Shipboard observations were conducted over the Shiriya Spur on 26th August 2021 with the research vessel (R/V) Wakataka-Maru, which is owned by the Japan Fisheries Research and Education Agency. Water sampling was performed using a carousel water sampler (SBE 32; Sea-Bird Scientific, Inc.; Bellevue, Washington, USA) at the start of the drift observation (41.60°N, 141.41°E) and at the end of the drift observation (41.69°N, 141.72°E). Notably, the starting point of the drift observation was located slightly further from the initial deployment point of turbulence measurement due to modifications in latitude. The samples were collected at 10, 20, 30, 50, 60, 80, 100, 125, 150, 200, and 300 dbar in addition to the surface. Moreover, during the drift observations, surface water samples were taken at approximately every other point to assess nutrient concentrations for vertical profiling by the nitrate sensor.
The microstructure measurements were repeated 24 times using a Turbulence Ocean Microstructure Acquisition Profiler (TurboMAP; JFE Advantech Co., Ltd.; Nishinomiya, Hyogo, Japan) that was loosely tethered during the drift observation. The calculation method was consistent with that in the western Tsugaru Strait9, whereby shear probe data were sampled at 512 Hz and processed using standard procedures41. For each 8-metre segment of each profile, the turbulent kinetic energy dissipation rate, ε, was estimated by integrating the shear spectrum from 1 cycle per metre to the Kolmogorov wavenumber, employing the Nasmyth universal spectrum42. Segments with spectral shape deviations from the Nasmyth universal spectrum due to factors, such as rapid changes in the fall rate, were excluded from ε evaluation. The turbulent vertical eddy diffusivity, Kρ, was calculated as Kρ = Γ × ε/N2, assuming Γ to be constant at 0.243. The buoyancy frequency, N, was derived from the temperature and conductivity sensors attached to the TurboMAP.
Simultaneously, with the turbulence observations, nitrate profiling was carried out using Deep SUNA v3 (Sea-Bird Scientific, Inc.). The profile offset was adjusted with reference to the bottle data. Unfortunately, SUNA's batteries ran out during the drift observations, so data for deployments east of 141.65°E could not be obtained. To estimate the vertical turbulent nitrate fluxes, we adopted an approach from a previous study9, 44:
$${F}_{{NO}_{3}}=-{K}_{\rho }\frac{\partial {NO}_{3}}{\partial z}=\frac{\partial {NO}_{3}}{\partial \rho }\frac{{\Gamma }\epsilon \rho }{g}$$
,
where g is the gravity acceleration and ρ is the potential density. In this study, we estimated the gradient of nitrate on the density surface for each profile using a fitting function. We utilized a multiple regression model with density as an explanatory variable, ranging from the first to the tenth order, and log-transformed nitrate concentration. The model with the smallest Akaike's information criterion was chosen. From this fitting function for density and nitrate concentration, we estimated the nitrate gradient in the density plane (Supplementary Fig. 1). Furthermore, we determined that flux was set to zero in regions where the corrected nitrate concentration was zero after 30 m running-mean filtering in the surface region. In the present study, the depth of the nitricline was defined as the depth at which the vertical gradient was greatest for each 30 m filtered nitrate vertical profile. This depth corresponded relatively well with the 2 mmol N m−3 nitrate concentration, except for 141.58°E–141.60°E immediately downstream of the topography (Fig. 2c). Fluxes were averaged over a range of ± 5 m at this depth and were defined as fluxes in the nitricline (Fig. 2d).
We utilized velocity data obtained from an acoustic Doppler current profiler mounted on the R/V Wakataka-Maru (SADCP, Ocean Surveyor, Teledyne RD instruments, 38 kHz, Poway, California) in our study. The data were corrected for misalignment45, and data with a percent good value of 80% or lower were excluded. Horizontal smoothing was applied to the data, which were subsequently divided into 3 km grids for comparison of the horizontal divergence of the zonal velocity (∂u/∂x) with that of the HFR velocity.
The HFR monitoring system employed in this study is the SeaSonde (13.9 MHz, CODAR, Mountain View, USA). The HFR data provided a spatial resolution of ~ 3 km covering an area of ~ 3 to 60 km from each antenna (Fig. 1c)3, 8. Approximately every 30 minutes, the system calculated the distribution of surface currents, which was uploaded to the MIO Ocean Radar data site for the eastern Tsugaru Strait (MORSETS; http://www.godac.jamstec.go.jp/morsets/e/top/). We excluded grids with data acquisition rates below 90% in 2021 with reference to the previous method8. Additionally, we verified the consistency of the HFR velocity with that obtained from SADCP data along the drift trajectory (Supplementary Fig. 2).
We analysed velocity fields for a larger region than the coverage of the HFR using surface velocity data from the Ocean Surface Current Analysis Real-time (OSCAR) dataset, which provides third-degree resolution of ocean surface currents (Ver. 1. Physical Oceanography Distributed Active Archive Center, National Aeronautics and Space Administration, California, USA.). The data, covering the period from 1992 to 2021, were accessed on 9 June 2022 at https://doi.org/10.5067/OSCAR-03D0146.
The surface distributions of chlorophyll-a were analysed using data from GlobColour (http://globcolour.info), which was developed, validated, and distributed by ACRI-ST, France24, 25. We selected data products with 4-km resolution and 1-day resolution, utilizing the level-three Garver-Siegel-Maritoren (GSM01) merged model47, 48 as well as that of 8-day resolution for Fig. 1c. The merged data encompassed information acquired from three satellite sensors: SeaWiFS (Sea-Viewing Wide Field-of-View Sensor) on GeoEye’s Orbview-2 mission, MODIS (Moderate resolution Imaging Spectroradiometer) on NASA’s (National Aeronautics and Space Administration) Aqua mission, and MERIS (Medium Resolution Imaging Spectrometer) on ESA’s ENVISAT (Environmental Satellite) mission. The period of study extended from 2014 to 2020.
Following the methodology presented by a previous study in the western Tsugaru Strait9, we determined the layer Froude number using the following formula:
$${Fr}_{l}=\frac{{U}_{l}}{{N}_{l}{h}_{l}}$$
,
where Ul, Nl, and hl represent the zonal velocity, vertically averaged buoyancy frequency, and thickness of the layer between the 24.0 σθ isopycnal surface and the observed bottom layer, respectively49.
We utilized the function previously suggested36 linking primary production and chlorophyll-a concentration, expressed as log(PP) = a + blog(Chl-a). In this equation, PP and Chl-a represent primary production and chlorophyll-a concentration, respectively, while constants a and b have fixed values of 2.793 and 0.559, respectively.
We utilized ETOPO150 for our topographic data.