predicting the electron requirement for carbon …fixation requires knowledge of the electron...

OPEN 3 ACCESS Freely available online © PLOSI o -

Predicting the Electron Requirement for Carbon Fixation in Seas and OceansEvelyn Lawrenz1*, Greg Silsbe2, Elisa Capuzzo3, Pasi Ylöstalo4, Rodney M. Forster3, Stefan G. H. Simis4, Ondrej Prásil1, Jacco C. Kromkamp2, Anna E. Hickman5, C. Mark Moore5, Marie-Hélèn Forget6, Richard J. Geider7, David J. Suggett71 Laboratory o f Photosynthesis, Institute o f Microbiology, ASCR (Academy o f Sciences o f the Czech Republic), Opatovickÿ rnlyn, Treboñ, Czech Republic, 2 Netherlands Institute o f Ecology, Centre fo r Estuarine and Marine Ecology (NIOO-CEME), Yerseke, The Netherlands, 3 Centre for Environment, Fisheries & Aquaculture Science (CEFAS), Lowestoft, Suffolk, United Kingdom, 4 Finish Environment Institute (SYKE), Helsinki, Finland, 5 Ocean and Earth Science, University o f Southampton, National Oceanography Centre, Southampton, Hampshire, United Kingdom, 6 Takuvic Joint International Laboratory, UMI 3376, Université Laval (Canada) CNRS (France), Department de Biologie and Québec-Océan, Université Laval, Québec City, Québec, Canada, 7 School o f Biological Sciences, University o f Essex, Colchester, Essex, United Kingdom

AbstractMarine phytoplankton account for about 50% o f all global net primary productivity (NPP). Active fluorom etry, mainly Fast Repetition Rate fluorom etry (FRRf), has been advocated as means o f providing high resolution estimates o f NPP. Plowever, not measuring C02-fixation directly, FRRf instead provides photosynthetic quantum efficiency estimates from which electron transfer rates (ETR) and ultim ately C02-fixation rates can be derived. Consequently, conversions o f ETRs to C02- fixation requires knowledge o f the electron requirement for carbon fixation (i>Siû ETR/C02 uptake rate) and its dependence on environmental gradients. Such knowledge is critical for large scale im plem entation o f active fluorescence to better characterise C 02-uptake. Flere we examine the variability o f experimentally determ ined values in relation to key environmental variables w ith the aim o f developing new working algorithms for the calculation o f @eC from environmental variables. Coincident FRRf and 14C-uptake and environmental data from 14 studies covering 12 marine regions were analysed via a meta-analytical, non-parametric, multivariate approach. Combining all studies, @eC varied between 1.15 and 54.2 mol e- (mol C)-1 w ith a mean o f 10.9±6.91 mol e- mol C)-1 . A lthough variability o f was related to environmental gradients at global scales, region-specific analyses provided far Improved predictive capability. Plowever, use o f regional 0>e_c algorithms requires objective means o f defining regions o f Interest, which remains challenging. Considering Individual studies and specific small-scale regions, temperature, nutrient and light availability were correlated w ith 0>eC albeit to varying degrees and depending on the study/region and the com position o f the extant phytoplankton community. At the level o f large blogeographlc regions and distinct water masses, 0>eC was related to nutrient availability, chlorophyll, as well as temperature and/or salinity In most regions, while light availability was also Im portant In Baltic Sea and shelf waters. The novel 0>e_c algorithms provide a major step forward for widespread fluorometry-based NPP estimates and h igh light the need for further studying the natural variability o f to verify and develop algorithms w ith Improved accuracy.

Citation: Lawrenz E, Silsbe G, Capuzzo E, Ylöstalo P, Forster RM, et al. (2013) Predicting the Electron Requirement for Carbon Fixation in Seas and Oceans. PLoS ONE 8(3): e58137. doi:10.1371/journal.pone.0058137

Editor: Lucas J. Stal, Royal Netherlands Institute o f Sea Research (NIOZ), The Netherlands

Received October 13, 2012; Accepted January 30, 2013; Published March 13, 2013

Copyright: © 2013 Lawrenz et al. This is an open-access article distributed under the terms o f the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This work and some o f the cruises have been made possible via funding through the EU FP7 project PROTOOL FP7 (grant number 226880). The funders had no role in study design, data collection and analysis, decision to publish, or preparation o f the manuscript.

Com peting Interests: The authors have declared that no competing interests exist.

* E-mail: [email protected]

Introduction

Accurately evaluating the im pact o f local environm ental and global clim ate change upon troph ic dynamics and biogeochemical nu trien t cycling is fundam entally tied to how well p rim ary productiv ity , defined here as carbon (C 0 2) fixation, is characterised. Follow ing the incorpora tion o f inorganic radio-labelled

C 0 2 in to algal cells has become a standard m ethod for quantify ing p rim a ry productiv ity . However, to this day, there still exists considerable uncerta inty as to w hether 1 4 C 0 2 uptake measures gross p rim a ry p roductiv ity (GPP) o r net p rim ary p roductiv ity (NPP). Follow ing the trad itiona l view, GPP refers to carbon fixa tion w ithou t accounting fo r any carbon losses due to respiration a n d /o r excretion, w h ile NPP represents the carbon

uptake rate after subtracting out any C 0 2 lost to oxidation o f organic carbon over a die l cycle [ 1 ],

O f all global NPP, m arine ecosystems account fo r ca. 50% [2], an am ount equivalent to ca. 51 • 1 0 15 g o f fixed carbon per year [2,3]; almost a ll o f this p roductiv ity is from phytoplankton. However, m arine ecosystem-scale p roductiv ity estimates contain a h igh degree o f uncertainty, since they are u ltim ate ly derived by extrapolation from discrete measurements o f NPP or GPP [4,5] w ith lim ited spatial and tem poral resolution. Remote sensing (ocean colour) p roductiv ity a lgorithm s are the most w idely used tool fo r m aking these extrapolations in order to better characterise the nature and extent o f va riab ility in NPP in m arine ecosystems. However, these p roductiv ity algorithm s are exp lic itly dependent on relatively few discrete, surface “ tru th ” 14C p rim a ry productiv ity

PLOS ONE I www.plosone.org 1 March 2013 | Volume 8 | Issue 3 | e58137

mailto:[email protected]

http://www.plosone.org

Electron Requirement for Carbon Fixation

T a b le 1. Studies included in the meta-analysis.

Study/Cruise ID Geographical area D ate <rpsnspec npsn ETR/14C N References

In situ1/PE;AtlanticAMT 11 15 Sep-09 Oct 00 0.003/0.002 [24,25]66

In situVPE5, SISCeltic SeaD246 17-28 May 00 [27,28]0.002

In situVin situ'Gulf o f FinlandTime series 11 Apr-15 Nov 00 [29]0.002

Ariake Bay, JapanTime series 26 Jun-04 Oct 06 [31]1.75 0.002

On deck2/FASTNorth SeaCEND0811 08-12 May 11

AMT15 Tropical Atlantic 17 Sep-29 Oct 04 0.003 In sltuVPE" [26]

Time series Massachusetts Bay 27 Feb-29 Nov 00 1.75 In situ1/PE; 37 [30]

JR98 Celtic & Irish Sea 31 Jul-11 Aug 03 measured On deck1/PE; 28

D366 UK-OA UK & European shelf 06 Jun-12 Jul 11

SYNTAX Baltic Sea 21 - 30 Jul 2010 0.003 On deck2/FAST 24

AMT6 Atlantic 14 May-12 Jun 98 0.002 In situ1/PE; 40 [24,25]

Time series Bedford Basin, Canada 23 Feb-18 Apr 07 + measured On deck2/PE5 20 c

BIOSOPE2 Pacific Ocean 12 Sep-21 Nov 04 1.5 0.003 On deck3/PE5 25 d

SUPREM011 Gulf o f Finland 11-15 Apr 11 + 0.002 On deck2’7/PE5 42 e

Data were collected during cruises undertaken as part o f the Atlantic Meridional Transect (AMT) Program, Individual cruises to the Celtic Sea, Baltic Sea, North Sea, other UK and European shelf waters and the Pacific, as well as part o f time series studies In the Gulf o f Finland, Massachusetts Bay (USA), Ariake Bay (Japan) and Bedford Basin (Canada). Methodological differences existed In the way data were corrected for spectral discrepancies between the fluorometer and 14C-lncubator light source (ftpstopec), in the estimates o f the number o f photosystem II (nPS!!), and In the approach used to compare electron transport rates (ETR) and 14C fixation.+ spectral corrections applied according to Moore et al. [11 ]. In some samples, a spectral correction factor o f 1.75 was used while a spectral correction was n.a. denotes not applicable because primary productivity was measured on samples Incubated In the cuvette holder o f a FASTact Fluorometer.ripsH was assumed to be constant (0.0033 or 0.0020 mol RCII (mol chia) ') or calculated either according to it ) nPSu = 500 (FyFm)/0.65 or (f) Oxborough et al.[7]. ETRs were either measured in situ or on discrete samples using a 1FASTtrecka, or 2FASTact FRR fluorometer, a 3FIRe benchtop fluorometer or a 4FRRFDlvin9 Flash. Superscript 5 denotes photosynthesis Irradiance curves measured on samples Incubated for 1-4 hours using a photosynthetron [32] or equivalent Incubator, 6 denotes 2 hour in situ bottle Incubations, while 7 Indicates rapid light curves carried out In on-deck laboratories under temperature controlled conditions. SIS stands for 24 hour on deck 'simulated in situ ' Incubations, and N Is the number o f observations for each study, a) Silsbe and Ylöstalo unpublished, b) Lawrenz, Capuzzo, Forster, Suggett unpublished, c) Suggett and Forget unpublished, d) Prásll, Gorbunov, Babin, Huot unpublished, e) Ylöstalo et al. unpublished. dol:10.1371 /journal.pone.0058137.t001

measurements. M any researchers have, therefore, turned to high resolution bio-optical-based approaches in order to meet this challenge.

Pulse A m plitude M odula ted (PAM ; [6 ]) and Fast Repetition Rate (FRR; [7,8]) fluorom etry provide the potentia l to dram atica lly increase the num ber o f estimates o f NPP in m arine ecosystems [8 ] , As w ith other b io-optica l sensors, active fluo rom etry can be utilised in situ and thus, measurements o f p roductiv ity can be linked d irectly to measurements o f physica l/chem ical variables at the tim e o f sampling [9 -1 1 ]. In addition, data can be collected at h igh tem poral (seconds) and spatial resolution a n d /o r over large scales [12]. I f direct, in situ measurements o f NPP using active fluorescence could be achieved, the advantages o f this approach w ou ld represent a step change in operational capacity, in terms o f accuracy and resolution, compared to ‘conventional’ 1 4 C-based approaches, w h ich are lim ited by the need to incubate relatively large water samples, often fo r long durations [13], w ith attendant potentia l problems due to ‘bottle effects’ [14,15]. Thus, the widespread im plem entation o f active fluorometers on broad- scale tem poral or spatial sampling platform s, such as ships o f opportun ity and moorings w ould represent a m ajor advance in evaluating the va riab ility o f p rim a ry p roductiv ity in seas and oceans.

Active fluorescence can be used to estimate the rate at w hich electrons flow from water th rough photosystem I I to N A D P H (the so-called linear photosynthetic electron transfer rate, E TR ), and other electron acceptors. E T R is most closely related to the rate o f gross 0 2 evolution, and potentia lly less d irectly to the subsequent p roduction o f energy (ATP) and reductant (N A D P H ) used to fix C 0 2. Thus, measurements o f E T R by active fluorescence, at best,

provide accurate estimates o f the rate o f gross 0 2 evolution from PSII [16,17] and, hence, GPP. Ehifortunately, m any applications (e.g. climate models, fisheries assessments) require that p rim ary p roductiv ity be expressed not as E T R o f 0 2 evolution bu t rather in the photosynthetic currency (sensu Suggett et al. [18]) o f fixed C 0 2, that is, NPP. As such, app licab ility o f broad-scale active fluorescence-based measurements o f E T R are potentia lly lim ited i f ETR s cannot be easily converted to GPP and eventually NPP. Such conversion requires knowledge o f the E T R to C 0 2 fixation ratio , i.e. the electron requirem ent fo r carbon fixa tion (0 >fjc) w ith units m ol e (mol C) . W ith E T R and short-term (hours) C O " fixa tion incubations representing GPP rather than NPP [19,20], 0 r c also captures gross rather than a net efficiency.

