We established a long-term study at the Blanes Bay Microbial Observatory (BBMO), a well-studied (cf. Gasol et al. 2012) NW Mediterranean shallow (25 m) open bay with a sandy bottom. It is located in front of the town of Blanes, between the submarine Blanes Canyon to the north and the mouth of the La Tordera River to the south (see a map and further details in Guadayol et al. 2009). Determinations of gross primary production and net community production by means of the dissolved oxygen method and short-term enrichment bioassays indicated that the bay is net heterotrophic and oligotrophic, with production being limited most of the year by phosphorus (Duarte et al. 2004, Lucea et al. 2005, Pinhassi et al. 2006). A relatively good knowledge of carbon and sulfur cycling has accumulated over the last few years (Alonso-Sáez et al. 2008, Romera-Castillo et al. 2013, Simó et al. 2009). Most terrestrial inputs arrive as runoff from the surrounding coastal area, but there is no direct river input to the site.

Surface waters were sampled for a minimum of once a month at about 800 m offshore (41°39.9′N, 2°48.3′E), filtered through a 200-µm mesh net, and transported to the lab under dim light (within 1.5 h) in 20-L polycarbonate carboys. Only one surface sample per date was considered for this study. In situ water salinity and temperature were measured with a CTD model SAIV A/S SD204. Secchi disk depth was measured as an integrated estimate of water column turbidity and was used to derive the light diffusion coefficient following the relationship Zsd×Kt=1.7 (Kirk 1994), where Zsd is the depth of the Secchi disk and Kt is the light diffusion coefficient. Total irradiance was recorded hourly by a pyranometer at the nearby meteorological station of Malgrat de Mar (Catalan Meteorological Service, SMC). In situ irradiance was measured with a spherical downwelling irradiance Li-Cor sensor (Li-193S) and a PAR-UV radiometer (PUV-2500, Biospherical). An index of stratification was estimated as the temperature difference between the surface and values close to the maximal depth (20 m).

Dissolved inorganic nutrients (NO₃^–, NO₂^–, NH₄⁺, PO₄^3– and SiO₂) were analysed with an Alliance Evolution II autoanalyser following the methods described by Hansen and Koroleff (1999) with minor modifications. Samples were kept frozen at –20°C until analyses. Chlorophyll a (Chl a) concentration was measured by fluorometry (Turner Designs fluorometer) in 6.5 ml extracts (90% acetone, 4°C, overnight) of 150 mL of seawater filtered through GF/F (Whatman). Parallel filtration onto polycarbonate, 3-μm pore filters (Poretics) provided the Chl a concentrations associated with larger organisms.

Abundance of cyanobacteria (Prochlorococcus spp. and Synechococcus spp.), autotrophic picoeukaryotes and cryptomonads were determined by flow cytometry in a BectonDickinson FACSCalibur instrument following standard methods. Small picoeukaryotes (≤2 µm) were separated from large picoeukaryotes. Autotrophic nanoplankton were enumerated using epifluorescence microscopy. Abundances were converted to biomass using average C:cell conversion factors from the literature: 51±18 fg C cell^–1 for Prochlorococcus spp., 175±73 fg C cell^–1 for Synechococcus spp., 1319±813 fg C cell^–1 for picoeukaryotes (see references in Simó et al. 2009), and 220 fg C µm^–3 for nanophytoplankton (Børsheim and Bratbak 1987). Microphytoplankton were identified and counted with an inverted microscope after sedimentation in Utermöhl chambers. Width and length were measured for each cell and cell volumes were calculated as described in Guadayol et al. (2009). Conversion to C was done by applying the formula C=0.109 (cell vol.)^0.991 fg C cell^–1 (Montagnes et al. 1994). For the period March to May 2003 and then after 2004, microphytoplankton biovolume data were not available, and microphytoplankton carbon was estimated from size-fractionated Chl a. We observed a good linear relationship (N=9, R²=0.88; p<0.05) between the percentage of total Chl a in the fraction >3 μm and the percentage of biomass comprised by microphytoplankton plus one-half of the nanophytoplankton and one- fourth of the picoeukaryotes in the months with complete biomass data. Thus, the Chl a>3 μm fraction and the nano- and picophytoplankton biomass values were used to estimate microphytoplankton biomass for the dates on which we had microphytoplankton abundance but not biomass data (see details in Simó et al. 2009).

Monthly from 2003 to 2010 and once per season from 2011 to 2014, samples were retrieved from the surface and used to determine particulate primary production (PPp) using the ¹⁴C method (Steeman-Nielsen 1952) from P-E curves. We obtained 114 estimates of ¹⁴C incorporation and photosynthetic parameters during this 12-year period. In total the number of monthly samples ranged from 7 (January, August, December) to 15 (May), with an average of 10.

