1. Introduction

The wintertime mean atmospheric circulation over the Far East and northwestern (NW) Pacific is characterized by a low-level northwesterly monsoonal flow between the Siberian high and Aleutian low and also by an extremely intensified westerly jet aloft. With this steady monsoonal flow, this region marks the strongest poleward heat transport over the Northern Hemisphere in the climatological mean (van Loon and Williams 1979; Higuchi et al. 1991; see also Fig. 1a). Over this region, the near-surface meridional temperature gradients are extremely high in midlatitudes (Fig. 1e), where the cold, dry monsoonal air encounters a warm air mass to the south. In addition, an abundant supply of heat and moisture occurs from the warm ocean surface to the monsoonal air (e.g., Esbensen and Kushnir 1981; see also Fig. 1g). These two effects sustain high lower-tropospheric baroclinicity (Hoskins and Valdes 1990; Nakamura 1992), which feeds migratory baroclinic eddies to form a well-defined storm track¹ downstream (Blackmon et al. 1977, 1984; Wallace et al. 1988; see also Fig. 1b). It is marked with a belt of local maxima in precipitation almost across the Pacific basin (Xie and Arkin 1997; Trenberth and Guillemot 1998; see also Fig. 1h). These migratory eddies act to relax the temperature gradients by systematically transporting sensible heat to higher latitudes (e.g., Blackmon et al. 1977; Nakamura 1992; see also Fig. 1b).

It became apparent that the east Asian winter monsoon and associated meridional heat transport both undergo substantial modulations with interannual timescales (van Loon and Williams 1979; Higuchi et al. 1991; Zhang et al. 1997). It has now started to be realized that the winter monsoon is modulated also on decadal timescales (Koide and Kodera 1999; Nakamura and Yamagata 1999), with its weakening since the late 1980s as a distinct example (Watanabe and Nitta 1999). These modulations could potentially alter the activity of migratory disturbances along the storm track through the associated changes in the near-surface baroclinicity and jet intensity aloft over the Far East, the entrance region of the Pacific storm track. The anomalous growth of the eddies in the storm track entrance may influence the wintertime-mean circulation downstream through anomalous eddy vorticity flux convergence (Peng and Whitaker 1999). Anomalies in the monsoon and storm track may also accompany significant changes in the hydrological cycle over the North Pacific through anomalous evaporation from the warm ocean surface and precipitation along the storm track as well. Therefore, the interannual and decadal modulations in the winter monsoon and Pacific storm track could have significant climatic implications.

The relationship is, by nature, interactive between the activity of the eddies migrating along a storm track and the mean flow in which the eddies are embedded. In fact, stationary anomalies in the basic flow are subject to the continual feedback forcing from those eddies (e.g., Lau 1988; Lau and Nath 1991). Still, a significant part of the interaction may be understood at least qualitatively, once anomalies in the basic flow are specified. Lau (1988) showed that the observed anomalous behavior of the major Northern Hemisphere storm tracks, in general, can be interpreted reasonably well in view of how an anomalous basic flow reorganizes the storm track. A linear stability analysis applied by Robertson and Metz (1989) to several anomalous basic flows suggests that the reorganizing effect by an anomalous mean flow may be interpreted reasonably well within the framework of the linear theory of baroclinic instability. The effect includes the modification of the mean baroclinicity and the steering effect acting on the synoptic-scale eddies. Interpretation of the relationship observed between changes in the Pacific storm track and east Asian monsoon, however, may not necessarily be straightforward. This is because not only of the presence of the feedback forcing from the storm track, but also of the presence of a complex relationship between mean baroclinicity and eddy amplitude inherent to the Pacific storm track (Nakamura 1992). This relationship is manifested as a striking midwinter minimum in the climatological-mean storm track activity in spite of locally enhanced mean-flow baroclinicity (Nakamura 1992). This phenomenon has been reproduced in general circulation models (GCMs; Christoph et al. 1997; Zhang and Held 1999) and even in a linear stochastic model (Zhang and Held 1999). Very recently, Chang (2001) investigated seasonal and interannual variations in the Pacific storm track activity based on a GCM simulation and by comparing observational circulation statistics between a couple of contrasting winter seasons. He found that fairly systematic changes in the eddy structure and group velocity are important in the interannual variations. He also found that fall–winter differences in the precipitation processes associated with individual cyclones are more important in the particular seasonal cycle in the Pacific storm track.

In the present study, we document the interannual and (quasi) decadal changes in the storm track activity observed over the NW Pacific for 17 recent years. We also document the associated changes in the hydrological cycle over the North Pacific basin. As shown in the previous studies mentioned above, this period includes a profound wintertime climatic change over the Far East and North Pacific, associated with the weakening of the monsoonal flow and upper-tropospheric jet that occurred in the late 1980s. Specifically, we are interested in what kind of changes occurred in the Pacific storm track in its activity and feedback forcing over its entrance region and farther downstream as well in association with the changes in the monsoonal flow. We are also interested in any indication of the associated changes, as ensemble, in the structure of the eddies migrating along the storm track. In specific, we are interested in zonal phase alignment between eddy velocity components and temperature fluctuations, which could influence the efficiency for the eddies in converting energy from the mean flow for their growth. A brief summary of the present study is found in Nakamura and Izumi (1999).

2. Data and analysis procedures

In the present study, we use 6-hourly global circulation data based on the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalyses (Kalnay et al. 1996) from 1979 (autumn) to 1995 (summer). The data quality should be better for this period than for the earlier period, because of satellite-derived information available for the reanalysis scheme.

