For the first time, using a high-resolution atmosphere-ocean coupled general circulation model (CGCM), we succeed in reproducing the far-reaching effects of the Hawaiian Islands, recently showed by satellite observations. The model reproduces the distributions of sea surface temperature (SST), surface winds and cloud liquid water (CLW) in the wake of the Hawaiian Islands. It is revealed that these distributions are caused by the Hawaiian Lee Counter Current (HLCC) and that this current is driven by the wind-curls induced by the orographic effect of the islands, as suggested from an observational study. It is also shown that wind changes around the Hawaiian Islands can further affect the speed of the North Equatorial Current (NEC) and SST over the current, and intra-annual variability in CLW to the west of the islands is governed, not only by SST but also by wind speed.
 Recently, satellite observations showed that the wake of the Hawaiian Islands, characterized by a zonal band of high sea surface temperature (SST) and co-located meridional surface wind convergence together with high cloud liquid water (CLW), could trail more than 3000 km to the west of the islands [Xie et al., 2001]. The extent of the Hawaiian lee wake is many times longer than that estimated by a theoretical study [Smith et al., 1997].
Xie et al.  explained these far-reaching effects of the Hawaiian Islands by an air-sea interaction mechanism, resulting from an eastward ocean current, called the Hawaiian Lee Counter Current (HLCC) [Flament et al., 1998]. In the low latitudes of the North Pacific, since the upper ocean is warmer in the western region, the HLCC advects warmer water from the western basin. This advected warmer water forms an anomalously high-SST band to the west of the islands, and consequently, surface-wind convergence and a high-CLW band appear over the high-SST band.
Xie et al.  also proposed a theory of HLCC formation as follows: as an orographic effect of the islands under the northeasterly trades, a dipole wind-curl is induced in the lee of the islands. These wind-curls make the depth of the ocean thermocline shoal northward by Ekman pumping. This thermocline slope propagates westward as baroclinic Rossby waves, and is thereby maintained over a long zonal distance to the west of the Hawaiian Islands. Simultaneously, an eastward current, that is the HLCC, appears in geostrophic balance with the thermocline slope. This explanation means that the Hawaiian Islands, although composed of small islands, trigger and sustain this air-sea interaction, which extends far to the west.
 Recent studies provide evidence in support of some elements of the above interaction processes. Xie et al.  and Yu et al.  showed that the local wind-curls around the islands created the HLCC using numerical ocean models, constrained by the wind stresses from the European Center for Medium-range Weather Forecast (ECMWF). Hafner and Xie , using a fine-resolution regional atmospheric model, showed that SST west of the islands maintained the high-CLW band. However, the complete air-sea interaction loop proposed by Xie et al.  has not yet reproduced in an atmosphere-ocean coupled model. As Xie et al.  and Hafner and Xie  pointed out, surface heat flux does not favor the maintenance of the high-SST band west of Hawaii. Therefore, it is necessary to confirm that warm water advection by the HLCC is required to maintain the high-SST band in an atmosphere-ocean coupled system.
 Not only the effects of the Hawaiian Islands, but also most of the air-sea interactions over cool oceans, that is not associated with atmospheric deep convections, have not been successfully reproduced in the hitherto atmosphere-ocean coupled models. The main reason is that the spatial resolution and the computer power have not been sufficient. However, the air-sea interactions over cool oceans have been revealed from the satellite observations recently [e.g., Xie, 2004], and it has been considered as a key to understand long-term climate variability. Therefore, the success of reproducing the cool ocean-atmosphere interaction is required for the climate models to improve the accuracy of the future climate projection.
 In the present study, using the Earth Simulator that is an upper-most powerful super-computer, it is shown that the reproduction of the Hawaiian effects in a high-resolution global coupled general circulation model (CGCM), and it is revealed that the existence of the Hawaiian Islands is the trigger for all the air-sea interaction processes west of the islands. Also investigated was the effect of the islands on the Western Pacific region and the intra-annual variability of CLW west of Hawaii.
2. Model and Data
 The CGCM used in the present study is Model for Interdisciplinary Research on Climate (MIROC) version 3.2, which has been developed by the Center for Climate System Research, University of Tokyo (CCSR), National Institute for Environmental Studies (NIES), and Frontier Research Center for Global Change (FRCGC). This model consists of five components; atmosphere, land, river, ocean and sea ice. The atmospheric component is based on CCSR/NIES AGCM 5.7b [Numaguti et al., 1997] and is a T106 global spectral model with 56 vertical sigma levels. The land component is the Minimal Advanced Treatments of Surface Interaction and RunOff (MATSIRO) [Takata et al., 2003] which has a 0.56° × 0.56° resolution. The river component is the Total Runoff Integrating Pathways (TRIP) [Oki and Sud, 1998] whose resolution is 0.5° × 0.5°. The ocean component is the CCSR Ocean Component Model (COCO) [Hasumi, 2000] in which horizontal resolution is zonally 0.28° and meridionally 0.19° with 48 vertical levels. The sea ice component is based on a 0–layer thermodynamics model [Semtner, 1976] and an elastic-viscous-plastic rheology [Hunke and Dukowicz, 1997] which has the same resolution as the ocean component.
