∘C
in the dark for several days until analyses.
The two phases of the experiment
In the results presented below, the 23-day mesocosm experiment was
separated into two periods: P1, from day 5 to day 14, and P2, from day 15 to
day 23. P1 and P2 denote the two phases of the experiment when the
diazotrophic community was dominated by diatom-diazotroph associations
(DDAs), more specifically heterocyst-forming Richelia associated with
Rhizosolenia and unicellular cyanobacteria group C (UCYN-C), respectively (Berthelot et
al., 2015; Turk-Kubo et al., 2015).
Absorption of CDOM and particulate matter
Measurement
Absorption coefficients of CDOM and CDOM + particulate matter
[ag(λ) and ag+p(λ)] were determined by measuring
absorption of 0.2 µm filtered and unfiltered samples using a
point-source integrating-cavity absorption meter (PSICAM) instrument as
described by Röttgers et al. (2007) and Röttgers and Doerffer (2007).
The cavity of the PSICAM was filled with purified water (Milli-Q
water), air bubbles were removed from the cavity wall and the central light
sphere by gentle shaking, and a reference intensity spectrum was recorded
between 370 and 726 nm. Afterwards, sample water was poured into the cavity
in the same way, and a sample intensity spectrum was recorded. The cavity
was rinsed and filled with purified water again, and a second reference
intensity spectrum was recorded. The two reference spectra were used to
calculate two “transmissions” (sample/reference) and, further, two
absorption coefficient spectra. The mean of these two spectra was taken as
the real absorption coefficient spectrum. The calibration of the PSICAM
consisted of determinations of the total cavity reflectivity spectrum by
using solutions of the dye nigrosine (Certistain®,
Merck) with maximum absorption between 1 and 3 m-1. Absorption
spectra were corrected for salinity and temperature differences between
sample and reference water according to Röttgers and Doerffer (2007).
The mean precision of the PSICAM within the range 370–700 nm is ±0.0008 m-1,
whereas its accuracy here is ±2 %, even for
absorption values < 0.1 m-1.
Particulate absorption and CDOM spectral slope determination
Absorption coefficients of particulate matter [ap(λ)] were
determined by subtracting ag(λ) from ag+p(λ) over
the range 370–720 nm. Spectral slope of ag(λ),
Sg (in nm-1), was computed by applying a nonlinear (exponential),
least-squares fit to the ag(λ) values between 370 and 500 nm in
accordance with the following formula:
ag(λ)=ag(λ0)×e-Sg(λ-λ0).
The fit was conducted on raw (i.e. not log-transformed) data according to
the recommendations by Twardowski et al. (2004). The average correlation
coefficient (r) of the exponential least-squares fits was 1.00 (n= 72). The
spectral range used here for the slope determination (370–500 nm) was close
to that employed in previous studies for different oceanic waters (i.e.
350–500 nm) (Babin et al., 2003; Röttgers and Doerffer, 2007; Bricaud et
al., 2010; Para et al., 2010; Organelli et al., 2014).
Fluorescence of DOM
Measurements
FDOM measurements were performed on 0.2 µm filtered samples using a
Hitachi F-7000 spectrofluorometer. The correction of spectra for
instrumental response was conducted from 200 to 600 nm according to the
procedure recommended by the manufacturer (Hitachi F-7000 Instruction
Manual) and fully described in Tedetti et al. (2012). The excitation (Ex)
and emission (Em) correction curves were applied internally by the
instrument to correct each fluorescence measurement acquired in signal over
reference ratio mode. The samples were allowed to reach room temperature in
the dark and transferred into a 1 cm pathlength far-UV transparent silica
quartz cuvette (170–2600 nm; LEADER LAB). The sample in the cuvette was kept
at 20 ∘C inside the instrument using a circulating water bath
connected to the cell holder. The cuvette was cleaned with 1 M HCl and
ultrapure water, and triple rinsed with the sample before use. EEMs were
generated over λEx between 200 and 500 nm in 5 nm intervals,
and λEm between 280 and 550 nm in 2 nm intervals, with 5 nm
slit widths on both Ex and Em sides, a scan speed of 1200 nm min-1, a
time response of 0.5 s and a PMT voltage of 700 V. Blanks (ultrapure water)
and solutions of 0.1 to 10 µg L-1 quinine sulphate dihydrate
(Fluka, purum for fluorescence) in 0.05 M sulphuric acid were run with each
set of the samples. Two replicates were run for each sample.
