Effects of elevated CO₂ and temperature on phytoplankton community biomass, species composition and photosynthesis during an experimentally induced autumn bloom in the western English Channel

The combined effects of elevated pCO₂ and temperature were investigated during an experimentally induced autumn phytoplankton bloom in vitro sampled from the western English Channel (WEC). A full factorial 36-day microcosm experiment was conducted under year 2100 predicted temperature (+4.5 ^∘C) and pCO₂ levels (800 µatm). Over the experimental period total phytoplankton biomass was significantly influenced by elevated pCO₂. At the end of the experiment, biomass increased 6.5-fold under elevated pCO₂ and 4.6-fold under elevated temperature relative to the ambient control. By contrast, the combined influence of elevated pCO₂ and temperature had little effect on biomass relative to the control. Throughout the experiment in all treatments and in the control, the phytoplankton community structure shifted from dinoflagellates to nanophytoplankton . At the end of the experiment, under elevated pCO₂ nanophytoplankton contributed 90 % of community biomass and was dominated by Phaeocystis spp. Under elevated temperature, nanophytoplankton comprised 85 % of the community biomass and was dominated by smaller nanoflagellates. In the control, larger nanoflagellates dominated whilst the smallest nanophytoplankton contribution was observed under combined elevated pCO₂ and temperature (∼ 40 %). Under elevated pCO₂, temperature and in the control there was a significant decrease in dinoflagellate biomass. Under the combined effects of elevated pCO₂ and temperature, dinoflagellate biomass increased and was dominated by the harmful algal bloom (HAB) species, Prorocentrum cordatum. At the end of the experiment, chlorophyll a (Chl a) normalised maximum photosynthetic rates ($P_{\mathrm{m}}^{\mathrm{B}}$) increased > 6-fold under elevated pCO₂ and > 3-fold under elevated temperature while no effect on $P_{\mathrm{m}}^{\mathrm{B}}$ was observed when pCO₂ and temperature were elevated simultaneously. The results suggest that future increases in temperature and pCO₂ simultaneously do not appear to influence coastal phytoplankton productivity but significantly influence community composition during autumn in the WEC.

1 Introduction

Oceanic concentration of CO₂ has increased by ∼ 42 % over pre-industrial levels, with a continuing annual increase of ∼ 0.4 %. The current CO₂ level has reached ∼ 400 µatm and has been predicted to rise to > 700 µatm by the end of this century (IPCC, 2013), with estimates exceeding 1000 µatm (Matear and Lenton, 2018; Raupach et al., 2007; Raven et al., 2005). With increasing atmospheric CO₂, the oceans continue to absorb CO₂ from the atmosphere, which results in a shift in oceanic carbonate chemistry resulting in a decrease in seawater pH, or ocean acidification (OA). The projected increase in atmospheric CO₂ and corresponding increase in ocean uptake is predicted to result in a decrease in global mean surface seawater pH of 0.3 units below the present value of 8.1 to 7.8 (Wolf-Gladrow et al., 1999). Under this scenario, the shift in dissolved inorganic carbon (DIC) equilibria has wide ranging implications for phytoplankton photosynthetic carbon fixation rates and growth (Riebesell, 2004).

Concurrent with OA, elevated atmospheric CO₂ and other climate active gases have warmed the planet by ∼ 0.6 ^∘C over the past 100 years (IPCC, 2014). Atmospheric temperature has been predicted to rise by a further 1.8 to 4 ^∘C by the end of this century (Alley et al., 2007). Phytoplankton metabolic activity may be accelerated by increased temperature (Eppley, 1972), which can vary depending on the phytoplankton species and their physiological requirements (Beardall et al., 2009; Boyd et al., 2013). Long-term data sets already suggest that ongoing changes in coastal phytoplankton communities are likely due to climate shifts and other anthropogenic influences (Edwards et al., 2006; Smetacek and Cloern, 2008; Widdicombe et al., 2010). The response to OA and temperature can potentially alter the community composition, community biomass and photophysiology. Understanding how these two factors may interact, synergistically or antagonistically, is critical to our understanding and for predicting future primary productivity (Boyd and Doney, 2002; Dunne, 2014).

Laboratory studies of phytoplankton species in culture and studies on natural populations in the field have shown that most species exhibit sensitivity, in terms of growth and photosynthetic rates, to elevated pCO₂ and temperature individually. To date, only a few studies have investigated the interactive effects of these two parameters on natural populations (e.g. Coello-Camba et al., 2014; Feng et al., 2009; Gao et al., 2017; Hare et al., 2007). Most laboratory studies demonstrate variable results with species-specific responses. In the diatom Thalassiosira weissflogii, for example, pCO₂ elevated to 1000 µatm and +5 ^∘C temperature synergistically enhanced growth, while the same conditions resulted in a reduction in growth for the diatom Dactyliosolen fragilissimus (Taucher et al., 2015). Although there have been fewer studies on dinoflagellates, variable responses have also been reported (Errera et al., 2014; Fu et al., 2008). In natural populations, elevated pCO₂ has stimulated the growth of pico- and nanophytoplankton (Boras et al., 2016; Engel et al., 2008) while increased temperature has reduced their biomass (Moustaka-Gouni et al., 2016; Peter and Sommer, 2012). In a recent field study on natural phytoplankton communities, elevated temperature (+3 ^∘C above ambient) enhanced community biomass but the combined influence of elevated temperature and pCO₂ reduced the biomass (Gao et al., 2017).

Phytoplankton species composition, abundance and biomass has been measured since 1992 at the time series station L4 in the western English Channel (WEC) to evaluate how global changes could drive future shifts in phytoplankton community structure and carbon biogeochemistry. At this station, sea surface temperature and pCO₂ reach maximum values during late summer and start to decline in autumn. During October, mean seawater temperatures at 10 m decrease from 15.39 ^∘C (±0.49 SD) to 14.37 ^∘C (± 0.62 SD). Following a period of CO₂ oversaturation in late summer, pCO₂ returns to near-equilibrium at station L4 in October when mean pCO₂ values decrease from 455.32 µatm (± 63.92 SD) to 404.06 µatm (± 38.55 SD) (Kitidis et al., 2012).

From a biological perspective, the autumn period at station L4 is characterised by the decline of the late summer diatom and dinoflagellate blooms (Widdicombe et al., 2010) when their biomass approaches values close to the time series minima (diatom biomass range: 6.01 (± 6.88 SD)–2.85 (± 3.28 SD) mg C m⁻³; dinoflagellate biomass range: 1.75 (± 3.28 SD)–0.66 (± 1.08 SD) mg C m⁻³). Typically, over this period nanophytoplankton becomes numerically dominant and biomass ranges from 20.94 (± 33.25 SD) to 9.38 (± 3.31 SD) mg C m⁻³, though there is considerable variability in this biomass.

