In Situ Evidence of Low-Level Atmospheric Responses to the Oyashio Front in Early Spring

Previous modeling studies have indicated that the Oyashio front in the subarctic Pacific Ocean significantly affects the atmosphere on mesoto basin scales; however, there were no in situ observations that captured oceanic imprints on the atmosphere in this region as far as the authors know. We present in situ evidence of atmospheric responses to the Oyashio front by using a total of 103 radiosondes launched around the Oyashio front in April 2013 with continuous surface meteorology and ceilometer observations. Composite profiles showed that the lowlevel atmosphere below 1000 m was statically stable on the cold side of the Oyashio front, but unstable and mixed on the warm side. In the atmosphere on the warm side, the relative humidity dropped sharply at an altitude of around 1000 m, an indication that the mean cloud top was at this altitude. While the frequency of cloud base height peaked at 50 – 100 m in the cold areas, cloud bases were distributed at higher altitudes in the warm areas. These differences in the atmospheric boundary layer and cloud base heights across the front were clearer under conditions of southerly winds compared with those of northerly winds. Above a local sea surface temperature minimum with a width of approximately 400 km, where the ocean mixed layer depth is known to reach a local maximum, a large horizontal air temperature gradient was observed below an altitude of 1000 m. This horizontal gradient corresponded to a sea level pressure (SLP) anomaly of 1.2 hPa, comparable to observations of SLP anomalies in the Kuroshio Extension region. Furthermore, we found that narrow warm ocean streamers moistened the overlying atmosphere, affecting downward longwave radiation. Over the wide streamer located between 146.4°E and 147.0°E on 5 April, the near-surface atmospheric properties were largely different over the western half and the eastern half.


