Underwater light climate and wavelength dependence of microalgae photosynthetic parameters in a temperate sea

Sampling area and strategy

Data were collected during a combined sampling campaign of the JERICO-NEXT program and the 2018 ECOPEL cruise in the EC, from the Strait of Dover (50°58.7′ N, 1°36.64′ E) to Brest (48°20.59′ N, 5°25.03′ W) from 18 April to 2 May 2018 (Fig. 1). Water was sampled at 19 locations at a depth of 2 m from inshore to offshore water (using a 20 L Niskin bottle) at different times of day. This zigzag sampling strategy was chosen to consider hydrobiological gradients between coastal/offshore and east/west waters (Vantrepotte et al., 2007). To characterize the sampling hour, which can influence phytoplankton physiology, we calculated the number of hours that had elapsed since sunrise (i.e. time since sunrise (TSS) in h). The EC is an epicontinental macrotidal temperate system with strong hydrodynamics and substantial river inputs, which provide contrasting light climates that are useful for testing the wavelength-dependence hypothesis of photosynthesis in natural phytoplankton communities. The water bodies sampled were thus used to experiment with different light climates along environmental gradients and with changes in phytoplankton community structure.

Controlling variables of photosynthesis

Two types of variables that could control photosynthesis were analyzed as ex-situ experimental conditions under which the communities grew: (i) abiotic variables and (ii) biotic variables that describe the phytoplankton community structure.

Wavelength-dependent photosynthesis parameters and functional absorption cross section of PSII

Wavelength-dependent PET_λ was studied using the MULTI-COLOR-PAM, which is particularly suitable for studying the PET_λ of phytoplankton (Schreiber, Klughammer & Kolbowski, 2012). It provides pulse-modulated measuring light, continuous actinic light, single-turnover light pulses and multiple-turnover or saturation pulses with peak wavelengths at 440 (bright blue), 480 (light blue), 540 (green), 590 (amber) and 625 nm (red light). See Schreiber, Klughammer & Kolbowski (2012) for a full description.

Before measuring PET_λ, samples were first dark-acclimated for 2.5 h (a compromise between the analyses and sampling strategy), without far red exposure (that would have locked the device for too long with respect to the many measurements required). The time of dark acclimation aims to optimize the maximum quantum yield of PSII measurements and neutralize the recent light history of cells (sampled in water columns of different depths and optical properties). It has been shown that there is not an universal protocol and the time required can exceed the classically considered time of 30 minutes and, in certain circumstances, durations of more than 2 hours are necessary (From et al., 2014). First, each dark acclimated sample was homogenized within an optical quartz cuvette with a magnetic stirrer then the light sensor US-SQS/WB Spherical Micro Quantum Sensor (Heinz Walz, Germany) was placed into the center of the cuvette to measure the photon flux density at each wavelength. This step provided the “PAR-list” file for each sample, which was used for all later measurements of that sample. Water samples filtered at 0.2 µm were used to determine the zero offset (i.e. the background signal to subtract from the total fluorescence signal at each wavelength). Next, a subsample of each dark acclimated sample was placed in a 2.5 mL cuvette with a 1 cm path length to adjust the measuring light and gain settings to it in order to obtain the same current fluorescence (Ft) level of 0.5 ± 0.05 (relative units) for all wavelengths and to get a good signal-noise ratio. This last step was used to compare fluorescence-rise kinetics (Szabó et al., 2014a).

Then, fast kinetic photosynthesis was measured to determine the wavelength-dependent functional absorption cross section of PSII (i.e. Sigma(II)_λ) by measuring O-I₁ fluorescence-rise kinetics repeatedly, as described by Schreiber, Klughammer & Kolbowski (2012). Sigma(II)_λ was estimated using the pre-programmed fast kinetic trigger file “Sigma1000.FTM”, in the same way as Szabó et al. (2014a, 2014b) and Schreiber & Klughammer (2013) did. In this phase of fast fluorescence, “O” was minimal fluorescence yield corresponding to all PSII reaction centers open. The full closure of PSII reaction centers τ (i.e. that of light-driven Q_A reduction during the O-I₁ rise) was obtained during a standard one ms long actinic illumination. Sigma(II)_λ was calculated according to Schreiber, Klughammer & Kolbowski (2012) as:

(4) $$\mathrm{S}\mathrm{i}\mathrm{g}\mathrm{m}\mathrm{a}(\mathrm{I}\mathrm{I}{)}_{\lambda }=\mathrm{l}\text{/}(\tau \times \mathrm{L}\times \mathrm{P}\mathrm{A}\mathrm{R})$$

where τ is the time constant (expressed here in seconds) of light driven Q_A reduction determined from the fast fluorescence kinetics measurements, L the Avogadro’s constant (6.022.10²³mol⁻¹), and PAR is the quantum flux density (that must be expressed here in mol quanta m⁻²s⁻¹) of the light driving the O–I₁ fluorescence rise.