Based on the existing literature, the working hypotheses of this study are that (1) community biomass will increase differentially under individual treatments of elevated temperature and pCO₂, (2) elevated pCO₂ will lead to taxonomic shifts due to differences in species-specific CO₂ concentrating mechanisms and/or RuBisCO specificity, (3) photosynthetic carbon fixation rates will increase differentially under individual treatments of elevated temperature and pCO₂, (4) elevated temperature will lead to taxonomic shifts due to species-specific thermal optima, and (5) temperature and pCO₂ elevated simultaneously will have synergistic effects.

The objective of the study was therefore to investigate the combined effects of elevated pCO₂ and temperature on phytoplankton community structure, biomass and photosynthetic carbon fixation rates during the autumn transition from diatoms and dinoflagellates to nanophytoplankton at station L4 in the WEC.

2 Materials and methods

2.1 Perturbation experiment, sampling and experimental setup

Experimental seawater containing a natural phytoplankton community was sampled at station L4 (50^∘15^′ N, 4^∘13^′ W) on 7 October 2015 from 10 m depth (40 L). The experimental seawater was gently pre-filtered through a 200 µm Nitex mesh to remove mesozooplankton grazers, into two 20 L acid-cleaned carboys. While grazers play an important role in regulating phytoplankton community structure (e.g. Strom, 2002), our experimental goals considered only the effects of elevated temperature and pCO₂, though the mesh size used does not remove microzooplankton. In addition, 320 L of seawater was collected into sixteen 20 L acid-cleaned carboys from the same depth for use as experimental media. Immediately upon return to the laboratory the media seawater was filtered through an in-line 0.2 and 0.1 µm filter (Acropak^™, Pall Life Sciences) and then stored in the dark at 14 ^∘C until use. The experimental seawater was gently and thoroughly mixed and transferred in equal parts from each carboy (to ensure homogeneity) to sixteen 2.5 L borosilicate incubation bottles (four sets of four replicates). The remaining experimental seawater was sampled for initial (T0) concentrations of nutrients, Chl a, total alkalinity, dissolved inorganic carbon, particulate organic carbon (POC) and nitrogen (PON) and was also used to characterise the starting experimental phytoplankton community. The incubation bottles were placed in an outdoor simulated in situ incubation culture system and each set of replicates was linked to one of four 22 L reservoirs filled with the filtered seawater media. Neutral density spectrally corrected blue filters (Lee Filter no. 061) were placed between polycarbonate sheets and mounted to the top, sides and ends of the incubation system to provide ∼ 50 % irradiance, approximating PAR measured at 10 m depth at station L4 on the day of sampling prior to starting experimental incubations (see Fig. S1 in the Supplement for time course of PAR levels during the experiment). The media was aerated with CO₂ free air and 5 % CO₂ in air precisely mixed using a mass flow controller (Bronkhorst UK Limited) and used for the microcosm dilutions as per the following experimental design: (1) control (390 µatm pCO₂, 14.5 ^∘C matching station L4 in situ values), (2) high temperature (390 µatm pCO₂, 18.5 ^∘C), (3) high pCO₂ (800 µatm pCO₂, 14.5 ^∘C) and (4) combination (800 µatm pCO₂, 18.5 ^∘C).

Initial nutrient concentrations (0.24 µM nitrate + nitrite, 0.086 µM phosphate and 2.14 µM silicate on 7 October 2015) were amended to 8 µM nitrate + nitrite and 0.5 µM phosphate. Pulses of nutrient inputs frequently occur at station L4 from August to December following heavy rainfall events and subsequent riverine inputs to the system (e.g. Barnes et al., 2015). Our nutrient amendments simulated these in situ conditions and were held constant to maintain phytoplankton growth. Previous pilot studies highlighted that, if these concentrations are not maintained, the phytoplankton population crashes (Keys, 2017). As the phytoplankton community was sampled over the transitional phase from diatoms and dinoflagellates to nanophytoplankton, the in situ silicate concentration was maintained to reproduce the silicate concentrations typical of this time of year (Smyth et al., 2010). Nutrient concentrations were measured at time point T0 only.

Media transfer and sample acquisition was driven by peristaltic pumps. Following 48 h acclimation in batch culture, semi-continuous daily dilution rates were maintained at between 10 and 13 % of the incubation bottle volume throughout the experiment. CO₂-enriched seawater was added to the high-CO₂ treatment replicates every 24 h, acclimating the natural phytoplankton population to increments of elevated pCO₂ from ambient to ∼ 800 µatm over 8 days followed by maintenance at ∼ 800 µatm as per the method described by Schulz et al. (2009). Adding CO₂-enriched seawater is the preferred protocol, since some phytoplankton species are inhibited by the mechanical effects of direct bubbling (Riebesell et al., 2010; Shi et al., 2009), which causes a reduction in growth rates and the formation of aggregates (Love et al., 2016). pH was monitored daily to adjust the pCO₂ of the experimental media (+/−) prior to dilutions to maintain target pCO₂ levels in the incubation bottles. The seasonality in pH and total alkalinity (TA) are fairly stable at station L4 with high pH and low dissolved inorganic carbon during early summer and low pH, high DIC throughout autumn and winter (Kitidis et al., 2012). By maintaining the carbonate chemistry over the duration of the experiment, we aimed to simulate natural events at the study site.

To provide sufficient time for changes in the phytoplankton community to occur and to achieve an ecologically relevant data set, the incubation period was extended well beyond short-term acclimation. Previous pilot studies using the same experimental protocols highlighted that, after ∼ 20 days of incubation, significant changes in community structure and biomass were observed (Keys, 2017). These results were used to inform a more relevant incubation period of 30+ days.

2.2 Analytical methods

2.2.1 Chlorophyll a

Chl a was measured in each incubation bottle. Triplicate samples of 100 mL from each replicate were filtered onto 25 mm GF/F filters (nominal pore size 0.7 µm), extracted in 90 % acetone overnight at −20 ^∘C and Chl a concentration was measured on a Turner Trilogy^™ fluorometer using the non-acidified method of Welschmeyer (1994). The fluorometer was calibrated against a stock Chl a standard (Anacystis nidulans, Sigma Aldrich, UK), the concentration of which was determined with a PerkinElmer^™ spectrophotometer at wavelengths 663.89 and 750.11 nm. Samples for Chl a analysis were taken every 2–3 days.

