The role of the extratropical ocean in climate remains unclear due to the complexities in air-sea interaction processes. We have found robust evidence for the Tsushima Warm Current (TWC) preceding the western Pacific (WP) teleconnection pattern by conducting an analysis over the past 30 years. The WP index in winter sharply succeeds the volume transport of the TWC in autumn, but rather smoothly connects with the El Niño indices, indicating a considerable role of the ocean current in the climate system. Correlation patterns of seasonal precipitation over the Japanese Islands are also consistent with this relationship. The significant lead-lag correlations with the coherent structures of surface temperature indicate ocean-to-atmosphere feedback in which the interannual variation of the wind-driven current, represented by the TWC transport, influences the regional climate conditions associated with the WP pattern in winter.
 The heaviest urban snowfall in the world often occurs in the northwestern cities of the Japanese archipelago. Dry and cold winter monsoon absorbs a large amount of evaporation from the warm surface of the Japan Sea/East Sea (JES), and the mountain uplift forces the dense condensation and the continuous precipitation associated with cumulus convection. This local air-sea interaction has been studied for more than half a century [e.g., Manabe, 1957]. Accurate prediction of snowfall events can contribute to mitigating disruption and reducing damage [e.g., Sasaki and Yamakawa, 2004].
 Many researchers have discussed atmospheric factors such as monsoon variation or explosive cyclone development to explain the occurrence of heavy snowfall assuming the JES to be a heat reservoir. Recently, Hirose and Fukudome  found a close relationship between the volume transport of the Tsushima Warm Current (TWC) in autumn and the interannual variation in local evaporation and precipitation in winter. This warm current in its large part is a branch of the Kuroshio but is also in part an extension of the Taiwan Warm Current in the East China Sea. After entering into the JES, the TWC separates into a few branches with dominant mesoscale variability. Large amount of heat transported by this warm current is supplied to the lower atmosphere in autumn and winter. The remainders are named Tsugaru and Soya Warm Currents, and they merge into the subarctic circulations in the northwestern Pacific.
 The volume flux of the TWC can be a good predictor of the seasonal precipitation, owing to its strong connection with the local atmosphere. However, the analysis of Hirose and Fukudome  was limited to the JES region and also the data were collected only in recent years. Since the TWC carries significant amount of heat to the north (∼0.2 PW in summer and autumn, according to Isobe et al. ), its fluctuation impacts not only the local coastal area but possibly a broader climate system. In fact, the lead-lag correlation between the volume transport in autumn and the east Asian monsoon index in winter is non-negligible [Hirose and Fukudome, 2006].
 Many studies have emphasized the El Niño-Southern Oscillation (ENSO) event as a major factor controlling the winter climate over the North Pacific. Dominant teleconnection patterns of Pacific/North America (PNA) and western Pacific (WP) were also shown to be strongly influenced by the tropical variation [e.g., Horel and Wallace, 1981; Mo and Livezey, 1986]. Recently, Tanimoto et al.  revealed the active role of regional ocean variability in determining the atmospheric conditions. Isobe and Beardsley  studied the regional interaction between the northwestern marginal seas and winter monsoon intensity. Frankignoul and Sennéchael  were able to find strong evidence of sea surface temperature (SST) anomaly preceding the PNA pattern. Although many evidences of contributions of the extratropical ocean to the winter climate were found in these studies, the SST changes are not only attributed to the dynamic changes in the ocean but they can also be easily contaminated with atmospheric forcings. Statistical decomposition methods such as empirical orthogonal function (EOF) or singular value decomposition (SVD) analyses do not always separate the roles of atmosphere and oceans due to their complex interactions.
 Here, we use the results of long-term ocean current measurements carried out at the Tsushima/Korea Strait (TKS), which is directly related to the meridional heat transport. The analysis conducted by Hirose and Fukudome  is extended in terms of the data period and coverage area to find more strong and broad evidence of atmosphere-ocean interaction. Particular attention is paid to precipitation over the Japanese Islands, because it is largely attributed to the latent flux from the ocean. Furthermore, we discuss a possible connection between the ocean current and the seasonal climate conditions using SST and tropospheric geopotential height data.
2. Correlation Analysis
 First, we compared the TWC transport in autumn with the precipitation in winter at Japanese cities. The ocean current over the past nine years was measured using an acoustic Doppler current profiler (ADCP) mounted on a commercial ferryboat operating along the line shown in Figure 1a. In this study, a long time series of geostrophic transport was reconstructed using the sea-level differences across the TKS from the tide gauge data available since the 1960s [Takikawa and Yoon, 2005]. Precipitation data measured by the Japan Meteorological Agency were averaged over the three consecutive months of December, January, and February (DJF) from 1976/77 to 2005/06.
