1. Introduction

The winter stratosphere is occasionally strongly coupled to the troposphere. On the synoptic timescale, there are frequently upward-propagating planetary waves, both stationary and transient, transporting energy to the stratosphere, sometimes causing stratospheric sudden warming. On the monthly to seasonal timescale, there are thermal disturbances, which are generated by the imbalance of radiative cooling and wave-induced dynamical heating in the upper stratosphere, propagating downward to affect tropospheric circulations. This coupling has been demonstrated as a natural mode of variability called the Arctic Oscillation (AO) in the Northern Hemisphere (Thompson and Wallace 1998, 2000). The upward-propagating effect has been well documented in theoretical and observational studies of stratospheric sudden warming (Matsuno 1971; Holton 1976; Labitzke 1982; Quiroz 1986; Pierce and Fairlie 1993). However, there have been fewer studies addressing the downward-propagating effect, probably because the effect is less clear, indirect, or not readily separated from the high-frequency fluctuations of tropospheric weather systems.

Recently, Kodera et al. (2000) have studied slowly propagating zonal-mean zonal wind anomalies associated with stratospheric sudden warmings. They found that, in the seasonal timescale, the zonal wind anomalies appear first in the subtropical upper stratosphere, and propagate poleward and downward. Baldwin and Dunkerton (1999) found downward propagation of AO anomalies in low-pass-filtered data with a propagating time of 3 weeks from 10 hPa to the surface. They also noticed that even in the low-pass-filtered data, not all AO anomalies in the upper stratosphere propagate down. Only those with large amplitude and persistence have a clear signature through the troposphere. This phenomenon has also been observed in numerical model simulations. For instance, Yoden et al. (1999) investigated stratospheric sudden warming from a multiyear perpetual January integration of the Berlin Troposphere–Stratosphere–Mesosphere general circulation model. They made composites of many warming events in the model results according to the relative strength of planetary waves of zonal wavenumber 1 and 2 in the stratosphere. They found that the wavenumber-1 warmings were different from those of wavenumber 2 in many respects. In particular, signals of sudden warming descend to the upper troposphere only in the wavenumber-2 cases. Perlwitz and Graf (2001) used a single wave analysis (SWAN) approach on the strength of the 50-hPa polar vortex and 500-hPa height to study the connection of the stratosphere and troposphere. They found that the strength of polar vortex is a key factor for stratospheric downward effects on tropospheric circulations, particularly for zonal wavenumber 1. For instance, zonal wavenumber 2 at 500 hPa has a significant lag correlation with that at 50 hPa regardless of the strength of polar vortex. However, for zonal wavenumber 1, similar correlations exist only in the case of a strong polar vortex. To distinguish those different downward-propagating processes, Kodera and Kuroda (2000) separated vertical structures of AO to two types, depending on whether it is induced by a downward propagation of zonal-mean zonal wind anomalies from the stratosphere (type S) or it is generated within the troposphere (type T).

The purpose of this study is to determine the dynamical conditions necessary for an upper-stratospheric anomaly to propagate downward from the stratosphere to the troposphere in the synoptic- to planetary-scale waves. In addition, when this happens, we aim to determine the impacts on the tropospheric circulations. Instead of using empirical orthogonal function (EOF) or single wave analysis, we chose to use polar temperature anomalies as an indicative feature of downward propagation. In order to understand the dynamical processes, we use diagnostics of Eliassen–Palm (E–P) fluxes and theories of wave propagation to examine wave–mean flow interactions, because they can illustrate more clearly the physical mechanisms in the connection of stratosphere and troposphere (Andrews and McIntyre 1976; Edmon et al. 1980). In the next section we briefly describe the data and composite technique. Results are presented in section 3. In section 4 we focus on the propagating cases and investigate mechanisms involved. The impact of stratospheric anomalies on the tropospheric circulation is examined in section 5. A summary is given in section 6.

