LETTER Contrasting effects of rising CO2 on primary production and

ecological stoichiometry at different nutrient levels

Jolanda M. H. Verspagen,1 Dedmer

B. Van de Waal,1,2 Jan F. Finke,1†

Petra M. Visser1 and

Jef Huisman1*

Abstract

Although rising CO2 concentrations are thought to promote the growth and alter the car-bon : nutrient stoichiometry of primary producers, several studies have reported conflictingresults. To reconcile these contrasting results, we tested the following hypotheses: rising CO2 levels(1) will increase phytoplankton biomass more at high nutrient loads than at low nutrient loads,but (2) will increase their carbon : nutrient stoichiometry more at low than at high nutrient loads.We formulated a mathematical model to predict dynamic changes in phytoplankton populationdensity, elemental stoichiometry and inorganic carbon chemistry in response to rising CO2. Themodel was tested in chemostat experiments with the freshwater cyanobacterium Microcystis aeru-ginosa. The model predictions and experimental results confirmed the hypotheses. Our findingsprovide a novel theoretical framework to understand and predict effects of rising CO2 concentra-tions on primary producers and their nutritional quality as food for herbivores under differentnutrient conditions.

Keywords

Carbon dioxide, elemental stoichiometry, eutrophication, global change, harmful algal blooms,lakes, Microcystis aeruginosa, nutrient limitation, ocean acidification.

Ecology Letters (2014) 17: 951–960

INTRODUCTION

Primary producers assimilate CO2 by the RubisCO enzyme,which generally features very low affinities for CO2 (Badgeret al. 1998; Giordano et al. 2005). Moreover, in aquatic eco-systems, dissolved CO2 concentrations can be stronglydepleted during phytoplankton blooms (Talling 1976; Maberly1996; Balmer & Downing 2011). Several studies have thereforepredicted that the current rise in atmospheric CO2 concentra-tions will increase photosynthetic carbon fixation and phyto-plankton growth rates (Riebesell et al. 1993, 2007), and hencethe occurrence and magnitude of phytoplankton blooms(Schippers et al. 2004). However, this prediction may notalways hold. For instance, in many waters, phytoplanktongrowth rates and bloom development are constrained bynutrient availability (Dzialowski et al. 2005; Xu et al. 2010;Lewandowska et al. 2014). If nutrients are strongly limiting,rising CO2 concentrations are unlikely to have a major impacton phytoplankton growth.An increase in CO2 availability may also affect the elemen-

tal stoichiometry of primary producers (Bellerby et al. 2008;Finkel et al. 2010; Van de Waal et al. 2010). In particular,several studies have shown that rising CO2 concentrationsincrease the carbon : nutrient ratio of phytoplankton (Riebe-sell et al. 2007; Verschoor et al. 2013), which may reduce theirnutritional quality for zooplankton (Urabe et al. 2003; Schooet al. 2013). However, some experimental studies do not sup-

port this expectation, as they did not find an increase in car-bon : nutrient stoichiometry in response to rising CO2

concentrations (Montechiaro & Giordano 2010; Gutow et al.2014). Elevated CO2 concentrations only seem to change phy-toplankton stoichiometry under specific conditions, forinstance, at low nutrient availability (Gervais & Riebesell2001; Leonardos & Geider 2005; Li et al. 2012).These contrasting predictions and results from previous

studies indicate that the response of primary producers andtheir elemental stoichiometry to rising CO2 may depend onthe nutrient status of ecosystems. However, a coherent theo-retical framework to reconcile the different observations isstill lacking. The nutrient-load hypothesis of Brauer et al.(2012) may offer a useful starting point to develop such aframework. The nutrient-load hypothesis describes resourcelimitation patterns as a function of the input of twopotentially limiting nutrients. Here, we consider one limitingnutrient while the role of the other nutrient is replaced byinorganic carbon (Fig. 1a). This modified nutrient-loadhypothesis predicts that, at low pCO2 levels but high nutrientloads, carbon will be the limiting resource for phytoplanktongrowth. At high pCO2 levels but low nutrient loads, thenutrient becomes limiting. Systems with both high pCO2

levels and high nutrient loads can develop dense phytoplank-ton populations, which increase the turbidity of the watercolumn and thereby generate light-limited conditions(Fig. 1a).

1Department of Aquatic Microbiology, Institute for Biodiversity and

Ecosystem Dynamics, University of Amsterdam, P.O. Box 94248, 1090 GE,

Amsterdam, The Netherlands2Department of Aquatic Ecology, Netherlands Institute of Ecology, P.O. Box

50, 6700 AB, Wageningen, The Netherlands

†Present address: Department of Earth, Ocean and Atmospheric Sciences,

University of British Columbia, 6339 Stores Road, Vancouver, British

Columbia, Canada

*Correspondence: E-mail: j.huisman@uva.nl

© 2014 John Wiley & Sons Ltd/CNRS

Ecology Letters, (2014) 17: 951–960 doi: 10.1111/ele.12298

In systems with low nutrient loads, rising pCO2 levels are pre-dicted to shift phytoplankton growth from carbon-limited tonutrient-limited conditions (indicated by the black arrow inFig. 1a). Such a shift towards nutrient limitation will not allowa large increase in phytoplankton biomass (Fig. 1b), but willlikely enhance the carbon : nutrient ratio of the cells (Fig. 1c).

In systems with high nutrient loads, rising pCO2 levels will shiftphytoplankton growth from carbon-limited to light-limitedconditions (indicated by the white arrow in Fig. 1a). Hence, athigh nutrient loads, rising pCO2 levels may allow a much largerincrease in phytoplankton biomass (Fig. 1b), but are unlikelyto lead to major changes in the carbon : nutrient stoichiometry(Fig. 1c). Based on these graphical arguments, we put forwardtwo pairs of hypotheses. Our first pair of hypotheses is that thedevelopment of phytoplankton biomass will increase substan-tially in response to rising pCO2 levels in eutrophic waters(Hypothesis 1a), but to a much lesser extent in oligotrophicwaters (Hypothesis 1b). Our second pair of hypotheses is thatthe carbon : nutrient stoichiometry of phytoplankton willincrease substantially in response to rising pCO2 levels in oligo-trophic waters (Hypothesis 2a), but to a much lesser extent ineutrophic waters (Hypothesis 2b).In this study, we combine theory and experiments to test

these hypotheses. To establish a firm theoretical basis, we firstbuild a stoichiometrically explicit model that allows for car-bon-, nutrient- and light-limited growth. For this purpose, weextend earlier models of multiple resource limitation (e.g. Ster-ner & Elser 2002; Klausmeier et al. 2004; Passarge et al. 2006;Brauer et al. 2012) by the explicit incorporation of inorganiccarbon chemistry (Stumm & Morgan 1996). We will testthe model predictions in laboratory chemostats with the fresh-water cyanobacterium Microcystis aeruginosa, a widespreadnuisance species that develops dense and often toxic bloomsin many lakes worldwide (Verspagen et al. 2006; J€ohnk et al.2008; Xu et al. 2010; Michalak et al. 2013). To address ourhypotheses, the theory and experiments will investigate theimpact of rising CO2 levels on phytoplankton growth andstoichiometry at both low and high nitrogen loads.

