Disparities between in situ and optically derived carbon biomass and growth rates of the prymnesiophyte Phaeocystis globosa

The oceans play a pivotal role in the global carbon cycle. It is not practical to measure the global daily production of organic carbon, the product of phytoplankton standing stock and its growth rate using discrete oceanographic methods. Instead, optical proxies from Earth-orbiting satellites must be used. To test the accuracy of optically derived proxies of phytoplankton physiology and growth rate, hyperspectral reflectance data from the wax and wane of a Phaeocystis bloom in laboratory mesocosms were compared with standard ex situ data. Chlorophyll biomass could be estimated accurately from reflectance using specific chlorophyll absorption algorithms. However, the conversion of chlorophyll (Chl) to carbon (C) was obscured by the non-linear increase in C : Chl under nutrient-limited growth. Although C : Chl was inversely correlated (r = 0.88) with the in situ fluorometric growth rate indicator Fv /Fm (Photosystem II quantum efficiency), none of them was linearly correlated to growth rate, constraining the accurate calculation of Phaeocystis growth or production rates. Unfortunately, the optical proxy φph (quantum efficiency of fluorescence: the ratio of the number of fluoresced photons to the number of photons absorbed by the phytoplankton) did not show any correlation with Phaeocystis growth rate, and therefore it is concluded that φph cannot be applied in the remotely sensed measurement of this species’ carbon production rate.


Introduction
Approximately half of the global photosynthetic CO 2 to organic carbon conversion takes place in marine waters (Field et al., 1998).Unfortunately, global daily CO 2 fixation, the product of phytoplankton standing stock and growth rates cannot be measured directly for the world oceans.Phytoplankton biomass and growth rates can be assessed directly and accurately by standard oceanographic techniques, but these miss the spatial coverage of the optical instruments onboard Earth-orbiting satellites.On the other hand, optically derived estimates of phytoplankton biomass and growth rates are less accurate than shipboard data (Abbott and Letelier, 1999;Carder et al., 2003;Behrenfeld et al., 2005;Huot et al., 2005;Astoreca et al., 2009;Martinez-Vicente et al., 2013).
Here we report, to our knowledge for the first time ever, on the simultaneous evaluation of standard oceanographic and state-of-the-art optical techniques for gauging both phytoplankton biomass and carbon growth rates.
In "standard" oceanographic measurements, carbon concentration, carbon fixation, chlorophyll and other photopigment concentrations are analyzed in discrete water samples (ex situ), as is the quantum efficiency of Photosystem II (Fv / Fm) that can be considered an indicator for phytoplankton growth rate (Kromkamp and Foster, 2003;Ly et al., 2014).
Optical estimates of the oceanic carbon concentration for growth rate estimations can be made from the particulate backscatter coefficient (b bp ;Behrenfeld et al., 2005), but this coefficient is non-specific for phytoplankton or valid only for low chlorophyll a concentrations (Martinez-Vicente et al., 2013).Alternatively, the phytoplankton-specific chlorophyll concentration can be estimated from water-leaving radiance as absorbance (Carder et al., 2003).However, the carbon to chlorophyll ratio (C : Chl) that is then needed to convert chlorophyll into carbon is not a constant (Sathyendranath et al., 2009).