Since the in troduction o f active fluorescence to m arine research, a num ber o f studies have attempted to compare measurements o f E T R w ith (quasi-)simultaneous measurements o f C 0 2 uptake, the la tte r representing GPP, NPP or something in between depending on incubation times (1 hour to a fu ll diel cycle) [10,21-23] (Table 1, references therein). In itia lly , these exercises aimed to evaluate whether ETR s provided a robust quantifica tion o f GPP [32 -35 ]. M ore recent studies have sought to understand the extent and nature o f varia tion between the E T R and GPP a n d /o r NPP [10,11,18,36], Numerous b iochem ical processes other than C 0 2

fixa tion can act to consume electrons, A T P a n d /o r reductant; fo r example, the oxygenase activ ity o f ribulose-l,5-bisphosphate carboxylase/oxygenase (Rubisco) v ia photorespiration [32], chlor- orespiration via a plastid te rm ina l oxidase (P TO X ) [33,34], M eh le r Ascorbate Peroxidase (MAP) activ ity [32] and nutrien t assim ilation [35,37]. A ll these processes are expected to exhib it bo th taxon-specific and environm ental dependencies w ith corre




sponding va riab ility o f 4>e c [11,18]; indeed, a recent com pilation o f published FRR-based <Fe>c data sets suggested the existence o f some general taxonom ic patterns in the va riab ility o f <Fec [38]. However, as yet, no single (global) systematic evaluation attempted to consider how environm enta l gradients regulate

Clearly, active fluorom etry could become a m uch more pow erfu l too l fo r estimating GPP (or even NPP) in carbon currency i f generic relationships describing the dependency o f <Fe>c upon rou tine ly measured variables can be established. Therefore, we constructed a database o f FRR-based measures o f <Fec and associated environm ental variables know n to regulate p rim ary productiv ity , (e.g. temperature, nutrients and light) from both previously published and unpublished data sets and used a meta- analytical approach to determ ine the p red ic tab ility o f <Fec from environm ental variables. Specifically, we addressed the fo llow ing three questions: (1) T o w hat extent does <Fe>c vary w ith in and between oceanic areas and water masses? (2) Can a ‘g lobal’ <Fec ever be applied o r should region-specific values always be used instead? (3) H o w is the va riab ility o f <Fe>c related to that o f environm ental factors o r combinations thereof (light attenuation, nutrients, temperature, salinity, etc.)? O u r analysis demonstrated tha t va riab ility in <Fe c was strongly correlated w ith the ava ilab ility o f lig h t and nutrients, as w ell as temperature a n d /o r salinity, albeit to vary ing degrees and depending on how data are organized in to region-specific subsets. Predictive algorithms based on these relationships are presented. T rea tm ent o f data at the regional scale provided m uch im proved predictive capability o f these algorithm s relative to large scale o r global algorithms. A lthough m any o f the algorithms still need fu rther verification and revision, this approach demonstrates strong relationships between <Fe c and environm ental gradients in m any areas o f the ocean.

Materials and Methods

Data compilationA comprehensive assessment o f the va riab ility in <Fec from

published literature and previously unpublished data was performed. W eb o f Science and J S T O R search engines were used to retrieve published data on coincident F R R fluorescence-based ETR s and carbon fixa tion rates from m arine habitats, y ie ld ing 17 studies. However, on ly 7 o f these studies reported key environmental variables in add ition to c o r both ETR s and C 0 2-fixation rates (resembling either NPP or GPP depending on the incubation lengths and grow th conditions), from w hich <FeC could be calculated. Previously unpublished data corresponding to parallel E T R and C 0 2 fixa tion measurements from an add itiona l 7 fie ld campaigns in 2004-2011, some o f w h ich were undertaken as part o f P R O T O O L (h ttp ://w w w .p ro to o l-p ro je c t.e u /), were also in cluded (Table 1). Together, these data sets included ten research cruises and fou r tim e series studies covering a range o f d ifferent geographical areas (Fig. 1), inc lud ing the temperate, trop ica l and subtropical A tlan tic Ocean (A M T 6 , A M T 11, A M T 15), Massachusetts Bay (USA), Bedford Basin (Canada), A riake Bay (Japan), the Celtic and Irish Sea (D246, JR98), the N o rth Sea (C E N D 0 8 11/P R O T O O L , G u lf o f F in land (including SU PR EM O 11/P R O T O O L ), the Baltic Sea (S Y N T A X 2 0 1 0 /P R O T O O L ), U K and European shelf waters (D 3 6 6 /P R O T O O L ) and the Pacific Ocean (BIOSOPE).

A comprehensive data m a trix w ith 333 d ifferent samples and the ir corresponding physico-chemical and m ethodological va riables was created. Physico-chemical variables included salinity, temperature, nu trien t concentrations ( N 0 3- and P 0 43 - ), the diffuse vertica l a ttenuation coefficient (Kd) o f photosynthetically active rad ia tion (PAR), optical depth (Q and ch lorophyll a (chia),

w ith the la tte r being used as a p roxy fo r phytop lankton biomass. W e fu lly acknowledge tha t other environm enta l variables, such as the silicate o r iron are often also key in determ in ing phytoplankton com m unity composition and physiological responses and, hence, influence <Fe>c- However, these data were either not collected as pa rt o f the studies included here o r no t available. M ethodologica l variables included differences in F R R protocols and 14C incubation techniques (see below). Locations o f samples were characterized by la titude and longitude, while seasonal differences were characterized by converting sampling dates to Ju lian day. A ll the environm ental data were either k ind ly provided by the authors o f the published work, the British Oceanographic D ata Centre (B O D C , www.bodc.ac.uk) or were taken from the orig ina l publications and dig itized (Plot D ig itize r 2.5.0, Free Software Foundation) from relevant figures.

Values o f Kd (in units o f m *) were derived from vertical irradiance profiles fo r the A M T cruises, Bedford Basin, the Celtic Sea (JR98), Baltic Sea (SYN TAX 2010), A riake Bay, the G u lf o f F in land [29], the U K -O cean A c id ifica tion cruise (D366) and the Pacific Ocean (BIOSOPE2) dataset.

For the rem ain ing studies, M O D IS 4 km satellite products o f euphoric zone depth (zeu), here defined as 1% o f surface PAR, produced by the G iovanni online data system (N ASA G oddard Earth Science Data and In fo rm a tion Services Center) averaged over 8 days were used to calculate Kd by solving

E 1% = E 0 x e - ZeuKd (1)

fo r Kd w ith E 0 set to 100%. M u ltip ly in g Kd by the actual sampling depth then gives C (dimensionless), w ith C<4.6 corresponding to irradiance levels > 1 % [39]. For data collected p rio r to 2002 (Massachusetts Bay), SeaWIFS 9 km zeu products were used for calculating Kd and subsequently

FRRF measurementsF R R fluorescence transients were either measured in situ o r on

discrete samples on-board using F A S T tracka I or F A S T tracka I I fluorometers w ith FA S T act systems (Chelsea Technologies Group, L td ., W est Molesey, U K ), F IR e benchtop instruments (Satlantic, LP, H a lifax , Canada) o r a F R R Dlvmg Flash fluorom eter (K im oto Electric Co., L T d ., Osaka, Japan) (see Table 1). F R R fluorometers were rou tine ly program m ed to generate a standard protocol w ith 50-100 single turnover (ST) saturation flashlets o f 1.1-3.3 ps dura tion at 1-3.6 ps intervals [18,40]. Each induction curve was separated by ~ 1 0 ms. T o increase signal to noise, between 5 and 160 sequential induction curves were averaged per acquisition.

The biophysical m odel o f K o lb e r et al. [41] was fitted to all fluorescence transients to derive the in itia l fluorescence (F¿¡ and m axim al fluorescence (Fm) yields measured in the dark, the m in im a l (Fo’), steady state (F’) and m axim al (Fm’) fluorescence yie ld measured under am bient irradiance, as well as the functional absorption cross section o f photosystem I I (PSII) in the dark (ffpsu) and lig h t (opsii ) (in units o f A 2 quanta-1 ). For most studies, the photosynthetic electron transfer rate through P S II (units o f m ol e- (mg chia)- h - ) [42,43] was calculated as:

F 'E T R = E x G p sn x n P S ii X -T 7 X <S>R C X 2.43 x IO “ 5 (2)

where E is lig h t intensity, npsn the ratio o f functiona l P S II reaction centres to ch lorophyll (in units o f m ol R C II (mol chia)"1), @pc aa assumed constant o f 1 electron yielded from each reaction centre I I (R C II) charge separation and 2.43 • IO" 5 is the factor that


http://www.protool-project.eu/

http://www.bodc.ac.uk



♦ Bedford Basin

^ O AMT 6I @ AMT 11* • AMT 15

□ Pacific B10S0PE 2

I o Ariake Bay

♦ Massachusetts Bay

I 0 Gulf o f FinlandA Baltic Sea SYNTAX2010

I A UK-OA D366

I A North Sea CEND0811V Celtic Sea D246 . V Celtic Sea JR98

60° N

3 0 ° N

9 0 °W 90°E

Figure 1. Data sets used in the meta-analysis. G u lf o f F in land inc ludes b o th th e SUPREMO2011 and th e Raateoja e t al. [29] s tu d y (See Tab le 1 fo r details).d o i:1 0.1371 /jou rna l.pone .0058137 .g001

accounts fo r the conversion o f Á 2 quantum " 1 to m 2 (mol R C II) ’ 1, m ol chi« to m g chi«, seconds to hours and pm ol quanta to m ol quanta [13,29]. Measurements o f <Jpsn account fo r transient nonphotochem ical quenching in the antenna bed as a result o f exposure to transient lig h t [44], and thus, values o f oPSI/ were typ ica lly taken from the F R R f dark chamber to increase signal to noise.

The PS II efficiency factor, term ed F q’/ F / (was calculated as (Em’- F ) / ( E m’- F 0) either from fluorescence emissions measured sequentially on dark acclimated samples and then under actinic ligh t, o r estimated as the difference in the apparent PSII photochem ical efficiency between the F R R ligh t and dark chamber quasi-simultaneously, as in Suggett et al. [25]. In this case, no b lank correction to the fluorescence yields is required because any con tribu tion o f background fluorescence (to F \ F q or F m) effectively cancels between lig h t and dark cham bers/ conditions [25,44], For two studies [25,28], the electron transport rate was evaluated using the equivalent equation:

F ' FE T R = E x ops ii x t ip s i i x — ^ - 7 x -E - x % x 2.43 x 10 5 (3)

i m i y

where values o f F v/F m and (Jpsn are measured in the dark or assumed dark values based on measurements from deeper in the water column, tha t is, from depths where E < E K and where nonphotochem ical quenching can be assumed to be negligible [25]. In the la tter case, the influence o f both photochem ical and nonphotochem ical quenching are all accounted fo r by changes in Fq / F m\ This slightly alternative approach was employed to enable ETRs to be estimated in the absence o f a ’dark chamber’ , and consequently, corresponding ligh t and dark acclimated samples from all measurement depths, w h ich would be required in order to apply Eq. 2. Both, Eqns. la n d 2, appear to re tu rn consistent estimates fo r quenching [8 ], and hence, E T R . I t should fu rthe r be noted that derivation o f F q’/ F / requires a measurement to be

PLOS ONE I www.plosone.org 4

made fo llow ing b rie f dark exposure. Consequently, under circumstances where rap id reversal o f certain components o f N P Q occurs during the dark measurements, Fq /F - f may be overestimated po tentia lly causing overestimates o f E T R evaluated using Eqn. (2) by up to 30% in phytop lankton cultures [8 ].

A p a rt from the b lank correction o f the absolute fluorescence yields (see above and [45]), the accuracy o f the E T R also depends on how variable some assumed ‘constants’ are and how well certain corrections are applied; specifically, (1) nPsn [46], (2) the spectral correction o f oPSn, and (3) differences between ETR-based and C 0 2-based values o f lig h t harvesting efficiency (a) and m axim um photosynthesis rates (Pmax). F irstly, nPsn is known to vary between taxa and environm ental conditions by up to a factor o f 5 [8 ] bu t is rare ly measured in the context o f FRR-based p roductiv ity studies. In fact, o f the studies used here only M oore et al. [11] and Suggett & Forget (unpubl.) used d irect measurements o f npsn- The m a jo rity o f the other studies assumed values o f 0.002 and 0.003 m ol R C II (mol chia) 1 fo r populations dom inated by eukaryotes and prokaryotes, respectively [47], o r calculated nPsn using a newly developed a lgorithm , w hich w ill also have associated caveats [7]. In evaluating the use o f this assumed constant nPsn in some o f our data sets, we compared @e c derived w ith constant rtpsu to <7C/, based on measured nPs n {see discussion). W e re tu rn to the issue o f spectral corrections in the section “ m atching ETRs w ith C -uptake” .

A lthough the F R R f approach provides a general estimate o f a and Pmax, the ligh t saturation po in t (EK) fo r the E T R m ay be lower than tha t from 14C uptake because of, fo r example, electron consuming processes [48 -50 ]. The la tte r can cause the turnover tim e fo r Q A to decouple from that o f the whole chain PSII turnover, so tha t the EK fo r the E T R may not be w ho lly indicative o f where ligh t is lim itin g or saturating fo r C 0 2 uptake.

14C02 fixation ratesC 0 2 fixa tion was measured by either in situ incubations [29],

simulated in situ incubations [27,28,30,31] or photosynthesis versus irradiance (PE) relationships [11,25,26] in a ‘photosynthetron’

March 2013 I Volume 8 I Issue 3 I e58137



according to Lewis and Sm ith [51] o r an equivalent thereof (see Table 1 fo r details). C 0 2 fixa tion in A riake Bay (Japan) was measured using the stable isotope 1 3 C-labelled N a H 1 3 C 0 3 [31]. Incubation lengths varied between the d ifferent studies, w ith the vast m a jo rity incubating fo r a few (1 M ) hours and thus capturing p roductiv ity somewhere between GPP and NPP [19,20].

Matching ETRs w ith C-uptakeIdeally, measurements o f photosynthetic carbon fixa tion and

E T R should be made simultaneously on the exact same sample to reduce errors arising from differences in sample treatm ent and handling [8,18]. M oreover, the methods used to determ ine C 0 2

fixa tion differed considerably between studies (Table 1). Some o f the most recent studies in the Baltic Sea (S YN TAX 2010) and N o rth Sea (CEND0811) fo llowed recommendations o f Suggett et al. [8 ] and measured 14C uptake by placing the radioactive sample d irectly in the cuvette holder o f the FASTact fluorom eter. This simultaneous incubation technique avoided discrepancies in the intensity and spectral qua lity o f the actin ic lig h t sources. In this way, E T R and GPP were measured fo r 1 h on the same sample at in situ temperatures and at a lig h t intensity corresponding to either h a lf o r twice the value o f E k (as determ ined from F R R f rap id ligh t curves); these lig h t intensities relative to E K were chosen to y ie ld a measure o f <&e c corresponding to an irradiance fo r ligh t-lim ited and lig h t saturated photosynthesis.