As in Nakamura and Wallace (1990) and Nakamura (1992), fluctuations associated with migratory synoptic-scale disturbances whose time periods are typically shorter than a week have been extracted by means of digital high-pass filtering with a half-power cutoff period of 8 days.² In recognition of the baroclinic nature of those disturbances particularly in the lower troposphere, an instantaneous storm track activity at a particular location is measured as the 8-day low-pass-filtered meridional heat flux (υ_hT_h) at the 850-mb level. The flux was further smoothed by 31-day moving averaging, in order for emphasizing its seasonal modulation. At the upper troposphere, a local, instantaneous activity of a storm track is measured by what may be called “envelope function” of 250-mb geopotential height (Z_env) as in Nakamura and Wallace (1990) and Nakamura (1992). It is defined as (the square root of) the squared time series of the 8-day high-pass-filtered geopotential height, which was then smoothed by 8-day low-pass filtering after being multiplied by 2. The square-rooted envelope function corresponds to the local, instantaneous amplitude of 250-mb height fluctuations with periods shorter than 8 days associated with the migratory eddies. A component of the feedback forcing from eddies migrating along the storm track has been evaluated as the 250-mb westerly tendency (∂U₂₅₀/∂t) the eddies act to induce through their vorticity flux at that level. The flux based on the 8-day high-pass-filtered data has been exposed to the 8-day low-pass filtering, so as to represent the feedback forcing independent of the phase of the individual eddy components (Nakamura 1992).

A stationary component that includes the mean westerly jet, low-level monsoonal circulation, and associated vertical motion and thermal fields has been extracted through 31-day averaging. In this study the meridional heat transport due to the persistent, stationary eddies including the monsoonal flow is evaluated at each location as the product of the 850-mb temperature and meridional velocity (υ*T*) based on their departures from the corresponding zonal means in the 31-day averaged fields. The transport averaged over the midlatitude Far East may be used as an indicator of the monsoon activity. In addition, we identified stationary disturbances with submonthly timescales in the 8-day low-pass-filtered fields of wind and temperature from which their corresponding 31-day running mean values had been removed. A local meridional heat flux associated with the submonthly stationary disturbances was evaluated by computing the product of the 850-mb temperature and meridional wind velocity in their difference fields and then applying 31-day averaging to it (υ_lT_l).

In our brief examination of the hydrological cycle over the North Pacific, we use not only the precipitation data based on the NCEP–NCAR reanalyses but also monthly data taken from the so-called Climate Prediction Center (CPC) Merged Analysis of Precipitation dataset (CMAP; Xie and Arkin 1997). In constructing the dataset, rain gauge measurements over land were combined with estimates over the oceans based on infrared and microwave radiation measurements by satellites, while the data from the NCEP–NCAR reanalyses were substituted for data missing grids mainly over the Arctic and southern oceans. Hence, the CMAP data are believed to bear higher credibility than the reanalysis precipitation data themselves that depend on model parameterization schemes (Trenberth and Guillemot 1998). Likewise, in our examination of turbulent fluxes of sensible and latent heat over the ocean surface, we use the reanalysis products and estimates based on in situ ship measurements in the Comprehensive Ocean–Atmosphere Data Set (COADS) archive as well. The former depends on the parameterization scheme of the surface boundary layer in the reanalysis, whereas the latter estimation suffers from insufficient time intervals in measurements that are likely to result in the underestimation of the contribution from submonthly transients to the fluxes.

The climatological mean of a given variable including the eddy heat fluxes is defined as the 16-yr mean of the 31-day moving-averaged value for each calendar day. Anomalies associated with the interannual variability are then represented as the 31-day averaged deviations from the climatological mean for a given day, and the overall strength of the variability is measured as its standard deviation for a given calendar day. Bimonthly mean anomalies for January and February of the satellite precipitation and COADS-based surface heat fluxes were simply calculated from their monthly anomalies, whereas a bimonthly anomaly field for any variable based on the reanalyses was constructed as the simple algebraic mean of the two 31-day running mean anomaly fields, one for the mid-January and the other for mid-February.

3. Interannual variability in the storm track and monsoon activities

Our main analysis domain is limited within the Far East and the NW Pacific (20°–60°N, 100°E–180°), where both the northwesterly monsoonal flow (υ*T*) and storm track activity (υ_hT_h) in the lower troposphere are particularly pronounced in their climatological means for January and February (Figs. 1a and 1b, respectively). We did not include the eastern Pacific in the main analysis domain, because the monsoonal flow from the Asian continent does not spread eastward beyond the date line (Fig. 1a) and its interannual variability is minimized between 150°E and the date line (Fig. 1c). These statistics are consistent with those shown by Higuchi et al. (1991). Our domain includes the core of the climatological-mean Pacific storm track (Fig. 1b), where the interannual modulation in the activity is strongest (Fig. 1d). The domain also includes the core of a lower-tropospheric baroclinic zone (Fig. 1e). Under the strong advective effect of the Pacific jet, amplitude of migratory transient eddies and their feedback forcing in the form of the meridional heat flux (Fig. 1b) and vorticity flux (Fig. 1f) as well tend to be maximized around 170°E, to the east of the core of the baroclinic zone (145°E; Fig. 1e). In fact, as evident in the map of 500-mb vertical motion (ω₅₀₀), a thermally indirect overturning extends zonally along the storm track eastward of 160°E (Fig. 1e). In contrast, no indication of such a localized Ferrel-type cell is found to the west. Instead, the western half of our analysis domain is covered almost uniformly with subsidence whose strong center is located over the Yellow Sea (Fig. 1e), slightly poleward of the jet axis. Nakamura (1993) demonstrated that the upper-level convergence associated with this subsidence is of the primary importance in the vorticity balance within the entrance of the Pacific jet. There, a poleward ageostrophic flow of 5 ∼ 7 m s⁻¹ across the extremely tight absolute-vorticity gradient yields the strong advection of anticyclonic vorticity at the tropopause level. The region of this midtropospheric subsidence overlaps the southern half of the region of the enhanced 850-mb υ*T* over the Far East. This overlapping is consistent with the lower-tropospheric vorticity balance; that is, the divergence effect associated with the subsidence acts to compensate for the planetary-vorticity advection by the monsoonal northerlies.

