1. Introduction
The Kuroshio is detached from the western boundary of the basin at the eastern tip of Japan to form a strong eastward jet: the Kuroshio Extension (KE). The warm KE releases abundant heat and moisture into the atmosphere, especially in the cold seasons, implying significant impact of KE on the atmospheric circulation. Furthermore, frontal gradient in sea surface temperature (SST) associated with KE can affect atmospheric eddy activity aloft and then its large-scale circulation. These suggest KE’s potential importance in air–sea interactions (e.g., Kwon et al. 2010; Qiu et al. 2007). Additionally, recent studies have suggested that decadal changes in the speed of the Kuroshio and an upstream portion of KE can influence natural mortality of infant Japanese sardine (Nishikawa and Yasuda 2011; Nishikawa et al. 2011), which is known to yield several orders of interdecadal variability in its mass. Prediction of the path and intensity of KE is thus of social and scientific importance.
Through a linear vorticity model, Schneider and Miller (2001) have shown that interannual variations in thermocline depth [and thus sea surface height (SSH)] and in SST averaged over a relatively large area of the western North Pacific Ocean (35°–40°N,1 140°–170°E) are potentially predictable. They found the particular predictability attributable to propagation time for wind-driven oceanic Rossby waves (Deser et al. 1999) across the basin.
Through a long-term integration of an eddy-resolving ocean general circulation model (OGCM), Taguchi et al. (2007) have confirmed the finding by Schneider and Miller (2001) and satellite observations by Qiu and Chen (2005) that broad meridional-scale variability around KE can be explained by propagation of Rossby waves that are driven by large-scale atmospheric variability. Taguchi et al. (2007) have also shown that meridionally confined frontal-scale variability that contributes to changes in the KE jet speed can be generated by internal ocean dynamics. As previously pointed out through idealized model experiments (e.g., Pierini 2006), the frontal-scale variability is thus by nature unpredictable but, at the same time, it tends to be modulated synchronously with the broad-scale wind-driven variations (Taguchi et al. 2007; Qiu and Chen 2010). Recently, there have been several studies on predictability of variability in the North Pacific Ocean. Tanaka et al. (2004) assessed the predictability of the Kuroshio south of Japan, focusing on its wind-driven component by excluding its intrinsic variability. In more recent years, several groups conducted global climate prediction for the next several decades with coupled atmosphere–ocean models (e.g., Mochizuki et al. 2010), whose horizontal resolution is, however, not necessarily high enough to study predictability of the frontal-scale KE variability.
Combining the results of Schneider and Miller (2001) and Taguchi et al. (2007) mentioned above, we hypothesize that, though unpredictable by nature, the frontal-scale KE jet speed variability includes a potentially predictable component that is synchronous with the predictable wind-driven broad-scale variability (see also Miller et al. 1998; Deser et al. 1999; Qiu and Chen 2010). With this hypothesis, the present study explores the predictability of interannual variability in the KE jet speed by using the same eddy-resolving OGCM as used by Taguchi et al. (2007), which can represent the KE variability realistically (Nonaka et al. 2006; Taguchi et al. 2007, 2010).
2. Model and experiments
a. OFES
We use the Modular Ocean Model 3 OGCM (Pacanowski and Griffies 2000) with substantial modifications added for its optimal performance on the vector-parallel hardware system of Japan’s Earth Simulator. This ocean model for the Earth Simulator (OFES) (Masumoto et al. 2004; Sasaki et al. 2008) covers a near-global domain (75°N–75°S), with a horizontal resolution of 0.1°. The model has 54 vertical levels with 5-m resolution just below the surface, and the maximum depth is 6065 m.
Following a 50-yr OFES integration with climatological monthly-mean forcing, we conducted a 61-yr OFES hindcast integration with daily-mean atmospheric fields taken from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data (Kalnay et al. 1996) from 1950 to 2010. This hindcast simulation successfully captures interannual-to-decadal variability in the western North Pacific (Nonaka et al. 2006, 2008; Taguchi et al. 2007, 2010), including the upper-layer thermal conditions averaged in a relatively large area to the south of the KE region (Fig. 1).2 Although it is still difficult to reproduce interannual-to-decadal modulations in eddy activity, especially in the upstream KE region (Taguchi et al. 2010), the simulation reproduces variability in the KE jet speed fairly well, as shown in section 3b. This gives us an unprecedented opportunity to investigate predictability of the KE jet over several phases of its decadal variability.
