How a Small Reef in the Kuroshio Cultivates the Ocean

Vertical nitrate fluxes associated with turbulent mixing and upwelling around a small reef in the Kuroshio are quantified by continuously deploying a turbulence microstructure profiler with an attached submersible ultraviolet nitrate analyzer while drifting from the upstream to the downstream of the reef. Flow separations and trains of Kelvin‐Helmholtz billows (thickness = 60 m) are identified using a shipboard ADCP and an echo‐sounder. The turbulence diffusivity associated with the vigorous turbulent mixing reaches up to O(10−1 m2 s−1), resulting in strong nitrate fluxes of O(1–103 mmol m−2 day−1). In addition, large differences between the upstream and downstream density profiles suggest a strong upwelling velocity of O(10−3 m s−1), as well as an upwelling nitrate flux of O(102 mmol m−2 day−1) in the entire subsurface layer.

occurrence of strong upwellings (e.g., Chang et al., 2019Chang et al., , 2013Hsu et al., 2017, and others). Nevertheless, such upwelling studies remain somewhat descriptive given the absence of methods for directly measuring vertical velocity.
Under the OMIX project (2015( -2020, interdisciplinary studies attempting to quantify the effects of turbulent mixing and upwelling caused by small islands and seamounts in the Kuroshio Current have been carried out. Nagai et al. (2019) demonstrated the enhanced turbulent vertical nitrate flux in the downstream of the Tokara Strait based on results of Underway VMP measurements with the indirectly estimated nitrate concentration using the local relationship between salinity and nitrate concentration. Hasegawa et al. (2019) developed a method for measuring the turbulent vertical nitrate flux in which an in situ nitrate sensor was attached to a turbulence profiler. Tanaka et al. (2019) used this method to demonstrate that the vertical mixing associated with rough topography at the Izu ridge elevated the nitrate flux and contributed to the nitrate component of the Kuroshio nutrient stream (Guo et al., 2013). Furthermore, Kobari et al. (2020) conducted nutrient addition experiments in oligotrophic Kuroshio water to determine the effects of turbulent nitrate flux elevation in the Tokara Strait. They found a significant increase in microzooplankton grazing on phytoplankton, suggesting that such rapid and systematic trophodynamics support biological productivity in the Kuroshio.
These remarkable scientific efforts have provided a glimpse of the spectacular Kuroshio IMEs, but some details are still missing, and therefore so is the complete picture. Herein, we focus on the detailed views of IMEs in the Kuroshio, in particular the questions of how and how much nitrate is supplied to the upper layers associated with the flow disturbance at Hira-se, a small reef in the Kuroshio.

Survey and Instrumentation
In November 2016, we conducted a survey using the TRV Kagoshima Maru (1,284 ton) around Hira-se while the northern edge of the Kuroshio was located on this reef (Figure 1a). Hira-se is the apex of an eroded conic volcano. It measures 2.5 km north-south at the surface, and it stands 500 m above its 15 km submarine foundation ( Figure 1c). From 13:45 JST (GMT+9) through 23:36 JST on November 16, 2016, we passively drifted the vessel from the upstream of Hira-se with the surface current along the northern passage between Hira-se and Uwa-no-se while continuously collecting measurements of microstructure turbulence (TurboMAP-L; JFE Advantech), horizontal current (75 kHz shipboard ADCP with 16 m bin; RDI), and HASEGAWA ET AL. echogram (EK60; Simrad) from station M01 to station M40 (refer to the colored circles in Figure 1). The shipboard wind speed was weak (average speed = 2.6 m s −1 ) during the survey; therefore, it was assumed that the vessel drifted along the surface current.
We also observed the nitrate concentrations and the light intensities in the water column simultaneously with the deployment of the microstructure profiler to precisely quantify the turbulent vertical nitrate flux to the euphotic layer. The system observing the turbulent nitrate flux was recently developed by Hasegawa et al. (2019) and consists of a submersible nitrate sensor (Deep SUNA V2; Sea-Bird Scientific) and a PAR sensor (DEFI2-L; JFE Advantech) attached to a microstructure profiler (TurboMAP-L). The proportionality coefficient of the nitrate sensor was calibrated against the bottle-sampled nitrate profiles collected during the same cruise, and the sensor offset for each cast was determined in order to set the surface minimum value to zero (mean offset is 0.78 ± 0.22 mmol m −3 ), since the isopycnal associated with the nitracline did not outcrop at the surface during the survey.
The vertical turbulent nitrate flux NO 3 F was quantified by multiplying the estimated vertical turbulent diffusivity from the microstructure measurements by the vertical gradient of the observed nitrate concentration, that is, where NO 3 is the nitrate concentration.

