The preconditioning of winter mixed layer in the formation of the eastern subtropical mode water (ESTMW) in the North Pacific Ocean is investigated using an eddy-permitting ocean general circulation model. The result shows that convergence in the northward Ekman transport of saltier water and weak summertime heat fluxes due to the presence of stratus cloud in the ESTMW formation region are both central to the initiation of significant mixed layer deepening that subsequently evolves into the ESTMW pycnostad in the main thermocline. These two factors originate from air-sea interactions inherent in the region of Northeast Pacific Basin. The approach described here has significant potential for determining the water-mass distribution and may lead to an explanation of the Sverdrup's Eastern Gyral in the subtropical North Pacific.
 Since the discovery of the “Eighteen Degree Water” in the North Atlantic by Worthington , analogous water masses, referred to as subtropical mode waters (STMWs), have been identified in the North Pacific Ocean [e.g., Masuzawa, 1969]. Recent studies have increasingly highlighted the importance of accurate descriptions of the physical processes connected with STMWs. In fact, the transport of both heat and watermass in the upper North Pacific Ocean is largely characterized by the behavior of STMWs because they are actually distributed over wide areas as surface water masses with volumetric modes. Clarification of the formation and transport mechanisms responsible is, therefore, one of the central requirements in our understanding of the dynamic nature of the North Pacific climate.
 In the North Pacific Ocean, three STMWs have been identified so far: (1) the classical Western Subtropical Mode Water (WSTMW [Masuzawa, 1969]) formed just south of the Kuroshio front at a potential density of ∼25.2σθ (15–19°C); (2) the Central Subtropical Mode Water (CSTMW) at a potential density of ∼26.2σθ (9–12°C) between the Kuroshio Extension and Kuroshio Bifurcation fronts [Suga et al., 1997]; and (3) the Eastern Subtropical Mode Water (ESTMW) which was most recently identified on a 24.0–25.4σθ contour (16–22°C) in the Northeast Pacific Basin [Hautala and Roemmich, 1998, hereinafter HR98].
 Previous studies revealed that both WSTMW and CSTMW are formed in the localized area of the Kuroshio Extension region. However, in detail the formation processes differ from each other. In the western part of this region, strong convection produces considerable mixed-layer deepening by strong wintertime cooling due primarily to intense continental air flowing eastward over the warm western boundary current region. On the other hand, in the CSTMW formation region, [Ladd and Thompson, 2000, hereinafter LT00] showed that the deeper stratification is quite weak and that once the seasonal pycnocline has been eroded, it becomes susceptible to further penetration.
HR98 identified the ESTMW centered at (140°W, 30°N) in the Northeast Pacific Basin as a local potential vorticity (PV) minimum and confirmed that sea-surface heat fluxes during winter in this region are not strong enough to induce the observed deep mixed layer, in support of Talley . Typical wintertime heat loss reaches only 130 Wm2 in the ESTMW formation region, while losses in the WSTMW and the CSTMW regions are greater than 400 and 175 Wm2, respectively. This fact implies that convection capable of creating the actual ESTMW occurs only when combined with other priming mechanisms.
HR98 discussed several possible ways for this priming to take place. The role played by depression of the permanent thermocline associated with stronger southward transport than expected on the basis of a Sverdrup balance in the Northeast Pacific Basin was emphasized in reducing the stability of upper water columns. Though this lowering of deeper stratification could activate the wintertime convection, the relationship between the presence of such a thermocline structure and the ESTMW formation remains uncertain, because a depression of the permanent themocline is likely to result from mixed layer deepening (i.e., formation of the ESTMW). In this regard, the result of LT00 is particularly relevant. By comparing numerical experiments with various combinations of initial background stratification and forcing fields, they concluded that the summertime background stratification in the Northeast Pacific Basin is quite effective in forming deep mixed layers. Thus, the primary focus of our investigation of the preconditioning mechanism of the ESTMW formation is to clarify the physical process responsible for the locally weak stability of the shallow seasonal pycnocline for both summer and autumn.
LT00 used a one-dimensional model and ignored any three-dimensional effects. A number of other factors including advection and Ekman pumping can influence the mixed layer depth. Most recently, simulation experiments using Ocean General Circulation Models (OGCMs) [Ladd and Thompson, 2001; Hosoda et al., 2001] have emphasized the importance of the wide spacing between outcrop lines to the subduction of large volumetric modes (see also Wong and Johnson  on the South Pacific ESTMW). However questions still remain concerning the formation of the mode water because this depends on low PV water formation and the generation process responsible for this is still unclear.
