Seasonal effects of photophysiology and chlorophyll a abundance on phytoplankton group-specific primary production in the Kuroshio region as revealed by SeaStar / SeaWiFS

To evaluate the group-specific primary production of diatoms, haptophytes, and cyanobacteria in the Kuroshio region, a novel satellite observation methodology using the SeaStar/SeaWiFS satellite 10 instrument was developed. The method used bio-optical relationships between the group-specific production and bio-optical properties such as the photosynthetic quantum yield and chlorophyll a specific optical absorption coefficient of phytoplankton, the last two of which were also estimated together with the group-specific production rather than assumed a priori. A global property of the absorption coefficient of phytoplankton, that the coefficient value at the wavelength of 510 nm was 15 close to their spectral average, was highlighted in the method for a use of multi-spectral ocean color satellite data. Our results showed that the derived quantum yield index was higher for diatoms than haptophytes and cyanobacteria. Furthermore, intraspecific variation in the index, emerged as a latitudinal gradient: the values for cyanobacteria increased towards the higher latitudes. The groupspecific primary production in the Kuroshio region showed that the climatological average of 134, 72 20 and 40 mg C m –2 day –1 for diatoms, haptophytes, and cyanobacteria, respectively. A comparison among variability of the group-specific primary production, the quantum yield index, and the absorption coefficient suggested that, in the Kuroshio region, the primary production due to diatoms was driven by their abundance through the year, whereas that due to cyanobacteria by photophysiology. The production due to haptophytes was seasonally regulated by both abundance and photophysiology. 25 Biogeosciences Discuss., doi:10.5194/bg-2017-164, 2017 Manuscript under review for journal Biogeosciences Discussion started: 8 May 2017 c © Author(s) 2017. CC-BY 3.0 License.


Introduction
The Kuroshio Current is the western boundary current carrying warm oligotrophic waters from the subtropic to the sub-arctic in the southwest region of the North Pacific (Fig. 1).Despite the oligotrophy of the current, higher fishery production has been recognized in and around the current, which is a paradox of the Kuroshio ecosystems.Understanding the physical-biogeochemical-biological 30 mechanisms of the Kuroshio ecosystems is crucial to not only explain this paradox as a science, but also provide a scientific basis for so-called ecosystem-based fishery management (Pikitch et al., 2004).
Recent advances in fishery sciences have revealed that the Kuroshio Current provides habitats for the larvae of pelagic fish as well as spawning grounds upstream (e.g., Sassa et al., 2004).In addition, the oligotrophic Kuroshio Current meets a cold and nutrient-rich subarctic current, the 35 Oyashio Current, forming a heterogeneous sub-polar front and eddies where primary production is fueled by vertical nutrient fluxes associated with physical instabilities at the front via upward transport of nutrients (Nagai et al., 2012;Clayton et al., 2014).In such a turbulent environment, diatoms with higher nutrient and/or low grazing mortality rates can become predominant in the phytoplankton assemblages.Because diatoms have a dense silica wall and sink well in calm water, it is clear that 40 turbulent upward motion is needed to keep them in the water column and is a prerequisite for the formation of dense populations of diatoms (Mann and Lazier, 2006).It is well known that many great fisheries are dependent on a grazing food chain that starts at diatoms and proceeds via copepods to young fish (Cushing, 1989).Therefore, understanding the community structure of primary producers provides clues of the underlying ecological structure.45 In a case study in the North Atlantic, Claustre et al. (2005) derived primary production parameters by the multiple regression analysis of in situ data and demonstrated that they were dependent on the phytoplankton community structure, and Uitz et al., (2008) further parameterized them.However, the Kuroshio region was not included in their analysis.Remote estimations of sizefractionated primary production have also been conducted (Kameda and Ishizaka, 2005;Uitz et al., 50 2010) covering larger geographical extents to support intermittent efforts of in situ observations of size-fractionated or taxon-specific primary production (Owens et al., 1993;Jochem et al., 1995;Lee et al., 1996;Bury et al., 2001;Claustre et al., 2005;Moran et al., 2004;Barnes et al., 2014;Robinson et al., 2017).While remote sensing of size-specific primary production revealed a synoptic view of community-resolved primary production, uncertainty remains due to regional variability in the 55 Biogeosciences Discuss., doi:10.5194/bg-2017Discuss., doi:10.5194/bg- -164, 2017 Manuscript under review for journal Biogeosciences Discussion started: 8 May 2017 c Author(s) 2017.CC-BY 3.0 License.physiological properties in association with the ambient environment such as the sea temperature (Carr et al., 2006).
