1. Introduction

The Southern Ocean (SO), defined here as the waters south of the Subtropical Front, is a significant CO$_{2}$ sink [1,2] which is undergoing physical and chemical changes from ocean warming, acidification, and freshening from increased sea-ice melt [3,4]. These environmental pressures will in turn alter the present and future physiology, abundance and community structure of phytoplankton [3,5], the photosynthetic microbes underpinning the biological carbon pump (BCP). Given the vastness and remoteness of the SO, ocean colour remote sensing is an important tool for observing changes in SO phytoplankton [6–9]. However, interpreting ocean colour data relies on robust relationships between the remotely measured signal, the in situ optical properties and the biogeochemical properties of interest. Consequently, substantial field sampling and the collection of both in situ biological and bio-optical datasets are essential to the application of ocean color remote sensing approaches. An additional benefit of establishing robust relationships is that by measuring one observable variable–for example the chlorophyll a concentration at the surface which can be achieved with ocean color remote sensing–further information can be inferred about correlated parameters like phytoplankton community composition, average size, or vertical distribution [10,11], all of which have implications for ecosystem function and the efficiency of the BCP.

When sunlight enters the surface ocean, it is absorbed and scattered by particles (phytoplankton and other non-algal particles), coloured dissolved organic matter (CDOM) and water, and partially scattered back out of the surface ocean with a spectrum that is different to that of the incoming sunlight [12]. The spectra of light exiting the surface ocean and its overall intensity, commonly referred to as ocean colour, is therefore dependent on the relative composition, abundance, and size (for particles) of the dissolved or particulate material in the surface ocean [13]. Empirical relationships between the ocean colour spectrum and bio-optical properties of optically active constituents (phytoplankton, CDOM, non-algal particles), or biogeochemical properties, enable us to deconvolve the ocean colour spectrum into useful quantities such as chlorophyll a concentration. Relationships between bio-optical properties (e.g., absorption), biogeochemical quantities (e.g., chlorophyll a concentration), and community characteristics (e.g., phytoplankton species or size), are however not uniform throughout the World Ocean [14]. Variability in these relationships can be expected on both regional and temporal scales driven by differences in phytoplankton species composition and changes in the availability of essential resources for phytoplankton growth such as light and nutrients [15,16].

Generally there is a paucity of in situ bio-optical and matched biogeochemical data in the SO, as compared to other ocean basins. Measurements are temporally constrained to the summer season and spatially constrained to a few areas of interest. Areas such as the Antarctic Peninsula, Kerguelen Plateau, Sub-Antarctic and Polar Frontal Zones of the Indian Ocean, Ross Sea and Scotia Seas are comparatively well sampled for properties such as pigment concentration and particulate absorption compared to other areas of the Southern Ocean [17–21]. However these areas represent only a fraction of the spatial extent of the SO and a mere glimpse into the optical properties of the many different biomes within the SO [6]. Areas such as open ocean waters of the high latitudes (South of 60$\mathrm {^{\circ }}$S) have been particularly undersampled, specifically the Bellinghausen and Amundsen Seas [22], and also the low latitudes of the Atlantic Ocean sector.

Past studies have shown that the bio-optical properties of phytoplankton within the SO are different to other oceans [18,23,24]. For example the chlorophyll-specific absorption coefficient ($\textit {a}^{\textrm {*}}$) has been reported to be considerably lower than values in low- and mid-latitude oceans at comparable chlorophyll a concentrations [18]. The lower $\textit {a}^{\textrm {*}}$ was typically due to increased pigment packaging, resulting from acclimation to low in situ light intensities and/or larger cell size [18,25]. However, large variations in both the absolute values and range of values of $\textit {a}^{\textrm {*}}$ has been observed in the Southern Ocean [26], sometimes within the same region [20,25]. Some of this disparity is explained by methodological differences such as the choice of pathlength amplification factors for the particulate absorption measurements [18,20]. Other factors, such as spatial and seasonal availability of light and nutrients [27], and phytoplankton species composition [26,28–30], leading to variations in cell sizes and pigment composition [31,32], have been identified as being important determinants. Overall, there are still fundamental gaps in our understanding of if, how and why Southern Ocean bio-optical and biogeochemical relationships differ from other oceans.

The SO is a unique environment for phytoplankton due to its strong circumpolar ocean currents and seasonal or permanent ice zones which create hydrographic boundaries that divide the surface ocean [33], where phytoplankton reside, into oceanic zones of distinct physical and chemical properties [34]. Environmental factors which influence phytoplankton growth and species selection such as temperature, light availability and dissolved nutrient composition and concentrations differ [35], both within and between regions. For example in the Subantarctic Zone, low silicic acid favours a more diverse community of typically nano- (2-20 $\mathrm {\mu }$m) and pico- (<2 $\mathrm {\mu }$m) sized phytoplankton [36,37], while further south, below the Polar Front, increased nitrate, phosphate and silicic acid encourage the growth of microphytoplankton (>20 $\mathrm {\mu }$m), specifically diatoms, which tend to dominate the phytoplankton community [28,36,38,39]. Across the entire SO, phytoplankton are typically co-limited by light, dissolved iron and silicic acid [40]. The dominant species of a bloom in response to increases in light, iron or silicic acid, can vary with location and change over time [41].

The Antarctic Circumnavigation Expedition (ACE) presented an opportunity to collect the most spatially extensive SO dataset for investigating the relationship between phytoplankton particulate absorption, pigments and particle size. Specifically, we investigated the variations in phytoplankton pigments and community composition with chlorophyll a and pheopigment concentration and the impact on the phytoplankton absorption coefficients. We also looked into the changes in particle volume as a function of chlorophyll a. The results from this study provide a set of chlorophyll-specific particulate absorption coefficients for modelling phytoplankton abundance and particulate absorption in this region.

2. Methods

2.1 Study area and sampling

Samples were collected during legs 1-3 of the ACE expedition (see Walton and Thomas [42]) aboard the RV Akademik Tryoshnikov during the Austral Summer from 20 December 2016-19 March 2017. Legs 1-3 of the expedition travelled east around the Southern Ocean from Cape Town, South Africa, through the Indian Ocean to Hobart, Australia (leg 1), via the Pacific Ocean to Punta Arenas, Chile (leg 2), and through the Atlantic Ocean back to Cape Town (leg 3). During legs 1 and 3, the ship’s meridional position remained within approximately 40 to 60$\mathrm {^{\circ }}$ S, venturing further south to almost 80$\mathrm {^{\circ }}$ S during leg 2 (Fig. 1). In addition, a number of subantarctic islands and the Antarctic continent were visited. Most locations are highlighted in Fig. 1. The voyage GPS track and other datasets cited here are available at https://zenodo.org/communities/spi-ace. Samples for particulate absorption, pigment analysis and particle size distribution were collected from the underwater seawater supply (Grundfos Lenntech centrifugal pump) fixed at 4.5 m every 3-6 hours.

 

Fig. 1. Map of underway sampling points during ACE legs 1-3, colourmapped with [Tchla] (sum of monovinyl-chlorophyll $\textit {a}$, divinyl-chlorophyll $\textit {a}$, chlorophyll $\textit {a}$ epimers, chlorophyll $\textit {a}$ isomers and chlorophyllide $\textit {a}$ concentrations). Grey points indicate location of samples extracted from the NOMAD dataset. Red points indicate island locations. Locations of the major fronts along the track identified in situ data [43] are marked with ’X’ markers. Dashed lines in background are the climatological positions of the major Southern Ocean fronts from Orsi et al. [33]. STF = Subtropical Front, SAF = Subantarctic Front, PF = Polar Front, SACCF = Southern Antarctic Circumpolar Current Front.

Download Full Size | PDF

2.2 Physical oceanography

A thermosalinograph (Chelsea Technologies AquaLine Ferrybox system) connected to the underway seawater supply collected a semi-continuous record of sea surface temperature (SST; $\mathrm {^{\circ }}$C) and salinity of the surface ocean (data available at [43]).