Sigma(II)_λ was calculated by the user software interface (PAM-Win-3, Heinz Walz) based on the fitted value of the time constant τ obtained from three consecutive measurements separated by 10 s dark intervals, according to (Klughammer & Schreiber, 2015). Following this method, the estimate of Sigma(II)_λ is independent of Chla concentration. Sigma(II)_λ was determined from six subsamples to estimate the mean and variance of each natural community accurately (Supplementary Material, Fig. S1).

Next, automated rapid light curves (RLC) of the PET_λ were determined in triplicate (i.e. three independent samples) at each of the five wavelengths. For each RLC, samples were exposed to 14 actinic increasing light intensity levels, each 20 s long, as defined in the PAR-list file for each wavelength and sample. Hereafter, “PAR” refers to the photon flux measured at each wavelength, and the same wavelength was always used for the measuring light and actinic light. Saturation pulse settings were defined at a width of 300 µs. The effective quantum yield of PSII (Y(II)) was calculated at each step (Eq. 5). An initial step at 0 µmol quanta m⁻² s⁻¹ was used to determine F_v/F_m for samples acclimated to the dark for a long period (2.5 h) (Eq. 6). The relative electron transport rate (r.ETR) was then determined using Y(II), the PAR intensity of the corresponding wavelength (from the PAR-list file) and an arbitrary factor of 0.5 to indicate that PSI and PSII absorb light equally (Eq. (7)). The wavelength-dependent absolute electron transport rate of PSII (ETR(II)) reported in electrons (PSII s)⁻¹) was then calculated from Sigma(II)_λ (nm⁻²), the Avogadro’s constant (L, 6.022.10²³ mol⁻¹) and PAR(II) which is the rate of quantum absorption in PS II, in units of quanta (PS II s)⁻¹ according to Schreiber, Klughammer & Kolbowski (2012) at each of the five wavelengths (Eqs. (8) and (9)).

(5) $$\mathrm{Y}(\mathrm{I}\mathrm{I})=({\mathrm{F}}_{\mathrm{m}}^{\mathrm{\prime }}-\mathrm{F})/{\mathrm{F}}_{\mathrm{m}}^{\mathrm{\prime }}$$

(6) $${\mathrm{F}}_{\mathrm{v}}/{\mathrm{F}}_{\mathrm{m}}=({\mathrm{F}}_{\mathrm{m}}–{\mathrm{F}}_0)/{\mathrm{F}}_{\mathrm{m}}$$

(7) $$\mathrm{r}.\mathrm{E}\mathrm{T}\mathrm{R}=\mathrm{Y}(\mathrm{I}\mathrm{I})\times \mathrm{P}\mathrm{A}\mathrm{R}\times 0.5$$

(8) $$\mathrm{P}\mathrm{A}\mathrm{R}(\mathrm{I}\mathrm{I})=\mathrm{S}\mathrm{i}\mathrm{g}\mathrm{m}\mathrm{a}(\mathrm{I}\mathrm{I})\times \mathrm{L}\times \mathrm{P}\mathrm{A}\mathrm{R}$$

(9) $$\mathrm{E}\mathrm{T}\mathrm{R}(\mathrm{I}\mathrm{I})=\mathrm{P}\mathrm{A}\mathrm{R}(\mathrm{I}\mathrm{I})\times [\mathrm{Y}(\mathrm{I}\mathrm{I})/{\mathrm{F}}_{\mathrm{v}}/{\mathrm{F}}_{\mathrm{m}}]$$

NPQ (Bilger & Björkman, 1990) was calculated as the normalized Stern-Volmer quenching coefficient (Eq. (10)), according to Lavaud (2007):

(10) $$\mathrm{N}\mathrm{P}\mathrm{Q}={\mathrm{F}}_{\mathrm{m}}/{\mathrm{F}}_{\mathrm{m}}^{\mathrm{\prime }}-1$$

All photosynthetic parameters were obtained in triplicate for the five wavelengths of MULTI-COLOR-PAM for each sample.