Thirteen 70-mL bottles (VWR) and one dark control (bottle wrapped with aluminium foil) were filled with seawater and inoculated with 10 µCi NaH¹⁴CO₃ (International Agency for ¹⁴C Determination, VKI, Denmark). The incubation was carried out in a water bath at in situ temperature for 2 h in a gradient of artificial PAR (10-1500 µmol photons m^–2 h^–1). Circulating water connected to the water bath maintained the temperature as close as possible to the one measured in situ. Illumination was provided by Phillips halogen lamps (50 W). The irradiance level at each position was directly measured with an Illuminova AB (Sweden) spherical PAR light-sensor placed inside the bottles. An intercalibration with a spherical Li-Cor quantum sensor proved that both sensors gave similar values (Morán and Estrada 2001). After the incubation, the samples were filtered through 0.2-µm pore size cellulose ester Millipore filters (GSWP02500). These filters were put into scintillation vials and left for 24 h in an HCl saturated atmosphere. Scintillation cocktail (4 ml, Optiphase Hisafe 2, Perkin Elmer) was then added to each vial, and the radioactivity was measured in a liquid scintillation counter (TriCarb2810, Perkin Elmer). Particulate primary production (PPp) was determined from the P-E curve and the in situ (sampling time) irradiance.

Chl a-normalized PPp measurements were initially fitted to the exponential model of Platt et al. (1980) using nonlinear least-squares regression (KaleidaGraph software):

P^B = P^B_s [1 – exp(–α E/P^B_s)] [exp(–ß E/ P^B_s)],

where P^B [mg C (mg Chl a)^–1 h^–1] is the Chl a-normalized photosynthetic rate, P^B_s [mg C (mg Chl a)^–1 h^–1] is the maximum Chl a-normalized photosynthetic rate without photoinhibition, α [mg C (mg Chl a)^–1 h^–1 (µmol photons m^–2 s^–1)^–1] is the initial slope of the P-E curve and ß (same units as α) is the photoinhibition parameter.

As fitted ß values were 0 in almost all cases, P^B_s in the model of Platt et al. was equivalent to P^B_max in the model proposed by Webb et al. (1974):

P^B = P^B_max [1– exp(–α E/ PBm)],

where P^B_max [mg C (mg Chl a)^–1 h^–1] is the maximum Chl a-normalized photosynthetic rate. The saturating irradiance or light saturation parameter (EK) was estimated as P^B_max/α.

In situ light levels and Chl a concentrations at the sea surface were used to estimate in situ surface primary production at the time of sampling. The values here should be taken as estimates of net particulate phytoplankton primary production since no correction for dark respiration was attempted. There has been some discussion about the use of photosynthetic parameters derived from artificial light sources for estimating in situ primary production, yet the limited number of studies comparing in situ vs P-E–derived carbon fixation rates fail to indicate significant biases with the latter method (see discussion in Morán and Estrada 2005). Daily estimates of total particulate production at the surface were obtained from the monthly average light irradiance in the sampling site, and the monthly-averaged P-E equations and these values were used to estimate average productions for all all the light hours. The integration of these values provided monthly-averaged daily production rates that were then compared with the oxygen-derived gross primary production values measured by Duarte et al. (2004) at the same site but for the period 1992-1998.

Phytoplankton ¹⁴C-derived specific growth rates (GR, d^–1) were calculated from primary production (PP) and phytoplankton carbon biomass (PC) assuming exponential growth as GR=ln[1+(PP/PC)]. Maximum growth rates (GR_max) were estimated from P^B_max values and C:Chl a ratios as GR_max=ln[1+(P^B_m/PC:Chl a)]. No significant differences were observed when a linear, not a logarithmic, model was assumed for the calculation.

We used a basic data analysis strategy in which we averaged all measurements done for each month of the year and we present averages plus standard errors as well as maximum and minimum values for each month. This is the type of data presented in Figures 1 to 6.

In addition, we analysed the variability from season to season, and for this the seasons were defined depending on water temperature and number of hours of light as follows: winter was considered to start when temperatures dropped below 17°C and the number of hours of light below 10 h d^–1. Spring from 12 h d^–1 of light and temperatures >13°C. Summer was considered from temperatures >18°C and more than 15 h d^–1 of light. When temperatures decreased below 23°C (from the maximum of 26°C) and illumination was <12 h d^–1, we considered that it was Autumn. These values coincided approximately with the beginning of November (winter), mid-March (spring), the beginning of June (summer) and mid-September (autumn). Differences were tested with one-way ANOVA in JMP software version 9.0 (SAS Institute).

