Phytoplankton Succession Explains Size-Partitioning of New Production Following Upwelling-Induced Blooms

Journal of Marine Systems 148 (2015) 14–25

Contents lists available at ScienceDirect

Journal of Marine Systems

journal homepage: www.elsevier.com/locate/jmarsys

Phytoplankton succession explains size-partitioning of new production following upwelling-induced blooms

N. Van Oostende a,⁎,J.P.Dunneb, S.E. Fawcett a, B.B. Ward a a Department of Geosciences, Princeton University, USA b NOAA Geophysical Fluid Dynamics Laboratory, USA article info abstract

Article history: Large and chain-forming diatoms typically dominate the phytoplankton biomass after initiation of coastal Received 1 October 2014 upwelling. The ability of these diatoms to accelerate and maintain elevated nitrate uptake rates has been Received in revised form 23 January 2015 proposed to explain the dominance of diatoms over all other phytoplankton groups. Moreover, the observed Accepted 30 January 2015 delay in biomass accumulation following nitrate supply after initiation of upwelling events has been Available online 7 February 2015 hypothesised to result from changes in the diatom community structure or from physiological acclimation. To in- Keywords: vestigate these mechanisms, we used both numerical modelling and experimental incubations that reproduced Phytoplankton the characteristic succession from small to large species in phytoplankton community composition and size Cell size structure. Using the Tracers Of Phytoplankton with Allometric Zooplankton (TOPAZ) ecosystem model as a Community succession framework, we find that variations in functional group-specific traits must be taken into account, through adjust- −1 New production ments of group-dependent maximum production rates (PCmax,s ), in order to accurately reproduce the Coastal upwelling observed patterns and timescales of size-partitioned new production in a non-steady state environment. Repre- Nitrate uptake sentation of neither nutrient acclimation, nor diatom diversity in the model was necessary as long as lower than Biogeochemical model theoretical maximum production rates were implemented. We conclude that this physiological feature, PCmax,is critical in representing the early, relatively higher specific nitrate uptake rate of large diatoms, and explains the differential success of small and large phytoplankton communities in response to nitrate supply during upwelling. © 2015 Elsevier B.V. All rights reserved.

1. Introduction improved by incorporating explicit representation of microbial community structure. A disproportionate fraction of global primary production (10–30%) Hydrodynamic turbulence, and hence the light and nutrient regime, and carbon sequestration (40–85%) – relative to their ocean area exerts strong controls on the phytoplankton assemblage (Margalef, (5–19%) – occurs on continental shelves, especially in areas of coastal 1978). Some phytoplankton species, notably larger diatoms, are espe- upwelling (Dunne et al., 2007; Longhurst et al., 1995; Muller-Karger cially adept at exploiting the higher nutrient conditions that character- et al., 2005). Wind-driven upwelling of nutrient rich water into the eu- ise upwelling events compared to other, often smaller phytoplankton photic zone along western continental margins drives the development species, whose growth rates saturate at lower nutrient and light levels of phytoplankton blooms, which are often dominated by large and (Barber and Hiscock, 2006; Finkel, 2001; Key et al., 2010; Litchman chain-forming diatoms (Estrada and Blasco, 1985; Kudela et al., 2008). et al., 2007). Field and modelling studies have shown that as the total Because of their relatively large size and biomineral content, these phytoplankton biomass increases, the biomass of larger phytoplankton bloom-forming diatoms efficiently transfer newly produced biomass increases relatively more than that of the smaller species. Across the to higher trophic levels and to the deep ocean and seafloor, where car- resource concentration gradient, this means that disproportionally bon sequestration occurs (Stock and Dunne, 2010; Thunell et al., 2007). more biomass is added to the larger phytoplankton at higher resource Because of this interconnection between phytoplankton community concentrations (Goericke, 2011; Irigoien et al., 2004; Li, 2002; Poulin composition and biogeochemical cycles, numerical models used to and Franks, 2010). understand elemental cycles in the context of climate change can be Phytoplankton cell size is often used as a key functional characteristic. Size distribution of phytoplankton communities can be modelled without invoking grazing control via bottom-up control of phytoplankton growth through maximum growth rates that increase with cell size ⁎ Corresponding author at: Princeton University, Guyot Hall, Department of Geosciences, Princeton NJ 08544, USA. Tel.: +1 609 258 1052. (Irwin et al., 2006). However, this approach contradicts observed E-mail address: [email protected] (N. Van Oostende). allometric relationships, which find decreasing maximum growth

http://dx.doi.org/10.1016/j.jmarsys.2015.01.009 0924-7963/© 2015 Elsevier B.V. All rights reserved. N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25 15 rates with increasing cell size (Edwards et al., 2012), but see Marañón of seawater upwelled from 70 m depth in Monterey Bay. See Fawcett et al. (2013). To resolve this contradiction, it has been proposed that and Ward (2011) for nutrient concentration data and mass balance further phylogenetic constraints be included in models. For example, validation. S M L maximum potential growth rates can be assigned to broad taxonomic Ambient uptake rates of NH4 and NO3 by the P ,P and P size frac- groups to impose controls on phytoplankton community structure tions were determined daily starting on the second day using isotopic (e.g., diatoms vs. green algae cf. Litchman et al. (2010)). tracer (15N) incubations (3 h) of 1.5 l subsamples from the mesocosms The differentiation between growth traits of functional groups in and size-fractionation of particulate nitrogen (PN) by filtration. PN and current generation global models is still poorly constrained (Finkel 15N content were measured using an elemental analyser coupled to a et al., 2010). Coastal ecosystem models using at least two, often Europa Scientific 20/20 mass spectrometer, as described in Fawcett size-based, functional phytoplankton groups have been used to improve and Ward (2011) and in the supplementary methods section. The up- −1 −1 our understanding of patterns of primary production (Goebel et al., take rates (ρi, μmol l h ) were calculated according to the equation 2010; Klein, 2002; Li et al., 2010; Moloney et al., 1991). The dynamic na- of Dugdale and Goering (1967).Specific rates of uptake of NH4 (VNH4) ρ ture of coastal ecosystems creates transient environmental conditions and NO3 (VNO3)werecalculatedbynormalising i to [PN] and averaging to which individual phytoplankton cells rapidly adjust in terms of over the length of the photoperiod during the experiment (13.9 ± 0.1 h, resource allocation to nutrient uptake and cell growth (Morel, 1987). based on continuous measurements of photosynthetic active radiation Some of the earlier coastal ecosystem models invoked physiological (PAR) performed by the Moss Landing Marine Laboratories Weather mechanisms, such as acclimation to changing light and/or nutrient Station; http://weathernew.mlml.calstate.edu). i conditions, or ecological succession to explain the apparent acceleration The relative specific uptake rates of NO3 (N ) in each PN size fraction of nitrate uptake rate by the whole phytoplankton community follow- (Pi) were calculated by normalising the specific nutrient uptake rate in a Pi ing an upwelling event (Dugdale et al., 1990; Wilkerson and Dugdale, particular size fraction (VNi) to the combined specific uptake in all size 1987; Zimmerman et al., 1987), but see Garside (1991).Inaddition, fractions, according to: many instances of acclimation of nitrate uptake or assimilation and the uncoupling of nutrient uptake and cell growth have been docu- X X RelV Pi ¼ V Pi = n ρPi = n ½PN PiÞ: ð1Þ mented experimentally (Collos, 1986; Collos et al., 2005; Jochem et al., Ni Ni i¼1 Ni i¼1 2000; Smith et al., 1992). However, it has not been possible to quantita- tively link these ecophysiological processes to the observed timing and fi partitioning of new production into different phytoplankton size and The equivalent model-derived relative speci c uptake rates were taxonomic groups following an upwelling event. calculated, using daily-averaged values, as: This study had two main objectives: First, to further evaluate the X X phytoplankton community composition and succession in a previously Pi ¼ Pi =½Pi= n Pi = n ½PiÞ : ð Þ RelV Ni JuptakeNi PN JuptakeNi PN 2 described mesocosm experiment (Fawcett and Ward, 2011) and sec- i¼1 i¼1 ond, to determine which ecophysiological processes in phytoplankton communities regulate the observed patterns in the timing and size- Total and size-fractionated f-ratios were then calculated according partitioning of new production following the initiation of an upwelling to the formulation of Eppley and Peterson (1979): event. The mesocosm experiment simulated growth conditions after coastal upwelling initiation and measured production and nutrient uptake into three size fractions (small (PS)=0.7–5 μm; medium f Pi ¼ V Pi = V Pi þ V Pi : ð3Þ NO3 NH4 NO3 (PM)=5–20 μm; large (PL)=N20 μm) during the subsequent development of a phytoplankton bloom. We hypothesised that, to explain the observed patterns in new production and shifts in community composition, i) the phytoplankton size groups had different intrinsic 2.2. Plankton community composition maximum production rates, and that the timing of new production in such a transient environment could be explained alternatively by ii) Plankton cells were identified and counted in preserved samples (1% the physiologically-limited ability of phytoplankton groups to acclimate paraformaldehyde) using light microscopy (Garrison et al., 2005). to the improved nutrient conditions, or iii) by allowing for greater reso- Smaller cells (b5 μm) were classified as autotrophic or heterotrophic lution in taxonomic and subsequently physiological diversity. To test flagellates based on the presence or absence of a chloroplast; those hypotheses we calibrated the ecosystem model “Tracers Of Phyto- picocyanobacterial cells were not counted. At least 270 smaller cells plankton and Allometric Zooplankton” (TOPAZ) (Dunne et al., 2005, and 730 larger cells were counted each day by microscopy, yielding a 2013) using a coastal upwelling species assemblage, and used it as a minimum counting accuracy of 50% at the 95% confidence level for the framework for evaluating the results of the mesocosm experiment. species that became most abundant throughout the experiment. Larger cells were identified to the species level, except for ciliates, which were 2. Methods classified as Strombidinium sp., Oligotrichous, Tintinnid or aloricate (Hasle and Syvertsen, 1997; Steidinger and Tangen, 1997; Throndsen, 2.1. Field experiment overview 1997). Biovolumes were estimated using cellular geometrical shapes and average cellular dimensions (Olenina et al., 2006), and biovolume Phytoplankton blooms were initiated by simulating coastal post- was then converted to carbon using the relationships of Menden- upwelling conditions in three 200 l mesocosms (hereafter, B1, B2, B3) Deuer and Lessard (2000). as reported by Fawcett and Ward (2011). Briefly, water was collected Growth rates of the plankton species or groups that were most abun- during the upwelling season in central Monterey Bay, California dant at the start or end of the experiment were estimated using an (36.85°N, 121.97°W; bottom depth = 250 m) from 70 m depth, exponential growth model where net specific growth rate (μ, day−1) where the nitrate concentration ([NO3]) was sufficiently high equals the slope of the linear regression of the natural logarithm of (~20 μmol l−1) to initiate a bloom. The mesocosms were inoculated cell abundance with time (Wood et al., 2005). Differences between with 2 l of surface water and incubated for 8 days at the light and tem- mesocosms in the net growth rates and initial summed abundance of perature conditions of surface water. The water in the mesocosms was the main small or large phytoplankton species or groups were deter- mixed three times each day to minimise particle settling. The initial am- mined by analysis of covariance (ANCOVA) and post hoc Tukey's honest monium (NH4)andNO3 concentrations in the mesocosms were typical significant difference (HSD) test. 16 N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25