2. Data and methods

We first calculate temperature anomalies in the polar region (70°–90°N) at different pressure levels from 1000 to 10 hPa, using 22 yr (1978–99) of daily National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data (Kalnay et al. 1996). The anomalies are calculated by subtracting the 22-yr average of daily values from the temperature data so that the mean annual cycle is removed. We focus on the winter and spring seasons because stratospheric temperature anomalies in the summer are relatively small. There are very large interseasonal and interannual variabilities in the temperature variation. A typical example of 1987–88 winter–spring is given in Fig. 1, which shows the temperature anomaly as a function of time and pressure. The anomaly is normalized by its standard deviation at different levels to exclude the density effect. It seems that the timescales of temperature variation are different between the stratosphere and the troposphere. In the troposphere the temperature seems to be affected by synoptic waves with a timescale of 1–2 weeks, while in the stratosphere it is driven by large-scale planetary waves with a timescale of 1–2 months. Here we pay attention to three major episodes in the stratosphere. The first one is a major sudden warming event in early December. Its warm anomaly propagated from the upper stratosphere downward to the lower stratosphere and upper troposphere. It is followed by a cold anomaly also propagating downward, but the amplitude is smaller. In early March there is another warming event that hardly propagated downward at all.

We then select stratospheric warming or cooling episodes in the 22-yr period, and divide them into three categories. The time periods selected for the three categories are listed in Table 1. There were seven episodes each for propagating and nonpropagating warm anomalies, and six episodes for cold anomalies. To ensure statistical significance, only those with very large amplitudes in the upper stratosphere are selected. For instance, a propagating warming category is defined here as the case in which a temperature anomaly is greater than 2 standard deviation at 10 hPa and followed by a temperature anomaly greater than 1.5 standard deviation at 200 hPa. A nonpropagating warming category is defined as that where a temperature anomaly is also greater than 2 standard deviation at 10 hPa but followed by a temperature anomaly smaller than 1 standard deviation at 200 hPa. These definitions are somewhat arbitrary, but they separate propagating and nonpropagating cases very well. There are fewer cold anomaly events originating in the upper stratosphere and their magnitudes are smaller. Those cooling events usually follow a large warming event like the one in 1987–88, and they usually do not propagate down to the tropopause. Therefore, we only choose cooling events with a temperature anomaly greater than 1.5 standard deviation at 10 hPa, and do not separate them into propagating or nonpropagating categories.

Instead of studying individual episodes, we use a composite method to study common characteristics of each category. Noting that the lasting time varies in different episodes, without loss of generality, we select each period consisting of 40 days and arrange those periods according to their maximal amplitudes at 10 hPa. The composite diagnostics include such quantities as temperature, zonal-mean zonal wind, E–P flux, and other quantities in the Transferred Eularian Mean (TEM) framework (Edmon et al. 1980). We use E–P flux (F) and its divergence to examine the role of wave forcing on the mean flow. The components of E–P flux are

View Expanded

where ρ is air density, a is the radius of the earth, φ is latitude, R is the gas constant, f is the Coriolis parameter, H is a constant-scale height (7 km), N is the buoyancy frequency, u and υ are zonal and meridional wind, T is temperature, primes denote zonal deviation, and overbars denote zonal average. Here N is calculated from the temperature data, that is, N² = (g/T)(γ_d – γ), where g is gravity, γ_d is the dry adiabatic lapse rate, and γ is the lapse rate of temperature. In addition, we use the refractive index of planetary waves to examine the role of mean flow on waveguide modulations. The refractive index (n²) for quasigeostrophic stationary waves are defined as

View Expanded

where q is the quasigeostrophic potential vorticity, s is the zonal wavenumber, and other variables are the same as above (Chen and Robinson 1992). All of those quantities, are calculated from the daily NCEP–NCAR reanalysis data and averaged for each category.

3. The composites

Composites of propagating warm temperature anomalies (Fig. 2a) show a clear downward-propagating feature that seems to consist of two stages. The first stage is fast, taking just a few days to propagate downward from 10 to 50 hPa. The second stage is slower, with about 3 weeks of propagating time from 50 to 200 hPa. The associated zonal-mean zonal wind change in the polar region is more dramatic. In the upper stratosphere the strong westerlies reverse to easterlies in just a few days. The “critical line” (zero wind line) descends below 50 hPa (Fig. 2b). Because waves cannot propagate in easterly winds (Charney and Drazin 1961), the altitude of wave transport also descends with the critical line (Fig. 2c). The largest zonal wind deceleration takes place at the same time as the maximum wave forcing. The strong wave forcing is caused by the upward-propagating planetary waves that transport heat poleward and energy upward. Figure 2d shows the vertical component of the E–P flux, which is proportional to the poleward heat flux. Very large heat transport to the stratosphere occurred before the wind changed sign. After the polar wind reversed to easterly, the tropospheric wave could not propagate upward so that the heat transport was interrupted. However, the interruption did not last very long. Several days later, when polar westerlies recovered a second pulse of wave fluxes took place, which again forced the polar wind to change to easterly and the critical line to descend. Although the second pulse of wave fluxes is not as strong as the first pulse, it plays a deciding role for warm temperature propagating downward, as will be discussed later.

