Scholarly article on topic 'Mixed layer depth and chlorophyll a: Profiling float observations in the Kuroshio–Oyashio Extension region'

Mixed layer depth and chlorophyll a: Profiling float observations in the Kuroshio–Oyashio Extension region Academic research paper on "Earth and related environmental sciences"

CC BY
0
0
Share paper
Academic journal
Journal of Marine Systems
Keywords
{"Mixed layer depth" / "Algal blooms" / "Chlorophyll a " / Turbidity / Phytoplankton / "Subsurface drifters" / "Kuroshio–Oyashio Extension region"}

Abstract of research paper on Earth and related environmental sciences, author of scientific article — Sachihiko Itoh, Ichiro Yasuda, Hiroaki Saito, Atsushi Tsuda, Kosei Komatsu

Abstract Variability in the chlorophyll a concentration (Chl) in relation to fluctuations in the mixed layer (ML) was investigated together with turbidity (Tur) in the Kuroshio–Oyashio Extension region, using profiling floats. A particular focus was the validity of two hypotheses concerning the spring bloom: the critical depth hypothesis (CDH) and the recently proposed alternative, the disturbance-recovery hypothesis (DRH). During the period from winter to early spring, Chl and Tur integrated over the photosynthetically active layer (PL; defined as the greatest depth of the ML and the euphotic layer) increased with increasing PL depth (PLD), indicating an increase in the phytoplankton biomass. This result is partly consistent with the DRH in that the observed increase in biomass was not explained by an increase in production. Instead, it was more likely attributable to a reduction in the loss rate. However, theoretical analyses revealed that grazer dilution alone could not cause this increase in biomass because such an increase in the ML in the real ocean (as opposed to a dilution experiment within a bottle) would cause a reduction in the mean light intensity. Despite the loss-controlled fluctuation in biomass during the period of low light, a production-driven fluctuation in biomass was also revealed. This occurred when the light intensity was elevated, particularly after late spring, and was consistent with the CDH. Thus, the present study suggests that both the production-driven and loss-driven hypotheses are responsible for the dynamics of the phytoplankton dynamics from winter to spring in the Kuroshio–Oyashio Extension region.

Academic research paper on topic "Mixed layer depth and chlorophyll a: Profiling float observations in the Kuroshio–Oyashio Extension region"

ELSEVIER

Contents lists available at ScienceDirect

Journal of Marine Systems

journal homepage: www.elsevier.com/locate/jmarsys

JOUHNAL OF

MARINE SYSTEMS

Mixed layer depth and chlorophyll a: Profiling float observations in the Kuroshio-Oyashio Extension region

Sachihiko Itoh a'*, Ichiro Yasuda a, Hiroaki Saito a, Atsushi Tsuda a, Kosei Komatsu a,b

a Atmosphere and Ocean Research Institute, The University of Tokyo, Kashiwa, Chiba, Japan b Graduate School of Frontier Sciences, The University of Tokyo, Kashiwa, Chiba, Japan

ARTICLE INFO ABSTRACT

Variability in the chlorophyll a concentration (Chl) in relation to fluctuations in the mixed layer (ML) was investigated together with turbidity (Tur) in the Kuroshio-Oyashio Extension region, using profiling floats. A particular focus was the validity of two hypotheses concerning the spring bloom: the critical depth hypothesis (CDH) and the recently proposed alternative, the disturbance-recovery hypothesis (DRH). During the period from winter to early spring, Chl and Tur integrated over the photosynthetically active layer (PL; defined as the greatest depth of the ML and the euphotic layer) increased with increasing PL depth (PLD), indicating an increase in the phyto-plankton biomass. This result is partly consistent with the DRH in that the observed increase in biomass was not explained by an increase in production. Instead, it was more likely attributable to a reduction in the loss rate. However, theoretical analyses revealed that grazer dilution alone could not cause this increase in biomass because such an increase in the ML in the real ocean (as opposed to a dilution experiment within a bottle) would cause a reduction in the mean light intensity. Despite the loss-controlled fluctuation in biomass during the period of low light, a production-driven fluctuation in biomass was also revealed. This occurred when the light intensity was elevated, particularly after late spring, and was consistent with the CDH. Thus, the present study suggests that both the production-driven and loss-driven hypotheses are responsible for the dynamics of the phytoplankton dynamics from winter to spring in the Kuroshio-Oyashio Extension region.

© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license

(http://creativecommons.org/licenses/by/4.0/).

CrossMark

Article history:

Received 17 April 2015

Received in revised form 22 June 2015

Accepted 24 June 2015

Available online 2 July 2015

Keywords:

Mixed layer depth

Algal blooms

Chlorophyll a

Turbidity

Phytoplankton

Subsurface drifters

Kuroshio-Oyashio Extension region

1. Introduction

In middle- and high-latitude oceans, the abundance and vertical distribution of phytoplankton from winter to spring are strongly linked to the variability in the surface mixed layer (ML). During winter, the phy-toplankton concentration is generally low and homogeneous within a deep ML, in which the mean light intensity is low. In spring, the surface phytoplankton concentration increases markedly (called the "spring bloom") in association with ML shoaling, which increases the mean light intensity. Sverdrup (1953) hypothesized that the spring bloom occurs when the ML becomes shallower than a critical depth, defined as the depth above which the areal gross production and loss of biomass are in balance. This is called the "critical depth hypothesis" (hereafter referred to as the CDH; nonstandard abbreviations are listed in Table 1).

The CDH has been used in various marine areas to investigate the conditions that initiate the spring bloom. For example, using climatological temperature and satellite ocean color data, Obata etal. (1996) investigated

* Corresponding author at: Atmosphere and Ocean Research Institute, The University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa, Chiba 277-8564, Japan. Tel.: +81 4 7136 6326, fax: +81 4 7136 6327.

E-mail address: itohsach@aori.u-tokyo.ac.jp (S. Itoh).

the global validity of the CDH based on a constant compensation irradiance (the irradiance level at which gross production balances the loss in biomass) of 1.5 Wm-2 (~0.43 mol quanta m-2 day-1). They found that the CDH explains basin-scale features of the spring bloom in the North Atlantic and western North Pacific, where the chlorophyll a concentration (Chl) at the surface typically doubles from March to May, when the mixed layer depth (MLD) becomes shallower than the critical depth. Whereas Obata et al. (1996) verified the CDH using a given compensation irradiance based on a phytoplankton culture experiment, Siegel et al. (2002) assumed the CDH and estimated the spatial distribution of the compensation irradiance (corresponding to the irradiance that balances the total loss) at the time of bloom initiation (defined as the time when the surface Chl exceeds the median value by 5%) in the North Atlantic by means of satellite remote sensing and hydrographic datasets. The estimated values were typically within the range of 1-1.5 mol quanta m-2 d-1 (range of zonal mean values) in the region 40-75°N in February to May.

Many studies have identified problems with the CDH (e.g., Smetacek and Passow, 1990; Nelson and Smith, 1991; Townsend et al., 1992; Stramska and Dickey, 1993; Marra and Barber, 2005). These problems were particularly related to several assumptions made by Sverdrup (1953) in the simple water column model that was used to assess the balance between production and loss. These assumptions included a

http://dx.doi.org/10.1016/j.jmarsys.2015.06.004

0924-7963/© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/40/).

Table 1

Nonstandard abbreviations and variables.

CDH Critical depth hypothesis proposed by Sverdrup (1953)

DRH Dilution recoupling hypothesis proposed by Behrenfeld (2010)

Aae Drop in potential density from 10m depth that defines the

mixed layer depth

Z0.415 Depth at which photosynthetically available radiation reaches

0.415 mol quanta m-2 d-1 (assumed irradiance limit for

photosynthesis)

PL; PLD Photosynthetically active layer, defined as either MLD or Z0.4i5,

whichever is deeper; PL depth

M-Chl (I-Chl) Mean (integrated) chlorophyll a concentration within the PL

M-Tur (I-Tur) Mean (integrated) turbidity within the PL

M-PAR Mean photosynthetically available radiation within the PL

F96 and F97 The two profiling floats used in the present study

WES Period from winter to early spring when the M-Chl was

observed to be stable

TR Period of transition following the WES period until the spring

fully turbulent surface ML, a constant light attenuation coefficient with respect to depth, a linear relationship between light level and photosyn-thetic production rate, and a known (fixed for the water column concerned) compensation irradiance. It is noteworthy that the third and fourth assumptions are used to derive the fixed respiration (loss) rate, which is considered constant within the ML. These issues related to the definition of the surface ML and the fixed loss rate have been highlighted in recent studies, including those of Taylor and Ferrari (2011) and Behrenfeld (2010), respectively. Below, we briefly review the background and studies related to these two issues ofthe definition of the surface ML and the fixed loss rate.