The mixed layer depth (MLD; m) at each sampling point was retrieved by linearly interpolating the climatological monthly MLD means from de Boyer Montégut et al. [44] to the sampling dates. The MLDs were derived using the density threshold method (depth where $\mathrm {\sigma }_{\theta }$=$\mathrm {\sigma }_{{\theta }_{\textrm {10m}}}$+/-0.03 kg $\textrm {m}^{-3}$; [44]; Fig. 2). The MLD was additionally calculated using density profiles derived from conductivity-temperature-depth (CTD; [45]) casts. However those measurements were sporadic and cannot be interpolated meaningfully along the complete ship transit. A comparison of the CTD derived estimates and climatology estimates show a very similar pattern across the timeseries (CTD=1.1x - 12, $\textit {R}^{\textrm {2}}=0.6$, ${p}<0.05$).

 

Fig. 2. Timeseries of leg 1 (top), leg 2 (middle) and leg 3 (bottom) showing latitude ($\mathrm {^{\circ }}$S; dotted grey line), mixed layer depth (MLD (m) from the de Boyer Montégut et al. [44] density threshold climatology; black circles), MLD from in situ CTD casts; open black circles, sea surface temperature ($\mathrm {^{\circ }}$C; blue circles) and salinity (red circles) from the underway Ferrybox. Some island visits are indicated by labelled blue lines. Background colours of indicate the ocean zones between major ocean fronts from [43] where red = STZ, orange=SAZ, yellow=PFZ, cyan=AZ, purple=MIZ. See text in Physical Oceanography section for definitions.

Download Full Size | PDF

The depth of the euphotic zone ($\textit {Z}_{\textrm {eu}}$; m) was derived as a polynomial expression of [Tchl$\textit {a}$] (sum of monovinyl-chlorophyll $\textit {a}$, divinyl-chlorophyll $\textit {a}$, chlorophyll $\textit {a}$ epimers, chlorophyll $\textit {a}$ isomers and chlorophyllide $\textit {a}$ concentrations) at the surface [Tchl$\textit {a}_{\textrm {surf}}$] (mg $\textrm {m}^{-3}$; equation from Fig. 9 in Morel et al. [46]):

Frontal positions from in situ data [43] and the climatological means of Orsi et al. [33] are plotted to aid interpretation of the results (Figs. 1 and 2). The positions of the Subtropical Front (STF), Subantarctic Front (SAF), Polar Front (PF) and the Southern Antarctic Circumpolar Current Front (SACCF) were derived in situ by identifying strong lateral gradients in temperature, salinity or sea surface height and were in close proximity to their climatological mean positions [33]. Oceanic zones are defined as the Subtropical Zone, north of the STF, the Subantarctic Zone (SAZ) between the the STF and SAF, the Polar Frontal Zone (PFZ) between the SAF and PF, the Antarctic Zone (AZ) between the PF and SACCF, and the Marginal Ice Zone (MIZ), south of the SACCF.

The median light intensity within the mixed layer, $\textit {I}_{\textrm {g}}$ ($\mu$mol photons m$^{2}$ s$^{-1}$) was estimated for each sample. First the light attenuation coefficient $\textit {K}_{\textrm {d}}$PAR ($m^{-1}$) was calculated as per Morel [47]. For each sample, the average instantaneous PAR over the previous 72 hours (avPAR(0)) in $\mu$mol photons m$^{2}$ s$^{-1}$ was calculated from 1 min resolution above-water PAR measurements [48]. At each sample point, the average PAR at 1 metre (z) intervals between the surface and mixed layer depth was calculated as per [49]

2.3 Phytoplankton pigments

Particles from up to 2 L of seawater were collected on 0.7 $\mathrm {\mu }$m pore size 25 mm diameter glass fibre filters (GF/F; Whatman) under low vacuum pressure, immediately frozen in liquid nitrogen and stored at -80 $\mathrm {^{\circ }}$C for post-voyage analysis. Phytoplankton pigment concentrations were determined using high performance liquid chromatography (HPLC) following Ras et al. [50]. A full description of the pigment analysis and the data are available from Antoine et al. [51].

Following Ras et al. [50], total chlorophyll $\textit {a}$ [Tchl$\textit {a}$] is the sum of monovinyl-chlorophyll $\textit {a}$, divinyl-chlorophyll $\textit {a}$, chlorophyll $\textit {a}$ epimers, chlorophyll $\textit {a}$ isomers and chlorophyllide $\textit {a}$ concentrations. [Pheo$\textit {a}$] is the sum of the pheopigments pheophorbide $\textit {a}$ and pheophytin $\textit {a}$, and [Tchl$\textit {a}$ + Pheo$\textit {a}$] is the sum of [Tchl$\textit {a}$] and [Pheo$\textit {a}$]. [Tchl$\textit {b}$] is the concentration of chlorophyll $\textit {b}$ only. [Tchl$\textit {c}$] is the sum of chlorophyll $\textit {c}$3 and chlorophyll $\textit {c}$1$\textit {c}$2. Concentrations of total photosynthetic [PSC] and photoprotective [PPC] pigments were also calculated where [PSC] is the sum of fucoxanthin, peridinin, 19’hexanoloxyfucoxanthin and 19’butanoloxyfucoxanthin, and [PPC] is the sum of diadinoxanthin, alloxanthin, zeaxanthin, lutein, $\alpha$-carotene and $\beta$-carotene.

2.4 Filter-pad absorption measurements

Particulate absorption measurements were made (first) on the same filters analysed for pigments. A full description of the measurement protocols and the data are available at [52]. Briefly, measurements of optical density from 350-750 nm (1 nm interval) were blank corrected and converted into particulate absorption ($\textit {a}_{\textrm {p}}(\lambda )$; $\textrm {m}^{-1}$). Unless specified, all chlorophyll-specific coefficients presented in this paper have been derived by dividing by [Tchl$\textit {a}$ + Pheo$\textit {a}$].

2.4.1 Decomposition of phytoplankton and non-algal particle absorption

The phytoplankton ($\textit {a}_{\mathrm {\phi }}(\lambda )$; $\textrm {m}^{-1}$) and non-algal particle ($\textit {a}_{\textrm {nap}}(\lambda )$; $\textrm {m}^{-1}$) absorption spectra were numerically decomposed using a semi-analytical inversion of the particulate absorption spectra using Bricaud et al. [53] and Bricaud et al. [54] phytoplankton chlorophyll-specific spectral shapes as the eigenvectors. Equation (3) was fitted to the particulate absorption spectra with a least squares fit with trust-reflective algorithm [55,56] in Matlab 2019b.

2.5 Particle size distribution

Measurements of the particle size distribution were made on a Beckmann Multisizer 3 Coulter counter with a 100 $\mu$m aperture tube detecting particles in the range of 2-60 $\mu$m across 400 bins. Each sample constituted a cumulative analyses of 20 runs of 2000 $\mu$L each. Blank samples of 0.2 $\mu$m filtered seawater were also measured and the median blank subtracted from all whole samples (mean of 20 subsample measurements). Counts were converted to particles per m$^{\textrm {3}}$ and (equivalent spherical) particle volume ($\mu \textrm {m}^{3}$ $\textrm {m}^{-3}$). The diameter bins were restricted to 2-30 $\mu$m as no particles were observed in larger bins. Plots of particle concentration vs bin diameter were visually inspected and samples with high noise or particle concentrations constrained to just a few bins were flagged as poor quality and removed. The total particle volume in each sample ($\mu \textrm {m}^{3}$ $\textrm {m}^{-3}$) was derived by integrating the particle volume distribution for each sample. The mean (equivalent spherical) particle volume of each sample ($\mu \textrm {m}^{3}$) was calculated by first dividing the particle volume distribution ($\mu \textrm {m}^{3}$ $\textrm {m}^{-3}$) by the bin width (in m), before integrating to total particle volume and dividing by the particle concentration (also corrected for bin width).

2.6 Data analysis

2.6.1 Pigment and packaging indices

The contribution of three pigment-based sized classes [57,58], microphytoplankton ($\textit {f}_{\textrm {micro}}$; > 20 $\mu$m, unitless), nanophytoplankton ($\textit {f}_{\textrm {nano}}$; 2-20 $\mu$m, unitless), and picophytoplankton ($\textit {f}_{\textrm {pico}}$; < 2 $\mu$m, unitless) to the pigment biomass were calculated according to [10].

