Seasonal differences in extratropical potential vorticity variability at tropopause levels

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109, D17102, doi:10.1029/2004JD004639, 2004

Seasonal differences in extratropical potential vorticity variability at tropopause levels M. A. Liniger Climate Services, Federal Office of Meteorology and Climatology, Zu¨rich, Switzerland

H. C. Davies Institute for Atmospheric and Climate Science, Eidgenossische Technische Hochschule, Zu¨rich, Switzerland Received 13 February 2004; revised 4 June 2004; accepted 21 June 2004; published 10 September 2004.

[1] The distribution of extratropical potential vorticity (PV) on isentropic surfaces that

transect the tropopause is a key feature of planetary-scale teleconnection patterns, synoptic-scale weather systems, and mesoscale stratosphere-troposphere exchange. Here a Northern Hemisphere January and July climatology is presented for the mean and variability patterns of the PV using the so-called European Centre for Medium-Range Weather Forecasts reanalysis-15 (ERA-15) data set. It is derived taking into account the strong seasonal cycle of the tropopause height and the sharp quasi-latitudinal gradient of PV on the isentropic surfaces together with the accompanying bimodality in the scale and amplitude of positive and negative anomalies. The bimodality is both emphasized and circumvented by comparing conventional standard deviation measures of the variability with those of separate depictions of the probability density function structures of positive and negative anomalies. It is shown that in winter the zonal heterogeneity is pronounced and the variability (e.g., storm track) patterns exhibit a rich spatial structure with marked differences between the Pacific and Atlantic. In contrast, in summer the heterogeneity on the lower stratospheric portion of the isentropic surfaces is much weaker, but there remain regions of high variability over oceanic regions on the tropospheric portion of the surfaces. The results relate directly to the structure and dynamics of storm INDEX TERMS: 3319 Meteorology and tracks and their spatial and seasonal variation. Atmospheric Dynamics: General circulation; 3309 Meteorology and Atmospheric Dynamics: Climatology (1620); 3362 Meteorology and Atmospheric Dynamics: Stratosphere/troposphere interactions; KEYWORDS: reanalysis, isentropes, storm tracks, Rossby waves, PDF, climatology Citation: Liniger, M. A., and H. C. Davies (2004), Seasonal differences in extratropical potential vorticity variability at tropopause levels, J. Geophys. Res., 109, D17102, doi:10.1029/2004JD004639.

1. Introduction [2] In the extratropics, synoptic-scale variability is high throughout the troposphere. At lower and midtropospheric levels the variability is both well known and widely documented and led to the identification of confined regions of high cyclone frequency, referred to as ‘‘storm tracks’’ across the northern Atlantic and Pacific [Ko¨ppen, 1881; Petterssen, 1956]. An associated structure is also found in variance fields of geopotential height [see, e.g., Blackmon, 1976]. This variability has been interpreted in terms of finite amplitude baroclinic waves whose structure and evolution vary with geographical location [Wallace et al., 1988]. [3] At tropopause levels the corresponding anomalies are associated with undulations of the tropopause, Rossby waves, and Rossby wave breaking. From a potential vorticity (PV) – potential temperature (TH) perspective, there is a strong link of the tropopause level variability with the Copyright 2004 by the American Geophysical Union. 0148-0227/04/2004JD004639

underlying synoptic surface structures [e.g., Hoskins et al., 1985; Davis and Emanuel, 1991; Morgan and NielsenGammon, 1998]. [4] Tropopause undulations can take the form of meridionally elongated structures, also referred to as streamers [Appenzeller and Davies, 1992]. These stratospheric intrusions are associated with strong positive PV anomalies and are often related to nascent surface cyclones and extreme weather events [Massacand et al., 1998]. Furthermore, tropopause folds and subsequent PV cutoffs contribute to cross-tropopause transport [Vaughan et al., 2001; Liniger and Davies, 2003; Sprenger et al., 2003]. The dynamical counterparts, anticyclonic structures of low PV values, also constitute a significant part of the flow structure in the extratropics. In particular, quasi-stationary ‘‘blocking’’ anomalies are associated with positive tropopause height anomalies [Pelly and Hoskins, 2003; Schwierz et al., 2004]. The strong correlation of tropopause-level PV to quantities such as ozone [e.g., Danielsen, 1968; Danielsen et al., 1987; Vaughan and Price, 1991; Beekmann et al., 1994; Rao et al., 2003] and upper tropospheric humidity [Appenzeller et