Our objective in this study is to estimate the regional, geographical patterns of the groupspecific primary production (i.e., diatoms, haptophytes, and cyanobacteria) and the related photosynthetic properties in the Kuroshio Current and its adjacent waters in the North Pacific.We have 60 developed a novel remote sensing methodology by integrating an image processing technique with a bio-optical model, and relaxing the assumptions of a certain physiology-environment association made in the previous approaches: the group-specific photophysiological properties, such as the quantum yield index (i.e. a scaled quantum yield) and the chlorophyll-specific absorption coefficient, are also obtained from space in this paper.65 2 Material and Method

In situ data
In this analysis, we defined the Kuroshio region as the area between 120 to 160°E and 24 to 44°N, 70 which includes the Kuroshio Current itself (shown schematically as a white arrow in Fig. 1) and adjacent waters.The southern side of the Kuroshio is characterized as a subtropical water whereas the northern side as a sub-polar water.We used in situ measurements of phytoplankton pigments measured with High Performance Liquid Chromatography (HPLC) and the spectral optical absorption coefficient of phytoplankton,  ℎ () where a wavelength is denoted by .The HPLC pigments were used to 75 calibrate a satellite algorithm, OC-PFT (Hirata et al., 2011), to derive the phytoplankton composition in the Kuroshio region, while the absorption coefficient was used in a bio-optical model to estimate the group-specific primary production.

Biogeosciences
The absorption data was only available for the R/V Soyo-Maru SY-13-04 cruise.To increase robustness of our analysis, all absorption data from the surface to the maximum of 59 m were included 85 in our analysis (n = 17).As shown in Section 3.1, that did not affect our result but rather supported the general applicability of our bio-optical model over different spaces (horizontally and vertically).The absorption coefficient was measured using the filter pad technique (Mitchell et al., 2000) with a pathlength correction scheme (Cleveland and Weidemann, 1993).We also used surface absorption data from geographically distinct regions, such as the Benguela upwelling waters (Fishwick et al., 2006) and 90 other globally located regions (Werdell and Bailey, 2005), to support robustness of a bio-optical model.

Satellite data
Our estimation of the group-specific primary productivity requires satellite data of (1) the chlorophyll a concentration, Chla [mg m -3 ]; (2) the absorption coefficient of the total phytoplankton 95 community at 510 nm, [m -1 ]; (3) the primary production of the total phytoplankton community, PP [mg C m -2 day -1 ]; and (4) the Photosynthetically Available Radiation, PAR [μmol m -2 day -1 ] as inputs.The Level 3 SeaWiFS (Sea-viewing Wide Field of View Sensor) monthly 9 km remote sensing reflectance, R rs [sr -1 ], at wavelengths of 412 nm, 443 nm, 490 nm, 510 nm, and 555 nm, Chla, and PAR for the period of 1998-2007 were obtained from NASA.The remote sensing reflectance was inverted 100 to derive using the inversion model of Smyth et al. (2006).In addition, Chla derived from R rs (O'Reilly et al., 1998) was used to derive the relative abundance of the three phytoplankton taxonomic groups (i.e., diatoms, haptophytes, and cyanobacteria [%]) following the procedure of Hirata et al. (2011), which was specifically calibrated for in situ data of the phytoplankton community structure in the Kuroshio region described in the previous section; see Fig. 2. The relative abundance of each 105 phytoplankton group was subsequently multiplied by Chla to obtain the group-specific Chla, that is, Values of PP for the total phytoplankton community derived from the Carbon-based Production Model (CbPM, Westeberry et al., 2008)

Bio-optical model of taxon-specific primary production 115
Because most of the satellite signal originates from the surface of a water column, the depth dependency of the quantities described hereafter will be omitted for brevity, unless otherwise stated.