For most studies, 1 4 C-specific PE experiments were perform ed on discrete water samples w hilst F R R data (and hence ETRs) were determ ined from in situ casts o r on deck (Table 1). Incubation lengths varied between the d ifferent studies, w ith the vast m ajority incubating fo r a few (1-4) hours and thus capturing p roductiv ity rates somewhere between GPP and NPP [19,20]. For this same reason, adhering to strict definitions o f NPP and GPP throughout the m anuscript is not always possible. W e have therefore specified NPP o r GPP where possible bu t otherwise use the terms C 0 2

fixa tion o r p rim a ry p roductiv ity . In situ F R R data were compared to the PE data from the same depth as the discrete water sample. For on-deck based E T R measurements du ring S U P R E M O ll, actin ic lig h t was provided by an external program m able ligh t source. The ligh t intensity measured du ring the F R R data acquisition (used to calculate the E T R ) was then applied to the 1 4 C-PE equation to y ie ld a measure o f the instantaneous 14C uptake fo r m atching w ith the E T R . For some studies, simulated in situ 14C data were collected (corrected using 1 4 C-PE to y ie ld ‘gross’ 14C uptake [27,28] (Table 1). Here, we adopted a sim ilar approach b u t fu rther normalised da ily 14C uptake rates to hourly rates based on knowledge o f the ligh t regime.

D irec t comparison o f the ETR s and C uptake requires that the spectral qua lity o f the actin ic lig h t o f the F R R f and 14C incubator is e ither equivalent o r corrected for. The spectral values o f gpsn (measured w ith a blue LE D ) also needed to be scaled to the ligh t qua lity o f the actin ic source used fo r the 14C incubations [40]. In most cases, in situ F R R values o f ffpsu and the 14C actin ic ligh t source were bo th spectrally corrected to m atch the spectral qua lity corresponding to the sample depth [40]. In cases where both FR R - and 1 4 C- PE curves were measured on discrete samples, data were spectrally corrected fo llow ing M oore et al. [11]. For the studies that d id not employ a spectral correction (Table 1) we assumed a constant factor o f dpSn/ \ J b [11,31] and GPsn/ 1.5 (Prásil et al. unpubl.) based on approxim ate values from the other studies fo r the water types in question.

T ak ing in to account the ligh t h istory o f samples from different studies was the greatest challenge in com piling the database due to considerable inconsistencies in the ava ilab ility and qua lity o f ligh t data. Thus, irradiances were expressed as optical depths, and

instantaneous ligh t was standardised by norm aliz ing E relative to the saturating lig h t intensity, E k , as determ ined from 1 4 C-specific PE curves, i.e. E :EK (dimensionless). In this way, E can be considered relative to the lig h t h istory to w hich cells are acclimated [52] p rov id ing in fo rm a tion as to whether the values o f <&ec correspond to lig h t lim ited photosynthesis (E:EK< 1 ) or ligh t saturated photosynthesis (E:EK >1). Note, however, that E comes from in situ P A R measurements, while E k was usually based on measurements made in the laboratory. In this case, the spectral qua lity o f E and E k m ay d iffe r and values o f E / E k ratios from different studies must be treated w ith caution.

A ll C 0 2 fixa tion rates (in m g C L 1 h *) were normalised to the corresponding chi a concentration to yie ld C 0 2 uptake rates as m ol C 0 2 (mg chia) 1 h *; thus values o f the <&e c (in m ol electrons m ol C 0 2 ’ ) were determ ined as

<$eC = E T R /C 0 2 u p ta k e ra te (4)

G iven the large array o f data sets and the various differences in approach fo r determ in ing bo th 14C uptake and F R R fluorescence ETR s [8,18], we recognise that a m ajor assumption inherent to our analysis is that environm ental influences outweigh m ethodological influences on <£eC, both, w ith in and between studies. Fortunately, a ll studies employed sim ilar F R R f protocols (above). Even so, in fo rm a tion on core variables associated w ith the E T R determ inations were in itia lly included in our database and examined alongside the environm ental variables to verify any role o f m ethod upon apparent va riab ility in Û>e c (see “ statistical approach” section). N ote tha t the ‘m ethodological’ data across studies is categorical rather than continuous, (e.g. spectral correction o f Gpsn applied o r no t applied); therefore, the available in fo rm a tion was converted in to a Boolean code w ith numbers one and zero representing use and non-use o f a particu la r method, respectively. In total, 3 m ethodological variables were u ltim ate ly included in the in it ia l data set: nPSi i assumed or measured, spectral correction o f (Jpsii assumed o r measured, and E :EK > 1 orE:E k < 1 .

A t this po in t i t should also be noted tha t apart from environm ental factors, differences in phytop lankton com m unity composition m ay also influence & t c- U n fortunate ly, taxonom ic data was no t consistently available fo r the m a jo rity o f studies. However, we assume tha t phytop lankton com m unity composition w ill pa rtia lly reflect environm ental characteristics and some influence o f com m unity composition on & ec w ill like ly be im p lic itly accounted fo r in our analyses pure ly based on environm ental variables.

Statistical approachM any o f the fo llow ing analyses were carried out on bo th the

entire dataset o r on ind iv idua l subsets thereof, which, fo r example, represent d ifferent oceanic regions. A lthough the m a jo rity o f our samples were collected in the A tlan tic Ocean and adjacent shelf waters while other regions (e.g. Pacific, Ind ian Ocean, and M editerranean Sea) are underrepresented/not represented at all, we still use the term ’global’ fo r analyses carried out on the dataset as a whole.

Spearman Rank O rde r C orre la tion analysis on the ind iv idua l data sets (i.e. samples grouped by study) were used to identify key variables that m ay be associated w ith <J>e c. Correlations, evaluated in SPSS 15.0 (SPSS Inc., Chicago, IL , USA) were considered significant when j&<0.05. These correlations were carried out i) on the ind iv idua l data sets (i.e. samples grouped by study), ii) on the global dataset, iii) on studies pooled according to regions (e.g.




N o rth Sea, Baltic Sea, A tlan tic Ocean etc.) and iv) according to shelf/oceanic waters, thus p rov id ing an overview o f w hich variables may be im portant. A ll data were then combined in to one large database to carry ou t a ll other statistical analyses using non-param etric m ultivaria te techniques in P R IM E R -E version 6

(P R IM E R -E , L td . Ivybridge, Devon, U K ) [53], unless noted otherwise.

Inter-corre lated and right-skewed physico-chemical variables (e.g. salinity, temperature, chirr, N 0 3 —, and P 0 4 3 ) were identified using D raftsm an plots and then square-root transformed to stabilise the variance and ensure that Euclidean distance could be used as an appropriate s im ila rity measure (see below). T o account fo r differences in scales and units between d ifferent variables, the la tter were norm alized by subtracting the mean from each entry o f a single variable and d iv id ing it by the variable ’s standard deviation.

N on-m etric M ultid im ensiona l Scaling (nM DS) was used as a means to map environm enta l characteristics o f samples in a low dimensional space based on a triangu lar resemblance m a trix that was created by calculating Euclidean distances between every possible pa ir o f samples. Hence, Euclidean distances between samples represented the dissim ilarities in the suite o f the ir physicochemical properties. P rincipal Com ponent Analysis (PCA) was then used to identify variables accounting for the differences between samples. Variables in itia lly included in the P C A were salinity, chi a, N 0 3 , P 0 4 3 , temperature, K¿, and C, as w ell as a Boolean-m atrix o f m ethodological differences w ith regard to nPsn, apsii and E :EK (as described above, Table 1).

Inclusion o f the m ethodological and location variables (latitude, longitude (absolute values) and Ju lian day) usually increased dissim ilarities, i.e. Euclidean distances, between samples o f d ifferent studies compared to analyses from w hich m ethodological and location variables had been excluded. Because the ‘m ethodological choices’, in particu lar, cannot be effectively incorporated in to predictive algorithm s fo r <&e c, the results we show in the m ain text here are based on PCAs w ithou t methodological and location variables. Nevertheless, the results o f the P C A inc lud ing these m ethodological and location variables have been included in Append ix S I (Fig. S I). O n ly those variables w ith the highest eigenvectors were included in any fu rther analysis to assess the va riab ility in & t c-

Both nM D S and P C A generate low dimensional ordinations. Here we on ly show the n M D S plots, w h ich matched the P C A plots well, because nM D S better preserves the distances between samples when m apping them onto a 2D o r 3D space [53]. In add ition n M D S provides a measure, the stress value, o f how well the low dimensional ordinations represent the distances between samples. In all cases the stress value o f the n M D S was considerably lower in the 3D ord ina tion relative to the 2D ord ination. Thus, the 3D-images are presented here where the ind iv idua l planes (x-y, x- z, y-z) are separate panels.

Because environm enta l factors showed considerable spatial and tem poral variation , which, in tu rn , influence physiological responses o f phytop lankton and subsequendy í>e>c to varying degrees, hierarchical agglomerative cluster analysis and a sim ilarity pro file (S IM PR O F) test were used to fin d and define groups o f samples w ith sim ilar physico-chemical properties [53]. S IM P R O F tests were used as a stopping rule o f the cluster analysis, so that successive partitions along the branches o f the dendrogram are only perm itted i f the nu ll hypothesis o f ‘no structure between samples’ was rejected. T o reduce the num ber o f clusters to a manageable size and to ensure tha t a sufficient num ber o f samples large enough fo r a lgorithm development were included in each cluster, the j6 -value fo r the S IM P E R O F test was reduced to 0.005.

Thus, once a non-significant test result was obtained (fi>0.005), samples below that s im ila rity level were no longer partitioned in to clusters and could be regarded as homogeneous [53].

A ll data were then re-grouped according to the results o f the cluster analysis and S IM P R O F test. The P R IM E R -B E S T match perm utation test was then used to iden tify variables and variable combinations that best “ expla in” the va riab ility in c [53]. T ha t is, the a lgorithm random ly permutes a resemblance m atrix generated from & ec (based on all possible pair-wise sample combinations) relative to a resemblance m a trix generated from subsets o f the environm enta l data m a trix searching for h igh rank correlations between the two and generating a correlation coefficient (p). Repeated (99) perm utations o f random ly ordered environm ental data fo llow to test the significance o f the results at a level o f j&CO.Ol [53],

One o f the m ain goals o f this meta-analysis was to generate algorithm s to p red ic t c from environm ental variables for specific regions. Therefore, m athem atical relationships between the environm ental data and & e>c were produced fo r each cluster tha t contained at least 5 data points using m ultip le linear regression (M LR ). O n ly those variables and variable combinations that were significandy correlated w ith c according to the P R IM E R -B E S T-test were entered in to the M L R in SPPS.

Results

Spatial and temporal variability in < P e CM ean values for & ec in all b u t four studies were < 1 0 m ol e

(mol C) *, resulting in a global mean (± standard deviation) o f 10.9± 6.91 m ol e- (mol C )- 1 . Values o f 0 eC less than the theoretical ra tio o f 5 m ol e (mol C) 1 (see discussion) were observed in the G u lf o f F in land 2000 and in pa rt o f the A M T 15 data. V a r ia b ility o f í>e>c w ith in and between ind iv idua l studies was considerable, w ith a to ta l range o f 1.15-54.2 m ol e- (mol C ) - 1 fo r a ll studies com bined (Fig. 2). W ith in -s tudy varia tion o f & e>c was largest in the Pacific Ocean (7.9-54.2 m ol e (mol C) *), Massachusetts Bay (8.5-50.1 m ol e- (mol C)-1 ), and the A M T 15 data (1.1-28.2 m ol e (mol C) *); in contrast, least w ith in - study variance was typ ica lly observed fo r tim e series’ at a single location (Ariake Bay: 5.1-5 .7 m ol e (mol C) , Bedford Basin: 3 .6-10.3 m ol e (mol C) *, and the G u lf o f F in land 2000: 2.0— 9.4 m ol e- (mol C )-1 ). As there were no d istinct differences in the num ber and types o f ffpsiispec and nPSn corrections between the tim e series studies and the other cruises, i t w ou ld appear that va riab ility o f <&e c is generally greater spatially than tem porally w ith in the included locations. The tim e series studies included here were o f relatively short dura tion (1.5-7 months), however, tem poral va riab ility m ay fu rther increase i f long-term studies (m ultip le years) are included.

Spearman Rank O der Correlations could be found between <&e c and every environm ental variable included in the study, albeit to vary ing degrees and depending on how data were grouped. O n a global scale, i.e. com bin ing the entire data set, <&e c exhibited significant positive correlations w ith Ju lian day (i.e. time/season) and salinity, while significant negative correlations existed w ith la titude, longitude, c h k concentrations Kd and Ç (Table 2). These correlations explained at least 12% o f the variance in a ll cases. T o consider i f the global data pool may obscure regional-scale trends, correlations were also carried ou t (i) on the ind iv idua l studies and (ii) by poo ling studies representative o f particu la r regions, such as, the Baltic Sea (including the G u lf o f Finland), the A dan tic Ocean (all A M T cruises combined) and by com bin ing data from shelf and oceanic waters, respectively.