2.3. Overview of the model's structure and assumptions Box 1 Nitrogen sources and sinks. The TOPAZ biogeochemical model (Dunne et al., 2005, 2013) was used as a framework to reproduce and evaluate the nitrogen dynamics The main nitrogen sources and sinks as depicted in the diagram in in the experimental simulation of a coastal upwelling event as described Fig. 1 are represented by the equations below. See Dunne et al. above. Nitrogen is the biomass currency in the model, which uses 30 (2013, Supplementary material section) for more details. state variables to resolve the cycling of nitrogen, carbon, phosphorus, silicate, iron, calcium carbonate and oxygen. The differential equations ∂NPS ¼ JuptakePS þ JuptakePS −JgrazPS ð4Þ encompassing the sources and sinks for each state variable are outlined ∂t NO3 NH4 N in Dunne et al. (2013, Supplementary material section). The planktonic food web comprises 10 nitrogen state variables, whose interactions are PM ∂N PM PM PM illustrated in Fig. 1 and described in Box 1.Inbrief,light(Irrmem)regu- ¼ Juptake þ Juptake −Jgraz ð5Þ ∂ NO3 NH4 N lates phytoplankton production (Geider et al., 1997), the rate of which t is determined by the most limiting nutrient (nitrogen, phosphorus, or iron (Fe)) sensu Liebig (NutLim) and the metabolic cost of biosynthesis PLi ∂N PLi PLi ζ ¼ Juptake þ Juptake ð6Þ ( )(Geider, 1992)(Box 2). Nitrogen and phosphorus limitation of ∂t NO3 NH4 phytoplankton (Lim) are hyperbolic functions of their ambient concentration and the nutrient uptake half-saturation constants of a particular Zμ phytoplankton size group (Edwards et al., 2012). ∂N PS PM Det DONi Zμ ¼ Jgraz þ Jgraz −Jprod −Jprod −Jloss ð7Þ ∂ N N N N N Uptake of NO3 and NH4 by phytoplankton is calculated as a fraction t of total nitrogen uptake (Box 2, E12 and E13). Ammonium and NO3 limitation are calculated following Frost and Franzen (1992), where NO 3 ∂NDet uptake is inhibited when NH concentrations are high relative to the ¼ JprodDet−JreminDet ð8Þ 4 ∂t N N half-saturation constant for NH4. By contrast, iron limitation (defFe)de- pends on the internal cellular iron quota (Sunda and Huntsman, 1995). DONi Phosphate uptake and limitation (defPO ) are patterned after the optimal ∂N 4 ¼ JprodDONi−JreminDONi ð9Þ allocation theory of Klausmeier et al. (2004) in which N:P is determined ∂t N N by the allocation of resources to photosynthetic machinery, nutrient uptake, assembly, and other biomass constituents. The P:N ratio that ∂ NH4 ¼ Zμ þ Zμ þ DONi becomes limiting for phytoplankton growth is taken from the fraction Jexcret Jloss JreminN ∂ NH4 N ð10Þ of the cell dedicated to assembly after Klausmeier et al. (2004).See t −JuptakePS−JuptakePM−JuptakePL Dunne et al. (2013, Supplementary material section) for further details. N N N The growth rate of phytoplankton cells is a function of the maximum carbon-specific rate of photosynthesis (PCm) at a given temperature and ∂NO 3 ¼ −JuptakePS−JuptakePM−JuptakePL ð11Þ level of nutrient and light limitation, and the cost of biosynthesis (Box 2, ∂t N N N E16, E19 and E20). PCm, in turn, is set by the maximum carbon-specific rate of photosynthesis at a reference temperature (0 °C) (PCmax) and kT is a function of temperature (i.e., PCm =PCmax ⋅ e ,withk= −1 0.0631 °C and T in °C) according to Eppley (1972).ThePCmax values in the initial model were set to yield a theoretical maximum growth Niskin bottles, which most likely excluded large grazers (feeding on rate (0.94 ⋅ 10−5s−1) following Bissinger et al. (2008). large or chain-forming diatoms) but included protozoan grazers Several assumptions were made to compare the modelled data with (feeding on smaller phytoplankton); 2) the mesocosm water was the measurements from the mesocosm study: 1) mesozooplankton well-mixed (because it was stirred); 3) iron was non-limiting since grazers were ignored in the model because the water was collected by the mesocosm water was collected near-shore and the experiment was not performed in an iron-clean environment; 4) fluxes of nitrogen fixation and denitrification were not significant during the incubations, because inorganic nitrogen concentrations were almost always high, a nitrogen mass balance between nutrient loss and biomass gain was observed, the water was well-oxygenated, and no sediment layer devel- oped. Because of these assumptions, and for better comparison with the experimental results of Fawcett and Ward (2011), the diazotrophs in the original TOPAZ formulation were replaced by an additional large (N20 μm) phytoplankton size group (PL). This size group encompassed mainly large diatoms such as Thalassiosira spp. and smaller chain- forming diatoms such as Chaetoceros spp.. Thus, in analogy to the mesocosm experiment 3 phytoplankton size groups were considered in the model. The two smaller phytoplankton size groups (PS and PM) were subject to loss through grazing by microzooplankton (Zμ)by means of a Type II Holling functional response (Holling, 1959). The clearance rate (λ,[molNm−3 s]−1) was estimated based upon a gross growth efficiency (gge) of 0.3 (Hansen et al., 1997; Straile, 1997), microscopy measurements of microzooplankton growth rate (μZμ) (see above), and small and medium phytoplankton biomass (NPS Fig. 1. Summary of nitrogen (N) state variables and their interactions in the modified PM Zμ PS PM μ and N ): λ = μ /gge/(N +N ). Microzooplankton loss was TOPAZ model (Table 1, configuration d). P: phytoplankton; Z : microzooplankton; Det: governed by quadratic loss representing cannibalism (Steele and detritus; S: small; M: medium; L: large, r: fast-growing, K: slow-growing; NO3: nitrate;

NH4: ammonium; DON: dissolved organic N, l: labile, sl: semi-labile. Henderson, 1992). N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25 17

