4.1. Case of July 2003
 The model domain is set to be 24°–67°N and 120°–170°E, which covers the Okhotsk Sea. The lateral boundary condition is set using 6 hourly zonal wind, meridional wind, temperature, specific humidity, and cloud liquid water (Qc) from JRA25 data. These lateral boundary conditions are linearly interpolated in time at every time step. National Oceanic and Atmospheric Administration (NOAA) optimum interpolation sea surface temperature (OISST, daily), with an original resolution of 0.25° × 0.25°, is used as the lower boundary condition over the ocean. We integrate the model from 00:00 UTC on 1 January 2003 for 1 year with the initial condition based on JRA25 data. Note that no nudging or data assimilation is performed in this model, except for the boundary conditions. The initial condition of the land process model, BATS, is horizontally uniform; however, the land surface temperature is set to be the atmospheric temperature at the lowest level. Hereafter, this experiment is referred to as the control run (CR).
 Figures 3a and 3b show the monthly mean SLP of JRA25 and CR, respectively, in July 2003. The Okhotsk high covers the entire Okhotsk Sea in both CR and JRA25, although its amplitude is slightly larger in CR. The Okhotsk high also tends to be slightly stronger in CR than in JRA25 in other years (not shown). Figures 3c and 3d illustrate vertical-meridional sections of the geopotential height anomaly from the meridional mean (between 35° and 60°N) at 150°E of JRA25 and CR, respectively, in July 2003. A positive anomaly is detected between 46°N and 58°N below 600 hPa, indicating that the Okhotsk high is limited to the lower troposphere in both JRA25 and CR. As in the case of SLP, the positive geopotential height anomaly is slightly larger in CR than in JRA25. The July 2003 mean vertically integrated Qc amounts from SSMI and CR are shown in Figures 3e and 3f, respectively. Although the integrated Qc amount is slightly larger in CR than in SSMI, the distribution is qualitatively similar; the maximum is along the Kuroshio Extension (30°–35°N and 140°–155°E), and the Qc amount gradually decreases northward, with a slight increase along the northwestern coast of the Okhotsk Sea. Figure 3g illustrates the time sequence of the area-averaged SLP (45°–60°N and 140°–160°E) in CR (solid line, daily) and JRA25 (dashed line, 6 hourly) in July 2003. SLP increases and reaches its maximum on 11 July and varies significantly after 16 July. As can be seen in Figures 3a and 3b, the SLP is slightly higher in CR than that in JRA25. Nevertheless, we believe that the model represents the time variation of the Okhotsk high reasonably well because the difference is partly due to the presence (or absence) of the low-level clouds in CR (or JRA25), as is discussed in section 4.2.
Figure 3. Sea level pressure of (a) JRA25 and (b) CR. Height-latitude sections of the geopotential height anomaly from the meridional average (35°–60°N) at 150°E of (c) JRA25 and (d) CR. The contour interval is 10 m, and positive and negative values are denoted by solid and dashed lines, respectively. Also shown is vertically integrated cloud liquid water of (e) SSMI and (f) CR. Figures 3a–3f show monthly mean values in July 2003. (g) Time sequence the area-averaged sea level pressure (45°–60°N and 140°–160°E) in JRA25 (dashed line, 6 hourly) and CR (solid line, daily) in July 2003.
Download figure to PowerPoint
 Figures 4a and 4b illustrate the time-height sections of temperature and cloud amount averaged over the Okhotsk Sea in CR. Figure 4a shows a strong inversion layer around σ0.96, which is a typical feature of a cloud top boundary layer (CBL) [Sedlar and Tjernström, 2009]. The vertical structure of the temperature is relatively stable in the first half of the month (before 16 July) during which the Okhotsk high strengthens before weakening gradually (Figure 4c). In this period, the cloud amount is largest around σ0.97, above which the inversion layer is located. The low-level clouds reach the surface and are accompanied by a weak drizzle (up to 1–2 mm d-1) over the Okhotsk Sea (not shown). On the contrary, the inversion layer at the lower levels almost disappears and temperature fluctuates greatly in the second half of the month (after 16 July). Clouds form at higher levels, and low-level cloud amounts become lesser than that seen in the first half of the month. This rise in cloud height and fluctuations in temperature are caused by synoptic-scale cyclones that pass over the Okhotsk Sea in the second half of the month. Although the Okhotsk high occurs in the same period, it is distorted and shifted northward by the synoptic-scale cyclones (not shown). We, therefore, focus on the first half of the month when both the Okhotsk high and low-level clouds are relatively stable.
Figure 4. Time-height sections of (a) temperature (K) and (b) cloud amount (percent) averaged over the Okhotsk Sea (45°–60°N and 140°–160°E) in CR. (c) Time series of the area-averaged sea level pressure in CR. (d) Temporal mean (1–16 July) cloud amount at σ0.977 in CR.