OÖE

oE

S

40

(S)

Figure 2. Variability in the electron requirement of carbon fixation (0 eci derived from corresponding FRRf-ETR and 14C- prim ary productivity measurements. Boxes rep resen t th e m ed ian , 0.25 and 0.75 q u a rtile , w hiskers are th e 1.5 in te rq u a rtile range. O u tlie rs are ind ica te d b y o pe n circles. The dashed g rey line is th e th e o re tica l re fe rence ra tio o f 4 m o l e~ [m o l C U 1. A M T = A tla n tic M erid iona l Transects, CS = C eltic Sea, NS = N orth Sea, GoF = G u lf o f F in land 2000 (tim e series), U K-OA = UK O cean A c id if ic a tio n c ru ise , M A = M assachusetts Bay (tim e series). SYNTAX2010 is a Baltic Sea cruise, G oF2011 is th e SUPREMO2011 s tu d y and th e Pacific Ocean s tu d y is th e BIOSOPE- cru ise to th e S ou theast Pacific (see Table 1, F igure 1 fo r deta ils).d o i:1 0.1371/jo u rn a l.pone.0058137.g002

A t the level o f the ind iv idua l study, d ifferent environm ental variables showed varying degrees o f corre lation depending on the region. M u ltip le notable trends were evident between <PfC, sampling depth, L j, 'Ç and nu trien t availab ility, including: 1 ) changes o f <Pf Cw ith depth, w h ich were observed du ring A M T 1 5 , in the Celtic Sea, the N o rth Sea (N S -C E N D 0811) and the G u lf o f F in land /B a ltic Sea; 2) a decline in & r c w ith increasing K j during A M T 6 and in the Baltic Sea, and w ith increasing ç during A M T 15, in the Celtic Sea, Pacific Ocean and in the G u lf o f F in land /B a ltic Sea; and 3) a change in & r c w ith nu trien t concentrations, du ring the A M T cruises, one o f the Celtic Sea cruises (JR98) and in the G u lf o f F in lan d /B a ltic Sea. In the Baltic Sea (SYN TAX 2010) and Bedford Basin significant correlations existed between & e c and temperature and salinity, respectively. N o corre lation o f i f , c w ith com m on environm ental variables were found in Massachusetts Bay or du ring the U K -O A cruise. Note also, that depth-dependent trends in <X>r c could only be assessed where m ultip le depths had been sampled at each station across large proportions o f the cruise transect/tim e series, such as in the open ocean studies where data were available up to a depth o f ~ 2 00 m (e.g. A M T cruises, Pacific Ocean, Celtic SeaJR98) or the G u lf o f F in land (S U P R E M O ll) studies where up to four depths were sampled across the euphotic zone at each station. Thus, a lack o f relationship between i f , c and depth may simply reflect lim ited sampling depths or, in shallow waters w ith rap id vertical m ixing, acclim ation to an average water co lum n irradiance.

A notable feature fo r the Pacific and A tlan tic transects as w ell as the Celtic Sea (JR 98, data not shown) was a pronounced increase o f i f c surface waters w ith low N 0 3 a n d /o r P 0 4 3 ava ilab ility (Fig. 3, Table 2). In fact, the highest i f c values across a ll studies

corresponded to a surface water lens in the H N L C and H um bo ld t upwelling region o f the Pacific Ocean. For the other A tlan tic studies, A M T 6 and 11, i f ̂ also declined w ith increasing N 0 3

a n d /o r P 0 43 - , bu t as the depth o f the surface m ixed layer a n d /o r n itracline varied between stations, the relationship between depth and i f c was not linear.

Negative correlations between i f c and 'Ç fo r A M T 1 5 , the Celtic Sea (JR98), the Pacific Ocean and the G u lf o f F in land indicated that i f , c often increased towards depths o f greater ligh t ava ilab ility in surface waters o r w ith declin ing K¿ values from coastal to offshore regions. Negative correlations o f & e c w ith depth, A /a n d Ç were also found bu t on ly in Baltic Sea and shelf waters.

Regional differences in environmental conditionsThe in itia l correlative exercise for each separate study

demonstrated that the relationships between i f ca n d environm ent are dependent upon how data w ith in and between data sets are grouped; therefore ‘choice’ o f grouping w ill inevitab ly influence the outcome o f em pirica l algorithm s generated to pred ict $>e c from physico-chemical variables. A m ajor objective o f this study was to iden tify environm enta l predictors o f i f therefore P C A in com bination w ith cluster analysis was subsequently used to 1 ) iden tify the p rinc ipa l differences in environm enta l condition between sites regardless o f the study to w hich they belonged and 2 ) to fo rm clusters o f sites w ith sim ilar environm ental conditions, w h ich could then be analysed fu rthe r w ith respect to the ir association w ith i f c.The first three p rinc ipa l components o f the PC A combined accounted for 8 6 % o f the cumulative varia tion in environm ental variables (PC I 47% , PC2 24% and PC3 15%). Tem perature, chia and K j had the highest coefficients for the linear com bination o f variables com prising P C I, i.e. they were associated w ith the separation o f samples along P C I (Table 3). Salinity, N 0 3_ and P 0 43 - differentiated samples along PC2, while a single variable, Ç, had the highest Eigenvectors o f PC3.

Cluster analysis in com bination w ith a S IM P R O F test generated 15 significantly d ifferent clusters (ƒ><().005) labelled alphabetically from a-o (/><().005) across a range o f Euclidean distances (representing dissim ilarities between samples) from 0 — 4.4; these clusters often overlapped in the n M D S ord ina tion o f the Euclidean distance m a trix derived from the environm ental variables, ind ica ting that there existed sim ilarities in some o f the environm ental conditions between samples from different studies (Table 4, Fig. 4).

Some o f the resulting clusters corresponded to obvious biogeographic regions a n d /o r seasonal (i.e. temperature dependent) groupings, while others represent more o r less d istinct water masses (Table 4). A ll the samples collected in the Baltic Sea and G u lf o f F in land fe ll in to biogeographically distinct clusters (b, c, f, g and h) characterised by low salinities (Fig. 5). The four G u lf o f F in land clusters (b, c, g, h) clearly represented tem poral changes in environm ental conditions w ith samples in cluster b being characterised by extremely low temperatures du ring late w in te r (SETPREMO11 cruise) coupled w ith h igh nu trien t availability. Low temperature and h igh nutrients set this cluster apart from spring cluster c w ith still relatively low tem peratures(<5 UC) and clusters g and h, w h ich contain all the summer and autum n samples characterized by w arm temperatures (~ 15 UC), low N 0 3

ava ilab ility and h igh Kd and t values. Samples from the Baltic Sea proper (cluster f) were characterised by almost undetectable nu trien t concentrations. O the r clusters w ith samples representative o f d istinct regions include the N o rth Sea cluster i and cluster k containing Celtic Sea samples and samples from the European shelf edge collected du ring A M T 6 . L ike in the Baltic Sea, deplete




T a b le 2 . Spearman Rank Order Correlation Coefficients for correlations between <PeC and environmental variables.

Lat Lon JD Depth Tem p Sal N 0 3 P 0 43 Chia Kd Ç

Studies combined

All Studies -0.272 -0.450 0.297 - - 0.226 - - -0 .130 -0.244 -0.123

Shelf -0.431 -0.546 0.214 0.149 0.376 0.229 0.186 -0 .1 36a -0.511 -0.155

Oceanic -0.324 -0.320 - 0.201 - - - -0.243 -0 .1 67a - - -

AMT combined - 0.212 0.250 0.276 -0.361 -0.417 -0.1 75a

Baltic com binedf ' 0-347 0.530 -0 .337 0.271 -0 .438 0.276 0.269 0.320 -0.562 -0.569

AMT 6 0.322 - 0.322 - 0.520 - -0.374 -0.406 -0.577 -0.415 -

AMT 11 -0.344 -0.431 0.344 - - 0.329 - -0.338 - - -

AMT 15 - - - -0 .709 - - -0.765 - -0.681 - -0.756

CS-D246 . . . . . . . . .

CS-JR98 - - - -0 .564 - - -0.391 - - - -0.482

CS-DS46+JR98 - - 0.377 -0.614 0.423 - - - -0 .374 - -0.416

PO-BIOSOPE -0.482

NS-CEND0811 - - - 0.616 - - - - - - -

UK-OA D366 . . . . . . . . .

SYNTAX2010 - - - 0.568 - - - - - -

G0F2OOO n.a. n.a. n.a.

SUPREM011 - - - -0.645 - - -0 .494 - - - -0.653

GoF +SUPREM011 0.352 0.752 -0.737 — 0.310a -0.683 0.668 0.582 0.586 -0.452 -0.705 -0.684

Bedford Basin n.a. n.a. - n.a. - 0.582 - - - - -

MA Bay . . . _ _ _ _ _ _

Values in bold or normal font indicate significant correlations where p<0.01 and p<0.05, respectively.JD is Julian Day. f denotes Baltic proper and the Gulf o f Finland combined. - non-significant correlations (p>0.06). b denotes correlations w ith 0.05>p>0.06. doi:10.1371 /journal.pone.0058137.t002

nutrien t concentrations in cluster k sets this cluster apart from cluster i.

Samples from Bedford Basin fe ll in to two clusters: cluster d containing samples collected from m id -M arch to m id -M ay when ch lo rophyll concentrations increased and nu trien t concentrations and water c la rity dropped from previous levels that characterized samples from cluster e, w h ich were collected in late w in ter (February to m id-M arch). Interestingly, cluster e also contained almost all o f the Massachusetts Bay samples, some N o rth Sea samples from fron ta l regions as well as A M T 6 and Pacific Ocean samples from the H u m bo ld t and Benguela upw elling areas, respectively. Th is represents sim ilarities in physico-chemical properties o f these d ifferent water masses rather than biogeographical regions. D u rin g late w in ter, salinity in the ice covered Bedford Basin was sim ilar to that in o ff shore waters, while ch lo rophyll and nu trien t concentrations as w ell as optical properties matched those o f upw e lling /fron ta l regions; hence, the ir grouping together.

The rem ain ing clusters (j, 1, m, n, o) also contained samples from a varie ty o f d ifferent cruises/regions. Samples collected from the deep ch lorophyll m axim um (D C M ) in o ff shore regions du ring the A M T cruises, the Pacific Ocean cruise and the Celtic Sea (JR98) fell in to one cluster 0 , w h ich was characterized by high temperatures as well as higher nu trien t ava ilab ility and 'Ç relative to the other offshore samples. Cluster 1 contained samples from interm ediate depths in temperate waters on the European shelf and the shelf edge, w h ich had relatively low temperatures. W hat set cluster 1 apart from cluster n, w h ich also contained samples from the European shelf, was its overall greater optical depth. Open ocean samples w ith high water c la rity and temperatures, as

well as extrem ely low ch lorophyll and nu trien t concentrations fell in to clusters m and o. W h ile cluster m contained samples from interm ediate depths collected during A M T 6 and A M T 11 and from the surface o f the South Pacific Gyre (SPG) and H N L C region in the Pacific, cluster o was comprised o f surface samples from the open ocean (all A M T cruises), central N o rth Sea and Celtic Sea JR98).

Clearly, temperature, salinity, ch lorophyll, nu trien t concentrations and ligh t ava ilab ility were the m ain environm ental factors responsible fo r the d istinct g rouping o f samples in to clusters. Given that cluster analysis in strongly stratified, off-shore waters resulted in d istinct groupings corresponding to D C M , interm ediate depth and surface samples and due to the pronounced effects o f water co lum n stratification on lig h t and nu trien t availab ility, we also p lo tted the mean <£f C, grouped according to samples from the surface m ixed layer (SML) and D C M against depth and N 0 3 fo r each biogeographic province sampled du ring the Pacific and A M T cruises (Fig. 6 ). As such, distinct differences in & eC existed between the S M L and D C M samples from the Pacific and A M T 15 cruise, where 0 r c was m uch higher in the S M L than at the D C M and the increase in & r c coincided w ith a drop in nu trien t ava ilab ility or even depletion o f nutrients in surface waters. D u rin g the A M T 6

and A M T 11 cruises, on the other hand, no such pronounced differences in & eC and nu trien t concentrations between S M L and D C M was observed fo r the sampled locations in most o f the biogeographic provinces.

In summary, there existed distinct groupings for m any o f the cruises and regions based on differences in environm ental gradients w ith in and between regions and water masses. This also

PLOS ONE I www.plosone.org March 2013 | Volume 8 | Issue 3 | e58137



30

30

25

20

S- 100

150

NAG EM200

40

30Ha- loo

—150

CZ NAGSAG EM

8- 100

150

200EQ EQ

Figure 3. Changes in the electron requirem ent o f carbon fixation (<Pe,à and N 0 3 across 4 cruise transects. (A) and (E) Pacific Ocean (BIOSOPE), (B) and (F) AMT 6, (C) and (G) AMT 11, and (D) and (Fl) AMT 15. Units of <PeC (left panels) are in mol e~ (mol C F 1), N03~ concentrations (right-hand panels) are in pmol L_1. Dashed lines denote transitions between different biogeographical provinces, where SPG is the South Pacific Gyre, FINLC FHigh Nutrient Low Chlorophyll areas, UPW denotes upwelling regions, AB is the Angola Basin, SAG South Atlantic, NAG the North Atlantic Gyre, EM the Eastern Margin off Western Europe, and CZ is the Suptropical Convergence zone. Note, difference in scales on both axes and the colour contours.doi:10.1371/journal.pone.0058137.g003

confirms that user specific differences in m ethodology d id not appear to have a systematic influence.

Relationships between environmental conditions and®e,c

Clustering all available data based on the inherent environm ental characteristics demonstrated that the dependence o f & e c on environm ental gradients cannot be resolved at the level o f the ind iv idua l studies. C luster analysis therefore enabled us to objectively identify how to pool data across the m ultidim ensional environm ental variable m atrix (Table 4) to fu rthe r examine va riab ility o f & e c- Perm utation tests conducted on the ind iv idua l clusters (i.e. clusters a -q , Table 4) indeed showed that often m ultip le variables com bined were significantly correlated w ith & e c (BEST, p < 0.05 or ƒ><().01), and that d ifferent variable com binations were d riv ing c in d ifferent clusters and regions (Table 5). These ‘best’ variable combinations were entered in to m ultip le linear regression to derive algorithms fo r the p red ic tion o f ib .c ,

T able 3. Eigenvectors o f Principal Component Analysis (PCA).

Variable PCI PC2 PC3

Temperature -0.476 0.111 -0.184

Salinity — 0.325 -0.509 0.054

N OF 0.337 0.541 -0.087

po43~ 0.399 0.491 - 0.020

Chia 0.464 0.232 0.020

Kd 0.424 0.373 -0.091

Ç 0.001 0.023 -0.973

Values in bold represent variables w ith the highest coefficients for each PC. Kd is the vertical attenuation coefficient o f photosynthetically available radiation and £ is optical depth.doi:10.1371 /journal.pone.0058137.t003




Table 4. Samples grouped according to similarities in environmental conditions as determ ined by cluster analysis and SIMPROF.