Fluorescence data processing
Different processing steps were carried out on the fluorescence data: (1) all
the fluorescence data were normalized to the intensity of the ultrapure
water Raman scatter peak at λEx/λEm of
275/303 nm, measured daily as an internal standard (Coble, 1996). This value varied
by 4 % (n= 20). (2) The mean, normalized EEM of ultrapure water was
subtracted from normalized sample EEMs to eliminate the water Raman scatter
signal. (3) These blank-corrected sample EEMs were converted into quinine
sulphate unit (QSU), where 1 QSU corresponded to the fluorescence of 1 µg L-1
quinine sulphate at λEx/λEm
of 350/450 nm (5 nm slit widths) (Coble, 1996; Murphy et al., 2008). The
conversion in QSU was made by dividing each EEM fluorescence data by the
mean slope of a linear regression of fluorescence vs. concentration for the
different quinine sulphate solutions (i.e. 8.4 arbitrary fluorescence
intensity units/QSU). r values of these linear regressions were on average
0.99 and the detection and quantification limits of the fluorescence
measurements were 0.19 and 0.63 QSU, respectively. The water Raman scatter
peak was integrated from λEm 380 to 426 nm at λEx of 350 nm for ultrapure water samples. The mean value was used to
establish a conversion factor between QSU and Raman unit (RU, nm-1),
based on the Raman-area normalized slope of the quinine sulphate linear
regression. The conversion factor was 0.025 RU per QSU. Considering the low
ag(λ) values, samples were not corrected for inner filter
effects (Stedmon and Bro, 2008).
Parallel factor analysis (PARAFAC)
In this work, a PARAFAC model was created and validated for 130 calibrated
EEMs according to the method by Stedmon et al. (2003). The EEM wavelength
ranges used were 210–500 and 280–550 nm for Ex and Em, respectively. EEMs
were merged into a three-dimensional data array of the form: 130 samples × 59
λEx × 136 λEm. The
PARAFAC program was executed using the DOMFluor toolbox v1.6 (Stedmon and
Bro, 2008) running under MATLAB® 7.10.0 (R2010a).
The full analysis showed that no outliers were present in the data set. The
validation of the PARAFAC model (running with the non-negativity constraint)
and the determination of the correct number of components (from 2 to 6
components tested) were achieved through the examination of (1) the
percentage of explained variance, (2) the shape of residuals, (3) the split
half analysis and (4) the random initialization using the Tucker Congruence
Coefficients (Tedetti et al., 2012). The fluorescence intensities of each
component found are given in QSU. The fluorescence intensities in QSU
provided for each sample is the mean of the two replicates with a
coefficient of variance (CV) < 10 %.
Biogeochemical and biological analyses
Filters for the determination of the TChl a concentration were collected by
filtering 550 mL of sample water onto a GF/F filter (Whatman). The filters
were directly shock-frozen and stored in liquid N2. TChl a was extracted
in methanol and measured by fluorometry (Le Bouteiller et al., 1992). The
precision of the measurement was ±0.005 µg L-1.
For the determination of phycoerythrin concentration, water samples (3–4 L)
were filtered onto 0.4 µm Nuclepore polycarbonate filters and
immediately frozen in liquid N2 until analysis. Phycoerythrin was
extracted in a 4 mL glycerol-phosphate mixture (50/50) according to Neveux
et al. (2009) after vigorous shaking for resuspension of particles (Wyman,
1992), and then quantified by fluorometry using a Perkin Elmer LS55
spectrofluorometer (λEx: 450–580 nm at λEm of
605 nm) (Lantoine and Neveux, 1997). The measurement precision was
∼ 16 %.
Pico- and nano-phytoplankton abundances were analysed by flow cytometry.
Samples (1.8 mL) were collected from the mesocosm everyday from 1, 6 and 12 m
depth in cryotubes, fixed with 200 µL of paraformaldehyde (4 %
final concentration), left 15 min at ambient temperature, flash frozen in
liquid N2 and stored at -80 ∘C until analysis on a
FACSCalibur (BD Biosciences) flow cytometer as described in Marie et al. (1999).
Before analysis, samples were thawed at ambient temperature in the
dark. 600 µL of each sample were mixed and homogenized with 25 µL
of TrueCount beads and 10 µL of 2 µm diameter beads
(Fluoresbryte™, Polysciences) used as a reference for size
discrimination between pico- and nano-phytoplankton. Phytoplankton
communities were clustered as Prochlorococcus spp. cell-like,
Synechococcus spp. cell-like,
nanoeukaryotes-cell-like and picoeukaryotes-cell-like according to their
optical properties (light-scattered and fluorescence emission by the cells)
(Marie et al., 1999).
For the determination of microphytoplankton community composition (diatoms),
water samples (250 mL) were taken every day by pumping and preserved with
formalin. In the laboratory, samples were sedimented and microphytoplankton
species were identified and enumerated under-inverted microscope.