Box 2 2.4. Model initialization and forcing Nitrogen uptake, nutrient and light limitation, growth, grazing, production of DON and detritus, and remineralisation. The model was initialised using the initial concentrations measured

in each mesocosm for NO3 and NH4, PN in each size group, and dissolved Processes controlling the main nitrogen sources and sinks as organic nitrogen (DON). The initial concentration of other nutrients depicted in the diagram in Fig. 1 are represented by the equations such as phosphate and silicate were based on local climatologies below. See Dunne et al. (2013, Supplementary material section) (Garcia et al., 2010). The initial dissolved [Fe] was set in excess as −1 for an overview of the parameter values used in the TOPAZ model. 16.45 nmol l , which corresponds to 5 times the upwelling value measured by Johnson et al. (2001), to account for the non iron-clean LimPi environment of the experimental mesocosms. Because of the very low JuptakePi ¼ μPi NPi NO3 ð12Þ NO3 Pi þ Pi surface [NO3] at the sampling site prior to the upwelling experiment LimNO LimNH 3 4 (K. Johnson, personal communication), we set the initial values of the

nutrient limitation memory (NutLimmem) to be low, corresponding to − Pi a Fe:N molar ratio of 10 2, which is equivalent to ambient concentra- LimNH Pi ¼ μPi Pi 4 ð Þ −1 −1 JuptakeNH N 13 tions of dissolved Fe of 1 nmol l and NO of 100 nmol l .The 4 LimPi þ LimPi 3 NO3 NH4 model was forced using the water temperature measured in the mesocosms and the time-series of local photosynthetic active radiation Pi Pi Pi Pi Pi (PAR) measurements from Moss Landing Marine Laboratories Weather NutLim ¼ min Lim þ Lim ; min def ; def ð14Þ NO3 NH4 PO4 Fe Station, scaled to measured PAR intensity inside the mesocosms. Allocation of initial PN into non-living detritus, heterotrophic and Pi ðÞ¼þ Pi ðÞ autotrophic biomass was constrained by microscopy estimates of plank- NutLimmem t 1 NutLimmem t ð15Þ ton biomass. We estimated the initial detritus content of the PN by þ PiðÞ− ðÞ γ assuming that the exponential rate of biomass increase derived from NutLim t NutLimmem t Nutmem the microscopy data (Nmicroscopy) should equal that of PN minus an initial amount of non-proliferating detrital PN (PN ), according to: detritus Pi ¼ Pi Pi; Pi kT ð Þ PCm PCmax min NutLim NutLimmem e 16 dN microscopy ¼ N et ð28Þ dt microscopy Pi ¼ Pi Pi kT ð Þ PCmθ PCmax NutLimmem e 17

! dPNbiomass ¼ t ð Þ 0:5 PNbiomass e 29 θPi ¼ θPi −θPi = 1 þ θPi −θPi αPi Irr : dt max min max min mem Pi ð Þ PCmθ 18 þ Pi; Pi θPi þ θPi min NutLim NutLimmem NutLim min ðÞ¼ t þ : ð Þ PN t PNbiomass e PNdetritus 30

−αPiIrrθPi Pi PPi IrrLim ¼ 1−e Cm ð19Þ Detritus initially constituted 38%, 68%, and 39% of the largest PN size

fraction, and 54%, 71%, and 64% of the summed PNb5 +PN5–20 size PPi fractions in B1, B2, and B3, respectively. μPi ¼ Cm ð20Þ ðÞ1 þ ζ IrrLimPi In accordance with the community succession patterns that emerged from the microscopy data (Fig. 2), we further divided the larg- L ! est phytoplankton subgroup (P ) into an early-dominant, low growth Pi LK Lr 1 N μ rate group (P ), and a later-dominant, high growth rate group (P ). JgrazPi ¼ min ; λ ekT NPi NZ ð21Þ N Δt Pi þ Pi Based on the microscopy data, members of PL were operationally N Nmin deﬁned as large autotrophic cells (equivalent spherical diameter (ESD) N20 μm or known to bear spines and/or to be colony-forming) Zμ ¼ γ kT Zμ Zμ ð Þ JlossN NZμ e N N 22 contributing at least 5% to total plankton biomass on at least one sampling time point during the experiment. An exponential growth rate cut-off of 1.0 day−1 was then used to allocate the PL species to Det ¼ PS PS ðÞ−ϕ −ϕ kreminT ð Þ Lr LK JprodN f Det JgrazN 1 DONl DONsl e 23 the P and P subpopulations. To constrain the differential growth L characteristics of these two P subpopulations in the model (3 PCmax & 2PL groups), we used the ratio of the exponential growth rates of the DONsl ¼ ϕ PS ð Þ JprodN DONsl JgrazN 24 biomass of PLr species to the biomass of PLK species derived from the

microscopy counts. It was thus only necessary to optimise for PCmax in L DONl ¼ ϕ PS Zμ ð Þ one of the two P subpopulations. JprodN DONl JgrazN LimP 25 2.5. Model configurations for hypothesis testing Det ¼ γ Det ð Þ JreminN NDet N 26 To test our hypotheses about the ecophysiological mechanisms governing the timing and partitioning of new production into different DONi ¼ γ DONi ð Þ fi JreminN NDONi N 27 phytoplankton size groups, we adapted alternative con gurations of the model. In addition to adding the largest phytoplankton size group to the baseline TOPAZ model, the following parameters and tracers were modified in the different model configurations (Table 1): the number 18 N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25

6 10 A Chaetoceros debilis Chaetoceros curvisetus ) 1 5 Thalassiosira anguste lineata 10 Eucampia zodiacus Pseudonitzschia sp.

4 10

3 10 cell abundance (cells l

2 10

6 10 B small pennate diatom auto. Gymnodinoid )

1 auto. flagellate hetero. flagellate 5 10

4 10 cell abundance (cells l

3 10 0 1 2 3 4 5 6 7 8 days

Fig. 2. Abundance (cells l−1) of large (A) and small (B) plankton species or groups that were most numerous at the start or end of the ﬁeld experiment in mesocosm B3 (see Fig. S1 for the results of B1 and B2). auto.: autotrophic, hetero.: heterotrophic. Note the difference in the y-axis scale between large and small groups at the start of the incubation.

of different optimised PCmax values (a. 1 PCmax or b. 3 PCmax), the rate of [DON], and the f-ratios taking measurement error into account. The f- nutrient acclimation (γNutmem)(c.3PCmax &nutrientacclimation)and ratio values for the first 3 days only were used in the optimisation L L the number of P groups (from one to two; d. 3 PCmax &2P groups). since uptake rates of ammonium were prone to overestimation once The choice of parameters (PCmax, γNutmem) to be tested was based on the ambient NH4 pool was depleted (Fawcett and Ward, 2011). For their influence on NO3 consumption from results of sensitivity analyses the optimisation of γNutmem (see below), the minimum rate was set to (see Supplementary methods). Nutrient acclimation (Box 2, E15) was 0.5 day−1 in accordance with the findings of Sakshaug and Holm- introduced into the model by including a simple temporal dependence Hansen (1977) regarding the time lag of PN accumulation by of the maximum photosynthetic carbon fixation rate (PCm)onthe Skeletonema costatum in response to a phosphate pulse. The maximum −1 multi-nutrient limitation status of the cell (NutLimmem). acclimation rate was set to 24.0 day , which is essentially instanta- The ability of the different model configurations to reproduce the neous and based on the time necessary for bacterial DNA synthesis mesocosm data was used as a measure of their mechanistic explanatory and chromosome replication induced by a nutritional shift-up potential. We employed an optimisation routine to calibrate the phyto- (Ingraham et al., 1983). plankton group parameter values of PCmax from the original TOPAZ The degree of mismatch between the model and the measurements model and γNutmem in the configuration that included nutrient acclima- of the above-mentioned variables was represented by the root mean tion. In effect, the model was fitted to measured values of [NH4], [NO3], squared error normalised to 1 standard deviation of the measurements [PN] and chlorophyll a concentration [Chla] for each size fraction, (NRMSE; table S1), where a NRMSE value of 1 or less represents a

Table 1

Normalised cumulative error (ΣNRMSE) and optimised parameter values of PCmax and γNutmem for each phytoplankton size group (±1 SD) in the different model conﬁgurations from the 95% conﬁdence interval of bootstrapping results (N = 500) of mesocosm B3. ΣNRMSE values represent the relative deviation of modelled values from measurements of nutrients, Chla and

PN concentrations, and f-ratios within 1 SD (see Table S1 for ΣNRMSE and optimised parameter values of PCmax and γNutmem from single optimizations for each mesocosm).