THE MODEL

Population dynamics and elemental stoichiometry

We develop a model that considers a well-mixed watercolumn, illuminated from above, with a homogeneouslydistributed phytoplankton population. Phytoplankton fix CO2

during photosynthesis and many species, including Microcys-tis, can also utilise bicarbonate as a carbon source (Giordanoet al. 2005; Sandrini et al. 2014). The water column isprovided with a limited influx of nitrogen, which is consumedby the phytoplankton population. The growing phytoplanktonpopulation absorbs light, thereby reducing light availability.Hence, the model may generate carbon-limited, nitrogen-lim-ited or light-limited conditions, depending on the environmen-tal conditions. Here, we introduce the key assumptions of themodel, while a complete description is given in Appendix S1of the Supporting information.Our model assumes a variable carbon and nitrogen content

of the phytoplankton, as in the classic Droop model andrelated approaches in ecological stoichiometry (Droop 1973;Sterner & Elser 2002; Klausmeier et al. 2004). Let X denotethe population density of the phytoplankton, QC its carboncontent and QN its nitrogen content. Dynamic changes inphytoplankton population density and its carbon and nitrogencontents can be summarised by three differential equations:

Nut

rient

load

(a)

Pop

ulat

ion

dens

ity

High nutrient loadLow nutrient load

pCO2

Car

bon

: nut

rient

ratio

(b)

(c)

Hypothesis 1a

Hypothesis 1b

Hypothesis 2a

Hypothesis 2b

Lightlimited

Carbonlimited

Nutrient limited

Figure 1 Hypothesised patterns. (a) Resource limitation at different pCO2

levels and nutrient loads. The arrows indicate changes in resource

limitation with rising pCO2 levels, at low nutrient loads (black arrow) and

high nutrient loads (white arrow). (b) Phytoplankton population density,

and (c) carbon : nutrient stoichiometry of phytoplankton cells at different

pCO2 levels. Dashed lines represent low nutrient loads, solid lines

represent high nutrient loads.

© 2014 John Wiley & Sons Ltd/CNRS

952 J. M. H. Verspagen et al. Letter

dX

dt¼ lðQC;QNÞX�DX ð1Þ

dQC

dt¼ uCO2

þ uHCO3� r� lðQC;QNÞQC ð2Þ

dQN

dt¼ uN � lðQC;QNÞQN ð3Þ

where l(QC,QN) is the specific growth rate of the phytoplank-ton as function of its carbon and nitrogen content, D is thedilution rate, uCO2

, uHCO3and uN are the uptake rates of CO2,

bicarbonate and nitrogen, respectively, and r is the respirationrate.The specific growth rate increases with the carbon and

nitrogen content of the phytoplankton, and depends on whichof these two elements is most limiting for growth:

lðQC;QNÞ ¼ lMAX minQC �QC;MIN

QC;MAX �QC;MIN;

QN �QN;MIN

QN;MAX �QN;MIN

� �

ð4Þwhere lMAX, is the maximum specific growth rate, min[. . .] isthe minimum function, QC,MIN and QN,MIN are the minimumcellular carbon and nitrogen content required for a cell tofunction, and QC,MAX and QN,MAX are the maximum carbonand nitrogen content of the cells.We assume that uptake rates of CO2, bicarbonate and nitro-

gen are increasing but saturating functions of the ambientconcentrations of these resources described by Michaelis–Menten kinetics, and are suppressed when phytoplankton cellsbecome satiated (Morel 1987). The energy required for carbonand nitrogen uptake comes from the light reactions of photo-synthesis, which depend on light availability. We assume thatthe respiration rate increases with the cellular carbon content,approaching maximum values when cells become satiated withcarbon (Visser et al. 1997). The mathematical equationsdescribing these relationships are presented in Appendix S1.

Dissolved inorganic carbon

On the timescales used in our model (ranging from minutes todays) the relative availability of CO2 and bicarbonate isessentially in equilibrium with pH, whereas the dissolved CO2

concentration is usually not in equilibrium with the atmosphere.Therefore, let [DIC] denotes the total concentration of dissolvedinorganic carbon. Changes in [DIC] can be described by:

d½DIC�dt

¼ Dð½DIC�IN � ½DIC�Þ þ gco2zMAX

þ ðr� uCO2� uHCO3

ÞXð5Þ

The first term on the right-hand side of eqn 5 describeschanges in the DIC concentration due to the influx ([DIC]IN)and efflux of water containing DIC. The second termdescribes exchange of CO2 gas with the atmosphere, wheregCO2

is the CO2 flux across the air–water interface and divi-sion by the water-column depth zMAX converts the CO2 fluxper unit surface area into a volumetric CO2 change. The thirdterm describes how the DIC concentration increases throughrespiration (r) and decreases through uptake of CO2 ðuCO2

Þand bicarbonate ðuHCO3

Þ by phytoplankton.

We assume that the CO2 gas exchange ðgCO2Þ between air

and water is proportional to the concentration gradientacross the air–water interface, which can be quantified asthe difference between dissolved CO2 in equilibrium with thegas pressure ([CO2*]) and the actual dissolved CO2 concen-tration (Siegenthaler & Sarmiento 1993; Cole & Caraco1998):

gCO2¼ mð½CO�

2� � ½CO2�Þ ð6Þwhere v is the gas transfer velocity. The equilibrium value[CO2*] is calculated from Henry’s law, i.e. [CO2*] = K0 pCO2,where pCO2 is the partial pressure of CO2 in air and K0 is thesolubility constant of CO2 gas in water.Uptake of CO2 and bicarbonate not only depletes the DIC

pool but also increases pH, which in turn affects the relativeavailability of the different forms of inorganic carbon (Stumm& Morgan 1996). CO2, bicarbonate and carbonate concentra-tions were calculated from pH by numerical solution of theunderlying chemical reactions, taking into account dynamicchanges in total DIC and alkalinity due to carbon and nutri-ent assimilation by the growing phytoplankton population(see Appendix S1 for details).