2.2.2 Carbonate system

Samples of 70 mL for total alkalinity (TA) and dissolved inorganic carbon analysis were collected from each experimental replicate, stored in amber borosilicate bottles with no head space and fixed with 40 µL of supersaturated Hg₂Cl₂ solution for later determination (Apollo SciTech^™ Alkalinity Titrator AS-ALK2; Apollo SciTech^™ AS-C3 DIC analyser, with analytical precision of 3 µmol kg⁻¹). Duplicate measurements were made for TA and triplicate measurements for DIC. Carbonate system parameter values for media and treatment samples were calculated from TA and DIC measurements using the programme CO2SYS (Pierrot et al., 2006) with dissociation constants of carbonic acid of Mehrbach et al. (1973) refitted by Dickson and Millero (1987). Samples for TA and DIC were taken for analysis every 2–3 days throughout the experiment.

2.2.3 Phytoplankton community analysis

Phytoplankton community analysis was performed by flow cytometry (Becton Dickinson Accuri^™ C6) for the 0.2 to 18 µm size fraction following Tarran et al. (2006) and inverted light microscopy was used to enumerate cells > 18 µm (BS EN 15204, 2006). For flow cytometry, 2 mL samples fixed with glutaraldehyde to a final concentration of 2 % were flash frozen in liquid nitrogen and stored at −80 ^∘C for subsequent analysis. Phytoplankton data acquisition was triggered on both chlorophyll fluorescence and forward light scatter (FSC) using prior knowledge of the position of Synechococcus sp. to set the lower limit of analysis. Density plots of FSC vs. Chl fluorescence, phycoerythrin fluorescence vs. Chl fluorescence and side scatter vs. Chl fluorescence were used to discriminate between Synechococcus sp., picoeukaryote phytoplankton (approx. 0.5–3 µm), coccolithophores, cryptophytes, Phaeocystis sp. single cells and nanophytoplankton (eukaryotes > 3 µm, excluding the coccolithophores, cryptophytes and Phaeocystis sp. single cells; for further information on flow cytometer calibration for phytoplankton size measurements, see the Supplement). For inverted light microscopy, 140 mL samples were fixed with 2 % (final concentration) acid Lugol's iodine solution and analysed by inverted light microscopy (Olympus^™ IMT-2) using the Utermöhl counting technique (Utermöhl, 1958; Widdicombe et al., 2010). Phytoplankton community samples were taken at T0, T10, T17, T24 and T36.

2.2.4 Phytoplankton community biomass

The smaller size fraction identified and enumerated through flow cytometry; picophytoplankton, nanophytoplankton, Synechococcus, coccolithophores and cryptophytes were converted to carbon biomass (mg C m⁻³) using a spherical model to calculate mean cell volume,

$$\begin{array}{}\text{(1)}& \left({\displaystyle \frac{\mathrm{4}}{\mathrm{3}}}\cdot \mathit{\pi }\cdot r^{\mathrm{3}}\right),\end{array}$$

and a carbon conversion factor of 0.22 pg C µm⁻³ (Booth, 1988). A conversion factor of 0.285 pg C µm⁻³ was used for coccolithophores (Tarran et al., 2006) and cell a volume of 113 µm³ and carbon cell⁻¹ value of 18 pg applied for Phaeocystis spp. (Widdicombe et al., 2010). Phaeocystis spp. were identified and enumerated by flow cytometry separately to the nanophytoplankton class due to high observed abundance in the high-pCO₂ treatment. Mean cell measurements of individual species/taxa were used to calculate cell bio-volume for the 18 µm + size fraction according to Kovala and Larrance (1966) and converted to biomass according to the equations of Menden-Deuer and Lessard (2000).

2.2.5 POC and PON

Samples for particulate organic carbon (POC) and particulate organic nitrogen (PON) were taken at T0, T15 and T36. Samples of 150 mL were taken from each replicate and filtered under gentle vacuum pressure onto pre-ashed 25 mm glass fibre filters (GF/F, nominal pore size 0.7 µm). Filters were stored in acid-washed petri-slides at −20 ^∘C until further processing. Sample analysis was conducted using a Thermoquest elemental analyser (Flash 1112). Acetanilide standards (Sigma Aldrich, UK) were used to calibrate measurements of carbon and nitrogen and also used during the analysis to account for possible drift in measured concentrations.

2.2.6 Chl fluorescence-based photophysiology

Photosystem II (PSII) variable chlorophyll fluorescence parameters were measured using a fast repetition rate fluorometer (FRRf) (FastOcean sensor in combination with an Act2Run laboratory system, Chelsea Technologies, West Molesey, UK). The excitation wavelengths of the FRRf's light-emitting diodes (LEDs) were 450, 530 and 624 nm. The instrument was used in single turnover mode with a saturation phase comprising 100 flashlets on a 2 µs pitch and a relaxation phase comprising 40 flashlets on a 50 µs pitch. Measurements were conducted in a temperature-controlled chamber at 15 ^∘C. The minimum (F_o) and maximum (F_m) Chl fluorescence were estimated according to Kolber et al. (1998). Maximum quantum yields of PSII were calculated as

$$\begin{array}{}\text{(2)}& F_{\mathrm{v}}/F_{\mathrm{m}}=(F_{\mathrm{m}}-F_{\mathrm{o}})/F_{\mathrm{m}}.\end{array}$$

PSII electron flux was calculated on a volume basis (JV_PSII; $\mathrm{mol}{\mathrm{e}}^{-}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{d}}^{-\mathrm{1}}$) using the absorption algorithm (Oxborough et al., 2012) following spectral correction by normalising the FRRf LED emission to the white spectra using Fast^PRO 8 software. This step required inputting the experimental phytoplankton community fluorescence excitation spectra values (FES). Since we did not measure the FES of our experimental samples, we used mean literature values for each phytoplankton group calculated proportionally (based on percentage contribution to total estimated biomass per phytoplankton group) as representative values for our experimental samples. The JV_PSII rates were converted to chlorophyll-specific carbon fixation rates (mg C (mg Chl a)⁻¹ ${\mathrm{m}}^{-\mathrm{3}}{\mathrm{h}}^{-\mathrm{1}}$), calculated as follows:

$$\begin{array}{}\text{(3)}& {\text{JV}}_{\mathrm{PSII}}\times {\mathit{\phi }}_{\mathrm{E}:\mathrm{C}}\times {\text{MW}}_{\mathrm{C}}/\text{Chl}a,\end{array}$$

where φ_E : C is the electron requirement for carbon uptake (molecule CO₂ (mol electrons)⁻¹), MW_C is the molecular weight of carbon and Chl a is the Chl a measurement specific to each sample. Chl a-specific JV_PSII-based photosynthesis–irradiance (PE) curves were conducted in replicate batches between 10:00 and 16:00 to account for variability over the photoperiod at between 8 and 14 irradiance intensities. The maximum intensity applied was adjusted according to ambient natural irradiance on the day of sampling. Maximum photosynthetic rates of carbon fixation ($P_{\mathrm{m}}^{\mathrm{B}}$), the light-limited slope (α^B) and the light saturation point of photosynthesis (E_k) were estimated by fitting the data to the model of Webb et al. (1974):

$$\begin{array}{}\text{(4)}& P^{\mathrm{B}}=(\mathrm{1}-e\times (-\mathit{\alpha }\times I/P_{\mathrm{m}}^{\mathrm{B}}\left)\right).\end{array}$$

Due to instrument failure during the experiment, samples for FRRf fluorescence-based light curves were taken at T36 only.