L. Peperzak et al.: Disparities between in situ and optically derived carbon biomass
A second optical growth rate proxy is the phytoplanktonspecific red chlorophyll fluorescence relative to absorbance (ϕ ph ).By definition this "quantum efficiency of fluorescence" is the ratio of the number of fluoresced photons to the number of photons absorbed by the phytoplankton, i.e. by all cellular photopigments (Abbott and Letelier, 1999;Huot et al., 2005).According to Falkowski and Kolber (Falkowski and Kolber, 1995), the quantum efficiency of photosynthesis varies inversely to the quantum efficiency of fluorescence.If the production of chlorophyll stops under nutrient limitation, it is expected that C : Chl, fluorescence and ϕ ph will increase (Kiefer, 1973;Falkowski et al., 1992;Behrenfeld et al., 2009).
Besides the lack of specificity, an inherent problem in the optical approach of organic carbon production is that estimates of carbon and chlorophyll are used in both biomass and growth rate proxies.Moreover, doubt has been raised whether the variability in remotely sensed phytoplankton physiology (ϕ ph ) is due to physiological changes in the phytoplankton or due to environmentally driven biases in algorithms needed to estimate ϕ ph (Huot et al., 2005).
In order to study the variability in phytoplankton biomass, growth rate, absorbance and fluorescence under variable but fully controlled conditions, a mesocosm experiment was conducted where detailed "standard" oceanographic measurements were combined with close-sensing hyperspectral measurements.Phytoplankton dynamics in the mesocosms were experimentally manipulated under semi-natural conditions of temperature, irradiance and turbulence (Peperzak et al., 2011).The prymnesiophyte Phaeocystis globosa, a key species in marine primary production, was used as test organism (Wassmann et al., 1990;Smith et al., 1991;DiTullio et al., 2000;Vogt et al., 2012).Our ambition was to use the optical signature of Phaeocystis globosa, which can now be detected by the MERIS and MODIS satellites (Kurekin et al., 2014), to better understand the wax and wane of its blooms.This optical signature includes light absorption, light emission and the quantum efficiency of the phytoplankton.In particular we would like to know how optical proxies compare to standard oceanographic techniques for estimating primary production, because this is still one of the key questions in ocean color research (Cullen and Lewis, 1995;Saba et al., 2010;Huot et al., 2013Huot et al., , 2005;;Behrenfeld et al., 2009).

Experimental
The flagellate, non-colony-forming strain Pg6-I of Phaeocystis globosa ("Phaeocystis") was inoculated in two duplicate 140 L mesocosms filled with 0.2 µm of filtered Atlantic Ocean water poor in organic and inorganic nutrients that had been diluted with Milli-Q ™ water to a salinity of 34 g kg −1 .A detailed description of the mesocosms is given in Peperzak et al. (2011).Temperature during Phaeocystis growth was kept at 15 • C. Irradiance was provided in a semi-sinusoidal light-dark (16 : 8 h) cycle with a maximum surface PAR of 41 W m −2 in mesocosm 1 and 45 W m −2 in mesocosm 2. Turbulence of the water was provided by pumping surface water to the bottom of the mesocosm at a turnover rate of 1 h.The water was enriched with macronutrients to 30 µM NO − 3 , 6.3 µM PO 3− 4 , and trace metals and vitamin B1 (Peperzak et al., 2011).On day 8 of the experiment, when cells were in the stationary growth phase, mesocosm 1 received enrichment with the initial nutrient concentrations to examine the effect of alleviation of nitrogen limitation on the physiological and optical properties of Phaeocystis.

Sampling
Water samples were taken in the middle of the light period (13:00, all times listed in CET) to measure salinity; pH; cell abundance; dissolved inorganic nitrogen (DIN); soluble reactive phosphorus (SRP); HPLC (high-performance liquid chromatography) pigments including chlorophyll a (Chl a), chlorophyll c2 and c3 (summed as Chl c) and carotenoids; particulate organic carbon (POC) and nitrogen (PON), and PAM-derived (Walz, Water PAM ™ ) Photosystem II quantum efficiency (Fv / Fm) on dark-adapted (> 20 min) samples.A detailed description of the analyses is provided elsewhere (Peperzak et al., 2011).See Table 1 for a list of measured and derived variables.
Surface irradiance (W m −2 nm −1 ), used to convert radiance (W m −2 nm −1 sr −1 ) to reflectance (R, sr −1 ), was measured prior to and after the experiment.In addition, phytoplankton absorption was measured daily at 13:00 using a 0.55 L integrating cavity absorption meter (ICAM; asphere ™ , HOBI Labs, Tucson, AZ, USA).ICAM absorption data (a ph , m −1 ) were blank-corrected daily by subtracting the absorption of filtered seawater, then divided by chlorophyll a or c concentrations to obtain the chlorophyll-specific absorption coefficients (a * Chl , m 2 (mg chlorophyll) −1 ) in both the exponential and the stationary Phaeocystis growth phase.Phaeocystis spectra of a * Chl , together with reflectance data, were used to determine the appropriate wavelengths in algorithms for the estimation of chlorophyll a (c) absorption from reflectance spectra.Details of the ICAM absorption, irradiance and radiance measurements are provided elsewhere (Peperzak et al., 2011).