Figure 3a shows temperature anomalies in the composites of the nonpropagating category. Compared to the propagating case, there is no significant warming below 100 hPa although the warm anomaly at 10 hPa is about the same size as in the other composite. Even above 100 hPa there is no clear downward-propagating feature, rather, the warm anomaly occurred at almost the same time throughout the upper stratosphere. The zonal wind weakened but did not change direction (Fig. 3b). This is consistent with the weaker wave forcing in the upper stratosphere. As shown in Fig. 3c, the convergence of the E–P flux is only half as large as in the propagating case. It is also limited to higher altitudes and lasts for a shorter period. The initial vertical component of the E–P flux (Fig. 3d) has the same amplitude as that in the propagating case, but fades away earlier near the tropopause. There is no second pulse of wave fluxes following up, so that the polar westerlies recover their strength in a short period of time. Without continuous pump up of wave energy into the stratosphere, therefore, the upper warm anomaly could not extend to the lower altitudes.

The composites of the cold anomaly are shown in Figure 4. The selected cold anomaly events mostly resulted from the imbalance of radiative cooling and dynamical warming. They usually followed immediately stratospheric warming events (except for the spring of 1997). During the cooling periods the polar vortex was restored from the previous broken or weakened conditions. Zonal wind was strengthened (Fig. 4b) and planetary wave activity was quiet (Fig. 4d). Consistent with the wind acceleration, the E–P flux divergence at 10 hPa is positive (Fig. 4c). However, the wave forcing was probably generated from instability of the stratosphere rather than propagation from the troposphere. Since the cold anomaly does not propagate down to the tropopause level (Fig. 4a), the influence on tropospheric circulations is assumed to be small. In this paper, we will only discuss those propagating and nonpropagating warm anomalies.

4. The mechanisms for downward propagation

Comparing Figs. 2 and 3, we see that whether an upper-stratospheric warm anomaly propagates down depends on how planetary waves interact with the stratospheric zonal flow. Specifically, there are three elements that are important for downward propagation. First, the wave forcing must be strong enough to reverse the polar westerlies below certain levels such as 10 hPa. Second, after the wind reverses, the critical line must descend to lower altitudes. Because waves cannot propagate upward in easterly winds, the descending critical line forces wave energy to be deposited at lower altitudes. Consequently, the warm anomaly gradually propagates downward. But this mechanism of wave–mean flow interaction only works above the lowest altitude of the critical line. For further downward propagation, the third element that is required is continuous wave transport into the polar stratosphere. Because of its importance in the second stage of warm anomaly propagation, the source of continuous wave transport needs to be understood. It is well known that large-scale planetary waves originate in the lower troposphere. The largest wave amplitudes are found in the subpolar latitudes centered near 60°N. When those waves propagate upward they are usually refracted by zonal flow either poleward or equatorward. The equatorward-propagating waves tend to transport momentum poleward, thereby maintaining or strengthening both the polar jet and the polar vortex. On the other hand, poleward-propagating waves tend to transport momentum equatorward, which reduces polar westerly winds. In the stratospheric sudden warming events, the zonal-mean flow is preconditioned for large waves propagating upward and poleward. However, the condition of zonal-mean flow would certainly be changed during the warming events. The changed zonal-mean flow would consequently affect the way of wave propagation. For instance, a decrease in the speed of polar zonal-mean flow not only promotes poleward propagation of tropospheric planetary waves, but also generates medium-scale waves (Kodera and Chiba 1995).