The first issue concerns the assumption of a fully turbulent surface ML.The photosynthetically available radiation (PAR) for individual phy-toplankton cells decreases with increasing depth, even within MLs characterized by homogeneous water properties, and the differences in PAR for cells only become negligible if PAR is averaged over a period longer than the time scale of vertical mixing within the ML. If the time scale of vertical mixing is longer than that of the gross growth of phytoplankton, the difference in the light level can create a vertical gradient in phytoplankton growth rates. More specifically, if turbulence is weak within the ML, a phytoplankton bloom can occur in the upper part of the layer, as proposed by Huisman et al. (1999). Based on this "critical turbulence hypothesis", Taylor and Ferrari (2011) showed analytically and numerically that shutdown of atmospheric cooling can cause a spring bloom.

Although Taylor and Ferrari (2011) assumed a deep MLD with weakened turbulence, the MLD in the real ocean generally reflects the turbulence intensity and the surface heat flux. Their numerical modeling reproduced an increase in phytoplankton concentration within the upper part of the deep ML identified from density stratification. We assume, however, results would be different if they had considered mixing caused by wind. Because the turbulent kinetic energy input from the wind is converted into potential energy, the shutdown of atmospheric cooling alone cannot suppress increasing upper layer potential energy, i.e., mixed layer development. Cessation of the mixed layer deepening needs loss of potential energy to be balanced with the energy input from wind, and this is achieved by atmospheric heating. The mixed layer at this moment can be diagnosed from converted turbulent energy forced by wind, and the net loss of potential energy by heating (e.g., Kraus and Turner, 1967). If wind is weak enough, mixed layer depth can be shallow; however, temperature increase (or potential density decrease) within the mixed layer can also be small if heating is moderate.

Although the MLD has been defined in practical terms using differences in water properties from the surface to depth, the criteria for establishing these differences (i.e., the extent of the drop in temperature, AT, or the rise in potential density, Act0) have been arbitrary. The most commonly used parameter is Aa9 = 0.125 kg m-3, which not

only has mainly been used to calculate the mean of multiple profiles (e.g., Levitus, 1983), but also has been applied to individual profiles (e.g., Suga et al., 2004). Because this value of Aue was originally proposed for climatological profiles, its application to individual or averaged profiles in the absence of ensemble profiles could result in it failing to identify the shoaling of the ML with small density difference caused by moderate heating, as discussed in the previous paragraph. For example, based on observations in the vicinity of New Zealand, Chiswell (2011) found elevated Chl in near-surface sublayers when it shoaled, defined using 0.025ct0. At the same time, the MLD defined using 0.125ct0 remained deep (Fig. 4 in Chiswell, 2011). Based on this observation, a stratification onset model was proposed for the spring bloom. However, we assume that the results of Chiswell (2011) would be consistent with the CDH when Aoe = 0.025 kg m-3. Therefore, defining the ML to makes it equivalent to the fully turbulent surface layer renders the "critical turbulence hypothesis" consistent with the CDH.

The second issue concerns the assumption of a fixed loss rate. Among many criticisms of this assumption, the one raised by Behrenfeld (2010) is fundamental. Based on analyses of satellite-derived phytoplankton data for the subarctic North Atlantic Ocean, Behrenfeld (2010) found that an increase in the surface Chl occurred before the MLD shoaled, at approximately the time when the ML ceased deepening. More importantly, the concentration ofphytoplankton-based carbon, integrated over the ML, increased in mid-winter when the MLD was increasing. Because the increase in biomass at the time of the lowest light levels cannot be explained by an increase in gross production (as suggested by the CDH), it has likely been caused by a reduction in losses. By analogy with the dilution experiment within a bottle (Landry and Hassett, 1982), Behrenfeld (2010) attributed the biomass increase to a reduction in grazing pressure resulting from the dilution of the ML by water from a deeper layer (originally termed the "dilution recoupling hypothesis", later expanded by Behrenfeld and Boss, 2014 as "disturbance-recovery hypothesis", and hereinafter referred to as DRH). The occurrence of positive net growth in mid-winter was supported by the results of Boss and Behrenfeld (2010), who used data from a fluorometer attached to a profiling float deployed in the western subarctic North Atlantic Ocean. An early increase in the integrated biomass was also reproduced using a biogeochemical model of the subarctic North Atlantic Ocean (Behrenfeld et al., 2013a).

More recently, field surveys and incubation experiments for the winter-spring bloom in the Long Island Sound, northeast of US also suggested the importance of zooplankton grazing (George et al., 2015). The onset of the bloom occurred when phytoplankton growth rates were greater than grazing mortality rates by microzooplankton, before thermal stratification was achieved. In the western Pacific subarctic gyre, Matsumoto et al. (2014) found that areal chlorophyll a concentration remained high despite reduced light availability caused by the ML deepening, suggesting the consistency with DRH.

A loss-rate-driven increase in the vertically integrated Chl (I-Chl) in mid-winter had been reported before the studies discussed above, in areas other than the North Atlantic Ocean. Based on analyses using a one-dimensional ecosystem model for the western subarctic gyre of the North Pacific Ocean, Yoshie et al. (2003) suggested that the vertically integrated Chl increases in mid-winter because of the dilution effect. However, as the observational data were very limited (eight occasions, once per year from 1991 to 1998), the relationship between the MLD and I-Chl remains to be confirmed. Yasuda and Watanabe (2007) measured Chl within the subtropical North Pacific Ocean gyre using a profiling float. They observed a positive relationship between the MLD and I-Chl during winter, when the MLD was 100-200 m, but found that I-Chl was very low when the MLD exceeded 250 m. This result was interpreted in the context of the CDH. Marra and Barber (2005) analyzed Joint Global Ocean Flux Study (JGOFS) data for the Arabian Sea to highlight the importance of the dilution effect, and noted that their results reflected a "natural dilution experiment".

As briefly reviewed above, although there has been substantial evaluation of the validity of the CDH (Sverdrup, 1953), the implication of the

DRH (Behrenfeld, 2010; Behrenfeld and Boss, 2014) is clear. The DRH proposes that a seasonal increase in the phytoplankton biomass is not caused by an increase in gross production with increasing light availability in spring, as proposed by the CDH, but by a reduction in the loss rate through a dilution effect, which can occur in mid-winter. Whereas the light availability for phytoplankton cells should be related more to the intensity of turbulence than to the MLD practically defined by a density profile (Huisman et al., 1999; Taylor and Ferrari, 2011), the critical turbulence hypothesis also considers the production-driven onset of the phytoplankton bloom. A comprehensive definition of the MLD should include the turbulent surface layer in the real ocean, where surface stirring by the wind usually contributes to mixing.

Although Behrenfeld (2010) proposed that the DRH should replace the CDH, a number of subsequent studies did not assess these two hypotheses in a manner comparable to that of Behrenfeld (2010). One important difference was the definition of "bloom initiation". Behrenfeld (2010) regarded the increase in the integrated biomass (based on Chl), measured in mg m-2, as indicating bloom initiation, whereas some other studies measured the increase in surface Chl in mg m- 3 (Suzuki et al., 2011; Chiswell, 2011; Mahadevan et al., 2012; Lavigne et al., 2013; Shiozaki et al., 2014). Although these studies generally did not support the DRH, it could not be rejected because their analyses were not conducted using ¡-Chl. Therefore, the issues of whether the phytoplankton biomass increases in mid-winter or not, and the

Fig. 1. Monthly sea surface chlorophyll a concentrations in the Kuroshio-Oyashio Extension region, determined with the SeaWiFS product: (a) January, (b) February, (c) March, (d) April, and (e) May. In (a), the paths of the Kuroshio Extension and the Oyashio Currents are drawn schematically in (a).

importance of loss effects, remain to be fully investigated. In this study, we considered these two variables using different units, and the term "bloom" is used to describe a prominent seasonal peak in the surface (or mean) Chl.

To investigate the variability in the Chl and its relationship to MLD, we made continuous observations of chlorophyll a fluorescence in the Kuroshio-Oyashio Extension region (Fig. 1) using profiling floats with a fluorescence sensor attached. In the Kuroshio-Oyashio Extension region in the western North Pacific, defined here as the region between the Kuroshio Extension and the extension (or return flow) of the Oyashio Current, an increase in the surface Chl resulting from a phyto-plankton bloom is generally observed in spring, especially along the northern flank of the Kuroshio Extension jet (Fig. 1). The Kuroshio-Oyashio Extension region is a key area for the recruitment of fishery resources, including the Japanese sardine (Nishikawa et al., 2011). The main sardine spawning grounds are located in southern coastal areas of Japan, and large numbers of their eggs and larvae are entrained and transported along the northern flank of the Kuroshio Extension and the transition region between the Oyashio (Itoh et al., 2009).

Shiozaki et al. (2014) recently investigated the distribution of the surface Chl in the western North Pacific, including the Kuroshio-Oyashio Extension region, using satellite-based data. They reported for the region north of the Kuroshio Extension, that the onset of the spring bloom was consistent with the CDH west of 150°E, whereas the eastward advection of Chl probably controlled the distribution of Chl east of 150°E. The validity of the DRH was not assessed, because they only analyzed the surface Chl.

In the present study, therefore, we assessed the variability of the Chl in areas north of the Kuroshio Extension from winter to spring using profiling floats, with particular reference to the validity of the CDH and the DRH in this region.