Bacterial production (BP) was estimated using the 3H-leucine
incorporation technique (Kirchman et al., 1985), adapted to the
centrifugation method (Smith and Azam, 1992). Radioactivity was counted
using a Liquid Scintillation Analyzer Packard 2100 TR and the 3H
counting efficiency was corrected for quenching. BP was calculated from
leucine incorporation rates using the conversion factor of 1.5 kg C mol-1
leucine, and is shown here in ng C L-1 h-1.
Samples for total organic carbon (TOC) concentrations were collected in
duplicate in precombusted (4 h, 450 ∘C), 12 mL sealed glassware
flask, acidified with orthophosphoric acid and stored in dark at 4 ∘C
until analysis. Samples were analysed by using a TOC-5000
total carbon analyser (Sohrin and Sempéré, 2005). The average TOC
concentrations in the Deep Atlantic Water and low carbon water reference
standards were 45 ± 2 µM C, n= 24 and 1 ± 0.3 µM C,
n= 24, respectively. The analytical precision of the procedure was ≤ 2 %.
Dissolved organic nitrogen (DON) concentrations were calculated from total
nitrogen (TN) concentrations subtracted by particulate organic nitrogen
(PON) and dissolved inorganic nitrogen (DIN) concentrations. Samples were
collected in 50 mL glass bottles and stored at -20 ∘C until
analysis. The samples were divided in two parts after a rapid thaw for
analysis of both organic and inorganic concentrations. TN concentration was
determined according to the wet oxidation procedure described in Pujo-Pay
and Raimbault (1994). Samples for PON concentrations were collected by
filtering 1 L of water on GF/F filters and analysed according to the wet
oxidation protocol (Pujo-Pay and Raimbault, 1994) with a precision of
0.06 µM. DIN concentration was determined according to Aminot and
Kérouel (2007). Measurements were conducted using a segmented flow
auto-analyser (AutoAnalyzer AA3 HR, SEAL Analytical).
Statistics
Linear regression analyses and one-way analyses of variance (ANOVA) were
performed with StatView 5.0 and the statistics package provided in Microsoft
Excel 11.0. ANOVA was used to compare the means of independent data groups
(normally distributed). For the different analyses and tests, the
significance threshold was set at p < 0.05.
Results
Evolution of the core parameters in the mesocosm
The detailed description of temperature, salinity and nutrient concentrations
in the three mesocosms is provided in Bonnet et al. (2016). Briefly, water
temperature progressively increased inside and outside the mesocosms from
25.4 to 26.2 ∘C over the course of the 23-day experiment. Salinity
also progressively increased from 35.1 to 35.5 but this increase was less
pronounced in the surrounding waters with salinities of 35.4 at day 23.
Temperature and salinity were homogeneous over depth in the mesocosms, the
water column having been well mixed throughout the experiment. In the
mesocosms, average concentrations of NO3-+ NO2- were
< 0.04 µM before the DIP fertilization (day 4) and decreased to
0.01 µM at the end of the experiment. In contrast, NH4+
concentrations were ∼ 0.01 µM up to day 18, and then
increased up to 0.06 µM at day 23. DIP concentrations increased from
0.02–0.05 µM before the fertilization to 0.8 µM just after, and
decreased gradually over time to return to their initial concentrations at
day 23 (0.02–0.08 µM). In the surrounding waters, NO3- remained
< 0.20 µM and DIP was 0.05 µM all over the experiment
(Berthelot et al., 2015; Bonnet et al., 2016).
For all the parameters described below, including CDOM and FDOM data, no
significant difference was found with depth, except for TChl a and PON whose
concentrations were higher at 12 m depth than at 1 and 6 m depths (ANOVA,
n= 20–22, p= 0.003–0.04). Therefore, in the following paragraphs, the
parameter descriptors are generally given in term of depth-averaged values.
Evolution of phytoplankton biomass, bacterial production and organic N
and C pools in the mesocosm
TChl a, PON concentrations and BP in the mesocosm M1 and in the surrounding
waters (OUT) generally increased throughout the experiment, with a decrease
from day 4 to day 9 and then an increase from day 9 to the end of the
experiment (Fig. 3a, d, f). This increase was more pronounced in M1, where
TChl a, PON concentrations and BP varying from 0.12 to 0.55 µg L-1,
0.65 to 1.31 µM and 85 to 681 ng C L-1 h-1,
respectively. TChl a, PON concentrations and BP were significantly higher
inside M1 during P2 (day 15 to day 23) than inside M1 during P1 (day 5 to
day 14), and than outside during P1 and P2 (ANOVA, n= 25–30, p < 0.0001–0.004)
(Table 1). Phycoerythrin concentration decreased from day 4
(0.36 µg L-1) to day 9 (0.05 µg L-1), increased
towards day 16 (0.34 µg L-1) and then oscillated to return to
the value of 0.34 µg L-1 at day 23 (Fig. 3b). In contrast, in
OUT, phycoerythrin concentration increased from day 9 to the end of the
experiment, showing a strong raise at day 21 (0.85 µg L-1).