Flow Perturbation and Turbulent Mixing
During the survey, the surface current was eastward upstream of Hira-se, then flowed into the channel between Uwa-no-se and Hira-se (see Figure 1b). Further, the flow in the intermediate layer (264 m, Figure 1c) was southward upstream of Hira-se, going around the shallow reefs, with some portion flowing into the channel along with the surface flow. In the channel, the surface flow turned southeast while passing over the sill extending from the northeastern edge of Hira-se. Thereafter, the surface flow turned south at the east side of Hira-se. The intermediate layer flow direction was opposite to that of the surface flow just in the lee of the sill and then oscillated between west and east downstream of the reef.
When the flow surmounted the shallow sill, the surface flow accelerated to 0.8 m s −1 (Figure 2a). A strong mean shear layer, namely the separated bottom boundary layer (SBBL), was evident between the surface jet and the sill peak. Profiles of microstructure shear (yellow plots in Figure 2b) indicate the development of the SBBL from the sill peak downstream, with the thickness of the turbulent layer increasing to more than 100 m by the time it reached station M15. The intensified wavy echo band in the echogram also indicates the growth of the SBBL along this trajectory ( Figure 2b).
In the layer located 40 m above the seafloor at station M13 (see Figure 1b for the location), the gradient Richardson number, Ri = N 2 /S 2 , in the SBBL measured ∼1 (black line in Figure 2e), which is above the critical value of 0.25, where N 2 = −g/ρ 0 ∂σ θ /∂z is the buoyancy frequency squared, σ θ is the potential density, ρ 0 is the reference density, g is the gravitational acceleration, and S 2 is the mean squared vertical shear. We cannot confirm the instability condition for Ri at the beginning of the SBBL close to the sill peak because the velocity data are missing due to acoustic interference at the seafloor, and the relatively coarse vertical resolution (16 m) of the velocity measurement is possibly overestimating Ri. However, the profile of the Thorpe length scale,  2 T T L d , shows 10 m-scale thermal inversions in the SBBL (green line in Figure 2f), where d T is the Thorpe-scale displacement. It should be noted that because the salinity quality was not sufficiently high to be used for this analysis, a temperature-based Thorpe-scale displacement d T was used instead of a density-based value. The kinetic energy dissipation rate (ϵ; magenta line in Figure 2e) in the same depth range reaches O(10 −5 W kg −1 ), and the corresponding vertical turbulent diffusivity (K ρ = 0.2⋅ϵ/N 2 Osborn, 1980; blue line in Figure  The vertical section of ϵ (Figure 2d) shows highly turbulent layers associated with the SBBL above the sill as well as at the KH billows. A remarkably turbulent patch in the surface layer also is evident near the end of the survey (stations M36 to M38; see Figure 1b for the location). This high-turbulence patch was also associated with shear instability, as indicated by the velocity profiles near those stations recording high turbulence (Figure 2a). This patch of enhanced turbulence appeared suddenly 14 km downstream of the sill without visible connection to the upstream instability phenomena. For clarifying the cause of the isolated high-turbulence patch, further investigations are required.
HASEGAWA ET AL.