 In the control-run, the atmospheric component was coupled with the ocean and sea ice coupled components, which were in a state attained after spin-up for 16 years from the state of rest, based on the climatological stratification [Levitus and Boyer, 1994; Levitus et al., 1994]. While spun-up, the ocean and sea ice components are constrained by Ocean Model Inter-comparison Project (OMIP) datasets [Röske, 2001]. This control-run was continued for 56 years after coupling the atmosphere and the ocean.
 In order to clear the effects of the Hawaiian Islands, a no-Hawaii-run was also conducted, in which the Hawaiian Islands were removed. The initial state of the no-Hawaii-run was selected from that of the control-run which had already integrated for 30 years after atmospheric and ocean components were coupled. From this initial state, integration continued for further three years. This three year calculation was long enough to see the HLCC change, because the first baroclinic mode of Rossby waves, which is strongly related to the thermocline change, would travel far to the west of the islands within three years in this region.
 In order to match the season covered by the observational study [Xie et al., 2001], the averaged fields from July to October of the 33rd year were use in the simulations. For SST and surface winds, high-pass filtered anomalous fields were extracted from the background fields, defined by 8° moving averages in the meridional direction, following Xie et al. .
 The high-pass filtered SST map shows that a positive anomalous zonal band appears at 18°–20°N west of the Hawaiian Islands, extending beyond 160°E (Figure 1a). Over the positive anomalous SST, surface winds converge in the meridional direction (Figure 1a) and CLW is relatively large (Figure 1b). These distributions of SST, surface winds and CLW are in good agreement with the observational study [Xie et al., 2001].
 On the other hand, there are some differences from the observations. At first, the high-SST band extends from the Hawaiian Islands to the west-southwest in the observation, but almost westward in this model. Also, the positive anomalous SST is larger than that observed. These differences may be caused by the excessive speed HLCC in this model. These points will be considered later.
 In Figure 1c, the HLCC is reproduced as an almost eastward current at 18°–20°N. The area where the HLCC flows is in good agreement with the high-SST band, and it seems to support the suggestion by Xie et al.  that the HLCC is the thermal source of the high-SST band. However, the current velocity of the HLCC in this model, at 30–40 cm sec−1, is larger than those in ocean models driven by wind product of ECMWF analysis [Xie et al., 2001; Yu et al., 2003]. Kobashi and Kawamura  estimated that the surface westward geostrophic velocity in HLCC region is 6–9 cm sec−1, using hydrographical data, and Yu et al.  showed that the maximum speed of the HLCC is under 10 cm sec−1, using drifting buoy data. The velocities of the HLCC revealed by these studies are also smaller than in the current model. Yu et al.  suggests that eddy activity to the west of the Hawaiian Islands can affect the current velocity and the extent of the HLCC as a result of baroclinic instability. In their results, the low eddy activity can cause a strong HLCC. In the present study, however, the root-mean-square (rms) of sea surface height (SSH) to the west of Hawaii (figure not shown) is consistent with the analysis of the TOPEX/Poseidon altimetric data [e.g., Qiu, 1999; Kobashi and Kawamura, 2001]. Thus the eddy activity to the west of the islands is not low in the present model. Considerate is postulated that the excessive speed of the HLCC results from the stronger wind-curl induced by the Hawaiian Islands in the CGCM used in the present study, rather than the product of ECMWF (figure not shown) and that this excessive wind-curl is caused by the over-reduced wind speed caused by surface drag of the islands. This should be improved in the future.
 We consider that this too fast HLCC is the reason for the high-SST band extending almost zonally, together with the enhanced anomalous SST as mentioned above. Hafner and Xie  speculated that the tilted high-SST band, noted from the satellite observation, was caused by the negative feedback from the atmosphere that forces the high-SST band to move southward. The rapidly flowing HLCC in the present model will advect too great heat, giving rise to the differences in anomalous SST between this model and the satellite observation.
 In the no-Hawaii-run, the anomalous SST pattern seen in the control-run disappears, and the convergence of surface winds and high-CLW band also disappears (Figures 1d and 1e). The eastward ocean current corresponding to the HLCC, as seen in the control-run, also disappears (Figure 1f). Although an eastward current exists at 18°–19°N, 170°E–170°W, this current is a temporary feature.