To elucidate the typical seasonal march and interannual variability in the monsoonal heat transport and storm track activity, the climatological means and standard deviations of the 850-mb υ*T* and υ_hT_h were averaged zonally within the main analysis domain. The mean υ*T* exhibits a single, well-defined maximum around 50°N in January (Fig. 2a). Its strongest interannual variability is observed around 60°N in December, and the secondary maximum appears around 45°N in the midwinter months of January and February (Fig. 2c). The climatological mean and interannual variability of υ_hT_h are both confined around the storm track axis between 40° and 45°N throughout the cold season (Figs. 2b and 2d). As found by Nakamura (1992), the mean υ_hT_h is markedly suppressed in midwinter between its fall and spring maxima. Still, the most vigorous interannual variability in υ_hT_h tends to occur in midwinter. This tendency may be a manifestation of the fact that the interannual variability in the seasonal-mean westerlies and baroclinicity is maximized in midwinter. At the same time, it implies that the feedback forcing from the storm track, in turn, on the anomalous seasonal-mean circulation is also maximized in midwinter. We henceforth focus on the midwinter situation where the storm track activity and monsoonal heat transport over the midlatitude NW Pacific both undergo the most vigorous interannual fluctuations in the year.

In order to identify the most prevailing interannual variability in the midwinter storm track activity within the analysis domain (Fig. 1d), a conventional empirical orthogonal function (EOF) analysis was applied to bimonthly (January–February) anomalies of υ_hT_h from 1980 to 1995. The first EOF accounts for nearly 50% of the total interannual variance within the domain. This mode represents anomalous υ_hT_h along the climatological-mean storm track over the NW Pacific (Fig. 3a), where its inteannual variability is the strongest (Figs. 1d and 2d). Typical seasonal-mean circulation anomalies associated with the first mode of υ_hT_h were identified in their linear regression maps with the principal component time series of that mode (PC1; see Fig. 5a). They represent a typical midwinter situation with enhanced storm track activity over the NW Pacific. The sign of each anomaly is reversed when the storm track activity is abnormally suppressed. The enhancement of the midwinter υ_hT_h tends to occur in conjunction with the lower-tropospheric anticyclonic anomalies that cover the entire midlatitude North Pacific and weaker cyclonic anomalies over Siberia as well (Fig. 4a). In other words, the Aleutian low and Siberian high both tend to weaken when the Pacific storm track is intensified, as shown in Fig. 5b, although the anomaly centers in Fig. 4a do not exactly coincide with the climatological-mean centers of the low and high. These low-level anomalies in Fig. 4a are accompanied by upper-tropospheric pressure anomalies that resemble the Pacific–North American (PNA) teleconnection pattern (Fig. 4b). Unlike the corresponding pattern identified by Wallace and Gutzler (1981), its anticyclonic “tail” extends as far westward as China, forming a seesaw with cyclonic anomalies over Siberia. The corresponding lower-tropospheric anticyclonic anomalies do not extend as westward as China (Fig. 4a). Though less obvious, a similar westward shift of the upper-level anomalies relative to their near-surface counterpart is found along the west coast of North America. This baroclinic structure is manifested also as significant anomalies in the 850-mb temperature (T) and υ*T* (Figs. 4c and 4d, respectively) over the midlatitude Far East and northeastern (NE) Pacific. The stationary anomalies shown in Fig. 4 are in reasonable agreement with van Loon and Williams (1979), who examined the stationary circulation anomalies associated with the anomalous υ*T* in winter over the Far East during the period of 1949–78.

It is hinted in Figs. 3a and 4a that the relationship between the anomalous westerlies and anomalous storm track activity is different between the NW and NE Pacific. This difference is more apparent in the corresponding maps for the upper troposphere (Figs. 3b and 4b). Over the NE Pacific, the storm track activity tends to be deflected poleward at the upper and lower levels in the presence of stationary anticyclonic anomalies in midlatitudes. In the situation shown in Figs. 4a and 4b, the axis of the mean westerlies that steer synoptic-scale eddies is also shifted poleward, and the mean baroclinicity is anomalously high and low along the northern and southern flanks of the anticyclonic anomalies, respectively, whose vertical structure is approximately equivalent barotropic. In the upper troposphere, another significant anomaly in the storm track activity is found to the southwest of Hawaii (Fig. 3b), where the anomalous seasonal wind tends to be westerly. The relationship between the anomalous storm track activity and stationary anomalies over the NE Pacific is in agreement with Lau (1988), and it is understandable in the framework of linear theory of baroclinic instability as analyzed in Robertson and Metz (1989). Over the NW Pacific, in contrast, the storm track activity tends to increase in spite of the anomalous easterlies in the upper troposphere (Figs. 3 and 4b) and the reduced lower-tropospheric baroclinicity along the storm track (Fig. 4c). This result is in good agreement with the empirical relationship between baroclinic eddy amplitude and mean westerly speed obtained for the NW Pacific by Nakamura (1992, his Figs. 10a and 12a) and in a GCM simulation by Carillo et al. (2000). The relationship found between the anomalous storm track activity and basic-state anomalies over the NW Pacific cannot be interpreted in the framework of the linear theory of baroclinic instability.