Between-variable Bonferroni-corrected Pearson correlations between untransformed data were produced in SYSTAT (Systat Software Inc, London, UK) after assuring normality and variance homogeneity. We used R (R Core Team 2014) and the package vegan (Oksanen et al. 2016) to perform a principal components analysis (PCA) with the independent variables (i.e. photosynthetic parameters, growth and photosynthesis rates) after centering and scaling all variables. The physico-chemical and phytoplankton data were then correlated to the PCA-determined main axes. We also ran simple and partial Mantel tests between three normalized data dissimilarity matrices: the environmental one (temperature, nutrients, salinity, etc.), the phytoplankton community composition one, and a third one with the independent variables of interest. The Pearson probabilities were determined after 10000 permutations. The partial Mantel tests were used to determine how much the phytoplankton matrix explained the photophysiology and production values after accounting for the environmental effects on both matrices.

The Blanes Bay Microbial Observatory is a shallow site (max. 25 m depth) close to the coast (800 m) and relatively unaffected by river discharges and human impacts (Guadayol et al. 2009) that we have identified as a reference for coastal oligotrophic stations. The seasonal cycle of oceanographic variables is typical for a temperate coastal system. June and July showed the highest number of light hours (ca. 15 h) and December and January the lowest (ca. 9 h), and this resulted in temperature (Fig. 1A) varying from a minimum of 12°C (February/March) to a maximum of 24°C (August). The site was therefore stratified in summer (April-October) and well mixed between November and March (Fig. 1B). Salinity varied from a minimum of 37.5 in March to a maximum of 38.0 in September (Fig. 1B).

Inorganic nutrients had maximum concentrations in winter and were lower from May to September/October (Fig. 2). Nitrate and silicate were below 1 µmol L^–1 in summer and reached 3 µmol L^–1 on average in March (Fig. 2A). Ammonia remained at 1 µmol L^–1 in summer and increased slightly in the spring. Concentrations of phosphate and nitrite in summer were near 0.1 µmol L^–1 and increased to 0.2-0.4 µmol L^–1 in winter (Fig. 2B).

Average (SE) Chl a concentration was 0.64 µg L^–1 (±0.04) (Table 1), with winter values around 1 µg L^–1 (Fig. 3A) and summer values of 0.3 µg L^–1. The fraction of Chl a<3 µm varied from ca. 0.2 µg L^–1 in summer to 0.5 µg L^–1 in winter, and was a larger fraction of total Chl a in summer. There was a marked seasonality of the various components of the picophytoplankton, with larger abundances of small (Fig. 3B) and large (Fig. 3C) picoeukaryotes in December-March, of Synechococcus between April and October, and of Prochlorococcus between September and December (Fig. 3B), yet the latter were never very abundant. Plastidic nanoflagellates enumerated by epifluorescence microscopy followed a very marked seasonality almost parallel to that of total Chl a (Fig. 3C). Finally, among the microphytoplankton, dinoflagellates were more abundant in early summer (May-August, Fig. 3D) and diatoms from mid-autumn to late spring (Fig. 3D). In terms of contribution to photosynthetic biomass (Fig. 4A), microphytoplankton dominated in late winter/early spring, picophytoplankton from July to October, and nanophytoplankton in late spring/early summer and then again in early winter. The ratio of carbon biomass to Chl a revealed a very clear seasonal cycle (Fig. 4B), with low C:Chl a values in autumn and early winter (<50), and values >100 from June to September.

We obtained 114 estimates of ¹⁴C incorporation and photosynthetic parameters during a 12-year period, from March 2003 to November 2014. While during the first 8 years we sampled at monthly intervals, we decreased the frequency after 2010 to every 3 months. Photoinhibition was observed only twice during this period, in May 2003 and in February 2012, with ß values of 0.001 and 0.046 mg C mg Chl a^–1 h^–1 (μmol photons m^–2 s^–1)^–1 respectively. The maximal photosynthetic rate P^B_max ranged one order of magnitude (0.5-15.0 mg C mg Chl a^–1 h^–1, Table 1, Fig 5A), and reached the maximum values in August and the minimum values in April and December. The initial or light-limited slope of the P-E relationship, α^B, averaged 0.007 (±0.001, range 0.001-0.045) mg C mg Chl a^–1 h^–1 (μmol photons m^–2 s^–1)^–1 but was rather stable seasonally, with somewhat lower values in spring (Fig 5A). The light saturation parameter or saturation irradiance, E_K, averaged 712±58 μmol photons m^–2 s^–1 and tended to be higher in the months with higher total irradiance (Fig. 5B). We compared the value of E_K with the radiation measured at the water surface and at 5 m depth. In winter, the observed values of E_K were close to the surface irradiance at the time of sampling. In summer, they were close to the irradiance values observed at 5 m depth. In all cases, the observed E_K were within the values delimited by the surface irradiance and that at 5 m depth.