Nitrate Flux
Throughout the drifting survey, the nitrate concentration in the surface layer was depleted below 1 mmol m −3 (Figure 3c). At station M13, where the SBBL was evident, the nitrate concentration gradually increased with increasing depth starting at 90 m depth (Figure 2f) with a vertical gradient magnitude of O(10 −2 mmol m −4 ; cyan line in Figure 2g). In the subsurface layer below 150 m depth at M18, the vertical gradient steepened to O(10 −1 mmol m −4 ; Figure 2j).
In the highly turbulent SBBL,  hanced nitrate flux reached O(1-10 3 mmol m −2 day −1 ), and the maximum fluxes of O(10 3 mmol m −2 day −1 ) were located at stations M21 and M22, at 210 and 200 m, respectively. High fluxes of O(10 2 mmol m −2 day −1 ) were also found in the subsurface layer at stations M36 and M37 where enhanced mixing was observed (Figure 2d). On the other hand, a low-flux band was observed along the surface jet where turbulence was inactive. The observed nitrate flux of O(10 3 mmol m −2 day −1 ) represents an extremely large turbulent flux value, apparently larger than any reported for the open ocean, but is of the same order as that of an energetic tidal estuary reported by Cyr et al. (2015).
In terms of the direct effect of nutrient flux to the ecosystem, the important flux is that to the layer above the compensation depth, that is the euphotic layer, where net photosynthesis is positive (Cullen, 2015). The compensation depth coincides with the depth where the relative light level, PAR rel. (z) = PAR water (z)/ PAR surf × 100%, is 0.1%-1% (Laws et al., 2014). Here, PAR water (z) is the vertical profile of the Photosynthetically Active Radiation in the water, and PAR surf. is the surface radiation. The mean radiation profile based on the mean attenuation coefficient ( d K = −0.048 m −1 ) obtained by a PAR sensor during the survey (gray line in Figure 3e) indicates a compensation depth of 143 m for a light level of 0.1%, and a compensation depth of 96 m for a light level of 1%. The enhanced flux did not affect the surface layer, as the shallowest depth observed during the survey for the nutricline defined by a 1 mmol m −3 nitrate concentration was 46 m (red line in Figure 3c). The layer-averaged flux for depths from 50 to 150 m (hereafter we call the layer the subsurface layer), corresponding with the lower euphotic layer, shows high values of O(10 mmol m −2 day −1 ) in the lee of the sill peak (station M13) associated with the SBBL. Values of O(10 2 mmol m −2 day −1 ) were observed at stations M36 and M37 where an isolated turbulent patch was observed (red circles in Figure 3b). These enhanced fluxes in the euphotic layer are expected to affect the downstream ecosystem.
The nitrate flux associated with KH billows was limited in the layer below the euphotic layer between 150 and 300 m (hereafter, we call the layer as the intermediate layer). The layer-averaged flux for the intermediate layer (blue circles in Figure 3b) was mostly far higher than O(1 mmol m −2 day −1 ) for the section with the KH billows (stations M17 to M26), and a peak value of O(10 3 mmol m −2 day −1 ) was reached at station M21. The averaged flux profile of the section (red dashed line with blue circles in Figure 3g) indicates a turbulent layer around the KH billows with a peak NO 3 F value of ∼ O(10 3 mmol m −2 day −1 ) at 220 m.
To evaluate the contribution of the turbulent vertical nitrate flux to the change in nitrate concentration in the water column, we performed a simplified 1D mixing model calculation for the section between stations M17 and M26 as follows: where T R (z) = L/|u(z)| is the layer residence time in the section between stations M17 and M26 (yellow line in Figure 3g), u(z) is the mean horizontal velocity profile, and L = 5.0 km is the cumulative distance of the section. For the computational simplicity, a vertically averaged value ( R T ) was used.    ( ) K z is the section-averaged turbulent diffusivity profile. A 50 m scale vertical smoothing was applied to    ( ) K z to study the overall effect of the turbulent mixing. We used the most upstream profile (M17) of the section as the initial condition. During the calculation, the flux was limited to maintain minimal stratification (i.e., −∂NO 3 /∂z > 0) in order to prevent the turbulent flux from causing concentration inversions.
The dashed magenta line with blue circles in Figure 3h represents the model result. In the layer between 175 and 215 m, the estimated mixing effect indicates the nitrate increase of 55.9 mmol m −2 associated with strong turbulent mixing by the KH billows. However, the increased nitrate is supplied from the layer below the flux peak, so the column-total change for the intermediate layer (150-300 m) is small as +7.0 mmol m −2 . The estimated changes due to mixing in the subsurface layer (50-150 m) are generally small, and the total change for this depth range is only +0.2 mmol m −2 . On the other hand, the actual change in the nitrate profile for this section (i.e., the increase from station M17 to station M26, plotted as a magenta line in Figure 3h) is positive for depths of 40 m and beyond. The increases in column-total nitrate concentration for the subsurface and intermediate layers are 186.0 and 263.6 mmol m −2 respectively. Therefore, mixing effects are almost hidden for the column-total changes. Focusing solely on the KH billow layer, the mixing HASEGAWA ET AL.
10.1029/2020GL092063 6 of 9 contribution to the actual concentration change is 40.1%. Indeed, this mixing contribution can be seen as a maximum peak on the observed profile change around 200 m, where the mixing contribution profile also takes the peak (solid magenta line on Figure 3h).
As above, turbulent mixing can still be an important process in explaining the local increase in nitrate concentration in the intermediate layer, but not in the subsurface nor for the column-total increases for this section. Therefore, other processes must be responsible for explaining the nitrate concentration increases.
The vertical section of nitrate concentration anomaly (Figure 3c) shows domings of the nitrate 1, 5, and 15 mmol m −3 isolines downstream from the sill, along with isopycnals of roughly 23, 25, and 26 σ θ , respectively. Fluctuations in the nitrate isoline depths with respect to the values measured at station M17 ( Figure 3d) indicate rises in the isolines along the section from station M17 to station M26. Meanwhile, the column-total concentration for the euphotic layer roughly doubles from 60.0 mmol m −2 at station M17 to 131.0 mmol m −2 at station M26. The maximum rises for the 1 and 5 mmol m −3 isolines at station M26 are 51 and 54 m, respectively. For the 15 mmol m −3 isoline, the maximum rise is 29 m at station M22.
As an alternative to vertical mixing, "upwelling" may be responsible for the nitrate concentration increases. The vertical velocity w required for the density change (Δσ θ ) can be estimated as The estimated w (green line on Figure 3f) reaches a peak value of 0.25 × 10 −2 m s −1 at 70 m but approaches zero at the surface and at 210 m. At depths deeper than 220 m, w becomes negative. The w profile suggests the existence of a surface divergence layer up to the depth of the vertical velocity peak, as well as a convergence layer below.
The nitrate flux associated with upwelling can be obtained as NO 3 .  Figure 3h show the residual (unexplained) changes calculated by subtracting the mixing and upwelling contributions from the observed change. The column-total residual nitrate increase in the subsurface layers is 51% (95.9 mmol m −2 ). Because the effect of the nutricline uplift has been already accounted for as the upwelling contribution, the residual increase could be attributed to changes along the isopycnal surfaces. These changes, occurring between M17 and M26, were analyzed in the density coordinate and then projected back on the depth coordinate using the density-depth relationship at M26 (gray bars in Figure 3h). The column-total change along the isopycnal surfaces is 82.9 mmol m −2 , which explains 87% of the residual change.
The increase in the concentration along the isopycnal surfaces is generally caused by a diapycnal mixing; however, no sufficiently strong turbulence was observed in the subsurface layer of the monitored section can explain this increase. The increase along the isopycnal surfaces can otherwise be attributed to (1) horizontal dispersion and mixing, (2) data limitation (e.g., errors in the nitrate measurements or missing a strong mixing), or (3) analysis errors such as inaccurate vertical second derivatives and isopycnal projection. Here, we propose (1) as the reason for this phenomenon because the sill depth closer to the reef is shallow; therefore, it is perfectly possible that the strong mixing in the shallow layer near the reef enriches the subsurface nitrate and then the enriched water masses are advected to the observed section with the complicated wake flow.
For the intermediate layer (150-300 m), the nitrate increase along the isopycnal surfaces is dominant (216.7 mmol m −2 ), which explains 82.1% of the total change in this layer, whereas the upwelling contribution is 40.2 mmol m −2 (15.2%). Even though the mixing contribution is large in the KH billow layer, the column-total contribution of the mixing is small (7.1 mmol m −2 ) and cannot explain the large increase along the isopycnal surfaces. Thus, here too, horizontal processes were suggested as the reason for the nitrate increase in the intermediate layer.
HASEGAWA ET AL.