Our data from the Kuroshio region show a hinge point at approximately 510 nm in the  ℎ () spectra (Fig. 3a).(510) are approximately 0.973, close to unity.The coefficient of variation (CV) at 510 nm also displays one of the lowest values, 0.035 (Fig. 3).The same analysis for the biologically productive Benguela upwelling waters, which are geographically and ecologically different from the Kuroshio region, shows  ℎ (510) = 0.882 (Fig. 3b).The CV (0.035) 145 at 510 nm is close to the minimum (0.034) found at 501 nm (the wavelength of 501 nm is however not used by historical ocean colour instruments).Furthermore, the global dataset (the NASA bio-Optical Marine Algorithm Dataset, NOMAD, Werdell and Bailey 2005) returned  ℎ (510) = 0.992 with one of the lowest CV values of 0.087, when discrete measurements of  ℎ () is interpolated over the spectrum (Fig. 3c).The small variability in  ℎ (510)(= 0.949 ± 0.06) over the three different 150 datasets shows the quasi-constant characteristic of  ℎ (510) despite the large geographical differences between these datasets.Subsequently,  ℎ () ̅̅̅̅̅̅̅̅̅ can often be approximated by  ℎ (510) ⋅  ℎ (510) worldwide with a quasi-constant value of  ℎ (510).Note that our data from the Kuroshio included not only horizontal but also vertical distribution of a ph (λ), supporting this approximation.The small variability in  ℎ (510) over large geographical and temporal scales implicitly indicates that 155  ℎ (510) is also quasi-constant irrespective of phytoplankton community structure.
When Eq. ( 1) is applied to each phytoplankton group, the group-specific primary production PP i is expressed by where the quantum yield index is given by Φ  # = Φ  ⋅   * and the absorbed PAR by   510 = 300 ⋅  ⋅  ℎ (510) ⋅  ℎ, (510).Meanwhile, the primary productivity of the total phytoplankton community PP is a linear sum of those of each taxonomic group   (i.e. = ∑    =1 ), where i is the index for phytoplankton taxonomic groups and Np is total number of phytoplankton groups.When is the inverse of the matrix   .While the primary production of total phytoplankton community can be obtained using satellite data (Behrenfeld et al., 2005;Westberry et al., 2008), determination of the 180 matrix of the absorbed PAR  requires  ℎ (510), PAR and  ℎ, (510).We use the value 0.949 for  ℎ (510) in the virtue of its quasi-constant characteristic.Satellite data of PAR is also available (e.g.Frouin et al., 2012).Hence, only  ℎ, (510) needs to be known.Sections 2.2.2 explains how to obtain  ℎ, (510).Once  ℎ, (510) is derived,  will be known.Using  and ,   # can be derived with Eq. ( 3), with which the group-specific primary production for diatoms, haptophytes and cyanobacteria 185 is finally obtained by Eq. (2).
Hereafter, the quantum yield index Φ  # will be invoked to represent Φ  or the state of the photophysiology of each phytoplankton group, even though Φ  # (= Φ  ⋅   * ) is not precisely the same quantities as Φ  .We will also invoke  ℎ, (510) to represent the group-specific Chla biomass due to the tight correlation between  ℎ (510) and Chla found in the local and global data (r 2 = 0.90, p < 190 0.001), despite the fact that 510 nm is located within a carotenoid absorption band(s) rather than a chlorophyll absorption band.

Spatial data sub-sampling 220
Eqs. ( 3) and ( 5) cannot be preformed for each grid or pixel in a satellite image because only one value is available at each pixel for each input variable (e.g., only one value of Chla i is available for (x, y), where x and y represent the longitudinal and latitudinal coordinates, respectively).However, they are solvable when N number of neighboring pixels are used.By defining a geographically small region (or 225 "window") consisting of N=n × n neighboring pixels in a satellite image of each input variable (i.e. N=n 2 satellite data samples are contained in the window), we can sub-sample the satellite data from the window to collect N measurements.We selected n=5 (i.e.N = 5 × 5 = 25 pixels) for a square window in the present analysis in order to achieve a balance between the statistical robustness of the regression and the resulting degradation of the spatial resolution that was tolerable for our later analysis.With 230 n=5, the system of the simultaneous equations (i.e.Eq. ( 5)) established for the window is usually overdetermined as there are only three unknowns (i.e. ℎ, * (510) for the three phytoplankton groups).