Using the in situ irradiance and the measured photosynthetic parameters, we estimated in situ primary production (Fig. 5C). The hourly primary production rates varied between 0.07 and 7.34 mg C L^–1 h^–1, with an average value of 1.45±0.13. Primary production was higher in winter (Table 1), particularly in March, and lower in the summer months. Conversely, chlorophyll-specific primary production varied 200-fold between 0.05 and 12.6 mg C mg Chl a^–1 h^–1 (Table 1), with an average value of 2.49±0.19. Chlorophyll-specific primary production was higher in summer (average for July and August, 4.5 mg C mg Chl a^–1 h^–1) and lowest in winter (1.2 in December, Table 1).

With the estimated total phytoplankton carbon, we computed actual (i.e. at the in situ irradiance) and maximal (based on P^B_max) phytoplankton growth rates (Fig. 6). With an average of ca. 0.4-0.5 d^–1, the growth rates showed little seasonality (Fig. 6 and Table 1), with phytoplankton growing slightly faster in summer (~0.5 d^–1) than in winter (~0.35 d^–1). Maximum growth rates were on average 1.4-fold the in situ rates of phytoplankton growth.

With all data pooled, primary production was significantly related to chlorophyll (N=113, Pearson R=0.67, P<0.007), but the ordinary least square slope between these two variables was low (log-log slope of PPp as a function of Chl a, 0.53±0.11, P<0.001, N=113) as compared with previous global estimates (de Lafontaine and Peters 1986, Baines et al. 1994). Interestingly, the relationship changed (Fig. 7) from having a higher slope in winter (N=30, log-log slope: 0.95±0.36, r²=0.47, P=0.01) to having a lower one in spring (N=28, slope: 0.59±0.15, r²=0.53, P<0.01) and in the autumn (N=23, slope: 0.64±0.31, r²=0.37, P=0.04), and a non-significant value in summer (N=32, P=0.54). Production values in summer were above the general trend (i.e. more production per unit of chlorophyll), and this generated the observed seasonality in chlorophyll-specific production (Fig. 5C).

Given the strong seasonal component, with parallel changes in temperature, stratification, light penetration, nutrient levels and phytoplankton, it is no surprise that the parameters of the photosynthesis-irradiance model were correlated in some way or another to seasonally-varying variables. P^B_max, for example, was positively related to variables increasing in summer (i.e. temperature, stratification and less strongly Synechococcus abundance, Table 2). The initial slope of the P-E curve (α^B) was positively related to particulate production, and strongly related to diatom abundance (Pearson’s r=0.65, Table 2). The saturation irradiance E_K was only weakly correlated with phytoplankton maximum growth rate.

Total particulate production was positively correlated with Chl a (r=0.67) and with the abundance of phytoplankton groups developing in wintertime, such as picoeukaryotes, cryptomonads and photosynthetic flagellates (all r values >0.45, Table 2). In contrast, Chl a–specific production was correlated with the "summer” variables, i.e. temperature, stratification and light, and strongly with P^B_max (r=0.83). Phytoplankton growth rates were also correlated with in situ light, albeit not strongly, and with P^B_max, total and specific production.

A principal component analysis of the determined photophysiology and production variables identified two main axes that together explained 64% of the variance (axis 1, 42%; axis 2, 22%, Table 3). The first axis was determined by specific production, E_K, P^B_max and growth rate, while the second axis was determined by α^B and E_K. Correlation of the physical and biological variables with these two factors identified the first axis as contributed by stratification and temperature (positively) and silicate and phosphate (negatively), clearly indicating the physically-enforced seasonality. The second axis was related to phytoplankton composition, with positive loadings of Chl a, diatoms, picoeukaryotes and photosynthetic nanoflagellates. To further test the role of community composition in determining the photophysiology parameters and the rates of primary production, we constructed three similarity matrices, one with the environmental data alone (PCm), one with the parameters (PHOm), and one with the phytoplankton composition data (PHYm). Partial Mantel correlation tests between the three matrices indicated that phytoplankton community composition and the photophysiology parameters were related to the environmental data (Mantel r=0.42, P<0.001, and r=0.16, P=0.02 respectively, Table 4), while these two matrices were not significantly correlated. These two analyses support the minor role of phytoplankton community composition in primary production and variability of photophysiology parameters.