Mesocosms, absorption and fluorescence algorithms
The mesocosm description and analysis of the spectra is based on the methodological paper of Peperzak et al. (2011), which contains an extensive description of the experiment, measurements, and validation of the analysis of the absorption and fluorescence signals.The mesocosm tank, height 0.75 m, diameter 0.5 m and water volume 0.14 m 3 was made of black high-density polyethylene and mounted in a black metal frame made of 30 mm square aluminum painted black (Fig. 1).To avoid light reflection from the walls, the interior of the tank was sand-blasted.The contents was mixed by pumping water at a turnover rate of 1 h from 0.05 m below the water surface (−0.05 m) to 0.10 m above the bottom (−0.65 m).A total of 25 Solux ™ MR16 halogen 4700 K "daylight" lamps of 50 W with a 24 • beam spread were used in a 5 × 5 matrix in a black-painted box that was mounted in a frame at 0.70 m above the water surface.A variable lightdark cycle with a semi-sinusoidal illumination was made possible by timers controlling all lamps.
Prior to and after the experiments of two weeks, surface irradiance (E 0 ) was measured every 15 min for at least 24 h from 320 to 950 nm in 190 channels (W m −2 nm −1 ) with a TriOS ™ RAMSES-ACC-VIS hyperspectral cosine ir-radiance sensor (TriOS, GmbH, Oldenburg, Germany) that was placed in the center of the mesocosm at the position of the water surface.During experiments the irradiance at the bottom (E b ) of the mesocosm (Fig. 1) was measured every 15 min with a similar TriOS ™ hyperspectral cosine irradiance sensor.Water-leaving radiance (L w ) was measured every 15 min with a TriOS ™ RAMSES-ACC-VIS hyperspectral radiance sensor (radiometer, 320-950 nm in 190 channels, W m −2 nm −1 sr −1 ) at an angle of 50 • nadir at 0.08 m above the water surface (Fig. 1).An integrating cavity absorption meter or ICAM (a-sphere spectrophotometer, HOBI Labs ™ , Tucson, AZ, USA) was used as an independent method to measure sample absorption (m −1 ).This type of instrument is very accurate, also at low concentrations, without interference from particle scattering.Based on the averaged spectra from the middle of the light period (13:00-14:00), four optical properties were derived: (1) the total number of photons absorbed by phytoplankton, (2) the total number of photons emitted by fluorescence, and (3) the chlorophyll c and (4) chlorophyll a concentration.The fifth quantity, the phytoplankton quantum efficiency (ϕ ph ), is defined as the ratio of moles of photons emitted as fluorescence divided by the moles of photons absorbed by the pigments and is therefore the ratio of property 2 over 1.
From a comparison of the irradiance sensor at the bottom of each mesocosm with the known irradiance at the water surface, the wavelength-dependent attenuation in the mesocosm was derived.This attenuation was corrected for the (small) effects of pure water and scattering effects at the mesocosm wall, and the total number of absorbed photons was calculated as the absorption times the illumination at each wavelength and integrated over the interval 400-672 nm.The stricter upper limit of 672 nm to the potential fluorescence radiation (PFR) is based on a central fluorescence emission at 682 nm and a Stokes shift of 10 nm that determined the minimum extra energy needed for the excitation of a chlorophyll molecule.The typical available PFR just below the wa-ter surface is 138 µmol photons m −2 s −1 for mesocosm 1, and slightly higher for mesocosm 2 (151 µmol photons m −2 s −1 ).
The classic fluorescence line height (FLH) algorithm (Abbott and Letelier, 1999) was applied on the remote sensing reflectance spectra (R), calculated as the ratio of the radiance spectra collected above water (S r , Fig. 1) divided by the illumination irradiance.
R max is at the fluorescence peak (λ = 682 nm) in the mesocosm reflectance spectra and R base is the baseline reflectance value at R max , calculated linearly from the reflectance between R b1 and R b2 , with b 1 = 650 nm and b 2 =710 nm.
In order to derive the number of emitted photons in the mesocosm, the FLH was first multiplied by the irradiance spectrum to obtain a baseline corrected radiance above water at 682 nm (W m −2 nm −1 sr −1 ).Then the signal was converted to photons and integrated over 4π sr (assuming isotropic emission) and integrated over the spectral range (650-710 nm), assuming a Gaussian distribution with a full width at half maximum (FWHM) of 25 nm.Subsequently, the emission was corrected for the water-air transition and the internal absorption in the mesocosm before it reaches the radiance sensor by water (0.43 m −1 at 682 nm) and selfabsorption by the phytoplankton (Huot et al., 2005).We refer the reader to the publication of Peperzak et al. (2011) for a more extensive description and validation of this conversion.The chlorophyll c concentration was calculated from reflectance (R) using a four-wavelength (at λ = 450, 466, 480 and 700 nm) absorption algorithm (ARP-4λ Chl c ) that was developed and positively applied by Astoreca et al. (2009) to detect Phaeocystis in the North Sea: with the absorption by pure water a w, 700 = 0.572 m −1 (15 • C) (Buiteveld et al., 1994).(2b) The weight (w) is determined by the position of the Chl c absorption maximum (466 nm) relative to the two reference (baseline) wavelengths (450 and 480 nm): A comparable absorption algorithm (ARP-4λ Chl a ) for chlorophyll a was derived after choosing the appropriate wavelengths, including the Chl a absorption maximum (438 nm): with water absorption given by Eq. ( 1b) and the weight (w) by

