Two eddy-resolving ocean simulations are carried out to study local air-sea interaction associated with the Hawaiian Lee Countercurrent (HLCC), one forced by the QuikSCAT satellite wind field (QSCAT run) and the other by the NCEP/NCAR reanalysis (NCEP run). The simulated HLCC in the QSCAT run extends west-southwestward from the Hawaiian Islands much farther than that in the NCEP run. We attribute this difference to difference in the wind fields. In the QSCAT wind field, there exist band-like structures of positive and negative wind stress curls induced by warm sea surface temperature band associated with the HLCC; however, these features are not apparent in the reanalysis wind field. The QSCAT run provides a good example of local two-way air-sea interactions in association with the HLCC. Furthermore, interannual variations are suggested in both the simulated HLCC and wind fields over the HLCC.
 The Hawaiian Lee Countercurrent (HLCC) is a narrow eastward current to the west of the Hawaiian Islands embedded in the wide westward North Equatorial Current, which is observed from the surface drifter [Qiu et al., 1997; Flament et al., 1998] and hydrographic data [Xie et al., 2001; Kobashi and Kawamura, 2001]. Xie et al.  propose that the HLCC and its far-reaching extension are responses to orographic wind-stress curls in the lee of the Hawaiian Islands. Yu et al. , however, suggest that westward extent of the influence of the Hawaiian Islands can be limited because of meso-scale eddies that can extract energy from the mean flow, and Kobashi and Kawamura  point that subsurface potential vorticity distributions can be responsible for eastward currents to the west of the dateline.
Xie et al.  further show that the HLCC advects western warm water eastward to form zonal band-like structure of warm sea surface temperature (SST), and meridional convergence of surface wind and high cloud water appear over the warm SST band, suggesting the existence of an ocean-to-atmosphere feedback following the atmosphere-to-ocean feedback triggered by the Hawaiian Islands. Using a coupled general circulation model (GCM), Sakamoto et al.  succeed in reproducing the far-reaching effects of the Hawaiian Islands and confirm that the islands trigger the air-sea interactions. However, if the wind convergence with high cloud water over the warm SST band further affects the ocean or not, has not been shown clearly, while Hafner and Xie  suggest thermal effects of wind-induced evaporation changes and negative feedback from the cloud band based on regional atmospheric simulations.
Figure 1a shows that to the west of the Hawaiian Islands, meridional convergence of surface wind is apparently observed by the QuikSCAT (QSCAT) satellite and it extends along the warm SST band observed by the Tropical Rain Measuring Mission (TRMM), as reported by Xie et al. . In contrast to it, the warm SST band in the NCEP/NCAR reanalysis [Kalnay et al., 1996] is broad and unclear, and the accompanied wind convergence is also not clearly found (Figure 1b), indicating difficulty to represent the ocean-to-atmosphere feedback over the HLCC in the reanalysis system.
 The differences in these wind fields suggest that we can investigate oceanic response to the local wind stress field induced by the warm HLCC by comparing two ocean simulations driven by the wind fields of the QSCAT satellite observation and the NCEP/NCAR reanalysis. In this study, we conduct this comparison with eddy-resolving ocean simulations, taking account of the influence of meso-scale eddies.
2. Model Descriptions
 In this study, we use the ocean GCM for the Earth Simulator (OFES [Masumoto et al., 2004]), based on the MOM3 [Pacanowski and Griffies, 1999] with substantial modifications for a vector-parallel hardware system of the Japan's Earth Simulator. The model domain covers from 75°S to 75°N with horizontal resolution of 0.1°. The model has 54 vertical levels, with resolutions from 5 m for the surface to 330 m for the bottom, and the maximum depth is 6,065 m.