Finally, we calculated trends in annual anomalies in the parameters of interest using the approach of Mackas et al. (2001), as explained in O’Brien (2012), to remove the seasonal signal from the annual calculations. None of the regressions between annual anomalies and year (from 2003 to 2010, when we had estimates for all months) were significant for the photosynthesis and production variables, except for E_K, which showed an increasing trend with time (N=8, F=7.4, P=0.034).

The maximal photosynthetic rates measured in Blanes Bay (P^B_max, 0.5-15 mg C mg Chl a^–1 h^–1) were within values measured in other seasonal studies, i.e. 0.1-17.4 for pico- and total phytoplankton in the Bay of Biscay (Morán 2007, Morán and Scharek 2015), 0.8-9.9 in the Wadden Sea (Tillmann et al. 2000), 1-11 in the Black Sea (Finenko et al. 2002) and 0.1-21.2 mg C mg Chl a^–1 h^–1 in the Rhode river estuary (Gallegos 2012), and within values previously measured in the western Mediterranean Sea (0.6-5, Morán and Estrada 2001) and the Atlantic Ocean (0.3-14, Marañón and Holligan 1999). P^B_max varied through our data set about the same magnitude as Chl a (coefficient of variation, CV: 6.8 vs 7.9%). The variation observed at the surface in an Atlantic Ocean transect by Marañón and Holligan (1999) was about one order of magnitude lower. Our measured variability in surface waters was of a similar magnitude to that observed throughout the Atlantic and at various depths.

Temperature has been shown to explain a large percentage of the value of P^B_max in various environments (e.g. Tillmann et al. 2000). Yet, it is commonly assumed that P^B_max is higher when there is no nutrient limitation. Our summer values for Blanes Bay were larger than winter ones (Fig. 5A), and were thus contrary to the distributions of nutrients (Fig. 2). P^B_max was statistically significantly related to phytoplankton growth rates and to chlorophyll-specific production, as expected from the way in which it is calculated. Only the abundance of Synechococcus was somehow correlated with P^B_max, but this was probably due to covariation with temperature and stratification.

The values we measured of α^B, the initial slope of the P-E curve, varied more than P^B_max (CV: 29%) (range 0.001-0.045 mg C mg Chl a^–1 h^–1 (μmol phot m^–2 s^–1)^–1). They were lower than those measured in the Wadden Sea (0.02, Tillmann et al. 2000) and, while being in general low (see list in Keller 1988), they were similar to those measured in the Bay of Biscay (0.004-0.049, Morán and Scharek 2015) and comparable to some of the values of the central Atlantic (Kyewalyanga et al. 1998, Marañón and Holligan 1999) and the coast of Japan (Furuya et al. 1998). In the Atlantic, the values were particularly low at low latitudes and in surface waters. In most studies in the Mediterranean Sea, the values of α^B were also low, similar to ours either in winter (Morán and Estrada 2005) or spring (Morán and Estrada 2001, Gutiérrez-Rodríguez et al. 2010, Estrada et al. 2014). Low values of the initial slope of the P-E curve are typical of oligotrophic or high-light-intensity environments (Babin et al. 1996, Sakshaug et al. 1997, Finenko et al. 2002). Nutrients are generally low in Blanes Bay, particularly in summer, and we have repeatedly measured P deficiency of microbial growth in the area (e.g. Lucea et al. 2005, Pinhassi et al. 2006). In addition, Blanes Bay receives relatively high light intensities, as reflected also in the measured values of E_K.

The initial slope of the P-E curve was significantly correlated with production and, interestingly, was also positively correlated with the abundance of diatoms (Table 2), indicating that phytoplankton composition could be relevant in determining the capacity of the natural communities to use light at sub-saturating levels. The correlation was negative in Morán and Scharek (2015), suggesting that even for a putatively ecologically coherent group, the actual species composition plays a fundamental role in photophysiological responses. In a seasonal study, such as this one, species replacement might be more important than photoacclimation, and we observed little seasonality in the evolution of α^B (Fig 5). Kyewalyanga et al. (1998) found the blue-to-red absorption ratio of phytoplankton to statistically explain αB, and one can assume that changes in this absorption ratio are related to changes in pigment content, and thus to phytoplankton community structure. Additional statistical evidence that the structure of the phytoplankton community also helped in part to explain the measured parameters was obtained by the PCA analyses, but the Mantel tests failed to indicate that the effect of community structure was additional to the effect of the seasonal variations in the environmental parameters.