In Figs. 5 and 6 we compare the quasigeostrophic refractive index for the wavenumber-1 stationary wave, as well as the E–P flux in the upper troposphere and lower stratosphere, in both the propagating and nonpropagating cases. As seen in the definition, the refractive index is basically determined by the structure of the zonal-mean zonal wind. Since waves propagate toward large positive values of the index and avoid negative values of the index, changes in zonal wind could affect the direction of wave propagation and the location of wave energy deposits. For easy comparison, we divide the 40-day composite time into two periods. The first half (day 1–20) corresponds approximately to the upper-stratospheric warming period, and the second half (day 21–40) is regarded as the postwarming period. In the first period the characteristics of the refractive index are quite similar in the propagating and nonpropagating cases, with maximum positive values in the polar upper troposphere. Midlatitude waves were mainly propagating through two waveguides, one toward subtropical tropopause and one toward subpolar stratosphere (Fig. 5). In the second half, however, the refractive index became very large in the 60°–70°N upper troposphere in the propagating case, so that the midlatitude waves were strongly refracted poleward (Fig. 6a). This feature is consistent with that of a low AO index (Limpasuvan and Hartmann 2000). It indicates a phase change of tropospheric annular mode. The increase in the value of the refractive index is mainly due to the large decrease in zonal-mean zonal wind associated with the reversal of the polar westerlies and the descent of the critical line. The weak westerly wind is a preferred condition for wave propagation (Dickinson 1968). There was little change of the refractive index in the nonpropagating case.

5. The impact on the troposphere

As shown in Fig. 2, the maximum warm temperature anomaly in the polar region takes about 2–3 weeks from 10 to 200 hPa in the propagating case. The corresponding spatial patterns of normalized geopotential height anomaly are shown in Fig. 7, in which consecutive 10-day averages are displayed in the selected 40-day period at 3 pressure levels from 10 to 200 hPa. The anomalies are approximately zonally symmetric with positive values in the polar region and negative values in the surrounding midlatitudes, which represents a typical AO signature. The maximum positive anomaly is about 2 standard deviation at 10 hPa in the day 21–30 composite. That pattern started at 10 hPa in day 1–10 and gradually propagated downward. Note that in day 21–30 the contour gradients at 200 hPa became very large in the North Atlantic section, which implies a large anomaly of geostrophic wind (easterly). Figure 8 shows the height anomaly in the nonpropagating case. The anomalies are generally zonally asymmetric, dominated by zonal wavenumber 1 in day 1–10 above 50 hPa. The amplitude of positive anomalies in the polar region is much smaller, with the maximum value of 1 standard deviation in the day 11–20 composite. The warm anomalies are less persistent even in the upper stratosphere, and the pattern does not propagate down below 50 hPa. In fact, the anomaly pattern at 200 hPa is almost out of phase with that at 10 hPa in the day 21–30 period.

The downward-propagating polar warm anomaly can affect the troposphere not only by changing its thermal state, but also by changing its dynamics and energetics. For instance, in the postwarming period the upper-tropospheric waves are refracted poleward, indicating equatorward momentum transport that could affect the upper-tropospheric zonal wind. Here we examine how subtropical jet streams respond to those propagating and nonpropagating polar warm anomalies. In Fig. 9 we compare composites of changes in wind speed at 200 hPa after upper-stratospheric warmings, for the propagating and nonpropagating cases, respectively. In the propagating case wind speed is generally reduced in the mid–high latitudes and enhanced in the low latitudes. The largest change is found in the North Atlantic region, where wind speed increases on the south side of the jet stream and decreases on the north side of the jet stream. As a result, the axis of the Atlantic jet stream shifts southward by ∼5° of latitude, and the alignment of the axis becomes more zonal, in the postwarming period. We should point out that this feature is similar to the negative phase of North Atlantic Oscillation (NAO). Similar changes are found in the Pacific jet stream near the exit of the jet, but the magnitude of change is smaller. On the other hand, in the nonpropagating case the pattern of wind change is very different. The strength of the jet stream somewhat weakens over North America, while the position of the Atlantic jet stream remains almost unchanged. The largest decrease in wind speed is in the North Pacific, and the Pacific jet stream moves somewhat southward near the exit. It seems that the Atlantic jet stream is more sensitive to downward-propagating stratospheric warm anomalies than the Pacific jet stream.