D17102

1 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

al., 1996a; Liniger and Davies, 2003] underlines the centrality of the PV pattern at upper tropospheric and lower stratospheric levels to atmospheric dynamics. [5] The strong linkage between the upper level PV distribution and the underlying weather systems and its strong correlation with other quasi-conserved tracers makes PV a suitable quantity to study the statistical properties of synoptic variability and the associated transport processes. Most studies of PV variability in upper levels have focused on winter months: The low- and high-frequency contributions of the dynamical and residual transient forcing were located by Brunet et al. [1995], and the PV distribution was linked to (conservative) Rossby waves and estimates derived of the diabatic, divergent, and rotational PV budgets [Edouard et al., 1997]. This technique was then applied to 39 winters of National Centers for Environmental Prediction reanalysis data with an additional focus on low-frequency variability [Derome et al., 2001]. Likewise, well-known indices such as the North Atlantic Oscillation and PacificNorth American pattern are evident in the monthly mean isentropic PV distribution and are related closely to Atlantic and European precipitation patterns [Massacand, 1999; Massacand and Davies, 2001a, 2001b]. [6] However, it remains to determine the space-time climatology of the upper level PV variability. The compilation of statistics of local PV values is hampered by technical difficulties because of the bimodal structure found in their temporal probability density functions (PDF) [Swanson, 2001]. This bimodality exhibits an interannual and geographical variation and affects the physical interpretation of PV statistics. A related complication is the strong interseasonal variation of the tropopause and isentropic height [e.g., Appenzeller et al., 1996b]. [7] In this study, a climatology of local high-frequency isentropic PV variability is derived that treats positive and negative anomalies separately. The focus is not on a specific spatial or isentropic height but rather on the geographic location and the vicinity to the tropopause. This is achieved by selecting appropriate isentropes and allowing the corresponding PV anomaly thresholds to have a seasonal dependence. This technique delivers a unique data set to address various aspects of the local interseasonal variability of near-tropopause anomalies. [8] First, the data set used for this study is described (section 2), and the results from a conventional approach of mean and standard deviation are presented (section 3). Then, the technique for a climatology of positive and negative upper level PV anomalies is set out (section 4), and the interseasonal variation of the geographical anomaly distribution and its variability are examined (section 5). Thereafter the statistical PV distribution (section 6) and the special case of strong positive anomalies are investigated in more detail (section 7). Finally, an interpretation of the results is given in section 8.