Figure 1 shows the correlation coefficients between the volume transport of the TWC averaged from September to November (SON) and the precipitation in the following winter (DJF). Positive and negative correlations were clearly separated by mountain barriers regardless of the measured or estimated transport. The positive correlation in the northwestern side basically confirms the finding of Hirose and Fukudome : Large (small) TWC transport in autumn carries more (less) heat into the JES and thus the high (low) water temperature strengthens (weakens) evaporation and precipitation in winter. The lag correlation between the area-averaged precipitation and the measured or estimated transport is 0.60 or 0.45.
 The reason for the negative correlation in the Pacific side of the Japanese Islands could not be easily explained. The areal correlations of −0.62 and −0.47 satisfy the 95% significance level, both for the measured and estimated transports, respectively. The northwesterly winter monsoon is not likely to be the direct cause because the moisture is lost in the mountains and the wind eventually dries up on reaching the populated region between Tokyo and Osaka. The winter precipitation, occasionally brought to the Pacific coast by the southerly winds associated with extratropical cyclones, rather seems to be hardly influenced by the maritime conditions of the JES region. Moreover, the direct correlation of the winter precipitation averaged over the two regions was weaker than the significance level of 95%, implying the presence of a controlling factor other than the local air-sea interaction.
 Second, we investigated the correlation between the TWC transport and the atmospheric pressure field to find whether a strong connection between the warm current and the precipitation pattern exists. Figure 2 indicates a close relationship between the time-mean transport in SON and the tropospheric variation in the following DJF. Large (small) transport of the TWC leads to the development of a weak (strong) westerly jet and a weak (strong) low pressure anomaly over the Kamchatka Peninsula. The dipole correlation pattern closely resembles the western Pacific (WP) pattern, which is one of the typical winter teleconnection patterns found in the Northern Hemisphere 500hPa geopotential height field [Wallace and Gutzler, 1981; Barnston and Livezey, 1987]. A positive center of the action is located in the Kamchatka Peninsula, and the negative correlation extends zonally in the western subtropical Pacific. The lead-lag correlations with the measured transport data are as strong as ±0.9, suggesting a significant contribution of the warm current to the winter climate system.
 A direct comparison of the two measures, the TWC transport in autumn and the WP index in winter, reveals the considerable similarity (Figure 3a). The WP index was calculated by a rotated principal component analysis of the 500hPa height anomalies at National Centers for Environmental Prediction (NCEP). Its sign was reversed to fit the original definition given by Wallace and Gutzler . The WP index in winter shows a time variation similar to the TWC transport in the previous autumn. The lead-lag correlations with the measured and estimated transport are 0.90 and 0.57, respectively. The seasonal lag is crucial in predicting the WP pattern, because the correlation shows a prominent peak at a three-month difference (Figure 3b) but it is negligible simultaneously in autumn or in winter (0.13 or −0.18).
 On the other hand, the TWC transport does not appear to affect the other dominant structure of the PNA teleconnection pattern. The PNA index provided by the NCEP gives negligible correlations, regardless of the lead time (Figure 3b). The TWC and associated oceanic changes seem to influence a certain natural variability in the North Pacific in winter.
 Previous studies have indicated that the ENSO event is a major controlling factor of the PNA and WP patterns [Horel and Wallace, 1981; Mo and Livezey, 1986]. However, the correlations of the WP index in winter and the El Niño indices are weaker than the significance level for the past 30 years (Figure 3b). Besides, the correlation coefficients hardly depend on the lead-lag time. The PNA index shows a higher and smoother dependence on the preceding El Niño indices (Figure 3b). The tropical variation probably has a large-scale and long-term effect on the North Pacific teleconnection patterns, as explained in the previous studies. The transport variation of the TWC can be an additional sharp local predictor of the WP pattern in winter.
 Some studies also clarified that the interannual variation in the WP pattern largely affects the winter precipitation on the Japanese Islands; Lau  and Ueno  revealed a high (low) probability of rain or snow on the Pacific side during a negative (positive) phase of the WP pattern. Takano et al.  recently showed that the convective activity over the JES region is correlated with the WP index. This relationship is consistent with our finding of the north-south difference in the correlation pattern between the ocean current in autumn and the precipitation in winter (Figure 1). Thus, we are successful in linking the TWC transport in autumn and the precipitation pattern on the Japanese Islands in winter on the basis of the WP teleconnection pattern.