Thus, during P2, phycoerythrin concentration was significantly higher
outside M1 than inside (ANOVA, n= 9, p= 0.004) (Table 1). The TOC
concentration decreased from day 4 (70 µM) to day 11 (64 µM)
and increased from day 11 to day 22 (81 µM) (Fig. 3c). This increase
in the second part of the experiment was not observed in OUT. Although the
TOC concentration was significantly higher during P2 than during P1 in M1
(ANOVA, n= 9, p= 0.03), there was no difference between M1 and OUT
during P2 (ANOVA, n= 7–9, p= 0.2) (Table 1). The DON concentration was
rather constant and only tended to decrease during P2 (Fig. 3e). No
significant difference in DON concentrations was found between M1 and OUT
(ANOVA, n= 22–29, p= 0.07–0.7) (Table 1).
Evolution of (a) total chlorophyll a (TChl a) and (b) phycoerythrin
concentrations (µg L-1), (c) total organic carbon (TOC),
(d) particulate organic nitrogen (PON) and (e) dissolved organic nitrogen (DON)
concentrations (µM) and (f) bacterial production (BP) (ng C L-1 h-1)
in the mesocosm M1 and in the surrounding waters (OUT) at 1, 6 and
12 m depths (except phycoerythrin and TOC concentrations, determined only at
6 m depth) over the course of the 23-day experiment. Dots are mean values
with standard deviations from duplicate measurements, except for
phycoerythrin. For TChl a, standard deviations are comprised within dots.
Black line represents the depth-averaged values. P1: first part of the
experiment, from day 5 to day 14; P2: second part of the experiment, from
day 15 to day 23.
Mean values and associated standard deviations of chromophoric,
biogeochemical and biological parameters of samples collected in the
mesocosm M1 and in the surrounding waters (OUT) during the first part of the
experiment, i.e. from day 5 to day 14 (P1), and during the second part of
the experiment, i.e. from day 15 to day 23 (P2). The means which have
different letters (a, b, c or d) are significantly different (ANOVA, p < 0.05).
M1–P2 values in bold are significantly different from M1-P1, OUT-P1
and OUT-P2 values.
M1-P1 (n)
M1-P2 (n)
OUT-P1 (n)
OUT-P2 (n)
TChl a(µg L-1)
0.19 ± 0.05a (28)
0.42 ± 0.14b (27)
0.21 ± 0.03a (25)
0.30 ± 0.07c (25)
Phycoerythrin (µg L-1)
0.17 ± 0.09a (9)
0.24 ± 0.09a (9)
0.19 ± 0.08a (10)
0.42 ± 0.19b (9)
TOC (µM)
66.5 ± 2.1a (9)
69.7 ± 4.3b (9)
66.6 ± 2.8a (9)
67.7 ± 1.5a,b (7)
PON (µM)
0.81 ± 0.13a (30)
1.10 ± 0.21b (27)
0.71 ± 0.06c (30)
0.87 ± 0.13a (27)
DON (µM)
5.5 ± 1.4a (29)
4.8 ± 0.6a (22)
5.0 ± 0.4a (29)
5.3 ± 1.8a (23)
BP (ng C L-1 h-1)
157 ± 49a (30)
348 ± 142b (27)
135 ± 24a (30)
256 ± 60c (27)
DDAs (× 103 nifH copies L-1)
120 ± 45a,b (5)
54 ± 31a (6)
227 ± 189b,c (5)
200 ± 220a,c (3)
UCYN-C (× 103 nifH copies L-1)
4.5 ± 7.6a (4)
64 ± 24b (6)
1.2 ± 0.8a (5)
2.9 ± 1.7a (3)
Total diatoms (× 103 cell L-1)
17 ± 9a (5)
44 ± 37a (5)
nd
nd
Synechococcus (× 103 cell mL-1)
41 ± 20a (24)
88 ± 14b (23)
nd
nd
Prochlorococcus (× 103 cell mL-1)
12 ± 6a (24)
15 ± 3a (23)
nd
nd
Picoeukaryotes (× 103 cell mL-1)
1.5 ± 0.8a (24)
2.4 ± 0.6b (23)
nd
nd
Nanoeukaryotes (× 103 cell mL-1)
0.9 ± 0.4a (24)
1.5 ± 0.4b (23)
nd
nd
ag(370) (m-1)
0.046 ± 0.004a (30)