D17102

1999]. Model level data with a 6-hour time resolution were interpolated to a 1° grid and then used to calculate the PV and its interpolation onto isentropes. Instead of considering a 3-month season of distributions, 2 representative months are used for Northern Hemispheric winter and summer: January and July. This restriction results in richly structured patterns. [10] An inspection of TH and PV in the zonal monthly mean reveals a strong interseasonal variability of both the vertical position of the isentropes and the tropopause (Figure 1). Hence an individual isentrope’s (e.g., TH = 320 K) intersection with the dynamical tropopause (defined as 2 potential vorticity unit (PVU) surface, following Holton et al. [1995]) undergoes a strong meridional shift from 40°N in January to 70°N in July. Further, more interannual variability and trends have been observed in tropopause height over the ERA-15 period [Kiladis et al., 2001; Santer et al., 2003]. To partially account for these seasonal and interannual variabilities, a specific isentrope is selected for each individual month that intersects the tropopause at 45°N in the zonal monthly mean. In effect, this enables us to focus on extratropical tropopause variability by removing a portion of the interseasonal, interannual, and decadal variability. The resulting values selected for TH vary from 308 K in winter to 339 K in summer (Figure 2). Note that the interannual variability of the selected THs is
5 K for all months. Inspection of Figure 1 indicates that the selected isentropes cross the upper level region of highest baroclinicity (i.e., steep isentropes) that is located between 35° and 40°N at around 350 hPa in January and between 40° and 45°N at above 300 hPa in July. The remainder of the study is based on these selected isentropes. [11] To illustrate the reduction of interannual and decadal variability, two time series of PV values on the selected isentropes are examined (Figure 3). Two points are chosen in vicinity to the climatological tropopause and respectively strong and weak high-frequency variability (see sections 3 and 5). The first point is located in a region in the center of the Atlantic storm track (Figure 3a). The strong highfrequency variability is not reflected in an interseasonal or interannual variability. Further, there is no indication for long-term fluctuations or a trend. The second point is located over eastern Asia (Figure 3b). This region is characterized by a trough with relatively weak high- frequency variability. Here there is a clear seasonal variation with high PV values in January and low values in July. However, the interannual variability is even smaller than over the North Atlantic. [12] The investigated period of 15 years is relatively short from a climatological perspective. However, the high temporal resolution and the absence of long-term fluctuations and trends give an estimate about the statistical robustness of the following results.

3. Conventional Approach 2. Data Set and Procedure for Selection of Isentropes [9] This study makes use of the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA)-15 data that cover the period 1979– 1993 with a spectral resolution of T106 and 31 levels [Gibson et al.,

[13] In a first step, mean and standard deviation of the local PV values are analyzed on the selected isentropes. The climatological PV mean (PV) for January resembles and relates to the typical stationary wave pattern (Figure 4a). Strong troughs and PV gradients are found over eastern North America and are somewhat less pronounced over

2 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

D17102

Figure 1. Latitude-height cross section of Northern Hemispheric zonal mean of PV (shaded area) (in PVU) and TH (dashed curves from 280 to 400 K with a spacing of 10 K) for (a) January and (b) July from 1979 to 1993. The solid curve denotes the isentropes cutting the tropopause at 45°N. Pressure levels are indicated by solid horizontal lines (spacing of 100 hPa).

eastern Asia. They are located immediately upstream of the Atlantic and Pacific storm tracks. The ridges are positioned downstream of the troughs,
90° farther east over the Pacific and
60° over the Atlantic. Meridional gradients are weak on the tropospheric side of the ridges over California and Spain at the end of the storm tracks. The Eurasian continent is characterized generally by weak gradients, but some isentropes cut the boundary layer in the Himalayan region. For the Pacific storm track a region of very low meridional gradients is found between the trough and the ridge over Alaska. In contrast, the gradients along the Atlantic storm track weaken downstream of the ridge at the end of the storm track. The overall structure of the January mean is more pronounced but similar to the winter PV values shown by other studies [e.g., Derome et al., 2001; Massacand and Davies, 2001b]. [14] The highest values in the standard deviation are located between the 2 and 4 PVU contours of PV; that is, the maxima are not located at the climatological 2 PVU tropopause but are meridionally shifted into the stratosphere. Zonally, the maxima are collocated with the ridges at the end of the corresponding storm tracks. The PacificAmerican variability maximum is weaker and narrower, and there are indications that it consists of two relative maxima, one collocated with the trough over the western North American and a weaker one over the North Pacific upstream of the trough. At the end of the storm tracks the European maximum is much stronger and broader. Within the climatological troposphere the variability is rather small, and this is associated with the narrow range of PV values within the troposphere (see section 4). Overall, the meridional distribution of the standard deviation correlates well with PV with the maximum between 2 and 4 PVU. This relationship was noted and utilized for a statistical description of the PV variability as a function of the PV value given by Swanson [2001]. However, the zonal asymmetry

of the standard deviation illustrates the limited applicability of this approach. [15] In July the PV pattern exhibits a less pronounced wave structure (Figure 4b). In particular, the east Asian trough and the ridges over the Mid-Atlantic and eastern Pacific are no longer clearly identifiable. Nevertheless, there are some notable deviations from the zonal mean: Over the Pacific and North Atlantic, there is equatorward extension into the subtropics, and a small ridge can be identified over eastern Asia. [16] The July standard deviation is significantly higher than in January (note the different scale intervals). Again, there is a good correlation with PV, exhibiting a weak zonal variability. However, the maximum is shifted to values of

