Observing intermittent biological productivity and vertical carbon transports during the spring transition with BGC Argo floats in the western North Pacific

Abstract. To investigate changes in ocean structure during the spring transition and responses of biological activity, two BGC-Argo floats equipped with oxygen, fluorescence (to estimate chlorophyll a concentration – Chl a), backscatter (to estimate particulate organic carbon concentration – [POC]), and nitrate sensors conducted daily vertical profiles of the water column from a depth of 2000 m to the sea surface in the western North Pacific from January to April of 2018. Data for calibrating each sensor were obtained via shipboard sampling that occurred when the floats were deployed and recovered. During the float-deployment periods, repeated meteorological disturbances passed over the study area and caused the mixed layer to deepen. After deep-mixing events, the upper layer restratified and nitrate concentrations decreased while Chl a and POC concentrations increased, suggesting that spring mixing events promote primary productivity through the temporary alleviation of nutrient and light limitation. At the end of March, POC accumulation rates and nitrate decrease rates within the euphotic zone (0–70 m) were the largest of the four events observed, ranging from +84 to +210 mmol C m−2 d−1 and –28 to –49 mmol N m−2 d−1, respectively. The subsurface consumption rate of oxygen (i.e., the degradation rate of organic matter) after the fourth event (the end of March) was estimated to be –0.62 micromol O2 kg−1 d−1. At depths of 300–400 m (below the mixed layer), the POC concentrations increased slightly throughout the observation period. The POC flux at a depth of 300 m was estimated to be 1.1 mmol C m−2 d−1. Our float observation has made it possible to observed biogeochemical parameters, which previously could only be estimated by shipboard observation and experiments, in the field and in real time.


subtropical mode water (STMW) and is known to develop a deep mixed layer (300-400 m) during late winter (Hanawa and Tally, 2001;Qiu et al., 2006Qiu et al., , 2007. Moreover, formation of a deep mixed layer entrains nutrients from below the base of the seasonal pycnocline that support active primary production in the early spring as the mixed layer sporadically shoals and deepens, allowing for episodic particle subduction by the MLP. Although persistent seasonal shoaling of the mixed layer is caused by solar 85 radiation warming the ocean surface, Mahadevan et al. (2012) have shown that rapid restratification events in ocean frontal regions can also occur during the formation of mixed layer eddies. As a result, it is important to simultaneously analyse the physical and biogeochemical ocean conditions during the spring transition to achieve proper attribution of the processes driving carbon export.
Our field study aimed to observe the temporal transition in biological production from a well-mixed 90 winter water column to a warm, stratified spring water column that experiences high-light and nutrientreplete conditions. This period of the year is associated with the highest annual primary productivity rates ) as well as elevated particle fluxes from the euphotic layer to depth (Honda et al., 2016). Using chemical and biological measurements from daily-profiling BGC Argo floats, we isolated times when local physical processes were largely one dimensional to study changes in biological 95 and chemical parameters that occurred during rapid transitions from deep mixing to intermittent stratification. We quantified the production of POC in the euphotic zone and its fate in the mesopelagic to evaluate how these short periods of productivity between episodes of deep mixing contributed to vertical carbon exchanges during the spring transition.  Metrohm, Japan) and used to calibrate an SBE43 oxygen sensor. The water samples for measurements of nitrate concentrations were immediately frozen at −20 °C and kept in a freezer until the nutrient 140 concentrations were measured with an autoanalyzer (QuAAtro: BL Tec, Japan). The nitrate was reduced to nitrite using a cadmium-copper column, and the nitrite concentration was determined based on the absorption of an azo dye at a wavelength of 550 nm (Grasshoff, 1976).
In the case of DO, the sensor output values from floats were first converted to oxygen concentrations following standard protocols outlined in Thierry et al. (2018) and using adjusted salinity data based on 145 Wong et al. (2020). Oxygen concentrations from each float's first profile were compared with CTD observations made at the time of the float deployments. A linear regression between float and CTD values from a region of the water column in which the oxygen gradient was less than 0.2 µmol O2 kg −1 dbar −1 was used to determine the gain of the float sensor (Takeshita et al., 2013). We used the difference in oxygen concentration between the last float profile and the CTD profile at the time of float recovery to 150 determine the drift correction for each 0.01 kg m −3 of potential density (sq) larger than 27.5 sq. The gaincorrected float oxygen data were then adjusted at each time step by applying the average drift rate to the entire profile over time.
For nitrate, the sensor output values were converted to nitrate concentrations following standard protocols outlined in Johnson et al. (2016). The optical wavelength offset, which accounts for the 155 uncertainty in the wavelength registration of the diode array spectrometer, was adjusted to minimize the difference between the discrete and sensor nitrate concentration from the deployment cast. Corrected nitrate concentrations were calculated using the new optical wavelength offset. Subsequent data quality https://doi.org/10.5194/bg-2022-9 Preprint. Discussion started: 13 January 2022 c Author(s) 2022. CC BY 4.0 License.
control was implemented using the SAGE software of the Monterey Bay Aquarium and Research Institute (Maurer et al., 2021following Johnson et al., 2017. We used the empirical relationship between Chl a concentration and fluorescence based on data from 175 both cruises to estimate the Chl a concentration for each daily float profile: Chl a (mg m −3 ) = fluorescence value × 0.952 + 0.014, r 2 = 0.88. (1)