Time series of 3-month mean anomalies of ocean heat content (1020 J) in the top 700-m layer averaged within 30°–34°N, 140°–160°E, on the basis of the observational data (gray solid curve; Levitus et al. 2009) and OFES hindcast integration (black dashed curve). The correlation coefficient between the two time series is indicated at the top-right corner. The 95% confidence level is 0.53 with 12 degrees of freedom assumed for 33-month time scales of the time series.
Citation: Journal of Climate 25, 10; 10.1175/JCLI-D-11-00641.1
b. Experiments
To assess potential predictability of the KE jet speed, we have conducted a suite of prediction experiments using OFES, following Schneider and Miller (2001). The initial conditions for the experiment were obtained from the 60-yr hindcast integration for 1 January of 1968, 1976, 1980, 1984, 1992, and 2000. Five of these six years were chosen at intervals of 8 yr to pick up different phases of the decadal variability, while another year (1980) was chosen to represent a particular timing when the KE jet underwent a distinct southward shift. From each of these six dates, OFES was then integrated for 4 yr as a prediction experiment with atmospheric forcing that includes no interannual variability. The forcing field had been constructed from the NCEP–NCAR reanalysis as follows. In each year, the atmospheric forcing was divided into the seasonal variability and transient disturbances through a 31-day running mean. Amplitudes of transient disturbances were then defined as the square root of the 5-day running mean of the square of the corresponding high-frequency fluctuations. The forcing for the prediction experiments was finally constructed locally as the sum of the long-term mean of the seasonal variability and transient disturbances in 1979 with their amplitude adjusted to its long-term mean. We emphasize that the inclusion of transient atmospheric disturbances is critical for representing realistic surface heat fluxes. One of these experiments was used in Taguchi et al. (2007). Although the model is by no means perfect, comparison of the predicted fields with the corresponding hindcast fields will provide us with a measure of potential predictability of the KE jet speed as if there were a perfect model in representing the KE variability.
In addition to the model output, we analyze global gridded delayed-time merged (Ducet et al. 2000) SSH anomaly (SSHA) and sea surface geostrophic current field reference products (SSALTO/DUACS 2011) distributed by Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO). The SSH and current data products are available on a regular 0.25° grid and a ⅓° Mercator projection grid, respectively, and their monthly-mean values are used in this study.
3. Results
a. Predictability of broad-scale SSH variability
To examine whether the predictability of broad-scale variability shown by Schneider and Miller (2001) is also found in a realistic eddy-resolving OGCM, OFES, we plot the broad-scale SSH variability and its predictability in Fig. 2. Here, we apply a 5° × 5° horizontal running mean to the SSH fields to suppress signals of eddies and frontal-scale variability. Time series of the area-averaged SSHA to the south of KE (30°–34°N, 140°–160°E), which is highly correlated with that of the KE jet speed (as shown in section 3c), indicate that the predicted fields overall follow the corresponding fields in the hindcast integration (Fig. 2a). In fact, their correlation coefficient for the whole forecast period (r = 0.68) is statistically significant at the 95% confidence level. Predictive skill represented by the explained fraction of variance (Fig. 2b) is slightly lower than that for the persistence of the initial state for the first two years. After the third year, however, the skill for the forecast experiments is substantially higher than that for the persistence, and the explained fraction of the variance tends to be more than ⅓.
(a) Time series of 5° × 5° smoothed monthly SSHA (cm) averaged over 30°–34°N, 140°–160°E, on the basis of the OFES hindcast integration (black) and six forecast experiments (colors). (b) Time series of the corresponding fraction of variance explained by the forecast experiments represented by 1 − (forecast errors)2/(signal variance) (colors) in the area-mean SSHA in (a). The thick black curve is for the root-mean-square errors among the six forecasts, and the gray dashed line is the counterpart of the black thick line for the case in which persistence of the initial state is assumed for each experiment period. Color usage is as in (a). Positive values in (b) correspond to forecast errors being smaller than the standard deviation of the signal. All plots are based on a monthly mean, and no temporal filter is applied, but the mean seasonal cycle is removed. Horizontal smoothing is applied to reduce the influence of a strong gradient of SSH in the KE front.