Introduction
The responses of western boundary currents and oceanic eddies to the mid-latitude atmosphere have been intensively investigated for the past two decades, and their mechanisms and importance in the climate system have been increasingly revealed (see the reviews by Small et al. 2008;Kwon et al. 2010).Local imprints of spatial variations of sea surface temperature (SST) on the atmosphere have been confirmed by in situ and satellite observations as well as numerical models (e.g., Chelton et al. 2004;Spall 2007;Tokinaga et al. 2009).Furthermore, local responses induce even distant influences.Sato et al. (2014), for example, indicated that diabatic heating accompanied by upward motion over the Gulf Stream remotely affected the atmospheric circulation over the Barents Sea and Eurasia.Ma et al. (2015) showed that meso-scale SST anomalies in the Kuroshio-Oyashio Extension region altered storm development in the Northeastern Pacific, resulting in enhanced winter precipitation along the U.S. Northern Pacific coast.Model simulations have confirmed that low-level baroclinicity enhanced by the mid-latitude SST front is essential for anchoring and maintaining storm tracks and the polar-front jets (e.g., Nakamura et al. 2008;Ogawa et al. 2012).
For the local response, two mechanisms have been proposed and debated to explain the relationship between near-surface winds and SST fronts.One is conventionally referred to as the vertical mixing mechanism: downward momentum transport is intensified within the low-level atmosphere destabilized over the warm sea surface, resulting in stronger surface wind (e.g., Wallace et al. 1989;Chelton et al. 2004).Another mechanism is the pressure adjustment mechanism: high and low sea level pressure (SLP) anomalies are produced on the cold and warm sides across an SST front, respectively, and surface wind converges (diverges) over the warm (cold) side (e.g., Lindzen and Nigam 1987;Minobe et al. 2008).The vertical mixing mechanism becomes dominant when the cross-frontal wind is strong, but both mechanisms can be important under moderate cross-frontal wind conditions (Chelton and Xie 2010).In either mechanism, the atmospheric mixed layer becomes higher on the warm side due to instability, and the heights of the low-level cloud base and top also show contrasts across SST fronts (e.g., Tokinaga et al. 2009;Liu et al. 2014;Kawai et al. 2015).The vertical mixing mechanism and the pressure adjustment mechanism are compatible with each other.If atmospheric vertical mixing is strongly suppressed around an SST front area, the SST gradient cannot induce a horizontal temperature gradient in the atmosphere, and the pressure adjustment mechanism cannot be effective.Reinforcement of the vertical mixing over a warm area is also important to the pressure adjustment mechanism.
In the Kuroshio and Kuroshio Extension (KE) regions, turbulent heat fluxes to the atmosphere are the largest in the world, and sharp SST fronts are located on the northern sides of the Kuroshio and KE.The KE is the eastward strong current after the Kuroshio leaves the Japanese coasts, and flows approximately east of 141°E, south of 38°N.Various studies have investigated atmospheric responses to the Kuroshio and KE by using observations and numerical models, and it is widely recognized that the KE greatly affects the atmosphere on local to basin scales.Observational research using radiosondes has shown that the SST distribution across the KE front is locally responsible for a distinct meridional contrast in atmospheric warming from the sea surface.This contrast leads to meridional gradients of air temperature, turbulent heat fluxes, SLP, mixed layer height (MLH), and cloud height (e.g., Tokinaga et al. 2009;Kawai et al. 2014Kawai et al. , 2015;;Nishikawa et al. 2016).These observations confirmed that low SLP anomalies were located over the warm side, supporting the pressure adjustment mechanism.
The Oyashio (subarctic) front is another major SST front in the Northwestern Pacific Ocean.It is located at the southern edge of subarctic low-salinity water along the northeastward return current of the Oyashio.Yasuda (2003) defined the Oyashio front as the western part of the subarctic front in the vicinity of Japan, and Qiu et al. (2017) reported that the western part of the subarctic front fluctuates independently of the eastern part (east of 153°E); however, we do not distinguish between the western and eastern parts in this paper.Whereas the KE front is deep and has a modest SST gradient, the Oyashio front is shallow and has a stronger SST gradient, although the sea surface height gradient is not strong (Nonaka et al. 2006;Qiu et al. 2017).Air-sea interaction research based on in situ Citation Kawai, Y., H. Nishikawa, and E. Oka, 2019: In situ evidence of low-level atmospheric responses to the Oyashio front in early spring.J. Meteor. Soc. Japan, 97, 423-438, doi:10.2151/jmsj.2019-024.observations has been performed over the KE, the Gulf Stream (Sweet et al. 1981), and the Brazil-Malvinas confluence region (Pezzi et al. 2009).Compared with the fronts on these other western boundary currents, SST is lower and the surface turbulent heat fluxes are smaller around the Oyashio front.Also, since the KE front is located nearby, the role of the Oyashio front had not been seriously considered in air-sea interaction.Some studies, however, have recently indicated that the Oyashio front plays an important role in the climate system independently of the KE front.Frankignoul et al. (2011) and Taguchi et al. (2012) have shown that SST anomalies in the Oyashio frontal zone from fall to winter affect the wintertime Aleutian low on a decadal scale.Masunaga et al. (2015), who analyzed high-resolution satellite data and global atmospheric reanalysis data in which the resolution of SST was high enough to distinguish the Oyashio front from the KE front, have revealed that the Oyashio front enhances the meridional contrast in turbulent heat fluxes, resulting in a local SLP minimum that is associated with surface wind convergence, upward motion, and cloudiness on the southern side of the front.Shimada and Minobe (2011) showed, using satellite data including the Atmospheric Infrared Sounder (AIRS), that the Laplacian of atmospheric thickness between 1000 and 850 hPa was proportional to surface wind divergence and the Laplacian of SST around the Oyashio (their Fig. 1b).The indications by Masunaga et al. (2015) and Shimada and Minobe (2011) support the pressure adjustment mechanism.Model results, including reanalysis products, are not always right, and one must constantly verify them by observations.Satellite observations are very useful, but their accuracy and resolution are not as good as in situ observations.The nominal resolution of the AIRS temperature product is, for example, approximately 1 km vertically in the troposphere, not necessarily sufficient to capture the response of the atmospheric boundary layer.However, while in situ observations have been intensively carried out to examine the effects of the KE on the atmosphere, in situ evidence of atmospheric responses to the Oyashio front has been scarce, despite the recognized importance of the Oyashio front.
In this article, we present in situ evidence of atmospheric local responses to the Oyashio front captured during an early spring observation cruise, that is, unprecedented meteorological observations around the Oyashio front.The cruise and observations are described in Section 2, and the analysis results are presented in Section 3. We did not conduct observations to examine the temporal variations of the atmosphere.Only average vertical structures and snapshots across the front are shown, and temporal variations are not discussed in this paper.Section 3.1 shows the mean differences in vertical atmospheric profiles and cloud base height across the Oyashio front.In situ-based estimation of the SLP anomaly over a cold SST area is given in Section 3.2.We also demonstrate examples of imprints on the low-level atmosphere of warm ocean streamers with a width of several tens of kilometers in Section 3.3.A "streamer" is the oceanographic term that means a narrow strip of warm or cold water, often observed in mid-latitude oceans such as the Kuroshio-Oyashio Extension and the Gulf Stream (e.g., Sugimoto and Tameishi 1992).Observations of the atmospheric response to an ocean streamer have been scarcely reported.Section 4 is a summary.