Download figure to PowerPoint
 Figure 4d shows the time-averaged (1–16 July) cloud amount at σ0.977 in CR, which is estimated from the sum of mixing ratios of cloud liquid water and cloud ice water at this level and thus should not be compared directly with ISCCP low-level cloud data. Most of the Okhotsk Sea is covered by low-level clouds. Two local maxima of low-level cloud amount are located near the eastern part of the Kuril Islands (46°–50°N and 150°–156°E) and Shelikhov Bay (55°–60°N and 155°–160°E). In these regions, the oceanic tidal mixing is strong and SST insignificantly lower than that in other regions of the Okhotsk Sea [Nakamura et al., 2000], leading to enhanced low-level cloud occurrence [Tokinaga and Xie, 2009]. CR captures the response to the oceanic tidal cooling effects through the prescribed SST. Conversely, the low-level cloud amount is relatively small north of Sakhalin Island (53°–56°N and 138°–144°E) and northeast of Hokkaido Island (44°–49°N and 142°–147°E). The cause of this cloud distribution is investigated in section 5.
 Figures 5a–5d show time-height sections of the area-averaged heating rates attributed to longwave radiation (LW), large-scale condensation (LSC), turbulent mixing (TM), and shortwave radiation (SW). The cooling caused by LW is strong around σ0.97 near the cloud top. The absolute value of the LW cooling, whose contribution to the heat budget is the largest, is about 60% larger than that of LSC heating, which is the second-largest contributor. The strong LW cooling can enhance the Okhotsk high near the surface because colder air is denser and shallower. While the LW cooling rate in the present case is smaller than that attributed to stratus clouds over the North Sea [Nicholls, 1984] and Arctic Ocean [Curry, 1986], it is higher than that caused by stratocumulus clouds over the subtropical ocean [Wang et al., 2004a]. The LSC heating rate is strongest around σ0.97, and its distribution also coincides approximately with that of cloud amount. In contrast, cumulus convection hardly occurs over the Okhotsk Sea in CR because of the low SST, which reaches only ∼283 K. LSC causes cooling near the surface through evaporative cooling of drizzle and cloud water. In this case, this evaporative cooling tends to stabilize the CBL because the cloud base is close to the surface [Stevens et al., 1998]. This cooling is also favorable for the surface Okhotsk high. TM induces significant heating near the surface and weak cooling above σ0.96. A part of the latter cooling represents the effects of entrainment, which plays an important role in the formation of stratocumulus clouds [Stevens et al., 2003] and is included in TM. The heating due to SW is positive; however, its amplitude is significantly smaller than that in LW, LSC, and TM.
Figure 5. Time-height sections of the area-averaged (45°–60°N and 140°–160°E) heating rates attributed to (a) longwave radiation, (b) large-scale condensation, (c) turbulent mixing, and (d) shortwave radiation. (e) Sum of heating rates (Figures 5a–5d). (f) Area-averaged temperature tendency in CR. Solid contours represent the area-averaged cloud amount (contour interval is 20%).
Download figure to PowerPoint
 Figure 5e shows the sum of the four heating rates. Heating is dominant near the surface, while cooling is strong around σ0.96. The heating near the surface is caused mainly by TM, which is almost balanced by the cooling due to LSC. The net cooling occurs around σ0.96 because of the larger absolute value of LW cooling than heating due to LSC and SW. LW cooling is thus important for maintaining the CBL at a low temperature. Figure 5f shows the temperature tendency. In CBL, the tendency is significantly smaller than the sum of the four heating rates, and thus the temperature in CBL is almost stable. This is because heating by vertical advection due to subsidence associated with the Okhotsk high almost balances with the sum of the four heating terms due to low-level clouds, although it is less than 15% of LW cooling (not shown). Above CBL, temperature tendency is relatively large. At these levels, thermodynamic effects due to low-level clouds are quite small and horizontal advection has a vital role in determining temperature tendency (not shown).
 The upward sensible and latent heat fluxes are found in this period with temporal and spatial mean fluxes of 10.8 and 18.7 W m–2, respectively, over the Okhotsk Sea. Tachibana et al.  showed positive heat fluxes in the presence of low-level clouds with the shipboard observations. In addition, they suggested that the combination of these upward turbulent fluxes and the cloud top radiation cooling maintains the CBL and effectively forms the low-level clouds.
4.2. Thermodynamic Effects of Low-Level Clouds on Cloud Top Boundary Layer
 In section 4.1 we suggested that strong cooling at the tops of low-level clouds is favorable for strengthening the surface Okhotsk high. To investigate this effect, we conduct sensitivity experiments, in which we reduce the effect of cloud amount on radiation, which is achieved by multiplying cloud amount by a constant factor γ (0 < γ < 1) when the radiation flux is calculated. This method is almost equivalent to the reduction of heating and cooling effects due to LW and SW associated with clouds. Note that the cloud amount values used in other processes are unchanged by this reduction at the same time step. In addition, this reduction does not affect the estimation of cloud droplet effective radius and cloud optical thickness. Nevertheless, because the values are calculated online, the change in radiation flux results in changes to other variables, which in turn feed back to the radiation. We conduct two sensitivity experiments, in which γ is set to be 0.5 and 0.1, and refer to these experiments as reduced cloud runs (RCRs; RCR05 and RCR01, respectively). Other conditions are the same as those in CR. We explore the radiative effects of low-level clouds by comparing the results of RCRs and CR.