Citation: Journal of Climate 25, 10; 10.1175/JCLI-D-11-00641.1
Although the predictability of SSHAs to the north of KE influences the KE jet speed, it is rather difficult to identify the influence of Rossby wave propagation in the northern SSHA, which tend to be more sensitive to the latitudinal position of the KE jet than the southern ones. Farther to the north (38°–42°N, 140°–160°E), the predictive skill of SSHAs is found to be much lower than that in the region to the south of the KE jet, which is consistent with the result of a linear Rossby wave model by Qiu (2003).
b. Predictability of KE jet speed variability
As shown by Taguchi et al. (2007), KE jet variability tends to be modulated more or less synchronously with broader-scale variability induced by wind-driven Rossby wave propagation. The potential predictability of the broad-scale variability shown above (Fig. 2) implies that the frontal-scale KE jet speed variability may also be predictable to a certain level. To explore this possibility, we construct a similar plot to Fig. 2 for the KE jet speed zonally averaged over its upstream region (145°–155°E) (Fig. 3) in recognition of its potential importance for Japanese sardine’s variability. Here, the KE jet speed is defined as the current velocity at 100-m depth along its axis, which is defined at each zonal grid point as the latitude of the maximum 100-m current velocity within the latitudinal band 30°–40°N.
(a) As in Fig. 2a, but for the KE jet speed (cm s−1) averaged over 145°–155°E. (b),(c) As in Figs. 2a and 2b, but for the KE jet speed to which a 13-month running mean is applied. The mean seasonal cycle is removed in (c). The orange curve in (b) is for the KE jet speed derived from the satellite-observed SSH.
Citation: Journal of Climate 25, 10; 10.1175/JCLI-D-11-00641.1
Time series of the predicted KE jet speed and the corresponding hindcast field suggest that the former overall well corresponds to the latter (Fig. 3a). At the same time, high-frequency variability with periods shorter than 1 yr is substantial in the KE jet speed, probably because of the influence of mesoscale eddies. Indeed, nearly 60% of the total variance is accounted for by high-frequency fluctuations defined locally as instantaneous deviations from the 13-month running mean. After smoothed with 13-month running mean (Fig. 3b), the predicted time series follow the corresponding hindcast field well over 4 yr, except the particular experiment from 1984. Although the KE jet speed changes tend to occur earlier in the predicted time series, their correlation r = 0.76 for the entire forecast period is statistically significant at the 90% confidence level. Furthermore, the explained fraction of the variance as a measure of predictive skill of the forecast experiments remains higher than 0.5 even in the third year of the forecast (Fig. 3c). The fraction is apparently higher than the counterpart for the persistency,3 although their values are comparable for the first two years of the forecast. These results suggest that, at least in this particular OGCM, there is some predictability for not only broad-scale SSHAs around KE but also the frontal-scale KE jet speed, supportive of our hypothesis based on Schneider and Miller (2001) and Taguchi et al. (2007). Furthermore, except around 2003, the hindcast integrations can reproduce the phase of decadal variability in the KE jet speed fairly well in comparison with the satellite observations (orange curve in Fig. 3b). In fact, their correlation r = 0.62, with statistical significance at slightly less than the 90% confidence level, with 5 degrees of freedom assumed for the limited length of the satellite observations. The hindcast integrations tend to underestimate the amplitude of the variability, however.
c. KE jet speed predictability from lagged correlations with SSHAs
The prediction experiments discussed above suggest crucial importance of Rossby wave propagation in the KE jet speed variability and its predictability, which is consistent with previous studies (e.g., Qiu and Chen 2010). Because of limited computational resource available, however, the number of our integrations is too few to make robust assessment of predictability and the results must therefore be interpreted with caution. One can nevertheless expect the KE jet speed to be correlated significantly with SSHAs associated with Rossby waves in the central portion of the North Pacific basin several years in advance. To examine this, we plot lag-correlation maps (Fig. 4) between the KE jet speed averaged over 145°–155°E and SSHAs based on the 61-yr hindcast integration.
Lag-correlation maps of SSHAs with the mean KE jet speed averaged between 145° and 155°E as the reference index: SSHAs lead the index by (a) 6, (b) 5, (c) 4, (d) 3, (e) 2, (f) 1, and (g) 0 yr. (h) Time series of the KE jet speed (black; left axis), and 3- (green), 4- (red), and 5- (blue; sign reversed) yr-lead SSHAs (right axis) averaged over 30°–34°N and 170°E–175°W, 30°–34°N and 170°–155°W, and 37°–41°N and 170°E–175°W, plotted with the green, red, and blue rectangles in (d),(c), and (b), respectively. The mean seasonal cycle is removed, and a 13-month running mean has been applied to all of the variables. White contours in (a)–(g) indicate the 95% confidence level (0.44). We estimate the number of degrees of freedom for the smoothed KE jet speed to be 20, with 30-month time scale of the KE jet speed.