The Eilers & Peeters (1988) model was used to fit r.ETR vs. PAR and ETR(II) vs. PAR(II) curves to estimate three photosynthetic parameters for each wavelength: light-use efficiency (i.e. α, the initial slope of the ETR vs. PAR curve), the maximum electron transport rate (ETR_max) and the optimum light parameter (E_op) in relative (r) and absolute (II) units. The light saturation parameter (E_k) was also calculated in the two units as E_k = ETR_max/α (Talling, 1957). To estimate the degree of photoacclimation of the phytoplankton communities, the E_k,440/E_avg ratio in Z_umixl at sampling was calculated for each location. The E_k ratios (in relative and absolute units) of three pairs of wavelengths (625/440, 540/440 and 540/625 nm) were calculated in the same way as the three pairs of in situ spectral bands (i.e. R/B, G/B and G/R measured in a water layer of 2.5 m in surface waters).

The Michaelis–Menten model was used to fit NPQ vs. PAR curves. A linear regression that forced the intercept to zero was used when the kinetics of these curves differed from the Michaelis–Menten model. Since two models were used, NPQ values were back-calculated using the calibrated models at two irradiances (of PE curves) for each wavelength: low PAR (300 µmol quanta m⁻² s⁻¹) and, for saturating conditions, high PAR (1,200 µmol quanta m⁻² s⁻¹), according to Szabó et al. (2014a).

All PE curves were fitted using the “fitEP” function of the “phytotools” package of R software R Core Team (2020) specifically designed to fit phytoplankton photosynthesis curves using simulated annealing (Silsbe & Malkin, 2015). The curves for r.ETR vs. PAR, ETR(II) vs. PAR(II) and NPQ vs. PAR were fitted for the three aggregated replicates, and all photosynthetic parameters were obtained at each of the five wavelengths.

Statistical analysis

All statistical analyses were performed with R version 3.6.0. For abiotic variables, the expectation-maximization with bootstrapping algorithm of Amelia II (Honaker, King & Blackwell, 2011) was used to determine missing values (n = 2) of light quality data. Principal component analysis (PCA) (Legendre & Legendre, 2012) of abiotic and biotic variables was performed using the “PCA” function of the “FactoMineR” package (Lê, Josse & Husson, 2007) to determine the internal structure of locations that best explained the variance in each datasets. All data were centered and reduced before performing the PCA analysis.

Statistical analysis of photosynthetic parameters followed a three-step approach (Fig. 2). First, wavelength dependence of each parameter was analyzed using a linear mixed-effects model (LMEM). Mixed-Effects Models are commonly used to fit regressions to repeated (i.e. longitudinal) measures (over time and/or space) by separating the variance explained by the main effects from that explained by random sampling, while considering the wavelength dependence of individuals. The most parsimonious model was linear, and higher-degree polynomials were not significant. LMEMs (Bates et al., 2015) were thus used to analyze the population trend across wavelengths for each parameter. LMEMs were fitted using the “lmer” function of the “lmer4” package (Bates et al., 2015). Random effects, defined as differences of the locations from the population trend (intercepts and slopes), were used to study individual photoacclimation processes. Wavelengths from 440–625 nm were transposed to 0–185 nm to decrease uncertainty in the model intercept. Hypothesis tests were based on t-tests (for the intercept and slope of fixed effects) and likelihood-ratio tests based on the χ² null hypothesis (for random effects) (Pinheiro & Bates, 2000).

Second, each wavelength-dependent photosynthetic parameter was detrended by calculating individual differences from the population trend (from LMEMs) and then used in partial triadic analysis (PTA) (Thioulouse & Chessel, 1987). PTA analyzes several two-way tables simultaneously (i.e. K-tables method). Five tables (one per wavelength) that contained eight photosynthetic parameters (in columns) and 18 locations (in rows) were analyzed (location no. 8 was not considered on PTA analysis due to missing values at 590 nm). Before analysis, all parameter values were centered and reduced based on their overall ranges from all tables. PTA identifies structures that are the same in all tables and assesses their stability among wavelengths. PTA was performed using the “pta” function of the “ade4” package (Dray & Dufour, 2007), and related graphics were created with the “adegraphics” package (Siberchicot et al., 2017). PTA was applied in three steps—interstructure, compromise and intrastructure analysis—which correspond to co-variance, mean and variance structure analysis, respectively (Lavit et al., 1994; Mendes et al., 2010).