Cluster Description n

a Ariake Bay + Massachusetts Bay 2

b Gulf o f Finland (SUPREM011) 22

c Gulf o f Finland 2000 (mid-Apr to mid-May) 5

d Bedford Basin (mid-Mar to end o f Apr) 9

e Bedford Basin (Feb to mid-Mar), MA Bay, North Sea CEND0811, AMT6 UPW and Pacific Ocean UPW 67

f Baltic proper (SYNTAX2010) 22

g Gulf o f Finland 2000 mid May, mid and late Jul, early Aug 5

h Gulf o f Finland late May to early Jul, Aug to Oct 10

i North Sea CEND0811 (Pen Field, South Pen Field, D366 (Falmouth, Southern North Sea, Flelgoland) 5

j Offshore DCM o f AMT6, 11, 15 & Pacific Ocean, Celtic Sea JR98 DCM 49

k Celtic Sea D246, AMT 6 European shelf edge surface 9

I AMT6, 11 European shelf edge intermediate depths, Celtic Sea JR98 DCM, CEFAS0811 (North Dogger), D366 (Central North Sea) 17

m AMT6 open ocean, surface & intermediate depths, AMT 11 open ocean intermediate depths, Pacific Ocean SPG surface, FINLC surface 54

n Celtic Sea JR98 (Irish Sea), North Sea (D366 Falmouth, St George Straight, Mingulay, Bay o f Biscay 12

o AMT6, 11.15 open ocean surface, D366 Central North Sea, Celtic Sea (JR98) surface 45

SPG South Pacific Gyre, FINLC High Nutrient Low Chlorophyll region, DCM Deep Chlorophyll Maximum; n is the number o f samples per cluster. doi:10.1371/journal.pone.0058137.t004

3-D Stress: 0.06

Studies▼ AMT 6▼ AMT11

▼' V AMT15• CS-D246o CS-JR98

Pacific+ North SeaX UK-OA▲ Baltic Sea▲ GoF 2000A GoF 2011♦ MA Bay• Bedford B* Ariake B.

V ► ♦

/K» T oV

<*>

B

A A

i h

DClusters

• a• bo c• do e• fo go h• i• jo k• I• m0 no 0

Figure 4. Three-dimensional non-metric Multidimensional Scaling (nMDS) ordination of the environm ental conditions of all field campaigns. Panels (A), (B) and (C) are the x-y, x-z and y-z plane o f the nMDS 3D ordination, respectively. (D) is the x-y plane where the symbols denote the various clusters as determined by cluster analysis in combination with similarity profile (SIMPER) tests. Clusters were significantly different from another (SIMPROF, p<0.005). For an overview of sample groupings according to these clusters see details in text and Table 4. doi: 10.1371/journal.pone.0058137.g004




SalinityTemperature

H I r h r lm

r - < I ~ r - , < V .^1a b c d e f g h i j k l m n o a b c d e f g h i j k I m n o a b c d e f g h i j k l m n o

Figure 5. Variability in the electron requirement for carbon fixation (0 e,c) and environm ental conditions within d ifferent clusters of samples. Clusters were generated by cluster analysis combined with a SIMPROF test. For samples contained in each cluster see also Table 4. Values are means and error bars are standard deviations (with n of 2-67, see Table 4) of <Pe C (mol e~ mol CT1), temperature ( C), salinity, chlorophyll a (Chi a, mg nrT3), nitrate and phosphate (pmol L_1), the vertical attenuation coefficient of photosynthetically available radiation (Kd, nrT1), optical depth £, (dimensionless) and sampling depth (z, in meters). doi:10.1371/journal, pone.0058137.g005

which, in m any cases, resulted in significant relationships w ith <&e¡c (M L R , p < 0.05) (Table 5). S ignificant relationships (R2 <0.05) between <&e¡c and environm ental variables existed in the G u lf o f F in land (cluster b), Bedford Basin (cluster d), European shelf waters (cluster n) and offshore samples from interm ediate depths (cluster m), surface waters (cluster n) and the deep ch lorophyll m axim um (D C M ) (cluster j). Surprisingly, P O 4 0 , ra ther than N O 3 availability, was often pa rt o f the variable combinations showing the strongest association w ith 0 e c, usually in add ition to temperature a n d /o r salinity. The strong relationships between

and P O 4 0 and lack thereo f w ith N O 3 m ay be due to N O 3

concentrations being close to or below the detection lim it in open ocean waters and, during the summer months, often also in shelf waters, w h ich compromises the detection o f relationships w ith <Zh,c. Thus, N O 3 ava ilab ility m ay still be an im portan t determ inant o f & e c in these waters, bu t w ith PC) 1 ’ being the only m acronutrient le ft in our analysis, on ly the la tte r was pulled out. Despite the distinct clustering o f samples from the fron ta l and upwelling regions, the Baltic Sea and European shelf edge, no

significant M L R existed between <Pe c and environm ental variables in any o f these regions (clusters e—i, k, 1).

T o develop region-specific algorithms, we also grouped the clusters according to m eaningful water masses and biogeographic regions resulting in 11 significant algorithms covering the G u lf o f F in land, the Baltic Sea as a whole (i.e. inc lud ing the G u lf o f F inland), European shelf waters (including the Northeast A tlantic), N orthwest A tlan tic shelf waters, the equatoria l A tlan tic and the South A tlan tic Ocean. A lthough some o f these region-specific relationships exhibited a low I t . they were h igh ly significant. The relationships between <t>^c and environm ental variables were strongest in the G u lf o f F in land and the Pacific Ocean (i?"<0.05), fo llowed by the N orthwest A tlan tic shelf samples (i.e. Massachusetts Bay + Bedford Basin). Relationships fo r the A tlan tic were m uch less pronounced (R2 <0 .22), albeit h igh ly significant (/><0.01). Once again, P O 4 0 ra ther than N O 3 ava ilab ility appeared to p lay a greater role in these M LR s. O n a global scale, M L R yielded an R 2 o f 0.038, bu t w ith both N O 3 and P O 1 ’ con tribu ting to the va riab ility in <//,/,.



60

50

40

30

20

10

0

30

25

20

15

10

5

0

30

25

20

15

10

5

0

30

25

20

15

10

5

0

6. I


• SMLSPG O DCM SPG A SMLHNLC ' zl DCM HNLC

^ . ♦ SML UPW .O DCM UPW

ÍI

50 100 150 200 0 10 15 20 25

B

+

SML AB DCM AB SML Eq DCM Eq

SML NAG DCM NAG SML EM DCM EM

20 40 60 80 100 120 7 14 21 28

• SMLSTCZ o DCM STCZ A SMLSAG ¿I DCM SAG♦ SML Eq

O DCM Eq ▼ SMLNAG V DCM NAG ■ SMLEM

_ l__________ I__________ I__________ I__________ I__________ I_

20 40 60 80 100 120 140 8

D

i

■ SMLSTCZ Ö DCM STCZ• SMLSAG o DCM SAG♦ SML Eq o DCM Eq

50 100 Depth [m]

150 200 0 2 3NO ' [pmol L'

4 5

aan electron requirem ent for carbon fixation (<Pe,à across d ifferent biogeographical provinces for four cruises. AeCleft panels) and N03~ (right panels) are shown for the Pacific Ocean (A and E), AMT 6 (B and F), AMT 11 (C and G) and AMT 15 (D and H), standard deviations with n = 3 to 14. SML denotes surface mixed layer, DCM the deep chlorophyll maximum, SPG is the South Pacific gh Nutrient Low Chlorophyll areas, UPW upwelling regions, SA South Africa, AB Angola Basin, Eq equator, NAG North Atlantic Gyre, EM n in the North East Atlantic, and STCS the Subtropical Convergence Zone. Note change in x- and y-axis scale. iurnal.pone.0058137.g006

vww.plosone.org 12 March 2013 I Volume 8 I Issue 3 I e58137


T a b le 5. Multivariate correlations and regressions between environmental variables and <&eiC for different clusters and regions.

BEST MLR

Cluster P P Variables Model R2 P

a - - - -

b GoF 0.238 >0.05 T, S, Kc, Ç —3.82T+2.80S+3.01 Kd-1 .12Ç 0.735 < 0.01

c GoF 0.538 >0.05 S n.s. 0.701 >0.05

d Bedford Basin 0.303 >0.05 T —1.31 T+11.78 0.710 <0.05

e Fronts and UPW 0.234 < 0.01 S n.s. 0.085 >0.05

f Baltic proper 0.135 >0.05 T n.s. 0.103 >0.05

9 GoF 0.430 >0.05 N, P, Kd n.s. 0.676 >0.05

h GoF 0.195 >0.05 S, PO43- n.s. 0.369 >0.05

i North Sea 0.539 >0.05 s, P0 43- , ç n.s. 0.965 >0.05

j DCM offshore 0.402 < 0.01 T, S —8.05T+8.89S — 3.23 0.251 < 0.01

k Celtic Sea, EUR shelf edge -0 .05 >0.05 N O T, Chia n.s. 0.150 >0.05

I EUR shelf edge 0.254 >0.05 % n.s. 0.012 >0.05

m Inter m. depths offshore 0.201 <0.05 S, N, P, Chia —50.85S+4.25NO3_+0.675PO43_ -11.81 Chla+322.8 0.315 < 0.01

n EUR shelf, North Sea 0.468 >0.05 T 6.40T -13.46 0.434 <0.05

o Surface offshore 0.190 >0.05 S, N o T , Chia —4.09S —5.83N03_+12.09Chla+32.59 0.269 < 0.01

Biogeographic regions

GoF (b c g h) 0.447 < 0.01 T, S, N O fi, Kd -0.741 T+14.8S+0.206N03" -2 .69Kd -26.14 0.561 < 0.01

Baltic + GoF (b c g h e) 0.236 < 0.01 S, N O T, Chia, Kd, Ç 1.30S -5.21 N 03" -0.61 Chia -2 .53Kd+0.77Ç+5.75 0.266 < 0.01

Shelf (i k I n) 0.245 < 0.01 T, N O T, P043b Chia, Ç 3.39T+1.44N03" -2.20PO43_ -1.42Chla -0.57Ç-0.25 0.273 <0.05

Offshore SML (m o) 0.173 <0.05 N 03- , P O yT Chia —5.97NO3~+9.62PO43_+0.33Chla +11.35 0.083 < 0.01

Offshore SML+DCM (j m o) 0.296 < 0.01 T —5.45T+37.51 0.090 < 0.01

EUR Shelf/NE Atlantic 0.142 <0.05 T, P043T Chia 0.04T -11.53P043_ - 1 .09Chla+10.99 0.207 < 0.01

NW Atlantic Shelf 0.329 < 0.01 T, S, N O T, P043~ —0.57T+0.25S+1.36N03_ -5 .6 6 P043 -+22.58 0.322 < 0.01

Equatorial Atlantic 0.205 <0.05 S 0.47S -0.723 0.090 <0.05

South Atlantic 0.395 < 0.01 N O T, P043b Chia 0.32NO3" - 1 3.07P043_+0.78Chla+6.20 0.213 <0.05

Pacific SPG +HNLC 0.233 >0.05 S, P O yT Chia —25.27S -94.29PO43_+29.29Chla+920 0.589 < 0.01

Pacific SPG + HNLC + UPW 0.183 >0.05 S, P O yT Chia 17.53S+33.11P043 -+14.6 Chla+1397 0.554 < 0.01

Global 0.114 < 0.01 T, N O T, P043~ 0.31 T -0 .4 9 N 0 3"+6.34P043_+4.05 0.038 < 0.01

Results o f the BEST test are the variable combinations resulting in the highest correlation coefficient (p) between the resemblance matrices o f the environmental data and 0 eC. Abbreviations: GoF Gulf o f Finland, UPW Upwelling region, SML surface mixed layer, DCM deep chlorophyll maximum, EUR European SPG South Pacific Gyre, HNLC High nutrient low chlorophyll area,. Cluster a was excluded from the BEST test and MLR due to its small sample size (n = 2). doi:10.1371 /journal.pone.0058137.t005

Discussion

Reconciling electron transfer and carbon fixationW hile em pirica l evidence demonstrates that fluorescence-based

measures o f the P S II photochem ical efficiency, and hence ETRs, are linearly related to net o r gross carbon fixa tion rates under m any conditions [16,47,54], there are still considerable uncertainties about what drives & e C as well as differences between NPP and GPP. G row th rate dependent d ifferentia l a llocation o f fixed carbon and varying lifetimes o f interm ediate products may cause large discrepancies between NPP and GPP, w hich short term 1 4 C 0 2 uptake measurements may not capture [19,20], W h ile it is generally accepted that short-term 14C incubation techniques approxim ate GPP, Halsey et al. [19,20] demonstrated that this is on ly the case in fast growing, nutrient-replete phytoplankton. In nu trien t lim ited cells, on the other hand, short-term 14C incubations equate NPP o r p rim a ry p roductiv ity rates somewhere between NPP and GPP. Failure to accurately quantify GPP w ill inevitab ly affect M ost data published in the past and alsothose used in our meta-analysis compared E T R to relatively short

te rm 14co2 uptake rates, so that most <Pe c values available to this day w ould provide a conversion from E T R to GPP or to something between NPP and GPP. ETnder most conditions, NPP differs from GPP by a factor o f 2 to 2.5 [55,19,20], fo r w h ich the current $>r c values cannot account. Conversion o f E T R to NPP and assessments o f the potentia l e rro r in & e c and subsequently NPP due to em ploying short te rm 14C incubations w ill require fu rther investigation and is the focus o f ongoing work.