2. Materials and methods

2.1. Profiling floats

Two profiling floats (NINJA; Tsurumi-Seiki Co., Ltd.) were deployed near the separation point of the Kuroshio Current from the continental margin, during a research cruise (KT-08-7) on the R/V Tansei-maru at 26 April and 2 May, 2008. Each float was equipped with a conductivity-temperature-depth (CTD) profiler (SBE41, SeaBird Electronics Inc.) and a fluorometer with optical scattering measurement (FLNTU, WET Labs), which includes a wiper to keep the sensor clean. The floats were programmed to measure pressure, temperature, salinity, chlorophyll a fluorescence, and particle back-scattering profiles down to 500 dBar during the night, with a vertical measurement interval of approximately 2 dBar (10 dBar) in the upper 300 dBar (300-500 dBar), and a temporal interval of 5 days. During the periods between profiling and data transmission, the floats were parked at 40 dBar, where they measured the same parameters at 3 hourly intervals. In this study, the dBar pressure unit was assumed to be equal to the depth in meters (which is an appropriate assumption considering the vertical resolution of the data). The two floats measured profiles over a period of 14 and 21 months as they moved generally eastward, apparently in areas north of the Kuroshio Extension.

The raw data obtained were first corrected to exclude unrealistic errors, and filtered using a three-point median filter to remove noise. The profile data were then interpolated to 1 m intervals using the piecewise cubic Hermite interpolation method. Although profiles were measured up to 5 m from the surface, we assumed that temperature, salinity, Chl, and turbidity (Tur) were homogeneous near the surface, and applied the values for 5 m to the water column above this depth. From the pressure, temperature, and salinity data, the potential density was calculated and used to estimate the MLD. The values for 30-50 m were extracted from among the data obtained during the periods when the floats were parked, and subsequently used to investigate the diurnal cycle of chlorophyll a fluorescence.

The fluorometers were calibrated by the manufacturer using the coefficients of gain and offset (under dark conditions), but Chl estimated near the surface (~5 m) was not necessarily comparable to the estimates from satellite data, as reported by Boss et al. (2008). The gain coefficients of the fluorometers were adjusted to the satellite-derived data from the Sea-viewing Wide Field-of-view Sensor Project (SeaWiFS Project; see next subsection for the data specification), using least-squares fitting of log-transformed data (Fig. 2). Chl in the low-light layer (typically below 400 m depth) was approximately 0.1 mg m-3, with or without adjustment. Therefore, we treated values < 0.1 mg m-3 as zero measurements, to avoid overestimation of I-Chl.

Because the floats moved eastward for several hundred kilometers, profile measurements were not made at the same time on each measurement occasion. Thus, diurnal variations in the in vivo fluorescence were potentially present in the measured data. Nevertheless, based on data from the periods when the floats were parked, the mean amplitudes of diurnal variations were < 10% of the variations in the daily mean (data not shown). Although other factors such as nonphotochemical quenching of fluorescence (Marra, 1997) and diurnal-scale photoacclimation could potentially influence the observed chlorophyll a fluorescence, we assume that these effects were minor at least for periods from winter to spring on which the present study mainly focuses. Therefore, we used the chlorophyll a data obtained from fluorescence without considering the diurnal variations.

Tur was estimated from optical scattering measurements at 700 nm in nephelometric turbidity units (NTU). The typical range of Tur measured in the study region was 0.04-0.2 NTU. However, the sensitivity of the instrument used was 0.01 NTU, which produced low-resolution measurements, typically <0.1 NTU. Hence, the main results of the present study were based on Chl data. Nonetheless, the Tur data were similarly analyzed to investigate the link between Chl and phytoplankton biomass.

2.2. Satellite observations

The surface Chl and the daily integrated PAR determined for the wavelength range of400-700 nm (obtained, processed, and distributed by the SeaWiFS Project) were downloaded from the NASA SeaWiFS Project website (http://oceancolor.gsfc.nasa.gov/SeaWiFS). The eight-day mean, 9 km resolution, level 3 mapped products were used. Chl and PAR from the SeaWiFS Project were interpolated for each day and location of the profiles obtained by the floats. As precisely comparable temporal and spatial SeaWiFS Project data were not available for many of the float measurement occasions, the nearest SeaWiFS Project date was selected with respect to time, and Gaussian window averaging (radius 25 km) was used for spatial measurements. The SeaWiFS products were occasionally unavailable for periods exceeding 8 days, such as during the period from late April to mid-June 2009. In these cases, the nearest SeaWiFS Project data within ± 20 days were used. Data beyond this extrapolation were not used in detailed analyses, but are shown in subsequent figures, using linear interpolation.

2.3. M-Chl and I-Chl

To assess the variations from winter to spring, we analyzed the time series of the mean Chl (M-Chl) and I-Chl and the mean Tur (M-Tur) and integrated Tur (I-Tur) within a PL where phytoplankton photosynthesis was occurring, whereby M-Chl x PLdepth (PLD) = I-Chl and M-Tur x PLD = I-Tur (nonstandard abbreviations are listed in Table 1). The definition of PL was based on the integration depth defined by Boss and Behrenfeld (2010): this depth is from the surface to either the MLD or the depth at which PAR = 0.415 mol quanta m-2 d-1 (Z0.415 is an assumed irradiance limit for photosynthesis; see the following paragraph for the process of estimation), whichever is deeper. However, we used an MLD criterion Aa9 = 0.025 kg m-3; from 10 m depth) that differed from the conventional one (Aae = 0.125 kg m-3) used by Boss and Behrenfeld (2010), because we found that with the

S. Itoh etal./Journal of Marine Systems 151 (2015) 1-14

(a) 30096 (a) F96 o ^

SeaWiFS Chl [mg m-3]

100 -3]

(b) 30097

(b)F97 0 ' 100 ( )

°Q «P..

10-1 ^S-

• •

SeaWiFS Chl [mg m-3]

Fig. 2. Comparison of the chlorophyll a concentrations derived from the SeaWiFS product and from the profiling floats (at 5 m) for (a) float F96, and (b) float F97, in the present study. Open and solid circles indicate raw (factory-set) and adjusted data (see text), respectively.

latter, sublayers of significantly different Chl were occasionally included, as reported by Chiswell (2011). The criterion Act0 = 0.025 kg m-3 is similar to the widely-used value of 0.03 kg m-3 recommended for individual profiles (de Boyer Montegut et al., 2004), but was slightly modified to match the observed Chl profiles. The PLD was almost equivalent to the MLD (i.e., MLD > Z0.4i5) for the period before the bloom (winter to spring), except that short-term stratification occurred several times during this period.

For each profile, we estimated the light conditions within the water column using the empirical relationship between the 1% light level and the surface Chl, proposed by Morel et al. (2007). From the 1% light level (Z^%), the light attenuation coefficient within a water column was calculated as:

k = - log (0.01 )/Z

Using PAR data for the surface, obtained from the SeaWiFS Project, the PAR at depth z was calculated as:

PAR (z) = 0.98 x PAR (0) exp(-kz),

where the coefficient 0.98 is the transmission rate into the ocean used by Boss and Behrenfeld (2010). We calculate Z0.4i5 from Eq. (2). The mean PAR averaged over the PL (M-PAR) was also calculated. The PAR at the surface is hereafter referred to as S-PAR. By definition, M-PAR represents the mean light intensity for photosynthesis within the PL. If photosynthesis and the loss terms integrated over the PL are equal, the M-PAR of that water column can be interpreted as the compensation irradiance, in the context of the CDH.

3. Results

3.1. Movement of floats and variations in water mass and chlorophyll a profiles

The two floats (designated F96 and F97; World Meteorological Organization identities 30096 and 30097, respectively) migrated eastward within the latitudinal range of 32-40°N (Fig. 3). The general eastward movement would be caused by the eastward flow on the northern flank of the Kuroshio Extension, or eastward branch currents of the Kuroshio Extension. The cross-frontal locations of the floats relative to the current axis of the Kuroshio Extension can be determined from the indicative temperature of the current axis, which was empirically defined as 14 °C at 200 m (T200) (Kawai, 1969). Float F96 migrated to approximately 150-160°E over 14 months from late May 2008. The mean hydrographic features around the trajectory shown in Fig. 3, and the T200 value of 11-15 °C, indicate that F96 was in a cyclonic recirculation area north of the Kuroshio Extension (Fig. 4a). The eastward movement of F97 followed a more northerly path and was more straightforward. From July 2008 to January 2009 the T200 value for F97 was 8-11 °C, indicating that it moved in a zone to the north of the Kuroshio Extension. The temperature at 100 m was 12-15 °C, however, which is far warmer than the indicative temperature (4 °C) proposed for the subarctic front (Favorite et al., 1976) (Fig. 4b). The water mass containing F97 gradually approached to the main Kuroshio Extension stream, typically after June 2009, as indicated by the float positions and the increase in T200 (Figs. 3 and 4a).

Despite differences in the properties of the thermocline water, the seasonal cycle of the ML is evident in the data from each float (black dots in Fig. 4). In both 2008 and 2009, the ML began to deepen after

30°N 130°E

140°E

150°E

160°E

170°E

180°W

170°W

Fig. 3. Positions of the floats at 5-day intervals. Successive positions are connected by solid lines, and those determined at the beginning of each month are marked with circles and text indicating the month. The background gray contour lines are the mean absolute dynamic height from May 2008 to May 2009, obtained from Archiving, Validation and Interpretation of Satellites Oceanographic Data (AVISO).