Third, redundancy analysis (RDA) of the same five wavelength-detrended tables as for the PTA was performed to test for relationships between wavelength-dependent photosynthetic parameters and explanatory abiotic and biotic variables. Data were centered and reduced before analysis. Explanatory variables were selected for the model using an automatic stepwise model (the “ordiR2step” function of the “vegan” package (Oksanen et al., 2019)) that performs forward selection based solely on the adjusted R² and p-value (199 permutations). At each step, the variable with the highest additional fit was added to the model.

Abiotic and biotic variables

The experimental conditions determined by the abiotic PCA showed contrasting results. For abiotic variables, the first two axes of the PCA explained 66.1% of total inertia (49.5% and 16.6%, respectively) (Fig. 3A). The first axis distinguished samples based on their light-quality ratios (R/B, G/B and G/R), vertical light attenuation coefficient (K_d(PAR)), Z_eu, Z_umixl, salinity and PO₄³⁻ concentration (Fig. 3A). The second axis distinguished samples based on their DIN concentration, temperature, Si(OH)₄ concentration and light intensity at a depth of 2 m (PAR_2m) (Fig. 3A). Three groups of samples were distinguished. Group 1 (samples 12, 18, 22, 31, 33, 38 and 49) had intermediate-to-high DIN (>2 µmol L⁻¹, up to 30 µmol L⁻¹ near Seine Bay) and temperatures (9.5–12.0 °C), low-to-intermediate salinity (<34 PSU) and PAR_2m (20–182 µmol quanta m⁻² s⁻¹), the highest R/B and G/B ratios (mean of 0.9 and 1.9, respectively), the highest K_d(PAR) (0.2–0.5 m⁻¹), the shallowest Z_eu (9–23 m) and Z_umixl (6.5–21.0 m), and the lowest G/R ratio (±2). On the opposite side of the factorial map, group 2 (samples 28, 37, 43, 45, 46 and 53) had the highest PO₄³⁻ and Si(OH)₄ concentrations (>1 and 0.5–2.0 µmol L⁻¹, respectively) and salinity (mean of 35 PSU), the lowest R/B and G/B ratios, the highest G/R ratios, the lowest K_d(PAR) (0.07–0.17 m⁻¹), and the deepest Z_eu (26–63 m) and Z_umixl (17–78 m). Group 3 (samples 8, 10, 13, 25, 35 and 41) had intermediate salinity (33.–35.0 PSU), the lowest temperatures (<10 °C) and DIN concentration, intermediate light-quality ratios, and the highest PAR_2m, but with high variability (116–930 µmol quanta m⁻²·s⁻¹). Thus, PAR in the abiotic PCA did not distinguish sampling locations well, nor did TSS. Samples had TSS less than 2 h (samples 10, 18, 28, 33, 38, 46, 49 and 53), greater than 10 h (samples 22, 31, 37 and 45) or values between the two (samples 8, 10, 13, 25, 35, 41 and 43). Group 1 had locations near the coast, while group 2 had locations offshore. Detailed information on abiotic variables is shown in the Supplementary Material, Figs. S2 and S3.

The biotic PCA based on FluoroProbe measurements of the biomass of main phytoplankton groups (Fig. 3B) distinguished samples mainly based on the biomass of P. globosa; however, the range of variation was low (±25%). The first axis distinguished P. globosa from brown microalgae and cryptophytes, while the second axis distinguished Cyanobacteria. The results indicate that P. globosa co-dominated with diatoms at several locations, and P. globosa dominated only samples from locations 10, 12, 13, 18 and 22. Samples from three locations (41, 43 and 46) contained the most cyanobacteria. Details of phytoplankton groups by location are shown in the Supplementary Material, Fig. S4.

Wavelength-dependent photosynthetic parameters from Linear Mixed-Effects Models

Fixed effects of the LMEMs represented the population trend of each photosynthetic parameter once the spatial nature of the data sampling was considered (Fig. 4). All intercepts were significant, indicating that all parameters differed from zero in the bright blue wavelength (440 nm) (Table 2). Wavelength dependence led to a significant slope of the fixed effect for all parameters except ETR_max(II) and E_op(II), but the sign of each slope varied among parameters. While the slope of F_v/F_m was significant, its small decrease across wavelengths was considered null for simplicity (Fig. 4A).