Statistics
To test the null hypothesis that there is no difference between means of variables measured in the two mesocosms, two-sample t tests were performed in SYSTAT ™ version 12. Linear regression equations were calculated in SYSTAT ™ or Excel ™ 2003.The 95 % confidence intervals (±95 % CI) around a variable mean (m) were calculated from a t distribution using n observations (days), n-1 degrees of freedom (df) and the standard deviation of the mean SD as m ± 95 % CI = m ± t (0.05; n-1) × SD/ √ n.The standard error (i.e., SD/ √ n) provided in linear regression by SYSTAT ™ was used to calculate 95 % confidence intervals of regression slopes.

Phytoplankton dynamics (ex situ observations)
Inoculation of the mesocosms was followed by a 3-day exponential increase in Phaeocystis cell abundance and Chl a, Chl c, POC and PON concentrations (Fig. 2a, c-f).Compared to mesocosm 1, the higher surface irradiance in mesocosm 2 led to 17 % more cells on day 5, when the stationary growth phase was reached in both mesocosms due to nitrogen limitation (Fig. 2b).In both mesocosms, cell abundances in the stationary growth phase decreased at an average rate of −0.07 d −1 .The 30 µM nitrate in the nutrient spike added to mesocosm 1 on day 8 was already depleted by Phaeocystis on day 9 (Fig. 2b) and incorporated as PON (Fig. 2f).In addition, Phaeocystis cells and Chl a and Chl c concentrations increased after the nutrient spike (Fig. 2a, c-d).In a separate experiment (no data shown), in which a mesocosm 2 water sample on day 10 was spiked with only nitrate, the resumption of cell growth and an increase in Fv / Fm confirmed that nitrogen was the limiting element.