In Blanes Bay, α^B was not correlated with P^B_max. This has been named "E_K -dependent variability”, when physiological adjustments in response to changing light, i.e. photoacclimation, drive the relationship between production and Chl a (Behrenfeld et al. 2004). In other seasonally studied environments, the two parameters were correlated (Tillmann et al. 2000, Morán and Scharek 2015), yet Morán (2007) observed no correlation between these parameters for the picophytoplankton fraction, but a positive correlation for the large phytoplankton (Morán and Scharek 2015). The positive correlation has been observed at sites dominated by phytoplankton with a larger cell size (Claustre et al. 1997, Basterretxea and Arístegui 2000, Morán and Estrada 2001). We did observe a summer increase in E_k (the ratio between P^B_max and alpha) in Blanes Bay, although this was modest (see Fig. 5 and Table 2). Picophytoplankton contribution to total phytoplankton biomass is large in summer (Figs 2 and 4), and this could explain the lack of correlation observed between the two photophysiological parameters.

Our measured values of E_k were relatively high, comparable to those from the surface waters of the low-latitude Atlantic Ocean (Marañón and Holligan 1999) but within the limits of other studies (i.e. Morán and Scharek 2015). These relatively high values of E_k concur with the relatively low values of α^B and with the light intensity that the communities were receiving in situ. Photoacclimation of the algal communities also explained why we found only two instances of photoinhibition. The latter could probably be explained by deep-water intrusions into the continental shelf, bringing non-adapted communities from deeper waters. These intrusions have been described for the area of Blanes Bay (Masó and Tintoré 1991).

Peeters et al. (1991) suggested that the measured values of E_k could be compared with the average irradiance in the mixed layer (Eav) to detect instances of light limitation of phytoplankton growth. If Eav/E_k were <1, this would indicate that the algae had the potential for more photosynthesis but there was not enough irradiance reaching the cells. In the Wadden sea Tillmann et al. (2000) observed average values of the ratio well below one. In the Bay of Blanes we observed average values of Eav/E_K=2.3, summer values above 3, and values clearly below 1 only in November (0.69), indicating that light was not limiting photosynthesis in the Bay. In fact, the ratio Eav/E_K was significantly positively correlated with temperature, transparency and stratification, and negatively correlated with phosphate and silicate concentrations (r=–0.8, P<0.05), indicating that when more light is available and less nutrients are present, the irradiance available is enough to meet the needs of the phytoplankton community. Blanes Bay showed clear nutrient deficiency in summer (Fig. 2).

We observed marked seasonal patterns of total particulate primary production and of Chl a-specific production rates, which were remarkably opposed (Pearson r=0.41, N=114, P<0.001). The seasonal trends are more evident for surface values (Fig. 5C) than for depth-integrated values. Seasonality is a strong force governing the dynamics of temperate coastal environments, and we had previously observed this seasonality in carbon (Alonso-Sáez et al. 2008) and sulfur (Simó et al. 2009) cycling, as well as in microbial community structure (Alonso-Sáez et al. 2007). Here we show that both phytoplankton community structure (Fig. 3), and its function (Fig. 5) also vary strongly seasonally.

As a consequence, the values of the C:Chl a ratio also varied strongly with the season (Fig. 4), with values above 80 in summer and around 40-70 in winter. These values are in line to those estimated or measured in other studies (e.g. Arin et al. 2002). Gutiérrez-Rodríguez et al. (2010) estimated C:Chl a values ranging from 34-163 in bloom and non-bloom conditions in the NW Mediterranean and Gutiérrez-Rodríguez et al. (2010) values of 11-135 in Blanes Bay, also with maximum values in summer and minimum in winter. Marañón (2005) calculated values of 100-170 for central Atlantic surface waters and in a recent data review (Marañón et al. 2014) presented average values of 50-117 for a variety of marine ecosystem types. Physiological studies show that this ratio tends to increase when the irradiance is higher and nutrients are low (Geider et al. 1998), explaining the summer maxima at Blanes Bay.

The hourly values of primary production varied by two orders of magnitude (0.07-7 mg C m^–3 h^–1), with an average of 1.45±0.13, maximal values in late winter and lower values in summer. The Chl a-specific values ranged from 0.05-13 (average, 2.49±0.19 mg C mg Chl a^–1 h^–1), with maximum values in summer. These values are similar to some of those measured in coastal waters of the eastern Mediterranean (e.g. average 0.92 mg C m^–3 h^–1 and 3.6 mg C mg Chl a^–1 h^–1 in Azov 1986, Berman et al. 1984) and to those measured in winter in surface water of the NW Mediterranean (Morán and Estrada 2005), and are about one-third of total primary production in the Bay of Biscay (Morán and Scharek 2015). The values measured at Blanes Bay are also slightly lower than those measured in the nearby Banyuls Bay (10 and 70 mg C m^–3 d^–1), about 200 km north in the NW Mediterranean coast (Charles et al. 2005) but with rather similar seasonal trends.