0.058 ± 0.009b (27)
0.049 ± 0.005a,c (9)
0.052 ± 0.006c (9)
ag(442) (m-1)
0.013 ± 0.001a (30)
0.016 ± 0.003b (27)
0.015 ± 0.002a,c (9)
0.015 ± 0.001c,b (9)
Sg(nm-1)
0.0172 ± 0.001a,b (30)
0.0174 ± 0.001b (27)
0.0169 ± 0.001a (9)
0.0169 ± 0.001a (9)
ap(442) (m-1)
0.014 ± 0.004a (30)
0.022 ± 0.004b (27)
0.015 ± 0.002a (9)
0.018 ± 0.002c (9)
ap(676) (m-1)
0.005 ± 0.002a (30)
0.009 ± 0.002b (27)
0.005 ± 0.001a (9)
0.008 ± 0.001b (9)
Humic-like (QSU)
4.47 ± 0.76a (30)
4.45 ± 1.09a (18)
nd
nd
Tryptophan-like (QSU)
7.68 ± 1.17a (30)
8.07 ± 2.07a (18)
nd
nd
Tyrosine-like (QSU)
6.57 ± 1.21a (28)
5.49 ± 0.83b (18)
nd
nd
nd: not determined; TChl a: total chlorophyll a concentration; TOC: total
organic carbon concentration; PON and DON: particulate and dissolved organic
nitrogen concentrations; BP: bacterial production; DDAs: diatoms-diazotrophs
associations; UCYN-C: unicellular diazotrophic cyanobacteria Group C;
ag(370) and ag(442): absorption coefficients of CDOM at 370 and
442 nm; Sg: spectral slope of CDOM; ap(442) and ap(676): absorption
coefficients of particulate matter at 442 and 676 nm; Humic-like,
tryptophan-like and tyrosine-like: fluorescence intensity of humic-like,
tryptophan-like and tyrosine-like FDOM fluorophores. Detailed data about
diazotrophs (DDAs and UCYN-C) are found in Turk-Kubo et al. (2015).
Evolution of the abundance of (a) diatoms-diazotrophs associations
(DDAs) and (b) unicellular diazotrophic cyanobacteria Group C (UCYN-C)
(× 103 nifH copies L-1) in the mesocosm M1 and in the
surrounding waters (OUT), and (c) total diatoms (× 103 cell L-1),
(d) Synechoccocus spp., (e) Prochloroccocus spp.,
(f) picoeukaryotes and (g) nanoeukaryotes
(× 103 cell mL-1) in the mesocosm M1 only, over the
course of the 23-day experiment. Synechoccocus spp., Prochloroccocus spp., picoeukaryotes and
nanoeukaryotes were determined at 1, 6 and 12 m depths, while DDAs, UCYN-C
and total diatoms were determined solely at 6 m depth. For DDAs and UCYN-C,
dots are mean values with standard deviations from duplicate measurements.
Black line represents the depth-averaged values. P1: first part of the
experiment, from day 5 to day 14; P2: second part of the experiment, from
day 15 to day 23. Detailed data about diazotrophs (DDAs and UCYN-C) are
found in Turk-Kubo et al. (2015).
The abundance of diazotrophs DDAs inside M1 increased from day 3 (77 × 103 nifH copies L-1)
to day 9 (190 × 103 nifH copies L-1), decreased from day
9 to day 15 (5.4 × 103 nifH copies L-1) and finally
increased from day 15 to day 23 (78 × 103 nifH copies L-1). In OUT a quite similar pattern was observed
despite a high value of 450 × 103 nifH copies L-1 at day 18
(Fig. 4a). No significant difference in the abundance of DDAs was observed
in M1 between P1 and P2, and between M1 and OUT (ANOVA, n= 3–6, p= 0.05–0.8)
(Table 1). On the other hand, the abundance of diazotrophic Group
UCYN-C strongly increased from day 9 (0.54 × 103 nifH copies L-1)
to day 23 (110 × 103 nifH copies L-1) in M1, while
it increased much more slowly in OUT from day 10 (0.32 × 103 nifH copies L-1)
to day 22 (4.8 × 103 nifH copies L-1) (Fig. 4b).