ICAM absorption
The ICAM absorption spectra of mesocosm water samples contained three major peaks: at 438 (Chl a), 466 (Chl c) and 674 nm (Chl a).In the exponential growth phase, a * Chl was lower than in the stationary growth phase, due to the increase in carotenoids after nitrogen was depleted (Fig. 3d).These differences in a * Chl between the exponential and stationary growth phase were significant at 438 and 466 nm, but not at 674 nm (Table 2).

Reflectance absorption
The specific chlorophyll a and c absorption (a Chl a and a Chl c ) computed from reflectance spectra (Fig. 4a-b) closely resembled the development of Phaeocystis cell abundance and Chl a and c concentrations (Fig. 2a, c-d).In both mesocosms, total Chl absorption, a Chl a (c) correlated well with HPLC-measured Chl a and Chl c concentrations (Fig. 4cd) and the regression slopes of the two variables in the mesocosms were not significantly different (Table 3).When the data of both mesocosms were split by growth phase, the exponential-phase (day 1 to 4) regression equations accurately (both r 2 = 0.98) estimated both Chl a and Chl c (Fig. 4e-f).The stationary-phase (day 5 to 14) regression intercepts between a Chl a (c) and Chl a and Chl c concentrations were lower than in the exponential growth phase (Fig. 4ef), although not significantly (Table 3).This means that application of the regression equations combining both growth phases (Table 3) will lead to small underestimations of Chl a and Chl c concentrations in the exponential growth phase, and small overestimations of Chl a and Chl c concentrations in the stationary phase (Fig. 4e-f).

Fluorescence
Fluorescence emission estimated from the water-leaving radiance (Fig. 5a) resembled Phaeocystis cell dynamics (Fig. 2a) and was well correlated with Chl a (Fig. 5b; overall r 2 = 0.81, Table 4).When the data of both mesocosms were split by growth phase, the stationary-phase (day 5 to 14) regression slope and intercept were significantly different from  those in the exponential phase (day 1 to 4; Fig. 5c, Table 4).This means that, according to expectation, nutrient-stressed cells in the stationary growth phase have higher fluorescence intensity per unit chlorophyll.

Fluorescence quantum efficiency (optical observations)
The fluorescence efficiency (ϕ ph ), calculated as moles of photons emitted as fluorescence divided by the moles of photons absorbed by the phytoplankton pigments, increased during exponential growth, stabilized from day 5 to 8 and then decreased (Fig. 6).No apparent change in ϕ ph was observed in response to the nutrient spike on day 8 to mesocosm 1.

Carbon growth rate and proxy comparison
In order to relate dynamics in light absorption and fluorescence to Phaeocystis physiology in the different growth phases, the dynamics of carbon growth rate (µ POC ) was compared to Fv / Fm, C : Chl and ϕ ph (Fig. 7a-c).Because the cellular Chl c content of Phaeocystis is about the same as the cellular Chl a content (Fig. 4c, d) and total chlorophyll (Chl) was linearly correlated to Chl a (Chl = 2.28 × Chl a, r 2 = 0.99), C : Chl was used rather than C : Chl a and C : Chl c separately.
The proxy comparison showed hyperbolic relations of µ POC with C : Chl and Fv / Fm with highly variable values at µ POC ∼ 0.0 d −1 (Fig. 7a-b).As could be expected from Fig. 7a and b, Fv / Fm was inversely linearly correlated to C : Chl (r 2 = 0.88).The good correlation implies that, under the present experimental conditions, Fv / Fm and C : Chl, as measured either in water samples or derived from water-leaving radiance, are directly comparable physiological proxies.
Fluorescence quantum efficiency did not show any correlation with growth rate (Fig. 7c).It appears that ϕ ph is a poor proxy for Phaeocystis carbon production in both mesocosms.