Nutrients and light

Changes in the concentration of dissolved inorganic nitrogen([DIN]) depend on the influx ([DIN]IN) and efflux of nitrogenand the uptake of nitrogen by the phytoplankton population:

d½DIN�dt

¼ Dð½DIN�IN � ½DIN�Þ � uNX ð7Þ

The nitrogen concentration in the influx ([DIN]IN) will becalled the ‘nitrogen load’.Because carbon and nitrogen assimilation depend on light

availability, the model also tracks changes in the underwaterlight environment. Let I(z) denote the light intensity at depthz. According to Lambert–Beer’s law,

IðzÞ ¼ IIN expð�KBGz� kXzÞ ð8Þwhere IIN is the incident light intensity at the water surface,KBG is the background turbidity and k is the specific lightattenuation coefficient of the phytoplankton. This equationtakes into account that a growing phytoplankton populationincreases the turbidity of the water column through self-shading, and thereby reduces the light available for furtherphotosynthesis (Huisman & Weissing 1994; Diehl 2002;Passarge et al. 2006). Nitrogen is a key element in photosyn-thetic pigments. We therefore assume that the specific lightattenuation coefficient increases with the cellular nitrogencontent (Aguirre von Wobeser et al. 2011). The light inten-sity at the bottom of the water column is defined asIOUT = I(zMAX).

MATERIAL AND METHODS

Species and culture conditions

We tested the model predictions with two strains of the fresh-water cyanobacterium M. aeruginosa. Strain Microcystis

© 2014 John Wiley & Sons Ltd/CNRS

Letter CO2 effects on biomass and stoichiometry 953

CYA140 was obtained from the Norwegian Institute forWater Research (NIVA). Strain Microcystis HUB5-2-4 wasobtained from the Humboldt University of Berlin, Germany.The Microcystis strains were grown as single cells, not incolonies.The experiments were carried out in laboratory-built

chemostats specifically designed for phytoplankton studies(Huisman et al. 1999; Passarge et al. 2006). Each chemostatconsisted of a flat culture vessel with an optical path length(‘mixing depth’) of zMAX = 5 cm and an effective workingvolume of 1.7 L. The chemostat vessel was continuouslyilluminated from one side by white fluorescent tubes (PhilipsPL-L 24W/840/4P; Philips Lighting, Eindhoven, The Nether-lands) using a low incident light intensity of IIN =50 � 1 lmol photons m�2 s�1 to enable light-limited condi-tions. The chemostats were supplied with a nutrient-richmineral medium (Van de Waal et al. 2009), in which wemodified the nitrate concentration to induce nitrogen-limitedconditions (see below). The chemostats were maintained at aconstant temperature using a metal cooling finger connectedto a Colora thermocryostat, and were aerated with sterilised(0.2 lm Millex-FG Vent Filter; Millipore, Billerica, MA,USA) N2 gas enriched with different CO2 concentrationsusing Brooks Mass Flow Controllers (Brooks Instrument,Hatfield, PA, USA). The gas mixture was dispersed fromthe bottom of the chemostat vessel in fine bubbles at aconstant gas flow rate of 25 L h�1. Experiments withoutphytoplankton showed that the gas transfer velocity (v) wasproportional to the gas flow rate (a), i.e. v = b a, where theconstant of proportionality b was estimated for eachchemostat separately.

Experiments

First, we ran three chemostat experiments with MicrocystisCYA140 under carbon-limited, nitrogen-limited and light-limited conditions, respectively, to investigate whether themodel could capture the transient dynamics from inocula-tion at low phytoplankton population density to a full-grown steady state. Carbon-limited conditions were inducedby a high nitrogen concentration (12.0 mmol L�1 NaNO3)supplied by the mineral medium, but a low pCO2 of200 ppm in the gas flow. Nitrogen-limited conditions wereinduced by a low nitrogen concentration (0.4 mmol L�1

NaNO3) in the mineral medium and an intermediate pCO2

of 400 ppm. Light-limited conditions were induced by a highnitrogen concentration (12.0 mmol L�1 NaNO3) in the min-eral medium and a high pCO2 of 1200 ppm in the gas flow(Table S1). The chemostats had a dilution rate ofD = 0.011 h�1, were maintained at a temperature of 21 °Cand were sampled every other day.Next, we studied the steady states of 12 chemostats of

Microcystis HUB5-2-4. The chemostats were provided witheither a low (0.4 mmol L�1 NaNO3) or high (12.0 mmol L�1

NaNO3) nitrogen concentration in the mineral medium, and arange of different pCO2 concentrations (0.5, 50, 100, 400, 800,2800 ppm) in the gas flow (Table S1). The chemostats had adilution rate of D = 0.00625 h�1 and were maintained at a

temperature of 24 °C. The steady states were sampled everyother day and monitored for at least 10 days.

Measurements

The incident light intensity (IIN) and the light intensity trans-mitted through the chemostat vessel (IOUT) were measuredwith a LI-COR LI-250 quantum photometer (LI-COR Bio-sciences, Lincoln, NE, USA) at 10 randomly chosen positionson the front and back surface of the chemostat vessel respec-tively.DIC concentrations were determined by sampling 15 mL of

culture suspension, which was immediately filtered over0.45 lm membrane filters (Whatman, Maidstone, UK). DICwas subsequently analysed by phosphoric acid addition on aModel 700 TOC Analyser (OI Corporation, College Station,TX, USA), with a detection limit of 0.15 ppm. Temperatureand pH were measured with a SCHOTT pH meter (SCHOTTAG, Mainz, Germany). Concentrations of dissolved CO2,bicarbonate and carbonate were calculated from DIC and pH(Stumm & Morgan 1996), based on the dissociation constantsof inorganic carbon corrected for temperature and salinity(Table S2). Alkalinity was determined in a 50-mL sample thatwas titrated in 0.1–1 mL steps with 10 mmol L�1 HCl to apH of 3.0. The alkalinity was subsequently calculated usingGran plots (Stumm & Morgan 1996).Residual nitrate and phosphate concentrations in the chemo-