 In this study, we use an eddy-permitting OGCM capable of offering greater information on actual water mass formation and motion than earlier models. In this fashion, we attempt more realistic simulations of the North Pacific circulation and place particular emphasis on preconditioning of the seasonal thermocline in ESTMW formation.
 The model covers the entire North Pacific Ocean (see Figure 1) with horizontal resolution of 1/6° zonally and 1/8° meridionally. There are 78 vertical levels, 62 of which are set in the upper 500 m with a finer resolution spaced from 4 m near the sea surface to 20 m. The initial values of potential temperature and salinity are taken from the World Ocean Atlas 1998 Monthly Data compilation (WOA98 [Conkright et al., 1998]), to which simulated values are restored in layers deeper than 2000 m and at the southern open boundary. Sea surface fluxes are calculated with the bulk formula [Ikeda, 1989] using the Ocean Model Intercomparison Project (OMIP) dataset [Röske, 2001], although in the uppermost 4 m, the commonly used flux correction method is adopted for heat and freshwater fluxes, with a relaxation time scale longer than 30 days. The model is forced by these climatological daily data, and is integrated until an almost steady seasonal state is obtained. Data from the last year of the simulation are used for the analysis below.
 Our experiment successfully represents realistic features of seasonally-varying circulation and mixed layer variabilities in the North Pacific. For example, Figure 1 shows the simulated wintertime (March) mixed layer depth (MLD) distribution. Since the MLD distribution has a significant impact on the water mass formation process [Huang and Qiu, 1994], its assessment forms an appropriate benchmark from which to judge the validity of our simulation result. In Figure 1, the mixed layer deepening is most pronounced in the Kuroshio Extension region, where it reaches a maximum depth of 300 m and reflects the formation of both WSTMW and CSTMW. Another prominent deepening down to 100–150 m depth is found near (140°W, 30°N) in the Northeast Pacific Basin, which represents the formation of ESTMW in this study. These features are in good agreement with observational findings (WOA98; HR98). Figure 2 shows the simulated potential temperature distribution in a vertical cross section between (160°, 20°N) and (120°W, 38°N) obtained for August. The data are similar to those shown in Figure 8c of HR98. For example, Figure 2 exhibits a shallow mixed layer of less than 50 m. Also, a low PV water formed in the previous winter is found between the seasonal and permanent thermoclines (about 80–150 m depth), with a spatial pattern that is in good agreement with the data presented in HR98. These results represent an encouraging validation of our model.
LT00 has suggested that the local deepening of the winter mixed layer in the ESTMW formation region arises from weak stratification in both the seasonal and permanent thermoclines in the preconditioning phase, of which the latter is largely attributed to the presence of the mode water formed previously. To identify the strength of the stratification in the seasonal thermocline before winter convection, we calculate the vertical difference in density between the sea surface and water of 100 m depth in November (Figure 3). This density difference is considered to be a good indicator of the seasonal change in the intensity of upper layer stratification because, in this region, water at 100 m depth lies below the seasonal pycnocline in summer and autumn and above it in winter and spring. Figure 3 shows the presence of a local minimum of stability in the ESTMW formation region. What causes the development of such a vertical structure in the upper regions of the water column?
Figure 4 shows the sea surface salinity (SSS) distribution in November. Most noticeable is the presence of a band-shaped SSS maximum region located at approximately 30°N. Excess evaporation over precipitation is the most likely cause of local maxima in SSS (roughly equivalent to the mixed layer salinity). The map of the freshwater flux in August estimated from the OMIP dataset (Figure 5) shows, however, that excess evaporation over precipitation is largest to the south around 20–25°N (more than 150 mm/month). The August salinity balance at this location is dominated by the freshwater flux term (∼3.5 psu s−1) and the advection term (∼−3 psu s−1). These facts suggest that northward transport from this region plays an important role in the production of the band-shaped SSS maximum.
 Our detailed examination reveals that vigorous evaporation around 20–25°N can be mainly attributed to the effect of intense northeasterly winds blowing from the US coast, and that salinity enriched water caused by the rapid evaporation is subsequently carried northward by Ekman flow associated with the strong easterly winds around 25°N (2–3 Sv between 160°W and 140°W), as is clearly shown in Figure 5. On moving northward, the salty water becomes even saltier by the additional effect of evaporation (Figure 5) and then converges within the band-shaped region (∼30°N) to eventually produce a salinity maximum. The mixed-layer salinity field in the eastern part of the band-shaped region (150–140°W, 30°N) is determined by the relative contributions of the surface freshwater flux (corresponding to a net evaporation of ∼3 psu s−1), advection (1 ∼ 2 psu s−1), and diffusion (∼−2.5 psu s−1). (It should be noted that surface relaxation to the climatology does not create the type of distributions shown in Figures 4 and 6. The restoring contribution to the mixed-layer salinity field, for example, remains at 0.5–1 psu s−1 and its seasonal variation is small.) This result is consistent with the results of Talley  who discussed the presence of significant Ekman transport convergence there. Associated with this northward transport, the SSS value in this region varies from 34.7 psu in May to 35 psu in November. Consequently, the value of SSS in autumn is larger by up to 0.8 psu than that at 100 m depth, leading to a weak stratification in the seasonal pycnocline. This situation is similar to that found in the WOA98 data. However, the band-shaped high SSS region is much broader than the formation region of ESTMW (centered on 140°W). Therefore some other factor arising from temperature effects on much weaker stratification is required to produce the localized formations in ESTMW.