5 −1 Model conﬁguration ΣNRMSE 10 PCmax at 0 °C (s ) PS (b5 μm) PM (5–20 μm) PL(r) (N20 μm) PLK (N20 μm)

a. 1 PCmax 3.49 ± 0.47 0.45 ± 0.04 0.45 ± 0.04 0.45 ± 0.04 –

b. 3 PCmax 2.75 ± 0.23 0.39 ± 0.07 0.46 ± 0.11 0.54 ± 0.08 –

c. 3 PCmax & nutrient acclimation 2.66 ± 0.23 0.49 ± 0.06 0.53 ± 0.10 0.67 ± 0.06 – L d. 3 PCmax &2P groups 2.77 ± 0.29 0.39 ± 0.07 0.46 ± 0.11 0.83 ± 0.08 0.35 ± 0.03

−1 γNutmem (day ) PS (b5 μm) PM (5–20 μm) PL (N20 μm)

c. 3 PCmax & nutrient acclimation 2.66 ± 0.23 1.5 ± 0.6 1.8 ± 0.8 1.0 ± 0.6 – N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25 19

2 10 25 AB

1 20 10 ) ) 1 1 15 0 10 mol l µ mol N l ( 3 10 NO PN ( µ 1 10 total >20 µm 5 5 20 µm µ 2 0.7 5 m 10 0

1 CD >20 µm 5 20 µm 1 0.8 0.7 5 µm 0.8 ) 1 0.6 0.6 mol l µ ( 0.4 4 0.4 NH fraction of total PN

0.2 0.2

0 0

14 1 total F µ E 12 >20 m 5 20 µm 0.8 10 0.7 5 µm ) 1 0.6 8 g l µ ( ratio

6 f 0.4 Chl a 4 0.2 2

0 0 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 days days

−5 −1 Fig. 3. Comparison of modelled nitrogen dynamics and mesocosm measurements using the same theoretical PCmax value (0.94 ⋅ 10 s at 0 °C) for each phytoplankton size group (original TOPAZ conﬁguration; ΣNRMSE = 6.36). Measurements from mesocosm B3 are denoted by dots with error bars representing 1 SD, while lines represent modelled values.

(A) [PN] in each size fraction, (B) [NO3], (C) proportion of total PN in each size fraction, (D) [NH4], (E) [Chla] in each size fraction, (F) f-ratio based on NO3 and NH4 uptake by the total PN pool. perfect match within measurement error. A nonlinear optimisation rou- larger cells such as the diatoms species Eucampia zodiacus and tine was used to search the parameter space, using a bounded version of Pseudonitzschia sp., and was very similar among mesocosms (Fig. 2 MATLAB's Nelder–Mead simplex direct search algorithm fminsearch and Fig. S1). However, there were inter-mesocosm differences in the (Lagarias et al., 1998), by minimising the sum of NRMSE for each vari- initial relative abundance of certain species, including: higher abun- able. Because small errors in the initial model values propagate in this dance of autotrophic Gymnodinoid dinoflagellates in B1 and B2, hetero- mesocosm experimental design, and to allow for robust comparison trophic flagellates in B2 and B3, Pseudonitzschia sp. in B1 and B3, the between the results from different model configurations we repeated dinoflagellate Cochlodinium sp. in B1 and B3, and the diatom genus the optimisation routines 500 times using a bootstrapping procedure Dactyliosolen in B1 (data not shown). Microzooplankton comprised where the measured values of nutrients, particulate and dissolved nitro- mainly heterotrophic flagellates and ciliates (e.g., Strombidinium sp.), gen, Chla concentrations and estimates of initial detritus content were both of which were more abundant in B2 than in B1 and B3 at the end randomly perturbed around their mean using a normal distribution of of the incubation (data not shown). The starting [PN] was initially their uncertainties. very low in all mesocosms (1.00 ± 0.07 μmol N l−1; Figs. 3 to 5, Figs. S5 and S6), and increased considerably more in B3 (18.4 fold) 3. Results than in B1 and B2 (12.8 and 13.4 fold) during the experiment. Most of

the accumulated PN derived from NO3 assimilation, since the initial 3.1. Species and size structure succession in the phytoplankton community [NH4] was low and rapidly consumed, [NO2] was negligible, [DON] − during experimental upwelling-induced bloom development remained roughly constant at ~3 μmol l 1, and a nitrogen mass balance was maintained in all mesocosms (Fawcett and Ward, 2011). The phytoplankton community was initially composed mainly of Community succession was most obvious among diatoms where small autotrophic and heterotrophic ﬂagellates (b5 μm), Gymnodinoid Chaetoceros spp. and Thalassiosira spp. (N20 μm, in B3) came to domi- dinoﬂagellates (b20 μm), small pennate diatoms (b20 μm), along with nate the assemblage (Fig. 2). An increase in the abundance of large 20 N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25

2 10 25 A B

1 20 10 ) ) 1 1 15 0 10 mol l µ mol N l ( 3 10 NO PN ( µ 1 10 total >20 µm 5 5 20 µm µ 2 0.7 5 m 10 0 1 C >20 µm D 5 20 µm 1 0.8 0.7 5 µm 0.8 ) 1 0.6 0.6 mol l µ ( 0.4 4 0.4 NH fraction of total PN

0.2 0.2

0 0 14 1 total µ EF 12 >20 m 5 20 µm 0.8 10 0.7 5 µm ) 1 0.6 8 g l µ ( ratio

6 f 0.4 Chl a 4 0.2 2

0 0 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 days days

Fig. 4. Comparison of modelled nitrogen dynamics and mesocosm measurements using a single optimised PCmax value for all of the 3 phytoplankton size groups (Tables 1 and S1, conﬁg- uration a). Measurements from mesocosm B3 are denoted by symbols with error bars representing 1 standard deviation, while lines represent modelled values. (A) [PN] in each size fraction, (B) [NO3], (C) proportion of total PN in each size fraction, (D) [NH4], (E) [Chla] in each size fraction, (F) f-ratio based on NO3 and NH4 uptake by the total PN pool.

cells (N20 μm) was apparent after the first day of incubation for 3.2. Mechanisms governing the timing and partitioning of new production e.g., Chaetoceros curvisetus, while an early increase in cell abundance into different phytoplankton size groups of small flagellates was not observed. It is noteworthy that no changes in NH4 and NO3 concentrations were detected during at least the first The original model (Dunne et al., 2013), which used the theoretical −5 −1 2 days in all mesocosms, probably because of the low initial cell abun- maximum PCmax value of 0.94 ⋅ 10 s sensu Bissinger et al. (2008), dance. Although [PN] increased in each size fraction over the course of strongly mischaracterised the observed rates of phytoplankton biomass the experiment, the highest increase in [PN] was invariably observed accumulation, nutrient consumption and community structure ob- in the largest size fraction (31.0, 37.4, 48.9 fold in B1, B2, and B3). As a served in the mesocosms (Fig. 3). Optimising the value of a single consequence, a gradual shift in proportional contribution to PN from PCmax for all phytoplankton size groups resulted in a downward revision the smallest size fraction to the largest size fraction occurred in all of the theoretical maximum PCmax value by 50%, yielding an improved mesocosms, and was most pronounced in B3 (Fig. 5). The difference in fit between the modelled and measured values of biomass accumula- timing of bloom development between mesocosms was likely due to a tion and nutrient consumption (Table 1,configuration a). Although higher initial abundance of the main large phytoplankton species in the overall rate of NO3 and NH4 consumption and total PN accumulation B3 (ANCOVA and Tukey's HSD test: F =6.51,p = 0.011 and p b 0.05). could be reasonably well represented when all phytoplankton groups

In subsequent analyses we will mainly focus on B3 because of its more had the same PCmax value, community succession could not (Fig. 4 and complete succession sequence and the lack of signiﬁcant differences be- Table 1, Tables S1 and S2). The shift in community size structure and tween the growth rates of the dominant small and large phytoplankton the associated transfer of NO3 into each phytoplankton size group were species in the different mesocosms (ANCOVA: large: F = 0.33, p = signiﬁcantly better explained by the model output when optimised

0.725; small: F =0.1,p =0.908). phytoplankton group-speciﬁc values of PCmax were implemented N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25 21

2 10 25 AB

1 20 10 ) ) 1 1 15 0 10 mol l µ mol N l ( 3 10 NO PN ( µ 1 10 total >20 µm 5 5 20 µm µ 2 0.7 5 m 10 0

1 CD >20 µm 5 20 µm 1 0.8 0.7 5 µm 0.8 ) 1 0.6 0.6 mol l µ ( 0.4 4 0.4 NH fraction of total PN

0.2 0.2

0 0

14 1 total µ EF 12 >20 m 5 20 µm 0.8 10 0.7 5 µm ) 1 0.6 8 g l µ ( ratio

6 f 0.4 Chl a 4 0.2 2

0 0 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 days days

Fig. 5. Comparison of modelled nitrogen dynamics and mesocosm measurements using optimised PCmax values for each of the 3 phytoplankton size groups (Tables 1 and S1, conﬁguration b). Measurements from mesocosm B3 are denoted by symbols with error bars representing 1 standard deviation, while lines represent modelled values. (A) [PN] in each size fraction,

(B) [NO3], (C) proportion of total PN in each size fraction, (D) [NH4], (E) [Chla] in each size fraction, (F) f-ratio based on uptake NO3 and NH4 uptake by the total PN pool.