Discussion
The aim of the mesocosm experiments was to investigate a relation between optical remote sensing and "standard" oceanographic measurements of phytoplankton physiology during different growth phases (here: nitrogen-controlled growth) of Phaeocystis and to infer possible implications for estimates of primary productivity.the effect of a nutrient spike to one mesocosm, proved that growth of Phaeocystis was indeed nitrogen-limited during the experiments.By measuring the in situ fluorescence (F ) increase due to nitrogen limitation, and the increase in photons absorbed by phytoplankton (PFR), an optical estimate of the quantum efficiency of fluorescence ϕ ph (i.e., F / PFR) could be made.It is shown that of the physiological diagnostics neither ϕ ph , nor Photosystem II quantum efficiency (Fv / Fm) nor C : Chl are reliable estimators of Phaeocystis growth rates.This may have consequences for global carbon fixation estimates using remote sensing data assessing phytoplankton physiology.

Phytoplankton dynamics
Temperature, salinity, irradiance and pH were at or near values for optimum Phaeocystis growth (Peperzak, 2002).The exponential-phase growth rate (µ = 0.7 d −1 ) and stationaryphase mortality rate (d = −0.07d −1 ) were equal to the rates obtained in cultures of P. globosa strain Ph91 (Peperzak et al., 2000a, b).The carbon and photopigment contents of Phaeocystis in the mesocosms were comparable to published values, although cellular Chl a and Chl c content was relatively low (Table 5).On the other hand, the fucoxanthin to Chl a ratio was high, which is probably caused by (1) an adaptation to the low-irradiance environment where this flagellate can thrive (Peperzak, 1993;Seoane et al., 2009) and/or (2) the effect of nitrogen-limited growth on the carotenoids : Chl ratio (Fig. 3d).In mesocosm 2, Phaeocystis in the stationary phase reached a C : N of 20, which is equal to the subsistence quota of 0.05 mol N mol C −1 in diatoms (Edwards et al., 2003).The rapid depletion of nitrate during the initial days of the experiment and the constant increase in C : N -combined with the decrease in C : N, resumption of cell growth and increase in Fv / Fm after the nutrient spikeconvincingly showed that Phaeocystis was nitrogen-limited in the stationary phase.
The physiological indicator Fv / Fm declined when nitrogen had been depleted on day 4.In addition, C : Chl increased.Both indicators responded directly following the nutrient spike to the nitrogen-depleted Phaeocystis on day 8. C : Chl was inversely linearly with Fv / Fm, but carbon growth rate was not.This can be explained by the fact that both Fv / Fm and C : Chl declined continuously after nitrogen depletion while cell division immediately halted on day 5.As a consequence, Fv / Fm and C : Chl not only signal physiological change but are also indicative of the persistence of nitrogen depletion in Phaeocystis.A comparable conclusion was reached for the decline of Fv / Fm and the duration of nitrogen depletion in the diatom Thalassiosira pseudonana (Parkhill et al., 2001).On the other hand, under balanced growth conditions, i.e. steady-state nitrogen-limited growth, the value of Fv / Fm in T. pseudonana was high and comparable to the value in nutrient-replete cultures (Parkhill et al., 2001).In other words, the steady 10-day change after an abrupt nitrogen depletion shows that Fv / Fm and C : Chl are not good indicators of short-term nutrient-limited phytoplankton growth rates.In the early stationary phase (day 4-8), the 10 % lower surface irradiance in mesocosm 1 led to a slightly lower (94 ± 21) but not significantly different C : Chl than in mesocosm 2 (106 ± 28).Comparable minor effects on cellular chlorophyll contents have been measured in Phaeocystis cultured at 10 and 100 µmol photons m −2 s −1 (Astoreca et al., 2009).The reduction of water column irradiance due to selfshading by increased chlorophyll concentrations during exponential growth would therefore have little effect on C : Chl.Far more important than the (relatively weak) effect of irradiance on C : Chl was the factor of 10 variability in C : Chl when Phaeocystis went from the exponential (C : Chl = 30) to the late stationary growth phase (C : Chl = 200, Fig. 3c and Table 5).This variability confirms that chlorophyll concentration is not a reliable indicator of phytoplankton biomass (Behrenfeld et al., 2009;Kruskopf and Flynn, 2006), which has implications for the correct conversion of chlorophyll to carbon in chlorophyll-based primary production models (Cloern et al., 1995;Sathyendranath et al., 2009).