The overall relationship between chlorophyll and primary production had a relatively low global log-log slope (0.53) compared with previous studies (slopes usually >0.7, de Lafontaine and Peters 1986, Baines et al. 1994, Gasol and Duarte 2000; or even >1, Marañón et al. 2014). Yet, a closer inspection of the data shows that the global low slope is driven by the summer data and, to a lesser extent, by the spring data. The summer Chl a and production were not significantly related, and at low Chl a values particulate production was maintained at a relatively large value (Fig. 7). These findings contrast with the winter data, in which Chl a and production were related with a log-log slope close to 1. This varying dependency of primary production on chlorophyll is well known from other sites (e.g. Steele and Baird 1962) and can be related to the higher efficiency of the phytoplankters that dominate the system in the summer, reflected in their higher P^B_max values and higher light utilization. It also casts doubts on the usage of Chl a as an estimator of production (see also Morán and Scharek 2015), especially in oligotrophic coastal regions such as the one studied here.

The estimated in situ (semi-instantaneous) rates of gross growth of the bulk phytoplankton community (average 0.42±0.03 d^–1, range 0.03-1.3, Table 1) were lower than those measured by the dilution method in Blanes Bay during the period 2003-2007 (range 0.3-1.9 d^–1, Gutiérrez-Rodríguez et al. 2011), which also showed a stronger seasonality than the one we observed (Fig. 6). Short-term (few days) variability in growth rates (from 0.1 to 0.8 d^–1) was observed in the SW Mediterranean Sea by Arin et al. (2002), suggesting that this large variability might end up averaged out in a long-term study such as ours, whereas it might be more evident in a one-per-month, one-year study.

Lignell et al. (2003) found that nutrient-replete growth rates of the whole phytoplanktonic community in the Baltic coastal zone ranged from 0.1 to 0.7 d^–1, with small (2 to 10 μm) algae growing faster (up to 1.5 d^–1) than cyanobacteria. Latasa et al. (2005) additionally argued that the relatively high growth rates they measured in summer in the Mediterranean were related to the overwhelming presence of eukaryotes within a community dominated by small phytoplankton. In Blanes Bay in summer, when the growth rates are higher, picophytoplankton dominate the phytoplankton community (Fig. 4), thus agreeing with the study by Lignell et al. (2003), but cyanobacteria dominate among the picophytoplankters (Fig. 3), thus disagreeing with the Latasa et al. study.

The rates of growth observed by Gutiérrez-Rodríguez et al. (2011) with the dilution method in spring and summer were >1.3 d^–1, while in the winter they were ~0.65 d^–1, although these are rates integrated through 24 h of incubation. Using data of ¹⁴C-based primary production for estimating phytoplankton growth rates has sometimes resulted in much lower values when compared with dilution-growth experiments (Moigis 1999, but see Figure 1 in Kirchman 2016). Marra (2002) proposed that short (<6 h) radiocarbon incubations would yield production rates closer to net rather than gross growth, yet Latasa et al. (2005) showed a good correspondence between growth rates estimated from primary production and short P-E experiments like the ones used here. Our measured growth rate values are substantially higher than those obtained in the oligotrophic Atlantic by Marañón et al. (2003) (0.2 to 0.3 d^–1), who also used fractionated ¹⁴C uptake and biomass measurements for estimating growth rate, and more similar to those estimated by Marañón (2005) from data compiled for the Atlantic Ocean (0.2-0.5 d^–1). There is ample evidence that low nutrient supply, such as that estimated for summer Blanes Bay communities, limits phytoplankton growth. The values we measured were much lower than those measured in nutrient-replete conditions in culture (Cullen et al. 1992). They were also relatively far from the maximum possible according to photophysiology (the rates reached were ca. 70% of the maximum, Table 1). Relatively modest growth rates would concur with the observed values of P^B_max (average 3.5), and also concur with the relatively high C:Chl a values measured in the Bay (average 65.2±3.7, summer values ca. 100, Fig. 4).