Pigment packaging was quantified following the pigment reconstruction technique [53] to derive $\textit {a}_{\textrm {sol}}({\lambda })$, the absorption coefficient of the same material dispersed into solution. The $\textit {a}_{\textrm {sol}}({\lambda })$ is calculated as

2.6.2 Comparison data

For comparison, data from the Southern Ocean (south of 40$\mathrm {^{\circ }}$S) within the NASA bio-Optical Marine Algorithm Dataset v2.a (NOMAD) [63] are presented as grey symbols in some figures. The locations of these samples are constrained primarily to Drake’s Passage, Scotia Sea and Antarctic Peninsula waters and along parallels between 80-90 $\mathrm {^{\circ }}$E in the Indian Ocean sector (Fig. 1).

The Bricaud et al. [53,54,64] datasets are used herein as comparison with our dataset. They include data from the Mediterranean Sea, North Atlantic Ocean, Benguela Upwelling and Pacific Ocean. A significant fraction of the 1998 dataset, especially at high chlorophyll concentration are from the Gulf of St-Lawrence. The pigment and particulate absorption data from [53] are also included as mauve symbols in some figures for comparison. For ease of interpretation, the following acronyms have been assigned to these datasets: B98 for Bricaud et al. [64], B04 for Bricaud et al. [53] and B10 for Bricaud et al. [54].

It is noted that many bio-optical papers presenting chlorophyll-specific parameterisations and statistics vary in their definition of total chlorophyll a and the abbreviations or symbols used to represent the inclusion of chlorophyll a derivatives and pheopigments. For example the chlorophyll-specific parameters and functions of B98, B04 and B10 include pheopigments in the calculations or normalisations, whereas Uitz et al. [10] does not. As we show below, the concentration of pheopigments, pheophorbide $\textit {a}$ and pheophytin $\textit {a}$ ([Pheo$\textit {a}$]) in our ACE dataset vary from 5 % to > 100 % of the [Tchl$\textit {a}$]. In contrast pheopigments contributed < 5 % in the B04 dataset. Additionally, when describing the "trophic status" of a water body using the chlorophyll a concentration, typically this constitutes monovinyl-chlorophyll $\textit {a}$ and divinyl-chlorophyll $\textit {a}$ [57]. For these reasons, we present most bio-optical, pigment, and particle relationships as a function of, or normalised to, [Tchl$\textit {a}$] only ([Tchl$\textit {a}$] = monovinyl-chlorophyll $\textit {a}$, divinyl-chlorophyll $\textit {a}$, chlorophyll $\textit {a}$ epimers, chlorophyll $\textit {a}$ isomers and chlorophyllide $\textit {a}$). However, for comparison, we do present some relationships as a function of, or normalised to, [Tchl$\textit {a}$ + Pheo$\textit {a}$], and explicitly specify this in the results and figures.

3. Results and discussion

3.1 Physical and biological oceanography during sampling

Sea surface temperature (SST) decreased with increasing latitude (Fig. 2) as expected [34], transitioning from warmer subtropical waters with a maximum of 22$\mathrm {^{\circ }}$C north of the STF in the Atlantic Ocean sector, to -2$\mathrm {^{\circ }}$C in Antarctic waters south of the SACCF. SST ranged from 5-16$\mathrm {^{\circ }}$C in the SAZ, 4-10$\mathrm {^{\circ }}$C in the PFZ, 2-4$\mathrm {^{\circ }}$C in the AZ and subzero south of the SACCF (Fig. 2), typical of Austral summer conditions [65].

The MLDs from in situ CTD casts and from climatological monthly means of de Boyer Montégut et al. [44] are presented in Fig. 2. Overall the trends in the in situ and climatology datasets agree well (See section Physical Oceanography). Surface mixed layers ranged from 15 to 100 m in situ and 20 to 80 m in the climatology. The MLDs are noticeably shallower in the Pacific sector of the Southern Ocean, typically < 40 m, where the ship predominately surveyed waters in the MIZ south of the SACCF. Increasing sea-ice meltwater in December is known to form shallow surface mixed layers of around 20 m in the Ross and Amundsen Seas [66], which can however deepen with changes in cross-shelf advection. The deepest MLDs were observed during the time the ship was near Kerguelen and Bouvetoya Islands (Fig. 2). At both islands, a combination of changes to the bathymetry and surface advection, which can also be influenced by nearby frontal boundaries, typically results in a deepening of the MLD relative to surrounding open waters [67–69]. A deepening of the MLD at Siple Island in the climatological database were not observed in the CTD temperature profiles, which remained shallow before and after Siple Island.

The [Tchla] concentrations along the ACE transit were generally low (< 0.4 $\textrm {mg m}^{-3}$; 75 % percentile), representative of the typical "High Nutrient Low Chlorophyll" (HNLC) conditions well reported in the Southern Ocean [70–72] (Fig. 1). The lowest values are similar to those typically observed in the most hyper-oligotrophic waters of the South Pacific [73]. The [Tchla] concentrations were highest at known biological ’hotspots’ within the Southern Ocean including near Subantarctic islands, Kerguelen [74,75], South Georgia-South Sandwich and Bouvet Islands [76], in the Scotia Sea [77,78], over the Patagonian shelf [79], near Mertz Glacier and within a known polynya in the MIZ of the Ross Sea and waters off the continental shelf leading to Siple Island [8]. In these locations, alleviation from iron-limitation is expected to occur due to shallow bathymetry, ice-melts and deposition from the islands [67,80,81].

3.2 Very high pheopigment concentrations at high latitude

In our dataset, pheopigments (pheophorbide $\textit {a}$ and pheophytin $\textit {a}$) make up a significant proportion of the total phytoplankton pigment concentrations, especially at latitudes in the MIZ (Fig. 3). At low latitudes 40-60$\mathrm {^{\circ }}$S, pheopigments were on average 10% of the [Tchl$\textit {a}$], but at latitudes higher than 60$\mathrm {^{\circ }}$S their contribution increased drastically (up to > 100 %, mean 27 % +/- 18 %; Fig. 3). Pheopigment concentrations in the surface waters of oligotrophic ocean regions (or hyperoligotrophic ocean regions) have typically not been reported at levels higher than 5 % of [Tchl$\textit {a}$] [53,54,82] (See mauve symbols in Fig. 3 from B04). The studies of Mendes et al. [83] and Wright et al. [84] which conducted North to South transects of the Indian Ocean sector of the SO both reported increasing pheopigment concentrations with increasing latitude. In Mendes et al. [83], surface concentrations of pheopigments were on average 2 % of [Tchl$\textit {a}$] in subtropical waters, increasing to 12 % in antarctic waters where a maximum of 20 % was observed. Overall Wright et al. [84] observed average integrated pheopigment concentrations at 15 % of integrated [Tchl$\textit {a}$]. Reynolds et al. [23] showed that pheopigment concentrations in surface waters of the PFZ and Ross Sea were strongly and linearly correlated with [Tchl$\textit {a}$]. The exact formation pathway of these chlorophyll a degradation pigments is unclear, although in the Southern Ocean, pheophytin $\textit {a}$ has been linked to increased grazing pressure [83,85] and pheophorbide $\textit {a}$ to either grazing, cell senescence or maybe from non-living particulate matter released from nearby melting sea-ice [84].

 

Fig. 3. Concentrations of [Pheo$\textit {a}$] as a function of [Tchl$\textit {a}$], colourmapped by latitude ($^{\textrm {o}}$S). Grey symbols indicate samples from NOMAD Southern Ocean subset. Mauve symbols are data from temperate oceans spanning a wide range of [Tchl$\textit {a}$] from Bricaud et al. [53] (B04; most points are much lower than 1 % and do not appear on the graph). The dashed lines represent where [Pheo$\textit {a}$] is 1 %, 10 % and 100 % of [Tchl$\textit {a}$].