stats were determined in triplicate by sampling 15 mL of culturesuspension, which was immediately filtered over 0.45 lm mem-brane filters (Whatman, Maidstone, UK) and the filtrate wasstored at �20 °C. Nitrate concentrations were analysed using aSkalar SA 400 autoanalyser (Skalar Analytical B.V., Breda,The Netherlands), and phosphate concentrations were analysedspectrophotometrically (Murphy & Riley 1962).Phytoplankton population density, both as cell numbers

and total biovolume, was determined in triplicate using aCasy 1 TTC cell counter with a 60 lm capillary (Sch€arfeSystem GmbH, Reutlingen, Germany). Cell size varied consid-erably during the experiments, ranging from 31 to 66 lm3

cell�1 for Microcystis CYA140 and from 21 to 74 lm3 cell�1

for Microcystis HUB5-2-4. We therefore used the total biovo-lume (i.e. the summed volume of all cells per litre of water) asa measure of phytoplankton population density.Samples for cellular carbon, nitrogen, phosphorus and sul-

phur content were pressurised at 10 bar to collapse the gasvesicles of Microcystis and subsequently centrifuged for15 min at 2 000 g. After discarding the supernatant, the pelletwas resuspended in demineralised water, and centrifuged for5 min at 15 000 g. The supernatant was discarded, pelletswere stored at �20 °C and subsequently freeze dried andweighted to determine dry weight. The carbon, nitrogen andsulphur content of homogenised freeze-dried cell powder wereanalysed using a Vario EL Elemental Analyser (ElementarAnalysensysteme GmbH, Hanau, Germany). To determinethe phosphorus content, cells were oxidised with potassiumpersulfate for 1 h at 100 °C (Wetzel & Likens 2000), andphosphate concentrations were subsequently analysed spectro-photometrically (Murphy & Riley 1962).

© 2014 John Wiley & Sons Ltd/CNRS

954 J. M. H. Verspagen et al. Letter

Parameter estimation

System parameters such as incident light intensity, composi-tion of the mineral medium, dilution rate, and CO2 concentra-tion in the gas flow were experimentally controlled (Table S1).Growth parameters of the phytoplankton were estimated byfitting the time courses predicted by the model to the timecourses of the variables measured during the experiments.These measured variables included population size (X), cellu-lar carbon and nitrogen content (QC and QN), light transmis-sion (IOUT), DIC concentration, alkalinity and pH. In theexperiments with Microcystis HUB5-2-4, the specific lightattenuation coefficient (k) was found to increase linearly withthe cellular nitrogen content QN (Table S3 and Fig. S1). Themodel fits were based on minimisation of the residual sum ofsquares, following the same procedures as in earlier studies(Huisman et al. 1999; Passarge et al. 2006). The parameterestimates are summarised in Table S3.

RESULTS

Dynamics under carbon, nitrogen and light limitation

We first studied growth of Microcystis CYA140 under car-bon-, nitrogen- and light-limited conditions. This showed thatthe changes in population density, light penetration, carbonand nitrogen availability, pH and C : N stoichiometry werespecific for each limitation (Fig. 2). Population densitiesincreased and reached steady state within 15–30 days(Fig. 2a–c). The steady-state population density was lowest inthe nitrogen-limited and highest in the light-limited chemostat(Fig. 2b,c). Dense phytoplankton populations absorbed morelight. Hence, light intensity transmitted through the water col-umn (IOUT) was highest in the nitrogen-limited and lowest inthe light-limited chemostat (Fig. 2b,c).Assimilation of CO2 by the growing phytoplankton popula-

tions decreased the dissolved CO2 concentration and increasedpH (Fig. 2d–f). Under carbon-limited conditions, the dis-solved CO2 concentration was depleted by 98% to < 0.2 lmolL�1 and the pH increased to 10 (Fig. 2d). In contrast, the dis-solved CO2 concentration was reduced by only 30–40% to~10 lmol L�1 while the pH increased by only 0.5 units undernitrogen- and light-limited conditions (Fig. 2e,f).The growing phytoplankton populations consumed nitrate

(Fig. 2g–i). Under carbon- and light-limited conditions, theresidual nitrate concentration remained high (Fig. 2g,i),whereas under nitrogen-limited conditions the nitrate con-centration was depleted to ~5 lmol L�1 at steady state(Fig. 2h).The availability of nitrogen, carbon and light had a major

effect on the C : N stoichiometry of the phytoplankton cells.Their molar C : N ratio decreased to 5.7 � 0.2 undercarbon-limited and to 4.9 � 0.1 under light-limited conditions(Fig. 2g,i), whereas it increased to 10.9 � 0.6 under nitrogen-limited conditions (Fig. 2h).The model not only predicted the steady-state conditions

but also adequately captured the transient dynamics underthese wide-ranging growth conditions (compare symbols andlines in Fig. 2).

Steady-state patterns along a pCO2 gradient

We studied steady-state patterns in 12 chemostats with Micro-cystis HUB5-2-4, supplied with a range of pCO2 levels in thegas flow at either a low or high nitrate concentration in themineral medium (Fig. 3). The steady-state population densityincreased with rising pCO2 levels. At low nitrogen loads, thepopulation density levelled off at pCO2 > 100 ppm, whereasat high nitrogen loads the population density increased furtherand became twice as high at pCO2 > 300 ppm (Fig. 3a). Dueto the increasing phytoplankton population, light penetrationIOUT decreased with rising pCO2 (Fig. 3b). At low nitrogenloads, IOUT remained fairly high at ~17 lmol photons m�2

s�1 for pCO2 > 100 ppm, whereas at high nitrogen loads IOUT

was reduced to only 3 lmol photons m�2 s�1.At low pCO2, phytoplankton photosynthesis depleted the

dissolved CO2 concentration to 0.1 lmol L�1 and raised pHto 10 (Fig. 3c,d). The control that phytoplankton exerted onthe dissolved CO2 concentration and pH decreased at athreshold of pCO2 > 100 ppm. Beyond this threshold, the dis-solved CO2 concentration increased while the pH decreasedwith rising pCO2. Interestingly, the nitrogen load had onlymarginal effects on the response of dissolved CO2 and pH torising pCO2 levels.In the chemostats supplied with high nitrogen loads, the

residual nitrate concentration remained high and the molarC : N ratio of the phytoplankton remained low irrespective ofpCO2 (Fig. 3e,f). In contrast, in chemostats supplied with lownitrogen loads, the residual nitrate concentration and C : Nstoichiometry depended on pCO2. Residual nitrate wasdepleted to 10 lmol L�1 at pCO2 > 100 ppm, while the C : Nratio increased from ~6 at pCO2 < 50 ppm to almost 14 atpCO2 > 100 ppm.The model described the steady-state data well (Fig. 3). It

predicts that phytoplankton growth switched from carbon tonitrogen limitation at a pCO2 of 80–100 ppm in theexperiments supplied with low nitrogen, while it switchedfrom carbon to light limitation at a pCO2 of 100–300 ppm inthe experiments with high nitrogen loads.