 The strong statistical relationship implies the existence of an ocean-to-atmosphere feedback in the extratropics; the wind-driven ocean variation may significantly affect the climate teleconnection pattern. Although many studies concluded the dominance of atmospheric forcing of the extratropical oceans, the present work suggests the ocean current forcing of the wintertime climate to a certain extent. However, the strong correlation does not ensure that the TWC transport in autumn solely controls the WP pattern in winter because the TWC is only a part of the western boundary current system in the North Pacific. It may be more rational to consider that the TWC variation represents large-scale variation in the northwestern Pacific with the marginal seas, based on their strong connectivity [Minato and Kimura, 1980; Toba et al., 1982].
 Assuming the existence of regional ocean-to-atmosphere feedback, the SST can be regarded as an interface variable between the ocean current in autumn and the climate pattern in winter. Many studies have already established the fact that the SST anomaly precedes the regional climate in winter by one or two months [e.g., Tanimoto et al., 2003; Hirose and Fukudome, 2006; Frankignoul and Sennéchael, 2007]. However, the accuracy of the historical SST data was not sufficiently high either due to the limited number of in-situ observations or the errors in remote-sensing measurements. Here, we use a combination of the two SST datasets of Reynolds and Smith  and Minobe and Maeda  to reduce the uncertainty. The former showed a large-scale variation after an optimal interpolation, but the latter retained small-scale features with many missing values. The combined SST anomaly of November and December (ND), which gives a smooth estimate of December 1st, is then compared with the TWC transport in autumn or the climate index in winter for the period from 1976/77 to 2005/06.
 Although covariance magnitudes are small and the patterns show noisy patches (Figure 4), some similarities are observed at mid-latitude; warm anomalies extend eastward from the eastern JES, and a cold anomaly is located in the Yellow Sea. The positive and negative anomalies observed in Figure 4a imply coherence of the TWC transport in autumn with the northwestern boundary current system. For instance, a large (small) transport through the TKS might be counterbalanced with a small (large) transport of the Yellow Sea Warm Current, resulting in a negative correlation between the JES and the Yellow Sea. Such coherent variation in the surface thermal conditions possibly contributes to the activity of the WP teleconnection pattern, as indicated by the similar covariance structures (Figures 4a and 4b).
 On the other hand, a large-scale negative covariance between the subtropical SST anomaly and the WP index in winter (Figure 4b) could not be explained by the TWC transport. There seem to be independent factors existing in the subtropical gyre related to the WP variability, which are partially supported by the weak correlation between the Kuroshio and TWC transports. The Kuroshio transport estimated in the East China Sea [Kawabe, 1995] lead the TWC transport in autumn less significantly (strongest correlation of −0.32 at a four-month lag). The lead-lag correlations between the Kuroshio transport and the winter WP index are also negligible within ±0.2. The interannual changes in the TWC transport in autumn may be related to the subarctic variations (35–50°N) rather than the subtropical ones, as illuminated by Figure 4.
4. Concluding Remarks
 The autumn transport of the Tsushima Warm Current (TWC) can be an excellent predictor of the western Pacific (WP) teleconnection pattern associated with the precipitation pattern in the Japanese archipelago in the following winter. Since the three-month lag between the autumn current and the climate index in winter cannot be explained by short atmospheric timescales, we interpret the strong preceding correlation as an indicator of ocean-to-atmosphere feedback in the extratropics. The interannual fluctuations in the WP pattern can be at least partly attributed to the variations in the ocean currents in the northwestern Pacific, especially in the marginal seas. However, the TWC transport in autumn shows no correlation with the other dominant pattern of PNA. Variations in the regional ocean contribute to the difference in the behavior of the WP and PNA teleconnection patterns.
 The correlation of the WP index in winter with the El Niño indices was more stationary and weaker than that with the TWC transport in autumn. In addition to the tropical factors controlling the North Pacific climate, the local oceanic changes can modify the seasonal condition. Coupling the two independent factors for a further accurate prediction of the WP teleconnection pattern, owing to their statistical independence, is recommended.
 Analyzing ocean currents can be a promising approach in clarifying the role of oceans in the climate system. The SST is easily influenced by both active and passive roles of oceans, resulting in a small covariance as shown in this study. Vertically-integrated ocean currents give a direct measure of the meridional heat transport, which is more robust than the surface variable due to high-frequency atmospheric variabilities. Long-term and accurate measurement of ocean currents at various locations is thus desired for better understanding of the climate system.
 The authors greatly appreciate the voluntary contribution of Camellia Line Co. to the long-term ADCP measurement. We also thank Shoshiro Minobe and Mototaka Nakamura for their useful comments.