Local wind effect on the Kuroshio path state off the southeastern coast of Kyushu

Abstract

Small meanders of the Kuroshio southeast of Kyushu, Japan, are conventionally known to form in winter or early spring, but rarely in other seasons. This study examines the seasonal nature of the small meander formation as it relates to the seasonal monsoon winds over the East China Sea, based on observational data analyses, theoretical considerations and numerical experiments. Positional data of the Kuroshio path indicate that the small meander experiences phase locking with the seasonal cycle, as is conventionally thought, although its intensity is modulated on decadal time scales. Monthly mean data of the surface geostrophic current and wind stress fields indicate that the southwestward wind blowing against the Kuroshio in the autumn and early winter causes an inshore-ward shift of the Kuroshio in the northern Okinawa Trough. This effect acts to strengthen the convoluted pattern of the Kuroshio path around Kyushu, and to develop the small meander southeast of the same island. Assuming that the Kuroshio is a surface geostrophic jet with a double-exponential velocity profile, the observed evidence is well explained by a combination of Ekman layer dynamics and quasi-geostrophic dynamics. More specifically, Ekman pumping due to nonlinear Ekman divergence over the jet decreases (increases) the surface velocity of the offshore (inshore) side of the jet, resulting in the inshore-ward shift of the jet. A two-layer shallow water model with idealized topography shows that the response of the modeled Kuroshio to the local wind stress is consistent with observational and theoretical results.

Appendix

1.1 Method used to derive the sea surface geostrophic velocity dataset

Three datasets were used to derive this data. The first is altimeter-derived geostrophic velocity anomalies, which were obtained from the Maps of Sea Level Anomaly (MSLA) product (provided by Aviso of Collecte Localisation Satellites, France). The second is positional data of drogue surface drifting buoys (central drogue depth 15 m) assembled by the Surface Velocity Program (SVP) of the World Ocean Circulation Experiment (WOCE), which were interpolated every 6 h by the Atlantic Oceanographic and Meteorological Laboratory (AOML). The drifting buoy data were low-pass-filtered to remove short-term variations due to ageostrophic velocity using the Hamming window with a cutoff period of 30 h based on Ambe and Ichikawa (2008). The third is the 0.5° gridded weekly surface wind data that were measured by ERS-1/2 and QuikSCAT (IFREMER 2002a, b).

The positional data of surface drifting buoys were processed as follows. Absolute velocities were first calculated using the centered difference scheme along the drifter trajectory at the sampling position and time. Ekman flow velocities at 15-m depth were then computed using Eq. 1 of Niiler et al. (2003) from the wind data that were tri-linearly interpolated onto the sampling position and time. Next, Ekman flow velocities were subtracted from absolute velocities to obtain geostrophic velocities, assuming that vertical shear in the velocity from the sea surface to 15-m depth can be considered negligible.

The method used to estimate a mean sea surface geostrophic velocity field is based on Uchida and Imawaki (2003), but the matching scheme for satellite altimeter and drifter data was improved, primarily as follows. Drifter-derived absolute geostrophic velocities were spatially averaged so that they could be allocated at each grid point of the altimeter-derived geostrophic velocity anomalies using a Gaussian weighting function with a latitude-dependent influence radius:

$$R = \alpha \cdot L.$$

(32)

Here, L (km) is the spatial correlation scale used in the mapping process of MSLA data, which is defined as

$$L = 50 + 250 \times \left[ {900/(\phi^{2} + 900)} \right],$$

(33)

where ϕ indicates latitude in degree. A factor $\alpha = 0.342$ was introduced in order for L to be applicable to the correlation scale (R) between two vectors of MSLA-derived geostrophic velocity. First, we defined an ensemble mean for the cosine of angle between two vectors $\vec{A}_{i}$ and $\vec{A}_{i - j}$ in the zonal direction, where i (1 ~ N) indicates the grid number in the zonal direction and j $(0 \le j < N)$ represents their zonal shift:

$$r = \frac{1}{N - j}\sum\limits_{i = 1}^{N - j} {\frac{{\left( {\vec{A}_{i} \cdot \vec{A}_{i - j} } \right)}}{{\left| {\vec{A}_{i} } \right|\left| {\vec{A}_{i - j} } \right|}}} .$$

(34)

Here, r is $1/\sqrt 2$ for the case where the ensemble mean for the angle between two vectors is 45°. Therefore, we determined that the two velocity fields are coherent for $r \ge 1/\sqrt 2$ (the ensemble mean angle is less than 45°). Finally, α was empirically determined as an average of the zonal shift ratios with $r = 1/\sqrt 2$ to L for all zonal bands from 25°N to 45°N. When compared to Uchida and Imawaki (2003), this improvement effectively removed ageostrophic velocity components and reduced regions of missing data near the coast.