The Kuroshio south of Japan is illustrated by a sharp offshoreward increase in SSH (Fig. 3a). A tongue-like structure of warm sea surface water has been observed along the Kuroshio current by satellite SST measurements (Figs. 1 and 3a). The SD observations found the SST gradient to be greater on the inshore side of the current axis than the offshore side (Figs. 3a and 3b), as reported by Taft (1978). Throughout the observation period, the SD-observed SST change is mostly attributable to passages of the SD across the temperature front of the Kuroshio but not to periodic variations such as diurnal cycle (Fig. 4a). The SST in the Kuroshio region is higher than air temperature during the winter; in other words, air-sea temperature difference (SST minus air temperature) is positive (Fig. 3d). Thus, heat is released to the atmosphere (Fig. 3g), as observed in previous studies (e.g., Hsiung 1986; Konda et al. 2010).
Additionally, spots of very high SST are found along the Kuroshio at approximately 50–100 km intervals (Figs. 1 and 3a), as Fuglister and Worthington (1951) observed in the Gulf Stream in the North Atlantic. The SD crossed twice the Kuroshio just north of the center of the MWS (31.5∘N, 135.8∘E) and crossed the southern edge three times (white lines in Fig. 3a). The dimensions of the MWS parallel and perpendicular to the Kuroshio current direction were approximately 100 and 50 km, respectively, making it the largest of the MWSs observed by satellite SST measurements in the region south of Japan during the study period. The SST around the center of the MWS is higher than 23∘C (Fig. 3a). The MWS continued to exist in almost the same location for approximately two weeks from late December 2018 to early January 2019, as denoted in the map of satellite-observed SST averaged from December 28, 2018, to January 8, 2019, in Fig. 3h.
Meanwhile, atmospheric conditions over the region south of Japan greatly varied due to the passages of low-pressure systems during the observation period (Figs. 4b and 4c). Therefore, air temperature (Fig. 3c) and specific humidity (Fig. 3e) were different each time the SD crossed the Kuroshio. In particular, air temperature and specific humidity were low on December 28–30, 2018. During this period, SD-observed air pressure continued to increase (Fig. 4c) and cold dry northerly or northwesterly winds blew from the continent under typical winter weather conditions around Japan, as indicated by SLP maps based on the JRA-55 data in Fig. 5. The winter weather condition was persistent while the SD crossing the Kuroshio. The influence of the Kuroshio on the atmosphere is identified (Fig. 3f), and the winds were remarkably intensified over the Kuroshio from December 28 to December 30, exceeding 12 m s−1.
Similar enhancements of winds over high-SST areas have been reported in the tropical regions (e.g., Lindzen and Nigam 1987; Wallace et al. 1989; Hayes et al. 1989) as well as over the Kuroshio and the Kuroshio Extension (e.g., Tanimoto et al. 2011; Tomita et al. 2013; Kawai et al. 2014). Because of the low air temperature and strong winds in this period, sea surface turbulent (sensible plus latent) heat flux (Fig. 3g) was estimated to be noticeably high (>700 W m−2) in comparison with other periods. Moreover, the strengthened northerly wind over the Kuroshio enforced sea surface upward turbulent heat flux.
Turbulent heat flux and its related atmospheric and oceanic variables along the track during the high heat flux period from December 28 09:00 to December 29 20:00 (UTC) (track A in Fig. 3a) are shown with respect to latitude by red lines in Fig. 7. These values are averaged within intervals of approximately 2 km, which is equivalent to a time average of approximately 30 min. The A track crossed just north of the center of the MWS (Fig. 3a). For comparison, the variables observed along the track in the case of low turbulent heat flux from January 3 03:00 to January 5 03:00 (track B) are indicated by blue lines in Fig. 7. The B track is along the southern edge of the MWS. When the SD began to cross the B track, JRA-55-based SLP was high west of the study region and low to the east and the background large-scale wind south of Japan was northerly (Fig. 6a and b). These background SLP and wind conditions are like those in the period of the A track; however, after that, the atmosphere rapidly became warm (Fig. 4b) in association with an SLP decline (Fig. 4c) due to the approach of a low-pressure system to the B track (Fig. 6c, d). Along the A and B tracks, the Kuroshio current axis, defined as the maximal sea surface current observed by the ADCP, was present at latitudes of 32.05∘ N and 31.60∘ N, respectively. To compare the atmospheric and oceanic parameters around the Kuroshio current axis, the ordinates in Fig. 7 were adjusted relative to the location of the current axis to match the locations of the Kuroshio current axis along both tracks.