Figure 2. TH (in K) of isentrope that cuts the tropopause in monthly zonal mean at 45°N in dependency of month. Each line represents a year from 1979 to 1993, with earlier years shaded lighter.

3 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

D17102

Figure 3. Time series of PV values at the selected isentropes at (a) 40°W, 47°N and (b) 130°E, 40°N. Shown are 6-hour values (light shaded curve), a running average with a window size of 3 months (dark shaded curve), and a running average with a window size of 1 year (dashed curve). PV between 4 and 6 PVU, clearly higher values than in January. [17] The maxima of variability is located within the climatological stratosphere, both in January and (even more pronounced) in July. The lack of variability within the climatological troposphere stands in contrast to a high variability of synoptic weather systems in the storm track regions and illustrates the complications arising from the

application of conventional statistical measures to PV values.

4. Identification Technique for PV Anomalies [18] Here an alternative approach is set out to describe the PV variability. The technique is designed to identify positive anomaly (PA) and negative anomaly (NA) frequencies

Figure 4. Monthly averaged PV (contour lines of 0.2, 0.8, 2, 4, 6, and 8 PVU) from 1979 to 1993 for (a) January and (b) July and standard deviation (shaded area) (note the different contour spacing between Figures 4a and 4b). 4 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

Figure 5. CDFs of area weighted PV values from the entire Northern Hemisphere from 1979 to 1993 on the selected isentropes for all time steps for January (shaded curve) and July (solid curve). See text for further explanations.

D17102

that the area enclosed by the ‘‘2 PVU + dPV’’ contour deviates from the area enclosed by the 2-PVU contour by d = 10% of the hemispheric area. This procedure is illustrated in Figure 5 by the horizontal lines at a distance d below (above) the ratio of the hemispheric area enclosed by 2 PVU for the PA (NA) threshold. [24] The conditions and resulting threshold values for NA and PA, both for January and July, are summarized in Table 1. Note that all PV values are considered that differ from the climatological mean PV by more than the threshold dPV. Thus, in a region with a climatological PV value of 6 PVU an instantaneous PV value of 3 PVU is counted as an NA. [25] The threshold values for the NA (PA) are now applied to every grid point of the instantaneous 6-hour isentropic PV distributions within the climatological stratosphere (troposphere). The local frequency of occurrence is then constructed by counting all grid points fulfilling the outlined conditions for every January (July) from 1979 to 1993 at the same location.

5. Positive and Negative Anomalies separately with a seasonal varying threshold to ease interseasonal comparisons. [19] A PV value is identified as an anomaly if it deviates from the climatological mean PV by a certain threshold PV value (dPV). The local PV anomaly frequency is then calculated for the positive and negative PV anomalies separately. Positive (negative) anomaly frequencies are only considered within the climatological troposphere (stratosphere). [20] The threshold dPV is determined on the basis of the seasonal characteristics of the PV values. For example, there is a stronger vertical stratification and larger vertical PV gradient within the polar stratosphere in the zonal mean in July than in January (Figure 1). This results in a much broader tail toward high PV values in the hemispheric PDF of PV values. [21] To account for such features, the cumulative distribution functions (CDFs) of PV for the different months are first investigated in detail to establish objectively defined threshold values. The PDFs of area weighted PV values are compiled from the entire Northern Hemisphere over all years and time steps and are then integrated to obtain the CDFs for each selected isentrope. [22] The resulting CDFs for January and July are shown in Figure 5. This depiction relates PV values to the geographical area covered by them. The weak gradients of PV within the troposphere between 0 and 1 PVU correspond to the steep slopes of the CDFs for both January and July. For PV values higher than 1 PVU the CDF is less steep, representing the stronger PV gradients. For PV values above
4 PVU the differing PV gradients in January and July (see Figure 1) can be identified by the weaker slope of the CDF for July. In July, PV values above 8 PVU contribute more than 5% to the hemispheric area but only a negligible fraction in January. The distribution for January is consistent with both the instantaneous global CDF [Ambaum, 1997] and the local PDF of PV values in midlatitudes in a winter climatology [Swanson, 2001]. [23] The values for dPV are now chosen to insure an equal weighting of both NA and PA in regard to the area covered by the anomalies. The dPVs are determined such