4. Interannual variability in poleward heat transport over the Far East

A striking feature found in comparison between Figs. 3a and 4d is that the enhanced heat transport by migratory eddies (υ_hT_h) along the Pacific storm track tends to be accompanied by the suppressed monsoonal heat transport (υ*T*) over the Far East and vice versa. For a more quantitative argument, anomalies of each of υ_hT_h, υ*T*, and υ_lT_l at the 850-mb level regressed linearly upon PC1 of υ_hT_h at the same level were averaged longitudinally from 100°E to 180°. Thus obtained anomalous heat fluxes are shown in Fig. 6a as functions of latitude, together with anomalies in the net poleward sensible heat flux defined as the sum of the three terms. Between 35° and 50°N in midwinter, 50%–65% of the anomalous monsoonal heat transport (υ*T*) tends to be compensated by the anomalous heat transport by migratory eddies along the storm track (υ_hT_h) in the opposite sense. The anomalous heat flux due to the submonthly stationary disturbances (υ_lT_l) is negligible. Thus, the compensation between υ*T* and υ_hT_h leaves rather small changes in the net atmospheric sensible heat transport, defined as above, over the midlatitude NW Pacific from one winter to another. The sign of the anomalous net transport tends to be determined by that of υ*T* that dominates over υ_hT_h. In fact, anomalous υ_hT_h and υ*T* are correlated with PC1 in midlatitude with high statistical significance, whereas the correlation between the net transport and PC1 is not significant at any latitude (Fig. 6b). It should be noted that no such compensation between υ*T* and υ_hT_h anomalies occurs at higher latitudes where the former dominates the latter (Figs. 2c and 2d).

In recognition of some ambiguities in defining a local poleward heat flux associated with the stationary eddies (or monsoonal flow), it is necessary to confirm the above findings from a different viewpoint. Specifically, for each midwinter season we evaluated the local heating rate due to the temperature advection associated with the 850-mb seasonal-mean monsoonal flow (−v · ∇T) and that due to the 850-mb horizontal heat flux convergence by migratory eddies (−∇·v_hT_h). Unlike υ*T*, these two terms appear, as they are, in the thermodynamic equation that describes the local heat balance in the seasonal mean. Although they emphasize smaller-scale features, their interpretations are more straightforward than those of the fluxes. Maps of −∇·v_hT_h and −v · ∇T regressed linearly upon PC1 of 850-mb υ_hT_h (not shown) indicate that, in association with the enhanced storm track activity in the NW Pacific, anomalous heating and cooling due to −∇·v_hT_h tends to occur to the north and south of the climatological-mean storm track axis, respectively, whereas −v · ∇T tends to exhibit the opposite tendency acting to cancel the anomalous −∇·v_hT_h. Their cancellation, however, is by no means complete, as −v · ∇T is twice as strong as −∇·v_hT_h. Yet, the overall tendency is consistent with our findings shown in Fig. 6.

5. Modulated seasonal march

In this section, seasonal dependency of the out-of-phase relationship between interannual fluctuations in meridional heat transport associated with the Pacific storm track and east Asian winter monsoon is highlighted in their composite latitude–season sections. Our selection of winters used for the compositing was based on the PC1 time series for midwinter υ_hT_h at the 850-mb level shown in Fig. 5a. In each of the composites, two sets of five winters (i.e., 1986/87, 1988/89, 1989/90, 1991/92, and 1993/94; and 1980/81, 1982/83, 1983/84, 1985/86, and 1994/95) are contrasted, during which PC1 for midwinter was most strongly positive and negative, respectively. In most of the figures presented below, the 31-day running mean anomalies were averaged zonally between 100°E and 180° for each calendar day from October to April and then composited for each set of the five winters.

In the composite maps of the 850-mb υ_hT_h (Fig. 7) and υ*T* (Fig. 8), their significant differences between the two types of winter are observed mainly in midwinter, which lead to their mutual cancellation as shown in Fig. 6 only during that time of the year. In its seasonal march (Fig. 8), the monsoonal heat transport υ*T* exhibits a stronger and much sharper midwinter peak in the “negative” years than in the “positive” years. When υ*T* is strong (weak), its midwinter peak tends to occur to the south (north) of its climatological mean position (cf. Fig. 2a). More striking distinctions appear in a pair of the υ_hT_h composites in Fig. 7. In the negative winters, the storm track activity is strongly suppressed in midwinter, which yields a marked minimum in υ_hT_h between its fall and spring maxima as in the climatology shown by Nakamura (1992). In the positive winters, in contrast, the midwinter suppression does not occur, yielding almost constant υ_hT_h throughout the cold season with no midwinter minimum. This distinction is confirmed in Fig. 9, where upper-tropospheric eddy amplitude measured as Z_env was sampled daily in its 31-day running mean maps along the instantaneous storm track axis defined as the maximum υ_hT_h at each meridian and then composited separately for the positive and negative winters. In the negative winters, Z_env exhibits a marked midwinter minimum all the way across the North Pacific. In contrast, the midwinter minimum almost disappears in the positive winters. Significant difference is observed in midwinter all way across the Pacific storm track, with the largest difference in the central Pacific.

The seasonal march in the mean state is shown in the composites of U₂₅₀ (Fig. 10) and ΔU_700−1000 (Fig. 11). These composites indicate that both the upper-level jet and the associated lower-tropospheric baroclinicity over the NW Pacific tend to be stronger significantly in the negative winters than in the positive winters. At the same time, the jet and baroclinic zone tend to be meridionally narrower in the negative winters, which is apparent in the difference in their poleward extent (in Figs. 10a,b and 11a,b) and in the dipolar patterns in their anomalies (in Figs. 10c and 11c). These changes in the jet and baroclinicity occurred in conjunction with more enhanced southward spreading of a cold air mass in the negative winters than in the positive winters (Fig. 8). It is evident in Fig. 7 that the storm track axis tends to be shifted southward in the negative winters, relative to the situation in the positive winters. This shift is much more apparent than the corresponding shifts in the axes of the jet and baroclinic zone. Under the enhanced monsoonal flow, the storm track tends to be closer to the jet axis and low-level baroclinic zone whereas the activity of the eddies is strongly suppressed.