Sigma(II)_λ is a key parameter since it connects relative and absolute parameters such as α, ETR_max, E_k and E_op. Population trend of Sigma(II)_λ decreased by a factor of 3 across wavelengths (from 6 to 2 nm², Fig. 4B). Conversely, r.ETR_max trend increased by a factor of three (from 50 to 150, Fig. 5C), which may have counteracted the decrease in Sigma(II)_λ trend and led to the null slope of ETR_max(II) trend (Fig. 4D). Trends of r.α and α(II) increased as wavelength increased (Figs. 4E and 4F), and sharply for the latter, which increased by a factor of two. Since E_k is the ratio of ETR_max to α, r.E_k increased and E_k(II) decreased as wavelength increased (Figs. 4G and 4H). Since E_op values were highly scattered, they were not considered in later analyses (Figs. 4I and 4J). The decreasing trend in NPQs was higher at 1,200 than at 300 µmol quanta m⁻² s⁻¹ (Figs. 4K and 4L). Since NPQ was estimated at low and high PAR from the PE curves, it is interesting to note that the NPQ trends were opposite to those in r.ETR_max and r.α (Figs. 4C and 4E respectively), meaning that NPQs could have more influence on the values of these parameters under the blue wavelengths (NPQ₄₄₀ and NPQ₄₈₀ ranged from 1–2) than under the light red wavelength (NPQ₆₂₅ reached 0.2 at PAR = 300 µmol quanta m⁻² s⁻¹). The increasing trend in α(II) across spectrum could thus correspond to a strong decrease in α(II) under the blue wavelengths and not to an optimization under the light red wavelength.

For the random effects, intercepts of all parameters were highly significant, but their slopes were not, except for ETR_max and E_k in relative (r) and absolute (II) values (Table 2). Thus, parameter values differed among samples but, except for ETR_max and E_k, had the same trend across wavelengths. This resulted in spatial differences in the spectral balance between bright blue and light red wavelengths for ETR_max and the photoacclimation parameters E_k (in relative and absolute units for the both).

When examining detrended values of absolute parameters among locations (Fig. 5), those of ETR_max(II) and E_k(II) tended to differ among the five wavelengths by sampling location (Figs. 5C and 5E), unlike those of the other parameters, which were generally more similar among the five wavelengths by sampling location. This was especially true for detrended values of α(II) and F_v/F_m, which differed little and almost not at all, respectively, among the five wavelengths by location (Figs. 5A and 5B). Because values of Sigma(II)_λ under the light red wavelength were not always the lowest across wavelengths (i.e. a slightly non-linear distribution) (Fig. 4B), its detrended values under the light red wavelength were higher than the population trend (Fig. 5A). According to the statistical analyses, however, a linear model fit best to Sigma(II)λ values.

Sample wavelength dependence of samples from the PTA

The first two axes of the PTA interstructure explained 85.38% of total inertia (Supplementary Material, Fig. S5), and the five wavelength tables had similar weights (0.38–0.47; Table 3) and a significant representation (cos² close to 1; Table 3). The PTA was thus adequate overall and highlighted similarities among the wavelengths. All five wavelengths were positively correlated and positively projected on the first axis (ca. 71.24% of the total inertia; Supplementary Material, Fig. S5). The second axis separated the bright blue and green wavelengths from the amber and light red wavelengths. The amber wavelength differed the most from the others and had the same correlation with the first and second axes.

When projecting the wavelength-dependent photosynthetic parameters on the compromise coordinates it explained 66.49% of total inertia (Table 4, Fig. 6A), the relative positions of polygons indicated that PTA results generally met our expectations: opposition between the group of F_v/F_m, α(II) and ETR_max(II) vs. the group of Sigma(II)_λ and NPQ_300–1,200, with parameters related to photoacclimation (E_k and E_op) between these two groups. We observed the well-known relationships between variables related to energy flows (F_v/F_m and NPQ_300–1,200), related to Sigma(II)_λ, and the major parameters that control PE relations in absolute units—α(II) and ETR_max(II)—which had a positive overall correlation.

The intrastructure of the PTA showed differences between photosynthetic parameter patterns among the five wavelengths (Figs. 6A to 6E). Patterns for the main parameters, such as ETR_max(II), α(II), Sigma(II)_λ and NPQ_300–1,200, changed from the bright blue wavelength (440 nm) to light red wavelength (625 nm). The most evident change was the rotation of α(II) and NPQ300–1,200 respected to the spectral pattern of Sigma(II)_λ. At blue wavelengths Sigma(II)_λ and NPQ_300–1,200 were inversely correlated to α(II) (Fig. 6A) but at amber and light red wavelengths there were any correlation (Figs. 6D and 6E). Sigma(II)_λ was incorrectly represented in the main plane at 625 nm (Fig. 6E). The PTA intrastructure analysis also showed patterns for the locations among the five wavelengths (Fig. 7). Wavelength dependence differed greatly among locations: polygons were largest for locations 33, 35, 43, 37 and 45, and smallest for locations 13, 18, 31, 46 and 53 (Fig. 7). In addition, the blue wavelengths (440 and 480 nm) displayed a general circular change among locations, moving from the right of the polygon for location 53 to the left for location 45 (Fig. 7).