Hence, the abundance of UCYN-C was much higher in M1 during P2 than in
M1 during P1 (14 times higher) and than in OUT during P1 and P2 (22–53 times
higher) (ANOVA, n= 3–6, p < 0.0001) (Table 1). It should be noticed
that the abundances of DDAs and UCYN-C are reported as nifH (gene) copies L-1
rather than cells L-1 because there is currently little
information about the number of nifH copies per genome in these diazotroph
targets (Turk-Kubo et al., 2015). Total diatoms in M1 decreased from day 2
(47 × 103 cell L-1) to day 9 (6 × 103 cell L-1)
and then oscillated to reach 41 × 103 cell L-1
at the end of the experiment, with a maximum value of 120 × 103 cell L-1
at day 15 (Fig. 4c). This was essentially due to the
large diatom Cylindrotheca closterium (data not shown). No difference in abundance of total diatoms
was observed between P1 and P2 (ANOVA, n= 5, p= 0.2). The abundances of
Synechococcus spp., Prochlorococcus spp., picoeukaryotes and nanoeukaryotes decreased from day 4
(∼ 43, 16, 2.2 and 0.9 × 103 cell mL-1,
respectively) to day 9 (∼ 18, 5, 0.8 and 0.6 × 103 cell mL-1,
respectively) (Fig. 4d–g). From day 9 to the end of
the experiment, the abundance of Synechococcus spp. and picoeukaryotes noticeably
increased to reach ∼ 90 and 3.4 × 103 cell mL-1
at day 23 respectively, whereas the increase in Prochlorococcus spp. and
nanoeukaryotes was much less (to ∼ 20 and 1.3 × 103 cell mL-1
at day 23, respectively). The abundance of
Synechococcus spp., picoeukaryotes and nanoeukaryotes was significantly higher in P2 than
in P1 (ANOVA, n= 23–24, p < 0.0001–0.002), while that of
Prochlorococcus spp. was not different (ANOVA, n= 23–24, p= 0.07) (Table 1).
Absorption spectra of CDOM and particulate matter
CDOM absorption spectra of samples collected in M1 and OUT were quite
similar, displaying an exponential decrease in ag(λ) without
any significant shoulder (Fig. 5). ap(λ) spectra, which reflect
the absorption by both phytoplankton and NAP, were characterized by two main
Chl a peaks, one between 432 and 442 nm (at 436 nm on average) and one
between 672 and 682 nm (at 676 nm on average), while several shoulders also
emerged at 376, 416, 464, 490 and 550 nm (Fig. 5). Hereafter,
ag(λ) is presented at 370 and 442 nm, while ap(λ)
is given at 442 and 676 nm, the two latter wavelengths corresponding to the
absorption maxima of Chl a.
Absorption spectra of chromophoric dissolved organic matter (CDOM)
and particulate matter over the ranges 370–720 nm of samples collected in
the mesocosm M1 at 1, 6 and 12 m depths and in the surrounding waters at 1 m
depth. Black lines represent the average of all spectra and shaded areas
represent the measured minimal and maximal values. Peaks and shoulders are
reported for particulate matter.
Evolution of absorption coefficients, spectral slope in the
mesocosm
In M1, absorption coefficients decreased from day 4 to day 9 and then
increased from day 9 to the end of the experiment (day 23), leading to
variations in the ranges 0.041–0.067 m-1 for ag(370),
0.011–0.020 m-1 for ag(442), 0.009–0.025 m-1
for ap(442) and
0.003–0.012 m-1 for ap(676) (Fig. 6a, b, d, e). In OUT, these
parameters also increased from day 9 or 10 to day 23 but with lower
amplitude. Inside M1, all these absorption coefficients were significantly
higher during P2 than during P1 (ANOVA, n= 27–30, p < 0.0001).
However, only ag(370) and ap(442) were significantly higher in M1
than outside during P2 (ANOVA, n= 9–27, p= 0.004–0.02) (Table 1).
Sg inside and outside M1, ranging from 0.0148 to 0.0188 nm-1, did
not display any clear pattern throughout the experiment (Fig. 6c).
Evolution of (a) absorption coefficient of CDOM at 370 nm
[ag(370) in m-1], (b) absorption coefficient of CDOM at 442 nm
[ag(442) in m-1], (c) spectral slope of CDOM absorption in the range
370–500 nm (Sg in nm-1), (d) absorption coefficient of particulate
matter at 442 nm [ap(442) in m-1] and (e) absorption coefficient of
particulate matter at 676 nm [ap(676) in m-1] in the mesocosm M1 at
1, 6 and 12 m depths and in the surrounding waters (OUT) at 1 m depth over
the course of the 23-day experiment. Dots are mean values with standard
deviations from duplicate measurements, except for Sg. Black line
represents the depth-averaged values. P1: first part of the experiment, from
day 5 to day 14; P2: second part of the experiment, from day 15 to day 23.
