Cold-air outbreaks frequently occur during the winter season and are investigated by observational and numerical methods (e.g. Brümmer, 1999; Pagowski and Moore, 2001; Kawase et al., 2005; Ninomiya et al., 2006). The outbreak over the eastern coast of the cold Eurasia Continent strongly influences the wintertime weather in Japan. The northwesterly stream over the Japan Sea is maintained by the strong pressure gradient induced by the fully developed Siberian High and Aleutian Low. As the cold and dry wind blows over the warm Japan Sea, water vapor and heat are supplied from the sea surface to the atmosphere. The sea surface temperature (SST), which is controlled by the Tsushima Warm Current that transports southern warm water to the Japan Sea through the Tsushima Strait between the Korean Peninsula and Kyushu Island (around 34°N, 128°E), is one of the most important meteorological factors that determines the wintertime weather in Japan (e.g. Hirose and Fukudome, 2006). In particular, as surface-sensible and latent fluxes are high in coastal areas (e.g. Yoshizaki et al., 2004; Ninomiya et al., 2006), the accurate estimation of coastal SST is critically important in determining the thermal and moisture budgets.
During cold-air outbreaks that lead to the widespread and continuous formation of clouds, however, it is difficult to estimate high-resolution SST in coastal areas because the cloud cover over the wintertime Japan Sea is problematic in terms of obtaining high-resolution IR data. Although microwave observations can be used to measure SST in all-weather conditions except for rain, the microwave imagers have a wider field of view (i.e. a lower resolution) than that of IR imagers. In addition, the effects of interpolation and deficiencies in the data cannot be neglected in satellite-based measurements of SST. To overcome these problems, SST data are improved by combining satellite measurements of not only SST but also sea surface height (SSH) into an eddy-resolving ocean model (e.g. Manda et al., 2005; Hirose et al., 2007). This assimilated product provides high-resolution SST data for the cloudy Japan Sea area and is consistent with the high-resolution eddies detected from SSH observations.
Yamamoto and Hirose (2007) used the high- resolution SST data assimilated using an ocean general circulation model (OGCM) to simulate a developing cyclone passing over the Japan Sea, and reported that the difference between the assimilated and interpolated SSTs could not be neglected in their simulation since it significantly influenced the cyclogenesis. In addition to an extratropical cyclone, cold-air outbreak also plays an important role in the weather system of the East Asia marginal seas. In the present study, we focus on a cold-air outbreak after a cyclone passed over the Japan Sea. Based on comparative simulations undertaken using assimilated and interpolated SSTs, we demonstrate that the simulation of the cold-air outbreak is improved by using the high-resolution assimilated SST.
2. Data and model
Two SST products are used in this study; one a direct interpolation from satellite SST measurement, the other an estimation from a model assimilating SST and SSH measurements. A daily digital map of New Generation Sea Surface Temperature (NGSST) for the open ocean ver.1.0 provided by Tohoku University, Japan, is used as the optimal interpolation product for satellite-measured SST obtained using infrared and microwave radiometers (Guan and Kawamura, 2004). Although NGSST is a high-quality and high-resolution (0.05°) dataset produced via an optimal interpolation scheme, the spatial and temporal interpolation with decorrelation scales of 200 km in latitude/longitude and 5 days in time influences the SST distribution in coastal areas of the wintertime Japan Sea. The assimilation of satellite SST and SSH data is undertaken using a Japan Sea ocean model produced at the Research Institute for Applied Mechanics (RIAM), Kyushu University, Japan (Hirose et al., 2007). The assimilated SST obtained from the model output (hereafter termed RIAMSST) has a 1/12° resolution in longitude/latitude. We can find significant differences between the two sets of the SSTs in Figure 1. NGSST is smoothed by spatiotemporal interpolation, while RIAMSST has mesoscale structures in oceanic front and coastal area.
We examine the influences of the two different SSTs (RIAMSST and NGSST) on a cold-air outbreak over the Japan Sea on 20–22 January 2005, as simulated using the PSU/NCAR mesoscale model (MM5V3, Grell et al., 1995) with 23 sigma layers. The two experiments are termed Exp. N and Exp. R for the NGSST and RIAMSST cases, respectively. The mother domain with a central latitude of 35°N and longitude of 135°E has a 103 × 121 horizontal grid resolution of 30 km; the two-way nested domain with 10 km resolution (130 × 130 grid) is located in the Japan Sea area. The initial (00 : 00 UTC, 20 January 2005) and boundary conditions were calculated from NCEP/GDAS data (ds083.0). The model setup is described more fully in Yamamoto and Hirose (2007).