Hence, the system is solved by the least square method.One value of  ℎ, * (510) of each phytoplankton group (3 groups in our case) is then obtained for that window.By repeating this operation for neighboring windows within the same satellite image, one can obtain a map of  ℎ, * , (510) as a 235 collection of output values from these windows, although the output images from this procedure has less spatial resolution than those of the input variables.Once  ℎ, * , (510) is derived,  ℎ, (510) is obtained, too, by multiplying  ℎ, " (510) by Chla i .Using  ℎ, (510) , the similar procedure is repeated with Eq. (3), which then gives a map of Φ  # , hence of PP i , for the 3 groups.
We further repeat these operations for monthly images to generate a monthly time series of 240 the derived variables ( ℎ, * (510) ,  ℎ, (510), Φ  # , and   ) for the period of 1998-2007, from which monthly climatological data is obtained.Note that, for each window, a correction of the degree of freedom may be required in the regression analysis when a significant spatial autocorrelation among data samples within a window of an input variable(s) is found.Also note that the degree of tolerable degradation of spatial resolution of derived quantities is application-specific, and one may change the 245 window size as appropriate.
As a result, the integration of the bio-optical theory and the multi-pixel image processing enables the derivation of the biological quantities  ℎ, * (510) ,  ℎ, (510) , Φ  # , and   for each phytoplankton group considered here.Multiple linear regression analysis was performed between the dependent variable PP i and the independent variables  ℎ, (510) and Φ  # using their monthly climatology over 1997-2007.All variables were standardized (i.e. a mean value was subtracted from the original data, and a resultant 255 value was divided by the standard deviation) prior to the regression analysis.The analysis was performed using spatial data within the Kuroshio region for each month.The multiple linear regression coefficients for the standardized  ℎ, (510) and Φ  # were defined as their contributions to PP i , and seasonal variation of the contributions was evaluated for each taxonomic group.and cyanobacteria of the entire region defined in our analysis were 0.21 mg Chla m -3 , 0.11 mg Chla m - 3 , and 0.03 mg Chla m -3 , respectively.While the spatial average of Chla i depends on the region 270 defined, we found the clear tendency that the diatom-and haptophyte-derived Chla i were higher on the northern side of the Kuroshio Current.Conversely, cyanobacterium-derived Chla i were rather uniform over the region.
The relative abundance of diatoms also showed a similar spatial pattern to Chla i.However, haptophytes and cyanobacteria did not.Haptophytes exhibited a higher relative abundance along the 280 Kuroshio Current (Fig. 4e), while cyanobacteria did so on the southern side of the Kuroshio Current (Fig. 4f) (see Fig. 1  haptophytes, and cyanobacteria over the region were 18%, 30%, and 17%, respectively, indicating again that haptophytes are the most dominant group in the region The climatological regional averages of  ℎ, * (510) for diatoms, haptophytes, and 285 cyanobacteria were 0.016 m 2 mg Chla -1 , 0.027 m 2 mg Chla -1 , and 0.040 m 2 mg Chla -1 , respectively.

The magnitude of 𝑎 𝑝ℎ,𝑖
The spatial distribution of Φ  # was classified into two patterns for the three phytoplankton groups.The diatom-derived Φ  # , with a climatological regional average of 0.8 × 10 -3 , was higher on the southern side of the Kuroshio Current (~0.2 × 10 -2 ) than on the northern side (< 0.1 × 10 -2 ).The haptophyte-derived Φ  # had a similar pattern to that of the diatoms but was relatively more uniform over the region with a climatological regional average of 0.2 × 10 -3 .Conversely, the cyanobacterium-300 derived Φ  # (climatological regional average of 0.2 × 10 -3 ) was to some extent higher on the northern side (~0.5 × 10 -3 ) of the Kuroshio Current than on the southern side (~0.2 × 10 -3 ), contrasting with the spatial patterns of the diatom-and haptophyte-derived Φ  # .