0 100 200 300 400 500

0 100 200 300 400

(a) Temperature (F96)

S. Itoh et al. / Journal of Marine Systems 151 (2015) 1-14

(b) Temperature (F97)

(c) Salinity (F96)

(d) Salinity (F97)

MJJASONDJFMAMJJA M J J A S O N D J F M AM J J A S O N D J F M

25 20 15 10 5

34.5 34 33.5 33

Time (year and month)

Fig. 4. Time-depth profiles determined from floats F96 and F97 for temperature (a and b, respectively) and salinity (c and d, respectively). The contour lines are 2 °C intervals (5 °C for thick lines) for (a) and (b), and 0.2 psu intervals (0.5 psu for thick lines) for salinity. Black dots are the MLD, defined as a 0.025 ae rise from 10 m depth.

summer; this occurred in October for F96 and in September for F97. Because the deepening of the ML caused mixing with the saline subsurface water, the salinity within the ML generally increased during winter.

Immediately following the deployment of the floats in late April for F96 and in early May for F97 in 2008, the surface Chl near the coast exceeded 0.5 mg m- 3, and then decreased in early summer. The subsurface chlorophyll a maximum (SCM; Fig. 5a and b) was evident in June for F96 and in July for F97, and was maintained during summer 2008 until the deepening of ML in late autumn. Chl was apparently homogeneous within the ML after December 2008 for F96, and after November 2008 for F97 and this remained the case until March 2009. Although the fluctuations in Chl in spring were complex, the maximum concentration (>0.5 mg m-3 near the surface) occurred in April 2009

for F96 and in May 2009 for F97. Detailed descriptions of these fluctuations are provided below. The surface concentrations exceeded 0.1 mg m-3 for 1-2 months following the maximum concentration, and then gradually returned to the SCM. For F97, the fluctuations in Chl observed during the second year following the formation of the SCM in 2009 (i.e., up to February 2010) were similar to those during the first year, as noted above (Fig. 5b).

Temporal variations in the profiles of Tur generally resembled those of Chl (Fig. 5c and d). Tur increased in the surface layer in spring, formed a subsurface maximum in summer, and deeply penetrated the ML in winter, as was also observed for the Chl. This indicates that phytoplank-ton contributed predominantly to the total particle concentrations in the upper layer.

J= 100

J= 100

MJJASONDJ FMAMJ J AMJ J ASONDJ FMAMJJASONDJ FM

x- -M-M2C

2009 2008

Time (year and month)

Fig. 5. Time-depth profiles of chlorophyll a concentrations determined with floats F96 and F97 for chlorophyll a concentrations (Chl; a and b, respectively) and turbidity (Tur; c and d, respectively). The levels of Chl and Tur are shown with color scales, whereas the concentrations 0.2 and 0.5 mg m-3 for Chl and 0.08 and 0.12 NTU for Tur are shown with contour lines. Black dots are the MLD, defined as a 0.025 as rise from 10 m depth.

The data for Chl and Tur are directly compared in Fig. 6. Because particles other than active phytoplankton cells possibly dominated below the photosynthetic layer, part of which could be detected as small-scale turbid patches (Fig. 5c and d), only the data within the PLD were used. Although Tur generally increased with increasing Chl (Fig. 6a and b), the relationship between Chl and Tur differed among seasons and the Chl range, as evident in the logarithmic plots (Fig. 6c and d). Chl and Tur were elevated in the subsurface layer in summer (June, July, and August) and autumn (September, October, and November) (Fig. 5), and the surface concentrations were very low. The relative levels of Tur were high in the low Chl range (corresponding to the layer above the SCM; typically <0.2 mg m-3), whereas the opposite trend was observed in the high Chl range (corresponding to the SCM layer), which is attributable to photoacclimation. In winter (December, January, and February) and spring (March, April, and May), Chl and Tur occurred in relatively narrow ranges. A more detailed discussion of Chl and Tur in these seasons is presented in the next subsection.

With respect to the Chl and Tur profiles from autumn 2008 to spring 2009, both temporal and vertical variability occurred in the upper layer (Fig. 7). During late autumn to early spring, Chl showed fluctuations that ranged from 0.2 to 0.4 mg m-3 (Fig. 7a and b). Whereas, Chl within the ML defined by Aa0 = 0.025 kg m-3 was homogeneous, which was not the case for Chl within the ML defined using Aa9 = 0.125 kg m-3. This result indicates that the conventional Aae = 0.125 kg m-3 criterion does not always represent the turbulent surface boundary layer. Whereas short-term variations in ML (defined by Aae = 0.025 kg m-3) sometimes detrained the lower layer that still contained Chl > 0.1 mg m-3 (e.g., February for F97), Chl in the detrained layer decayed before the ML became deep again.

For F96, there were marked increases in Chl in January and February, in response to the rapid shoaling of the ML. Nevertheless, Chl decreased

rapidly as the ML became deep again (Fig. 7a). Except for these cases, Chl did not always increase as the MLD decreased until mid-March, as described below. Moreover, deepening of the ML did not clearly cause a reduction in Chl, other than in the cases noted above.

Tur fluctuated from approximately 0.07 to 0.1 NTU until early spring (Fig. 7c and d). The profiles of Tur were less smooth than those of Chl, partly because the typical sensitivity of Tur with the instrument used is 0.01 NTU. However, Tur values measured within the ML did not show a marked vertical gradient if the ML was defined by Aoe = 0.025 kg m-3.

To assess the profiles of Chl and Tur in relation to stratification from winter to spring, we analyzed the temporal variability in M-Chl and ¡-Chl, together with M-Tur and ¡-Tur in relation to the fluctuations in MLD, as discussed below.

3.2. M-Chl and ¡-Chl from winter to spring

From early winter 2008 (early December for F96 and late November for F97) to early March 2009 (= winter to early spring; WES), Chl and Tur were homogeneous within the PL (effectively the ML during this period) and relatively stable for both F96 and F97 (Figs. 8c, d, 9c and d). The WES periods were determined diagnostically from the vertical and temporal variability in Chl and Tur. During WES, the standard deviations (SD) for Chl and Tur within the PL for individual profiles (error bars for M-Chl in Figs. 8c, d, 9c and d) were small. The coefficients of variation (CV = SD / mean) calculated for individual Chl profiles were 2%-9% and 2%-10% for F96 and F97, respectively. Similarly, for individual Tur profiles, the CV were 0%-8% and 3%-7%, respectively.

The temporal fluctuations in M-Chl and M-Tur (fluctuations of the time series in Figs. 8c, d, 9c, and d) were also small during WES. The temporal mean M-Chl ± SD and CV were 0.33 ± 0.045 mg m-3 and

i ~ A" "

V<* ^o Ad 4a &A Mft

v w v yw □ □tj: AAifr

^s^Ai ¿A

'V 0.15

A mam 0.1

□ son 0.05

(b) F97 0 v_

O VS v

0.2 1 Chl [mg m-3]

0.1 0.2

Chl [mg m-3]

Fig. 6. Comparisons of the chlorophyll a concentration (Chl) and turbidity (Tur) for float F96 (a: normal scale; c: log scale) and float F97 (b: normal scale; d: log scale). Data above the PLD are plotted at intervals of 2 m and with different colors and marks for different seasons. In (c) and (d), linear lines are drawn to highlight the different relationships between Chl and Tur with season and Chl level.

Fig. 7. Time-depth profiles of chlorophyll a concentrations determined with floats F96 and F97 for the period from October 2008 to May 2009: chlorophyll a concentrations (Chl; a and b, respectively) and turbidity (Tur; c and d, respectively). The solid and dashed black lines indicate the MLD, defined using Aae = 0.025 kg m-3 (MLD0025), and Aae = 0.125 kg m-3 (MLD0125), respectively, and the pink dashed lines are the depth at which the PAR = 0.415 mol quanta m-1 day-1 (Zo.4i5). The color scale is narrowed to 0.1-0.4 mg m-3 for Chl and 0.07-0.10 NTU for Tur to focus on the variability from winter to early spring.

13%, respectively, for F96, and 0.31 ± 0.026 mg m 3 and 8.3%, respectively, for F97. The temporal mean M-Tur ± SD and CV were 0.088 ± 0.007 NTU and 7.8%, respectively, for F96, and 0.083 ± 0.008 NTU and 9.0%, respectively, for F97. The estimates for M-Tur were calculated below the measurement sensitivity of 0.01 NTU because 20-70 independent data values (2 m interval measurements over 40-140 m in the PLD) were used to calculate the means and SDs. Despite the short-term increase in mid-winter, these CV values were lower than those for the PLD itself: 34% (96 ± 33 m) for F96, and 28% (63 ± 25 m) for F97. Although the difference was less for F97 in the second winter (from early December to the end of the record in early February), the CV for M-Chl (17%; 0.29 ± 0.049 mg m-3) was also less than that for the PLD (24%; 107 ± 26 m) (Fig. 5b). The high CVs for PLD during the WES period were not caused by variations in Z0.415, but by variations in MLD, which was usually deeper than Z0.415 during this period.