Spectral characteristics and identification of FDOM components
Three FDOM components (C1–C3) were identified by the PARAFAC model validated
on 130 EEM samples from M1 and OUT. The spectral characteristics of C1–C3
are reported in Fig. 7. These components exhibited one or two Ex maxima and
one Em maximum. C1, with a maximum at λEx/λEm of
230/476 nm, corresponded to the category of ultraviolet C (UVC) humic-like fluorophores,
referred to as peak A (Coble, 1996, 2007; Ishii and Boyer, 2012). C2 and C3
had two maxima each, located at λEx1, λEx2/λEm of 225, 280/344 and 225, 275/304 nm,
respectively (Fig. 7). They belonged to the group of protein-like
fluorophores, C2 being analogous to tryptophan-like fluorophore (peaks T)
and C3 being analogous to tyrosine-like fluorophore (peaks B) (Coble, 1996,
2007).
Spectral characteristics of the three FDOM components (C1–C3)
validated by the PARAFAC model for 130 EEMs of samples collected in the
mesocosm M1 at 1, 6 and 12 m depths and in the surrounding waters at 1 m
depth over the course of the 23-day experiment. Both contour (left column)
and line (right column) plots are depicted. The line plots show the
excitation (left side) and emission (right side) fluorescence spectra. The
dotted grey lines correspond to split half validation results. The
excitation and emission maxima (λEx and λEm) of
each component are given.
Evolution of FDOM components in the mesocosm M1
Inside M1, the fluorescence intensity of humic-like fluorophore decreased
from day 2 (∼ 5.3 QSU) to day 8 (∼ 2.7 QSU),
increased from day 8 to day 14 (∼ 4.8 QSU) and dropped down to
∼ 2.5 QSU at day 15. Then, it increased to reach
∼ 5.6 QSU at day 20 (Fig. 8a). The fluorescence intensity of
tryptophan-like fluorophore decreased from day 3 (∼ 9.1 QSU)
to day 8 (∼ 5.3 QSU) (Fig. 8b). At day 9, it increased up to
∼ 8.3 QSU and remained relatively stable up to day 14
(∼ 8.4 QSU). After a reduction at day 15 (∼ 5.9 QSU),
the fluorescence intensity increased up to the end of the experiment
(∼ 10.4 QSU at day 20). The fluorescence intensity of
tyrosine-like fluorophore decreased from day 5 (∼ 8.2 QSU) to
day 15 (∼ 3.9 QSU) and then slowly increased to day 20
(∼ 6.2 QSU) (Fig. 8c). While for humic- and tryptophan-like
fluorophores no differences in their fluorescence intensity were observed
between P1 and P2 (ANOVA, n= 18–30, p= 0.4–0.9), the fluorescence
intensity of tyrosine-like fluorophore was significantly lower during P2
(ANOVA, n= 18–28, p= 0.002) (Table 1). Overall, the FDOM pool was
dominated by protein-like material: the combined fluorescence of tryptophan
and tyrosine fluorophores ranged from 9.1 to 22.3 QSU, while the
fluorescence of humic fluorophore ranged from 1.9 to 6.2 QSU.
Evolution of the fluorescence intensities (QSU) of the three FDOM
components: (a) humic-like, (b) tryptophan-like and (c) tyrosine-like
fluorophores in the mesocosm M1 at 1, 6 and 12 m depths over the course of
the 23-day experiment (actually up to day 20 and not to day 23). Dots are
mean values with standard deviations from duplicate measurements. Black line
represents the depth-averaged values. P1: first part of the experiment, from
day 5 to day 14; P2: second part of the experiment, from day 15 to day 23.
Fluorescence intensities in the surrounding waters (OUT) at 1 m depth were
determined on only few samples at the beginning and the end of the
experiment and are thus not presented here.
Relationships between the chromophoric and the
biogeochemical/biological parameters
Table 2 presents r values of linear regressions between the chromophoric and
the biogeochemical/biological parameters for the samples collected in M1
from day 5 to day 20. Here we consider that only the correlations that are
very highly significant (p < 0.0001) reflect relevant linear
relationships. ag(370, 442) and ap(442, 676) were not that much
correlated to each other (r= 0.52–0.62, n= 36, p < 0.0001–0.002).
Sg was not correlated to ag(370, 442) (r= 0.15–0.22, n= 36,
p= 0.06–0.9). Even though humic- and tryptophan-like fluorophores were very
highly correlated (r= 0.67, n= 36, p < 0.0001), they did not show
any coupling with tyrosine-like fluorophore (r= 0.20–0.48, n= 36, p= 0.005–0.2).
Moreover, none of these three fluorophores was very highly
correlated to the absorption coefficients and spectral slope (r= 0.09–0.42,
n= 36, p > 0.5–0.05) (Table 2). These correlations
emphasize the decoupling between the CDOM and FDOM materials during the
experiment.