180
Water samples for POC were also collected at the float deployment and recovery stations. Seawater ( We used the empirical relationship between POC concentration and bbp based on data from both cruises to estimate POC concentrations for each daily float backscatter profile: As with the Chl a concentrations, data in the surface 100 m at the recovery stations were not used for the estimation of POC concentrations. The ranges of the POC concentrations and bbp values for equation (2) were 1-4 mmol C m −3 (15-48 mg C m −3 ) and 2.1-5.9 × 10 −4 . Cetinić et al. (2012) respectively. The coefficients in these equations are 30-40 % smaller than the corresponding coefficients calculated in this study. One possible reason for this difference may be that the prior studies focused on subarctic and polar regions inhabited by large phytoplankton, whereas our study was in a relatively oligotrophic subtropical area dominated by small phytoplankton (Fujiki et al., 2015). Small particles are 215 known to be associated with lower bbp index values (Morel and Ahn, 1991) and therefore tend to yield higher coefficients in the POC-bbp relationship.

Auxiliary data sets
We used daily mean global eddy-resolving ocean reanalysis data (1/12° horizontal resolution, Here, ∆ is the difference between two daily profiles and % | &'" is the surface heat flux due to the airsea heat exchange.
To consider the air-sea heat flux, we interpolated hourly float positions between daily positions and 250 then interpolated the ERA5 hourly reanalysis data to those positions. The interpolated hourly heat fluxes were integrated over 24 h and compared with the daily change of heat content.

255
In this study, we focused on the changes in biological and chemical parameters that occurred during periods of deep mixing and rapid stratification and the times when stratification persisted for several days.
In early February, a week or so of strong winds was followed by a few calm days. After mid-February, most of the calm intervals between periods of strong winds exceeded a few days (Fig. 4a). The primary period of mixed layer shoaling occurred during one day when it decreased from >200 m to <70 m ( depth was defined to be the shallowest depth at which the density was 0.0125 kg m −3 greater than the density at the sea surface (defined to be the average density over the shallowest observed layer to at most 10 dbar; de Boyer Montégut et al. 2004). Four stratification events were detected for each float (Table 1).

265
Because the density changed significantly before the end of events 3 and 4 observed by float 2903329 and event 4 observed by float 2903330, the end date was determined from the change in the density profile.
The average temperature of the water column from the surface to the 400-dbar level (Fig. 4e) increased until 10 March, decreased suddenly just before 11 March, and then started to increase slowly thereafter.
Because surface heat fluxes measured in February and March showed cooling of the ocean surface, we 270 hypothesized that the observed increases were likely caused either by lateral advection of warm water or by lateral movement of the floats. The fact that the daily change of heat content was 10 times (or even more) the time-integrated net heat flux (Fig. 4f) underscored the importance of changes due to horizontal temperature gradients. Our results suggested that among the post-storm events that we identified, Case 4 was relatively close to a one-dimensional exchange of heat in the upper 400 dbar of the water column.  Table 1). In April, the mixed layer shoaled in response to atmospheric conditions, and the positions of the two floats approached the Kuroshio Extension, which caused a rise of isopycnal surfaces at depths of 300-400 m. oxygen via respiration. The vertical profiles of DO from the surface to 400 m showed a temporal change similar to that of density until mid-March (Fig. 6ab). Oxygen was transported to deeper layers with the gradual deepening of the mixed layer during this period. In mid-to-late March, the DO concentration near the surface increased a few days after the deepening of the mixed layer. Atmospheric cooling and lower water temperatures may have caused the DO concentration to increase due to increased solubility and gas 320 exchange. Later, the surface mixed layer was stratified, and the water temperature began to increase, but the DO concentration remained high, perhaps because of photosynthetic production of oxygen in the euphotic zone.