M-PAR was < 4 mol quanta m-2 d-1 during the WES period, except for short-term peaks in the range of 4-6mol quanta m- 2 d -1 that were detected in January and February for F96 (Fig. 8a), and in February for F97 (Fig. 9a). These were caused by ML shoaling (Figs. 8b and 9b). In some of these cases, increases in M-Chl were observed (Fig. 8c, event I; Fig. 9c, event II). Although the amplitude was moderate, increases were also observed for M-Tur at the same time as those for M-Chl (Figs. 8d and 9d). However, correlation coefficients through the WES period were not significant either between PLD and M-Chl or PLD and M-Tur (not shown).

Despite the negative relationships between PLD and M-Chl or M-Tur, ¡-Chl and ¡-Tur responded positively to variations in PLD (Figs. 8b-d and 9b-d). Increases in ¡-Chl and ¡-Tur were observed several times when the PLD deepened (peaks exceeding 40 mg m-2 occurred from late January to early March for F96, and from early March to early April for F97) before the peaks in M-Chl and M-Tur in late spring (see below).

Following the stable WES period until early March, M-Chl and M-Tur began to increase, which was apparently related to an increase in the M-PAR corresponding to PL shoaling, which was caused, in turn, by ML shoaling. M-Chl for F96 exceeded 0.5 mg m-3 for the first time in late March 2009 (single events of >0.4 mg m-3 occurred twice in mid-winter; Fig. 8c, event I), and remained above 0.4 mg m-3

until the middle of May (Fig. 8c, event III), following an increase in the M-PAR to 6-8 mol quanta m-2 d-1 in mid-March (Fig. 8a). M-Tur also increased at the same time as M-Chl (Fig. 8d, event III). For F97, the transition of M-Chl from the average winter level of ~0.3 mg m-3 to >0.4 mg m-3 was observed as three peaks from March to April 2009 (Fig. 9c, event IV), corresponding to a short-term increase in M-PAR, which had maximum values of 6.8-8.5 mol quanta m-2 d-1 (Fig. 9a). These increases in M-PAR were primarily caused by shoaling of the ML (Fig. 9b). M-Tur also responded positively to these events, but with smaller amplitudes (Fig. 9d, event IV).

The maximum M-Chl in 2009 occurred as a sharp peak in early April for F96 (Fig. 8c, event V), and in early May for F97 (Fig. 9c, event VI). By convention and the definition adopted in the present study, these can be regarded as spring blooms. Increased carbon fixation by these blooms was suggested by increases in Tur simultaneously with increases in M-Chl (Figs. 8d and 9d). The interval following the WES period until the occurrence of the spring bloom is referred to as the "period of transition" (TR). For F96, the maximum ¡-Chl also occurred simultaneously with the maximum M-Chl (Fig. 8c), because the concentration of phytoplankton was high within the relatively deep PL (Fig. 7a). The peak ¡-Tur also occurred at the same time, but with a smaller amplitude (Fig. 8d). For F97, ¡-Chl and ¡-Tur during the spring bloom were lower than the short-term peaks of ¡-Chl that occurred from March to April, before the spring bloom (Fig. 9c and d).

The responses of M-Chl to the peaks of M-PAR during WES and TR are directly compared in Fig. 10a and b. The increases in M-Chl corresponded to M-PAR values typically >4 mol quanta m-2 d-1, which was responsible for a significant positive correlation between M-PAR and the M-Chl during WES for F96 (r = 0.59, p < 0.01, n = 21, Fig. 10a) and during TR for F97 (r = 0.77, p < 0.01, n = 13, Fig. 10b). However, no significant relationship was observed in the data obtained during WES for F97 and during TR for F96. Although the time series of M-Tur also indicated moderate but positive responses to M-PAR (Figs. 8d and 9d), the relationship was not statistically significant (Fig. 10c and d).

In contrast, ¡-Chl and ¡-Tur showed marked positive linear relationships with PLD (Fig. 11). The correlation coefficients between PLD and

60 40 20

0 50 100 150 200 250

0.8 0.6 0.4 0.2 0

(a) PAR (F96)

SONDJ FMAMJ J

Fig. 8. Time series of (a) light intensity, (b) MLD and Z0415, (c) M-Chl and ¡-Chl, and (d) M-Tur and ¡-Tur obtained with float F96 from September 2008 to July 2009. Error bars in for M-Chl (c) and M-Tur (d) indicate the standard deviations of the values within the PL. The WES and TR periods, which are further analyzed in (Figs. 10 and 11, are shown with thick arrows). The peaks of M-Chl are indicated by gray vertical lines with Roman numerals that are referred to in the text.

¡-Chl were r = 0.96 (p < 0.01) during WES for F96 (Fig. 11a), r = 0.98 (p < 0.01, n = 22) during WES for F97 (Fig. 11a), and r = 0.95 (p < 0.01) during TR for F97 (Fig. 11b). Only the data for F96 during TR were not significant (r = 0.81, p = 0.10, n = 5). Similarly, the PLD and ¡-Tur displayed high correlation coefficients: r = 0.98 (p < 0.01) during WES for F96 (Fig. 11c); r = 0.91 (p = 0.03) during WES for F97 (Fig. 11c); r = 0.97 (p < 0.01) during TR for F96 (Fig. 11d); and r = 0.88 (p < 0.01) during TR for F97 (Fig. 11d).

I— I

0 50 100 150 200 250

0.8 0.6 0.4 0.2 0

40 6 =r 30 [

u 20 [

SONDJ FMAMJ J

Fig. 9. As for Fig. 8, but for float F97.

These results allow us to assess the cause of the variations in the phytoplankton biomass from winter to spring, and the applicability of the CDH and/or DRH to this region. To arrive at our conclusions, we first discuss possible biases in our profiling float observations conducted with an interval of 5 days (Section 4.1) and the degree of variability in the chlorophyll-to-carbon (Chl:C) ratio caused by variations in light intensity (Section 4.2). Then, we assess the causes of the increases in ¡-Chl with increasing MLD, based on the balance between gross production and losses of phytoplankton biomass (Section 4.3). Finally, we consider the validity of the CDH and DRH (Section 4.4).

4.1. Observations based on the profiling floats

4. Discussion and conclusions

Profiling float observations of chlorophyll a fluorescence in the Kuroshio-Oyashio Extension region revealed variability in the vertical profiles of Chl and Tur over time scales ranging from weekly to seasonal, and showed a link between the fluctuations in Chl and Tur and the PLD from winter to spring. We found a significant positive relationship between the PLD and ¡-Chl and ¡-Tur, with correlation coefficients > 0.9 during winter. The strong linearities indicate that the variations in M-Chl and M-Tur were relatively small, despite large fluctuations in the PLD, which predominantly reflected the deep MLD. However, the positive responses of M-Chl to shoaling of the PLD were evident in several cases, even during the WES period.

The two profiling floats used in the present study were designed to park at 40 dBar to follow currents. To acquire the vertical profile data, the floats first descended from 40 dBar to 500 dBar and then ascended to the surface, recording the data. They stayed at the surface until the data ofeach profile were transmitted to the satellite. Because ofthe profiling and surfacing for data transmission, there may have been some differences between the movements of the floats and those of the water parcels at 40 dBar, where the floats were parking. Although we do not have data to evaluate this, biases would increase in the analyses of the time series within and around the surface ML if the flow of the deep layers had a substantial effect. We assumed, however, that the biases caused by profiling and surfacing were minor because the time that the floats spent below 150 dBar is estimated to have been 5 h in every 5 days, at most. Displacements of F96 and F97 from the upper

S. ¡toh et al. / Journal of Marine Systems 151 (2015) 1-14 1

0246 M-PAR [mol quonta m-2 d-1]

(b) - rR

c ) A,

(c) WE :s o

C&c c o A

!i A L)

0 2 4 6 8 10 M-PAR [mol quonta m-2 d-1]

Fig. 10. Plots of M-PAR and M-Chl (a and b, respectively) and M-PAR and M-Tur (cand d, respectively). Those for the WES periods are shown in (a) and (c), and those for the TR periods are shown in (b) and (d). Regression lines indicate that the relationship is significant.

(b) TR u

(c) WES

(d) TR

0 50 100 150

PLD [m]

0 50 100 150

PLD [m]

Fig. 11. As for Fig. 10, but for PLD and ¡-Chl (a and b, respectively) and PLD and ¡-Tur (c and d, respectively).

layer water columns caused by this 5 h migration are estimated to be 2.4 ±1.1 km and 2.6 ±1.3 km during WES and TR, respectively, with an assumption that difference between flow velocities within and below the ML is 50% of the 5-day mean velocity of the float movement. Since most of these displacements might occur in directions along synoptic-scale currents with a typical spatial scale of O (100 km), the floats were likely to capture same upper layer water columns in a seasonal time scale. Although temporal fluctuations in temperature and salinity were observed below the ML (Fig. 4), our assumption that the floats followed the upper-layer water masses within the ML was not violated.