Explanatory variables of wavelength dependence from RDA and linear regression

The RDA results showed that abiotic variables related to light emerged first as explanatory variables (Table 5). Euphotic depth (Z_eu) was selected the most often, for three of the five wavelengths (Table 5), from 440–540 nm. TSS, which represents the recent light history of cells, was selected twice, for the light blue wavelength (480 nm) and the amber wavelength (590 nm). The G/R light ratios were selected for the amber wavelength (590 nm). DIN concentration was the only non-light parameter selected, for the light blue wavelength. No variables were selected for the light red wavelength (625 nm), and PAR_2m did not seem to influence the parameters.

To further explore the influence of Z_eu and the difference in control of photosynthetic parameters under bright blue and light red wavelengths, we sought specific connections between the photoacclimation parameter E_k and Z_eu (in ratios with E_avg and Z_umixl, respectively), and between the ratio of E_k (in relative and absolute units) measured at 625 and 440 nm and the corresponding R/B light ratio (E_625/440) in water masses. Two significant linear trends were found between the E_k,440/E_avg ratio (in relative and absolute units) and the Z_eu/Z_umixl ratio for stratified water columns (Fig. 8A for graphs and correlation coefficients). Correlations were non-significant for non-stratified water columns for r.E_k,440 and E_k(II)₄₄₀, as well as under the other wavelengths. Considering all sampling locations revealed other significant correlations between the absolute ratio E_k(II)_625/440 and (1) the red/blue light ratios in surface waters E_625/440 (see Fig. 8B for correlation coefficients) and (2) the TSS factor (r = −0.65, n = 19, p < 0.05, Supplementary Material, Fig. S6).

Physiological meaning of wavelength dependence of photosynthetic parameters

Ecological meaning of sample spectral variability and controlling factors

To investigate the ecological meaning of our results for wavelength dependence, we addressed the variability in photosynthetic parameters outside of the population trends. This approach was based on the detrended measurements of the PTA and RDA by wavelength, and the relationships between the E_k(II) ratios and their controlling factors.

According to the PTA, the covariations observed among detrended values of photosynthetic parameters and their change over the spectrum are consistent with wavelength dependence at the population level of the parameters studied with the LMEMs. The PTA showed that the pattern of the parameters changed gradually and consistently between blue and red wavelengths, concretely for light absorption and photochemical quenching. Using the PTA to scan the intrastructure of the 19 samples revealed also that spectral variation patterns of the parameters of each sample differed from each other in size, shape and position on the factorial map. Sizes and shapes of these spectral polygons were not related to water column stratification (e.g. the polygon of the most stratified location (no. 33) was the same size as that of location 43, which was not stratified). According to the RDA, photosynthetic parameters measured by wavelength were influenced mainly by euphotic depth. The forward selection of the RDA first retained the most correlated variable in a group of abiotic variables that covaried with biotic variables. Consequently, since the euphotic depth (Z_eu), the light extinction coefficient (K_d(PAR)) and the three light-quality ratios were collinear or correlated variables (in the abiotic PCA), individual photosynthetic parameters could also be controlled by the underwater light quality and turbidity, since the RDA also selected K_d(PAR) and a wavelength ratio. Considering all photosynthetic parameters, the main RDA result is thus consistent with the significant relationship observed between the light saturation ratio (E_k(II)_625/440) and R/B light ratio (E_625/440). The RDA also selected the TSS variable discussed later. Mixing depth (Z_umixl) was another interesting ecological parameter selected by the RDA. The RDA confirms the hydrodynamic regime’s control of the photoacclimation index E_k previously displayed by the relationship between E_k (r and II) and Z_eu/Z_umixl. Since physical forcing in a given upper mixed layer controls certainly the level reached by E_k, wavelength dependency of photosynthetic parameters is generally a trade-off between the light quality of the different encountered water masses and changes in light quantity throughout the upper mixed layer due to the hydrodynamic regime. The RDA results and singular correlations with E_k (r and II) are consistent, but the correlations of E_k(II)_625/440 with critical environmental parameters provide better understanding of the ecological mechanisms that influence phytoplankton photoacclimation. The results are interesting because they were obtained from the field, where light-quality ratios (between red and blue wavelengths for instance) are known to change with depth (Supplementary Material, Fig. S7), as also described by Brunet et al. (2014) and Jaubert et al. (2017). It is likely that the influence of vertical changes in light-quality ratios on microalgae acclimation was small in the present study, especially because most upper mixed layers were shallow (6.5–14 m) and the residence times of the cells at a given depth were low. The spectral light saturation ratios E_k(II)_625/440 were therefore not correlated with the Z_eu/Z_umixl index.