Download Full Size | PDF

3.3 High fraction of microplankton compared to lower latitudes for the same trophic state

Microphytoplankton (largely diatoms) increasingly dominated the phytoplankton community composition at increasing latitude (Fig. 4), even at very low chlorophyll concentrations. In our data, we did not find an overall systematic trend between community size structure and chlorophyll concentration, except above 1 $\textrm {mg m}^{-3}$ where microphytoplankton made up most of the community (Fig. 4). It has previously been observed that a significant shift in phytoplankton species composition occurs at the PF which typically delineates waters of seasonal silicate limitation to the north [86]. Phytoplankton species composition in waters north of the PF, within the SAZ and PFZ, are typically more diverse, comprised of nanoeukaryotes, nanoflagellates including haptopytes, especially coccolithophores but also Phaeocystis sp., prasinophytes, euglenoids, cryptophytes, dinoflagellates (both autotrophic and heterotrophic), small lightly silicified diatoms, and cyanobacteria [36–38,87]. South of the PF however, diatom species typically dominate the composition with haptophytes, especially Phaeocystis sp., in high abundance [28,36,38,39]. Similar to Uitz et al. [10] data from the SO, the picophytoplankton fraction ($\textit {f}_{\textrm {pico}}$) in our dataset never exceeded 10 % of the total pigment biomass (Fig. 5), however, cyanobacteria Synechococcus sp. have been reported to contribute 20 % of cell counted biomass within Northern reaches of the SAZ (via flow cytometry) [87] and as such, their low contribution to the pigment biomass may simply be due to their low chlorophyll content [88]. In Uitz et al. [10], the authors suggest that the surface layer mixing regime may help explain the contribution of ($\textit {f}_{\textrm {micro}}$) to total pigment biomass. They found that in "mixed" waters of the Ross Sea (South of 60$\mathrm {^{\circ }}$S), i.e. where the euphotic depth was shallower than the mixed layer depth ($\textit {Z}_{\textrm {eu}}$/MLD < 1), microphytoplankton dominated at low chlorophyll concentrations, and increases in chlorophyll tended to be associated with nanophytoplankton, i.e. blooms of haptophytes such as Phaeocystis sp. Under stratified conditions ($\textit {Z}_{\textrm {eu}}$/MLD > 1) in the Ross Sea it is expected that microphytoplankton would dominate the pigment biomass based on previous observations [89–91], however, the Uitz et al. study did not include observations under those conditions. In other areas of the Southern Ocean, the dominant species or class of phytoplankton blooms varies between haptophyte species Phaeocystis sp., coccolithophores like Emiliania huxleyi, diatoms, and cryptophytes [92,93]. In our ACE dataset, the mixing regime alone did not explain much of the variability in microphytoplankton abundance (Fig. 4; $\textit {Z}_{\textrm {eu}}$/MLD indicated by marker size), and the $\textit {Z}_{\textrm {eu}}$/MLD was strongly dependent on [Tchl$\textit {a}$] with the mixed vs stratified regime boundary occurring roughly at 0.2-0.3 $\textrm {mg m}^{-3}$.

 

Fig. 4. Proportion of microphytoplankton ($\textit {f}_{\textrm {micro}}$) as a function of [Tchl$\textit {a}$] colourmapped by latitude ($^{\textrm {o}}$S). Datapoint size indicates the mixing regime experienced by phytoplankton, $\textit {Z}_{\textrm {eu}}$/MLD, see section Physical Oceanography. Datapoints where $\textit {Z}_{\textrm {eu}}$/MLD > 1 are classed as representing ’stratified’ conditions (indicated with black outer circles). Also presented on the plot are the $\textit {f}_{\textrm {micro}}$ vs [Tchl$\textit {a}$] relationships parameterised by Uitz et al. [10], see text for details. The black lines are the parameterisation for ’mixed waters’ in the ACE dataset using the same trophic categories as [10].

Download Full Size | PDF

 

Fig. 5. Proportion of picophytoplankton ($\textit {f}_{\textrm {pico}}$) as a function of [Tchl$\textit {a}$] colourmapped by latitude ($^{\circ }$S). Datapoint size indicates the mixing regime experienced by phytoplankton, $\textit {Z}_{\textrm {eu}}$/MLD, see section Physical Oceanography.

Download Full Size | PDF

3.4 Surface [Tchla] does not relate well to the dominant phytoplankton size class

In Uitz et al. [10], the authors presented a simple relationship between [Tchla] and $\textit {f}_{\textrm {micro}}$ for discrete trophic (chlorophyll a) categories or ranges as a means to infer community composition from the surface chlorophyll a, which are presented in Fig. 4 overlaying our ACE dataset. In addition to the relationship parameterised for well-mixed waters of the Ross Sea (Uitz "SO Mixed"), the relationships for well-mixed waters from 60$\mathrm {^{\circ }}$S to 60$\mathrm {^{\circ }}$N (Uitz "Other Mixed"), and stratified waters from 60$\mathrm {^{\circ }}$S to 60$\mathrm {^{\circ }}$N (Uitz "Other Stratified") are also included, noting though that in Uitz et al. [10] stratified waters were not observed in the Southern Ocean. Compared to the statistics compiled by Uitz et al. [10] from 60$\mathrm {^{\circ }}$S to 60$\mathrm {^{\circ }}$N, we observed much higher proportions of microphytoplankton in the phytoplankton pigment biomass within stratified waters, especially at low [Tchla]. However, the proportion of $\textit {f}_{\textrm {micro}}$ within each chlorophyll a range (see Uitz et al. [10] for the 7 stratified tropic categories) within stratified waters in the ACE dataset is highly variable, and cannot be parameterised as a function of [Tchla]. In our dataset, at [Tchla] between 0.2-0.8 $\textrm {mg m}^{-3}$ the average $\textit {f}_{\textrm {micro}}$ in mixed waters is close to that observed in the equivalent trophic categories in the Uitz et al. [10] SO mixed dataset, but also highly variable. At [Tchla] > 0.8, $\textit {f}_{\textrm {micro}}$ in our dataset continues to increase, contrary to the Uitz et al. [10] dataset. In fact at high [Tchla] > 0.8, $\textit {f}_{\textrm {micro}}$ is much higher than all of the statistics in Uitz et al. [10]. This may be a reflection of the greater importance of diatoms to Southern Ocean phytoplankton composition relative to other oceans. Interestingly the ACE samples with high [Tchla] and $\textit {f}_{\textrm {micro}}$ in "well-mixed" waters were sampled within the SAZ which is typically co-limited by iron and light [40,72]. These samples were predominately found on the Kerguelen Plateau (and also near Bouvetoya Island) where subsurface pools of dissolved iron on the Kerguelen Plateau within the upper 100 m were observed during ACE by Janssen et al. [94] and would be accessible to phytoplankton when waters were well mixed potentially alleviating iron limitation. However, the relationship between $\textit {f}_{\textrm {micro}}$ and [Tchla] in "well-mixed" waters would likely change with season where during winter for example mixed layer depths can reach hundreds of metres imposing chronic light limitation on phytoplankton growth [6,95]. The high proportion of $\textit {f}_{\textrm {micro}}$ in stratified waters south of the SACCF was unexpected, especially at such low chlorophyll a. These waters are often influenced by polynyas, featuring strong Phaeocystis sp. blooms [10,22,26], which in contrast we did not observe. We sampled the region south of the SACCF from late January to mid February and it is possible that we either observed a post-polynya bloom stage with species succession occuring from Phaeocystis sp to diatoms [5,8], were sampling too far off the continental shelf in waters up to 4000 m to observe the Phaeocystis sp dominated waters [8], or that the dissolved iron had been depleted below levels required by Phaeocystis sp, which can have a higher iron requirement than diatoms [96]. Similar to our study, the study by Hewes [97] in waters south of the ACC identified a low chlorophyll a scenario where phytoplankton in low-iron conditions in shoaled waters were dominated by microphytoplankton. Although $\textit {Z}_{\textrm {eu}}$/MLD was derived using a climatological MLD, this is not biasing the relationships between $\textit {f}_{\textrm {micro}}$ and [Tchla]. We observed the same trends in [Tchla] vs $\textit {f}_{\textrm {micro}}$ in a smaller sample set of n=28 using MLDs calculated from CTD density profiles, and corresponding pigment samples collected using niskin bottles on the same casts at surface depths < 10 m (data not shown). It is clear from this ACE dataset that the paradigm of phytoplankton size class increasing and decreasing proportionally with phytoplankton biomass (e.g., chlorophyll a concentration) [98,99], does not always apply to the SO, and as we show below has implications for the interpretation of bio-optical properties.