The discussion in the above paragraph, however, does not consider submesoscale variability along the current. As suggested by Mahadevan et al. (2012), small-scale features of O (1-10 km), called "mixed layer eddies" (ML eddies) with a shallow ML, are generated through slumping of horizontal density gradient, and make the spring bloom occur earlier than expected from the timing of ML shoaling after spring warming. The shoaling of the ML in the records of the floats could occur either through the interleaving of an upper layer water mass containing the floats, or the displacements of the floats from the upper layer water (with spatial scales of 1 -4 km, as mentioned above) mass caused by the profiling, which are further discussed in the following paragraphs.

The first case due to the interleaving is obviously not a one-dimensional process, but the floats kept capturing the upper layer water masses. Because the results of this study do not depend on the mechanism of ML shoaling but on the response of phytoplankton to fluctuations in the ML, the relationship between the PLD (and MLD) and Chl (and Tur) can be used to assess the response of phytoplankton to ML shoaling. An exception is the case that the increase in the surface Chl (and Tur) is caused by the advection of phytoplankton through the thin surface layer above the parking depth of the floats (40 dBar). The small difference in the M-Chl data for F96 and F97 during WES (0.33 ± 0.045 mg m-3 and 0.31 ± 0.026 mg m-3, respectively) suggests, however, that horizontal advection was not a primary factor controlling phytoplankton concentrations in this region. Furthermore, the increasing ¡-Chl and ¡-Tur with increasing PLD cannot simply be explained by eddy-driven stratification that was proposed by Mahadevan et al. (2012).

As in the second case, if ML eddies appeared or disappeared in the upper layer at the float positions (specified by the latitude and longitude) during when the floats were below the ML, observed variations in PLD, Chl and Tur in the upper layer were primarily attributed to differences in these properties inside and outside the ML eddies. More specifically, if spatial variations in PLD around the ML eddies were large but those of M-Chl (M-Tur) were moderate, profiling inside and outside the ML eddies could reproduce linear relationships between PLD and ¡-Chl (¡-Tur). Although we cannot exclude this possibility, it is not reasonable to attribute all observed PLD fluctuations solely to the submesoscale gradient. Moreover, nearly homogeneous phytoplankton concentration with a large PLD gradient within a narrow area, if there were (in this case with a horizontal scale of 1-10 km), is not trivial. Given phytoplankton communities were not isolated around the ML eddies, data sampled across the ML eddies could also be analyzed in the context of phytoplankton response to different environments.

The two profiling floats provided measurements of the variability in temperature, salinity, and Chl and Tur over time scales from weekly to seasonal, with profiling intervals of 5 days. However, as the generation time for phytoplankton is approximately 1 day, our sampling frequency was insufficient to directly investigate the growth rate of phytoplankton. Therefore, we did not conduct analyses for time derivatives of M-Chl, M-Tur, ¡-Chl, and ¡-Tur that would directly represent production and loss rates. Nevertheless, we assume that measured values of Chl and Tur at each instance reflected the rapid response of phytoplankton to physical fluctuations. Higher-frequency measurements will be conducted in future studies to clarify these rate processes.

42. Chlorophyll a concentration and phytoplankton biomass

As partly evident in Fig. 6, the phytoplankton Chl:C ratio changes in response to variations in light levels during growth optimization (photoacclimation). We had expected that Chl:C would even change during winter and spring in response to fluctuations in M-PAR, largely through the deepening and shoaling of the MLD. However, the positive responses of M-Tur to M-PAR were less evident than those of M-Chl (Fig. 10). Moreover, the correlation coefficients between PLD and ¡-Tur were as high as those between the PLD and ¡-Chl (Fig. 11). Because the measurement resolution of Tur was relatively low, less than approximately 0.1 NTU, some of the fluctuations in winter may not have been accurately detected. Therefore, to confirm the link between PLD and phytoplankton biomass, we used an empirical formula to estimate the influence of photoacclimation.

According to Cloern et al. (1995), the empirical formula (their Eq. (2)) for the Chl:C ratio is:

Chl: C = 0.003 + 0.0154 exp(0.05T) exp (-0.059 I)fif,

where T, I, and ff' are the temperature (°C), irradiance (mol quanta m-2-d-1), and normalized maximum growth rate as a function of nutrient concentration, respectively (f = 1 if the nutrient concentration is sufficiently high). Using this formula, the photoacclimation modeled as the carbon-based biomass within a deep ML was less than that expected for a constant Chl:C ratio. The modeling results (using Eq. (3)) revealed that the degree of change in the Chl:C ratio became increasingly important when the nutrient concentration was sufficient. Using M-PAR and the mean temperature within the ML, we obtained values for Chl:C of 0.003 + f' (0.033 ± 0.0027) for F96 and 0.003 + f (0.030 ± 0.0032) for F97; the values increased with decreasing M-PAR values. Although nutrient measurements were not made concurrently with the float profiling, the climatological mean winter surface nitrate concentration was 2-15 |jmol L-1 in the areas to which the floats migrated during the WES period (Fig. 12a). The minimum concentration in this range is more than twice that of the half-saturation constant (~1 ^mol L-1) estimated for the subarctic North Pacific Ocean (Kanda et al., 1985). This implies that nutrient levels did not strongly limit production during the WES period. Consequently, f' was assigned a value of 1, which caused the large variation in the Chl:C ratio. However, even assuming f' = 1, the CVs of the modeled Chl:C ratio during the WES period were 8.2% for F96 and 11% for F97, which do not explain the variation in ¡-Chl. The correlations between the MLD and the carbon-based integrated biomass in the float data were strong and statistically significant for both F96 (r = 0.92, p < 0.01) and F97 (r = 0.95, p < 0.01).

4.3. Balance between the production and loss terms

The results of this study for the Kuroshio-Oyashio Extension region are consistent with those of Behrenfeld (2010), and Boss and Behrenfeld (2010) for the subarctic North Atlantic, where it was found that the phytoplankton biomass within the ML increased with increasing MLD. We used equations for the mean and integrated biomass to assess whether the increase in the biomass was caused by grazer dilution, as hypothesized by Behrenfeld (2010), and considered whether variability in grazing pressure could control the winter fluctuations in Chl. However, we will find in the following that dilution alone could not increase ¡-Chl.

The variations in phytoplankton concentrations in winter were analyzed from a Lagrangian frame ofreference, which considered photosynthesis, loss effects from metabolism, grazing, and the dilution caused by the deepening of the MLD. A simple formula for the specific growth rate under light-limited conditions (without nutrient limitation) is:

1 dP APdh ak

Pdt= rM = - Phdt+ W-m-fiZ

(a) Surface nitrate concentration

°E 140°E 150°E 160°E 170°E 180°W 0 (b) Zonal-mean (150-160°E) nitrate concentration

[pmol L-1] 25 20 15 10 5 0

Latitude [°N]

Fig. 12. Winter (January to March) mean nitrate concentrations derived from the World Ocean Atlas 2009 (Garcia etal., 2010). (a) Mean surface concentration and (b) mean zonal-mean concentrations above 300 m, averaged over 150-160°E. The thick gray rectangles in (a) indicate that the area within which floats F96 and F97 moved from winter 2008 to spring 2009.

where P, AP, Z, h, I0, m, a, and 3 are the phytoplankton concentration, the difference in P between the ML and the layer below, the zooplankton concentration, the MLD, the light intensity (PAR) at the surface, the metabolic loss rate, the slope of the light-photosynthesis relationship normalized by the phytoplankton concentration, and the grazing rate, respectively. The term h was assumed to be deeper enough than 1/k (e-folding depth of light attenuation « MLD ~ PLD). We assumed that AP = P during deepening of the ML, when the phytoplankton-free lower water layer was entrained to the ML, and AP = 0 during ML shoaling, when the lower part of the ML, containing a homogeneous phytoplankton concentration, was detrained.

The equation for the integrated biomass is:

1 d(Ph) Ph dt

(P-AP) dh alo Ph dt + ~kh

The term (P - AP) / Ph x dh / dt equals zero when the MLD is deepening, but equals 1 / h dh / dt (<0) when the MLD is shoaling, which represents the cutting-off (detrainment) of the lower layer. For

a deepening ML, where initially rI = 0, h = h0, and Z = Z0: 0 - £-m-3z°- (6)

When the MLD increases to h1 = h0 + Ah, Eq. (6) becomes:

= Oc -m-№ - - ^m. (7)

kh1 h1 h1

Here, we assumed that the increase in h did not alter m, a, or 3, and that the grazer biomass did not change (it was simply diluted). Eq. (6) was substituted to obtain the right-hand side of Eq. (7). As in Eq. (7), based on the assumptions discussed above, the net growth during the deepening of the ML was not likely to become positive, even though grazing pressure was reduced, because photosynthesis similarly decreased at a rate inversely proportional to h. As a consequence of this reduced photosynthesis and grazing, the relative importance of the MLD-independent loss rate m increased and caused net negative growth. It is noted that the above inversely proportional relationship

between the specific photosynthetic rate and h also holds for nonlinear light-photosynthesis relationships, if h was deeper enough than 1/k.