[26] We consider the anomaly distributions for January and July. In January the NA frequency pattern is characterized by maxima between 2 and 4 PVU (Figure 6a). High values are zonally confined to the climatological ridges with a broad maximum over the mid – North Atlantic and a weaker one over western North America. The Atlantic maximum broadens toward Europe and separates from the tropopause slightly poleward over eastern Europe and Siberia. The American maximum consists of two relative maxima, collocated with the ones observed for the standard deviation (see Figure 4a). The minimum over the pole is extended equatorward toward eastern Asia and, much less pronounced, toward eastern North America. The overall structure is very similar to that of the stratospheric part of the standard deviation (see Figure 4a). The narrow band with very low frequency directly poleward of the climatological tropopause is due to the sparseness of PV values of PV + dPVPA PV > 2 PVU

January July

Threshold Values NA = 2.1 PVU dPVJan NA = 2.8 PVU dPVJul

PA dPVJan = 1.4 PVU PA dPVJul = 1.2 PVU

a NA, negative stratospheric anomalies; PA, positive tropospheric anomalies; PV, potential vorticity; PVU, potential vorticity unit.

5 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

D17102

Figure 6. Frequency of positive and negative anomalies (shaded area) and monthly averaged PV as isolines (0.2, 0.8, 2, 4, 6, and 8 PVU) for (a) January and (b) July from 1979 to 1993. Positive (negative) anomalies are shown only within the climatological troposphere (stratosphere).

downstream, collocated with the weaker meridional gradients in PV. [28] In July the regions of high NA frequency shift toward higher PV values to around 4 or 5 PVU, depending on the longitude (Figure 6b). Further, the anomaly maxima are found more upstream than in January; both maxima are now collocated with the oceans. Thus the Atlantic maximum is shifted only slightly, whereas the American one is located 60° eastward over the central Pacific. Again, the Atlantic maximum separates from the tropopause northward over eastern Europe. [29] The PA frequency in July exhibits pronounced southwestward tongues over the central North Atlantic and eastern North Pacific. These tongues extend from the end of the storm tracks into the subtropical domain around 30°N. A weaker signal of this feature can be discerned in the PV structure.

[30] In contrast to the standard deviation the anomaly frequency has a much more detailed structure within the climatological troposphere. Further, the values are within a very similar range for January and July, as desired by the choice of threshold values depending on the PV distribution of the corresponding month.

6. PDF Structures [31] The approach adopted also allows a detailed inspection of local PDFs along selected meridians. Two regions representing two different dynamical regimes are chosen: (1) the meridian 40°W that bisects the Atlantic storm track (Figure 7) and (2) the meridian across the trough of weak variability over China at 130°E (Figure 8). The meridians contain the two points used to illustrate the homogeneity of the time series of PV values (see Figure 3).