 Following a 50-year spin-up integration with climatological monthly fields [Masumoto et al., 2004], a hindcast simulation from 1950 to 2004 (NCEP run; see Sasaki et al. ; H. Sasaki et al., An eddy-resolving hindcast simulation of the quasi-global ocean from 1950 to 2003 on the Earth Simulator, submitted for publication in High Resolution Numerical Modelling of the Atmosphere and Ocean, edited by W. Ohfuchi and K. Hamilton, Springer, New York, 2006; and/or http://www.es.jamstec.go.jp/esc/research/AtmOcn/ofes/index.en.html for details) is driven by daily mean wind stress of the NCEP/NCAR reanalysis data, and forced by surface heat flux estimated from atmospheric field of the same reanalysis data and the simulated SST using bulk formulas. The salinity flux is obtained from evaporation associated with the heat flux and precipitation of the reanalysis data with an additional surface salinity restoring to the monthly climatology. A bi-harmonic operator and the KPP scheme [Large et al., 1994] are adapted for horizontal and vertical mixing, respectively.
 Another hindcast simulation (QSCAT run) is driven by daily mean surface wind stress data set based on the QSCAT satellite observation, which is constructed by weighted mean method with horizontal resolution of 1° [Kutsuwada, 1998] and provided from the J-OFURO data set [Kubota et al., 2002]. Other atmospheric data that force the QSCAT run are the same as those in the NCEP run. The QSCAT run starts from the oceanic field on July 20, 1999 in the NCEP run and is integrated to the end of 2004. In the following analyses, we use the simulated fields after January 2001 to exclude influences of the initial shock.
Figure 2 shows the current vectors in the QSCAT run averaged from August to October, the season of strong HLCC [Kobashi and Kawamura, 2002]. The simulated eastward current to the west of the Hawaiian Islands exhibits realistic representation of the HLCC. The far-reaching HLCC extends west-southwestward across the dateline that is consistent with the observation [Kobashi and Kawamura, 2001], and it is especially strong and extends long in 2003. In contrast to it, in the NCEP run, westward extent of the simulated eastward current is limited and it is less organized (Figure 3) compared to the counterpart in the QSCAT run in 2003 (Figure 2c), as an example.
Figure 2 also shows that to the west of the Hawaiian Islands, zonally elongated structures of positive and negative curls of the QSCAT wind stress extend west-southwestward. The structures correspond to the wind convergence induced by the warm SST (Figure 1a). In contrast to it, a pair of positive and negative wind stress curls based on the NCEP reanalysis in 2003 (Figure 3), for example, is broad and not extending far westward, although the dipole structure of wind stress curls is distinct to the east of 170°W.
 In Figure 2, it is found that the simulated HLCC in the QSCAT run is mostly distributed in regions of strong meridional gradient of wind stress curl between the positive and negative curls, especially in 2003, the year of the most distinct HLCC and band-like wind curl structure in the period of the QSCAT run. This relation is consistent with the linear vorticity balance and suggests that the local wind curl can drive the current. It is also noteworthy that significant interannual variations are apparent in both the simulated HLCC and wind stress field over the HLCC, although the period of the QSCAT run is rather short to discuss interannual variability. The distinct HLCC in 2003 is also captured in geostrophic velocity fields (not shown) derived from dynamic sea surface height (SSH) of the merged satellite altimeter distributed by Aviso (Archiving, Validation, and Interpretation of Satellite Oceanographic Data project), supporting the existence of interannual variability in the HLCC.
 To see the relative distributions of the current vectors and wind stress curl in the mean fields, we plot them in Figure 4 based on the two runs. Just to the west of the islands, the contrast of wind stress curls in the NCEP reanalysis (Figure 4b) is rather bold compared to that in the QSCAT wind stress field (Figure 4a), where both simulated HLCCs are represented clearly. Between 175°W and the Hawaiian Islands, the relatively strong HLCC is also captured in the mean geostrophic velocity field based on the Aviso satellite altimeter (not shown). To the west of 170°W, however, in the NCEP run, the contrast of wind stress curls is indistinct and, at the same time, the simulated HLCC does not continuously extend far westward compared to those in the QSCAT run. This result suggests that extension of the HLCC in the QSCAT run is associated with local wind stress field over the HLCC. Note that the continuous extent of the HLCC in the four-year mean field in the QSCAT run (Figure 4a) is also found in the three-year mean field, excluding 2003 (not shown). This means that although this feature is significant in 2003, it is found not only in 2003 but also in other years.