Cruise and data
Research cruise KH-13-3 of R/V Hakuho-maru took place in April 2013.Its main purpose was to investigate the formation of the denser type of central mode water (Nakamura 1996;Suga et al. 1997;Oka and Suga 2005) and of transition region mode water (Saito et al. 2007) on the southern side of the Oyashio front.The research vessel departed from Tokyo on 2 April; cruised around the Oyashio front from 5 to 27 April, stopping twice at Kushiro, Hokkaido (Fig. 1a); and arrived at Shimonoseki at the western tip of Honshu Island on 1 May.High-resolution hydrographic observations were carried out with a conductivity-temperature-depth profiler (CTD) and expendable CTDs along zonal sections across the Oyashio front.
Global positioning system (GPS) radiosonde observations were conducted with an RS-06G radiosonde and an RD-08AC receiver (Meisei Electric Co., Ltd) a total of 119 times along the ship route (Fig. 1a and Supplement).However, 16 radiosondes that were launched south of 36.5°Nacross the KE, on the way to Shimonoseki after the intensive observations in the Oyashio frontal region, were excluded from this analysis.Because no barometer was attached to the radiosondes, atmospheric pressure was calculated by using the hypsometric equation and the GPS-measured height.The radiosondes were sometimes launched in rapid succession while the ship was cruising at a speed of approximately 15 knots (27.8 km h −1 ) to obtain cross sections of temperature, humidity, and wind speed (Sections 3.2 and 3.3).At other times, they were launched at irregular intervals of more than 3 h between oceanographic observations.The radiosonde data were interpolated at 10-m-height intervals before analysis.SST and surface meteorological variables, including air temperature, dew point, wind speed and direction, SLP, and shortwave and longwave radiation, were recorded every minute.A CL31 ceilometer (Vaisala) was used to measure the cloud base height according to lidar backscatter intensity every 15 s; here, the median values at 5-min intervals were analyzed.The in situ observations used in this study are available at http://ocg.aori.u-tokyo.ac.jp/member/ eoka/cruises/kh-13-3/data/index.html.
To examine the two-dimensional distribution of SST, we used optimally interpolated SST (OISST) produced from microwave and infrared satellite data (Gentemann et al. 2009) (Fig. 1).However, the OISST cannot always capture small-scale structures because of optimal interpolation smoothing.We therefore also used Pathfinder SST, produced with a horizontal resolution of 4 km from satellite infrared data (Casey et al. 2010;Saha et al. 2018), to identify ocean streamers (Section 3.3).In the Pathfinder dataset, an overall quality flag, which varied from 0 to 7, was assigned to each SST value.The 3-day means presented here include SSTs of all flags; that is, they include even the lowest quality SSTs, which are inappropriate for most applications.Although these SST values are unreliable, these low-quality data are necessary to capture the spatial structure of SST.Here, only the spatial distribution of SST is relevant to our analysis and is considered; absolute SST values, which are unreliable, are ignored.