Because an increasing ML generally causes decreasing temperatures, the metabolic loss rate m is likely to decrease. However, our observations during the WES period revealed relatively small temperature variations: the mean and standard deviation during this period were 162 ± 1.1 °C for F96 and 14.4 ± 1.8 °C for F97. Based on the Q10 = 2re-lationship (i.e., a twofold increase in the rate of change with every 10 °C increase in temperature), m was set at 0.96-1.03 (96-103%) for F96 and 0.89-1.10 (89-110%) for F97, which were smaller than the MLD variations (30-70%).

Although the climatological surface nitrate concentrations in the areas covered by the float migration (Fig. 12a) were not lower than the estimated half-saturation constant (Kanda et al., 1985), the deepening of the ML may have caused the nutrient concentrations within the ML to increase through the entrainment of nutrient-rich water from the lower layer. It is possible that the increased nutrient concentrations enhanced photosynthesis, even if the nutrient concentrations were relatively high (typically above the half-saturation constant). However, we assumed that this effect was minor and did not explain the significant linear relationship between PLD and I-Chl. The climatological nutrient concentrations indicated that the vertical gradient in the nitrate concentrations in the upper layer (above ~150 m) was weak during winter (Fig. 12b). Thus, the supply of nitrate from short-term variations in the MLD during winter below and around this level were probably insignificant.

The results of the present study do not simply indicate that rI > 0 during the deepening of the ML, but also show a linear relationship between the PLD and I-Chl. This is equivalent to a nearly constant M-Chl, which requires that rM is approximately equal to 0 (Eq. (4)). Because this does not hold in all instances, we expect more general negative feedback for various disturbances of P; for example, an abrupt reduction caused by the entrainment of phytoplankton-free water (dilution), or an instantaneous increase caused by an abrupt increase in light intensity. However, a solution satisfying this requirement cannot be obtained from Eq. (4) if only grazer dilution is considered. With respect to the hypothesis of Behrenfeld (2010), we also observed an increase in biomass with increasing MLD, which should have been caused by the effect of reduced loss, but grazer dilution alone cannot explain this relationship.

Although an analogy to the dilution experiment in a bottle (Landry and Hassett, 1982) was proposed by Behrenfeld (2010), this may not be appropriate for the ocean. The deepening of the MLD causes grazer dilution, but in contrast to what occurs in the bottle, it also causes a reduction in the mean light intensity.

The general formula for P stabilization (negative feedback) requires that drM / dP < 0 for rM defined in Eq. (4). In addition to the grazer dilution that occurs during the deepening of the ML, which satisfies -dZ / dP < 0, the following can also apply (see Eq. (4)): da / dP < 0 for photoacclimation (in response to the deepening of the ML that instantaneously reduces P) and the self-shading of phytoplankton; dk/dP < 0 and -dm / dP < 0 for coagulation; and -dj3 / dP > 0 for the Holling type III functional response of grazers (Holling, 1966). Self-shading and coagulation are not usually likely to contribute to dr/dP becoming negative in winter, because P is relatively low. Although our data suggest that photoacclimation might alter the Chl-C ratio by approximately 10% in response to ML variations from winter to early spring, this effect alone cannot account for the stability of P. If grazers follow the Holling Type-III response, so that they reduce their attack rate with decreasing prey concentrations, this will become significant in conditions of low prey availability.

We assume that the results of the numerical experiments by Yoshie etal. (2003) and Behrenfeld etal. (2013a) reproduced the positive relationship between MLD and phytoplankton biomass because the grazers were diluted by ML deepening, but also because the models they used employed the Holling Type-III response. The NEMURO model used by Yoshie et al. (2003) assumes a grazing threshold (the minimum phyto-plankton concentration below which zooplankton grazing does not occur) (Kishi et al., 2007), and the CCSM-BEC model used by Behrenfeld et al. (2013a) uses a sigmoidal equation for grazing (Moore et al., 2004). However, because we have no evidence yet that this occurs in the Kuroshio-Oyashio Extension region, further observations and experiments are required to evaluate this possibility.

4.4. CDHorDRH

Based on our observations, the phytoplankton biomass represented by ¡-Chl increased before ML shoaling in spring in the Kuroshio-Oyashio Extension region. We agree with the main point of Behrenfeld (2010) that a reduction in loss is expected to play a major role in this phenomenon, which differs from the premise of the CDH that an increase in photosynthesis causes net growth. Nevertheless, the CDH should not be abandoned on this point alone. Because CDH is a hypothesis for the timing of the initiation of the spring bloom, in particular the rapid increase in the phytoplankton concentration that usually occurs in spring, this hypothesis should not be rejected completely unless it can be confirmed that increasing photosynthesis in spring is not the primary factor controlling the increase in net growth. Furthermore, our results do not entirely support the DRH, because dilution alone is unlikely to increase the integrated biomass.

As noted above, the linear relationship between the PLD and ¡-Chl or ¡-Tur during the WES period indicates that there are only small variations in M-Chl and M-Tur compared with the variations in MLD, which is thought to be stabilized by grazing pressure. Whereas, an increase in M-Chl in response to the increase in M-PAR (reduction in the MLD) was observed, typically when M-PAR increased beyond 4 mol quanta m-2 d-1. These events occurred mainly during TR, but on three occasions (two for F96 and one for F97), similar events were observed during the WES period (Figs. 8-9). We assume that negative feedback is dominant only if gross production is low. The two shoaling events for F96 during the WES period could not be detected using the conventional MLD criterion of Aae = 0.125 kg m-3. Therefore, the early bloom in surface Chl detected by Behrenfeld (2010) (besides the increase in integrated biomass in mid-winter) was possibly related to their criterion for calculating the MLD (Act0 = 0.125 kg m-3).

It has been suggested that grazing pressure controls the phytoplankton biomass from winter to early spring, whereas increased photosynthesis increases the phytoplankton concentration in both spring and winter. How then do these two effects contribute to the spring bloom? The complexity of this issue is related to an implicit, but not necessarily correct, assumption that the phytoplankton biomass increases consistently from winter to spring. Behrenfeld (2010) and Boss and Behrenfeld (2010) drew a distinction in the definition of net growth rate between the periods of MLD deepening and shoaling. The use of concentration (mg m-3 d-1) rather than integrated biomass (mg m-2 d-1) in the shoaling phase is partly justified (Behrenfeld et al., 2013b), because a cut-off lower layer generally does not contribute to the accumulation of biomass. However, these two rates represent different quantities (Chiswell, 2013). In the present study, ¡-Chl and ¡-Tur generally increased with increasing MLD, so the shoaling of the MLD in spring did not lead to a continuous increase in ¡-Chl or ¡-Tur. Therefore, it is inappropriate to simply regard the increase in biomass during winter as the initiation of the spring bloom, as Behrenfeld (2010) does.

Although the production-driven (e.g., the CDH and critical turbulence hypothesis) and loss-driven hypotheses (e.g., the DRH and the Holling Type-III response) are fundamentally different theories, the present study suggests that no single hypothesis exclusively explains the period from winter to spring in the Kuroshio-Oyashio Extension region. The loss term should contribute to the rapidity and magnitude of blooms, even during the late spring. Changes in community structure in response to light levels, for both phytoplankton and zooplankton, might occur during TR before the spring bloom, in which case it will be necessary to move beyond Sverdrup's CDH.

Acknowledgments

The authors thank members of the Fisheries Environmental Oceanography Group of the Atmosphere and Ocean Research Institute, The University of Tokyo, for their help in acquiring the profiling float data. This research was supported by the Population Outbreak of Marine Life (POMAL) program of the Agriculture, Forestry and Fisheries Research Council of Japan, and by the Ministry of Education, Science, Sports and Culture of Japan via Grants-in-Aid for Scientific Research (KAKENHI) (S) 20221002, (B) 23310002, (B) 26929099 and Innovative Areas 24121002.

References

Behrenfeld, M.J., 2010. Abandoning Sverdrup's critical depth hypothesis on phytoplankton blooms. Ecology 91, 977-989. Behrenfeld, M.J., Boss, E.S., 2014. Resurrecting the ecological underpinnings of ocean

plankton blooms. Ann. Rev. Mar. Sci. 6,167-194. Behrenfeld, M.J., Doney, S.C., Lima, I., Boss, E.S., Siegel, DA, 2013a. Annual cycles of ecological disturbance and recovery underlying the subarctic Atlantic spring plankton bloom. Glob. Biogeochem. Cycles 27,526-540. Behrenfeld, M.J., Doney, S.C., Lima, I., Boss, E.S., Siegel, DA, 2013b. Reply to a comment by Stephen M. Chiswell on: "Annual cycles of ecological disturbance and recovery underlying the subarctic Atlantic spring plankton bloom" by M. J. Behrenfeld et al. (2013). Glob. Biogeochem. Cycles 27,1294-1296. Boss, E., Behrenfeld, M., 2010. In situ evaluation of the initiation of the North Atlantic phytoplankton bloom. Geophys. Res. Lett. 37, L18603. http://dx.doi.org/10.1029/ 2010GL044174.