Figure 1 (a) MODIS weekly composite Chl a image of the western English Channel covering the period 30 September–6 October 2015 (coincident with the week of phytoplankton community sampling for the present study); processing courtesy of NEODAAS. The position of coastal station L4 is marked with a white diamond. (b) Profiles of weekly nutrient and Chl a concentrations from station L4 at a depth of 10 m over the second half of 2015 in the months prior to phytoplankton community sampling (indicated by the black arrow and text).

2.3 Statistical analysis

To test for effects of temperature, pCO₂ and possible time dependence of the measured response variables (Chl a, total biomass, POC, PON, photosynthetic parameters and biomass of individual species), generalised linear mixed models with the factors pCO₂, temperature and time (and all interactions) were applied to the data between T0 and T36. Analyses were conducted using the lme4 package in R (R Core Team, 2016).

3 Results

Chl a concentration in the WEC at station L4 from 30 September to 6 October 2015 (when sea water was collected for the experiment) varied between 0.02 and 5 mg m⁻³, with a mean concentration of ∼ 1.6 mg m⁻³ (Fig. 1a). Over the period leading up to phytoplankton community sampling, increasing nitrate and silicate concentrations coincided with a Chl a peak on 23 September (Fig. 1b). Routine net trawl (20 µm) sample observations indicated a phytoplankton community dominated by the diatoms Leptocylindrus danicus and L. minimus with a lower presence of the dinoflagellates Prorocentrum cordatum, Heterocapsa spp. and Oxytoxum gracile. Following decreasing nitrate concentrations, there was a P. cordatum bloom on 29 September, during the week before the experiment started (data not shown).

Figure 2 Calculated values of partial pressure of CO₂ in seawater (pCO₂) (a) and pH (b) from direct measurements of total alkalinity and dissolved inorganic carbon. (For full carbonate system values see Table S1 in the Supplement).

3.1 Experimental carbonate system

Equilibration to the target high pCO₂ values (800 µatm) within the high-pCO₂ and combination treatments was achieved at T10 (Fig. 2a, b). These treatments were slowly acclimated to increasing levels of pCO₂ over 7 days (from the initial dilution at T3), while the control and high-temperature treatments were acclimated at the same ambient carbonate system values as those measured at station L4 on the day of sampling. Following equilibration, the mean pCO₂ values within the control and high-temperature treatments were 394.9 (± 4.3 SD) and 393.2 (± 4.8 SD) µatm respectively, while in the high-pCO₂ and combination treatments mean pCO₂ values were 822.6 (± 9.4) and 836.5 (± 15.6 SD) µatm respectively. Carbonate system values remained stable throughout the experiment (for full carbonate system measured and calculated parameters, see Table S1 in the Supplement).

3.2 Experimental temperature treatments

Mean temperatures in the control and high-pCO₂ treatments were 14.1 (± 0.35 SD) ^∘C, and in the high-temperature and combination treatments the mean temperatures were 18.6 (± 0.42 SD) ^∘C, with a mean temperature difference between the ambient and high-temperature treatments of 4.46 (± 0.42 SD) ^∘C (Supplement, Fig. S2a, b).

Table 1 Results of generalised linear mixed model testing for effects of time, temperature, pCO₂ and all interactions on Chl a, phytoplankton biomass, and particulate organic carbon and nitrogen. Significant results are in bold; * p<0.05, ** p<0.01, *** p<0.001.

3.3 Chlorophyll a

Mean Chl a in the experimental seawater at T0 was 1.64 (± 0.02 SD) mg m⁻³ (Fig. 3a). This decreased in all treatments between T0 and T7 to ∼ 0.1 (± 0.09, 0.035 and 0.035 SD) mg m⁻³ in the control, high-pCO₂ and combination treatments, while in the high-temperature treatment at T7 Chl a was 0.46 mg m⁻³ (± 0.29 SD) (z = 2.176, p<0.05). From T7 to T12 Chl a increased in all treatments, which was highest in the combination (4.99 mg m⁻³ ± 0.69 SD) and high-pCO₂ treatments (3.83 mg m⁻³ ± 0.43 SD). Overall, Chl a was significantly influenced by experimental time, independent of experimental treatments (Table 1). At T36 Chl a concentration in the combination treatment was higher (6.87 (± 0.58 SD) mg m⁻³) than all other treatments while the high-temperature treatment concentration was higher (4.77 (± 0.44 SD) mg m⁻³) than the control and high-pCO₂ treatment. Mean concentrations for the control and high-pCO₂ treatment at T36 were not significantly different at 3.30 (± 0.22 SD) and 3.46 (± 0.35 SD) mg m⁻³ respectively (pairwise comparison t = 0.78, p = 0.858).

3.4 Phytoplankton biomass

The starting biomass in all treatments was 110.2 (± 5.7 SD) mg C m⁻³ (Fig. 3b). The biomass was dominated by dinoflagellates (∼ 50 %) with smaller contributions from nanophytoplankton (∼ 13 %), cryptophytes (∼ 11 %), diatoms (∼ 9 %), coccolithophores (∼ 8 %), Synechococcus (∼ 6 %) and picophytoplankton (∼ 3 %). Total biomass was significantly influenced in all treatments over time (Table 1), and at T10 it was significantly higher in the high-temperature treatment when biomass reached 752 (± 106 SD) mg C m⁻³ (z = 2.769, p<0.01). Biomass was significantly higher in the elevated pCO₂ treatment (interaction of time × high pCO₂) (Table 1), reaching 2481 (± 182.68 SD) mg C m⁻³ at T36, ∼ 6.5-fold higher than the control (z = 3.657, p<0.001). Total biomass in the high-temperature treatment at T36 was significantly higher than the combination treatment and ambient control (z=2.744, p<0.001), which were 525 (± 28.02 SD) mg C m⁻³ and 378 (± 33.95 SD) mg C m⁻³ respectively. Reaching 1735 (± 169.24 SD) mg C m⁻³, biomass in the high-temperature treatment was ∼ 4.6-fold higher than the control.