3.5 Particle volume increases with latitude for a given [Tchla]

The relationship between [Tchla] and particle volume varies between high and low latitude waters in the Southern Ocean. In latitudes south of 60$\mathrm {^{\circ }}$S, increases in [Tchla] are strongly associated with an increase in the total particle volume [Total particle volume=3.2×10$^{10}$[Tchla]$^{0.97}$, R$^{2}$ 0.76; see power law relationships in Fig. 6(a)] as well as an increase in the mean equivalent spherical particle volume [or size; Mean particle volume=141[Tchla]$^{0.42}$, R$^{2}$ 0.72; see section Particle Size Distribution and Fig. 6(b)]. At latitudes 50-60$\mathrm {^{\circ }}$S, total particle volume is less strongly related to increases in [Tchla] (Total particle volume=4.1×10$^{9}$[Tchla]$^{0.53}$, R$^{2}$ 0.43). The mean particle volume also appears to be less variable and overall smaller in size [Mean particle volume=60.9[Tchla]$^{0.33}$, R$^{2}$ 0.33; Fig. 6(b)]. At lower latitudes 40-50$\mathrm {^{\circ }}$S total particle volume and mean particle volume show no apparent increase with [Tchla] [R$^{2}$ of both relationships close to 0 and p>0.05; see Figs. 6(a) and 6(b)]. It is however difficult to fully interpret these trends as non-phytoplankton particles are also counted here and as shown earlier, pheopigment concentrations, and hence degraded or detrital particles are important components of the particulate matter. It is possible though that in the northern reaches of the SAZ, changes in chlorophyll a were manifested as adjustments in chlorophyll a per cell which could be an acclimation response to low dissolved iron concentrations or photoacclimation to changes in the mixing layer depth [100–102].

 

Fig. 6. (a) The total particle volume ($\mu \textrm {m}$ $\textrm {m}^{-3}$) of samples as a function of [Tchl$\textit {a}$]. (b) The mean (equivalent spherical) particle volume ($\mu \textrm {m}^{3}$) as a function of [Tchl$\textit {a}$]. Data points in a) and b) are colour mapped by latitude ($^{\circ }$S).

Download Full Size | PDF

3.6 Particulate absorption is tightly linked to [Tchla] with significant background of non-algal particles

The particulate absorption coefficient at 440 nm ($\textit {a}_{\textrm {p}}$(440), $\textrm {m}^{-1}$) in the ACE and NOMAD datasets increased with [Tchl$\textit {a}$] in a similar manner to B04 [53] and B98 [64] [Fig. 7(a); Table 1], but was systematically lower than B04 [53] at [Tchl$\textit {a}$] > 0.1 mg m$^{\textrm {-3}}$. Overall there was little variance in the relationship at any given [Tchl$\textit {a}$]. The coefficient and exponent for the $\textit {a}_{\textrm {p}}$(440) vs. [Tchl$\textit {a}$] was $\textit {a}_{\textrm {p}}$(440)=0.043[Tchl$\textit {a}$]$^{\textrm {0.61}}$ ($\textit {R}^{\textrm {2}}$=0.89; Table 1). The slope is similar to Reynolds et al. [23] for data for the Ross Sea (0.041) and PFZ (0.047), and the NOMAD dataset (0.044) from Drake’s Passage, Scotia Sea and Antarctic Peninsula waters. It is also in agreement with previous observations that chlorophyll-specific particulate absorption $\textit {a}^{\textrm {*}}_{\textrm {p}}$ is generally lower in the Southern Ocean [18,23] than values reported for other oceans. Lower slopes in the relationship between $\textit {a}_{\textrm {p}}$(440) and [Tchl$\textit {a}$] imply that a lower increase in absorption per unit [Tchl$\textit {a}$], or lower $\textit {a}^{\textrm {*}}_{\textrm {p}}$. Lower values of $\textit {a}^{\textrm {*}}_{\textrm {p}}$ per unit chlorophyll a may result from smaller contributions to particulate absorption by $\textit {a}_{\textrm {nap}}$, reduced contributions by pigments other than chlorophyll a to phytoplankton absorption, or variability in absorption per unit chlorophyll a resulting from pigment packaging effects [103]. An interesting feature in Fig. 7(a) is the change of slope of $\textit {a}_{\textrm {p}}$(440) at low [Tchl$\textit {a}$] (at around $\textit {a}_{\textrm {p}}$(440) = 0.006 $\textrm {m}^{-1}$ and 0.1 mg m$^{-3}$ [Tchla]) that may arise from an increased proportions of non-algal particles (e.g., heterotrophic bacteria, detritus, viruses, minerogenic particles; see also Fig. 16). This is reminiscent of the relationship between particulate backscattering and chlorophyll a seen in Brewin et al. [104] which was attributed to an increased background of non-algal particles and the observation by Zhang et al. [105] of a significant background signal from sub-micron particles in the particulate backscattering in low chlorophyll a waters of the North Pacific which become a more important part of the signal at low chlorophyll a compared to at high chlorophyll a.

 

Fig. 7. Variations in (a) particulate absorption [$\textit {a}_{\textrm {p}}$(440)] and (b) phytoplankton absorption ($\textit {a}_{\mathrm {\phi }}(440)$) at 440 nm as a function of [Tchl$\textit {a}$]. Purple line indicates parameterised ACE data relationships. Magenta line indicates relationship from the NOMAD dataset. Previously parameterised relationships of B98 [64] (solid grey line), B04 [53] (dashed grey line), B10 [54] (solid red line), B10+B04 [53,54] combined datasets (our fit to combined original datasets; dashed red line), Reynolds et al. [23] in the APFZ (solid blue line) and Ross Sea (dashed blue line) are also plotted.

Download Full Size | PDF

Table 1. Constants and Exponents From the Power Law Relationships Between Chlorophyll a (either [Tchla] or [Tchla+Pheoa]) and a) $\textit {a}_{\textrm {p}}$(440), b) $\textit {a}_{\mathrm {\phi }}$(440), or c) $\textit {a}_{\mathrm {\phi }}$(676) from the ACE dataset (ACE), B98 [64], B04 [53], B10 [54], B04+B10 [53,54] Combined Datasets (our fit to combined original datasets; dashed red line), and Reynolds et al. [23] in the APFZ (RAPFZ) and Ross Sea (RRS).^a

View Table

3.7 Phytoplankton absorption is lower than in other oceans for the same [Tchl$\textit {a}$]

The phytoplankton absorption coefficient at 440 nm [$\textit {a}_{\mathrm {\phi }}$(440)] as a function of [Tchl$\textit {a}$] [Fig. 7(b); Table 1] reflects the absorption by all photosynthetic and non-photosynthetic pigments in this waveband. It was 50% lower than the relationship derived by B04 [53] for the first optical depth in tropical and temperate oceans and seas (ACE data $\textit {a}_{\mathrm {\phi }}$(440)=0.036[Tchl$\textit {a}$]$^{\textrm {0.82}}$ $\textit {R}^{\textrm {2}}$=0.91). At [Tchl$\textit {a}$] greater than 0.1 $\textrm {mg m}^{-3}$ that ACE values were more similar to the NOMAD dataset and at >0.3 $\textrm {mg m}^{-3}$ the ACE values $\textit {a}_{\mathrm {\phi }}$(440) are more similar to the B98 [64] dataset, which included data from all depths up to 200 m (i.e., greater than 1 optical depth) where samples are strongly impacted by pigment packaging and hence substantially lower than the B04 [53] coefficients. As the exponent for the ACE data suggests, the relationship is non-linear, and at [Tchl$\textit {a}$] below 0.3 $\textrm {mg m}^{-3}$ $\textit {a}_{\mathrm {\phi }}$(440) is even lower and were more similar (though still lower) to those measured in the ultra oligotrophic waters of the south-east Pacific as reported in B10 [54].