Conclusions
The flow of the Kuroshio past a shallow sill on the flank of Hira-se is actively turbulent and exhibits bottom boundary layer separation, and Kelvin-Helmholtz instability. These processes produce patches of turbulent nitrate flux that are extremely strong for the open ocean. In addition, the observed nitracline rise with isopycnal doming suggests strong upwelling in the lee of the reef. The magnitude of the estimated upwelling nitrate flux is one order smaller than the peak value of the turbulent nitrate flux but encompasses the entire subsurface layer. The combination of mixing and the upwelling cultivates the water column and results in significant nutrient enrichment in the euphotic layer. Nevertheless, the observed turbulent flux and estimated upwelling cannot explain the total change in the nitrate profile. The large nitrate increase along the isopycnal surfaces suggested the occurrence of horizontal processes; though the enriched water is probably being originally introduced by the turbulent mixing near the reef.
Since the Kuroshio Current always passes through ridges with abrupt topographies, we expect constant existences of vigorous turbulent mixing, strong upwelling, and strong vertical and horizontal nutrient fluxes in those regions. Therefore, considerable efforts are still required for fully grasping the total effects of these flow-topography interactions on the Kuroshio Current and its ecosystem.

Data Availability Statement
The in situ observational data that support the findings of this study are openly available at http://dx.doi. org/10.17632/nx8tz6gcw2.1. The Ssalto/Duacs altimeter products are produced and distributed by the Copernicus Marine and Environment Monitoring Service (CMEMS) (https://marine.copernicus.eu).