The climatological regional averages of PP i for diatoms, haptophytes, and cyanobacteria were 134 mg C m -2 day -1 , 72 mg C m -2 day -1 , and 40 mg C m -2 day -1 , respectively.The values of PP i 305 showed a clear latitudinal gradient for all groups such that they were higher at higher latitudes and lower at lower latitudes.However, the gradient was largest for diatoms and smallest for cyanobacteria.
The latitudinal gradient of PP i was therefore in agreement with Chla i for diatoms but not necessarily so for cyanobacteria.As a result, the diatom-derived PP i had a spatial pattern very different from that of the diatom-derived Φ  # , whereas the cyanobacterium-derived PP i had a pattern closer to that of the 310 cyanobacterium-derived Φ  # .
3.2 Factors controlling group-specific primary production: which is more important, chlorophyll a biomass or physiology?

315
Figure 5 shows the relative contribution of variability in the group-specific phytoplankton abundance represented by  ℎ, (510) and the photophysiology represented by Φ  # to the variability in PP i .The contribution of the abundance and photophysiology were not equal in general and depended on the taxonomic group and the season.For diatoms, the contribution of the abundance to PP i was always larger than that of the photophysiology.The abundance contribution for diatom PP i was larger than the 320 physiological contribution throughout the year and was particularly large between May and November (0.77-0.85) compared to between December and April (0.38-0.64).Conversely, the photophysiological contribution to PP i if diatoms were small or even absent over the all seasons (0-0.24).In general, diatom PP i was derived by the abundance.For haptophytes, the physiological contribution remained large throughout the year, with a relatively smaller contribution (~0.58) in boreal summer (June-325 September) and a relatively larger contribution (~0.89) in winter (November-April).Conversely, the abundance contribution was highest (0.65) in June and lowest (0.25) in January.Thus, the two contributions were out of phase for haptophytes; however, this does not indicate a simple alternation between the abundance and physiological contributions over the year.Both the abundance and physiological contributions were found significant from May to November, whereas the physiological 330 contribution dominated from December to April.For cyanobacteria, the physiological contribution always prevailed throughout the year (> 0.65), while a biomass contribution was also found, but only to a small degree (< 0.26).

Discussion 335
The highest climatological regional average of primary production was found for diatoms in the entire Kuroshio region (134 mg C m -2 day -1 , 72 mg C m -2 day -1 , and 40 mg C m -2 day -1 for diatoms, haptophytes, and cyanobacteria, respectively).However, within the Kuroshio Current and its extension domain (the area between two dotted curves in Fig. 4p and 4q), the PP i for haptophytes was found to be 340 higher than that for diatoms.Even though a direct validation of the satellite estimate of PP i remains to be conducted, a recent in situ observation by Nishibe et al. (2015) within the Kuroshio domain showed that smaller phytoplankton (< 10 μm) have higher production (61-185 mg C m -2 day -1 ) than larger phytoplankton (> 10 μm) (82-871 mg C m -2 day -1 ).In addition, independent in situ measurements of molecular analyses using an Ion Torrent high-throughput sequencer indicated an elevated diversity of 345 haptophytes in the upstream Kuroshio Current (Suzuki et al., unpublished data).An additional calculation via light microscopy in the Kuroshio Current also found that haptophytes were most closely associated with the Kuroshio front itself (Clayton et al., 2014).Our analysis is consistent with in situ studies, both in identifying the taxonomic group and for estimating the corresponding PP i simultaneously with a larger geographical coverage.350 A high abundance of cyanobacteria was also reported by means of a cell count on the southern edge (sub-tropical side) of the Kuroshio front (Clayton et al., 2014).However, our estimates of the cyanobacterium-derived PP i may only be comparable to the haptophyte-derived PP i and may not be dominant.While the temporal and spatial scales between their analysis and ours are different, the gap can be explained by our finding that the cyanobacteria PP i were regulated by their photophysiology 355 rather than their abundance.Although phytoplankton community structure is not explicitly resolved, Behrenfeld et al. (2005) showed from a satellite data analysis a less co-variation or even a decoupling between phytoplankton biomass and chlorophyll concentration in oceanic regions where chlorophyll variance is small (i.e.sub-tropical waters).In addition, White et al. (2015) showed from field observations that the HPLC-based abundance of phytoplankton groups in the subtropical North Pacific 360 was not significantly correlated with the primary productivity in the upper layers of the ocean when picoplankton was dominant, suggesting a potential role of photophysiology on PP.We conclude that our methodology and analysis captured the major mechanism contributing to PP in the study area.