Model predictions at different pCO2 levels and nitrogen loads

We used the model to predict steady-state patterns along awider range of pCO2 levels and nitrogen loads. Steady-statephytoplankton population density increased with rising pCO2

levels and nitrogen loads (Fig. 4a). Absorption of light by theincreasing phytoplankton population decreased light penetra-tion (Fig. 4b). At high pCO2 levels and nitrogen loads (i.e.pCO2 > 400 ppm and [DIN]IN > 1.5 mmol L�1), the increasein population density and decrease in light penetration levelledoff, and phytoplankton growth became limited by the lowavailability of light.At low pCO2 levels but high nitrogen loads, phytoplankton

growth depleted dissolved CO2 concentrations and raisedthe pH to high levels (Fig. 4c,d). Under these conditions,phytoplankton growth was limited by the availability of inor-ganic carbon. At high pCO2 levels but low nitrogen loads,phytoplankton growth depleted dissolved nitrate concen-trations (Fig. 4e), while the carbon : nutrient ratio of the

© 2014 John Wiley & Sons Ltd/CNRS

Letter CO2 effects on biomass and stoichiometry 955

phytoplankton was high (C : N > 13; Fig. 4f). Under theseconditions, phytoplankton growth was limited by nitrogenavailability.

DISCUSSION

One of the major achievements of our study is the incorpora-tion of inorganic carbon chemistry into a stoichiometricallyexplicit model. The experimental results show that the modelsuccessfully predicted the population dynamics, nutrient andlight availability, carbon chemistry, pH and elemental stoichi-ometry during phytoplankton growth (Figs. 2–4). Moreover,the model predictions and experimental results are in good

agreement with the hypotheses put forward in Fig. 1. Morespecifically, in the experiments at high nitrogen loads, risingpCO2 levels shifted phytoplankton growth from carbon-lim-ited to light-limited conditions. This resulted in a largeincrease of phytoplankton biomass (Hypothesis 1a; Figs 1band 3a), while the carbon : nitrogen ratio of the phytoplank-ton remained low (Hypothesis 2b; Figs 1c and 3f). Con-versely, at low nitrogen loads, rising pCO2 shiftedphytoplankton growth from carbon-limited to nitrogen-limitedconditions. This increased phytoplankton biomass to only aminor extent (Hypothesis 1b; Figs 1b and 3a), but increasedthe carbon : nitrogen ratio of the phytoplankton substantially(Hypothesis 2a; Figs 1c and 3f).

(a)

Pop

ulat

ion

dens

ity (m

m3

L–1 )

0

250

500

750

1000

1250

1500

population densityIOUT

(d)

time (days)

[CO

2] (μ

M)

0.01

0.1

1

10

100

1000[CO2]pH

(g)

0 10 20 30 40

[DIN

] (m

M)

0

2

4

6

8

10

12

14

C:N

(mol :m

ol)

C:N[DIN]

(b) (c) IOU

T(μm

ol photons m–2 s –1)

0

5

10

15

20

25

(e) (f)

pH

7

8

9

10

11

(i)

0 10 20 30 40 50

Molar C

: N ratio

3

5

7

9

11

13

15

Carbon limitation Nitrogen limitation Light limitation

(h)

Time (days)0 10 20 30 40

Figure 2 Dynamics observed under carbon-limited (left column), nitrogen-limited (middle column) and light-limited conditions (right column). (a–c)Phytoplankton population density (expressed as biovolume) and light intensity penetrating through the chemostat (IOUT). (d–f) Dissolved CO2

concentration and pH. (g–i) Dissolved inorganic nitrogen concentration ([DIN]) and phytoplankton C : N ratio. The experiments were performed with

Microcystis CYA140. Symbols represent measurements, lines represent model fits. Parameter values of the model are provided in Tables S1–S3.

© 2014 John Wiley & Sons Ltd/CNRS

956 J. M. H. Verspagen et al. Letter

The conceptual framework of Fig. 1 not only allows astraightforward interpretation of our experimental results, butmay also help to reconcile contrasting results of several earlierlaboratory and field experiments. Different studies have com-pared different hypotheses. For example, Fu et al. (2010)investigated our Hypotheses 1a and 1b. In line with thesehypotheses, Fu and colleagues found that growth rates of theharmful dinoflagellate Karlodinium veneficum increased withincreasing pCO2 concentrations under phosphate-replete butnot under phosphate-limiting conditions. Li et al. (2012)investigated all four hypotheses using the diatom Phaeodacty-lum tricornutum. They found that the specific growth rate didnot respond to elevated pCO2 under nitrogen-limitedconditions, while it increased slightly but not significantlyunder nitrogen-enriched conditions. These results confirmHypothesis 1b, and are consistent with but not strong enough

to confirm Hypothesis 1a. Furthermore, the C : N ratio of P.tricornutum increased from 6.7 to 8.0 when pCO2 was raisedunder nitrogen-limited conditions, while it remained at 5.3–5.5when pCO2 was raised under nitrogen-enriched conditions.These results confirm Hypotheses 2a and 2b.Montechiaro & Giordano (2010) investigated our

Hypotheses 1a and 2b. They found that, under nutrientreplete conditions, elevated pCO2 concentrations increased thegrowth rates of the dinoflagellate Protoceratium reticulatumbut had only a minor impact on phytoplankton stoichiometry.In contrast, Riebesell et al. (2007) investigated Hypotheses 1band 2a. They spiked marine mesocosm experiments withnutrients at three different pCO2 levels. After an initialincrease in phytoplankton biomass, the nutrient concentra-tions were depleted and phytoplankton biomass returned backto normal levels irrespective of the pCO2 treatment. However,

Pop

ulat

ion

dens

ity (m

m3

L–1 )

0

200

400

600

800

1000Low NHigh N

(b)

I OU

T(μ

mol

pho

tons

m–2

s–1

)

0

10

20

30

40

(d)

pH6

7

8

9

10

11

12(c)

[CO

2] (μ

M)

0.001

0.01

0.1

1

10

100

1000

(e)

pCO2 (ppm)0.1 1 10 100 1000 10000

[DIN

] (m

M)

0.001

0.01

0.1

1

10

100(f)

pCO2 (ppm)0.1 1 10 100 1000 10000

Mol

ar C

: N

ratio

4

6

8

10

12

14

16

18

(a)

Figure 3 Steady-state patterns at different partial pressures of CO2 (pCO2), in experiments with either low nitrogen loads (open circles and dashed lines) or

high nitrogen loads (closed circles and solid lines). (a) Phytoplankton population density (expressed as biovolume), (b) light intensity penetrating through

the chemostat (IOUT), (c) dissolved CO2 concentration, (d) pH, (e) dissolved inorganic nitrogen concentration ([DIN]), and (f) phytoplankton C : N ratio.