 A second no-Hawaii-run was conducted in which the islands were left in the ocean, merely as sea-mountains, with no elevation above the sea surface. In this additional run, the HLCC also disappears (figure not shown). The missing HLCC in these no-Hawaii-runs shows that HLCC is governed by the wind-curl excited over the Hawaiian Islands [Xie et al., 2001]. That the high-SST and high-CLW bands are also missing implies that the thermal source of the high-SST band is nothing more than the HLCC.
 Compared with the ocean current field in the control-run, the North Equatorial Current (NEC) is weaker, especially east of 175°W in the no-Hawaii-run (Figures 1c and 1f). It seems follow on that the trade winds are enhanced to the south of the Hawaiian Islands in the control-run (Figure 1a), as mentioned by Xie et al.  and Hafner and Xie . The winds, being weaker than in the control-run to the south of the islands, consequently force the surface heat-content to increase. In agreement with this, the SST east of the Philippines in the no-Hawaii-run, is 0.2°–0.6°C higher than in the control-run, since the NEC advects warmer waters in the no-Hawaii-run (Figure 2). The weakened NEC in the no-Hawaii-run may also contribute to this higher-SST because of reduction in cool water transport by the NEC from the eastern basin. This means that the heat balance in the Western Pacific region will be affected by the winds from a long distance to the east, around the Hawaiian Islands.
3.3. Variability of CLW to the West of the Hawaiian Islands
 From the climatology, represented by time-averaged fields between the 21st year and the 56th year of the control-run, CLW west of the Hawaiian Islands exhibits two peaks, in spring and late summer to fall (Figure 3). This is in phase with wind speed, but out of phase with SST and high-pass filtered SST, although the high-CLW in fall is consistent with the high-SST (Figure 3).
 Cloud liquid water is considered to relate to latent heat flux on the sea surface and is a function of wind speed and SST in the bulk formula. In order to separate the contributions of wind speed and SST to the latent heat flux, we apply a linear expansion to the perturbative latent heat flux as equation (1) [cf. Hafner and Xie, 2003].
Q′EV is the perturbative latent heat flux, W wind speed, RH relative humidity, qs saturation specific humidity and CEV the bulk coefficient. Over-bar and prime indicate the time mean and perturbation, respectively.
 In the spring, latent heat flux is mainly due to wind speed, and in the fall, it is due to both wind speed and SST (Figure 4). Therefore, the zonal high-SST band is necessary for the high-CLW, but its variability is determined by the changes of not only SST, but also wind speed.
4. Summary and Discussions
 The present paper clearly demonstrates the reproduction of the air-sea interaction system which extends a few thousand kilometers to the west of the Hawaiian Islands, and reveals that the mountains on the islands are the trigger for the air-sea interaction system, just as suggested by observational study [Xie et al., 2001]. In addition, it is shown that the wind changes around Hawaii can affect the heat balance in the Western Pacific, and the wind speed plays an important role in the CLW variability to the west of Hawaii.
 The variability of the wind-curl generated by the Hawaiian Islands will induce meso-scale eddies to the west of the islands since the rms of SSH in the no-Hawaii run is reduced on the HLCC (figure not shown). These eddies may play a role in the variabilities seen around the HLCC region and in the south of Japan, where eddy activities are vigorous [e.g., Kobashi and Kawamura, 2001, 2002], through the long Rossby wave propagation. Additionally, in the Western North Pacific, meso-scale eddies can be caused locally by baroclinic instability [e.g., Qiu, 1999; Kobashi and Kawamura, 2002]. Therefore it is necessary to investigate the effect of the Hawaiian Islands on the eddy activities to the west of the islands.
 As has been shown in Section 3b, the HLCC disappears whether there are the Hawaiian Islands in the ocean component or not. However, the oceans to the west of the islands in the two no-Hawaii-runs are slightly different, and the existence of the islands in the ocean may play a role [cf. Qiu and Durland, 2002]. This is also a subject of interest.
 In the atmospheric component of the CGCM, the Hawaiian Islands are represented as a mountain whose height is only 103 m and the zonal scale is about 500 km after smoothing in the T106 spectral model. In spite of the rather poor representation of the islands, a realistic air-sea interaction system derived from meso-scale topography can now be demonstrated. This means that this CGCM is reliable in regard to dynamics and thermodynamics, and useful for climate studies. In the future, it is expected that a higher-resolution CGCM will contribute, not only to climate studies, but also to meso-scale air-sea interaction processes.
 The authors gratefully acknowledge valuable discussions and comments with Prof. Shang-Ping Xie, Dr. Masami Nonaka, and Dr. Fumiaki Kobashi. This work is supported by the first subject of the Kyousei project – “Project for Sustainable Coexistence of Human, Nature, and the Earth”, which is promoted by Ministry of Education, Culture, Sports, Science and Technology of Japan.