(Fig. 5 and Figs. S3 and S4; ΣNRMSE in Table 1, Table S1 and Table S2, To explain the observed delay in PN accumulation (Fig. 4)andlackof configuration b). Different phytoplankton groups had significantly growth in the initially abundant population of small cells (cell abun- L different PCmax values, with the P group having the highest PCmax value dance of autotrophic flagellates increased only after day 3; Fig. 2) during compared to the other groups and to the model configuration using a the first two to three days of the mesocosm experiment, we allowed the single uniform PCmax value (i.e., configuration a; comparison of the possibility for the phytoplankton groups to acclimate to improved nutri- bootstrapping results of B3 by ANOVA and Tukey's HSD test: F = ent conditions following upwelling simulation (configuration c). In 308.53, p b 0.001 and p b 0.001). effect, this introduced a damped growth response with a timescale de-

pendent on the optimised parameter γNutmem. The inclusion of nutrient acclimation improved the overall fit of the model with the observations (ΣNRMSE) compared to the model configuration that had phytoplank- Table 2 ton group-specificP values only (Table 1 and S1, and Fig. S2; Measured and modelled maximum VNO3 for each particulate matter size fraction or phy- Cmax toplankton size group from B3, averaged over the length of the photoperiod. Modelled ANOVA and Tukey's HSD test: F = 430.89, p b 0.001 and p = 0.015). fi fi values are derived from the model con guration using optimised, group-speci cPCmax Also, the [PN] in the smaller plankton size groups was better matched fi values (Table S1, con guration b). to the observations when nutrient acclimation was included PN size fraction Size cut-off Measured Modelled Time at (Table S2). These improvements were minor, however, because of the or phytoplankton or ESD (μm) max. V max. V max. V NO3 NO3 NO3 related effect of PCmax and γNutmem on the rate of nitrate consumption, −1 −1 group (h ) (h ) (d) and were not always apparent in mesocosms B1 and B2 (Table S1 and small, PS b5 0.11 ± 0.01 0.08 5 Table S2). M medium, P 5–20 0.11 ± 0.04 0.09 5 To address the discrepancy between the immediate growth of cer- large, PL N20 0.11 ± 0.10 0.09 4 tain large phytoplankton species as observed by microscopy (Fig. 2) 22 N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25

3 >20 µm the daily photoperiod). Focusing on B3, and using the most parsimoni- A ﬁ ﬁ 5 20 µm ous model con guration that yielded an improved t(3PCmax), maxi- 2.5

3 µ 0.7 5 m mum VNO3 values for the modelled bloom were of similar magnitude for all size groups, but generally lower than the VNO values derived 2 3 from the isotopic tracer incubations (Table 2). It is noteworthy that the maximum V of PL was reached one day earlier than in the smaller 1.5 NO3 groups, PS and PM. ﬁ 1 Phytoplankton groups with a higher than average relative speci c nitrate uptake rate (RelVPi N 1) will have a competitive advantage.

measured relative VNO NO3 0.5 The measured and modelled RelVNO3 for each PN size fraction and phytoplankton group, respectively, reﬂects the nitrate uptake velocity of a 0 particular size group normalised to the nitrate uptake velocity of the en- 1 PCmax L tire phytoplankton assemblage. The measured RelV for each PN size P NO3 2 B PM fraction (Fig. 6A), corrected for the initial detritus content, did not 3 show an obvious temporal trend, nor was there a clear difference in PS

1.5 RelVNO3 between the different PN size fractions, except for the high L RelVNO3 of P after the second day. In contrast to the measured uptake rates, the modelled speciﬁc nutrient uptake rates are internally consis- 1 tent (i.e., values at a given time point are dependent on previous values such that mass balance is achieved by deﬁnition). Moreover, the

modelled relative VNO 0.5 modelled values are not confounded by inclusion of initial detritus into the biomass pool or violation of the tracer level assumptions during experimental incubations (Lipschultz, 2008), as likely happened for the 0 L NH4 uptake rate measurements beyond day 3. The RelVNO3 of P was ini- 3 PCmax tially higher than that of the smaller groups in the model configurations 2 C with a single optimised PCmax value (configuration a) or group-specific

3 optimised PCmax values (configuration b) (Fig. 6B and C), and even 1.5 more pronounced in the model configuration including nutrient acclimation or resolving higher PL diversity (Fig. S6). This early potential dominance is caused in part by the nutrient uptake characteristics and 1 the potential maximum production rate of each phytoplankton group. The effect of the latter is reflected in the extension of the period of PL L modelled relative VNO 0.5 high RelV when the P group is attributed a higher P in the NO3 Cmax model configurations using optimised group-specificPCmax values PL 0 (Fig. 6C). Even though the RelVNO3 decreased to the level of the total 0 1 2 3 4 5 6 7 8 assemblage during bloom development and concomitant nutrient days exhaustion (Fig. 6 and Fig. S6), the PL group sustained its competitive advantage longer than the smaller phytoplankton size groups while Fig. 6. Relative specificNO3 uptake rates, (A) measured and corrected for initial detritus content in each PN size fraction in B3 and modelled for each phytoplankton group in nitrate was plentiful. model configurations (B) using a single optimised PCmax value or (C) group-specific optimised PCmax values. A relative VNO3 value of 1 (dashed grey line) indicates that the relative specific uptake rate of a particular group is equal to that of the total assemblage. 4. Discussion

The mesocosm experiment (Fawcett and Ward, 2011) reproduced the conditions of a coastal upwelling event and the ensuing phytoplank- and the apparent need for a growth lag period to match the observed ton bloom, as evidenced by the timescale of nutrient drawdown and the [PN] changes in B3 (Fig. 4), we divided the largest phytoplankton size succession in phytoplankton community size structure as a whole and group into two subgroups representing initially abundant, slower grow- the diatom assemblage in particular (Dugdale and Wilkerson, 1989; LK ing species (P ;e.g.,E. zodiacus) and initially rare, faster growing Dugdale et al., 2006; Wilkerson et al., 2006). Phytoplankton species dom- Lr species (P ; e.g., Chaetoceros spp. and Thalassiosira spp.). Adding this inating the community at the end of the experiment (e.g., Chaetoceros fi biological diversity to the model (con guration d) did not improve the debilis, C. curvisetus and Thalassiosira anguste-lineata) were typical of fi fi overall t to the observations compared to the model con guration those that dominate the phytoplankton bloom assemblage following fi fi with only phytoplankton group-speci cPCmax values (i.e., con guration coastal upwelling (Lassiter et al., 2006; Venrick, 2009). To simulate Σ b; Fig. S3 and NRMSE in Table 1; ANOVA and Tukey's HSD test: F = these characteristic patterns of nutrient drawdown and partitioning of b N fi 430.89, p 0.001 and p 0.05). This model con guration, however, nitrate uptake into different phytoplankton size groups, it was neces- was able to better explain the accumulation of PN in the largest size sary to differentiate phytoplankton group-specific maximum produc- fraction (Table S2) and demonstrated that cells of the initially rare PLr tion rates, through optimisation of PCmax values. The inclusion of population could indeed achieve near theoretical maximum growth neither nutrient acclimation (through γNutmem), nor resolution of rates (Table 1 and Table S1). higher diversity within the largest model size group, representing diatoms, further improved the overall model fit significantly beyond 3.3. Specific nitrate uptake rates and community structure shift utilisation of appropriate group-specific maximum production rate values. While this implies that inclusion of additional diversity within