Boss, E., Swift, D., Taylor, L., Brickley, P., Zaneveld, R., Riser, S., Perry, M.J., Strutton, P.G., 2008. Observations of pigment and particle distributions in the western North Atlantic from an autonomous float and ocean color satellite. Limnol. Oceanogr. 53,2112-2122. Chiswell, S.M., 2011. Annual cycles and spring blooms in phytoplankton: don't abandon

Sverdrup completely. Mar. Ecol. Prog. Ser. 443,39-50. Chiswell, S.M., 2013. Comment on "Annual cycles of ecological disturbance and recovery underlying the subarctic Atlantic spring plankton bloom". Glob. Biogeochem. Cycles 27,1291-1293.

Cloern, J.E., Grenz, C., VidergarLucas, L., 1995. An empirical model of the phytoplankton chlorophyll:carbon ratio — the conversion factor between productivity and growth rate. Limnol. Oceanogr. 40,1313-1321. de Boyer Montegut, C., Madec, G., Fischer, A.S., Lazar, A., Iudicone, D., 2004. Mixed layer depth over the global ocean: an examination of profile data and a profile-based climatology. J. Geophys. Res. Oceans 109.

Favorite, F., Dodimead, A.J., Nasu, K., 1976. Oceanography of the Subarctic Pacific region, 1960-1971. Bull. Int. North Pac. Fish. Comm. 33,1-187.

Garcia, H.E., Locarnini, R.A., Boyer, T.P., Antonov,J.I., Zweng, M.M., Baranova, O.K., Johnson, D.R., 2010. World Ocean Atlas 2009. In: Levitus, S. (Ed.), NOAA Atlas NESDIS. U.S. Government Printing Office, Washington, D.C.

George, J.A., Lonsdale, D.J., Merlo, L.R., Gobler, C.J., 2015. The interactive roles of temperature, nutrients, and zooplankton grazing in controlling the winter-spring phytoplankton bloom in a temperate, coastal ecosystem, Long Island Sound. Limnol. Oceanogr. 60,110-126.

Holling, C.S., 1966. The functional response of invertebrate predators to prey density. Mem. Entom. Soc. Can. 98, 5-86.

Huisman, J., van Oostveen, P., Weissing, F.J., 1999. Critical depth and critical turbulence: two different mechanisms for the development of phytoplankton blooms. Limnol. Oceanogr. 44,1781-1787.

Itoh, S., Yasuda, I., Nishikawa, H., Sasaki, H., Sasai, Y., 2009. Transport and environmental temperature variability of eggs and larvae of the Japanese anchovy (Engraulis japonicus) and Japanese sardine (Sardinops melanostictus) in the western North Pacific estimated via numerical particle-tracking experiments. Fish Oceanogr. 18 (2), 118-133. http://dx.doi.org/10.1111/j.1365-2419.2009.00501.x.

Kanda, J., Saino, T., Hattori, A., 1985. Nitrogen uptake by natural-populations of phytoplankton and primary production in the Pacific-ocean — regional variability of uptake capacity. Limnol. Oceanogr. 30, 987-999.

Kawai, H., 1969. Statistical estimation of isotherms indicative of the Kuroshio axis. Deep-SeaRes. 16,109-115.

Kishi, M.J., Kashiwai, M., Ware, D.M., Megrey, B.A., Eslinger, D.L., Werner, F.E., Noguchi-Aita, M., Azumaya, T., Fujii, M., Hashimoto, S., Huang, D.J., Iizumi, H., Ishida, Y., Kang, S., Kantakov, G.A., Kim, H.C., Komatsu, K., Navrotsky, V.V., Smith, S.L., Tadokoro, K., Tsuda, A., Yamamura, O., Yamanaka, Y., Yokouchi, K., Yoshie, N., Zhang, J., Zuenko, Y.I., Zvalinsky, V.I., 2007. NEMURO — a lower trophic level model for the North Pacific marine ecosystem. Ecol. Model. 202, 12-25.

Kraus, E.B., Turner, J.S., 1967. A one-dimensional model of seasonal thermocline II. The general theory and its consequences. Tellus 19, 98-106.

Landry, M.R., Hassett, R.P., 1982. Estimating the grazing impact of marine microzooplankton. Mar. Biol. 67, 283-288.

Lavigne, H., D'Ortenzio, F., Migon, C., Claustre, H., Testor, P., d'Alcala, M.R., Lavezza, R., Houpert, L., Prieur, L., 2013. Enhancing the comprehension of mixed layer depth control on the Mediterranean phytoplankton phenology. J. Geophys. Res. Oceans 118, 3416-3430.

Levitus, S., 1983. Climatological atlas of the world ocean. Eos. Trans. AGU 64, 962-963.

Mahadevan, A., D'Asaro, E., Lee, C., Perry, M.J., 2012. Eddy-driven stratification initiates north Atlantic spring phytoplankton blooms. Science 337, 54-58.

Marra, J., 1997. Analysis of diel variability in chlorophyll fluorescence. J. Mar. Res. 55, 767-784.

Marra, J., Barber, R.T., 2005. Primary productivity in the Arabian Sea: a synthesis of JGOFS data. Prog. Oceanogr. 65,159-175.

Matsumoto, K., Honda, M.C., Sasaoka, K., Wakita, M., Kawakami, H., Watanabe, S., 2014. Seasonal variability of primary production and phytoplankton biomass in the

western Pacific subarctic gyre: control by light availability within the mixed layer. J. Geophys. Res. Oceans 119,6523-6534.

Moore, J.K., Doney, S.C., Lindsay, K., 2004. Upper ocean ecosystem dynamics and iron cycling in a global three-dimensional model. Glob. Biogeochem. Cycles 18, GB4028. http://dx.doi.org/10.1029/2004GB002220.

Morel, A., Huot, Y., Gentili, B., Werdell, P.J., Hooker, S.B., Franz, BA, 2007. Examining the consistency of products derived from various ocean color sensors in open ocean (case 1 ) waters in the perspective of a multi-sensor approach. Remote Sens. Environ. 111,69-88.

Nelson, D.M., Smith, W.O., 1991. Sverdrup revisited — critical depths, maximum chlorophyll levels, and the control of southern-ocean productivity by the irradiance-mixing regime. Limnol. Oceanogr. 36,1650-1661.

Nishikawa, H., Yasuda, I., Itoh, S., 2011. Impact of winter-to-spring environmental variability along the Kuroshio jet on the recruitment of Japanese sardine (Sardinops melanostictus). Fish Oceanogr. 20 (6), 570-582. http://dx.doi.org/10.1111Zj.1365-2419.2011.00603.x.

Obata, A., Ishizaka, J., Endoh, M., 1996. Global verification of critical depth theory for phy-toplankton bloom with climatological in situ temperature and satellite ocean color data. J. Geophys. Res. Oceans 101, 20657-20667.

Shiozaki, T., Ito, S.-I., Takahashi, K., Saito, H., Nagata, T., Furuya, K., 2014. Regional variability of factors controlling the onset timing and magnitude of spring algal blooms in the northwestern North Pacific. J. Geophys. Res. Oceans 119, 253-265.

Siegel, D.A., Doney, S.C., Yoder, J.A., 2002. The North Atlantic spring phytoplankton bloom and Sverdrup's critical depth hypothesis. Science 296, 730-733.

Smetacek, V., Passow, U., 1990. Spring Bloom Initiation and Sverdrup Critical-Depth Model. Limnol. Oceanogr. 35 (1), 228-234.

Stramska, M., Dickey, T.D., 1993. Phytoplankton bloom and the vertical thermal structure of the upper ocean. J. Mar. Res. 51 (4), 819-842. http://dx.doi.org/10.1357/ 0022240933223918.

Suga, T., Motoki, K, Aoki, Y., Macdonald, A.M., 2004. The North Pacific climatology of winter mixed layer and mode waters. J. Phys. Oceanogr. 34, 3-22.

Suzuki, K., Kuwata, A., Yoshie, N., Shibata, A., Kawanobe, K., Saito, H., 2011. Population dynamics of phytoplankton, heterotrophic bacteria, and viruses during the spring bloom in the western subarctic Pacific. Deep Sea Res., Part I 58, 575-589.

Sverdrup, H., 1953. On conditions for the vernalblooming of phytoplankton. J. Cons. Int. Explor. Mer 18, 287-295.

Taylor, J.R., Ferrari, R., 2011. Shutdown of turbulent convection as a new criterion for the onset of spring phytoplankton blooms. Limnol. Oceanogr. 56, 2293-2307.

Townsend, D.W., Keller, M.D., Sieracki, M.E., Ackleson, S.G., 1992. Spring phytoplankton blooms in the absence of vertical water column stratification. Nature 360, 59-62.

Yasuda, I., Watanabe, T., 2007. Chlorophyll a variation in the Kuroshio Extension revealed with a mixed-layer tracking float: implication on the long-term change of Pacific saury (Cololabis saira). Fish. Oceanogr. 16,482-488.

Yoshie, N., Yamanaka, Y., Kishi, M.J., Saito, H., 2003. One dimensional ecosystem model simulation of the effects of vertical dilution by the winter mixing on the spring diatom bloom. J. Oceanogr. 59, 563-571.