The experiments were performed with Microcystis HUB5-2-4. Symbols represent measurements, lines represent model fits. Parameter values of the model

are provided in Tables S1–S3.

© 2014 John Wiley & Sons Ltd/CNRS

Letter CO2 effects on biomass and stoichiometry 957

the stoichiometry of carbon to nitrogen drawdown increasedfrom 6.0 at low pCO2 to 8.0 at high pCO2, thus exceeding theRedfield carbon : nitrogen ratio of 6.6 in today’s ocean (Rieb-esell et al. 2007). Similar results were obtained by Verschooret al. (2013) in freshwater mesocosm experiments with greenalgae and cyanobacteria. Growth at elevated pCO2 levels ledto depletion of the concentrations of dissolved inorganicnitrogen and phosphorus. Elevated pCO2 levels did not havea significant effect on phytoplankton biomass, but shiftedtheir elemental composition towards higher carbon : nutrientratios (Verschoor et al. 2013), thus confirming Hypotheses 1band 2a.

Of course, the hypotheses in Fig. 1 are an abstract presenta-tion of how rising pCO2 levels may affect phytoplankton bio-mass and stoichiometry, emphasising the contrast betweenoligotrophic and eutrophic ecosystems. In reality, there is agradual transition between the different resource limitations.This is also visible in our results (Fig. 4). Such gradual transi-tions are often due to colimitation, i.e. to interactive effects ofthe limiting resources. For instance, our model takes intoaccount that light energy is required for the assimilation ofcarbon and nitrogen, while nitrogen molecules are needed todevelop the photosynthetic machinery harvesting the availablelight energy. These interactions imply that, at intermediate

Population density(mm3 L–1)

Light penetration(μmol photonsm–2 s–1)

[CO2] (μM) p H

< 2020 - 100

300 - 500

500 - 700> 700

> 3025 - 30

15 - 257.5 - 153.5 - 7.5< 3.5

(a)

0.01

0.1

1

10(b)

0.0 1

0.1

1

10

< 0.1

10 - 100> 100

> 9.59 - 9.58.5 - 98 - 8.57.5 - 87 - 7.5

< 7

[DIN] (mM) Molar C:N ratio

< 77 - 88 - 1010 - 1212 - 13> 13

(d)

0.01

0.1

1

10(c)

Nitr

ogen

load

(mM

)

0.01

0.1

1

10

(f)

pCO2 (ppm)1 10 100 1000 10000

0.01

0.1

1

10< 0.010.01 - 0.10.1 - 1

1 - 33 - 10

> 10

(e)

pCO2 (ppm)1 10 100 1000 10000

0.01

0.1

1

10

100 - 300

0.1 - 0.30.3 - 11 - 10

Figure 4 Model prediction of steady-state patterns at different pCO2 levels and nitrogen loads. (a) Phytoplankton population density (expressed as

biovolume), (b) light intensity penetrating though the chemostat (IOUT), (c) dissolved CO2 concentration, (d) pH, (e) dissolved inorganic nitrogen

concentration ([DIN]) and (f) phytoplankton C : N ratio. The solid white line indicates the boundary between carbon- and nitrogen-limited growth

according to the minimum function in eqn 4. The dashed white line roughly indicates the transition from carbon to light limitation (vertical line), and from

nitrogen to light limitation (horizontal line). The model predictions are based on parameter estimates obtained from our experiments with Microcystis

HUB5-2-4 (Tables S1-S3).

© 2014 John Wiley & Sons Ltd/CNRS

958 J. M. H. Verspagen et al. Letter

nutrient loads, both the biomass and carbon : nutrient ratiosof phytoplankton can increase in response to rising pCO2.Indeed, when the green alga Scenedesmus acutus was grown atP concentrations corresponding to meso- to eutrophicconditions, elevated pCO2 levels stimulated algal growth andincreased algal C : P ratios (Urabe et al. 2003). Mesotrophicsystems may therefore be particularly sensitive to rising pCO2

levels, because they can suffer from the dual effects of bothbloom intensification and changes in phytoplanktonstoichiometry.Our results provide ‘proof of principle’, based on a

combination of theory and experiments. However, like allmodels, our model is based on a series of simplifying assump-tions that ignore many of the intriguing complexities of thenatural world. For example, we simplified cell physiology, inparticular the uptake and assimilation of inorganic carbon bycarbon concentrating mechanisms. Cyanobacteria have fivedifferent inorganic carbon uptake systems, a series of carbonicanhydrases, and specialised cellular compartments known ascarboxysomes where CO2 fixation takes place (Badger et al.1998; Giordano et al. 2005; Sandrini et al. 2014). Incorpora-tion of these physiological components, and their regulatorymechanisms, into our model may improve understanding ofthe physiological acclimation of phytoplankton to changingCO2 conditions.Furthermore, our study operated at the laboratory scale,

with model parameters tuned to the small dimensions andhighly idealised conditions of chemostat experiments. Thisimplies that the critical nitrogen loads and pCO2 levels atwhich we measured a transition from, say, carbon-limited tonutrient-limited conditions cannot be directly applied to lakesand reservoirs. For instance, light supply and CO2 influx areexpressed per unit surface area, whereas nutrient depletionoccurs per unit volume. Hence, the critical nutrient loads andpCO2 levels will be different in natural systems that obviouslyhave different dimensions than our chemostat vessels. Fur-thermore, we used relatively low and constant light levels, didnot investigate temperature effects and ignored the potentialimpacts of grazing by herbivores. Implementation of ourmodel in more extensive ecosystem studies will require properdimensionalisation and incorporation of additional processesrelevant for these ecosystems.Despite these caveats, extrapolation of our results to natural