Absorption by phytoplankton at 676 nm is predominantly due to chlorophyll a pigments, and so changes to the chlorophyll-specific absorption coefficient at 676 nm are generally mostly due to changes in pigment packaging only, rather than adjustments in carotenoid pigments from changes to the community composition and/or the photophysiological (pigment) response. The chlorophyll-specific absorption coefficient at 676 nm ($\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676)) normalised to [Tchl$\textit {a}$] was on average 0.020 $\textrm {m}^{\textrm {2}} \textrm { mg}^{\textrm {-1}}$ [Fig. 8(a); Table 1], similar to the in vivo specific absorption coefficient for unpackaged chlorophyll a of Bricaud et al. [82] (0.0207 $\textrm {m}^{2}$ $\textrm {mg}^{-1}$), Bidigare et al. [62] (0.0199 $\textrm {m}^{2}$ $\textrm {mg}^{-1}$) and Clementson and Wojtasiewicz [59] (0.0196 $\textrm {m}^{2}$ $\textrm {mg}^{-1}$). However, $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676) declined with [Tchl$\textit {a}$] ($\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676)=0.020[Tchl$\textit {a}$]$^{0.82-1}$) [and with increasing cell volume, see Fig. 8(b)], and was substantially lower than the B04 [53] and B04+B10 [54] relationships at Tchl$\textit {a}$] > 0.1, suggesting that this ACE dataset is more packaged than observed in other temperate and tropical oceans. The $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676) was most similar to the NOMAD data from the SO, and B98 [64] data from temperate and tropical oceans which has also been described as being impacted by packaging. Comparing Figs. 8(a) and 8(b) (which has been normalised to [Tchl$\textit {a}$ + Pheo$\textit {a}$]), there is noticeable scatter of $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676) in Fig. 8(a) that is absent in Fig. 8(b), which is attributable to increased absorption by pheopigments in samples with high [Pheo$\textit {a}$/Tchl$\textit {a}$] at high latitudes [see colourmap of Fig. 8(a)]. Although the overall relationship of $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676) (normalised to [Pheo$\textit {a}$/Tchl$\textit {a}$]) vs [Pheo$\textit {a}$/Tchl$\textit {a}$] is not different to the relationship above, $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676)=0.018[Tchl$\textit {a}$ + Pheo$\textit {a}$]$^{0.87-1}$ [CE relationship from Fig. 8(b); Table 1], it is clear that on a case by case basis, pheopigments may contribute substantially to $\textit {a}_{\mathrm {\phi }}$(676) and cannot be discounted as being an important part of the optical constituents in the Southern Ocean, especially at high latitudes. As seen in Fig. 9(a), $\textit {Q}^{\textrm {*}}_{\textrm {a}}$(676) coefficient decreases in value with increasing [Tchl$\textit {a}$] and increasing cell volume, confirming increased pigment packaging with increasing [Tchl$\textit {a}$] due to increasing cell size and increased intracellular chlorophyll content. By definition, $\textit {Q}^{\textrm {*}}_{\textrm {a}}$ should range from 1 to 0, and values > 1 as seen in Fig. 9(a) indicate the phytoplankton absorption is greater than that of the same pigment material in solution [103]. Other studies have observed $\textit {Q}^{\textrm {*}}_{\textrm {a}}$ > 1 [53,106–108] and have attributed this to missing pigments (e.g., phycobilins) or other light absorbing compounds not detectable by HPLC [109], cellular components and structure such as thylakoid stacking which contribute to $\textit {a}_{\mathrm {\phi }}$ [108], differences in in vivo absorption vs absorption by pigments in solution which may be due to pigment complex construction or is unexplained [110], and/or natural or species-specific variability in the weight-specific pigment absorption coefficients [110]. It is not clear which explanation is responsible for the discrepancies in this dataset, and as mentioned in the section Pigment and Packaging Indices, the $\textit {a}_{\textrm {miss}}$ term solution of Bricaud et al. [53] did not seem to be correct for our dataset. Nevertheless, the general trends in $\textit {Q}^{\textrm {*}}_{\textrm {a}}$ (regardless of the absolute magnitude) as a function of [Tchla] and cell volume are as expected.

 

Fig. 8. Variations in chlorophyll-specific a phytoplankton absorption at 676 nm ($\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(676)$). (a) $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(676)$ is normalised to and plotted as a function of [Tchl$\textit {a}$], and colour mapped by [Pheo$\textit {a}$/Tchl$\textit {a}$]. (b) $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(676)$ is normalised to and plotted as a function of [Tchl$\textit {a}$ + Pheo$\textit {a}$] as a function of [Tchl$\textit {a}$ + Pheo$\textit {a}$], and colourmapped by mean particle volume ($\mu$m$^3$). Purple line is the ACE dataset parameterisation. Magenta line is the NOMAD dataset parameterisation. Previously parameterised relationships of B98 [64] (solid grey line), B04 [53] (dashed grey line), and B04+B10 [53,54] combined datasets (our fit to combined original datasets; dashed red line) are also plotted.

Download Full Size | PDF

 

Fig. 9. Packaging indices calculated at (a) 440 nm ($\textit {Q}^{\textrm {*}}_{\textit {a}}(\textrm {440})$) and (b) 676 nm ($\textit {Q}^{\textrm {*}}_{\textit {a}}(\textrm {676})$) using the pigment reconstruction method, plotted as a function of [Tchl$\textit {a}$]. Both (a) and (b) are colourmapped by mean particle volume ($\mu$m$^{3}$).

Download Full Size | PDF

Very low $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) at chlorophyll a concentrations below 0.3 $\textrm {mg m}^{-3}$ were observed compared to previously described fits from temperate and tropical oceans [Fig. 10(a)] and also the SO (i.e., NOMAD). However, in the ACE dataset, although values of $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) measured on samples with > 0.3 $\textrm {mg m}^{-3}$ chlorophyll a still depart strongly from global surface waters values of B04 and B04+10 [53,54,64] they more closely resembles values that include deeper waters from B98 [64] and NOMAD SO data. Chlorophyll-specific absorption at 440 nm is typically lower in microphytoplankton [31]. This is clearly illustrated in this dataset which is dominated by microphytoplankton (Fig. 5). The $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) is also impacted by changes in the concentration of accessory pigments which overlap with chlorophyll a absorption at 440 nm. According to $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) vs. [Tchl$\textit {a}$] relationships parameterised in non-polar waters, $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) is typically much higher at low chlorophyll accounting for the predominance of picophytoplankton when chlorophyll is low in temperate and tropical waters. Pigment packaging at 440 nm does appear to explain the trends in $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) within the ACE dataset, similar to $\textit {Q}^{\textrm {*}}_{\textrm {a}}$(676), $\textit {Q}^{\textrm {*}}_{\textrm {a}}$(440) also decreases with [Tchl$\textit {a}$] and cell volume [Fig. 9(b)], as has been observed by others [18]. However, it does not appear to fully explain why $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) is so much lower than observations in temperate and tropical oceans and varies so little with [Tchl$\textit {a}$] compared to previously described trendlines. Unlike at 676 nm, pheopigments do not contribute substantially to absorption at 440 nm (Fig. S1 in Supplement 1), and so there is a smaller difference between Figs. 10(a) and 10(b), where Fig. 10(b) has been normalised to [Tchl$\textit {a}$ + Pheo$\textit {a}$].

 

Fig. 10. Variations in chlorophyll-specific a phytoplankton absorption at 440 nm ($\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(440)$). (a) $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(440)$ is normalised to and plotted as a function of [Tchl$\textit {a}$], and colourmapped by [Pheo$\textit {a}$/Tchl$\textit {a}$]. (b) $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(440)$ is normalised to and plotted as a function of [Tchl$\textit {a}$ + Pheo$\textit {a}$] as a function of [Tchl$\textit {a}$ + Pheo$\textit {a}$], and colourmapped by $\textit {f}_{\textrm {micro}}$. Purple line is the ACE dataset parameterisation. Magenta line is the NOMAD dataset parameterisation. Previously parameterised relationships of B98 [64] (solid grey line), B04 [53] (dashed grey line), and B04+B10 [53,54] combined datasets (our fit to combined original datasets; dashed red line) are also plotted.