Of the quantities obtained using the presented approach, only Φ  # does not precisely equal the definition of the variable it was meant to represent (i.e., the quantum yield of photosynthesis).Even 365 though Φ  # was designed to represent Φ  , it is only an index of Φ  .The treatment of the numerical value of Φ  # requires care, especially when it is compared with Φ  values obtained from other studies.
In addition, Φ  is defined here for carbon fixation not for oxygen evolution.The climatological regional averages of the derived Φ  # were 0.8 × 10 -3 , 0.2 × 10 -3 , and 0.2 × 10 -3 for diatoms, haptophytes, and cyanobacteria, respectively (Fig. 4m-o).Conversely, the maximum quantum yield of 370 carbon fixation for micro-, nano-, and picoplankton for the surface layer of other oceanic domains, determined from in situ databases, were reported as 46-88 × 10 -3 , 7-47 × 10 -3 , and 4-25 × 10 -3 , respectively (Cleveland et al., 1989;SooHoo et al., 1987;Uitz et al., 2008;Lindley et al., 1995).Therefore, our Φ  # and the reported quantum yield, as expected, differ from these over two orders of for diatoms, haptophytes, and cyanobacteria, which are in consistent with the previous reports mentioned above.In addition, the general tendency of our result, that Φ  # of diatoms is higher while Φ  # of cyanobacteria is lower, was also in consistent with Claustre et al. (2005).We conclude that our Φ  # 380 captured the nature of the quantum yield of photosynthesis both quantitatively and qualitatively.
Figure 4 demonstrated that there is large geographical variability in Φ  # even within a single taxonomic group.Our result showed that cyanobacteria had lower Φ  # in the subtropical waters south of the Kuroshio Current and higher Φ  # in the subarctic waters.Cyanobacteria are known to have higher contents of zeaxanthin (Suzuki et al., 2015) that can dissipate excessively absorbed light energy 385 to protect the photosynthetic apparatus from photooxidation due to over-excitation (Dall'Osto et al., 2012).The dissipation of the absorbed light energy results in less effective photosynthesis per unit mole of photons absorbed, leading to a smaller quantum yield at lower latitudes where light is usually not limited.This scenario may explain our result for cyanobacterium-Φ  # .
Class-dependent variation was found in the  ℎ, * (510) derived from the satellite observations 390 such that diatoms (cyanobacteria) have lower (higher)  ℎ, * (510) values, whereas haptophytes have an intermediate value.This result agrees with both theory (Morel and Bricaud, 1981) and observations for size-fractionated  ℎ * (510) for microplankton, nanoplankton, and picoplankton (Ciotti et al., 2002).In situ observations (the NOMAD dataset) indicate that the total  ℎ * ( 510) is significantly correlated with the total Chla at a global scale (r 2 = 0.60, p < 0.001, in log scale), despite that it is already divided by 395 Chla by definition, (i.e.,  ℎ * =  ℎ (510)/ℎ).This correlation not only indicates a non-linearity between  ℎ (510) and Chla but also implies a correlation between  ℎ (510) and  ℎ () in the chlorophyll a absorption bands.Indeed,  ℎ (510) were significantly correlated with  ℎ (670)(r 2 = 0.90, p < 0.001) and  ℎ (443)(r 2 = 0.99 and p < 0.001).Given that the wavelength of 510 nm falls in a carotenoid absorption band, it indirectly suggests that the carotenoids and Chla are correlated at larger 400 scales, at least to the first order.As a result,  ℎ * (510) still carries information of Chla.Because Chla (and  ℎ ( 443)) is known to contain information concerning the phytoplankton community structure at larger scales (e.g.Uitz et al., 2006;Hirata et al., 2008, Brewin et al., 2010, 2011), so does  ℎ * (510), which would partly explain its class-dependent variations.However, a care must be taken that smaller spatio-temporal scales (inc.coastal 405 Variability in  ℎ * () has often been attributed to a phytoplankton community change in association with the cell size via the packaging effect, which is the basis to estimate phytoplankton in different size classes (e.g., Ciotti and Bricaud, 2006;Mouw and Yoder, 2010).Because the package effect is a function of the optical thickness of a cell (i.e. the product between the cell size and the cell matter absorption including intercellular Chla (Morel and Bricaud, 1981), not only physical effects of 410 cell size but also the intercellular Chla abundance results in the package effect.It has been reported that higher bulk Chla was observed as larger cells become more abundant (e.g., Suzuki et al., 2011).When there is a correlation between the cell size and the cellular abundance of Chla, the indirect effect of cell size on  ℎ * () via the cellular abundance of Chla is also a source of variation in the package effect.