The Kuroshio MWS is found to not have a simple cross-current structure with a single SST peak (Fig. 7b). Large air-sea temperature differences (Fig. 7d) basically correspond to the high SST near the MWS. The maximum SST (∼23∘C) and large air-sea temperature difference on the A track (red line) within the MWS (∼31.9∘ N) largely corresponds to the maximum wind speed (>14 m s−1) (red line in Fig. 7e). The intensification of wind over the SST maximum relative to the surrounding regions exceeds 2 m s−1 and is associated with an SLP depression of approximately 1 hPa (red line in Fig. 7g). The turbulent heat flux reached a maximum of 1141 W m−2 (red line in Fig. 7a), which is much larger than wintertime climatological heat flux in the Kuroshio and Kuroshio Extension region (∼600 W m−2) (e.g., Konda et al. 2010; Sugimoto and Hanawa 2011).
Another SST peak around a latitude of ∼31.7∘N on the A track (red line in Fig. 7b) is also accompanied by a wind speed peak (red line in Fig. 7e) and pressure depression (red line in Fig. 7g). The southwestward sailing SD reached the northern primary SST peak at 05:40 and southern secondary peak at 12:20 in local sidereal time (LST) before the expected peak time of diurnal SST cycle (circa 13:00 LST) studied by Koizumi (1956). The SD-observed SST decreased with time toward noon. As described above, the diurnal SST cycle is also not clear throughout the study period (Fig. 4a). Therefore, the SST peaks are principally attributable not to the diurnal SST cycle but to the intrinsic sea surface structure of the Kuroshio.
These SST and wind speed peaks yield remarkable turbulent sea surface heat flux peaks (red lines in Fig. 7a). The enhancements of latent heat flux in association with strengthened evaporation correspond to peaks in specific humidity (red line in Fig. 7f). Note that the moist air masses are considered to be advected from a little upstream of the A track. Associated with the atmospheric response, a peak in air pressure was exhibited between the depressions (red line in Fig. 7g). This submesoscale structure of SLP intensified the northwesterly wind over the SST peaks. This is consistent with the slightly high SSS observed around latitudes of 31.7∘ N and 31.9∘N on the A track (red line in Fig. 7h), but differs from observed low SSS around the warm core during the summer by Taft (1978). This is probably because summer is not a season of active ocean heat release to the atmosphere.
It should be noted that the time series of air pressure (Fig. 4c) includes periodic variations probably due to the atmospheric tides. The power spectrum of this time series exhibits a significant peak at a semidiurnal period of 11.3 h, which exceeds the 95% confidence interval. Meanwhile, no clear such periodic variations are displayed in the time series of SST (Fig. 4a) and air temperature (Fig. 4b). On the other hand, the submesoscale structure of SLP over the MWS of the Kuroshio is accompanied by the corresponding spatial changes in SST and other atmospheric variables, as described above. For this reason, the submesoscale SLP change over the MWS is considered to be caused not by the atmospheric semidiurnal tide but by the air-sea interaction in the MWS.
On the B track, which is along the southern edge of the MWS, SST and air-sea temperature difference peak values were reduced (blue lines in Fig. 7b, d). As a low-pressure system approached to over the warm core of the Kuroshio, the southwestward sailing SD observed a steep pressure decrease of approximately 10 hPa (blue line in Fig. 7g). Associated with the SLP decrease, the background northerly wind changed to the southwesterly weak wind (Fig. 3f and blue line in Fig. 7e). Consequently, the upward sea surface turbulent heat flux on the B track was smaller while the SD crossing the warm core of the Kuroshio (<300 W m−2) than the surrounding regions (blue line in Fig. 7a). The weakening in turbulent heat flux over the Kuroshio is, therefore, mostly caused by the weakening of wind due to the approach of the low-pressure system. The B track being apart from the MWS center, the influence of the Kuroshio warm core on the atmosphere is insignificant in comparison with the A track near the MWS center.
Large fluctuations in SSS were observed on the B track (blue line in Fig. 7h). However, as noted in “Methods/experimental” section, the SD-observed SSS data includes the sensor drift. Because the exact function form of the sensor drift is unknown, we should be cautious to make a comparison between absolute values of SSS along the A and B tracks. SSS values on the B track such as at latitudes of 31.10– 31.25∘ N and 31.48– 31.55∘N are low and values between these areas are high. The magnitudes of the fluctuations are approximately 0.2–0.3 (psu), which are larger than the RMS difference. The high-salinity fluctuations are likely caused by evaporation and wind-driven mixing of underlying high salinity water derived from the North Pacific tropical water. Meanwhile, low-salinity fluctuations might be attributable to water freshened by rainfall due to the passage of an atmospheric low-pressure system on December 26, 2018, that was cooled by heat release and subducted to the sensor depth (53 cm).