The significant correlations between E_k(II)_625/440 and E_625/440 or TSS clearly highlight that natural phytoplankton communities can implement photoacclimation processes that are driven by the in situ light quality that also change during the daylight cycle. Most water masses had E_625/440 ratios less than one, which indicates water in which blue wavelengths dominate red wavelengths, and displayed higher E_k(II) phytoplankton photoacclimation index in blue wavelengths than in red wavelengths. This original result is valid only for absolute E_k(II)_λ parameters related to Sigma(II)_λ, not for relative parameters r.E_k. Since Sigma(II)_λ measured with the MULTI-COLOR-PAM is an intrinsic property of microalgae (Schreiber, Klughammer & Kolbowski, 2012), and since microalgae can increase pigment concentration to absorb more light (Lawrenz & Richardson, 2017), differences in pigment concentrations may change the level of Sigma(II)_λ measured during the day. Due to the differing variations in r.E_k between blue and red wavelengths, E_k(II)_625/440 ratios are therefore correlated with E_625/440 ratios in the water masses and with TSS, the time elapsed since sunrise. These relationships provide new information about the natural environment and are consistent with many experiments under controlled conditions. As several studies of monospecific cultures subjected to contrasting R/B ratios show (Schellenberger Costa et al., 2013a; Brunet et al., 2014), variations in the light spectrum and in blue vs. red wavelengths influence photoprotective capacity and the pigment composition of phytoplankton. Schellenberger Costa et al. (2013a) conclude that photoprotection is regulated more by light quality (especially blue wavelengths) than by the overall light intensity. Kirk (2011) stated that phytoplankton detect not so much the spectra of light, but rather differences between wavelength ratios received by PSI and PSII, using blue and red wavelength photoreceptors (Jaubert et al., 2017) that regulate photosynthesis and promote photoacclimation (Schellenberger Costa et al., 2013b; Petroutsos et al., 2016). This explains why the E_k(II) for the green wavelength did not correlate significantly with the in situ light-quality ratios of the green wavelengths. Photoacclimation mediation by in situ blue wavelengths, as discussed by Schellenberger Costa et al. (2013a), is thus consistent with our field study.

The E_k(II)_625/440 vs. E_625/440 correlation, like the abiotic PCA, indicates indirectly that variables such as temperature, PAR_2m, and DIN and Si(OH)₄ concentrations do not influence wavelength photoacclimation greatly. Previous studies in the EC showed that abiotic variables were the main variables that controlled spatial and/or temporal variations in relative photosynthetic parameters (Jouenne et al., 2007; Napoléon et al., 2013; Houliez et al., 2015). However, the variables that control these photosynthetic parameters may vary among geographic areas and/or seasons. In the present study, this correlation was determined over a large spatial scale.

Ecological implications and consequences

It is the general question of the absorption capacity of light in relation to the quality of light and its impact on primary production that is discussed here.