POC followed the same trends in all treatments between T0 and T36 (Fig. 3c) and was in close range of the estimated biomass (R²=0.914, Fig. 3d). POC was significantly influenced by the interaction of time × high pCO₂ and time × high temperature (Table 1). At T36 POC was significantly higher in the high-pCO₂ treatment (2086 ± 155.19 SD mg m⁻³) followed by the high-temperature treatment (1594 ± 162.24 SD mg m⁻³) – ∼ 5.4-fold and 4-fold higher than the control respectively, whereas a decline in POC was observed in the control and combination treatment. PON followed the same trend as POC over the course of the experiment, though it was only significantly influenced by the interaction between time and high pCO₂ (Fig. 3e, Table 1). At T36 concentrations were 147 (± 12.99 SD) and 133 (± 15.59 SD) mg m⁻³ in the high-pCO₂ and high-temperature treatments respectively, while PON was 57.75 (± 13.07 SD) mg m⁻³ in the combination treatment and 47.18 (± 9.32 SD) mg m⁻³ in the control. POC : PON ratios were significantly influenced by the interaction of time × high pCO₂ and time × high temperature (Table 1). The largest increase of 33 %, from 10.72 to 14.26 (± 1.73 SD) mol C : mol N, was in the high-pCO₂ treatment (73 % higher than the control), followed by an increase of 32 % to 9.83 (± 1.82 SD) mol C : mol N in the combination treatment (19 % higher than the control) and an increase of 17 % to 12.09 (± 2.14 SD) mol C : mol N in the high-temperature treatment (46 % higher than the control). In contrast, the POC : PON ratio in the control declined by 20 % from T0 to T36, from 10.33 to 8.26 (± 0.50 SD) mol C : mol N (Fig. 3f).

Table 2 Results of generalised linear mixed model testing for significant effects of time, temperature, pCO₂ and all interactions on phytoplankton species biomass. Significant results are in bold; * p<0.05, ** p<0.01, *** p<0.001.

Figure 3 Time course of Chl a (a), estimated phytoplankton biomass (b), POC (c), regression of estimated phytoplankton carbon vs. measured POC (d), PON (e) and POC : PON (f).

Figure 4 Percentage contribution to community biomass by phytoplankton groups/species throughout the experiment in the control (a), high-temperature (b), high-CO₂ (c) and combination treatments (d).

Figure 5 Response of individual phytoplankton groups to experimental treatments.

3.5 Community composition

From T0 to T24 the community shifted away from dominance of dinoflagellates in all treatments, followed by further regime shifts between T24 and T36 in the control and combination treatments. At T36 diatoms dominated the phytoplankton community biomass in the ambient control (Fig. 4a), while the high-temperature and high-pCO₂ treatments exhibited near-mono-specific dominance of nanophytoplankton (Fig. 4b, c). The most diverse community was in the combination treatment where dinoflagellates and Synechococcus became more prominent (Fig. 4d).

Between T10 and T24 the community shifted to nanophytoplankton in all experimental treatments. This dominance was maintained to T36 in the high-temperature and high-pCO₂ treatments whereas, in the ambient control and combination treatment, the community shifted away from nanophytoplankton (Fig. 5a). Nanophytoplankton biomass was significantly higher in the high-pCO₂ treatment (Table 2) with biomass reaching 2216 (± 189.67 SD) mg C m⁻³ at T36. This biomass was also high (though not significantly throughout the experiment until T36) in the high-temperature treatment (T36: 1489 (± 170.32 SD) mg C m⁻³, z = 1.695, p = 0.09) compared to the control and combination treatments. In the combination treatment, nanophytoplankton biomass was 238 (± 14.16 SD) mg C m⁻³ at T36, which was higher than the control, though not significantly (162 ± 20.02 SD mg C m⁻³). In addition to significant differences in nanophytoplankton biomass amongst the experimental treatments, treatment-specific differences in cell size were also observed. Larger nanoflagellates dominated the control (mean cell diameter of 6.34 µm) and smaller nanoflagellates dominated the high-temperature and combination treatments (mean cell diameters of 3.61 and 4.28 µm), whereas Phaeocystis spp. dominated the high-pCO₂ treatment (mean cell diameter 5.04 µm) and was not observed in any other treatment (Supplement, Fig. S3a–d).

At T0, diatom biomass was low and dominated by Coscinodiscus wailessii (48 %; 4.99 mg C m⁻³), Pleurosigma (25 %; 2.56 mg C m⁻³) and Thalassiosira subtilis (19 %; 1.94 mg C m⁻³). Small biomass contributions were made by Navicula distans, undetermined pennate diatoms and Cylindrotheca closterium. Biomass in the diatom group remained low from T0 to T24 but increased significantly through time in all treatments (Table 2), with the highest biomass in the high-pCO₂ treatment (235 ± 21.41 SD mg C m⁻³, Fig. 5b). The highest diatom contribution to total community biomass at T36 was in the ambient control (52 % of biomass; 198 ± 17.28 SD mg C m⁻³). In both the high-temperature and combination treatments, diatom biomass was lower at T36 (151 ± 10.94 SD and 124 ± 19.16 SD mg C m⁻³ respectively). In all treatments, diatom biomass shifted from the larger C. wailessii to the smaller C. closterium, N. distans, T. subtilis and Tropidoneis spp., the relative contributions of which were treatment specific. Overall N. distans dominated diatom biomass in all treatments at T36 (ambient control: 112 ± 24.86 SD mg C m⁻³, 56 % of biomass; high temperature: 106 ± 17.75 SD mg C m⁻³, 70 % of biomass; high pCO₂: 152 ± 19.09 SD mg C m⁻³, 61 % of biomass; and combination: 111 ± 20.97 SD mg C m⁻³, 89 % of biomass; Supplement, Fig. S4a–d).

The starting dinoflagellate community was dominated by Gyrodinium spirale (91 %; 49 mg C m⁻³), with smaller contributions from Katodinium glaucum (5 %; 2.76 mg C m⁻³), Prorocentrum cordatum (3 %; 1.78 mg C m⁻³) and undetermined Gymnodiniales (1 %; 0.49 mg C m⁻³). At T36 dinoflagellate biomass was significantly higher in the combination treatment (90 ± 16.98 SD mg C m⁻³, Fig. 5c, Table 2) followed by the high-temperature treatment (57 ± 6.87 SD mg C m⁻³, Table 2). There was no significant difference in dinoflagellate biomass between the high-pCO₂ treatment and ambient control at T36 when biomass was low. In the combination treatment, the dinoflagellate biomass became dominated by P. cordatum, which contributed 59 (± 12.95 SD) mg C m⁻³ (66 % of biomass in this group).

Synechococcus biomass was significantly higher in the combination treatment (reaching 59.9 ± 4.30 SD mg C m⁻³ at T36, Fig. 5d, Table 2) followed by the high-temperature treatment (30 ± 5.98 SD mg C m⁻³, Table 2). In both the high-pCO₂ treatment and control Synechococcus biomass was low (∼ 7 mg C m⁻³ in both treatments at T36), though an initial significant response to high pCO₂ was observed between T0 and T10 (Table 2). In all treatments and throughout the experiment, relative to the other phytoplankton groups, biomass of picophytoplankton (Fig. 5e), cryptophytes (Fig. 5f) and coccolithophores (Fig. 5g) remained low, though there was a slight increase in picophytoplankton in the combination treatment (11.26 ± 0.79 SD mg C m⁻³; Table 2).