ecosystems indicates that rising pCO2 concentrations willintensify phytoplankton blooms in eutrophic and hypertrophicwaters. This may further enhance the productivity of thesewaters, but is also likely to aggravate the problems associatedwith phytoplankton blooms, such as deterioration of waterquality, anoxia and fish kills (Hallegraeff 1993; Anderson et al.2002; Paerl & Huisman 2009). In particular, more intenseblooms by toxic phytoplankton species, such as the freshwatercyanobacteriumMicrocystis used in this study, can make watersunsuitable for recreation, fisheries, drinking water and agricul-tural purposes (Verspagen et al. 2006; J€ohnk et al. 2008; Xuet al. 2010; Michalak et al. 2013). In oligotrophic waters, risingpCO2 levels are less likely to intensify blooms, but will increasephytoplankton carbon : nutrient ratios. Such changes in ele-mental stoichiometry tend to reduce the nutritional quality ofphytoplankton (Sterner & Elser 2002), which may lead to

reduced growth rates and higher carbon excretion rates bynutrient-demanding zooplankton feeding on these phytoplank-ton populations (Urabe et al. 2003; Schoo et al. 2013).Increases in phytoplankton carbon : nutrient ratios can there-fore alter food-web structure and may reduce the secondaryproduction of oligotrophic waters (Van de Waal et al. 2010).Our results thus imply that rising atmospheric CO2 levels willhave very different consequences for eutrophic versus oligo-trophic ecosystems.

ACKNOWLEDGEMENTS

We thank the referees for their helpful comments, Ellen vanDonk for her encouragement and Leo Hoitinga and Bert deLeeuw for their contribution to the nutrient analyses. Theresearch of JMHV and JH was supported by the Earth andLife Sciences Foundation (ALW), which is subsidised by theNetherlands Organization for Scientific Research (NWO).

AUTHORSHIP

JMHV, DBVdW, PMV and JH designed the study. JMHV,DBVdW and JFF performed the experiments. JMHV and JHdeveloped the model. JMHV and JH wrote the manuscriptand all authors contributed to the final version.

REFERENCES

Aguirre von Wobeser, E., Ibelings, B.W., Bok, J., Krasikov, V., Huisman,

J. & Matthijs, H.C.P. (2011). Concerted changes in gene expression and

cell physiology of the cyanobacterium Synechocystis sp. strain PCC

6803 during transitions between nitrogen and light-limited growth.

Plant Physiol., 155, 1445–1457.Anderson, D.M., Glibert, P.M. & Burkholder, J.M. (2002). Harmful algal

blooms and eutrophication: nutrient sources, composition, and

consequences. Estuaries, 25, 704–726.Badger, M.R., Andrews, T.J., Whitney, S.M., Ludwig, M., Yellowlees,

D.C., Leggat, W. et al. (1998). The diversity and coevolution of

Rubisco, plastids, pyrenoids, and chloroplast-based CO2-concentrating

mechanisms in algae. Can. J. Bot., 76, 1052–1071.Balmer, M.B. & Downing, J.A. (2011). Carbon dioxide concentrations in

eutrophic lakes: undersaturation implies atmospheric uptake. Inland

Waters, 1, 125–132.Bellerby, R.G.J., Schulz, K.G., Riebesell, U., Neill, C., Nondal, G.,

Heegaard, E. et al. (2008). Marine ecosystem community carbon and

nutrient uptake stoichiometry under varying ocean acidification during

the PeECE III experiments. Biogeosciences, 5, 1517–1527.Brauer, V.S., Stomp, M. & Huisman, J. (2012). The nutrient-load

hypothesis: patterns of resource limitation and community structure

driven by competition for nutrients and light. Am. Nat., 179, 721–740.Cole, J.J. & Caraco, N.F. (1998). Atmospheric exchange of carbon

dioxide in a low-wind oligotrophic lake measured by the addition of

SF6. Limnol. Oceanogr., 43, 647–656.Diehl, S. (2002). Phytoplankton, light, and nutrients in a gradient of

mixing depths: theory. Ecology, 83, 386–398.Droop, M.R. (1973). Some thoughts on nutrient limitation in algae. J.

Phycol., 9, 264–272.Dzialowski, A.R., Wang, S.H., Lim, N.C., Spotts, W.W. & Huggins,

D.G. (2005). Nutrient limitation of phytoplankton growth in central

plains reservoirs, USA. J. Plankton Res., 27, 587–595.Finkel, Z.V., Beardall, J., Flynn, K.J., Quigg, A., Rees, T.A.V. & Raven,

J.A. (2010). Phytoplankton in a changing world: cell size and elemental

stoichiometry. J. Plankton Res., 32, 119–137.

© 2014 John Wiley & Sons Ltd/CNRS

Letter CO2 effects on biomass and stoichiometry 959

Fu, F.X., Place, A.R., Garcia, N.S. & Hutchins, D.A. (2010). CO2 and

phosphate availability control the toxicity of the harmful bloom

dinoflagellate Karlodinium veneficum. Aquat. Microb. Ecol., 59, 55–65.Gervais, F. & Riebesell, U. (2001). Effect of phosphorus limitation on

elemental composition and stable carbon isotope fractionation in a

marine diatom growing under different CO2 concentrations. Limnol.

Oceanogr., 46, 497–504.Giordano, M., Beardall, J. & Raven, J.A. (2005). CO2 concentrating

mechanisms in algae: mechanisms, environmental modulation, and

evolution. Annu. Rev. Plant Biol., 56, 99–131.Gutow, L., Rahman, M.M., Bartl, K., Saborowski, R., Bartsch, I. &

Wiencke, C. (2014). Ocean acidification affects growth but not

nutritional quality of the seaweed Fucus vesiculosus (Phaeophyceae,

Fucales). J. Exp. Mar. Biol. Ecol., 453, 84–90.Hallegraeff, G.M. (1993). A review of harmful algal blooms and their

apparent global increase. Phycologia, 32, 79–99.Huisman, J. & Weissing, F.J. (1994). Light-limited growth and

competition for light in well-mixed aquatic environments: an

elementary model. Ecology, 75, 507–520.Huisman, J., Jonker, R.R., Zonneveld, C. & Weissing, F.J. (1999).

Competition for light between phytoplankton species: experimental

tests of mechanistic theory. Ecology, 80, 211–222.J€ohnk, K.D., Huisman, J., Sharples, J., Sommeijer, B., Visser, P.M. &

Stroom, J.M. (2008). Summer heatwaves promote blooms of harmful

cyanobacteria. Glob. Change Biol., 14, 495–512.Klausmeier, C.A., Litchman, E. & Levin, S.A. (2004). Phytoplankton

growth and stoichiometry under multiple nutrient limitation. Limnol.