Following the passage of a developing cyclone (located near 42°N, 138°E at 00 : 00 UTC on 20 January 2005) through the Japan Sea area, a cold-air outbreak was maintained for about 2 days. A strong northwesterly wind was predominant over a large part of the Japan Sea area at 00 : 00 UTC on 21 January, shifting to the eastern part of the sea at 00 : 00 UTC on 22 January.
Figure 2(a) and (b) show the horizontal distributions of upward surface-turbulent heat fluxes and surface horizontal winds after 24 h (00 : 00 UTC on 21 January). The cold-air outbreak has northwesterly surface winds of more than 10 m s−1. The Japan Sea Polar air-mass Convergence Zone (JPCZ) is formed near the east coast of the Korean Peninsula (e.g. Nagata et al., 1986).
As the distributions of surface winds are similar for the two experiments, the differences in the surface-turbulent heat fluxes are mainly caused by differences between the two SSTs. In Exp. R, we found mesoscale distributions of high heat fluxes in the central Japan Sea (40°N, 133°E). This occurred because strong northwesterly winds blow over mesoscale oceanic eddies (Figure 1(b)).
The surface-turbulent heat fluxes in both experiments are low in the JPCZ, being less than 200 W m−2 in the area of 40°N, 130°E, since the horizontal wind speed becomes small along this convergence zone where upward motion is predominant. In contrast, we found that there are high heat fluxes near the north coast of the Japan Sea (42°N, 131°E) and the Tsushima Strait (34°N, 128°E). High sensible fluxes of > 500 W m−2 are observed in the coastal area of 42°N, 131°E in both experiments (not shown). The sensible fluxes in Exp. N are more than 150 W m−2 higher than those in Exp. R in the coastal areas, where NGSST is more than 4 K higher than RIAMSST. In the north coast of the Tsushima Strait, the sensible and latent fluxes in Exp. N are ∼150 and ∼200 W m−2 higher than those in Exp. R., respectively. Since SST in the narrow strait is smoothed by temporal and spatial interpolation in Exp. N, high SST (NGSST-RIAMSST ≈ 5 K) in the northern part of the strait leads to the high heat flux. We emphasize that the differences between the interpolated and assimilated SSTs have a significant influence on the estimates of heat and moisture fluxes at narrow straits such as the Tsushima Strait.
Figure 2(c) and (d) show the upward surface-turbulent heat fluxes in Exp. N and Exp. R, respectively, after 48 h (00 : 00 UTC on January 22), once the strong outbreak has shifted to the eastern Japan Sea area. The convergence zone of surface winds is located along the 140°E meridian between 42 and 45°N. Close to the 139°E meridian, we find the influence of the two different SSTs on the surface-turbulent heat fluxes from the coastal sea. The fluxes in Exp. N are higher than those in Exp R in the coastal regions of 41°N and 139°E where NGSST > RIAMSST. The large flux area of > 900 W m−2 is spread widely and continuously in Exp. N, whereas Exp. R records the inhomogeneous distribution of heat fluxes resulting from oceanic mesoscale eddies.
For the vertical structures of cold-air outbreaks at three locations (Akita, Wajima, and Yonago), the differences between observation and simulation can be evaluated by root mean squared differences of wind speed V, potential temperature θ, and water vapor mixing ratio q in Table I. Although the results are similar for both simulations, the differences of the wind speeds in Exps. N and R for 24 h forecast are quite substantial at Akita and Wajima. The root mean squared differences in Exp. N are 1.7 and 1.9 times larger than those in Exp. R at Akita and Wajima, respectively. Figure 3(a) shows vertical profiles of horizontal wind speeds recorded at Wajima Station (37.4°N, 136.9°E). The wind speed simulated in Exp. R on 21 January 2005 agrees well with the observed data, but the minimum wind speed at 700 hPa is not reproduced in Exp. N. The strong northwesterly wind is weakened by upward flow near the observation site (37.4°N, 136.9°E) in Exp. R (Figure 3(c)). In contrast, since the upward flow in Exp. N is weaker than that in Exp. R at 700 hPa, the strong northwesterly wind in Exp. N is only partly weakened by the upward motion near the observation site (Figure 3(b)). This leads to the differences between the vertical structures of cold-air outbreak in Exps. R and N. The inhomogeneous distribution of SST influences the location and development of convective motions in the coastal area. The differences in convective activity induced by the two SSTs lead to the significant differences in mesoscale circulation and outbreak structure near Wajima.