Boundary layer structure on the warm and cold
sides of the Oyashio front Cyclones, some of which were quite developed, frequently passed over the observation region in April 2013.In general, the radiosondes were launched before the cyclones approached the research vessel: we were forced to suspend observations during and just after the cyclones passed; as a result, 49 of the 103 radiosonde observations analyzed were conducted when the surface wind direction was approximately southeast (wind direction = 90 -180°).East of 150°E, the SST front and warm water extended northward, although a local SST minimum was located at around 42°N, 157°E (Fig. 1a).A deep oceanic mixed layer was observed in this area on our cruise, the existence of which is consistent with previous observations (e.g., Suga et al. 2004;Faure and Kawai 2015).The SST front extending northeast-southwest between 150°E and 155°E corresponds to a quasi-stationary jet trapped by the bottom topography; this jet plays an important role in transporting warm water toward the subarctic region (Isoguchi et al. 2006;Mitsudera et al. 2018).The Oyashio front has been defined as the 4°C isotherm at 100-m depth (Yasuda 2003).The large horizontal gradient of SST mostly corresponded to the 4°C SST isotherm during the cruise (Fig. 1b) as SST was close to the temperature at a depth of 100 m because of the weak early spring stratification.Therefore, in this paper, we define the 4°C SST contour between 142.5°E and 165.0°E, except for a circle around 45°N, 162.5°E, as the Oyashio front.
Figures 2 and 4a show individual and composite vertical profiles of the lower atmosphere above in situ SST higher and lower than 4°C.Radiosonde sites with lower SST were distributed in the western part of the observation region (Fig. 1a), and contrasts across the Oyashio front shown below reflected geographically east-west distribution.Below a height of 1000 m, the vertical gradient of potential temperature was smaller south of the Oyashio front.Although the low-level potential temperature was relatively close to constant vertically over areas with SST > 4°C, the atmosphere was stable over cold areas (Figs. 2, 4a).On the cold side of the front, the air in the vicinity of the sea surface was close to saturation, and it became drier than air on the warm side of the front between heights of 100 and 1000 m (Figs.3b, 4b).On the warm side, the atmosphere was humid below 1000-m height, but relative humidity dropped sharply above that; therefore, the mean cloud-top height was inferred to be approximately 1000 m (Figs.3a, 4b).The mean wind speed profiles also show that the boundary layer was well mixed, and the wind below 1000 m was stronger on the warm side and stable on the cold side (Figs.3c,  d, 4c).We examined the profiles separately when the winds were southerly (90 -270°, 65 %) and when they were northerly (270 -360° and 0 -90°, 35 %) (Figs. 2,  5).When the winds were southerly, the mean profiles were similar to the overall average profiles (Figs.2a, c, 5a), with enhanced differences between the southern and northern sides of the front.In contrast, when the winds were northerly, the differences in the mean profiles between the north and south of the front were small below a height of 1000 m (Figs.2b, d, 5b).On the basis of the radiosonde observations during this cruise, at least, we observed no clear correlation between the air-sea temperature difference and the surface wind direction.The small number of northerly cases and the absence of data during and just after the cyclones passed might have affected the results.
The temperature inversion at the top of the mixed layer was obscured in the mean vertical profiles in Fig. 4a.To preserve the temperature inversion, we attempted another composite procedure: all soundings were normalized by the MLH, and after averaging, the vertical axis of the composite sounding was rescaled to the mean MLH (Norris 1998).The MLH was determined by the method proposed by Wang and Wang (2014).Their method to determine the MLH is complicated, and we do not fully describe it in this paper.In short, we locate the altitudes of the 10 smallest vertical gradients of four variables: potential temperature, relative humidity, specific humidity, and refractivity that combines the information of temperature and humidity.All 10 altitudes are considered to meet the criterion of the MLH.Basically, the MLH is defined as the altitude where at least three of the four variables meet the criteria simultaneously.In their method, cloud layers estimated from radiosonde data are also considered to determine the MLH.We avoided using the sounding data below 50 m when the MLH was calculated.The normalized composite shows a clear temperature inversion at approximately 900 m on the warm side (Fig. 4d).The mean values and standard deviations of the MLH and surface fluxes are listed in Table 1.For the cold side, the distribution of the MLH was biased toward the lower side (Figs.2c,  d), and the median of the MLH was 170 m, although the median on the warm side (890 m) was nearly the same as the mean value.The mean MLH of approximately 900 m with a thermal inversion on the warm side and the stable profiles on the cold side are consistent with previous soundings over the KE in winter     (Tokinaga et al. 2009) and the Brazil-Malvinas confluence region in spring (Pezzi et al. 2009), although SST around the Oyashio front was lower than these regions by more than 4°C.The MLH on the warm side tended to be higher under southerly conditions than under northerly conditions (Fig. 5), but the difference was not statistically significant.The surface turbulent heat fluxes and momentum flux show a clear contrast across the front.On the cold side, strong stability inhibited the air-sea transfer of heat and momentum.Furthermore, downward longwave radiation was larger in the warm areas than in the cold areas probably due to higher air temperature and humidity, leading to suppression of the contrast of net heat flux across the front.This is consistent with in situ observation results across the KE front by Kawai et al. (2015).
The cloud height clearly differed between the cold and warm sides of the Oyashio front (Fig. 6): the cloud base height peaked at 50 -100 m in the cold areas, whereas clouds were distributed at higher altitudes in the warm areas.This is consistent with the observations obtained over the KE by Tokinaga et al. (2009) and Kawai et al. (2015).We do not have enough cloudtype data for discussion, but stratus and stratocumulus clouds were frequently observed during the cruise.On the warm side, clouds with a base height of 700 -1000 m appeared mainly under southerly cases, and the height tended to be lower under northerly cases (Figs.6b, c), in harmony with the difference in the MLH (Fig. 5).Spall (2007) indicated by numerical experiments that vertical turbulent mixing over the warm side is stronger when wind blows from the warm-to-cold side compared with an opposite wind direction.