Microzooplankton was dominated by Strombilidium spp. in all treatments throughout the experiment, though biomass was low relative to the phytoplankton community (Fig. 6). Following a decline from T0 to T10, microzooplankton biomass increased in all but the high-CO₂ treatment until T17 when biomass diverged. The biomass trajectory maintained an increase in the control when at T36 it was highest at ∼ 1.6 mg C m⁻³, 90 % higher than the high-temperature treatment (0.83 mg C m⁻³). Microzooplankton biomass was significantly lower in the high-CO₂ treatment at T36 ($z=-\mathrm{2.100}$, p=0.036) and undetected in the combination treatment at this time point (Table 2).

Table 3 FRRf-based photosynthesis–irradiance curve parameters for the experimental treatments on the final day (T36).

Table 4 Results of generalised linear model testing for significant effects of temperature, CO₂ and temperature × CO₂ on phytoplankton photophysiology at T36; $P_{\mathrm{m}}^{\mathrm{B}}$ (maximum photosynthetic rates), α (light-limited slope) and I_k (light-saturated photosynthesis). Significant results are in bold; * p<0.05, ** p<0.001, *** p<0.0001.

Figure 6 Microzooplankton biomass (dominated by Strombilidium sp.) relative to total phytoplankton biomass.

Figure 7 Fitted parameters of FRRf-based photosynthesis–irradiance curves for the experimental treatments on the final experimental day (T36).

3.6 Chl a fluorescence-based photophysiology

At T36, FRRf PE parameters were strongly influenced by the experimental treatments. $P_{\mathrm{m}}^{\mathrm{B}}$ was significantly higher in the high-pCO₂ treatment (18.93 mg C (mg Chl a)${}^{-\mathrm{1}}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{h}}^{-\mathrm{1}}$), followed by the high-temperature treatment (9.58 mg C (mg Chl a)${}^{-\mathrm{1}}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{h}}^{-\mathrm{1}}$; Fig. 7, Tables 3 and 4). There was no significant difference in $P_{\mathrm{m}}^{\mathrm{B}}$ between the control and combination treatments (2.77 and 3.02 mg C (mg Chl a)${}^{-\mathrm{1}}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{h}}^{-\mathrm{1}}$). Light-limited photosynthetic efficiency (α^B) also followed the same trend and was significantly higher in the high-pCO₂ treatment (0.13 mg C (mg Chl a)${}^{-\mathrm{1}}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{h}}^{-\mathrm{1}}$ $(\mathrm{\mu }\mathrm{mol}\mathrm{photon}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}{)}^{-\mathrm{1}}$) followed by the high-temperature treatment (0.09 mg C (mg Chl a)${}^{-\mathrm{1}}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{h}}^{-\mathrm{1}}$ $(\mathrm{\mu }\mathrm{mol}\mathrm{photon}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}{)}^{-\mathrm{1}}$; Tables 3 and 4). α^B was low in both the control and combination treatment (0.03 and 0.04 mg C (mg Chl a)${}^{-\mathrm{1}}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{h}}^{-\mathrm{1}}$ $(\mathrm{\mu }\mathrm{mol}\mathrm{photon}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}{)}^{-\mathrm{1}}$ respectively). The light saturation point of photosynthesis (E_k) was significantly higher in the high-pCO₂ treatment relative to all treatments (144.13 $\mathrm{\mu }\mathrm{mol}\mathrm{photon}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$), though significantly lower in the combination treatment relative to both the high-pCO₂ and high-temperature treatments (Tables 3 and 4).

4 Discussion

Individually, elevated temperature and pCO₂ resulted in the highest biomass and maximum photosynthetic rates ($P_{\mathrm{m}}^{\mathrm{B}}$) at T36, when nanophytoplankton dominated. The interaction of these two factors had little effect on total biomass with values close to the ambient control, and no effect on $P_{\mathrm{m}}^{\mathrm{B}}$. The combination treatment, however, exhibited the greatest diversity of phytoplankton functional groups, with dinoflagellates and Synechococcus becoming dominant over time.

Elevated pCO₂ has been shown to enhance the growth and photosynthesis of some phytoplankton species which have active uptake systems for inorganic carbon (Giordano et al., 2005; Reinfelder, 2011). Elevated pCO₂ may therefore lead to lowered energetic costs of carbon assimilation in some species and a redistribution of the cellular energy budget to other processes (Tortell et al., 2002). In this study, under elevated pCO₂ where the dominant group was nanophytoplankton, the most abundant species was the haptophyte Phaeocystis spp. Photosynthetic carbon fixation in Phaeocystis spp. is presently near saturation with respect to current levels of pCO₂ (Rost et al., 2003). Dominance of this spp. under elevated pCO₂ may be due to lowered grazing pressure since microzooplankton biomass was lowest in the high-CO₂ treatment throughout the experiment. The increased biomass and photosynthetic carbon fixation in this experimental community under elevated pCO₂ is due to the community shift to Phaeocystis spp. The increased biomass in the high-temperature treatment (where microzooplankton biomass remained stable between T17 and T36, though lower than the control) may be attributed to enhanced enzymatic activities, since algal growth commonly increases with temperature until after an optimal range (Boyd et al., 2013; Goldman and Carpenter, 1974; Savage et al., 2004). Optimum growth temperatures for marine phytoplankton are often several degrees higher than environmental temperatures (Eppley, 1972; Thomas et al., 2012). Nanophytoplankton also dominated in this treatment and, while Phaeocystis spp. was not discriminated, no further classification was made at a group/species level. Reduced biomass in the control from T24 onwards may be due to increased grazing pressure given the highest concentrations of microzooplankton biomass were observed in the control. Conversely, microzooplankton biomass declined significantly from T17 in the combination treatment, indicating reduced grazing pressure while phytoplankton biomass also declined from this time point. Nutrient concentrations were not measured beyond T0 and we cannot therefore exclude the possibility that differences in nutrient availability may have contributed to observed differences between control, high-temperature and high-CO₂ treatments.