The measured and modelled maximum NO3 uptake rates for each PN functionally homogeneous groups was not necessary to explain the size fraction during the development of the bloom in B3 are shown in rates of new production during coastal upwelling conditions, we also Table 2 as a specific uptake rate (i.e., nutrient transport rate normalised demonstrated that an initially very small fraction of the PL population to [PN] for each size fraction or phytoplankton group and averaged over could indeed achieve near theoretical maximum growth rates (Table 1 N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25 23 and Table S1). Furthermore, acclimation of phytoplankton cells' physio- group was reduced, either by incorporating a slower community-level logical condition to higher nutrient conditions, although experimentally nutrient acclimation timescale or by dividing the large-sized community documented, was not essential to simulate the evolution of size- into initially abundant, slower growing species (e.g., E. zodiacus and partitioned new production during the experiment. This relatively Pseudonitzschia sp.) and initially rare, faster growing species parsimonious mechanistic physiological functioning and resulting (e.g., C. debilis and T. anguste-lineata). The apparent difference in growth biogeochemical signature provides a tantalizingly simple community strategy of the two bloom-forming diatom species, T. anguste-lineata level description. In the next sections we reconcile the seeming contra- and the smaller C. debilis, is noteworthy. While T. anguste-lineata was diction between the relatively simple biogeochemical outcome and the not detected during the first 4 days of the experiment, after which it underlying complex competitive interactions and resulting composi- grew exponentially, C. debilis was growing exponentially by the second tional changes at the group level observed through biodiversity day (Fig. 2). This is consistent with the findings of Collos (1982, 1986), indicators. and suggests a decoupling of nitrate uptake and growth by the generally larger T. anguste-lineata cells compared to the reduced capacity for ni- 4.1. Group-specific maximum production rates trate storage by C. debilis. Both growth strategies were successful in this particular context. All phytoplankton cells generally respond to a relative improvement in nutrient and light conditions with an increase in growth rate, and the 4.2. Nutrient acclimation saturation level of this growth response is specific to cell size and phylogenetic group (Barber and Hiscock, 2006; Litchman et al., 2010; The temporal lag between the metabolic stimulus and the Marañón et al., 2013). Such a size-specific growth response partly biogeochemically-relevant responses, such as nutrient uptake rate and explains the additive nature of the phytoplankton size distribution as cell growth, can be affected by both the prior physiological status of a function of resource concentration (Poulin and Franks, 2010)and the organism and the magnitude of the environmental shift (Collos, the accumulation of large to intermediate-sized diatoms during upwell- 1986; Collos et al., 2005; Smith et al., 1992). Nutrient-limited cells ing conditions. Thus, the use of multiple parameter values to describe exposed to a sudden increase in ambient nutrient concentration often growth characteristics (e.g., PCmax and nutrient half-saturation coeffi- display a time lag in growth or physiological acclimation to the new cients) of phytoplankton cells in ecosystem models is fundamental for condition (Collos, 1986; Collos et al., 2005). Given the lack of upwelling the accurate representation of different phytoplankton functional in the two weeks preceding the experiment (www.pfeg.noaa.gov) groups, which often differ in size (Follows et al., 2007). and the extremely low surface [NO3] at the time of the experiment In this study, the original parameterisation of the TOPAZ ecosystem (K. Johnson, personal communication), the phytoplankton cells model, which used a universal parameter value for maximal production from the surface that were used to inoculate the nutrient-rich 70 m rate based on maximum growth rate as a function of temperature depth water were likely nutrient stressed. The growth lag of the initially (Bissinger et al., 2008), could not reproduce the rate and size- abundant small flagellates and the lack of a detectable change in [PN] partitioning of new production from experimental observations over the first three days of the experiment partially support this hy- (Fawcett and Ward, 2011)(Fig. 3). Instead, lower, size group-specific pothesis. We acknowledge, however, that such mesocosm experiments maximal production rates were necessary to achieve correspondence entail certain known (e.g., the die-off of the Pseudonitzschia sp. popula- between the modelled output and observations. Compared to open tions) and unknown artifacts that could not be covered when modelling ocean systems to which the TOPAZ model is currently being applied, such an idealised system. Nonetheless, a small percentage of the larger the lower PCmax values generated by this study likely result from the phytoplankton cells began growing as early as the second day of incuba- sub-optimal conditions that characterise transient environments such tion (Fig. 2), suggesting a size group-dependent growth response to a as coastal upwelling systems, or from physiological characteristics of shift from nutrient-limited to nutrient-replete conditions. Alternatively, the species adapted to these environments. Within each mesocosm, these cells may have originated from the deeper, nutrient-rich water the optimum PCmax value of the smaller phytoplankton size groups such that they were light-limited but not nutrient-limited at the begin- was lower than that of the largest size group (Table 1). This was the ning of the experiment, and were thus susceptible to a light acclimation case even though grazing pressure, which would act to elevate PCmax response rather than nutrient acclimation (data not shown). Transitions during the optimisation process, had been taken into account. Such in species' abundances, community composition, and nitrate uptake low PCmax values for smaller size groups such as flagellates and small rates in response to changes in nutrient supply (i.e., upwelling) have Gymnodinoids are consistent with data compiled by Geider et al. been documented in field and experimental studies (Berges et al., (1997). Diatoms typically dominated the largest size fraction in the 2004; MacIsaac et al., 1985). Using a similar experimental set-up to mesocosm experiment and generally have high PCmax and μmax in natu- that described here (Fawcett and Ward, 2011), Berges et al. (2004) ral environments or cultures (Furnas, 1990; Geider et al., 1997). The observed a slower response to nitrate enrichment of nitrate reductase optimised PCmax value of the large, fast growing phytoplankton group activity by a phytoplankton assemblage dominated by small flagellates (PLr) was more similar to the predicted values (Bissinger et al., 2008; (10–24 h) compared to one dominated by chain-forming diatoms Marañón et al., 2013) than that of the other size groups since it was (5–6 h). Unfortunately, we cannot assess such early-stage growth or rooted in the growth rates of fast growing species determined by nutrient uptake acclimation of the larger diatom cells because of the microscopy (Table 1 and Fig. 2). One example of a large, fast growing lack of observations during this early time window (0–24 h). Consider- species is the chain-forming diatom C. debilis. Even though it has a ation of functional group- or size-specific acclimation rates in transient modest cell size by diatom standards (ESD = ~13 μm) it was retained environments merits further scrutiny in the context of the most adap- in the N20 μm size fraction during the size filtration procedure in the tive strategy in cellular resource allocation in agreement with the mesocosm experiment. Its cell size places C. debilis in the top part of recommendations of Menge et al. (2011). the unimodal distribution of growth rate versus cell size in Fig. 1 of Marañón et al. (2013), which, together with its ability to avoid 4.3. Dissolved inorganic nitrogen uptake rates microzooplankton grazing by bearing setae and forming cell chains, renders C. debilis a successful bloom species in a coastal upwelling Early in the mesocosm experiment, the largest phytoplankton size fi environment. fraction showed signi cantly higher RelVNO3, although all size fractions Considering the growth of the community in the largest size fraction eventually achieved similar maximum specificuptakerates(Table 2 as a whole, the initial delay in PN accumulation in this fraction in B3 was and Fig. 6). This fast response coupled with the maintenance of high up- better reproduced by the model when the initial growth rate of this size take rates appears to be the mechanism by which diatoms exploit 24 N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25 upwelling conditions (Fawcett and Ward, 2011). Using the calibrated models to predict the dynamics of the substantial nutrient and organic model, we derived nutrient uptake rates for the various phytoplankton matter fluxes occurring on continental shelves. size fractions that were internally consistent and not confounded by the presence of detritus in the initial, newly upwelled PN pool (Garside, Acknowledgements