Thermally induced pressure anomaly
The vessel traveled nearly due east at around 41°00′ -41°30′N on the southern side of the Oyashio front from 0430 UTC 6 April to 2226 UTC 7 April (Period II) at a speed of approximately 15 knots; during this time, 21 radiosondes were launched at intervals of about 1.5 (3.0) h west (east) of 155°14′E (Fig. 7).Throughout Period II, the surface wind was constant at approximately 10 m s −1 and southeasterly (not shown).We traversed a local SST minimum at around 157°E; the SST difference on either side of the minimum was about 3°C (Fig. 7b).The zonal distribution of air temperature below 1000-m height mostly followed that of SST, but the air temperature peaks shifted slightly westward (Figs.7b, c), perhaps as a result of advection by the southeasterly winds.
The SST minimum produced a cold air pool between 154°E and 159°E with a width of approximately 400 km.During Period II, SLP was decreasing because of the approach of a cyclone, with short-term fluctuations of less than 2 hPa (Fig. 8).
A horizontal temperature gradient in the atmospheric boundary layer modifies SLP.The thermally induced pressure gradient at the surface P t can be estimated as follows (Mahrt et al. 2004;Kawai et al. 2014): where x is the horizontal distance and θ¢ is the devi- ation of virtual potential temperature from the basic state θ 0 that is not affected by surface cooling.To remove the effects of the large-scale temperature gradient, we set θ 0 to the sum of the along-track mean θ ( ) z and the along-track linear fitting of the near-surface virtual potential temperature averaged below 100-m height: that is, θ 0 (x, z) = θ ( ) z + ax + b, where a and b are the linear fitting coefficients (Kawai et al. 2014).The upper limit h of the integral in (1) represents the height below which all of the surface-induced cooling occurs; here, we set it to a constant value of 1000 m on the basis of the radiosonde-observed virtual potential temperature and relative humidity data (Figs.7c, d).The thermally induced component of SLP can be obtained by multiplying P t by air density and integrating it with respect to x.The surface cooling at around 157°E increased SLP by 1.2 hPa (Fig. 8b), an increase comparable to the SLP anomalies caused by the KE front (Kawai et al. 2014).This estimation cannot be done without such high-resolution in situ measurements.The local maximum of the detrended SLP can be quantitatively accounted for by the surface cooling, although the SLP also reflects other components such as atmospheric tides.The moisture also affects SLP, but in this case, its effect was negligible since the zonal gradient of water vapor was small (Fig. 7d).The mixed layer was locally higher over the high SST areas at 151.8°E and 160.0°E (Figs.7c, d), implying the vertical mixing mechanism.The high SLP anomaly over a cold SST area, however, means that the pressure adjustment mechanism was effective.Although it was impossible to examine surface wind divergence, these observation results suggested that both the mechanisms worked compatibly.