4.1 Chl a

Biomass in the control peaked at T25 followed by a decline to T36. Correlated with this, Chl a also peaked at T25 in the control and declined to 3.3 mg m⁻³ by T27, remaining close to this value until T36. Biomass in the combination treatment peaked at T20 followed by decline to T36, whereas Chl a in this treatment declined from T20 to T25 followed by an increase at T27 before further decline similar to the biomass. Chl a peaked in this treatment again at T36 (6.8 mg m⁻³). We attribute the increase in Chl a between T25 and T27 (coincident with an overall biomass decrease) to lower species-specific carbon : Chl a ratios as a result of the increase in dinoflagellates, Synechococcus and picophytoplankton biomass from T25. We speculate that the decline in biomass under nutrient replete conditions in the combination treatment was probably due to slower species-specific growth rates when diatoms and dinoflagellates became more prominent in this treatment. Carbon : Chl a ratios in diatoms and dinoflagellates have previously been demonstrated to be lower than nano- and picophytoplankton (Sathyendranath et al., 2009). This contrasts the results reported in comparable studies as Chl a is generally highly correlated with biomass (e.g. Feng et al., 2009). Similar results were reported however by Hare et al. (2007), which indicates that Chl a may not always be a reliable proxy for biomass in mixed communities.

4.2 Biomass

This study shows that the phytoplankton community response to elevated temperature and pCO₂ is highly variable. pCO₂ elevated to ∼ 800 µatm induced higher community biomass, similar to the findings of Kim et al. (2006), whereas in other natural community studies no CO₂ effect on biomass was observed (Delille et al., 2005; Maugendre et al., 2017; Paul et al., 2015). A ∼ 4.5 ^∘C increase in temperature also resulted in higher biomass at T36 in this study, similar to the findings of Feng et al. (2009) and Hare et al. (2007), though elevated temperature has previously reduced the biomass of natural nanophytoplankton communities in the western Baltic Sea and Arctic Ocean (Coello-Camba et al., 2014; Moustaka-Gouni et al., 2016). When elevated temperature and pCO₂ were combined, community biomass exhibited little response, similar to the findings of Gao et al. (2017), though an increase in biomass has also been reported (Calbet et al., 2014; Feng et al., 2009). Geographic location and season also play an important role in structuring the community and its response in terms of biomass to elevated temperature and pCO₂. (Li et al., 2009; Morán et al., 2010). This may explain part of the variability in responses observed from studies on phytoplankton during different seasons and provinces.

4.3 Carbon : nitrogen

In agreement with others, the results of this experiment showed highest increases in C : N under elevated pCO₂ alone (Riebesell et al., 2007). C : N also increased under high temperature, consistent with the findings of Lomas and Glibert (1999) and Taucher et al. (2015). It also increased when pCO₂ and temperature were elevated, albeit to a lesser degree, which was also observed by Calbet et al. (2014), but contrasts other studies that have observed C : N being unaffected by the combined influence of elevated pCO₂ and temperature (Deppeler and Davidson, 2017; Kim et al., 2006; Paul et al., 2015). C : N is a strong indicator of cellular protein content (Woods and Harrison, 2003) and increases under elevated pCO₂ and warming may lead to lowered nutritional value of phytoplankton which has implications for zooplankton reproduction and the biogeochemical cycling of nutrients.

4.4 Photosynthetic carbon fixation rates

At T36, under elevated pCO₂, $P_{\mathrm{m}}^{\mathrm{B}}$ was > 6 times higher than in the control, but only one time point was measured, so we are not able to make decisive conclusions. Riebesell et al. (2007) and Tortell et al. (2008) also reported an increase in $P_{\mathrm{m}}^{\mathrm{B}}$ under elevated pCO₂. By contrast other observations on natural populations under elevated pCO₂ reported a reduction in $P_{\mathrm{m}}^{\mathrm{B}}$ (Feng et al., 2009; Hare et al., 2007). Studies on laboratory cultures have shown that increases in temperature cause an increase in photosynthetic rates (Feng et al., 2008; Fu et al., 2007; Hutchins et al., 2007), similar to what we observed in this study. In the combined pCO₂ and temperature treatment, we found no effect on $P_{\mathrm{m}}^{\mathrm{B}}$, which has also been observed in experiments on natural populations (Coello-Camba and Agustí, 2016; Gao et al., 2017). This contrasts the findings of Feng et al. (2009) and Hare et al. (2007) who observed the highest $P_{\mathrm{m}}^{\mathrm{B}}$ when temperature and pCO₂ were elevated simultaneously. In this study, increases in α^B and E_k under elevated pCO₂ and a decrease in these parameters when elevated pCO₂ and temperature were combined also contrast the trends reported by Feng et al. (2009). However, we should stress that, while our photophysiological measurements support our observed trends in community biomass, they were made on a single occasion at the end of the experiment. Future experiments should focus on acquiring photophysiological measurements throughout.

Species-specific photosynthetic rates have been demonstrated to decrease beyond their thermal optimum (Raven and Geider, 1988), which can be modified through photoprotective rather than photosynthetic pigments (Kiefer and Mitchell, 1983). This may explain the difference in $P_{\mathrm{m}}^{\mathrm{B}}$ between the high-pCO₂ and high-temperature treatments (in addition to differences in nanophytoplankton community composition in relation to Phaeocystis spp. discussed above), as the experimental high-temperature treatment in this study was ∼ 4.5 ^∘C higher than the control.

There was no significant effect of combined elevated pCO₂ and temperature on $P_{\mathrm{m}}^{\mathrm{B}}$, which was strongly influenced by taxonomic differences between the experimental treatments. Warming has been shown to lead to smaller cell sizes in nanophytoplankton (Atkinson et al., 2003; Peter and Sommer, 2012), which was observed in the combined treatment together with decreased nanophytoplankton biomass. Diatoms also shifted to smaller species with reduced biomass, while dinoflagellate and Synechococcus biomass increased at T36. Dinoflagellates are the only photoautotrophs with form II RuBisCO (Morse et al., 1995), which has the lowest carboxylation : oxygenation specificity factor among eukaryotic phytoplankton (Badger et al., 1998), which may give dinoflagellates a disadvantage in carbon fixation under present ambient pCO₂ levels. Phytoplankton growth rates are generally slower in surface waters with high pH (≥ 9) resulting from photosynthetic removal of CO₂ by previous blooms and the associated nutrient depletion (Hansen, 2002; Hinga, 2002). Though growth under high pH provides indirect evidence that dinoflagellates possess carbon concentrating mechanisms (CCMs), direct evidence is limited and points to the efficiency of CCMs in dinoflagellates as moderate in comparison to diatoms and some haptophytes (Reinfelder, 2011, and references therein). Given that dinoflagellates accounted for just ∼ 20 % of biomass in the combination treatment, exerting a minor influence on community photosynthetic rates, further work is required to explain the lower $P_{\mathrm{m}}^{\mathrm{B}}$ under the combined influence of elevated pCO₂ and temperature compared to the individual treatment influences. We applied the same electron requirement parameter for carbon uptake across all treatments, though in nature and between species there can be considerable variation in this parameter (e.g. 1.15 to 54.2 $\mathrm{mol}{\mathrm{e}}^{-}(\mathrm{mol}\mathrm{C}{)}^{-\mathrm{1}}$; Lawrenz et al., 2013), which can co-vary with temperature, nutrients, Chl a, irradiance and community structure. Better measurement techniques at quantifying this variability are necessary in the future.