ETnder ‘op tim a l’ grow th (where NPP ~ GPP), and accounting fo r electron sinks associated w ith nu trien t reduction, the slope o f the linear relationship between E T R and NPP should y ie ld values fo r <Pe c o f 4—6 m ol e- (mol C ) _ 1 [16 -18 ], T o date, most FR R f- based studies use an electron requirem ent fo r carbon fixa tion o f 5 m ol e (mol C) 1 to convert ETR s to carbon fixa tion rates, assuming (i) that at least 4 e_ transported through PS II are required per 0 2 molecule produced and (ii) tha t 1-1.5 m ol o f 0 2

is produced fo r each m ol C 0 2 fixed, i.e. the photosynthetic quotient takes values o f 1-1.5 m ol 0 2 (mol C 0 2) 1 [56], Indeed, some studies here confirm ed $>e c equal (or close to) 4—6 m ol e-




(mol C) *). However, the vast m a jo rity o f estimates fo r <t>e c are considerably higher, sometimes reaching extremes o f > 5 0 m ol e (mol C) 1 (e.g. B IO S O P E Pacific Ocean Cruise) or a lternatively &e,c o f < 5 m ol e (mol C) 1 G u lf o f F in land 2000 and A M T 15. Thus, fo r m any cases, application o f an assumed value o f 4 -6 m ol e- (mol C ) - 1 to F R R data w ould yie ld erroneous estimates o f C- uptake (w ith in the lim ita tions o f the 1 4 C-uptake itse lf for quantify ing C -uptake).

Values o f 0 e c > 5 m ol e- (mol C ) - 1 have been observed previously and potentia lly reflect processes that act to decouple ETR s from C -fixa tion , such as photorespiration [32], ch loror- espiration v ia a plastid te rm ina l oxidase (P T O X ) [33,34] and M eh le r reaction [32]. W e re tu rn to this issue in the fo llow ing sections. In contrast, values o f 0 „ c < 5 m ol e (mol C) 1 are more d ifficu lt to reconcile w ith biophysical and physiological processes, potentia lly ind ica ting the magnitude o f rem ain ing m ethodological discrepancies in deriv ing such as incorrect assumptionsconcerning the value or va riab ility o f nPSn (see below), inadequate spectral correction o r rem ain ing non-systematic errors in both carbon fixa tion and fluorescence estimates. There were two studies w ith a high p roportion o f <&ec values < 5 m ol e (mol C) *: the A M T 15 cruise (H ickm an et al. unpublished) and the G u lf o f F in land 2000 study. U n like during the A M T 15 cruise where < 5 m ol e- (mol C )- occurred on ly in samples from the D C M , we could not identify any consistency in the occurrence o f such low values that could be related to sample handling and processing in the G u lf o f F in land 2000 data. Thus, the h igh p ropo rtion o f low <&e c values in the G u lf o f F in land m igh t po in t to systematic errors in the E T R calculations, i.e. in the component values fo r ffpsn, npsn a n d /o r E, values tha t are often assumed (or no t well measured). W h ils t values o f ffpsn measured by F R R f have been shown to m atch independent b io-optica l measurements well [46], we have to assume that the b io-optica l instrum entation used to quantify bo th E and ffpsn have been appropria te ly calibrated. In the G u lf o f F in land [29], values o f ffpsn typ ica lly range from ca. 150-300 A 2 quantum - 1 , w h ich is on the low er end o f the range o f values expected fo r assemblages dom inated by diatoms, dinoflagellates, cryptophytes and cyanobacteria (data not shown), b u t still w ith in the range com m only observed fo r such species [38]. Absolute underestimations o f ffpsn as a result o f erroneous instrum ent calibrations are therefore un like ly a significandy con tribu ting factor. However, in cases when cyanobacteria dominate, ETR s may be underestimated due to the saturating ligh t blue L E D pulse being ineffic ient at d riv ing reaction centre closure, resulting in low values. Furtherm ore, assumption o f a constant npsn w ill in troduce errors due to taxonom ic va riab ility in physiological traits [46]. M oreover, phytop lankton cells tend to change the size o f the ir photosynthetic units in response to ligh t and nu trien t ava ilab ility [57-59] and vertical m ix ing [60]. Hence npsn can vary among species by more than a factor o f four, from 0.0010 to 0.0042 m ol R C (mol Chia ) - 1 [46]. Raateoja et al. [29] assumed a constant npsn o í 0.002 m ol R C (mol Chia)-1 , w h ich is representative o f the eukaryote phytop lankton com m unity observed in the ir study (mostly diatoms, dinoflagellates and cryptophytes; b u t not typ ica lly when cyanobacteria dominate. Consequently, increasing assumed values o f nPSn in the study o f Rateeoja et al. [29] to account fo r the presence o f cyanobacteria [8,46] w ould increase ETR s and hence potentia lly $>e C to values > 5 m ol e- (mol C)- 1 .

W e assessed the potentia l e rro r in $>e C due to the use o f a constant nPSn in some o f the data sets presented in this study by com paring <£eC based on a constant nPS_n w ith c values calculated from E T R where npsn was either measured v ia oxygen flash yields (Bedford Basin data) o r calculated using a fluorescence

based a lgorithm [7] (CEND0811, U K -O A D366). These com parisons showed that a constant h rh j may lead to an underestimate o f <Pe>c by 28-47% (Fig. S2, Table S I in Append ix SI), w hich could thus, at least in part, explain the low in the G u lf o f F in land data. The npsn values estimated fo r the N o rth Sea spanned a range o f 0.0018 to 0.0056 m ol R C (mol chia) \ w hich is comparable w ith the range observed fo r eukaryotic and p rokaryotic phytop lankton [46,47,58], bu t higher, by ca. a factor o f 2, than npsn d irectly measured by M oore et al. [11] in European shelf waters at a sim ilar tim e o f year. C learly fu rther direct measurements o f npsn and valida tion o f algorithms proposed for estimating the value o f this variable [7] are required before a more robust assessment o f the e rro r in calculated w hich isassociated w ith npsn cun be achieved. Such measurements are the focus o f ongoing work.

O the r sources o f discrepancies between electron transfer and carbon fixa tion may stem from differences among protocols for E T R -lig h t response curves (rapid lig h t curves vs. steady state ligh t curves) [61], deviations in the timescales o f in situ E T R and laboratory 1 4 C-photoynthesis measurements [43,61], discrepancies in the E k values derived from E T R and C 0 2 uptake [52 -54 ], or the effects o f water co lum n structure and subsequent changes in lig h t ava ilab ility on the ‘shape’ o f E T R ligh t response curves [8 ] . For the data included in the present study and the associated differences in methodology w ith regard to E T R calculations (ffpsn, npsn) and 14C incubation techniques, we estimated that, in worst case scenarios, m ay be over- o r underestimated by as m uch as 53% (data no t shown). A detailed assessment o f these m ethodological differences between techniques and studies is beyond the scope o f this study, and we refer the reader to a previous comprehensive reviews o f these issues [8,38]. However, we note that va riab ility in estimated values o f (t>e c was lower than that observed across all studies (Fig. 2) fo r the two new studies included here, w h ich were perform ed under the most contro lled conditions (S Y N T A X 2010 and N o rth Sea C EN D0811), i.e. were E T R and C 0 2 fixa tion were simultaneously measured on the same sample. Thus, the extent to w hich the extreme values a n d /o r va riab ility in

w hich is observed in some other studies, represents rem ain ing errors in either carbon fixa tion or E T R estimates remains unclear.

Environmental regulation o f < P e CData presented in this study focused almost exclusively on the

apparent effect o f readily measurable environm enta l variables on In fo rm a tion on phytop lankton com m unity composition

could no t be rou tine ly included and we assumed that any change in taxonom y was inherently accounted fo r in the environm ental descriptors. Indeed, environm ent often leads to d istinct structural differences o f the photosynthetic apparatus and acclim ation responses amongst phytop lankton from different biogeographic regions [34,62-67] and, towards unique, often consistent fluorescence "signatures" o f PS II [10,11,38,68-70]. Future assessments o f the in ter- and intraspecific va riab ility in & ec w ill, o f course, be necessary to fu lly resolve the mechanisms responsible fo r changes in 0 e c and the derived algorithms (Table 5), w h ich are pure ly based on statistical models and should therefore no t be interpreted from a mechanistic perspective. F rom the relatively few available studies on phytop lankton cultures grow n under various conditions to date, (fre e usually does no t exceed values o f ~ 2 5 m ol e (mol C) 1 as opposed to estimates o f up to 54 m ol e m ol C 1 from natura l comm unities (reviewed in Suggett et al. [8 ]). Low er m axim al values for phytop lankton cultures could indicate a less extreme range o f grow th environments tested in the labora tory as compared to natura l conditions, such as, the h igh-ligh t, low- n u trien t conditions o f the surface open ocean that are d ifficu lt to




reproduce in the laboratory. Furtherm ore, labora tory studies often examine cells that are in ‘steady state’ grow th whereas natural comm unities m ay persist under (rapidly) changing environm ental conditions. In the la tter case, non-steady state conditions could lead to a strong and transient uncoup ling o f electron transfer and carbon fixa tion [10,71,72], w h ich w ould suggest tha t the p rim ary source o f va ria tion in & ec is the environm ent rather than taxonom ic variab ility . The effect o f grow th rate on our ab ility to measure GPP by short-term 14C incubations under natura l conditions may fu rther contribute to the w ider range o f <l>e>c values from fie ld observations relative to culture-based 0 , r [19,20],

Considerable va riab ility o f <l>e>c existed not on ly spatially bu t also tem porally. G iven the large range o f <&e c fo r the entire data set, i t is notable that mean & e c in all b u t one o f the tim e series studies (Massachusetts Bay) never exceeded 10 m ol e (mol C) ’ . The greater variance in the Massachusetts Bay tim e series study could, on the one hand, have been related to its longer dura tion (9 month) relative to the other tim e series (1.5-6 month) and, on the other hand, to its topography; while the G u lf o f F inland, Bedford Basin and Ariake Bay, are a ll surrounded by land fo rm ing at least pa rtia lly land-enclosed basins, Massachusetts Bay is relatively exposed and more strongly influenced by exchange o f water w ith the A tlan tic Ocean. Nevertheless, seasonal changes appear to influence the va riab ility in c a lthough m any areas are still under-sampled.

In the G u lf o f F inland, fo r example, vertica l m ix ing and the depth o f the surface m ixed layer change across d ifferent seasons [73-75] and w ith i t the optical and physicochemical properties o f the water column, that is, lig h t and nu trien t ava ilab ility o f the extant phytop lankton com m unity [29]. The clustering o f the G u lf o f F in land samples - in to a late w in te r (b), spring (c), summer (g) and sum m er/early autum n (h) m irro rs these physicochemical changes. However, the h igh ly dynam ic nature o f this system makes the development o f predictive algorithms fo r the entire Baltic Sea (including the G u lf o f F inland) challenging. The best predictors o f 0 „ c fo r the whole region were temperature, salin ity and P 0 4 3 , resulting in a relatively weak (R2 = 0.266) albeit h igh ly significant relationship. For the G u lf o f F in land on its own, however, the relationships were m uch stronger (R2 = 0.561, ƒ><().01), ind ica ting tha t the G u lf o f F in land should perhaps be treated as its own region.

D is tinc t tem poral changes in environm ental gradients also existed in Bedford Basin, where in itia l correlations (Table 2) showed significant relationships o f <&e c w ith salinity, w h ich may have bo th ind irec t (via alterations o f density and water colum n stratification) and d irect (osmotic) effects on c. The com bination o f low nutrients coupled w ith sudden h igh ligh t ava ilab ility due to elevated freshwater discharge after snow m elt and ice breakup could cause an up-regulation o f electron, A T P a n d /o r reductantconsuming pathways (e.g. photorespiration and M eh le r reaction). Conceivably, sim ilar changes in salinity could also d irectly affect <&e c by causing osmotic stress o r an increase in respiration o f the extant phytop lankton com m unity [76]. Bedford Basin phytop lankton com m unity composition often shifts from diatoms and dinoflagellates to chlorophytes after the snow m elt (Suggett & Forget unpublished) [77], m atching the increase o f í>e>c■ Thus, in Bedford Basin m ultip le environm ental factors (salinity, temperature, stratification) were like ly at play, poten tia lly exerting ligh t and osmotic stress on the extant, b u t changing, phytoplankton com m unity [77 -80 ]. Resolving the underly ing mechanisms responsible fo r seasonal changes in c in remains a challenge and w ill require teasing apart daylight hour effects from temperature effects and other physicochemical variables.

M any samples from Massachusetts Bay and other open ocean areas fell in to one cluster w ith the Bedford Basin data (Table 4, Fig. 5). The Bedford Basin spring cluster (d) was characterized by h igh ch lorophyll concentrations, whereas samples in the late w in te r cluster (e) had low ch lo rophyll concentrations and temperatures m ore sim ilar to those in fron ta l and upwelling regions o f the N o rth Sea, and the A tlan tic and Pacific Oceans. The w ide range o f environm ental gradients observed in these clusters may be the reason w hy M L R d id not re tu rn any significant relationships between c and environm ental variables. Furtherm ore, such relationships m ay not necessarily be linear [81] and other models w ill have to be tested in the future.

M ost o f the other studies, also extended across m ultip le biogeographic provinces (sensu Longhurst [82]) o r d istinct environments water masses (S M L /D C M ), and although there seemed to be consistent overlaps between nu trien t deplete regions and areas o f h igh & ec (Fig. 3), the absolute nu trien t concentration was often not correlated w ith & e c (Table 2). Im po rta n t examples are the open ocean regions, h igh ligh ting how environm ental fo rc ing can confound establishment o f strong relationships between & ec and measurable environm ental variables at the basin scale: & e>c often declined w ith increasing depth and towards the nu tridm e; even so, <&ec was often not significantly correlated w ith measured nu trien t concentration most like ly reflecting differences in nu trien t ava ilab ility between the gyres, H N L C areas and coastal upw elling regions. A dd itiona lly , nu trien t stocks are no t necessarily a good index o f the level o f nu trien t stress w ith in a population, w h ich w ill depend rather on the overall (re-)supply rates o f the nu trien t relative to the demands o f the extant com m unity. In the iro n and n itrogen deplete SPG [12] and the low -iron H N L C region o f the Pacific Ocean [83], <&ec declined when N 0 3_ concentrations were above the detection lim it (usually at the deep ch lorophyll m axim um ). A p a rt from the extreme outlie r o f <&e¿ > 50 m ol e (mol C) *, highest <&e c values were observed in the upwelling region and the gyre corresponding to samples from a lo w -N 0 3_ surface water lens w ith high irradiances. Beneath this surface water lens, N 0 3 increased while lig h t levels dropped and i>e>c declined.