We measured average values of primary production of 1.45±0.13 mg C m^–3 h^–1 (Table 1). To calculate depth-integrated yearly-averaged values of production, we used the monthly averages of the photosynthetic parameters, water transparency and surface irradiance to model average monthly values of surface (Fig. 8) and integrated production, accounting for the natural and depth variability in total irradiance and day length. Calculating daily production on the basis of one single surface measurement per day (or one monthly average) implies assuming constant conditions of the main parameters (transparency, Chl a, and P^B_max and α^B) in the whole water column and throughout the day, when diel variations in the photosynthetic parameters are well documented, spanning over ten-fold across small temporal and spatial gradients (Harding et al. 1982, Côté and Platt 1983, Morán and Estrada 2001), and that P^B_max and α^B tend to vary predictably along the vertical profile (Morán and Scharek 2015). For our goals of obtaining a rough estimate of annual average production, however, the assumption of monthly-averaged parameters should allow for a reasonable estimate.

The average seasonality of production was much smoother, since summer days are longer and irradiance is higher when production is lower, and this generates an average surface value slightly above 1.2 mg C m^–3 h^–1. Considering water transparency, which changes seasonally, and integrating daily (with the changing light during the day) and yearly, this value translated into 48 g C m^–2 yr^–1, equivalent to daily value of 0.13 g C m^–2 d^–1. This value would be only 7% larger (51 gC) if we considered the summer (April-October) stratification of the site (which in general is 2-3 degrees colder in the lower 10 m than in the first 10 m) and we assumed half the value of P^B_max and double α^B values in the deeper waters, as occurs at other coastal sites (Morán and Scharek 2015).

Coastal stations in the Mediterranean Sea, such as the Bay of Naples (over a deeper water column) or the Adriatic coast, over similar relatively shallow water columns, showed values above those we have measured in Blanes Bay (ca. 150 g C m^–2 yr^–1, Pugnetti et al. 2006). The Bay of Marseille and the Banyuls area also showed higher values (88 g C m^–2 yr^–1, Lefevre et al. 1997; and 86-142 g C m^–2 yr^–1, Jacques 1970). Similarly, some old estimates for the Catalan Sea were also higher (76-88 g C m^–2 yr^–1, Margalef and Ballester 1967, Margalef and Castellví 1967). In the classical compilation of Sournia (1973), the values for coastal Mediterranean sites ranged between 30 and 140 g C m^–2 yr^–1, with the lowest values in the eastern Mediterranean. Estrada (1996), on the bases of published data, assumed minimum values of about 80 g C m^–2 yr^–1 for the Catalan coast (but for stations deeper than that at Blanes). Open ocean stations in which production has been measured continuously also show higher values (Dyfamed station, 86-232 g C m^–2 yr^–1, Marty and Chiavérini 2002), although they have always integrated a much deeper water column than the one we integrated (only 20 m). In fact, the hourly rates of primary production in Blanes Bay (average 1.45 mg m^–3 h^–1) were similar to the values observed at the surface of the Dyfamed open ocean station (Fig. 6 in Marty and Chiavérini 2002), all of which confirm the true oligotrophic nature of the Blanes Bay site, and its utility as a reference for an oligotrophic NW Mediterranean model site.

Duarte et al. (2004) investigated community metabolism in the Bay of Blanes during the period 1992-1999 with the oxygen method. The values of gross primary production (GPP) that they measured in winter were higher than but comparable to (assuming a respiratory quotient, RQ of 1) our values of ¹⁴C based primary production (Fig. 8), but the summer values were between 2-fold and 3-fold higher (and in July >4-fold), coincident with the maximum respiration rates measured in Blanes Bay. Respiration covaried strongly with temperature, and net community production was very negative in summer (Duarte et al. 2004). In a compilation of published values, oxygen-based estimates of GPP were found to be on average 2.2 times higher than ¹⁴C-based particulate carbon production estimates (Regaudie-de-Gioux et al. 2014), a value similar to the average increase of 2 that we measured. Values derived using ¹⁴C-POC are thought (Marra 2002) to estimate something near net primary production (NPP, the production available to support phytoplankton growth after accounting for losses due to respiration and excretion) but also include the effects of trophic interactions within the incubated sample and do not account for DOC release, which may account for up to 40% of total primary production in the NW Mediterranean (Morán et al. 2010) and specifically 45% in Blanes Bay (Alonso-Sáez et al. 2008); nor do they account for respiratory losses by the whole community, which would be better captured with other methods. On the other hand, oxygen-based estimates are thought to approach GPP (Marra 2002) and are logically higher than the NPP values. It is interesting to note that we found seasonality in the difference between O₂-based production and ¹⁴C-PPp, with larger differences in the warmest months, and that this difference was well related to the mean community respiration measured by Duarte et al. (2004): Ratio GPP/¹⁴C-PPp=0.69+0.39×Resp, N=12, P<0.001, r²=0.49. In fact, discrepancies between these two estimates have often been identified, especially under conditions of low nutrients, high light, or low P/R ratios (Peterson 1980), which are the conditions characterizing Blanes Bay in summer.