Atmospheric responses to ocean warm streamers
From 2359 UTC on 4 April to 1618 UTC on 5 April (Period I), the vessel sailed eastward at a speed of approximately 15 knots.During this period, streamers with SST higher than 11°C were observed (Fig. 9b) in an area with a highly complicated thermal structure.The presence of a warm streamer between 146.4°E and 147.0°E was confirmed in the infrared SST image (Fig. 9a), though not in the OISST (not shown).We launched 12 radiosondes during Period I, and the 11th radiosonde measured the atmosphere directly above the narrow warm streamer.The temperature and MLH above the streamer largely differed from those measured by adjacent radiosondes near the surface, and became less stable below 1000 m (Fig. 9c).The relative humidity above the streamer was more than 80 % between heights of 300 and 1000 m, much higher than that measured by the adjacent radiosonde west of the streamer, where the relative humidity was about 40 % (Fig. 9d).In addition, the wind speed in the boundary layer was locally intensified (Fig. 9e), an effect of the vertical mixing mechanism.
Ceilometer backscatter also became large locally below 1000-m height between 146.4°E and 170.0°E, and cloud bases were detected at heights of 450 -770 m at 146.6 -146.9°E (Fig. 10a).Another streamer was located at 144.0°E (Figs.9a, b).The relative humidity (Fig. 9d) and backscatter (Fig. 10a) between heights of 100 and 500 m around 144.0°E were higher than in neighboring regions, although these atmospheric imprints were weaker than those of the warm streamer to the east.The greater backscatter above the streamers indicates larger amounts of clouds and aerosol particles and suggests the intensification of updrafts.Cloud bases were not detected over areas other than 145.6 -145.9°E and 146.6 -146.9°E, and we did not observe clouds by eyes around 144.0°E.Therefore, the greater backscatter around 144.0°E over the western streamer  was due to aerosol particles rather than clouds.The downward longwave radiation was remarkably magnified over the western half of the streamer around 146.5°E (Fig. 10b) by the warm, humid near-surface air (Fig. 10c) and the presence of low-level clouds (Fig. 10a).On the other hand, over the eastern half of that streamer, the latent and sensible heat fluxes were extremely large instead of longwave radiation due to strong surface wind (Fig. 10d).The near-surface mixing ratio over the eastern half was not as large as over the western half: stagnated water vapor near the surface might have been transferred upward due to the intensified wind.The reason for the surface wind change is not clear due to lack of data, and the following is one possible explanation.The surface wind direction around that streamer was nearly southeast (Fig. 10d), which suggests that the background nearsurface pressure gradient was westward.This is not contradictory to the weather chart during Period I (Fig. 11a).If the sharp SST fronts on both sides modified the pressure gradient, the wind was intensified (weakened) on the eastern (western) side (Fig. 11b).We estimated the thermally induced SLP anomaly along the observation line and the corresponding wind speed anomaly across the line (Fig. 12).The large anomalies west of 143.5°E were due to the descent of warm air at an altitude of around 600 m (Fig. 9c), and unrelated to sea surface heating.The decline of wind speed west of the streamer can be partly accounted for by the local modification of the pressure gradient field.We consider that both vertical mixing and pressure adjustment were enhanced in this case.The change of wind We then attempted to estimate a time scale on which the atmosphere responds to that streamer.The differences in the amount of water vapor below 1080 m (MLH at the 11th radiosonde) and the surface latent heat flux between the 10th and 11th radiosonde positions were 2.1 kg m −2 and 164.4 W m −2 , respectively.If it is assumed that this water vapor difference was caused only by the difference in supply from the sea surface, it took 8.8 h.Kawai et al. (2015) observed around the KE that it took about half a day for the atmosphere to adjust to the SST front after the background wind direction reversed.The estimated time scale of approximately 9 h will be within a reasonable range.