Pigments and absorption
Nitrogen depletion led to increases in carotenoid concentrations relative to chlorophyll.Comparable increases in light absorption under nitrogen limitation, due to increased carotenoid : Chl a ratios, have been observed in other phytoplankton species (Heath et al., 1990;Staehr et al., 2002).The increase in carotenoids : Chl ratio had a direct effect on the estimation of light absorption from the reflection spectra and ICAM measurements.The excellent correlations (Table 3) between a Chl a and a Chl c as well as Chl a and Chl c concentrations in the exponential phase (both r 2 = 0.98) were lower in the stationary phase (0.59 < r 2 < 0.82).Besides more variability in the stationary phase, a Chl was lower than in the exponential phase due to interference by carotenoids in the reflection spectrum.This interference was more pronounced for a Chl c than for the a Chl a (Table 3), because the a Chl c algorithm employs wavelengths from 450 to 480 nm, where carotenoid absorption is more pronounced (Fujiki and Taguchi, 2002;Lubac et al., 2008).
The interference of carotenoids in the stationary phase will increase if total pigment absorption (a ph ) is measured instead of specific chlorophyll absorption.It is not surprising, therefore, that by using the ICAM data (400 to 672 nm) the correlation of absorption with Chl was lower (r 2 = 0.74) than when using the Chl a-and Chl c-specific algorithms.Carotenoid interference in the stationary phase also explains the limited apparent linearity of chlorophyll detection by www.biogeosciences.net/12/1659/2015/Biogeosciences, 12, 1659-1670, 2015 ICAM absorption to a maximum of approximately 50 µg L −1 (Peperzak et al., 2011).At a high nitrogen-limited Phaeocystis biomass, the use of total absorption, including the carotenoids, leads to an overestimation of the chlorophyll concentration.

Fluorescence quantum efficiency
The optically measured fluorescence signal correlated well with the ex situ measured Chl a concentrations and, as expected, showed a relative fluorescence increase in the stationary phase.ϕ ph in mesocosm 2 ranged by more than 100 %, from ≈ 0.8 to ≈ 1.7 % (Fig. 6).Satellite estimates of ϕ ph have a corresponding range, 0-3 % (Huot et al., 2005;Behrenfeld et al., 2009).However, there was no correlation with µ (cell growth rate) or µ POC (carbon growth rate, Fig. 7c), due to the effect of a changing carotenoids : Chl ratio as a result of nitrogen limitation.This suggests that, in order to relate growth conditions and fluorescence signal strength, new optical proxies should be developed for the photon absorption and emission by individual pigments (Fawley, 1989).
Even though ϕ ph can be estimated using appropriate fluorescence and absorbance algorithms, its value is not a reliable indicator of actual nitrogen-controlled Phaeocystis growth rate.ϕ ph is also a diagnostic for the duration of nitrogen depletion in Phaeocystis, which adds to the discussion on the physiological significance of Fv / Fm and C : Chl.For instance, under steady-state nitrogen-limited growth, the value of Fv / Fm in T. pseudonana is as high as the value in nutrient-replete cultures (Parkhill et al., 2001).As the present investigation was deemed to be exemplary of the phytoplankton dynamics during the wax and wane of a short-term bloom, i.e. a fast reduction from a high concentration of the limiting nutrient towards depletion, a real-world estimate of ϕ ph might behave similarly to ϕ ph in the mesocosms.
On the other hand, in oceanic waters the supply of the limiting nutrient may be low but relatively more constant, such as by aeolian deposition of iron or by continuous heterotrophic remineralization of organic material in the water column.For iron-limited phytoplankton growth, ϕ ph derived from satellite data was elevated (Behrenfeld et al., 2009), so in 82 % of the oceanic regions with a low iron deposition rate, ϕ ph appears to be a reliable remote sensing physiology proxy.This applicability of ϕ ph corresponds with that of Fv / Fm as a good physiological proxy in iron-limitation studies (Timmermans et al., 2001(Timmermans et al., , 2008)).Iron limitation likely has a more pronounced effect on ϕ ph than limitation of the major nutrients (N, P).
The present Phaeocystis study is an example of how experiments can contribute to validation of assumptions on optical data that are being made in the estimation of global carbon production.More experimental data are needed from phytoplankton species that differ in their pigment composition and in the effects of short-and long-term nutrient (N, P, Fe) limitation so that new optical proxies for phytoplankton physiology can be examined.Until these issues have been resolved we should be aware of the obscured view of phytoplankton physiology and hence marine primary production estimates using remote sensing.