Provided that the total Chla in bulk water is a simple sum of the cellular Chla of the individual cells, 415 the relationship between the bulk Chla and the phytoplankton size structure (Hirata et al., 2011) is partly a reflection of the relationship between  ℎ * () and the community structure.
According to the mean theorem of integration, there must be at least one wavelength ′ where  ℎ () ̅̅̅̅̅̅̅̅̅ =  ℎ (′) .We found that  ℎ (510) was numerically close to unity, not only for the Kuroshio region (0.972) but also in other oceans (see section 2.2.1); therefore, 510 nm is often likely to 420 be one of the ′, even though it may not always be the case because  ℎ (510) does have variability as expressed by the CV.SeaWiFS had a waveband that has the central wavelength of 510 nm.
However, subsequent satellite instruments, such as MODIS onboard AQUA and VIIRS on Suomi NPP abandoned it.It is desirable to include the band in future satellite instruments.
In situ Rrs can also exhibits a hinge point at the wavelength of 490-520 nm (Reynolds et al. 2001).
Although Rrs spectrum, hence emergence of the hinge point as well as the wavelength at which the hinge point appears, is a result of inter-play among any optical constituents responsible for the spectra (e.g.colored dissolved organic matter, inorganic particles etc.), the hinge point at 510 nm in the  ℎ () 430 spectra raises the speculation that the hinge point of the normalized  ℎ () may be a basis of the hinge point seen in Rrs, as (i) the wavelength of the hinge points are relatively close and (ii) Rrs is also a normalized quantity in that it is defined by the upward radiance divided by the radiance integrated over a specific space (i.e. the downward irradiance).
The novelty of our approach is the integration of the image processing with a bio-optical 435 model.It allows a satellite derivation of new photophysiological quantities, such as Φ  # and  ℎ, * (510).As a consequence, the photophysiological and absorption properties of each phytoplankton group need not to be assumed a priori when the group-specific primary production.In satellite ocean colour remote sensing, it is known that the ocean color and bio-optics inversions are mathematically illposed problems (Sydor et al., 2004;Defoin-Platel and Chami, 2007), and the information available 440 (i.e., spectral measurements) is usually deficient in comparison with the number of unknowns, if only multi-spectral measurements, as those taken by conventional satellite observation, are available.The presented approach attempted to find additional information by spatial sampling of input variables, taking full advantage of the simultaneous acquisition of spatial data by the satellite remote sensing.