 Figures 6a–6d show the time- and area-averaged vertical profiles of heating rate due to LW, LSC, TM, and SW in CR (solid line), RCR05 (dot-dashed line), and RCR01 (dashed line). The cooling due to LW weakens significantly in RCR05 and almost disappears in RCR01. The heating due to LSC and TM also diminishes with decreasing γ, from CR to RCR05 to RCR01. This result implies that Qc is hardly generated, and the upward turbulent heat fluxes are weakened over the Okhotsk Sea in RCRs. As previously described, we reduced only the effects of cloud amount on the radiative process, and Qc is not reduced by this change in the model code. The reduced Qc in RCRs indicates that low-level cloud formation is not effective in RCRs. Although the heating due to SW almost disappears in RCRs around σ0.97, its relative importance increases near the surface as Qc decreases. Nevertheless, we can neglect the contribution of SW heating in this study because in CR, SW heating is considerably smaller than other heating terms. Figures 6e and 6f illustrate the time- and area-averaged vertical profiles of temperature and virtual potential temperature (θv), respectively, in CR (solid line), RCR05 (dot-dashed line), and RCR01 (dashed line). In CR, a strong inversion layer associated with the CBL is detected between σ0.977 and σ0.96 (Figure 6e), as described in section 3. As the cloud amount is reduced in RCRs, the temperature below σ0.977 increases remarkably. The largest temperature difference between CR and RCR01 is ∼4 K at σ0.977. Because the radiative cooling is dominant at this level in CR, the temperature difference is mainly caused by a reduction in radiation effects.
Figure 6. Time (1–16 July) and area-averaged (45°–60°N and 140°–160°E) vertical profiles of heating rates due to (a) longwave radiation, (b) large-scale condensation, (c) turbulent mixing, and (d) shortwave radiation and those of (e) temperature and (f) virtual potential temperature in CR (solid lines), RCR05 (dot-dashed lines), and RCR01 (dashed lines).
Download figure to PowerPoint
 The values of θv in CR (Figure 6f) are approximately constant between the surface and σ0.977, indicating the development of a mixed boundary layer. With low-level clouds, the CBL is warmed at the surface by turbulent mixing and cooled near the cloud top by LW, as shown in Figure 5e. These heating and cooling cycles tend to maintain the well–developed CBL, resulting in a vertically uniform θv. In RCRs, the boundary layer is stabilized, and θv increases vertically because both the cooling by LW and warming by turbulent mixing are reduced.
 Figures 7a–7c show the time-averaged horizontal distribution of SLP in CR, RCR05, and RCR01. The Okhotsk high is weakened in RCR05 and diminishes in RCR01 drastically; in particular, the area including SLPs higher than 1018 hPa becomes much smaller around the center of the Okhotsk Sea. In addition, SLP is also reduced over the North Pacific Ocean near the Kamchatka Peninsula. It should be noted that the SLP difference between CR and RCR01 is ∼2–3 hPa and is similar to that between CR and JRA25. As the Okhotsk high weakens, the easterly Yamase winds in northern Japan become weaker and suppressed in RCRs (not shown).
 Figures 8a and 8b show the time-mean sea surface sensible heat flux in CR and RCR01, respectively. Although the season is summer, the sensible heat flux in CR is upward over most of the Okhotsk Sea. This upward flux is large in the eastern part of the Okhotsk Sea, where the low-level cloud amount is also large (Figure 4d). In contrast, the sensible heat flux is downward in RCR01 over most of the Okhotsk Sea. This switch in the direction of surface sensible heat flux is attributed to the warmed surface atmosphere owing to suppressed LW cooling in RCR01. Figures 8c and 8d show the time-mean sea surface latent heat flux in CR and RCR01, respectively. Although the direction of the latent heat flux is the same, its amplitude is significantly weaker in RCR01. The sensible and latent heat fluxes in RCR01 are –5.8 and 9.4 W m–2, respectively, when averaged over time (1–16 July) and area (45°–60°N and 140°–160E°). The differences (CR minus RCR01) are 153 and 50% of the values in CR (see section 4.1).
Figure 8. Time-averaged (1–16 July) sensible heat flux in (a) CR and (b) RCR01 and latent heat flux in (c) CR and (d) RCR01. Positive values denote upward sensible and latent heat fluxes.
Download figure to PowerPoint
 It should be argued that the stochastic response due to synoptic-scale disturbances is insignificant in the previous analysis. In these experiments, the lateral boundary condition largely controls the synoptic and larger-scale atmospheric fields, including the Okhotsk high. In addition, the Okhotsk high persists and synoptic cyclones do not pass over the Okhotsk Sea in the analysis period. Actually, the SLP decrease in RCRs compared with CR is detected every day in the analysis period, which indicates that the response is not stochastic. Moreover, the RCRs show an almost linear response to decreasing γ, as previously described. We thus consider that the effects of stochastic response are small in the present model setting.