Download Full Size | PDF

Figures 11(a) and 11(b) show $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) and $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676), respectively, as a function of the median light intensity within the mixed layer ($\textit {I}_{\textrm {g}}$). Overall, there is very little change in the chlorophyll-specific absorption coefficient at 440 or 676 with increases in $\textit {I}_{\textrm {g}}$ except at $\textit {I}_{\textrm {g}}$ > 200. This suggests that variation in $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) and $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(676) in this dataset are due to changes in cell size and pigmentation driven by changes in taxa. It is noted though that when surveying bio-optical properties in surface waters, i.e., across the horizontal dimension, the dominant source of variation in $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$ tends to be changes in the phytoplankton size structure and pigment composition from changes in taxa, as opposed to the vertical dimension where photoacclimation become more important [111].

 

Fig. 11. Chlorophyll-specific a phytoplankton absorption at (a) 440 nm ($\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(440)$) and (b) 676 nm ($\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(676)$) as a function of the median light intensity within the mixed layer ($\textit {I}_{\textrm {g}}$), colourmapped by latitude ($^{\circ }$S). Note the different scale in (b).

Download Full Size | PDF

3.8 High photosynthetic and [Tchlc] together with low photoprotective and [Tchlb] pigments complement characterise the Southern Ocean in summer

Bricaud and coauthors [54] (B10) surveyed the ultra-oligotrophic waters of the South-Pacific gyre and found that for low chlorophyll samples $\textit {a}_{\mathrm {\phi }}$(440) (and hence $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440)) were 30%-40% lower than in other oceanic areas with the same chlorophyll. They attributed this to both a decrease in the concentration of accessory pigments, specifically photoprotective carotenoids, relative to chlorophyll a ([PPC/Tchla], and an increase in relative cell size. In the ACE dataset, the relative contribution of two of the major pigment groups, [Tchlb] [Fig. 12(a)] and [PPC] [Fig. 12(d)] are generally lower (compare mauve and ACE points in Fig. 12) for the same [Tchl$\textit {a}$] than in other oceans. In addition to the lower growth irradiance compared to other oceans, this likely due to the lower contribution of picophytoplankton to the Southern Ocean pigment biomass specifically chlorophytes and cyanobacteria which contain significant concentrations of chlorophyll b, zeaxanthin, lutein and violaxanthin. High chlorophyll c and [PSC] concentrations are due to the dominance of diatoms, haptophytes and dinoflagellates [Fig. 12(b)].

 

Fig. 12. Ratios of major accessory and degradation pigments to [Tchl$\textit {a}$], including (a) [Tchl$\textit {b}$], (b) [Tchl$\textit {c}$], c) [PSC], d) [PPC]. Grey symbols indicate samples from NOMAD Southern Ocean subset. Mauve symbols are data from Bricaud et al. [53]. In (b) we note the highly variable and sometimes very high [Tchl$\textit {c}$] / [Tchl$\textit {a}$] in the NOMAD dataset which may be artefactual.

Download Full Size | PDF

It should be noted that in the Southern Ocean literature, the relationship between $\textit {a}_{\mathrm {\phi }}$ and [Tchl$\textit {a}$] varies widely. For example $\textit {a}_{\mathrm {\phi }}$(440) vs. [Tchl$\textit {a}$] measured by Clementson et al. [20] in the Indian Ocean Sector of the SAZ in late summer were similar to the relationship of B04 [53]. In other studies, considerable variation has been observed within a single study, for example Kerkar et al. [25] also observed the Indian Ocean sector of the SAZ in summer and reported chlorophyll-specific $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$(440) ranging from 0.04-0.5 for fluorometrically derived chlorophyll a between 0.11-0.26. We have also observed a large variation in $\textit {a}^{*}_{\mathrm {\phi }}$(440 nm) with very little change in [Tchl$\textit {a}$] in the Atlantic and Indian Ocean sectors during the austral winter (data unpublished). Although it is acknowledged that methodological differences in the measurement (fluorometric vs. HPLC) and definition of chlorophyll a, application of integrating spheres for measuring optical density, pathlength amplification factors for calculating $\textit {a}_{\textrm {p}}$ and method for decomposing $\textit {a}_{\textrm {p}}$ and $\textit {a}_{\mathrm {\phi }}$ add uncertainties and variability to these measurements and relationships. We have limited these potential discrepancies in our comparison with the Bricaud datasets by: (1) running the HPLC samples in the same laboratory, (2) using both the [Tchl$\textit {a}$] and [Tchl$\textit {a}$ + Pheo$\textit {a}$] normalization approach, (3) verifying that the decomposition approach provided consistent results at 440 and 676 nm with the Bricaud and Stramski [112] approach used in some of campaigns of the Bricaud datasets. Both the inversion method presented here, and the Bricaud & Stramski approach produced reliable estimates of $\textit {a}_{\mathrm {\phi }}$ as compared to a small number of values from ACE derived using the traditional methanol extraction method of Kishino et al. [113]. For absorption measurements, some of the difference could arise from the measurement methods used such as using an integrating sphere with associated path-length amplification approach in our case. The use of an integrating sphere reduces the measurement error as more scattered photons are captured in the detector than other techniques (e.g., transmittance and transmittance-reflectance) [114] and path-length amplification factors for the integrating sphere technique have lower uncertainties than the other techniques [114,115]. However, as more scattered photons are captured by the detector, rather than be attributed to absorption, $\textit {a}_{\textrm {p}}$ could theoretically be lower when using this method as compared to others.

3.9 Flatter absorption spectrum and lower $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$ across the spectrum compared to other oceanic regions

The low chlorophyll-specific absorption coefficients are associated with a significant flattening of the phytoplankton absorption spectral shape as observed in the ratio $\textit {a}_{\mathrm {\phi }}$(440)/$\textit {a}_{\mathrm {\phi }}$(676) (Fig. 13). In other oceanic regions $\textit {a}_{\mathrm {\phi }}$(440)/$\textit {a}_{\mathrm {\phi }}$(676) increases strongly with decreasing chlorophyll a as pigment packaging decreases due to decreasing cell size, and picophytoplankton dominate (See B98, B04, B04+10 and NOMAD relationships in Fig. 13). As we show, such changes in the phytoplankton size distribution are not present in the same way in the Southern Ocean (Figs. 4 and 5). While there is a slight increase in the $\textit {a}_{\mathrm {\phi }}$(440)/$\textit {a}_{\mathrm {\phi }}$(676) with decreasing [Tchl$\textit {a}$] at lower latitudes, this is not observed at higher latitudes (Fig. 13). To represent these unique relationships, different spectral models are needed. Power law functions between log-transformed $\textrm {a}_{\phi }(\lambda )$ and log-transformed [Tchl$\textit {a}$] were fit at 2 nm wavelength intervals (see Data File 1, Data File 2, and Data File 3 and Supplement 1 for coefficients). The resulting phytoplankton absorption spectral shapes normalized to $\textit {a}_{\mathrm {\phi }}$(440) (Fig. 14), and their chlorophyll-specific equivalents (Fig. 15), show very different shapes and amplitudes compared to other oceanic regions. The equivalent shapes from the combined datasets of B04 [53] and B10 [54] are in mauve in Figs. 14(a) and Fig. 15(a). Overall the spectral shapes are noticeably different from the B04+B10 shapes, but differ again between low and high latitudes [Figs. 14(b), 14(c), 15(b), and 15(c)]. The shape and magnitude of the absorption spectra for phytoplankton reflect the pigment composition and concentration, which are in turn dictated by the species composition, size and environment. In addition to the higher values above $\sim {500}$ nm reflecting the flatter, more packaged spectra, the most prominent feature in the ACE spectral shapes are a shoulder at 460 nm and between 550-600 nm which intensify at low chlorophyll a concentrations. The spectral peak at 460 nm is likely due to increased chlorophyll c, a diagnostic pigment marker for diatoms. Similar spectral features between 550-600 nm have been observed in haptophytes (e.g., Phaeocystis sp.) and nanoeukaryotes of the ultraplankton (2-5 $\mu$m) or nanophytoplankton (5-20 $\mu$m) size [28,31]. Also different to the average global spectral shapes is the variability at 676 nm. In the high latitude samples [Figs. 14(c) and 15(c)] the spectral shapes are more similar to a microphytoplankton shape across all chlorophyll a concentrations [31], including features at 460 nm (where absorption at 460 nm is lower than 440 nm) and 500 nm due to chlorophyll c pigments and fucoxanthin pigments. There is also very little change in the $\textit {a}_{\mathrm {\phi }}$(440)/$\textit {a}_{\mathrm {\phi }}$(676) with chlorophyll a concentration. Whereas, in the low latitude samples, there appears to be a more significant taxonomic shift from high chlorophyll a to low chlorophyll a concentrations. These differences are particularly important when examining the chlorophyll-specific absorption. In most oceanic waters [see mauve lines in Fig. 15(a)], the chlorophyll-specific absorption does not change much above $\sim {540}$ nm when chlorophyll a concentration changes, while the shorter wavelengths increase rapidly with decreasing chlorophyll a concentration reaching as high as $\sim {0.16}$ $\textrm {m}^{2}$ $\textrm {mg}^{-1}$ at $\sim {450}$ nm. In the Southern Ocean, $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$ increases with decreasing chlorophyll a concentration at all wavelength, but much less than in other oceans, reaching $\sim {0.08}$ $\textrm {m}^{2}$ $\textrm {mg}^{-1}$ in the low latitude waters and about half of that in higher latitudes. Similarly, at 3 $\textrm {mg m}^{-3}$ of chlorophyll a the $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$ is about half of the $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}$ in other oceanic regions. The implication of these changes are important. From an ecological perspective, phytoplankton in the Southern Ocean exhibit higher pigment packaging implying that they must invest more energy into pigment synthesis. From a remote sensing perspective, the absorption in the blue band specific to phytoplankton is significantly lower (a half to a quarter) when compared to other oceanic regions.