Iro n and n itrogen lim ita tion may be expected to influence the photosynthetic apparatus in d ifferent ways, w ith iron deficiency causing a preferentia l decline in iron rich-ce llu la r components (e.g. PS I, P S II and cytochrom e b 6 f) [56 ,84-86 ]. Thus, d ifferent forms o f nu trien t lim ita tio n can induce differences in the fluorescence signatures o f the resident phytop lankton com m unity [67,72] and subsequently in <£e C. Iro n lim ited phytop lankton in H N L C regions, fo r instance, m y express the ch lo rophyll-b ind ing prote in IsiA , w h ich reduces the apparent PS II photo synthetic efficiency and p rim a ry p roductiv ity rates norm alized to ch lorophyll [87,88]. The la tte r w ould cause <&ec to increase. N u trien t lim ita tio n in com bination w ith h igh-ligh t stress in surface waters m ay hence have been responsible fo r the highest c obtained in these regions, suggesting a h igh degree o f uncoupling between electron transfer and carbon fixa tion. S im ilarly, variations in nitrogen assim ilation and n itrogen fixa tion require large amounts o f A T P also leading to an uncoupling o f electron transport from carbon fixa tion [67,89,90].

ConclusionsConversions o f FRRf-based E T R estimates to carbon-specific

rates o f NPP depend on our ab ility to accurately pred ic t <l>e>c fo r given environm ental condition. Th is present study shows, fo r the first tim e, the extent o f va riab ility o f <j, a lbe it w ith in the past m ethodological constraints o f accurate quantifica tion o f both E TR s and NPP. M ost data on & ec available to this day are based




on short-term 14C incubations which, depending on algal grow th rates, may capture GPP rather than NPP. Further studies are needed to fu lly resolve the discrepancies between derived from GPP o r NPP in bo th cultures and natura l communities. Nevertheless, our w ork shows tha t some o f the va riab ility in c can be linked to environm ental variables that are routine ly measured. G iven the observed variab ility , i t is h igh ly un like ly that a global approach o r a lgorithm can be produced w h ich captures a h igh p roportion o f the variance in c- However, w ith the present study we provide firm evidence that such algorithm s can in fact be generated for biogeographic regions and d istinct water masses, b ring ing us closer than ever to p red ic ting carbon uptake from ETRs. Independent va lidation o f these algorithm s is still required, and, we m ay expect they w ill be refined as more data accumulate. Im portandy, our study provides methodology fo r future data collection and integration to im prove <&e c algorithms. C learly, developing standardized protocols w ould greatly facilitate in te rcomparisons o f studies and reduce some o f the potentia l e rro r due to methodological differences. Assessing the role o f phytoplankton taxonom y, development o f non-linear models and testing o f the present algorithm s on novel, independent data represent necessary future steps to im prove the level o f accuracy.

O ne o f the m ajor rem ain ing challenges in u tilis ing these algorithm s is defin ing the (biogeographic) regions, while keeping in m ind tha t the ir boundaries m ight actually shift in space and time. Hence caution is still required in widespread application o f the current algorithms. Even so, the appearance o f clear patterns and p ro o f o f predictive power o f environm enta l variables in the present article provides strong support and m uch im proved confidence for successful conversion o f ETR s to carbon fixa tion rates at a much greater spatial and tem poral resolution than current 14C fixa tion approaches.

Supporting Information

Figure SI Principal component analysis including the environmental data, location data and methodological information (A) and with the methodological differences excluded from the analysis (B).

References1. L indem an R L (1942) T h e trophic dynam ic aspect o f ecology. Ecology 23:399—

418.2. F ield C B , Behrenfeld M J , Randerson J T , Falkowski P G (1998) Prim ary

production o f the biosphere: Integrating terrestrial and oceanic components. Science 281: 2 3 7 -2 4 0 .

3. C a rr M -E , Friedrichs A M , Schm eltz M , A k a M N , Antoine D , et al. (2006) A com parison o f global estimates o f m arine p rim ary production from ocean color. D eep Sea Res I I 533 , 741770, d o i:1 0 .1016 /j.d sr2 .2006 .01 .028 .

4. Longhurst A , Sathyendranath S, P latt T , C averh ill C (1995) A n estimate o f global p rim ary production in the ocean from satellite radiom eter data. J Plankton Res 17: 124 5 -12 7 1 .

5. Behrenfeld M J , Boss E , Siegel D A , Shea D M . (2005) Carbon-based ocean productiv ity and phytoplankton physiology from space. G lobal Biogeochem Cycles 19: G B 1006, d o i:1 0 .1 0 2 9 /2 0 0 4 G B 0 0 2 2 9 9

6 . Schreiber U , N eubauer C , Schuwa U (1993) P A M fluorom eter based on m edium -frequency pulsed Xe-flash measuring light: A highly sensitive new tool in basic and applied photosynthesis research. Photosynth Res 36:65—72.

7. O xbo ro ugh K , M o o re M G , Suggett D , Lawson T , C h a n H G , et al. (2012) D irect estim ation o f functional P S II reaction centre concentration and P S II electron flux on a volum e basis: a new approach to the analysis o f Fast Repetition Rate fluorom etry (F R R f) data. L im n o l O ceanogr M e th 10: 1 42 -1 54 .

8 . Suggett D J, M o o re M , G eider R J (2011) Estim ating aquatic productiv ity from active fluorescence measurements. In : Suggett D J, Prásil O , Borow itzka M A , editors. C h lorophyll a fluorescence in aquatic sciences: methods and applications. Springer D ordrech t, pp. 103—127.

9. Falkowski P G , K o lb e r ZS (1995) Varia tions in C h lorophyll Fluorescence Yields in Phytoplankton in the W o rld Oceans. Austr J P lant Physiol 22: 3 4 1 -3 5 5 .

10. M o o re M G , Suggett DJ, H o lligan P M , Sharpies J , Abraham s E R (2003) Physical controls on phytoplankton physiology a t a shelf front: a fast repetition-rate fluorom eter based field study. M a r Ecol Prog Ser 259: 29M-5.

(TIF)

Figure S2 Comparison o f (in mol e (mol C) ’)calculated with an »t/>s// = 0.0020 mol RC (mol chia)-1 and where nPsu was measured with oxygen flash yields (Bedford Basin) or by FRR fluorometry according to Oxborough et al. [11] (UK-OA D366 and North Sea CEND0811). Bold line represents the regression equation for all three studies combined: i>e>c constant nPsn = 0.617 x { f ie,c measured nPSII)+2.165 (R2 = 0.807, n = 1 1 0 , p<0 .05 ). Regression coefficients are shown in Table S I in Append ix S I.(TIF)

Appendix SI Results o f PCA including methodological and location variables and assessm ent o f the effect of variable nPSII on Phie,C.(D O C X )

AcknowledgmentsWe are extremely grateful to Sarat C. Tripathy and Prof. Jo ji Ishizaka (HyARC Nagoya University), Christopher Melrose (NOAA Northeast Fisheries Science Center, (Mika Raateoja (Finnish Environm ent Institute) and T im Smyth (Plymouth M arine Laboratory) who very graciously provided additional environmental data from their previously published work, as well as Rob Thomas a t the British Oceanographic D ata Centre for supporting the retrieval o f the A M T and JR 98 data and W endy Lea at the Massachusetts W ater Resources Authority for providing the Massachusetts Bay data. We also acknowledge the mission scientists and principal investigators of M O D IS and SeaWIFS who provided data via the Giovanni online data system, developed and maintained by NASA GES DISC. We are also grateful for the comments provided by AJ Milligan and an anonymous reviewer.

Author ContributionsAdvice with statistical analysis: RJG. Conceived and designed the experiments: EL J K DJS GS RM F. Performed the experiments: EL JK GS PY M -H F SS EC O P AEH C M M DJS RM F. Analyzed the data: EL DJS. Contributed reagents/m aterials/analysis tools: RM F. W rote the paper: EL DJS.

11. M o o re M C , Suggett DJ, H ic km a n A E , K im Y -N , Tw ed d le J F , et al. (2006) P hytoplankton photoacclim ation and photoadaptation in response to environm ental gradients in a shelf sea. L im nol. O ceanogr. 5: 936—949.

12. Behrenfeld M J , K o lb er Z S (1999) W idespread iron lim itation o f phytoplankton in the South Pacific O cean. Science 283: 840—843.

13. M a c In ty re H L , C u llen JJ (2005) Using cultures to investigate the physiological ecology o f microalgae. In: Andersen R A , editor. A lga l culturing techniques. Elsevier Academ ic Press, N e w Y ork . pp. 287—236.

14. V en ric k E L , Beers, J R , H e in b o ke lJF (1977) Possible consequences o f contain ing m icroplankton fo r physiological rate measurements. I E xperim ental M a r Biol Ecol 26: 5 5 -7 6 .

15. Fogg G E , C a lva rio -M a rtin e z O (1989) Effects o f botde size in determ inations o f p rim ary productivity by phytoplankton. Hydrob io log ia 173: 89—94

16. G enty B, Briantais J -M , Baker N R (1989) T h e relationship betw een the q u an tu m y ie ld o f photosynthetic e lectron transport and quench ing o f chlorophyll fluorescence. B iochim Biophys Acta 990: 87—92

17. Edwards G E , Baker N R (1993) C a n C O 2 assimilation in m aize leaves be predicted accurately from chlorophyll fluorescence analysis? Photosynth Res 37: 8 9 -1 0 2 .

18. Suggett DJ, M a c In ty re H L , K a n a T M , G eider RJ (2009) C om paring electron transport w ith gas exchange: param eterising exchange rates betw een alternative photosynthetic currencies fo r eukaryotic phytoplankton. A quatic M ic ro b ia l Ecol 56: 1 47 -1 62 .

19. Halsey K H , M illig an AJ- Behrenfeld M J (2010) Physiological optim ization underlies grow th rate-dependent chlorophyll-specific gross and net p rim ary production. Photosynth Res 103: 1 25 -1 37 .

20. Halsey K H , M illig an AJ- Behrenfeld M J (2011) L ink ing tim e-dependent carbon- fixation efficiencies in D unalie lla tertiolecta (Chlorophyceae) to underlying m etabolic pathways. J Phycol 47: 66—76.




21. Boyd P W , A iken J , K o lb e r ZS (1997) Com parisons o f radiocarbon and fluorescence based {pum p and probe) m easurem ents o f phy top lankton photosynthetic characteristics in the Northeast A dantic O cean. M a r Ecol Prog Ser 149: 2 1 5 -2 2 6 .

22. Corno G , Letelier R M , A bbott M R , K a r l D A (2006) Assessing p rim ary production variab ility in the north Pacific subtropical gyre: a com parison o f fast repetition rate fluorom etry and 14C measurements. J Phycol 42:51—60.

23. K ro m k am p J C , D ijkm a n N A , Peene J , Simis S G H , Gons H J (2008) Estim ating phytoplankton p rim ary production in Lake Ijsselmeer {The Netherlands) using variable fluorescence {P A M -F R R F ) and 14C -up take techniques. E u r J Phycol 43: 3 2 7 -3 4 4 .

24. Suggett DJ {2000) V aria b ility o f phytoplankton productiv ity rates in the Adantic Oceans observed using the Fast Repeddon R ate fluorom eter. University o f Southam pton, U n ite d K ingdom , D octo ra l Thesis, 184 pp.

25. Suggett DJ, M o o re M C , M a ra ñ ó n E , O m ac h i C , V are la R A , et al. {2006) Photosynthedc electron turnover in the tropical and subtropical A dantic O cean. D eep Sea Res I I 5 3 :15 73 -1 5 92 .

26. H ic km a n {2007) T h e photophysiology and p rim ary productiv ity o f phytoplankton w ith in the deep chlorophyll m axim um . University o f Southam pton, School o f O cean and E arth Sciences, D octora l Thesis, 237pp.

27. Pem berton K L , C larke R , J o in t I {2006) Q uan tify ing uncertainties associated w ith the m easurem ent o f p rim ary production. M a r Ecol Prog Ser 322: 51—59.

28. Sm yth TJ, Pem berton K L , A iken J , G eider RJ {2004) A methodology to determ ine p rim ary production and phytoplankton photosynthedc parameters from fast repeddon rate fluorom etry. Jo u rn al o f Plankton Research 26: 1337— 1350.

29. R aateo ja N , Seppälä J , Kuosa H {2004) B io-opdcal m odelling o f p rim ary production in the S W Finnish coastal zone, Baltic Sea: fast repeddon rate fluorom etry in Case 2 waters. M a r Ecol Prog Ser 267: 9—26.

30. M elrose D C , O v ia tt C A , O ’R e illy J E , B erm an M S {2006) Com parisons o f fast repeddon rate fluorescence estimated p rim ary production and 14C uptake by phytoplankton. M a r Ecol Prog Ser 311: 37—46.

31. T rip a th y SC , Ishizaka J , Fu jik i T , Shibata T , O ka m u ra K , et al. {2010) Assessment o f carbon- and fluorescence-based p rim ary productiv ity in Ariake Bay, southwestern Japan. Estuarine Coastal Shelf Sc 87: 163—173.