1991). The maximum modelled VNO3 of the large, medium, and small phytoplankton size groups (Table 2) was in the range of the values re- We thank B. Song and P. Bhadury for the assistance with the sample ported by Dugdale et al. (1990, 2006) based on the empirical relation- collection and incubation experiments. The staff of Moss Landing ship between nitrate uptake and initial nitrate concentration in an Marine Laboratory and Captain L. Bradford of the R/V ‘John Martin’ upwelling region (0.06–0.09 h−1). Although the maximum values of were most cooperative and helpful. We are grateful to A. Gibson for modelled VNO of the different phytoplankton groups were not signifi- 3 performing the phytoplankton microscopy counts, to A. Babbin and O. cantly different from each other, the large phytoplankton group Coyle for the measurements of ammonium concentrations, to C. Stock attained its higher VNO earlier than the other size groups (Table 2 and 3 and A. Babbin for the stimulating discussions and anonymous reviewers Fig. 6). The general trend of VNO in the model representation mainly re- 3 for their comments and suggestions which helped to improve the flects group-specific differences in nitrate uptake affinity and P Cmax manuscript. This research was supported by the NOAA Climate Program values, which drive the uptake rate. In coastal upwelling regions, Office grant n° NA110AR4310058. enhanced capacity for nitrate uptake confers a competitive advantage upon diatoms for which several mechanisms have been proposed: i) uptake inhibition decoupled from cellular nutrient quota due to vac- Appendix A. Supplementary data uolar nutrient storage (Lomas and Glibert, 2000; Stolte and Riegman, 1995), ii) lower temperature optima for nitrate and nitrite reductase Supplementary data to this article can be found online at http://dx. suggesting a possible physiological adaptation to thrive in cool, NO3- doi.org/10.1016/j.jmarsys.2015.01.009. rich environments (Lomas and Glibert, 1999), or iii) a higher density of transport sites inferred from the isometric scaling of maximum References uptake rates (Aksnes and Cao, 2011; Marañón et al., 2013). The lower cellular surface area per biomass and the similar VNO of the large phy- Aksnes, D.L., Cao, F.J., 2011. Inherent and apparent traits in microbial nutrient uptake. Mar. 3 – toplankton group compared to the smaller groups in our model favour Ecol. Prog. Ser. 440, 41 51. Barber, R.T., Hiscock, M.R., 2006. A rising tide lifts all phytoplankton: growth response of explanation iii) above but do not exclude the other options. Overall, other phytoplankton taxa in diatom-dominated blooms. Glob. Biogeochem. Cycle 20. we found that when considering the full model of ecological function- Berges, J.A., Gibson, C.E., Stewart, B.M., 2004. Physiological responses of phytoplankton communities in the Irish Sea to simulated upwelling. Hydrobiologia 517, 121–132. ing, phytoplankton group-specific differences in PCmax and nutrient af- fi fi fi Bissinger, J.E., Montagnes, D.J.S., Sharples, J., Atkinson, D., 2008. Predicting marine phyto- nities were suf cient to represent group-speci c nutrient dynamics, plankton maximum growth rates from temperature: improving on the Eppley curve avoiding the addition of an unnecessary layer of complexity derived using quantile regression. Limnol. Oceanogr. 53, 487–493. Collos, Y., 1982. Transient situations in nitrate assimilation by marine diatoms — short- from the interpretational framework of VNO . 3 term uncoupling of nitrate uptake and reduction. J. Exp. Mar. Biol. Ecol. 62, 285–295. Collos, Y., 1986. Time-lag algal growth dynamics — biological constraints on primary 5. Conclusions production in aquatic environments. Mar. Ecol. Prog. Ser. 33, 193–206. Collos, Y., Vaquer, A., Souchu, P., 2005. Acclimation of nitrate uptake by phytoplankton to high substrate levels. J. Phycol. 41, 466–478. Applying the TOPAZ ecosystem model to accurately reproduce the Dugdale, R.C., Goering, J.J., 1967. Uptake of new and regenerated forms of nitrogen in observed patterns and timescales of size-partitioned new production primary productivity. Limnol. Oceanogr. 12, 196–206. in a non-steady state environment identified the importance of group- Dugdale, R.C., Wilkerson, F.P., 1989. New production in the upwelling center at Point fi Conception, California: temporal and spatial patterns. Deep Sea Res. A Oceanogr. speci c functional traits, through adjustments of group-dependent Res. Pap. 36, 985–1007. maximum production rates. This feature of phytoplankton physiology Dugdale, R.C., Wilkerson, F.P., Morel, A., 1990. Realization of new production in coastal is thus recognised as critical in determining species' response and rela- upwelling areas — a means to compare relative performance. Limnol. Oceanogr. 35, – tive success in transient environments such as during coastal upwelling. 822 829. Dugdale, R.C., Wilkerson, F.P., Hogue, V.E., Marchi, A., 2006. Nutrient controls on new It allows the diatom-dominated large size fraction in particular to re- production in the Bodega Bay, California, coastal upwelling plume. Deep-Sea Res. II spond rapidly to an injection of nitrate into surface water and accounts Top. Stud. Oceanogr. 53, 3049–3062. for the ability of large cells to attain relatively higher specific growth Dunne, J.P., Armstrong, R.A., Gnanadesikan, A., Sarmiento, J.L., 2005. Empirical and mechanistic models for the particle export ratio. Glob. Biogeochem. Cycles 19, 16. rates on nitrate early on. The phytoplankton size group-dependent Dunne, J.P., Sarmiento, J.L., Gnanadesikan, A., 2007. A synthesis of global particle export extent of downward revision from the theoretical maximum growth from the surface ocean and cycling through the ocean interior and on the seafloor. rates originally used in TOPAZ reflects the differences in taxonomic Glob. Biogeochem. Cycles 21. Dunne, J.P., John, J.G., Shevliakova, E., Stouffer, R.J., Krasting, J.P., Malyshev, S.L., Milly, composition of each size group, the averaging effect of amalgamating P.C.D., Sentman, L.T., Adcroft, A.J., Cooke, W., Dunne, K.A., Griffies, S.M., Hallberg, closely related species adapted to thrive in different environments, R.W., Harrison, M.J., Levy, H., Wittenberg, A.T., Phillips, P.J., Zadeh, N., 2013. GFDL's and the sub-optimal growth of the phytoplankton community as a ESM2 global coupled climate-carbon earth system models. Part II: carbon system formulation and baseline simulation characteristics. J. Clim. 26, 2247–2267. whole during transient environmental conditions. Potential limitations Edwards, K.F., Thomas, M.K., Klausmeier, C.A., Litchman, E., 2012. Allometric scaling of the parsimonious representation of physiology and biodiversity and taxonomic variation in nutrient utilization traits and maximum growth rate of may arise in, for example, post-upwelling conditions when larger phytoplankton. Limnol. Oceanogr. 57, 554–566. grazers catch-up with the growth of large phytoplankton cells and Eppley, R.W., 1972. Temperature and phytoplankton growth in sea. Fish. Bull. 70, 1063–1085. diverse ecological strategies (e.g., mechanical and chemical grazing Eppley, R.W., Peterson, B.J., 1979. Particulate organic-matter flux and planktonic new deterrence, mixotrophy or spore formation) may help in shaping the production in the deep ocean. Nature 282, 677–680. community structure and export flux. While strong shifts in biodiversity Estrada, M., Blasco, D., 1985. Phytoplankton assemblages in coastal upwelling areas. In: Bas, C., Margalef, R., Rubias, P. (Eds.), International Symposium of Upwelling of were observed within the largest, most competitive phytoplankton Western Africa. Instituto de Investigaciones Pesqueras, Barcelona. group, representation of only an averaged version of the most compet- Fawcett, S.E., Ward, B.B., 2011. Phytoplankton succession and nitrogen utilization during – itive of species in this group (diatoms) was sufficient to fittheobserved the development of an upwelling bloom. Mar. Ecol. Prog. Ser. 428, 13 31. Finkel, Z.V., 2001. Light absorption and size scaling of light-limited metabolism in marine biogeochemical changes in this idealised, nutrient replete framework, in diatoms. Limnol. Oceanogr. 46, 86–94. the absence of large grazers. With the advent of physical models using Finkel, Z.V., Beardall, J., Flynn, K.J., Quigg, A., Rees, T.A.V., Raven, J.A., 2010. Phytoplankton higher spatial resolution more appropriate to representing coastal in a changing world: cell size and elemental stoichiometry. J. Plankton Res. 32, 119–137. regions, inclusion of size group- or phylogenetic group-dependent max- Follows, M.J., Dutkiewicz, S., Grant, S., Chisholm, S.W., 2007. Emergent biogeography of imum photosynthetic rates could benefit the ability of ecosystem microbial communities in a model ocean. Science 315, 1843–1846. N. Van Oostende et al. / Journal of Marine Systems 148 (2015) 14–25 25