Figure 6 shows the sea surface temperature (SST) distribution in November. SST values around 140°W in the salty band-shaped region are considerably lower (1–1.5°C) than those in the west, so that the temperature effect in reducing the stability of upper water columns at ∼30°N is much stronger to the east of 150°W than to the west. This additional effect means that intense convection can occur locally in the eastern part of the high SSS region caused even by the moderate and broad-scale wintertime heat loss inherent in the Northeast Pacific Basin. A robust comparison of the mixed layer heat balance reveals that colder SST values in the east are mainly the result of smaller short wave radiation fluxes relative to the western region. This difference arises from the presence of more extensive stratus cloud in the Northeast Pacific Basin (e.g., LT00), presumably due to the cold SST feedback suggested by Philander et al. . Composite satellite imageries support this fact as does the OMIP dataset. The weak seasonal stratification caused by reduced short wave radiation is consistent with the results of LT00. They focussed on meridional differences in the insolation while we emphasize the importance of its longitudinal variations.
Figure 7 summarizes the preconditioning mechanism proposed in this study. In summer and autumn, strong easterly winds centered on 25°N enhance evaporation and thereby enrich the salinity of the shallow mixed layer water. Subsequently, this saltier water is carried northward by Ekman flow associated with the easterly winds and then converges at a latitude of ∼30°N. Salinity contributions to the reduced stability of the summertime seasonal pycnocline remain, however, insufficient to cause the observed deep convection down to 130 m by moderate winter cooling that is inherent in the Northeast Pacific Basin. To the east of 150°W, the enhanced covering of stratus clouds due to air-sea interactions works to reduce the insolation and thereby produces much weaker stratification in the seasonal pycnocline. Consequently, the mixed layer deepening in the ESTMW formation region is enhanced, eventually forming the pycnostad in the permanent thermocline in winter, as is observed.
 Our modeling results provide, for the first time, a physical insight into the role of northward Ekman transport and subsequent convergence of salty water as a means of preconditioning the initial regime of significant wintertime convection localized in the ESTMW formation region. Furthermore, they provide a better approach to the understanding of the unknown mechanism of stronger southward transport than that derived from a Sverdrup balance within the upper 2000 m [Hautala et al., 1994], because the low PV input associated with the ESTMW formation has the tendency to produce an anticyclonic circulation. Figure 8 demonstrates the presence of an additional anticyclonic circulation (∼3 Sv) centered at (135°W, 27°N) in the eastern area of the Northeast Pacific Basin. The pattern and intensity vary seasonally and the basic features are thought to be in broad agreement with the Eastern Gyral explored by Sverdrup et al. [1942, Figure 205]. Our investigation suggests that an anticyclonic circulation is produced by similar dynamics to the PV-induced subsurface circulation proposed by Kubokawa and Nagakura , who investigated the wave dynamics induced by anomalous Ekman pumping using a 2.5-layer ventilated thermocline model. Also, the northward Ekman transport described here can create the front located to the north of the ESTMW, which is a common feature of the STMWs although in the case of both the WSTMW and the CSTMW the frontal boundary is associated with the Kuroshio Extension. Furthermore, the intense Ekman convergence and the resulting downwelling is consistent with the data reported in Talley . These facts imply that the major source of the ESTMW is the subtropical surface water.
 Our work is primarily a modeling study of one possible mechanism responsible for the formation of ESTMW. The relative importance of this proposed mechanism to other possible candidates for the preconditioning should be assessed in future intensive high-resolution observations. Such work contributes to our understanding of the dynamics of the Eastern Gyral and the actual role played by the formation and movement of ESTMW in intergyre exchange in the North Pacific Ocean. Both the close similarities found here with observational findings and the ability of our model to clarify aspects of the physics in the Northeast Pacific Basin highlight the importance of our simulation study.
 The authors wish to acknowledge Dr. J. P. Matthews for his critical reading and useful comments. Numerical calculations are done at the Academic Center for Computing and Media Studies of Kyoto University.