Figure 7. Local PV value histograms (vertical axis) along 40°W for (a) January and (b) July (contour values of 0.16, 0.5, 1, 3, and 6% and bin size of 0.2 PVU) for all latitudes (horizontal axis). The shaded line denotes the climatological mean PV, and solid lines correspond to the threshold PV + dPV used for positive (negative) anomalies within the climatological troposphere (stratosphere). The dashed line denotes the same for strong positive anomalies discussed in section 7. 6 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

D17102

Figure 8. Same as Figure 7 but along 130°E for January and July. [32] In general, the tropospheric part of the PDF is very narrow (between 0 and 2 PVU) and has a broad tail toward positive values. Within the stratosphere the distribution is wider (from 2 up to 10 PVU) and exhibits a confined tail toward low PV values. For both regions the higher values are evident in July. [33] The local mean PV values indicate a strong difference between the meridional PV gradient over the Atlantic and over China. In January the PV gradient is weaker over the Atlantic, whereas the Asian gradient is weaker in July. [34] The meridional gradients of the local mean values are linked to both higher PV values in the stratosphere (values are similar over the Atlantic and China) and the overlap of tropospheric and stratospheric values. This overlap is much stronger over the Atlantic because of the Atlantic storm track. In the dynamically calmer region over China in winter the overlap is very small (around 40°N). In summer, however, the PDFs over Asia between 50° and 70°N are extremely wide, exhibiting no clear bimodal structure. [35] To indicate what kind of PV values contribute most to the PV anomaly frequencies, the threshold for the identification of NA and PA is denoted in Figures 7 and 8. The PA frequency is dominated by stratospheric PV values, both over the Atlantic and Asia. Hence the PV frequency is very sensitive to stratospheric anomalies that can be associated with high PV values. [36] For the NA, however, tropospheric PV values contribute significantly only over the Atlantic in the region from 50° to 75°N in winter and from 50° to 65°N in summer. Over Asia the NA frequency is less influenced by tropospheric PV values because of the weaker bimodality of the PDF. This corresponds to the wide band of low NA frequencies found in this region. [37] From the PDFs over the North Atlantic it can be further inferred that the expansion of the PA toward the subtropics is mainly caused by stratospheric PV values. Analogously, the local NA maximum over Asia observed around 60° north of the small ridge (see Figure 6b) can be attributed to values with a significant portion of tropospheric PV values (Figure 8b). [38] Evidently, the PDFs are subject to a strong variability with regard to location and season. Note that the PV and the standard deviation do not capture all the features described in sections 5 and 6. Their values are strongly affected by specific features of the PV distribution, in particular the different range of values above and below the tropopause. Therefore the skewness Cs and the variation (also referred to

as ‘‘normalized standard deviation’’) Cv are introduced here to further illuminate the geographical characteristics of the PDF structures (with s denoting the standard deviation and N denoting the sample size): 3 N  1 X PV  PV Cs ¼ N t¼1 s Cv ¼

s : PV

The skewness Cs describes the asymmetry of the PDF. For a normal variable the skewness is zero, and distributions with a longer upper tail are said to be positively (right) skewed. Accordingly, it is strongly positive within the troposphere (Figure 9). The highest values are associated with PV below 0.8 PVU and a low (but not zero) frequency of positive anomalies. The region of high skewness exhibits a surprisingly strong zonal symmetry. Along the climatological tropopause the skewness decreases to zero since the tropospheric and stratospheric values build a bimodal (and symmetric) PDF (see Figure 7a). North of it, in particular north of the troughs, the skewness exhibits moderate negative values. An inspection of local PV PDFs reveals a lack of strong positive values and more tropospheric values (e.g., north of 50°N in Figure 8a and less pronounced in Figure 7a). [39] For July the asymmetry between the climatological troposphere and stratosphere is weaker but still pronounced. A higher number of strong positive values and fewer tropospheric values can be found in the local PDFs (as, e.g., north of 70°N in Figure 7b) resulting in less anomalous skewness within the stratosphere. [40] The variation (Cv) shows the variability relative to PV (Figure 10). The meridional shift of the maxima in the standard deviation away from the tropopause toward higher PV mean values is compensated for by the distribution of the variation: The variability within the troposphere is weighted much stronger. In winter the maxima are located south of the ridges along the 0.8-PVU PV mean isoline. The region at the end of the storm tracks with strong Rossby wave breaking exhibits the highest values. [41] The July maxima are found more upstream, with an extension toward the subtropics over both the Atlantic and Pacific. The structure is similar to the PA distribution, but the calculation is not very robust in the region of very low PV values, i.e., toward the subtropics, and thus causes high spatial variability in this region. The overall difference in