 In contrast to the relationship between eastward current and wind stress curls found in 2003 in the QSCAT run (Figure 2c), in the mean fields, eastward currents are located on the northern side of the peak of Ekman pumping gradient induced by wind stress curls (Figure 4a). Such obscure features may come from westward propagations of influence of Ekman pumping via Rossby waves. It is thus possible that there are time lag and/or discrepancy in the meridional distributions between the HLCC and local wind stress field. For example, in 2004 (Figure 2d), eastward currents generally do not appear to the south of peak of positive wind curl because the result of stronger wind curls in the previous season (not shown) did not dissipate yet. Their distributions, however, are rather consistent with each other when the local forcing is significant as found in 2003 (Figure 2c), suggesting that stronger contrast of local wind stress curls can cause more pronounced relation between the HLCC and local wind stress field.
Figure 5 shows that in the region just on the lee side of the Hawaiian Islands, 165–160°W, the tilt of thermocline (contours) and eastward current velocity (colors) are larger in the NCEP run than in the QSCAT run. This indicates that larger westward extent of the HLCC in the QSCAT run is not caused by stronger wind curl near the Hawaiian Islands, whose influence could then extend westward via Rossby waves. On the other hand, the limitation of the westward extent in the NCEP run is probably due to the influence of meso-scale eddies as shown by Yu et al. .
 It is also noteworthy that there are some other eastward currents, for example, between 175°E and 175°W at 21°N in 2003 in the QSCAT run (Figure 2c) and around 175°E at 20°N in the mean field in the NCEP run (Figure 4b). For all these eastward currents, corresponding meridional gradients of wind stress curl that might induce the currents are detected.
4. Summary and Discussion
 We have investigated dynamical oceanic response to atmospheric field associated with the far-reaching HLCC by comparing eddy-resolving ocean simulations driven by two different wind stress fields. In the QSCAT run, the far-reaching HLCC extends west-southwestward realistically across the dateline, while the far-reaching extent is not represented well in the NCEP run. The comparison suggests that zonally elongated structures of wind stress curls induced by the warm HLCC in turn play a role to drive further the HLCC, which follows the atmosphere-to-ocean and ocean-to-atmosphere feedbacks triggered by the Hawaiian Islands suggested by Xie et al. , and thus there are possible two-way air-sea interactions accompanying the HLCC. The result also implies that the far-reaching HLCC can be directly forced by winds even in the eddy-resolving simulation if the local air-sea interaction is taken into account. Although this study emphasizes a role of the local wind stress curl, it would be also worth to note a possible role of Rossby wave propagation and/or subsurface potential vorticity distribution. They probably set initial eastward current with warm water advection and subsequent changes in the wind stress, and therefore might trigger the two-way air-sea interactions.
 Interannual variability is suggested in the wind field and in the simulated HLCC in the QSCAT run, and it is also supported both by the satellite-observed SSH field and by the SST field observed by TRMM satellite (S.-P. Xie, personal communication, 2005). In 2003, the simulated HLCC is stronger, the observed SST along the HLCC is warmer, and the association between the HLCC and local wind stress curl is more apparent than those in other years. In addition to the interannual variations, there are also seasonal ones in the simulated HLCC (not shown). Detailed investigations for variability in the air-sea interaction system accompanying the HLCC are left for future works.
 We are grateful to OFES project members including Y. Masumoto, and H. Sakuma and researchers in Atmosphere and Ocean Simulation Research Group in Earth Simulator Center. Our thanks are extended to S.-P. Xie, Z. Yu, N. Maximenko, and T. T. Sakamoto for valuable discussions. QuikSCAT wind stress data in the J-OFURO data set (http://dtsv.scc.u-tokai.ac.jp/j-ofuro/) are provided by Kutsuwada. TRMM data are produced by Remote Sensing Systems (http://ssmi.com/). The OFES simulations were conducted on the Earth Simulator under support of JAMSTEC.