Temporal variations of biogeochemical parameters
Examination of the time-series profiles of DO percent saturation ( Fig. 6cd) revealed changes similar to those of the DO concentrations. By mid-March, the DO was slightly undersaturated in the mixed layer.

325
This undersaturation was likely due to the redistribution of low-DO water from depth into the mixed layer caused by deep mixing. Intermittent DO supersaturation was observed near the surface in late March.
These temporal variations of DO saturation imply that photosynthetic production of oxygen actively occurred or that water temperature increased during this period. At depths of 300-400 m, the DO concentration and percent saturation tended to decrease with time in April ( Fig. 6a-d) because the 330 isopycnal surface shoaled, and a water mass with a low oxygen concentration was lifted as the Kuroshio Extension was approached.

Chl a
The concentrations of Chl a ( Deep mixing increased nutrient concentrations in the euphotic zone, and the surface water was stratified for a few weeks. These conditions suggest that phytoplankton received a stable supply of light, which led 345 to a high rate of primary production. However, after the floats approached the Kuroshio Extension in April, Chl a values below the euphotic zone were lower than before. The implication is that the vertical distribution of Chl a from winter to spring in the deep mixing area was greatly influenced by mixing in the water column.

POC
The temporal variation of the POC profiles ( Fig. 6ij) was similar to that of the Chl a concentrations (Fig.   6ef), and a similar trend was also observed for the distribution of oxygen percent saturation (Fig. 6cd).
The POC concentrations in the euphotic zone exceeded 3 µmol C kg −1 throughout the observation period; between 30 and 130 g C g −1 Chl a throughout the observation period (Fig. 7ef). These values are within the range of previously reported C/Chl a ratios in the North Pacific (30-190 g C g −1 Chl a, Behrenfeld et al., 2005). for float 2903330 during this study and increased slightly throughout the observation period (Fig. 8ab).
The vertical distribution of POC (Fig. 7ij) indicated that the POC concentration was low in the highdensity water that shoaled at depths of 300-400 m in April. The

Case studies: biogeochemical processes during the transitions from mixing to restratification
We observed biological activity in the surface mixed layer associated with intermittent mixing and restratification events from winter to spring of 2018. The Chl a concentration in the euphotic zone 390 increased after the deepening and subsequent shoaling of the mixed layer (Fig. 7ab). This increase suggests that vertical mixing in this oligotrophic region transports nutrients (such as nitrate) from the aphotic zone to the surface and thereby increases primary production (reflected by increases of Chl a and https://doi.org/10.5194/bg-2022-9 Preprint. Discussion started: 13 January 2022 c Author(s) 2022. CC BY 4.0 License.
POC concentrations) when the mixing ceases (Carranza et al., 2018). We analyzed temporal variations of water mass structure and biogeochemical parameters in the euphotic zone and the mixed layer during 395 selected events (Table 1) to reveal the relationship between surface disturbances and biological activities.
Here, the temporal variation (rate) of each parameter is shown with + for increases and -for decreases. Concentrations of Chl a and POC during this period increased slightly, whereas the concentration of 405 nitrate decreased slightly (Fig. 7acg). The POC/Chl a ratio, which is a metric for the fraction of particles comprised of living phytoplankton, decreased (Fig. 7e, Table 2). The rates of POC increase and nitrate decrease were +0.28 µmol C kg −1 d −1 and -0.08 µmol N kg −1 d −1 , respectively, based on changes of these concentrations in the euphotic zone from the last day of the mixing period to the last day of the stratified period. The rate of increase of organic carbon, estimated from the rate of decrease of the nitrate 410 concentration and the Redfield C/N ratio (6.6 by atoms), was +0.53 µmolC kg −1 d −1 , almost twice as large as the observed rate of increase of the POC. Possible causes of this are that some of the produced organic matter was in the form of dissolved organic carbon (DOC), that zooplankton grazed phytoplankton, and/or that part of the POC sank immediately into the deeper layer (>70 m) after being produced.
For float 2903330 (7-9 February), deepening and cooling of the mixed layer after the storm on 5