4.5 Community composition

Phytoplankton community structure changes were observed, with a shift from dinoflagellates to nanophytoplankton which was most pronounced under single treatments of elevated temperature and pCO₂. Amongst the nanophytoplankton, a distinct size shift to smaller cells was observed in the high-temperature and combination treatments, while in the high-pCO₂ treatment Phaeocystis spp. dominated. Under combined pCO₂ and temperature from T24 onwards, however, dinoflagellate and Synechococcus biomass increased and nanophytoplankton biomass decreased. An increase in pico- and nanophytoplankton has previously been reported in natural communities under elevated pCO₂ (Bermúdez et al., 2016; Boras et al., 2016; Brussaard et al., 2013; Engel et al., 2008) while no effect on these size classes has been observed in other studies (Calbet et al., 2014; Paulino et al., 2007). Moustaka-Gouni et al. (2016) also found no CO₂ effect on natural nanophytoplankton communities but increased temperature reduced the biomass of this group. Kim et al. (2006) observed a shift from nanophytoplankton to diatoms under elevated pCO₂ alone while a shift from diatoms to nanophytoplankton under combined elevated pCO₂ and temperature has been reported (Hare et al., 2007). A variable response in Phaeocystis spp. to elevated pCO₂ has also been reported with increased growth (Chen et al., 2014; Keys et al., 2017), no effect (Thoisen et al., 2015) and decreased growth (Hoogstraten et al., 2012) observed. Phaeocystis spp. can outcompete other phytoplankton and form massive blooms (up to 10 g C m⁻³) with impacts on food webs, global biogeochemical cycles and climate regulation (Schoemann et al., 2005). While not a toxic algal species, Phaeocystis spp. are considered a harmful algal bloom (HAB) species when biomass reaches sufficient concentrations to cause anoxia through the production of mucus foam which can clog the feeding apparatus of zooplankton and fish (Eilertsen and Raa, 1995).

Recently published studies on the response of diatoms to elevated pCO₂ and temperature vary greatly. For example, Taucher et al. (2015) showed that Thalassiosira weissflogii incubated at 1000 µatm pCO₂ increased growth by 8 % while for Dactyliosolen fragilissimus, growth increased by 39 %; temperature elevated by +5 ^∘C also had a stimulating effect on T. weissflogii but inhibited the growth rate of D. fragilissimus; and when the treatments were combined growth was enhanced in T. weissflogii but reduced in D. fragilissimus. In our study, elevated pCO₂ increased biomass in diatoms (time dependent), but elevated temperature and the combination of these factors reduced the signal of this response. A distinct size shift in diatom species was observed in all treatments, from the larger Coscinodiscus spp., Pleurosigma and Thalassiosira subtilis to the smaller Navicula distans. This was most pronounced in the combination treatment where N. distans formed 89 % of diatom biomass. Navicula spp. previously exhibited a differential response to both elevated temperature and pCO₂. At +4.5 ^∘C and 960 ppm CO₂ Torstensson et al. (2012) observed no synergistic effects on the benthic Navicula directa. Elevated temperature increased growth rates by 43 % while a reduction of 5 % was observed under elevated CO₂. No effects on growth were detected at pH ranging from 8 to 7.4 units in Navicula spp. (Thoisen et al., 2015), while there was a significant increase in growth in N. distans along a CO₂ gradient at a shallow cold-water vent system (Baragi et al., 2015).

Synechococcus grown under pCO₂ elevated to 750 ppm and temperature elevated by 4 ^∘C resulted in increased growth and a 4-fold increase in $P_{\mathrm{m}}^{\mathrm{B}}$ (Fu et al., 2007) which is similar to the results of the present study.

The combination of elevated temperature and pCO₂ significantly increased dinoflagellate biomass to 17 % of total biomass. This was due to P. cordatum, which increased biomass by more than 30-fold from T0 to T30 (66 % of dinoflagellate biomass in this treatment). Despite the global increase in the frequency of HABs few studies have focussed on the response of dinoflagellates to elevated pCO₂ and temperature. In laboratory studies at 1000 ppm CO₂, growth rates of the HAB species Karenia brevis increased by 46 %; at 1000 ppm CO₂ and +5 ^∘C temperature its growth increased by 30 % but was reduced under elevated temperature alone (Errera et al., 2014). A combined increase in pCO₂ and temperature enhanced both the growth and $P_{\mathrm{m}}^{\mathrm{B}}$ in the dinoflagellate Heterosigma akashiwo, whereas in contrast to the present findings only pCO₂ alone enhanced these parameters in P. cordatum (Fu et al., 2008).

5 Implications

Increased biomass, $P_{\mathrm{m}}^{\mathrm{B}}$ and a community shift to nanophytoplankton under individual increases in temperature and pCO₂ suggest a potential negative feedback on atmospheric CO₂, whereby more CO₂ is removed from the ocean, and hence from the atmosphere through an increase in photosynthesis. The selection of Phaeocystis spp. under elevated pCO₂ indicates the potential for negative impacts on ecosystem function and food web structure due to the formation of hypoxic zones which can occur under eutrophication, inhibitory feeding effects and lowered fecundity in many copepods associated with this species (Schoemann et al., 2005; Verity et al., 2007). While more CO₂ is fixed, selection for nanophytoplankton in both of these treatments, however, may result in reduced carbon sequestration due to slower sinking rates of the smaller phytoplankton cells (Bopp et al., 2001; Laws et al., 2000). When temperature and pCO₂ were elevated simultaneously, community biomass showed little response and no effects on $P_{\mathrm{m}}^{\mathrm{B}}$ were observed. This suggests no change in feedback to atmospheric CO₂ and climate warming in future warmer high-CO₂ oceans. Additionally, combined elevated pCO₂ and temperature significantly modified taxonomic composition, by reducing diatom biomass relative to the control with an increase in dinoflagellate biomass dominated by the HAB species, P. cordatum. This has implications for fisheries, ecosystem function and human health.

6 Conclusion

These experimental results provide new evidence that increases in pCO₂ coupled with rising sea temperatures may have antagonistic effects on the autumn phytoplankton community in the WEC. Under future global change scenarios, the size range and biomass of diatoms may be reduced with increased dinoflagellate biomass and the selection of HAB species. The experimental simulations of year 2100 temperature and pCO₂ demonstrate that the effects of warming can be offset by elevated pCO₂, maintaining current levels of coastal phytoplankton productivity while significantly altering the community structure, and in turn these shifts will have consequences on carbon biogeochemical cycling in the WEC.