The precise examination of the E_k(II)_625/440 versus E_625/440 relationship in link with the reference values of 1 indicates that the two ratios matched each other well. For example, at location 38, the E_k(II)_625/440 and E_625/440 ratios equaled one, which could be because the water there was sampled before sunrise. However, other communities sampled before sunrise (e.g. at locations 28, 46 and 53) showed imbalances in their E_k(II)_625/440 ratios related to the E_625/440 ratios of water masses. The regression model indicates that to observe an E_k(II)_625/440 ratio of 1, a theoretical E_625/440 ratio of 1.29 would be required (which is close to our measurements). In this case, the blue wavelengths would decrease to 77 µmol quanta m⁻² s⁻¹ given a red light of 100 µmol quanta m⁻² s⁻¹. Thus, the E_625/440 ratio cited above indicates no strong imbalance in available energy and thus no stressful ecological situation for phytoplankton. In spring in temperate water, blue wavelengths are absorbed due to CDOM, in link with terrestrial discharge near estuaries and/or phytoplankton blooms themselves (Vantrepotte et al., 2007; Astoreca, Rousseau & Lancelot, 2009). Lawrenz & Richardson (2017) studied photoacclimation under extreme conditions, with a total absence of blue light (i.e. black water with high CDOM concentrations). They showed that, depending on the taxon, microalgae retain or lose their initial light absorption capacity on the spectrum, and their absorption capacity adapts to the light quality to which they were exposed, even with red wavelengths. Some species can survive under red light in the short term, but in the long term, cytoplasmic structures and chloroplast membranes degrade (Humphrey, 1983) and Chla concentrations decline (Forster & Dring, 1992). Rivkin (1989) showed the strong influence of blue light on carbon fixation and incorporation into amino acids and proteins. The E_k(II)_625/440 vs. E_625/440 relationship in the present study cannot be extrapolated beyond the measurement limits (i.e. to the completely unbalanced light ratios of Lawrenz & Richardson (2017)), but includes representative conditions generally found in the EC or temperate systems.

In comparison, E_k(II)_625/440 ratios near 0 would indicate that blue wavelengths are ultra-dominant in the water mass and could activate strong photoprotection mechanisms or even cause photodamage, which would decrease ETR_max(II) greatly under blue wavelengths. Under these conditions, given the intercept of the linear model, E_k(II) values under blue wavelengths would be only 25% of those under the red wavelength, which suggests that primary production would decrease greatly. However, electron flows under green, amber and light red wavelengths remained high, and NPQ was not as strong as under blue wavelengths. These results were consistent with the low light absorption (Sigma(II)_λ) under these wavelengths, as was the NPQ. However, this raises the issue of using the spectral approach to calculate primary production based on the photosynthetic parameters of RLC relationships.

Previous studies of the wavelength dependence of photosynthesis specified the systematic error produced by measuring α under white-light incubators when comparing incubation light climates (Laws et al., 1990) or primary production models (Kyewalyanga,Platt & Sathyendranath, 1992). Most classic incubators do not reproduce light spectra at the low intensities that phytoplankton encounter in the water column because of the high variability in light quality with depth, but also with time, due to vertical mixing in the upper part of the water column. Schofield, Prezelin & Johnsen (1996) examined the error caused by using the same or different α from cultures grown under different light qualities when calculating primary production in a theoretical and simplified water column. Depending on the species and growing conditions, differences between vertical primary production rates estimated by the two calculation methods ranged from 12–49%. Other studies showed that α values measured on board under artificial light (Irwin et al., 1990) could be corrected from the shape of the phytoplankton absorption spectrum (Kyewalyanga, Platt & Sathyendranath, 1997; Sathyendranath et al., 1999). This approach involves the field of remote sensing in particular and includes “optical” and “full spectral” models (Platt & Sathyendranath, 1988; Sathyendranath & Platt, 1993; Behrenfeld & Falkowski, 1997). Recent studies (Kovač et al., 2017; Sathyendranath et al., 2020) combined a spectral model of underwater light with a model of the integrated spectral response of algal photosynthesis consistent with photoacclimation processes. These studies recommend using the photosynthesis action spectrum or spectral correction of α in the water column, especially when differences in the shape of the action spectrum of α are larger than those in its magnitude at each wavelength (Sathyendranath & Platt, 1993). However, these studies did not consider that light-quality ratios in surface water can also greatly influence photoacclimation of microalgae. This was revealed in the present study by some individual wavelength-dependence phenomenon that differed significantly among the samples, and the E_k(II)_625/440 ratios, which involve ETR_max(II) and α(II) (i.e. the photosynthetic apparatus), especially the functional absorption cross section of PSII and the maximum rate of PET_λ, according to the optical definition of E_k (Falkowski & Raven, 2007). Parameters α(II) and Sigma(II)_λ showed no significant individual wavelength dependence among water masses, unlike ETR_max(II) and E_k(II), which are involved in primary production at high light intensities. The central issue is thus how photosynthetic activity induced by wavelengths beyond 480 nm can compensate for the decrease in photosynthesis under strong blue light when estimating primary production in different water masses. Sensitivity analysis of physicochemical properties of water would pave the way for future research on wavelength dependence of phytoplankton photosynthesis, as well as spectral dependence at the seasonal scale, using current active fluorescence measurement technologies.