Citation: Journal of Climate 25, 10; 10.1175/JCLI-D-11-00641.1
The simultaneous correlation map (Fig. 4g) shows that the KE jet speed exhibits high positive correlation with SSHAs to its south and slightly weaker negative correlation to its north, which is consistent with the geostrophy of the eastward KE jet. As the lead time for SSHAs increases, the domain of the high correlation extends eastward (Figs. 4c–f), corresponding to the westward propagation of oceanic Rossby waves as shown in previous studies (e.g., Qiu and Chen 2010, their Fig. 6). In fact, with their lead times of 3 and 4 yr to the KE jet speed, SSHAs exhibit their maximum correlation around 30°–34°N in the longitudinal spans of 170°E–175°W and 170°–155°W, respectively (Figs. 4d,c). Indeed, time series of area-mean SSHAs over these regions are highly correlated with the KE jet speed if the corresponding time lags are imposed (Fig. 4h). Specifically, interannual variability in the KE jet speed well followed the SSHAs in the central North Pacific 3 yr in advance (green curve), except in the early 1960s and mid-1990s, and their correlation r = 0.68 (0.55 for 4-yr-lead SSHAs). These correlation coefficients are statistically significant at the 95% confidence level and higher than the corresponding values for the persistent anomalies. For longer lead times, a region of organized negative correlations emerges in the central North Pacific to the north of KE, around 37°–40°N (Figs. 4b,a). Specifically, the SSHA averaged within the region 5 yr in advance (Fig. 4h, blue curve) exhibits significant negative correlation (r = −0.61) with the KE jet speed. The distribution of negative correlation is consistent with slower Rossby wave propagation at higher latitude, but the northern signal loses its significance in shorter lead time, as shown in Figs. 4e,f. How the northern signal propagates into the western portion of the basin needs to be investigated further.
The same lag-correlation analysis as above was applied also to satellite-observed SSHAs and the geostrophic KE jet speed estimated from them. In a manner consistent with the model hindcast, the observations also reveal apparent correlation between the anomalous KE jet speed and SSHAs in the central North Pacific 3 yr in advance (not shown). Having started in 1992, however, the satellite SSH observations are still limited in length and, thus, degrees of freedom. In fact, a significant change was observed only from the early to mid-2000s. Continuous satellite SSH monitoring is therefore necessary to obtain robust statistical relation in the real North Pacific Ocean.
In Fig. 5, we further examine the predictability of the latitudinal position of the KE jet. It is rather difficult to assess predictive skill of the KE jet position for the particular period that includes only a single major event of its axial shift in the early 1980s. The time series of several forecast integrations nevertheless follow that of the hindcast integration, implying some predictability. This is consistent with the result of Sasaki and Schneider (2011), which suggests significant influence of pseudowestward propagating signals on the latitudinal position of the KE jet axis. It is noteworthy that the first two forecast integrations are good in predicting the KE jet speed (Fig. 3) but not the jet latitude. This implies that signals causing variability in the KE jet speed and its latitude may be significantly different.
As in Fig. 3, but for the latitude of the KE jet axis (°N) averaged over 145°–155°E. The mean seasonal cycle is removed in (c).
Citation: Journal of Climate 25, 10; 10.1175/JCLI-D-11-00641.1
The result of this paper indicates that the synchronization between the broad-scale wind-driven variability and the frontal-scale KE jet variability is significant enough to induce the predictability of the KE jet speed as hypothesized in the introduction. This implies that coarse-resolution OGCMs and even layer models could be used for qualitative prediction of the KE jet speed, if they can represent wind-driven Rossby wave propagation realistically. The quantitative relationship between the broad- and frontal-scale variability, however, needs to be clarified by observations and/or model simulations with high horizontal resolutions.