Table 1. Root mean squared differences of wind speed (V), potential temperature (θ), and water vapor mixing ratio (q) for the levels of 1000–200 hPa at Akita, Wajima, and Yonago in Exps. R and N
Akita (39°43′N, 140°06′E)
Wajima (37°24′N, 136°54′E)
Yonago (35°26′N, 133°20′E)
V (24 h),
V (48 h),
θ (24 h),
θ (48 h),
q (24 h),
2.88 × 10−4
1.88 × 10−4
2.07 × 10−4
2.10 × 10−4
3.14 × 10−4
1.74 × 10−4
q (48 h),
7.50 × 10−4
7.26 × 10−4
1.51 × 10−4
1.47 × 10−4
7.18 × 10−4
7.03 × 10−4
Figure 4 shows the horizontal distributions of integrated cloud-water amount and an infrared image of clouds. A north–south cloud band along 140°E is simulated for both experiments. In addition, a cloud region near 37°N, 138°E is also predicted in Exp. R (red circle in Figure 4(b)), though the area is somewhat smaller than the observational one. In contrast, this is not reproduced in Exp. N. This cloud is formed by the vertical wind (Figure 3(c)), which leads to the weakness of the horizontal flow at the Wajima site in Exp. R (Figure 3(a)). Since the surface-latent heat flux (i.e. the surface moisture flux) in Exp. R is ∼100 W m−2 higher than that in Exp. N in the upstream sea area around 40°N, 135°E, where RIAMSST is ∼2 K higher than NGSST, the rich moisture flow enhances the convection with strong vertical flow in the downstream coastal region. This implies that the mesoscale SST structure significantly influences the coastal convective activity in the cold-air outbreak.
We conducted comparative simulations of a cold-air outbreak using interpolated and assimilated SSTs for the Japan Sea area. A cold-air outbreak with strong northwesterly wind is maintained over the sea area, and the JPCZ is formed near the east coast of the Korean Peninsula. The upward surface-turbulent heat fluxes (sensible and latent heat fluxes) associated with the outbreak are influenced by the SST differences. Mesoscale oceanic eddies reproduced in the assimilated SST lead to the mesoscale fluctuations of the heat fluxes in Exp. R. The differences in the surface heat flux magnitudes induced by the two SSTs are most pronounced in the Tsushima Strait and coastal areas of the Japan Sea.
Although the horizontal distributions of the surface-turbulent heat fluxes differ markedly between Exp. N and Exp. R, the distributions of surface winds are similar in the two experiments, since the high-resolution SST does not largely influence the air-sea momentum exchange (differently from the heat exchange) and surface pressure gradient force (i.e. sea-level pressure) in the cold-air outbreak. The vertical structures of the simulated northwesterly wind are also similar to those revealed in aerological data; however, the SST difference influences the convection in the coastal area. The difference in convective activities between the two experiments strongly influences the cold-air outbreak at 700 hPa and cloud distribution via the strong vertical flow on 21 January 2005. The results in Exp. R agree well with the aerological observation and IR cloud image in comparison with those in Exp. N.
Yamamoto and Hirose (2007) demonstrated that the SST assimilated using OGCM improved the cyclogenesis simulation in a marginal sea where it is difficult to accurately estimate high-resolution SST because of continuous and widespread cloud cover in the winter season. As the next step, we focused on the cold-air outbreak influencing wintertime weather in Japan. The assimilation of ocean data, which combines not only SST but also SSH into an OGCM, is expected to successfully capture high-resolution variations in SST in order to accurately evaluate heat and moisture budgets in narrow straits and coastal areas where remote-sensing observations are deficient. In addition, the high-resolution SST structure influences local convective motions in the coastal area. Thus, it is necessary to conduct further investigations to determine if weather predictions are improved for other cases and to confirm the validity of these predictions.
New Generation Sea Surface Temperature for Open Ocean Ver.1.0 data were provided by the NGSST development group (http://www.ocean.caos.tohoku.ac.jp/). We used aerological data provided by the Japan Meteorological Agency and the IR images provided by the Japan Meteorological Satellite Center. The NCEP/GDAS data were sourced from CISL's research data archive (http://dss.ucar.edu/).