6. Summary

From the Northern Hemisphere polar temperature anomaly data we find that the stratosphere occasionally allows a warm anomaly to propagate from the upper stratosphere to the troposphere, and occasionally prohibits downward propagation. The condition of the polar atmosphere is primarily determined by the strength of the wave activity and the structure of zonal-mean flow. If the initial wave forcing is strong and persistent enough to reverse the polar westerlies, then the warm anomalies can descend along with the critical line because waves cannot propagate into easterly winds. However, this type of mechanism usually only works for a few days and to a limited depth (depending on how deep the critical line descends) due to the exhaustion of wave energy. Additional downward propagation of the warm anomalies requires continuous energy supply by tropospheric waves, which can be induced, in principle, by positive feedbacks due to changes in zonal-mean flow in the mid- and high latitudes. This feedback mechanism has also been addressed in previous case studies (e.g., Kodera et al. 1991; Kodera and Chiba 1995).

Apparently, the downward propagation of upper-stratospheric warm anomalies is related to the upward propagation of planetary waves from the troposphere. The ultimate force or energy source is not in the stratosphere but in the troposphere. In other words, instead of the stratosphere forcing the troposphere, it is essentially the troposphere that drives the stratosphere through wave transports of heat and momentum. The convergence of wave fluxes, due to planetary wave breaking, critical line wave trapping, or other processes, generates an anomaly in the mean state, such as a midwinter stratospheric warming. The anomaly by wave forcing usually occurs first in the upper stratosphere because planetary waves can travel freely in the troposphere and lower stratosphere before breaking. If the altitude of wave forcing descends with time due to changes in the mean states of zonal wind, then it looks like downward propagation of an anomaly from above. If the change in the mean state is not large enough to affect wave propagation, we would not see a downward-propagating anomaly. Therefore, the role of the stratosphere is not like a thermal pump but like a regulator. It controls the path of planetary wave propagation and the location of the wave energy deposit. This mechanism dominates not only in the synoptic to planetary timescales but also at seasonal timescales.

As shown in the analysis of the 90-day low-pass-filtered data by Baldwin and Dunkerton (1999), the downward propagation of the negative AO phase is exclusively related to the stratospheric warming events that are caused by tropospheric waves. Here, in the propagating cases, the patterns of geopotential height anomaly also indicate downward propagation of AO mode. Unlike the AO anomaly in the low-pass-filtered data, polar warm anomalies could not propagate downward to the lower troposphere or the surface in the synoptic to planetary timescales, but the impact of downward-propagating warm anomalies could be felt in lower-tropospheric weather systems through the link of upper-tropospheric jet streams. Changes in the strength and position of subtropical jet streams are often indicative of changes in weather patterns. We also point out that the negative phase of NAO is characterized by a southward shift of the North Atlantic jet stream, and the variation of AO is in phase with NAO in the winter and spring seasons. Thus, the change of the North American and Atlantic jet streams in the poststratospheric warming period is consistent with the downward propagation of a negative stratospheric AO mode.

Although stratospheric cold anomalies usually do not propagate downward to the troposphere, positive AO signatures occasionally do, as shown by Baldwin and Dunkerton (1999). Since the AO signature is calculated from geopotential height data, we have examined polar anomalies of geopotential height, which also show occasional downward propagation of negative anomalies from the upper stratosphere to troposphere and the surface (figures are not shown here). The mechanism for such propagation is more implicit and difficult to explain in terms of wave–mean flow interactions discussed in section 4.

Generally, our results provide a legitimate explanation of downward propagation of the polar stratospheric warm anomalies. However, these results are based on a composite analysis of a relatively small number (seven) of samples. Only those very large anomalies in the upper stratosphere are chosen, and the criteria to define propagating and nonpropagating categories are somewhat arbitrary. Further investigations are needed by selecting more samples from longer datasets, and by carrying out sensitivity studies using different criteria. The downward propagation may be divided into more than two categories. Furthermore, the problem may also be studied in various longitude sections, instead of in zonal average. In that case, the generalized E–P flux (Plumb 1985) should be a very useful diagnostic tool, which should also be more appropriate to study the impact of stratospheric anomalies on regional subtropical jet streams.