415
February continued until 7 February (Fig. 9c). Then, during 8-9 February, the mixed layer shoaled and warmed. During 8-11 February, a warm front within the mixed layer was identified near the western end of a SeaGlider transect across the float trajectory (Figs. 9a and 10). We therefore concluded that the warmer water encountered by the floats as they drifted westward after deployment represented a different water mass. During this period, concentrations of Chl a and POC increased, whereas nitrate concentrations decreased (Fig. 7bdh). In contrast, the POC/Chl a ratio did not change (Fig. 7f). A comparison of rates of POC increase and nitrate decrease revealed that the latter exceeded the former. In this case, the C/N ratio may have differed significantly from the Redfield ratio because different surface water masses were The deep mixing continued for several days until 8 March, and the restratification period lasted for about 450 one week starting on 9 March. While the floats moved westward roughly along the SSHA contour during this period (Fig. 12ab), a cyclonic feature moved northward through the area, and the floats recorded a doming isotherm (Fig. 5) below the mixed layer (Fig. 12c). The isopycnal surfaces at depths of 200-400 m became uniformly shallower. It can be inferred from the time series of nitrate profiles (Fig. 6gh) that eddy upwelling supplied nutrients from the twilight layer to the euphotic zone. The doming isotherm was 455 also accompanied by salinity intrusions (Fig. 12d), which disappeared when the floats left the doming structure. This mesoscale doming and the accompanying salinity intrusions indicates the importance of lateral processes associated with a cyclonic feature.
During this period, the rate of increase of POC was +0.4 µmol C kg −1 d −1 (Table 2, Fig. 7cd). In contrast, the nitrate concentrations recorded by float 2903330 barely decreased, whereas those recorded by float 460 2903329 increased (Fig. 7gh). This difference was caused by the fact that the floats moved from the edge of the cyclonic eddy toward the center, so that isopycnal surfaces and the nitracline shoaled. The POC/Chl a ratios were high at the edge of the eddy and decreased toward the center (Fig. 7ef).

465
The northward movement of the floats during the time interval of Case 4 was related to the cyclonic structure of the large-scale circulation (Fig. 13ab). We observed the deepest mixed layer (~320 dbar) during this period, and potential temperature in the water column within the mixed layer was relatively uniform and then gradually stratified (Fig. 13cd). Case 4 was associated with the largest increases of Chl a and POC among the four events ( Fig. 7a-d). The increase of the POC might have been due to the post-470 storm formation of submesoscale cyclonic eddies in the mixed layer (Fig. 14), which trapped nutrients and phytoplankton near the surface during the relatively calm, warm weather. The deepening of the mixed layer ended on 25 March, and the restratification period lasted for 6 days from 26 March (Table 1).
It was determined that Case 4 ended on 28 and 27 March for float 2903329 and 2903330, respectively because these days corresponded to the maximum Chl a concentration after the storm (Fig. 7ab). The one-  . The fact that the satellite images also showed that the Chl a concentrations increased by factors of 3-5 (Fig. 14) suggests that the maximum rate of primary production in the study area during springtime can be greater than previously reported, at least temporarily.

Effects of mesoscale and submesoscale features
It is known that moving cyclonic eddies promote local upwelling (McGillicuddy et al., 1998). In We calculated the above variables by using ocean reanalysis data (described in Section 2.2) that did not include vertical velocity ( ) fields. Here, ( ) and ( ) are zonal and meridional velocities (directions), 500 respectively. We examined those variables in March, when cyclonic features and high surface concentrations of Chl a (e.g., Fig. 14c) were observed during Cases 3 and 4.
The Q vector at a depth of 380 m (below the mixed layer) indicated that the density field there was squeezed in early March when upwelling was observed at the edge of cyclonic features (Fig. 15ad). From the middle of March, the density field became relaxed, and the distribution of upwelling became patchy 505 (Fig. 15be). By 21 March, the mesoscale changes of the density field and upwelling had slowed (Fig.   15cf). We hypothesized that the patchy upwelling associated with mesoscale fluctuations of the density field in early and middle March increased local nutrient concentrations below the mixed layer. Then, during the sporadic storms toward the end of March, nutrients were entrained into the deepening mixed layer, and restratification due to mixed layer eddies created high surface concentrations of Chl a. respectively (e.g., Tandon and Garrett, 1995;Wenegrat et al., 2018). Here, " is surface buoyancy flux, which is approximated as " ≈ − 6 ! " ⁄ , (positive for cooling), is the mixed layer depth, and is the Coriolis frequency. For 789 -⁄ , the inertial period was used as the time scale. We set to 200 dbar to exclude the effects of changes in the main thermocline. During Case 4 (Fig. 16), when the two floats observed a similar water mass, we hypothesized that rapid restratification in the mixed layer could 520 have been caused by geostrophic adjustment. We speculated that the subsequent formation of ML eddies was also important for the restratification as well as biological activities in Case 4 because the intervals between storms were longer in late winter, and the satellite data (Fig. 14) showed high surface concentrations of Chl a within submesoscale cyclonic eddies at that time.