In the PC1 time series of the storm track activity shown in Fig. 5a, the decadal variability, which changed its polarity in the late 1980s, is apparent even without any smoothing imposed. Correspondingly, all of the five “positive winters” listed above are sampled from the period of the late 1980s to the mid-1990s, whereas 4 out of the 5 negative winters are in the early and mid-1980s. Though slightly less pronounced, the same kind of distinctions as above in the Pacific storm track activity and in the mean state were found, when the winter of 1994/95 was replaced with that of 1984/85 in the cluster of the negative winters for the compositing.³ It is suggested that substantial modulations occurred in the seasonal march of the Pacific storm track activity, in conjunction with the recent decadal-scale weakening tendency in the east Asian winter monsoon. In fact, the midwinter intensity of the Siberian high varied predominantly on decadal scales during the recent period, with a rapid weakening around 1987 (Fig. 5b). Though masked by the dominant year-to-year fluctuations associated mainly with ENSO events, a decadal-scale tendency is noticeable also in the intensity of the Aleutian low (Fig. 5b). It tended to be stronger until the late 1980s and weaker since then. The decadal modulation in the storm track activity is also evident in Fig. 12, where the 850-mb υ_hT_h sampled each day along the instantaneous storm track axis over the NW Pacific was plotted for individual cold seasons over the 17 years. In the first half of the 1980s, the midwinter suppression in the Pacific storm track activity was profound, leaving a marked minimum in the seasonal march. Since the 1986/87 winter the minimum almost disappeared due to the enhanced activity in midwinter, which continued until the 1993/94 winter. According to Fig. 5a, the situation in the 1994/95 winter was closer to the typical situation in the early to mid-1980s.

6. Possible physical mechanisms for the observed storm track changes

As shown in the preceding section, the observed midwinter suppression (enhancement) in the activity of the Pacific storm track occurred in spite of the intensified (relaxed) low-level baroclinicity and upper-level jet. The intensified baroclinicity favors the amplification of baroclinic eddies, but the intensified jet may counteract it by advecting the eddies too fast away from a zonally confined baroclinic zone (e.g., Lin and Pierrehumbert 1993). Nakamura (1992) speculated that the latter effect may contribute to the midwinter minimum in the climatological seasonal march in the Pacific storm track activity. In order to assess whether the interannual variations in that activity shown in the preceding section can be attributed to the excessive advective effect, we evaluated a parameter dG defined as dG = [f(ϕ_s)R^−1/2_i] × (U⁻¹_m a cosϕ_s dλ), where ϕ_s denotes the latitude of the storm track axis at a given meridian, a the radius of the earth, and dλ a longitudinal grid interval. In this definition, R^−1/2_i is the ratio of the thermal wind shear to the Brunt–Väisälä frequency, both of which were evaluated from the 700- and 850-mb temperature, and U_m is the zonal wind velocity averaged between the 250- and 700-mb levels. The evaluation was based on the 31-day running mean data. The quantity dG consists of two components. One of them proportional to R^−1/2_i represents a rough measure of the local growth rate of baroclinic eddies, because this parameter, when multiplied by the Coriolis parameter f as above, is proportional to the growth rates of the fastest growing normal modes given in the linear theories of Charney (1947) and Eady (1949). The other component including U_m corresponds to the residence time for the eddies in each grid interval, since U_m crudely measures the strength of the mean-flow advection acting on them. It may also provide us with a crude measure of the eastward group velocity of a baroclinic wave packet (Chang and Yu 1999). In case of a zonally confined baroclinic zone, the longer the residence time is, the more it favors the eddy amplification. The zonal integration of dG along the upstream half of the Pacific storm track from 100°E to 180° crudely measures the amplification that eddy components or their packets would undergo when traveling along the baroclinic zone. This integrated quantity is hereafter referred to as G.

The two components of dG were evaluated separately for each of the two sets of winters on the basis of the 5-yr composites of R^−1/2_i and U_m for each calendar day. Figure 13 shows the thus obtained seasonal marches of dG along the Pacific storm track. Interestingly, each of them exhibits a clear minimum in midwinter within the storm track entrance (around 120°E) in the course of winter. Between 100° and 145°E, dG in January and February is smaller for the strong monsoon years than for the weak monsoon years (Fig. 13). In each of the composites of R^−1/2_i and U_m (not shown), the midwinter peak is stronger in the former years than in the latter years. Figure 13 therefore suggests the overcompensating effect of the mean-flow advection in the former type of winter. The corresponding difference in G is also greatest in January and February (Fig. 14), during which the most distinct difference was observed in the Pacific storm track activity (Fig. 7). These results seem to indicate that the excessive advection by the Pacific jet contributed to the retarded eddy amplification in the first half of the 1980s, which is in good agreement with Chang (2001). It should be kept in mind, however, that as the residence time of eddies in the baroclinic zone is shortened by the enhanced mean-flow speed, a time period over which damping processes act on them is also shortened (Chang 2001). Since this effect counteracts the effect of the mean-flow advection discussed above, the latter should account for a smaller fraction of the observed changes in the storm track activity than Figs. 13 and 14 suggest.

Putting this possible overestimation aside, we find in Fig. 14 that G differs only about 5% between the composites for the strongest and weakest υ_hT_h. However, Fig. 7 shows that υ_hT_h in midwinter in the former composite was twice as much as in the latter composite, indicating that the observed difference in eddy amplitude was as much as ∼20% in the lower troposphere. The corresponding difference in the upper-tropospheric eddy amplitude was also about 20% (Fig. 9). Although a comparison between the observed changes in G and the eddy amplitude may not necessarily be straightforward, it is suggested that processes other than the mean-flow advection were probably involved in the observed changes in the storm track activity.