4. Summary and discussion
Potential predictability of interannual variability in the KE jet speed attributed to the westward propagation of wind-driven oceanic Rossby waves is investigated through prediction experiments with an eddy-resolving OGCM under the perfect-model assumption. Although based on a small number of the experiments because of limited computational resources available, our experimental results suggest some predictability not only in broad-scale SSHAs but also in anomalous KE jet speed associated with meridionally confined frontal-scale variability. Specifically, more than half of the variance in the 13-month running mean KE jet speed can be explained by the ensemble mean of the forecast integrations even in the third year (Fig. 3c). The predictability of the KE jet speed thus revealed has been confirmed in the hindcast integration as significantly high correlation (r = 0.68) between the smoothed interannual variability in KE jet speed and SSHAs over the central North Pacific (30°–34°N, 170°E–175°W) 3 yr earlier (Fig. 4h). Although limited in their availability, the satellite-observed SSHA data and the geostrophically derived KE jet speed are also correlated if the same time lag is assigned. Still, further accumulation of observation is required for more robust assessment of the potential predictability of the KE jet speed variability. It is also noteworthy that because of correlation of the jet speeds in KE and in the Kuroshio off the southern coast of Japan, 13-month running mean current speed averaged between 125° and 155°E along their axis also suggests some predictability (with correlation r = 0.58 with the 3-yr lead SSHAs in the central North Pacific), although the correlation is somewhat lower than that for the KE jet speed itself.
The suggested predictability of the KE jet speed is rather modest, since only 46% of its total variance of the smoothed time series can be explained by SSHAs over the central basin with 3-yr lead time. There are at least two possible reasons for the limited predictability. One is influence of wind variations in the western portion of the North Pacific basin, which can reduce predictability with longer lead time. The other reason is the presence of intrinsic, internal variability of KE jet that is uncorrelated with atmospheric forcing and can yield certain uncertainty in the KE jet variability. Significance of its impact is under investigation.
Acknowledgments
The OFES simulations were conducted on the Earth Simulator under the support of JAMSTEC. We thank Drs. Y. Masumoto, H. Sakuma, T. Yamagata, and the OFES group members for their efforts and supports in the model development. Comments from Drs. S. Ito, H. Saito, A. J. Miller, and anonymous reviewers were very helpful. This study is supported in part through the research project Population Outbreak of Marine Life (POMAL) by the Agriculture, Forestry, and Fisheries Research Council of Japan and through Grant-in-Aid for Scientific Research in Innovative Areas 2205 (Grant 22106006) by the Japanese Ministry of Education, Culture, Sports, Science and Technology, and also through Grant 21540458 by the Japan Society for the Promotion of Science.
REFERENCES
Deser, C., M. A. Alexander, and M. S. Timlin, 1999: Evidence for a wind-driven intensification of the Kuroshio Current Extension from the 1970s to the 1980s. J. Climate, 12, 1697–1706.
Ducet, N., P.-Y. Le Traon, and G. Reverdin, 2000: Global high resolution mapping of ocean circulation from TOPEX/Poseidon and ERS-1 and -2. J. Geophys. Res., 105, 19 477–19 498.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Kwon, Y.-O., M. A. Alexander, N. A. Bond, C. Frankignoul, H. Nakamura, B. Qiu, and L. Thompson, 2010: Role of the Gulf Stream and Kuroshio–Oyashio systems in large-scale atmosphere–ocean interaction: A review. J. Climate, 23, 3249–3281.
Levitus, S., J. I. Antonov, T. P. Boyer, R. A. Locarnini, H. E. Garcia, and A. V. Mishonov, 2009: Global ocean heat content 1955–2008 in light of recently revealed instrumentation problems. Geophys. Res. Lett., 36, L07608, doi:10.1029/2008GL037155.
Masumoto, Y., and Coauthors, 2004: A fifty-year eddy-resolving simulation of the World Ocean—Preliminary outcomes of OFES (OGCM for the Earth Simulator). J. Earth Simul., 1, 35–56.
Miller, A. J., D. R. Cayan, and W. B. White, 1998: A westward-intensified decadal change in the North Pacific thermocline and gyre-scale circulation. J. Climate, 11, 3112–3127.
Mochizuki, T., and Coauthors, 2010: Pacific decadal oscillation hindcasts relevant to near-term climate prediction. Proc. Natl. Acad. Sci. USA, 107, 1833–1837, doi:10.1073/pnas.0906531107.
Nishikawa, H., and I. Yasuda, 2011: Long-term variability of winter mixed layer depth and temperature along the Kuroshio jet in a high-resolution ocean general circulation model. J. Oceanogr., 67, 503–518.
Nishikawa, H., I. Yasuda, and S. Itoh, 2011: Impact of winter-to-spring environmental variability along the Kuroshio jet on the recruitment of Japanese sardine. Fish. Oceanogr., 20, 570–582.