Summary
Although previous modeling studies have indicated that the effects of the Oyashio front are imprinted in the subarctic atmosphere, these results had not been corroborated by in situ observations, unlike along the KE front.In April 2013, during a research cruise around the Oyashio front, a total of 103 radiosondes were launched to examine the atmospheric imprints of the front.Our observations have first shown the following local atmospheric responses to the Oyashio front: contrasts of the MLH and cloud base height across the front, a thermally induced SLP anomaly, and responses to oceanic streamers with a width of several tens of kilometers.It is almost impossible for satellite observations to capture them.Composite vertical atmospheric profiles showed that the stability and humidity below 1000-m height differed between the warm and cold sides of the front.The low-level atmosphere was stable on the cold side, but unstable and mixed on the warm side.In the atmosphere on the warm side, a sharp drop in the relative humidity at around 1000 m suggested that the mean cloud top was at that height.The cloud base heights measured by a ceilometer showed a frequency peak at 50 -100 m in the cold areas, whereas the clouds were distributed at higher altitudes in the warm areas.Turbulent heat  fluxes, momentum flux, and downward longwave radiation also showed differences across the Oyashio front.
We zonally traversed a local SST minimum with a width of approximately 400 km, where the ocean mixed layer depth is known to reach a local maximum, and we observed a large horizontal air temperature gradient below 1000-m height that corresponded to an SLP anomaly of 1.2 hPa.Our in situ observations confirm that the cold Oyashio front, like the warm KE front, can modify the SLP field.Seafloor topography affects the locations of the Oyashio front and the SST minimum (Isoguchi et al. 2006;Mitsudera et al. 2018), and also the low-level atmosphere, as Xie et al. (2002) have indicated for the East China Sea; however, the mechanisms are quite different from each other.
Furthermore, we captured examples of a low-level atmospheric response to warm ocean streamers.The narrow streamers moistened the air and increased the amount of aerosol particles and clouds; they thus affected even downward longwave radiation.For the streamer between 146.4°E and 147.0°E, the nearsurface atmospheric properties over the western half were quite different from those over the eastern half.The difference in surface wind on both the sides might be partly accounted for by the local modification of pressure gradient over the sharp SST fronts.The time scale of the atmosphere to respond to the streamer was estimated to be approximately 9 h.On the other hand, it appears that the atmosphere around the western end of the observation line was not strongly affected by high SST (Fig. 9).The atmospheric response to SST also depends on a large-scale condition: imprints of SST will be suppressed under a strong high pressure or disturbance.
SST is lower around the Oyashio front compared with other fronts such as in the KE, Gulf Stream, and Brazil-Malvinas confluence region.Contrasts of the MLH and cloud base height, however, have been clearly confirmed across the Oyashio front.Highresolution SST data are crucial for the accurate reproduction of atmospheric responses to SST fronts by numerical models.SST fields are particularly complicated in frontal areas, and actual SST differences within a few tens of kilometers can reach more than 10°C (Fig. 9b).The SST datasets widely used for atmospheric models, however, cannot capture these differences in frontal areas because of smoothing (Kawai et al. 2014(Kawai et al. , 2015)).Kawai et al. (2015) compared three kinds of objectively analyzed SST with those observed across the KE front by three research vessels over 5 days and found that none of the objec-tively analyzed SSTs successfully reproduced the fine frontal structure of the real SST.Recently, it has been recognized that meso-scale atmospheric responses to SST on a scale of several hundred kilometers can affect basin scale atmospheric circulations (Ma et al. 2015).On the other hand, there are no previous studies that focus on smaller scales yet, and one cannot know whether the air-sea interaction on a scale of a few tens of kilometers is really important for phenomena on a synoptic or basin scale without numerical experiments.This is beyond the scope of this study.A future study should investigate how complicated SST distributions on a scale of less than 100 km affect atmospheric simulation results.