 

Fig. 13. Ratio of $\textit {a}_{\mathrm {\phi }}$(440)/$\textit {a}_{\mathrm {\phi }}$(676) as a function of [Tchl$\textit {a}$]. Previously parameterised relationships of B98 [64] (solid grey line), B04 [53] (dashed grey line), and B10 [54] (solid red line) and B04+B10 [53,54] combined datasets (dashed red line) are also plotted. The relationship from the NOMAD dataset is indicated by the magenta line.

Download Full Size | PDF

 

Fig. 14. Phytoplankton absorption $\textit {a}_{\mathrm {\phi }}(\lambda )$ spectra normalised to $\textit {a}_{\mathrm {\phi }}$(440) for various values of [Tchl$\textit {a}$] from the ACE dataset. The $\textit {a}_{\mathrm {\phi }}(\lambda )$ have been reconstructed using constants and exponents from power law functions fitted between log-transformed $\textit {a}_{\mathrm {\phi }}(\lambda )$ and log-transformed [Tchl$\textit {a}$] at for each wavelength 400-700 nm at 2 nm wavelength intervals. The functions were fitted for (a) all latitudes, (b) low latitudes between 40 to 60$\mathrm {^{\circ }}$S, and (c) high latitudes southwards of 60$\mathrm {^{\circ }}$S. The $\textit {a}_{\mathrm {\phi }}(\lambda )$/$\textit {a}_{\mathrm {\phi }}$(440) from the power law functions of the combined B04+B10 [53,54] datasets are presented in (a) as mauve lines.

Download Full Size | PDF

 

Fig. 15. Chlorophyll-specific phytoplankton absorption $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(\lambda )$ spectra for various values of [Tchl$\textit {a}$] from the ACE dataset. The $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(\lambda )$ have been reconstructed using constants and exponents from power law functions fitted between log-transformed $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(\lambda )$ and log-transformed [Tchl$\textit {a}$] at for each wavelength 400-700 nm at 2 nm wavelength intervals. The functions were fitted for (a) all latitudes, (b) low latitudes between 40 to 60$\mathrm {^{\circ }}$S, and (c) high latitudes southwards of 60$\mathrm {^{\circ }}$S. The $\textit {a}^{\textrm {*}}_{\mathrm {\phi }}(\lambda )$ from the power law functions of the combined B04+B10 [53,54] datasets are presented in a) as mauve lines.

Download Full Size | PDF

3.10 Southern Ocean has high non-algal particulate absorption compared to other oceanic regions

Given the difference in relationship between $\textit {a}_{\mathrm {\phi }}$(440) and [Tchl$\textit {a}$] [Fig. 7(b)], relative to changes in $\textit {a}_{\textrm {p}}$(440) and [Tchl$\textit {a}$] [Fig. 7(a)] (particularly at low chlorophyll a), it is clear that the contribution of non-algal particulates ($\textit {a}_{\textrm {nap}}$(440)) to the total particulate absorption budget must also deviate from expected relationships. Previous reports suggest that $\textit {a}_{\textrm {nap}}$ contributes little to total particulate absorption in the Southern Ocean, except at locations close to landmasses or ice-shelves and is strongly related to chlorophyll a [18,20,23]. In contrast we find that $\textit {a}_{\textrm {nap}}$ can be responsible for up to 85% of particulate absorption, varying loosely in an inverse manner with phytoplankton biomass [Tchl$\textit {a}$] (Fig. 16). Overall the contribution is considerably higher than observed in temperate oceans at comparable chlorophyll a (See data from B04 [53] in Fig. 16). This is especially true in high latitude waters where increased pheopigment concentrations (Figs. 3 and 16) suggests that phytoplankton biomass is increasingly degraded and redistributed to the non-algal particulate pool.

 

Fig. 16. Ratio of non-algal particulate absorption to total particulate absorption at 440 nm ($\textit {a}_{\textrm {nap}}$(440)/$\textit {a}_{\textrm {p}}$(440)) as a function of [Tchl$\textit {a}$]. Datapoints are colourmapped by latitude. Data from B04 [53] are mauve points and from the NOMAD dataset are grey points.

Download Full Size | PDF

4. Conclusion

Southern Ocean waters are optically unique [116], although much of the in situ bio-optical data collected in the SO so far has been spatially constrained. The ACE voyage gave us the opportunity to collect a unique dataset of phytoplankton particulate absorption, pigments and particle size across the entire longitudinal and latitudinal range of the SO. From this dataset, we confirm strikingly different relationships between chlorophyll-specific absorption and community composition and size when compared to other oceans (particularly in low wavelengths and at low chlorophyll a concentrations). In particular, we find high fractions of microplankton across all chlorophyll a concentrations and as a result, the phytoplankton absorption spectrum is flatter than other oceans at all chlorophyll a, and overall phytoplankton absorption and chlorophyll-specific absorption is lower than other oceans. Furthermore, we find considerable differences in the bio-optical relationships of high and low latitude phytoplankton in the SO. It is apparent from our work that interpreting ocean colour or other bio-optical signals in the SO with paradigms and parameterisations from lower latitude temperate oceans, such as the phytoplankton absorption spectral shape or composition of the phytoplankton community as a function of chlorophyll a can lead to incorrect conclusions.

In terms of optical budgets, this study focused on absorption terms and only measured particulate properties. It can be assumed that if chlorophyll-specific particulate absorption is lower in the Southern Ocean than other ocean basins, and is variable with latitude, and yet that satellite chlorophyll retrievals appear to be within acceptable uncertainties [117,118], then there may be some deviations in the chlorophyll-specific and particle-specific backscattering relationships, which has implications for the potential accurate satellite retrievals of particle size [119]. Additionally, patterns in $\textit {a}_{\textrm {nap}}$ show that high latitude locations deviate from a ’Case 1’ waters optical scenario of $\textit {a}_{\textrm {nap}}$ varying strongly with chlorophyll a, and so it can be expected that there is also considerable spatial variability in absorption by coloured dissolved organic matter. We also show that pheopigments contribute substantially to the pigment biomass, especially at high latitudes, and may also contribute to particulate absorption at 676 nm.

It is clear from our work that relationships between phytoplankton pigments, particulate absorption and particle properties vary considerably within the SO, but can be parameterised with continued in situ observations across different biomes. We would also expect considerable intra- and inter-seasonal variability in these relationships as the phytoplankton composition and size responds to changes in mixing layer depth, light availability and access to dissolved iron and silicic acid, but capturing those changes in situ remains a logistical challenge.