Oceanogr., 49, 1463–1470.Leonardos, N. & Geider, R.J. (2005). Elevated atmospheric carbon

dioxide increases organic carbon fixation by Emiliania huxleyi

(Haptophyta), under nutrient-limited high-light conditions. J. Phycol.,

41, 1196–1203.Lewandowska, A.M., Boyce, D.G., Hofmann, M., Matthiessen, B.,

Sommer, U. & Worm, B. (2014). Effects of sea surface warming on

marine plankton. Ecol. Lett., 17, 614–623.Li, W., Gao, K. & Beardall, J. (2012). Interactive effects of ocean

acidification and nitrogen-limitation on the diatom Phaeodactylum

tricornutum. PLoS ONE, 7, e51590. doi: 10.1371/journal.pone.0051590.

Maberly, S.C. (1996). Diel, episodic and seasonal changes in pH and

concentrations of inorganic carbon in a productive lake. Freshw. Biol.,

35, 579–598.Michalak, A.M., Anderson, E.J., Beletsky, D., Boland, S., Bosch, N.S.,

Bridgeman, T.B. et al. (2013). Record-setting algal bloom in Lake

Erie caused by agricultural and meteorological trends consistent with

expected future conditions. Proc. Natl Acad. Sci. USA, 110,

6448–6452.Montechiaro, F. & Giordano, M. (2010). Compositional homeostasis of

the dinoflagellate Protoceratium reticulatum grown at three different

pCO2. J. Plant Physiol., 167, 110–113.Morel, F.M.M. (1987). Kinetics of nutrient uptake and growth in

phytoplankton. J. Phycol., 23, 137–150.Murphy, J. & Riley, J.P. (1962). A modified single solution method for

the determination of phosphate in natural waters. Anal. Chim. Acta, 27,

31–36.Paerl, H.W. & Huisman, J. (2009). Climate change: a catalyst for global

expansion of harmful cyanobacterial blooms. Environ. Microbiol. Rep.,

1, 27–37.Passarge, J., Hol, S., Escher, M. & Huisman, J. (2006). Competition for

nutrients and light: stable coexistence, alternative stables states, or

competitive exclusion. Ecol. Monogr., 76, 57–72.Riebesell, U., Wolf-Gladrow, D.A. & Smetacek, V. (1993). Carbon

dioxide limitation of marine phytoplankton growth rates. Nature, 361,

249–251.

Riebesell, U., Schulz, K.G., Bellerby, R.G.J., Botros, M., Fritsche, P.,

Meyerh€ofer, M. et al. (2007). Enhanced biological carbon consumption

in a high CO2 ocean. Nature, 450, 545–549.Sandrini, G., Matthijs, H.C.P., Verspagen, J.M.H., Muyzer, G. &

Huisman, J. (2014). Genetic diversity of inorganic carbon uptake

systems causes variation in CO2 response of the cyanobacterium

Microcystis. ISME J., 8, 589–600.Schippers, P., L€urling, M. & Scheffer, M. (2004). Increase of atmospheric

CO2 promotes phytoplankton productivity. Ecol. Lett., 7, 446–451.Schoo, K.L., Malzahn, A.M., Krause, E. & Boersma, M. (2013).

Increased carbon dioxide availability alters phytoplankton

stoichiometry and affects carbon cycling and growth of a marine

planktonic herbivore. Mar. Biol., 160, 2145–2155.Siegenthaler, U. & Sarmiento, J.L. (1993). Atmospheric carbon dioxide

and the ocean. Nature, 365, 119–125.Sterner, R.W. & Elser, J.J. (2002). Ecological Stoichiometry: The Biology

of Elements from Molecules to the Biosphere. Princeton University

Press, Princeton, NJ, USA.

Stumm, W. & Morgan, J.J. (1996). Aquatic Chemistry: Chemical

Equilibria and Rates in Natural Waters, 3rd edn. Wiley-Interscience,

New York, NY, USA.

Talling, J.F. (1976). The depletion of carbon dioxide from lake water by

phytoplankton. J. Ecol., 64, 79–121.Urabe, J., Togari, J. & Elser, J.J. (2003). Stoichiometric impacts of

increased carbon dioxide on a planktonic herbivore. Glob. Change Biol.,

9, 818–825.Van de Waal, D.B., Verspagen, J.M.H., L€urling, M., Van Donk, E.,

Visser, P.M. & Huisman, J. (2009). The ecological stoichiometry of

toxins produced by harmful cyanobacteria: an experimental test of the

carbon-nutrient balance hypothesis. Ecol. Lett., 12, 1326–1335.Van de Waal, D.B., Verschoor, A.M., Verspagen, J.M.H., van Donk, E.

& Huisman, J. (2010). Climate-driven changes in the ecological

stoichiometry of aquatic ecosystems. Front. Ecol. Environ., 8, 145–152.Verschoor, A.M., Van Dijk, M.A., Huisman, J. & Van Donk, E. (2013).

Elevated CO2 concentrations affect the elemental stoichiometry and

species composition of an experimental phytoplankton community.

Freshw. Biol., 58, 597–611.Verspagen, J.M.H., Passarge, J., J€ohnk, K.D., Visser, P.M., Peperzak, L.,

Boers, P. et al. (2006). Water management strategies against toxic

Microcystis blooms in the Dutch delta. Ecol. Appl., 16, 313–327.Visser, P.M., Passarge, J. & Mur, L.R. (1997). Modelling vertical

migration of the cyanobacterium Microcystis. Hydrobiologia, 349, 99–109.

Wetzel, R.G. & Likens, G.E. (2000). Limnological Analyses, 3rd edn.

Springer, New York, NY, USA.

Xu, H., Paerl, H.W., Qin, B., Zhu, G. & Gao, G. (2010). Nitrogen and

phosphorus inputs control phytoplankton growth in eutrophic Lake

Taihu, China. Limnol. Oceanogr., 33, 420–432.

SUPPORTING INFORMATION

Additional Supporting Information may be downloaded viathe online version of this article at Wiley Online Library(www.ecologyletters.com).

Editor, James GroverManuscript received 13 March 2014First decision made 11 April 2014Manuscript accepted 18 April 2014

© 2014 John Wiley & Sons Ltd/CNRS

960 J. M. H. Verspagen et al. Letter