Supplements
Supplement is the log of radiosonde observations conducted during the KH-13-3 cruise (including 16 observations around the KE front that were not analyzed in this paper).

Fig. 1 .
Fig. 1.(a) Ship track (white line) and radiosonde launch positions (dots: gray for in situ SST > 4.0°C; white for in situ SST ≤ 4.0°C) during the KH-13-3 cruise.The color scale shows the monthly mean OISST for April 2013.The thin and thick contour intervals are 1°C and 5°C, respectively.The radiosondes around the KE are not shown.(b) Horizontal gradient of the monthly mean OISST.The white line shows the 4°C SST contour (Oyashio front).

Fig. 2 .
Fig. 2. Individual vertical profiles of potential temperature for (a -b) SST > 4.0°C and (c -d) SST ≤ 4.0°C.Panels (a, c) and (b, d) show southerly and northerly cases, respectively.A red dot represents the MLH of each profile.

Fig. 3 .
Fig. 3. Individual vertical profiles of (a -b) relative humidity and (c -d) wind speed.Panels (a, c) and (b, d) show cases with SST > 4.0°C and SST ≤ 4.0°C, respectively.A red dot represents the MLH of each profile.

Fig. 4 .
Fig. 4. Composite vertical profiles of (a) potential temperature, (b) relative humidity, (c) wind speed, and (d) normalized temperature by the MLH.Solid and broken lines represent averages for SST > 4.0°C and SST ≤ 4.0°C, respectively.

Fig. 6 .
Fig. 6.(a) Frequencies of the lowest cloud base height north of 39.0°N counted in 50-m bins.Solid and dashed lines denote cloud base heights on the warm and cold sides of the Oyashio front, respectively.(b) Southerly and (c) northerly cases.

Fig. 7 .
Fig. 7. (a) Location of the section shown in (b) -(d) (red).The thin black line shows the entire ship track.(b) SST (blue) and surface air temperature (red) observed from 0430 UTC 6 April to 2226 UTC 7 April 2013 (Period II).(c) Virtual potential temperature (K) and (d) relative humidity (%) observed with radiosondes.Triangles at the bottom of panels (c) -(d) show radiosonde launch positions.

Fig. 8 .
Fig. 8. (a) Observed SLP and (b) detrended SLP (gray) during Period II.The dashed line in (a) is the linear fitting line.The bold black line and circles in (b) denote the thermally induced SLP anomaly estimated by Eq. (1).

Fig. 9 .
Fig. 9. (a) Pathfinder SST averaged over 4 -6 April 2013.White areas denote land or mean SST < 2.0°C.All data, including those with the lowest quality flag, were averaged, so only the spatial distribution pattern should be considered.The white line and dots denote the ship track and the radiosonde launch positions from 2359 UTC 4 April to 1618 UTC 5 April (Period I).(b) In situ SST (blue) and surface air temperature (red).(c) Virtual potential temperature (K), (d) relative humidity (%), and (e) wind speed (m s −1 ) observed with radiosondes.Triangles at the bottom of panels (b) -(e) show radiosonde launch positions.Red dots in (c) represent the MLH.

Fig. 11 .
Fig. 11.(a) Weather chart at 1200 UTC on 5 April 2013.(b) Schematic picture of the surface winds modified by the SST fronts of a warm ocean streamer.Upper white arrows represent surface winds without the modification by differential heating.Lower white arrows show the winds affected by the SST fronts.

Fig. 12 .
Fig. 12. Thermally induced SLP anomaly estimated by equation (1) along the observation line shown in Fig. 9a during Period I (black line).The blue line shows a geostrophic wind anomaly component perpendicular to the observation line corresponding to the SLP anomaly estimated on the assumption of no surface friction.