Nakamura (1992) showed that in climatology the correlation between the 850-mb temperature and meridional wind velocity in the high-pass-filtered fields tends to be lower in midwinter than in fall and spring. Similar changes in the temperature–velocity correlation associated with baroclinic eddies were found by Chang (2001) in his analysis of a GCM simulation and the observed data. As discussed in detail in the appendix, the velocity–temperature correlation in the high-pass-filtered data is an indicator of their phase relationship as an ensemble of migratory eddies. The correlation statistics are unlikely to influenced by the number of synoptic-scale disturbances or their amplitude observed within a particular season. One may wonder whether the observed changes in the midwinter storm activity as in Fig. 7 are attributable to the significant changes in eddy structure. In order to examine if that was really the case, a correlation analysis was applied at each grid point between the 8-day high-pass-filtered time series of the 850-mb temperature and meridional wind velocity for the January–February period separately for the five positive and negative winters defined in section 5. The same local correlation analysis was applied between the high-pass-filtered time series of the 500-mb temperature and vertical P velocity. The former and latter correlations are hereafter referred to as the 850-mb T–υ correlation and 500-mb T–ω correlation, respectively. For baroclinically growing eddies, the former correlation should be positive but the latter should be negative because of the sign convention.

Figure 15 contrasts each of the two types of correlation between the positive and negative winters. The 850-mb T–υ correlation is higher in the positive winters than in the negative winters. In the former winters the correlation exceeds 0.7 in many locations along the storm track (Fig. 15a), while in the latter winters it is generally below 0.7 and even below 0.6 in some locations along the storm track (Fig. 15b). The 500-mb T–ω correlation also exhibits similar differences. The negative correlation is −0.4 or even stronger throughout the Pacific storm track in the positive winters (Fig. 15c), whereas in the negative winters the correlation is substantially weaker in general and it is even weaker than −0.2 around Japan (Fig. 15d). From an energetic viewpoint, the stronger T–υ correlation means more efficient conversion of available potential energy from the mean westerlies to baroclinic disturbances that tend to be more baroclinic than in the climatological situation (see the appendix). Likewise, the stronger T–ω correlation means more efficient conversion from available potential energy to kinetic energy within those disturbances. The results shown in Fig. 15 suggest that the synoptic-scale eddies tended to lose their optimum structure for their baroclinic growth in the presence of the excessively strong westerlies in the negative winters. These results are consistent with those in the previous studies mentioned above.

7. Influence upon the hydrological cycle and atmosphere–ocean heat exchange

In order to examine whether the recent decadal variations in the storm track and monsoon activities over the NW Pacific as documented in the preceding sections were accompanied by any significant changes in the hydrological budget or atmosphere–ocean heat exchange over the North Pacific, January–February mean fields of precipitation, water vapor, and turbulent heat fluxes were linearly regressed upon PC1 of 850-mb υ_hT_h in Fig. 5a. The linear regression maps plotted in Fig. 16 represent typical midwinter anomalies in several specific aspects of the energy and hydrological cycles over the North Pacific, when the eddy activity along the storm track is intensified as in the late 1980s and early 1990s. The following argument is concerned with this particular situation. Sign of each anomaly should be reversed when arguing a situation as in the early to mid-1980s, during which the storm track activity was suppressed in the NW Pacific under the intensified monsoon. As shown in Figs. 3 and 4, the decadal modulation in the storm track activity observed in the NW Pacific was accompanied by significant changes in the storm track activity and seasonal-mean stationary circulation in other parts of the North Pacific. Accordingly, the corresponding anomalies in the heat and moisture budget also spread widely over the extratropical North Pacific. Midwinter precipitation significantly increases by 10%–15% along a belt lying in the midlatitude NW Pacific and curving northward into the Aleutian region (Fig. 16c), where the lower-tropospheric storm track activity is enhanced (Fig. 3a). Precipitation also increases significantly by 30%–60% in the tropical/subtropical NE Pacific, where the high-frequency disturbances tended to be more active particularly in the upper troposphere (Fig. 3). The precipitation increase there seems to be associated with enhanced convective rainfall, as suggested by significant negative anomalies in the corresponding regression map of outgoing longwave radiation (OLR; not shown). At the same time, precipitation decreases by 10%–15% in the midlatitude NE Pacific, where the storm activity tended to be suppressed in the presence of the weakened westerlies. Nearly the same anomaly pattern in the precipitation as in Fig. 16c can be obtained when the precipitation data in the NCEP–NCAR reanalyses was used in place of the CMAP data. The significance of the decadal-scale precipitation anomalies shown in Fig. 16c has been pointed out by Lukas (2001), who applied an EOF analysis to the wintertime CMAP data over the North Pacific. In his analysis a pattern essentially the same as in Fig. 16c was obtained as the second EOF, whose PC time series is dominated by the decadal variability that changed the sign in 1989. His first EOF, of course, captures the prevailing remote influence of ENSO on shorter timescales.

The decadal changes in the midwinter atmospheric circulation and storm track activity are accompanied by a coherent pattern of anomalous latent heat release from the ocean surface (Fig. 16a). Weakened spills of cold, dry air from the Asian continent in the winter of the weak monsoon suppress latent heat release by 5%–10% over the warm Kuroshio, its extension, and branches. Suppressed latent heat release farther to the east is due to the reduced surface westerlies associated with the weakened Aleutian low. The weakened low also reduces the transport of relatively warm and moist air from the south along the west coast of Canada, which acts to increase the evaporation from the ocean surface by 10%–15%. In the subtropics, situated to the south of the anticyclonic anomalies covering over the midlatitude North Pacific, the latent heat release from the ocean surface is enhanced by as much as 15% under the intensified northeasterly trades. Sensible heat flux anomalies are distributed in a manner very similar to that of the latent heat flux anomalies (Fig. 16b), but the former is about a half in magnitude of the latter in midlatitudes and even about 30% in the subtropics. Again, the above argument signifies a typical situation since the late 1980s, and the sign of each anomaly should be reversed in discussing the counterpart for the early to mid-1980s. These results based on the surface flux fields of the NCEP–NCAR reanalysis data are reproduced reasonably well on the basis of anomalous flux fields estimated from the COADS archive.