A drawback of the presented approach is the degradation of the spatial resolution of outputs 445 (i.e.PP i ,  ℎ, * (510), and Φ  # ).In this study, a 5 × 5 window was used to derive these quantities resulting in the spatial resolution being 5-times coarser than the original satellite image of the input variables.Future ocean color satellite sensors will have finer spatial resolution than the earlier satellites used in this analysis (e.g., a nominal spatial resolution of 300 m was achieved by OLCI (Ocean and Land Color Instrument) onboard Sentinel-3 and 250 m by SGLI (Second-generation Global Imager) on 450 GCOM-C (Global Change Observation Mission -Climate).However, it should be emphasized that a further improvement of the spatial resolution of the PP i ,  ℎ, * (510), and Φ  # , requires continuous development of advanced ocean color instruments with finer spatial resolutions, even if the spatial resolution of the satellite signal and the conventional variables, such as Chla, reach a sufficient level for the observation of pelagic oceans.455 Our approach internally involves a statistical procedure (i.e., multiple linear regression analysis), and the analysis is strongly constrained by, or based on, a bio-optical principle or model.Therefore, it is not similar to the conventional "empirical" methodologies that do not necessarily involve bio-optical mechanisms in them.As a result, in theory, it is expected that our approach can be applied to larger or smaller spatial scale analyses (e.g., basin/global and local scales).However, 460 geographical maps of the derived quantities (e.g., Φ  # and  ℎ, * ( 510)) tend to be spatially noisy due to the introduction of the statistical procedure.Because the spatial resolution is already reduced due to the methodological principle, temporal smoothing of the derived quantities may be a better way to reduce the noise at a cost of the temporal resolution.Hence, our methodology would be more robust for climatological analyses.The Advanced Himawari Imager (AHI) on board the recently launched 465 geostationary satellite Himawari-8 observes the ocean color every 2.5 min or 10 min, depending on the focused region within the field of view.Therefore, noisy outputs are expected to improve with the high temporal frequency observations of AHI, which would allow for the application of the presented methodology at finer temporal scales.In summary, the presented approach has room to improve further via advances in the spatial and temporal resolutions of satellite observations.470

Conclusions
In situ measurements of the optical absorption coefficient of phytoplankton showed that the inversed data of the normalized spectrum often exhibited a hinge point at 510 nm which may be a basis of a hinge point of Rrs spectra observed.In addition, the absolute magnitude (i.e.un-normalized) of the 475 absorption coefficient at this wavelength often corresponded to the magnitude of spectral average over 400 and 700 nm.As a result, the importance of wavelength of 510 nm was highlighted for ocean colour and photosynthetic study, and an inclusion of this wavelength in the future satellite and/or in situ observation mission of ocean colour was recommended.
In the virtue of the representativeness of the phytoplankton absorption coefficient at 510 nm as 480 its spectral average over 400-700nm, and by taking advantage of spatial bulk sampling by satellite remote sensing, a proof of concept of satellite-based bio-optical inversion was provided by utilizing spatial variability in satellite-derived phytoplankton absorption data, rather than spectral variability.
The new methodology demonstrated a possibility of derivation of the new satellite products such as the group-specific quantum yield index, the group-and the chlorophyll-specific absorption coefficients at 485 510 nm and the group-specific primary production.Among those quantities, the index of the quantum yield for photosynthesis was of particular importance, as it opened a door to access to photophysiological information of marine phytoplankton from space at large scales so that more direct assessment of primary production may become possible without assuming the photo-physiology.
Analysis of the satellite data of the group-specific quantum yield index and the primary 490 production showed that variability of primary production by diatoms in the Kuroshio region was driven due to the variability of their abundance, whereas variability of the cyanobacteria production was driven by physiology.In addition, the variability of haptophytes production showed comparable effects from both abundance and physiology in northern summer.
While validation of the derived quantities, as well as applicability test of the methodology to 495 temporal/spatial scales especially smaller than those considered in our study, remains to be done.More frequent simultaneous observation of a suit of photosynthetic parameters and the ocean colour variables should be carried out.

Competing interest 500
The authors declare that they have no conflict of interest.

Figure 4
Figure4shows the climatological distribution of the absolute abundance of Chla i , the relative abundance of Chla i ,  ℎ,(510),  ℎ, * (510), Φ  # , and PP i of the three taxonomic groups for the period of 1998-2007.The climatological average of the absolute abundance of Chla i of diatoms, haptophytes, also).The climatological averages of the relative abundances of diatoms, Biogeosciences Discuss., doi:10.5194/bg-2017-164,2017 Manuscript under review for journal Biogeosciences Discussion started: 8 May 2017 c Author(s) 2017.CC-BY 3.0 License.
with the same temporal and spatial resolutions as (http://www.science.oregonstate.edu/ocean.productivity/).1102.2Methodology It corresponds to the central wavelength of one of the SeaWiFS wavebands