32. Badger M R , von C aem m erer S, Ruuska S, N akano H {2000) E lectron flow to oxygen in higher plants and algae: rates and control o f direct photoreduction {M eh ler reaction) and rubisco oxygenase. Phil T rans R Soc Lond D 355: 1433— 1446.

33. Bailey S, M ehlis A , M ackey K R M , C ard o l P, F inazi G , et al. {2008) A lternative photosynthetic electron flow to oxygen in m arine Synechococcus. B iochim Biophys A cta 1777: 2 6 9 -2 7 6 .

34. M ackey K R M , Paytan A , Grossman A R , Bailey S {2008) A photosynthetic strategy fo r coping in a high-light, low -nu trien t environm ent. L im n o l O ceanogr53: 9 0 0 -9 1 3 .

35. Holm es JJ, W eger H G , T u rp in D H {1989) C h lorophyll a fluorescence predicts total phytosynthetic electron flow to C O 2 or N O 3 / N O 2 under transient conditions. P lant Physiol 91:331—337.

36. Estévez-Blanco P, Cerm eño P, Espiñeira M , Fernández E {2006) Phytoplankton photosynthedc efficiency and p rim ary production rates estim ated from fast repetition rate fluorom etry at coastal em baym ents affected by upw elling {Rias Baixas, N W o f Spain). J Plankton Res 28: 115 3 -11 6 5 .

37. Behrenfeld M J , Prásil O , Babin M , Bruyan t F {2004). In search o f a physiological basis for covariations in light-lim ited and light-saturated photosynthesis. J Phycol 4 0 :4 -2 5 .

38. Suggett D J, M o o re M C , H ic km a n A E , G eider RJ {2009) In terpretation o f fast repetition rate {F R R ) fluorescence: signatures o f phytoplankton com m unity structure versus physiological state. M a r Ecol Prog Ser 376: 1—19.

39. K irk J T O {2010) L ig h t and photosynthesis in aquatic ecosystems, 3r edition. Cam bridge: C am bridge University Press. 662 p.

40. Suggett DJ, K ra a y ,G , H o lligan P, D avey M , A iken J , et al. {2001) Assessment o f photosynthesis in a spring cyanobacteria! b loom by use o f a fast repetition rate fluorom eter. L im n o l O ceanogr 46: 802—810.

41. K o lb e r Z S , Prásil O , Falkowski P G {1998) Measurem ents o f variable chlorophyll fluorescence using fast repetition rate techniques: defining m ethodology and experim ental protocols. B iochim Biophys Acta 1367: 88—106.

42. Suggett DJ, O xbo ro ugh K , Baker N R , M a c In ty re H L , K a n a T M , et al. {2003) Fast repetition rate and pulse am plitude m odulation chlorophyll a fluorescence measurements for assessment o f photosynthedc electron transport in m arine phytoplankton. E ur J Phycol 38: 3 71 -3 84 .

43. K ro m k a m p J C , Forster R M {2003) T h e use o f varia b le fluorescence measurements in aquatic ecosystems: differences between m ultiple and single turnover protocols and suggested term inology. E u r J Phycol 38: 103—112.

44. Suggett DJ, M a b e rly SC, G eider RJ {2006) Gross photosynthesis and lake com m unity m etabolism during the spring phytoplankton bloom . L im n o l O ceanogr 51: 2064—2076.

45. C u llen JJ, Davis R F {2003) T h e b lank can m ake a b ig difference in oceanographic measurements. L im n o l O ceangr B ull 12:29—35.

46. Suggett DJ, M a c In ty re H L , G eider RJ {2004) Evaluation o f biophysical and optical determinations o f light absorption by photosystem I I in phytoplankton. L im n o l O ceanogr M e th 2: 3 1 6 -3 3 2 .

47. K o lb e r Z S , Falkowski P G {1993) Use o f active fluorescence to estimate phytoplankton photosynthesis in situ. L im n o l O ceangr 38: 1646—1665.

48. T in g CS , O w ens G {1994) T h e effects o f excess irradiance on photosynthesis in the m arine d iatom Phaeodactylum tricornutum. P lant Physiol 106: 763—770.

49. Laney SR , Letelier R M , Desiderio R A , A bbott M R , K ie fe r D A , et al. {2001) M easuring the natura l fluorescence o f phytoplankton cultures. J Atm os O cean Techno l 18: 1 92 4 -34 .

50. M illig an AJ, Aparicio U A , Behrenfeld M J {2011) Fluorescence and nonphotochem ical quenching responses to simulate vertical m ixing in the m arine d iatom Thalassiosira weissflogii. M a r Ecol Prog Ser 448: 67—78.

51. Lewis M R , S m ith JG {1983) A Sm all V o lum e, S hort-Incubation -T im e M e th o d for M easurem ent o f Photosynthesis As A Function o f Inc iden t Irrad iance. M a r Ecol Prog Ser 13: 9 9 -1 0 2 .

52. M a c ln trye H L , K a n a T M , A n n in g T , G eider RJ {2002) Photoacclim ation o f photosynthesis irradiance response curves and photosynthetic pigments in m icroalgae and cyanobacteria. J Phycol 38: 17—38.

53. C larke K R , Som erfield IJ , G orley R N {2008) Testing o f null hypotheses in exploratory com m unity analyses: s im ilarity profiles and bio ta-environm ent linkage. J E xp M a r B io l Ecol 3 66 :5 6 -69 .

54. Baker N R , O xborough K {2004) C h lorophyll fluorescence as a probe o f photosynthetic productivity. In: Papageorgiou G C , G ovindjee, editors. C h lo ro phyll fluorescence: a signature o f photosynthesis. Springer, Dordrech t, pp. 65— 85.

55. W illiam s PJ le B, Robinson C , Sondergaard M , Jespersen A -M , Bentley T L , et al. {1996) A lgal 14C and total carbon m etabolism. 2. E xperim ental observations w ith the d iatom Skeletonema costatum. J Plankton Res 18: 1961—1974.

56. Laws E A {1991) Photosynthetic quotients, new production and net com m unity production in the open ocean. Deep Sea Res 30: 143—167.

57. Falkowski P G , O w ens T G {1980) Light-shade adaption - two strategies in m arine phytoplankton. P lant 66: 5 9 2 -5 9 5 .

58. Dubinsky Z , Falkowski P G , W y m a n K {1986) L ig h t harvesting and utilization by phytoplankton. P lant C e ll Physiol 27: 1 33 5 -13 4 9 .

59. Berges J A , Gharlebois D O , M a u ze ra ll D C , Falkowski P G {1996) D ifferentia l effects o f nitrogen lim itation on photosynthetic efficiency o f photosystem I and I I in microalge. P lant Physiol 110: 6 8 9 -6 9 6

60. K ro m k am p J , L im beek M {1993) Effect o f short-term variation in irradiance on light-harvesting and photosynthesis o f the m arine diatom Skeletonema costatum : a laboratory study sim ulating vertical m ixing. J G enera l M ic ro b io l 139: 2277— 2284.

61. Serôdio J , V ie ira S, C ru z S, Coelho H {2006) M icrophytobenthos vertical m igratory photoresponse as characterized by ligh t response curces. PhotosynthRes 90: 2 9 ^ 3 .

62. Sunda W G , H untsm an SA {1995) Photosynthetic architecture differs in oceanic and coastal phytoplankton. M a r C h em 50: 1 89 -2 06 .

63. Scanlan DJ {2003) Physiological diversity and niche adaptation in m arine Synechococcus. A d v M ic ro b Physiol 47: 1—64.

64. Strzepek R F , H arrison PJ {2004) Photosynthetic architecture differs in coastal and oceanic diatoms. N a tu re 431: 689—692.

65. C ard o l P, Bailleui B, R appaport F , Derelle E , Beal D , et al. {2008) A n original adaptation o f photosynthesis in the m arine green alga Ostreococcus. Proc N a tl A cad Sei U S A 105: 7 8 8 1 -7 8 8 6 .

66. Bibby T S , Zhang Y , C hen M {2009) Biogeography o f photosynthetic light- harvesting genes in m arine phytoplankton. PloS O N E 4: e4601.

67. Grossman A R , M ackey K R M , Bailey S {2010) A perspective on photosynthesis in the oligotrophic oeans: hypotheses concerning alternate routes o f electron flow. J Phycol 46: 6 2 9 -6 3 4

68. Behrenfeld M J , W orth ing ton K , Sherrell R M , Chavez FP, S trutton P, et al. {2006) Controls on tropical Pacific O cean productiv ity revealed through nutrient stress diagnostics. N a tu re 442: 102 5 -10 2 8 .

69. Behrenfeld M J , Halsey K H , M illig an AJ {2008) Evolved physiological responses o f phytoplankton to their integrated grow th environm ent. Phil T rans R Soc L on d B 363: 2 68 7 -27 0 3 .

70. L iu S -W , B-S Q iu {2012) D ifferen t responses o f photosynthesis and flow cytom etric signals to iron lim ita tion and nitrogen source in coastal and oceanic Synechococcus strains {Cyanophyceae). M a r Biol 159:519—532.

71. M asojidek J , G robbelaar J U , Pechar L , K ob lizek M {2001) Photosystem I I electron transport and oxygen production in natural waterbloom s o f freshwater cyanobacteria during a diel cycle. J P lankton Res 23: 57—66.

72. W agner H , Jakob T , W ilh e lm C {2006) Balancing the energy flow from captured light to biomass under fluctuating ligh t conditions. N e w Phytol 169: 95—108.

73. T a m m in e n T , Andersen T {2007) Seasonal phytoplankton nutrient lim itation patterns as revealed by bioassays over Baltic Sea gradients o f salinity and eutrophication. M a r Ecol Prog Ser 340: 1 21 -1 38 .

74. Lessin G , Lips I , Raudsepp U {2007) M ode llin g nitrogen and phosphorus lim itation on phytoplankton grow th in N a rva Bay, south-eastern G u lf o f Finland. O ceanología 49: 2 59 -2 76 .

75. van Beusekom JJE, M engedoht D , Augusting C B , Schilling M , Boersma M {2009) P hytop lank ton , p ro tozo op lank ton and n u trie n t dynam ics in the B ornholm Basin {Baltic Sea) in 2002—2003 during the G erm an G L O B E G Project., In t J E arth Sei 98: 2 5 1 -2 6 0 .

76. F lam eling IA , K ro m k am p J {1984) Responses o f respiration and photosynthesis o f Scenedesmus protuberans Fritsch to gradual and steep salinity increases. J Plankton Res 16: 1 78 1-17 9 1 .

77. L i W K W , H arrison W G , H e a d E JH {2006). C oherent assembly o f phytoplankton com m unities in diverse tem perate ocean ecosystems. Proc R oya l Soc B BiolSei 273{1596) 1 95 3 -19 6 0 .




78. W u Y , Peterson IK , T a n g C C L , P latt T , Sathyendranath S, et al. (2007) Th e im pact o f sea ice on the in itiation o f the spring b loom on the N ew foundland and Labrador Shelves. J P lankton Res 29: 5 0 9 -5 1 4 .

79. W u Y , P la tt T , T a n g C C L , Sathyendranath S (2008) Reg ional differences in the tim ing o f the spring b loom in the Labrador Sea. M a r Ecol Prog Ser 355: 9—20.

80. Z h a i L , P la tt T , T a n g C , Sathyendranath S, W alls R H (2011) Phytoplankton phenology on the Scotian Shelf. IC E S J M a r Sei 6 8 : 7 8 1 -7 9 1 .

81. N apo léon C , C laqu in P (2012) M u lti-p ara m etric relationships betw een P A M measurements and carbon incorporation, an in situ approach. PLoS O N E 7: e40284.

82. Longhurst A R (1998) Ecological geography o f the sea. Academ ic Press, 398 p.83. Bonnet S, G uieu C , Bruyan t F , Prasil O , V a n W am beke F et al. (2008) N u trie n t

lim itation o f p rim ary productiv ity in the Southeast Pacific {B IO S O P E cruise). Biogeosciences 5: 2 1 5 -2 2 5 .

84. G reene R M , G eider RJ, Falkowski P G (1991) E ffect o f iron lim itadon on photosynthesis in a m arine d iatom . L im n o l O ceanogr 36: 1772—1782.

85. G reene R M , G eider RJ, K o lb er Z , Falkowski P G (1992) Iron-induced changes in light harvesdng and photochem ical energy conversion processes in eukaryodc m arine algae. P lan t Physiol 100: 5 6 5 -5 7 5 .

8 6 . G eider RJ, LaR oche J , G reene R M , O la izo la M (1993) Response o f the photosynthetic apparatus o f Phaeodactylum tricornutum (Bacillariophyceae) to n itrate , phosphate, or iron starvadon. J Phycol 29: 755—766.

87. Schrader PS, M illig an AJ, Behrenfeld M J (2011). Surplus photosynthedc antennae complexes underlie diagnostics o f iron lim itadon in a cyanobacterium . PLoS O N E 6 : e l8 7 5 3 .

8 8 . R y an -K eogh TJ , M acey A I , Gackshutt A M , M o o re M G , Bibby T S (2012) T he cyanobacteria! chlo rophyll-binding-protein Is iA Acts to increase the in vivo effective absorption cross section o f P S I under iron lim itation . J Phycol 48: 145— 154.

89. Levitan O , Rosenberg G , Sedik I , Sedokova E , G rig e lJ , et al. (2007) E levated C O 2 enhances nitrogen fixation and growth in the m arine cyanobacterium Trichodesmium. G lobal C hange B iol 13: 531—538.

90. Levitan O , K ra n z SA, Spungin D , Prásil O , Rost B, et al. (2010) Com bined effects o f C O 2 and ligh t on the N 2 -fix in g cyanobacterium Trichodesm ium IM S 1 0 1 : A m echanistic view. P lant Physiol 154: 3 4 6 -3 5 6 .



predicting the electron requirement for carbon …fixation requires knowledge of the electron...

Documents