With respect to the hydrological cycle, the increased rainfall over the subtropics that occurs in the proxy of the region of enhanced evaporation implies the availability of a local moisture supply. Indeed, no significant anomalous convergence of an atmospheric moisture flux is found in that subtropical region (Fig. 16d). In midlatitudes, the increased precipitation along the intensified storm track in the region of the suppressed evaporation over the Kuroshio and the opposite situation over the NE Pacific yield the anomalous net moisture sink and source in these respective regions. Accordingly, we observe significant anomalous convergence and divergence of the moisture flux along the western half of the Pacific storm track and over the extreme NE Pacific, respectively. Correspondingly, a clear signal of the anomalous westward transport of moisture is found along the storm track across the Pacific. Though less pronounced, another anomalous moisture transport into the entrance of the intensified storm track is present from the subtropics. Figure 16d represents a typical situation since the late 1980s, and those anomalous moisture transports tended to be reversed in the first half of the 1980s. The amount of the anomalous zonal moisture transport is such that it could compensate most of the net anomalous moisture deficit around Japan as inferred from Figs. 16a and 16c if the transport as in Fig. 16d occurs within the lowest 3 km of the atmosphere. It is concluded that the observed modulation in the storm track activity over the North Pacific in the presence of the decadal weakening of the winter monsoon and Aleutian low was accompanied by significant changes in precipitation, evaporation, moisture transport, and atmosphere–ocean heat exchange over the entire North Pacific. It should be kept in mind that the anomalous moisture transport shown in Fig. 16d is probably more or less overestimated, because the precipitation decrease along the climatological-mean storm track in the NE Pacific since the late 1980s appears to be somewhat overestimated in the reanalyses compared to the CMAP data.

8. Summary and discussion

We have shown that a profound modulation in the seasonal march in the NW Pacific storm track activity occurred concomitantly with the decadal weakening of the east Asian winter monsoon (Siberian high) and Aleutian low during the 1980s and 1990s. The most notable signature of this modulation is the tendency that the midwinter minimum in storm track activity became less apparent in the second half of this period than in the first half, in association with a increasing tendency in the midwinter activity. This enhancement of the storm track activity occurred despite the reduced mean-flow baroclinicity and jet intensity aloft, which is opposed to linear theory of baroclinic instability and is analogous to the eddy–mean-flow relationship in the midwinter suppression within the climatological-mean seasonal cycle (Nakamura 1992). At this stage, no solid argument can be made on what mechanism caused that particular decadal enhancement in the midwinter Pacific storm track activity. Yet, our analysis suggests that, as the excessively strong westerlies weakened, the eddies changed their structure so as to be more efficient in converting the mean-flow available potential energy into their kinetic energy for their growth. Our analysis also implies that the weakening of the Pacific jet in the late 1980s prolonged the residence time for growing eddies within the zonally confined baroclinic zone beneath the jet. This particular effect favored the amplification of those eddies and seemingly dominated over the effect of the reduced mean baroclinicity, although this effect appears to be of secondary importance. Our findings are in good agreement with Chang (2001), who showed that the reduced energy conversion from the mean-flow available potential energy and increased eastward group velocity both contributed to the suppression of downstream eddy growth along the Pacific storm track in a winter when the Pacific jet is particularly strong.

This observed relationship between the anomalous storm track activity and mean baroclinicity shown in this paper is found unique to the NW Pacific, as the corresponding relationship is found different for another major storm track over the NW Atlantic. The same analysis as in section 3 applied to the January–February anomaly fields within (30°–60°N, 100°–30°W) reveals that the strongest interannual variability in υ_hT_h over New England and Newfoundland (Fig. 17a). The variability is accompanied by a meridional dipole of T anomalies whose node coincides with the belt of the υ_hT_h anomalies (Fig. 17c). The dipole is confined over the NW Atlantic and eastern North America. The associated geopotential height anomalies at the 850- and 250-mb levels also form meridional dipoles (e.g., Fig. 17d), but they are shifted downstream covering the entire North Atlantic. Their nodes lie along the axis of the strongest anomalies in the 250-mb storm track activity (not shown). Their polarities are such that the intensified storm track appears to be in response to the increased baroclinicity around the storm track entrance in a manner consistent with the linear theory and to a better-defined waveguide associated with the enhanced westerly jet as well. The eastward shift of the geopotential dipoles and their barotropic structure both suggest that they are generated and maintained, at least in part, by a barotropic feedback from the anomalous storm track activity in the upper troposphere. The weak baroclinic component in the seasonal-mean circulation anomalies gives rise to weak anomalies in υ*T* over the NE part of the United States, as shown in Fig. 17b. Since their sign is opposite to that of the corresponding υ_hT_h anomalies, these two anomalies act to compensate each other as observed in the NW Pacific. However, mechanisms behind such compensation in the two components of the meridional heat transport are not the same between the two major storm tracks.

As mentioned earlier, the presence of the feedback forcing from eddies migrating along the storm track renders it difficult to identify any causal relationship between changes in the mean flow and eddy activity as obtained through our analysis. Over the central and eastern Pacific, the difference map of ω₅₀₀ indicates a clear signature of the enhancement of a thermally indirect cell to the north of the mean storm track axis (Fig. 18a), which is consistent with the northward deflection of the storm track extracted in the first EOF (Fig. 3a). In addition, the westerly acceleration through eddy vorticity flux tended to be significantly stronger just to the north of the storm track axis and weaker over the subtropics in the positive winters than in the negative winters (Fig. 18b). This upper-level barotropic forcing includes a component that is in phase with the observed U₂₅₀ anomalies, and it acts to vary the low-level wind accordingly via ageostrophic circulation. Thus the system appears to be interactive over the central and eastern North Pacific, and hence the anomalous storm activity was likely to contribute to the observed anomalous Aleutian low and associated anomalies aloft as in Figs. 4a and 4b. Over the midlatitude western Pacific and Far East, in contrast, no significant difference was found in the barotropic feedback forcing from the anomalous storm track activity (Fig. 18b). Furthermore, a significant anomaly pattern evident over the midlatitude Far East in the difference map of ω₅₀₀ is in thermally direct sense (Fig. 18a), which cannot be attributed to the enhanced storm track activity in the positive winters. Therefore, the anomalies in temperature and vertical motion observed over the Far East are unlikely to be generated in direct response to the feedback forcing from the local storm track. One may argue that quasi-stationary anomalies generated in response to the anomalous feedback forcing from migratory eddies in the core and exit regions of the Pacific storm track may retrograde slowly and possibly give rise to the temperature anomalies in the entrance region of the storm track over the Chinese continent as in Fig. 4c. However, no such indication of retrogression was found in a Hovmöller diagram of 8-day low-pass-filtered 250-mb height anomalies for the January–February period for any of the five “positive” or five “negative” winters (not shown). As discussed later, the anomalous vertical motion over the Far East might occur in association with an upper-level stationary wave train propagating across Eurasia.