Degradation and sinking processes of POC in the twilight zone
The transport and degradation of POC formed in the euphotic zone may be observed in the temporal variation of POC concentrations at sq values of 25.25-25.50 (Fig. 8c-f). This density layer sank from the During the observation period, mixing of surface water due to cooling and subsequent restratification occurred repeatedly. Four mixing-restratification events were detected by each float, and enhancement of 565 Chl a concentrations was observed during the restratification phase of all events. Because Argo floats do not always track the same water mass, we determined whether the observed phenomena were occurring within the same water mass based on the water mass structure, heat flux, and SSHA. Our analyses indicated that cases 1, 2, and 4 for float 2903329 and cases 2 and 4 for float 2903330 were events that occurred in the same water mass. In those cases, POC concentrations increased and nitrate concentrations 570 decreased during the restratification phase, and the C/N ratios were close to or slightly lower than the Redfield ratio. The reason for the relatively (to Redfield) low C/N ratios may have been that only suspended POC could be observed by the backscatter sensor in this study, and DOC and fast-sinking POC produced by primary production were not included. The C/Chl a ratio tended to decrease during the restratification period compared to the mixing period in most of these cases. This pattern suggests that 575 production of organic matter was taking place in the euphotic zone after mixing with subsurface water containing particles with high C/Chl a ratios. During the restratification period of Case 4, a significant https://doi.org/10.5194/bg-2022-9 Preprint. Discussion started: 13 January 2022 c Author(s) 2022. CC BY 4.0 License.
increase of Chl a and POC concentrations and an equivalent decrease in nitrate were observed in the euphotic zone within a few days. These changes were caused by deep mixing that alleviated nutrient limitation prior to stratification alleviating light limitation. In addition, the floats were located at the edge 580 of a mesoscale cyclonic feature, and the heaved isopycnal surface may also have contributed to the increased nutrient availability. The largest rate of POC increase, +210 mmol C m −2 d −1 (+2520 mg m −2 d −1 ) during Case 4, was observed by float 2903330. This rate of POC increase was larger than the rate of net primary production reported in this area. The results of this study indicate that intense biological activity events, although short lived and highly localized, occur repeatedly in the western North Pacific 585 subtropical region during winter and spring. There is also a suggestion that cyclonic features increase the nutrient concentrations in the euphotic zone and enhance primary productivity.
In this study, we were also able to use BGC floats to observe biological activities and material transport in the twilight layer. When the mixed layer deepened, concentrations of Chl a, oxygen, nitrate, and POC became uniform in the mixed layer. Concentrations of POC in the twilight layer then decreased as mixing 590 stopped. Decreases of POC (-0.05 to -0.09 µmol C kg −1 d −1 ) were observed in the density range of sq 25.25-25.50 that were exposed to the surface and then capped by less dense waters and isolated below the euphotic zone. Decreases of oxygen (-0.62 µmol O2 kg −1 d −1 ) were also observed for float 2393330; however, the estimated POC consumption based on Redfield stoichiometry exceeded the observed POC decreases. A possible explanation is that the substrate for the decomposition contained not only POC but 595 also DOC, which cannot be observed by a backscatter sensor. At depths below the mixed layer (300-400 m), the averaged POC concentrations increased slightly throughout the observation period. This rate of increase could be attributed to a flux of slowly sinking particles at depths of ~300 m. The estimated rate was 1.1 mmol C m −2 d −1 , which is one-third of the sinking flux estimated by sediment traps in a neighboring area. Most organic matter, whether particulate or dissolved, that accumulated in the euphotic 600 layer during the deep mixing and subsequent restratification events was decomposed at depths shallower than the winter mixed layer (~300 m) and will likely be re-entrained the following winter. However, the particles that sunk deeper than 300 m may be exported and isolated from the atmosphere on the longer https://doi.org/10.5194/bg-2022-9 Preprint. Discussion started: 13 January 2022 c Author(s) 2022. CC BY 4.0 License.
time scale. In the future, particle observation with many BGC Argo floats would be essential to accurately estimate the flux of particles sinking deeper than the winter mixed layer.

605
BGC Argo floats have enabled high temporal and spatial resolution measurements of BGC parameters that were previously observed only discretely by onboard water sampling observations. This capability allowed us to quantitatively evaluate the processes of particle (1) production, (2) decomposition, and (3) transport to depth. It is also clear that these processes are greatly influenced by the mesoscale structure of the ocean and transient weather changes. The use of two floats in this study also revealed differences in     Table 1.         Table 1 Table 1