Nonaka, M., H. Nakamura, Y. Tanimoto, T. Kagimoto, and H. Sasaki, 2006: Decadal variability in the Kuroshio–Oyashio Extension simulated in an eddy-resolving OGCM. J. Climate, 19, 1970–1989.
Nonaka, M., H. Nakamura, Y. Tanimoto, T. Kagimoto, and H. Sasaki, 2008: Interannual-to-decadal variability in the Oyashio and its influence on temperature in the subarctic frontal zone: An eddy-resolving OGCM simulation. J. Climate, 21, 6283–6303.
Pacanowski, R. C., and S. M. Griffies, 2000: MOM 3.0 manual. Geophysical Fluid Dynamics Laboratory/National Oceanic and Atmospheric Administration, 680 pp.
Pierini, S., 2006: A Kuroshio Extension system model study: Decadal chaotic self-sustained oscillations. J. Phys. Oceanogr., 36, 1605–1625.
Qiu, B., 2003: Kuroshio Extension variability and forcing of the Pacific decadal oscillations: Responses and potential feedback. J. Phys. Oceanogr., 33, 2465–2482.
Qiu, B., and S. Chen, 2005: Variability of the Kuroshio Extension jet, recirculation gyre, and mesoscale eddies on decadal timescales. J. Phys. Oceanogr., 35, 2090–2103.
Qiu, B., and S. Chen, 2010: Eddy-mean flow interaction in the decadally modulating Kuroshio Extension system. Deep-Sea Res. II, 57, 1098–1110, doi:10.1016/j.dsr2.2008.11.036.
Qiu, B., N. Schneider, and S. Chen, 2007: Coupled decadal variability in the North Pacific: An observationally constrained idealized model. J. Climate, 20, 3602–3620.
Sasaki, H., M. Nonaka, Y. Masumoto, Y. Sasai, H. Uehara, and H. Sakuma, 2008: An eddy-resolving hindcast simulation of the quasi-global ocean from 1950 to 2003 on the Earth Simulator. High Resolution Numerical Modelling of the Atmosphere and Ocean, K. Hamilton and W. Ohfuchi, Eds., Springer, 157–185.
Sasaki, Y. N., and N. Schneider, 2011: Decadal shifts of the Kuroshio Extension jet: Application of thin-jet theory. J. Phys. Oceanogr., 41, 979–993.
Schneider, N., and A. J. Miller, 2001: Predicting western North Pacific Ocean climate. J. Climate, 14, 3997–4002.
SSALTO/DUACS, 2011: SSALTO/DUACS user handbook: (M)SLA and (M)ADT near-real time and delayed time products. Centre National D’études Spatiales Tech. Rep. CLS-DOS-NT-06-034, Issue 2.4, 49 pp.
Taguchi, B., S.-P. Xie, N. Schneider, M. Nonaka, H. Sasaki, and Y. Sasai, 2007: Decadal variability of the Kuroshio Extension: Observations and an eddy-resolving model hindcast. J. Climate, 20, 2357–2377.
Taguchi, B., B. Qiu, M. Nonaka, H. Sasaki, S.-P. Xie, and N. Schneider, 2010: Decadal variability in the Kuroshio Extension: Mesoscale eddies and recirculations. Ocean Dyn., 60, 673–691.
Tanaka, K., M. Ikeda, and Y. Masumoto, 2004: Predictability of interannual variability in the Kuroshio transport south of Japan based on wind stress data over the North Pacific. J. Oceanogr., 60, 283–291.
The meridional span of the linear model used for their prediction is 35°–40°N, although its results are compared with SST at 40°N in their Fig. 1.
As shown in Fig. 5, the simulated KE jet axis tends to be situated slightly to the north of its counterpart in observations. As a result, the interannual variability in Fig. 1 is weaker in OFES, especially in 1990–95, as the area-mean heat content is less influenced by strong variability of the KE jet. This bias in the KE latitude, however, does not affect our analysis of predictability of the KE jet speed, which is defined at the jet axis rather than at prescribed latitude.
The corresponding skills based on the first- and second-order autoregression simulations are found to be lower than that for the persistent anomalies. Although the KE jet speed time series shows low-frequency variability, its phase changes because of the nonlocal Rossby wave propagation and not because of some stochastic process. Then, the autoregressive models cannot represent the phase changes.