Nakamura and Yamagata (1999) pointed out that 2 of the 3 dominant patterns of the decadal variability in the NW Pacific wintertime SST changed their polarities in the later half of the 1980s. One of them is what may be called a “midlatitude pattern,” characterized by SST anomalies along the subarctic front associated with the anomalous Aleutian low and the PNA teleconnection pattern aloft. Their polarities changed around 1988/89. The other represents SST anomalies in the East China Sea, which is considered to be generated in thermal response to the decadal modulation of the east Asian winter monsoon. The associated atmospheric pattern changed its polarity around 1986/87,⁴ just when the weakening of the Siberian high occurred (Fig. 5b). Our analysis seems to capture some systematic changes in the Pacific storm track activity associated with the polarity changes of those two patterns of the decadal climate variability. In fact, the unsmoothed time series of PC1 of υ_hT_h is dominated by a decadal signal and its sign was reversed around 1987 (Fig. 5a). Moreover, the January–February Z₂₅₀ anomalies correlated with PC1 (Fig. 4b) are dominated by a combined signal of the PNA pattern and the upper-level signature of the decadal modulation in the winter monsoon, both of which are shown in Nakamura and Yamagata (1999). The latter is characterized by a stationary wave train across the northern Eurasian continent that resembles the Eurasian (EU) teleconnection pattern defined by Wallace and Gutzler (1981). In order for estimating where the wave train originates, a phase-independent flux of wave activity pseudomomentum defined for stationary Rossby waves on a zonally asymmetric climatological-mean flow by Takaya and Nakamura (2001) was applied to the January–February mean anomalies of 500-mb height that had been regressed upon PC1. The flux, which is parallel to the local group velocity of the stationary Rossby waves, appears to originate from the North Atlantic reaching into the Far East (Fig. 19). It may suggest a possible Atlantic influence upon the Far East and NW Pacific on decadal timescales as postulated by Xie et al. (1999). In Fig. 19, the wave activity flux is converging into the anticyclonic anomalies over the Far East, implying that the particular anomalies may form at the leading edge of the stationary Rossby wave train. Since the stationary anomalies are situated in the strong upper-level westerlies, anomalous midtropospheric uprising must occur along the upstream flank of those anomalies to meet the upper-tropospheric vorticity balance, which is consistent with the anomalous midtropospheric vertical velocity in the “positive winters” shown in Fig. 18a. From the viewpoint of the lower-tropospheric vorticity balance, the associated anomalous low-level convergence is also consistent with the weakening of the monsoonal northerlies as shown in Fig. 4a.

Figure 16 indicates that the decadal-scale modulation observed in the east Asian winter monsoon and the associated modulation in the storm track activity influenced the atmosphere–ocean heat exchange and the hydrological cycle over the entire North Pacific as well. The weakening of the winter monsoon in the late 1980s caused the reduced evaporation over the extreme NW Pacific around Japan, which contributed to the observed warming in this maritime region (Nakamura and Yamagata 1999). The weakening of the Aleutian low contributed to the reduction and enhancement in surface evaporation over the midlatitude North Pacific and off the west coast of North America, respectively, which reinforced the warming and cooling of SST observed in the respective regions (Nakamura et al. 1997; Nakamura and Yamagata 1999). The enhancement and suppression in the storm track activity observed over the NW and NE midlatitude Pacific, respectively, were accompanied by the increased and decreased precipitation in these respective regions (Fig. 16c). These decadal anomalies in precipitation may be influential on the ocean circulation through changing density structure in the upper ocean. In fact, Lukas (2001) observed a marked increase in salinity in 1990/91 at the main thermocline level and above (between 150 and 350 m in depth) near the Oahu Island of Hawaii. Isopycnal surfaces crossing those depths at that particular location are ventilated within the oceanic subtropical frontal zone located to the north. This observed salinity change seems consistent with the below-normal precipitation during the first half of the 1990s over the oceanic subtropical frontal zone (Lukas 2001; see also Fig. 16c), allowing for a time lag with which the salinity increase thus generated in the mixed layer over the frontal zone was subducted and then advected southward by a geostrophic current. It should be mentioned that the characteristics of deepest mixing also depend on intensity and frequency of strongest storms in each winter season. Hence, the storm track variations as shown in the present study may be important in the oceanic variability not only through changing the net supply of freshwater at the ocean surface but also through altering the strength of subduction.

Although the aforementioned profound signatures were found in the present study, our analysis is based on a data period that is obviously too short to capture the robust characteristics of the decadal-scale storm track variability. Therefore, our analysis should be regarded as a case study concerning a single transition event associated with the decadal weakening of the winter monsoon that occurred in the late 1980s. An attempt is currently under way to examine whether any coherent pattern as identified in the 1980s and 1990s can be extracted for the wintertime North Pacific from the entire 40-yr period of the NCEP–NCAR reanalyses. A preliminary result suggests that the leading EOF of the variability in the NW Pacific storm track activity in midwinter and associated seasonal-mean anomalies in the wind and thermal fields are very similar to those presented in the current study.