Pearson's correlation coefficients (r) of linear regressions between
the chromophoric and the biogeochemical/biological
parameters of samples collected in the mesocosm M1 from
day 5 to day 20, i.e. from the day after the dissolved inorganic phosphorus
fertilization to almost the end of the experiment (P1 + P2) (n= 36).
ag(370)
ag(442)
Sg
ap(442)
ap(676)
Humic
Trypto.
Tyrosine
ag(442)
0.90
Sg
0.22
-0.15
ap(442)
0.61
0.52
0.23
ap(676)
0.62
0.53
0.30
0.93
Humic
0.42
0.36
0.13
0.22
0.10
Trypto.
0.28
0.24
0.28
0.17
0.16
0.67
Tyrosine
-0.09
-0.25
0.11
-0.28
-0.39
0.48
0.20
TChl a
0.68
0.60
0.32
0.86
0.88
0.22
0.21
-0.33
Phyco.*
0.45
0.42
0.11
0.74
0.73
0.05
0.00
-0.35
TOC*
0.35
0.16
0.63
0.57
0.59
0.52
0.43
0.28
PON
0.71
0.58
0.29
0.75
0.70
0.43
0.29
0.04
DON
-0.30
-0.23
-0.13
-0.14
-0.04
-0.26
-0.10
-0.14
BP
0.75
0.72
0.10
0.78
0.72
0.43
0.32
-0.12
DDAs*
-0.44
-0.38
-0.52
-0.85
-0.78
0.20
0.05
0.60
UCYN-C*
0.73
0.67
0.55
0.90
0.85
0.15
0.23
-0.47
Diatoms*
-0.07
-0.08
0.40
0.49
0.47
-0.85
-0.74
-0.88
Synecho.
0.76
0.76
0.08
0.83
0.76
0.35
0.29
-0.27
Prochlo.
0.42
0.47
0.08
0.57
0.50
0.13
0.03
0.00
Picoeuka.
0.52
0.62
-0.07
0.71
0.58
0.40
0.34
-0.25
Nanoeuka.
0.48
0.45
0.01
0.65
0.58
0.11
0.01
-0.35
Correlation coefficients (r) in bold are very highly significant
(p < 0.0001). * Correlations determined on a lower number of samples (n): 15 for
Phyco. and TOC, 10 for DDAs, 9 for UCYN-C and 8 for diatoms. ag(370) and
ag(442): absorption coefficients of CDOM at 370 and 442 nm (m-1);
Sg: spectral slope of CDOM; ap(442) and ap(676): absorption
coefficients of particulate matter at 442 and 676 nm (m-1); Humic:
fluorescence intensity of humic-like fluorophore (QSU); Trypto.:
fluorescence intensity of tryptophan-like fluorophore (QSU); Tyrosine:
fluorescence intensity of tyrosine-like fluorophore (QSU); TChl a: total
chlorophyll a concentration (µg L-1); Phyco.: phycoerythrin
concentration (µg L-1); TOC: total organic carbon concentration
(µM); PON and DON: particulate and dissolved organic nitrogen
concentrations (µM); BP: bacterial production (ng C L-1 h-1);
DDAs: diatoms-diazotrophs associations (nifH copies L-1);
UCYN-C: unicellular diazotrophic cyanobacteria Group C (nifH copies L-1);
Diatoms: total diatoms (cell L-1); Synecho.: Synechococcus spp. (cell mL-1);
Prochlo.: Prochlorococcus spp. (cell mL-1); Picoeuka.: Picoeukaryote (cell mL-1);
Nanoeuka.: Nanoeukaryote (cell mL-1). Detailed data about diazotrophs
(DDAs and UCYN-C) are found in Turk-Kubo et al. (2015).
All absorption coefficients were very highly positively correlated to
Synechococcus spp. abundance (r= 0.76–0.83), BP (r= 0.72–0.78), TChl a concentration
(r= 0.60–0.88), PON concentration (r= 0.58–0.75) and picoeukaryote
abundance (r= 0.52–0.71) (n= 36, p < 0.0001). Linear
relationships between ag(370) or ap(442) and Synechococcus spp. abundance are
presented in Fig. 9. Sg as well as the three FDOM fluorophores did not
present any highly significant correlation with the
biogeochemical/biological constituents. Phycoerythrin, TOC, DDAs, UCYN-C and
total diatoms did not display any very highly significant correlations with
the chromophoric parameters, although some r values were quite high (for
instance 0.90 between UCYN-C and ap(442)). This is because these
correlations were determined for a lower number of samples (8–15).
Nonetheless these relationships
highlighted interesting, albeit not very highly significant, trends such as positive (negative) relationships
between absorption coefficients (tyrosine-like fluorophore) and UCYN-C
abundance as well as negative (positive) relationships between absorption
coefficients (tyrosine-like fluorophore) and DDA abundance (Table 2).