Figure 1 .
Figure 1.Schematic representation of the mesocosm system.A: side view; B: view from below.The mesocosm vessel (M, height 0.75 m, diameter 0.50 m) is placed inside a metal frame (F) that also holds the illumination box (I).The illumination box contains 25 Solux ™ lamps (L) and is height-adjustable.Two hyperspectral sensors were installed: one for bottom irradiance (S i ) and one for water-leaving radiance (S r ).Water was pumped round through an outlet at −0.10 m (O) and an inlet (I) at +0.10 m from the bottom.An overflow (OF) at −0.05 m was used to keep the water level constant.The mesocosm is emptied with a drain (D) in the bottom.For clarity, the construction holding the radiance sensor (S r ) and the electrical wiring, tubing, valves and pump are not shown.
Figure 3. (a-d) Phaeocystis physiology and pigment ratios in two mesocosms in time.(a) Photosystem II efficiency (Fv / Fm), (b) carbon to nitrogen ratio (C : N, mol mol −1 ), (c) carbon to chlorophyll a + c ratio (C : Chl a + c, g g −1 ), and (d) carotenoids to chlorophyll a + c ratio (g g −1 ).The arrow indicates the nutrient addition to mesocosm 1 after sampling on day 8.
Figure 4. (a-f) Absorption characteristics in Phaeocystis.(a-b) Temporal development (days) of absorption by chlorophyll a (a, a Chl a , m −1 ) and c (b, a Chl c , m −1 ) calculated from reflectance spectra in both mesocosms.(c-f) Linear regression of absorption on chlorophyll a and c (m −1 ) against Chl a and Chl c concentrations (µg L −1 ) was performed separately for both mesocosms (c, d) and for exponential and stationary growth phases (e, f).

Figure 6 .
Figure 6.Development in time (days) of quantum efficiency ϕ ph derived from total phytoplankton absorption and fluorescence in two mesocosms.

Table 1 .
List of used variables, measurements and computations.

Table 3 .
Linear regression equations of Phaeocystis absorption on HPLC-measured chlorophyll a and c concentrations.Absorption was calculated with the ARP-4λ-Chl a and ARP-4λ-Chl c algorithms (Eq.1).Regressions were made for the mesocosms separately, for exponential (day 0-4) and stationary (day 5-14) growth phases.Indicated are slope and intercepts ± 95 % confidence interval.

Table 4 .
Linear regression equations of Phaeocystis fluorescence on HPLC-measured chlorophyll a concentrations.Fluorescence was calculated with the FLH-H algorithm (Eq.2).Regressions were made for the mesocosms separately, for exponential (day 0-4) and stationary (day 5-14) growth phases.Indicated are slopes and intercepts ±95 % confidence intervals.

Table 5 .
Biochemical characteristics of Phaeocystis in the mesocosm compared to published data from cultures, unless otherwise indicated.Chl is the sum of chlorophyll a and c; Fuco: fucoxanthin.For larger non-flagellated Phaeocystis cells.b Range of three species cultured at different irradiances.c C : Chl a for prymnesiophytes in field samples determined by regression analysis.d High value at low irradiance.e In Marsdiep during Phaeocystis blooms (Wadden Sea tidal inlet). a