Frost, B.W., Franzen, N.C., 1992. Grazing and iron limitation in the control of phytoplank- Lomas, M.W., Glibert, P.M., 2000. Comparisons of nitrate uptake, storage, and reduction in ton stock and nutrient concentration — a chemostat analog of the pacific equatorial marine diatoms and flagellates. J. Phycol. 36, 903–913. upwelling zone. Mar. Ecol. Prog. Ser. 83, 291–303. Longhurst, A., Sathyendranath, S., Platt, T., Caverhill, C., 1995. An estimate of global prima- Furnas, M.J., 1990. In situ growth-rates of marine-phytoplankton — approaches to ry production in the ocean from satellite radiometer data. J. Plankton Res. 17, measurement, community and species growth-rates. J. Plankton Res. 12, 1117–1151. 1245–1271. Garcia, H.E., Locarnini, R.A., Boyer, T.P., Antonov, J.I., Zweng, M.M., Baranova, O.K., Johnson, MacIsaac, J.J., Dugdale, R.C., Barber, R.T., Blasco, D., Packard, T.T., 1985. Primary production D.R., 2010. World Ocean Atlas 2009: Nutrients (Phosphate, Nitrate, Silicate) vol. 4. cycle in an upwelling center. Deep-Sea Res. A Oceanogr. Res. Pap. 32, 503–529. U.S. Government Printing Office, Washington, D.C. Marañón, E., Cermeño, P., López-Sandoval, D.C., Rodriguez-Ramos, T., Sobrino, C., Huete- Garrison, D.L., Gibson, A., Coale, S.L., Gowing, M.M., Okolodkov, Y.B., Fritsen, C.H., Jeffries, Ortega, M., Blanco, J.M., Rodriguez, J., 2013. Unimodal size scaling of phytoplankton M.O., 2005. Sea-ice microbial communities in the Ross Sea: autumn and summer growth and the size dependence of nutrient uptake and use. Ecol. Lett. 16, 371–379. biota. Mar. Ecol. Prog. Ser. 300, 39–52. Margalef, R., 1978. Life-forms of phytoplankton as survival alternatives in an unstable Garside, C., 1991. Shift-up and the nitrate kinetics of phytoplankton in upwelling systems. environment. Oceanol. Acta 1, 493–509. Limnol. Oceanogr. 36, 1239–1244. Menden-Deuer, S., Lessard, E.J., 2000. Carbon to volume relationships for dinoflagellates, Geider, R.J., 1992. Respiration — taxation without representation. In: Falkowski, P.G., diatoms, and other protist plankton. Limnol. Oceanogr. 45, 569–579. Woodhead, A.D. (Eds.), Primary Productivity and Biogeochemical Cycles in the Sea, Menge, D.N.L., Ballantyne, F., Weitz, J.S., 2011. Dynamics of nutrient uptake strategies: Book 43. lessons from the tortoise and the hare. Theor. Ecol. 4, 163–177. Geider, R.J., MacIntyre, H.L., Kana, T.M., 1997. Dynamic model of phytoplankton growth Moloney, C.L., Field, J.G., Lucas, M.I., 1991. The size-based dynamics of plankton food and acclimation: responses of the balanced growth rate and the chlorophyll a:carbon webs. II. Simulations of three contrasting southern Benguela food webs. J. Plankton ratio to light, nutrient-limitation and temperature. Mar. Ecol. Prog. Ser. 148, 187–200. Res. 13, 1039–1092. Goebel, N.L., Edwards, C.A., Zehr, J.P., Follows, M.J., 2010. An emergent community ecosys- Morel, F.M.M., 1987. Kinetics of nutrient-uptake and growth in phytoplankton. J. Phycol. tem model applied to the California Current System. J. Mar. Syst. 83, 221–241. 23, 137–150. Goericke, R., 2011. The structure of marine phytoplankton communities-patterns, rules Muller-Karger, F.E., Varela, R., Thunell, R., Luerssen, R., Hu, C.M., Walsh, J.J., 2005. The im- and mechanisms. Calif. Coop. Ocean. Fish. Invest. Rep. 52, 182–197. portance of continental margins in the global carbon cycle. Geophys. Res. Lett. 32. Hansen, P.J., Bjørnsen, P.K., Hansen, B.W., 1997. Zooplankton grazing and growth: scaling Olenina, I., Hajdu, S., Edler, L., Andersson, A., Wasmund, N., Busch, S., Göbel, J., Gromisz, S., within the 2–2,000 μm body size range. Limnol. Oceanogr. 42, 687–704. Huseby, S., Huttunen, M., Jaanus, A., Kokkonen, P., Ledaine, I., Niemkiewicz, E., 2006. Hasle, G.R., Syvertsen, E.E., 1997. Marine diatoms. In: Tomas, C.R. (Ed.), Identifying Marine Biovolumes and size-classes of phytoplankton in the Baltic Sea. HELCOM Baltic Sea Phytoplankton. Academic Press Inc., San Diego. Environment Proceedings 106, p. 144. Holling, C.S., 1959. The components of predation as revealed by a study of small-mammal Poulin, F.J., Franks, P.J.S., 2010. Size-structured planktonic ecosystems: constraints, predation of the European pine sawfly. Can. Entomol. 91, 293–320. controls and assembly instructions. J. Plankton Res. 32, 1121–1130. Ingraham, J.L., Maaløe, O., Neidhardt, F.C., 1983. Growth of the Bacterial Cell. Sinauer Sakshaug, E., Holm-Hansen, O., 1977. Chemical composition of Skeletonema costatum Associates. (grev) cleve and Pavlova (monochrysis) lutheri (droop) green as a function of Irigoien, X., Huisman, J., Harris, R.P., 2004. Global biodiversity patterns of marine phyto- nitrate-limited, phosphate-limited, and iron-limited growth. J. Exp. Mar. Biol. Ecol. plankton and zooplankton. Nature 429, 863–867. 29, 1–34. Irwin, A.J., Finkel, Z.V., Schofield, O.M.E., Falkowski, P.G., 2006. Scaling-up from nutrient Smith, G.J., Zimmerman, R.C., Alberte, R.S., 1992. Molecular and physiological responses of physiology to the size–structure of phytoplankton communities. J. Plankton Res. 28, diatoms to variable levels of irradiance and nitrogen availability — growth of 459–471. Skeletonema costatum in simulated upwelling conditions. Limnol. Oceanogr. 37, Jochem, F.J., Smith, G.J., Gao, Y., Zimmerman, R.C., Cabello-Pasini, A., Kohrs, D.G., Alberte, 989–1007. R.S., 2000. Cytometric quantification of nitrate reductase by immunolabeling in the Steele, J.H., Henderson, E.W., 1992. The significance of interannual variability. In: Evans, marine diatom Skeletonema costatum. Cytometry 39, 173–178. G.T., Fasham, M.J.R. (Eds.), Towards a Model of Ocean Biogeochemical Processes. Johnson, K.S., Chavez, F.P., Elrod, V.A., Fitzwater, S.E., Pennington, J.T., Buck, K.R., Walz, Springer-Verlag, Heidelberg, Germany. P.M., 2001. The annual cycle of iron and the biological response in central California Steidinger, K.A., Tangen, K., 1997. Dinoflagellates. In: Tomas, C.R. (Ed.), Identifying Marine coastal waters. Geophys. Res. Lett. 28, 1247–1250. Phytoplankton. Academic Press Inc., San Diego. Key, T., McCarthy, A., Campbell, D.A., Six, C., Roy, S., Finkel, Z.V., 2010. Cell size trade-offs Stock, C., Dunne, J., 2010. Controls on the ratio of mesozooplankton production to primary govern light exploitation strategies in marine phytoplankton. Environ. Microbiol. 12, production in marine ecosystems. Deep-Sea Res. I Oceanogr. Res. Pap. 57, 95–112. 95–104. Stolte, W., Riegman, R., 1995. Effect of phytoplankton cell-size on transient-state nitrate Klausmeier, C.A., Litchman, E., Daufresne, T., Levin, S.A., 2004. Optimal nitrogen-to- and ammonium uptake kinetics. Microbiology-UK 141, 1221–1229. phosphorus stoichiometry of phytoplankton. Nature 429, 171–174. Straile, D., 1997. Gross growth efficiencies of protozoan and metazoan zooplankton and Klein, L., 2002. An Ecosystem Dynamics Model of Monterey Bay, California. MSc. Univer- their dependence on food concentration, predator–prey weight ratio, and taxonomic sity of Maine. group. Limnol. Oceanogr. 42, 1375–1385. Kudela, R.M., Banas, N.S., Barth, J.A., Frame, E.R., Jay, D.A., Largier, J.L., Lessard, E.J., Peterson, Sunda, W.G., Huntsman, S.A., 1995. Iron uptake and growth limitation in oceanic and T.D., Woude, A.J.V., 2008. New insights into the controls and mechanisms of plankton coastal phytoplankton. Mar. Chem. 50, 189–206. productivity in coastal upwelling waters of the Northern California Current System. Throndsen, J., 1997. The planktonic marine flagellates. In: Tomas, C.R. (Ed.), Identifying Oceanography 21, 46–59. Marine Phytoplankton. Academic Press Inc., San Diego. Lagarias, J.C., Reeds, J.A., Wright, M.H., Wright, P.E., 1998. Convergence properties of the Thunell, R., Benitez-Nelson, C., Varela, R., Astor, Y., Muller-Karger, F., 2007. Particulate Nelder–Mead simplex method in low dimensions. SIAM J. Optim. 9, 112–147. organic carbon fluxes along upwelling-dominated continental margins: rates and Lassiter, A.M., Wilkerson, F.P., Dugdale, R.C., Hogue, V.E., 2006. Phytoplankton assem- mechanisms. Glob. Biogeochem. Cycles 21. blages in the CoOP-WEST coastal upwelling area. Deep-Sea Res. II Top. Stud. Venrick, E.L., 2009. Floral patterns in the California Current: the coastal-offshore boundary Oceanogr. 53, 3063–3077. zone. J. Mar. Res. 67, 89–111. Li, W.K.W., 2002. Macroecological patterns of phytoplankton in the northwestern North Wilkerson, F.P., Dugdale, R.C., 1987. The use of large shipboard barrels and drifters to Atlantic Ocean. Nature 419, 154–157. study the effects of coastal upwelling on phytoplankton dynamics. Limnol. Oceanogr. Li, Q.P., Franks, P.J.S., Landry, M.R., Goericke, R., Taylor, A.G., 2010. Modeling phytoplank- 32, 368–382. ton growth rates and chlorophyll to carbon ratios in California coastal and pelagic Wilkerson, F.P., Lassiter, A.M., Dugdale, R.C., Marchi, A., Hogue, V.E., 2006. The phyto- ecosystems. J. Geophys. Res. Biogeosci. 115. plankton bloom response to wind events and upwelled nutrients during the CoOP Lipschultz, F., 2008. Isotope tracer methods for studies of the marine nitrogen cycle. In: WEST study. Deep-Sea Res. II Top. Stud. Oceanogr. 53, 3023–3048. Capone, D.G., Bronk, D.A., Mulholland, M.R., Carpenter, E.J. (Eds.), Nitrogen in the Wood, A.M., Everroad, R.C., Wingard, L.M., 2005. Measuring growth rates in microalgal Marine Environment. cultures. In: Andersen, R.A. (Ed.), Algal Culturing Techniques. Elsevier Academic Litchman, E., Klausmeier, C.A., Schofield, O.M., Falkowski, P.G., 2007. The role of functional Press, London. traits and trade-offs in structuring phytoplankton communities: scaling from cellular Zimmerman, R.C., Kremer, J.N., Dugdale, R.C., 1987. Acceleration of nutrient-uptake by to ecosystem level. Ecol. Lett. 10, 1170–1181. phytoplankton in a coastal upwelling ecosystem — a modeling analysis. Limnol. Litchman, E., Pinto, P.D., Klausmeier, C.A., Thomas, M.K., Yoshiyama, K., 2010. Linking Oceanogr. 32, 359–367. traits to species diversity and community structure in phytoplankton. Hydrobiologia 653, 15–28. Lomas, M.W., Glibert, P.M., 1999. Temperature regulation of nitrate uptake: a novel hypothesis about nitrate uptake and reduction in cool-water diatoms. Limnol. Oceanogr. 44, 556–572.