Latent (sensible) heat flux, FL (FS), is a function in terms of SST TS, air temperature TA, specific humidity q, and wind speed U. The spatial heat flux variation, ΔF, is approximated as
$$ {}{\Delta}F \: \!{\approx}\! \: \left(\frac{\partial F}{\partial T_{\mathrm{S}}} \right) {\Delta}T_{\mathrm{S}} + \left(\frac{\partial F}{\partial T_{\mathrm{A}}} \right) {\Delta}T_{\mathrm{A}} \,+\, \left(\frac{\partial F}{\partial q} \right) {\Delta} q \!+ \left(\frac{\partial F}{\partial U} \right) {\Delta}U, $$
(1)
where ΔTS, ΔTA, Δq, and ΔU are spatial variations in SST, air temperature, specific humidity, and wind speed, respectively. To examine which variables dominantly contribute to the turbulent heat flux, we decomposed the heat flux variations into the variation components due to the four variables on the basis of Eq. (1). Because the turbulent heat flux function form is not simple, each term of the right hand side (RHS) of Eq. (1) in this study was evaluated as the rate of the change of a variable by using the mean values for the other three variables on each SD track.
Figure 8a and b show the respective contributions of the variables for latent and sensible heat fluxes on the A track. The sums of the contributions to the variations in both latent and sensible heat fluxes from the four variables, the RHS of Eq. (1), are in good agreement with the variations in ΔF, the left hand side (LHS) of the equation (Fig. 9). Wind speed intensification is found to be the largest positive factor for both latent and sensible heat fluxes around the warm core of the Kuroshio (orange lines in Fig. 8). Remarkably, spatial variation of the heat fluxes on scales of approximately 0.1∘ latitude (∼10 km) and smaller are almost solely attributable to wind speed variations. In particular, the widths of wind peaks are approximately 10 km or slightly more. The peak winds primarily contributed to the latent (sensible) heat flux in the MWS around latitudes of 32.0∘ N and 31.6∘ N by approximately 170 (70) and 70 (30) W m−2, respectively. The contribution of peak SST to the latent (sensible) heat flux, which is approximately 70 (30) W m−2, is secondary (magenta lines). Although air temperature and specific humidity have less effect on the heat fluxes than wind speed and SST near the current axis of the Kuroshio, a cold air mass from the continent intensifies the sensible heat flux on the onshore side of the Kuroshio (green line in Fig. 8b).
On the B track (Fig. 10a, b), latent and sensible heat fluxes are reduced in the Kuroshio owing principally to the attenuation of wind due to the approach of the low-pressure system. In comparison with the A track, the RHS of Eq. (1) on the B track underestimates the attenuation of the latent and sensible heat releases to the atmosphere due to the large wind speed change while the low-pressure system approaching to the south of 32.0∘ N (Fig. 11). The discrepancies for the latent and sensible heat fluxes are not more than approximately 50 W m−2 and 40 W m−2, respectively. Despite the discrepancies, the spatial patterns of the RHS and LHS of Eq. (1) on the B track are similar to each other. Thus, it is certain that the wind variation is responsible to the attenuation in turbulent heat fluxes.
Despite such a large heat release observed on the A track, the Kuroshio MWS continued to exist for at least approximately 2 weeks, as mentioned previously. If the MWS is 50 m thick, a monthly mean value of winter net heat flux (∼600 W m−2) near the Kuroshio current axis (e.g., Tomita et al. 2019) is evaluated to reduce the water temperature at a rate of ∼0.3 K d−1. This evaluated temperature decrease rate is too large to allow the very high SST of the MWS to be maintained for 2 weeks. Additionally, because the Kuroshio current speed (>2 m s−1) is very fast, disturbances are passively advected by the Kuroshio (e.g., Nagano and Kawabe 2005), unlike the MWS. Therefore, the heat content in the MWS should be supplied by the continuing local horizontal current convergence. As shown in the map of horizontal divergence of currents based on OSCAR data (Fig. 12), the MWS centered at approximately 31.5∘ N, 135.8∘ E is found to be in the convergence (or no significant divergence) regions. Although the examination cannot be conclusive evidence as OSCAR data are derived from spatially low-resolution satellite data, the MWS might be maintained by the convergence of warm surface water. This can be clarified by future well-organized observations using unmanned surface vehicles such as SDs.