7 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

D17102

Figure 9. Same as Figure 4 but for skewness (shaded area). magnitude of the standard deviation disappears between the winter and summer. Therefore the larger standard deviation in July than in January (see Figure 4) is caused not by a stronger relative variability but by a shift toward higher PV values.

7. Strong Positive Anomalies [42] Precursors of surface cyclogenesis can often be identified by anomalous high PV values of the order of 4 – 6 PVU in midlatitudes. They are identified here as strong positive anomalies (SPA) within the climatological stratosphere. Here we employ a special anomaly selection threshold to directly identify SPA. [43] It is unclear to what extent these anomalies contribute to the standard deviation. Moreover, SPAs are to be

expected mainly within the climatological stratosphere; the analysis need not be constrained to the climatological troposphere as it was for the PA frequency. [44] The previous technique (see section 4) used to establish the threshold value for PA within the troposphere is now used to define SPA within the stratosphere (illustrated in Figure 11). As reference a PV value of 4.8 PVU is chosen (see the value of 2 PVU in section 4). The threshold PV value dPVSPA is selected to identify anomalies such that the area enclosed by the 4.8 PVU + dPV contour deviates by d = 10% of the hemispheric area, from the area enclosed by the 4.8-PVU contour. This procedure results in threshold values of dPVSPA Jan = 1.8 PVU for January and dPVSPA Jul = 2.8 PVU for July. [45] The resulting structures (Figure 12) are very similar to the standard deviation and NA distribution both for

Figure 10. Same as Figure 4 but for normalized standard deviation (shaded area). 8 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

D17102

the amplitude of the frequencies is in the same order of magnitude.

8. Synthesis and Further Comments

Figure 11. Same as Figure 5 but with labels for strong positive anomalies. See text for further explanations.

January and July as shown in Figures 4 and 6, respectively. In particular, the Pacific maxima consist again of two relative maxima, one over western North America and a weaker one over the central North Pacific. Both in January and July the differences between the Atlantic and Pacific maximum are similar for SPA and for NA. However, the detailed structure of the storm tracks reveals some differences: (1) The SPA Atlantic maximum extends farther eastward into the Eurasian continent than the NA maximum and (2) during July the SPA maxima are more confined. [46] The magnitudes, locations, and patterns of the stratospheric NA and SPA frequency are found to be very similar. Thus the storm tracks can also be seen as regions of high frequency of anticyclonic anomalies. With the threshold values being different for positive and negative anomalies,

[47] The combination of conventional statistical measures, inspection of climatological and local PDFs of PV, and a subjective positive and negative anomaly identification technique was employed to study PV variability near the extratropical tropopause. The PV distribution is studied on isentropic surfaces that are selected for each January and July separately. This enables direct consideration of interseasonal and interannual variability. [48] In January the region of lowest variability is located upstream of the Pacific storm track and thereby separates the track from that over Asia. In contrast, the Atlantic storm track is linked upstream to the Pacific storm track. [49] Indeed, the Pacific storm track differs significantly from the Atlantic one. It is less confined and can be subdivided into three regions. Region 1 is upstream, over eastern Asia, with extremely low variability and strong zonally aligned PV gradients. Region 2 is downstream, over the central Pacific, and the PV gradients are reduced on the stratospheric side. Close to the tropopause, a secondary maximum can be identified in all fields describing the stratospheric variability. Region 3 is the climatological ridge farther downstream over western North America and is associated with maximal variability within both the climatological stratosphere and the troposphere. The structure observed for the Pacific storm track can be related to a separation between several regions of Rossby wave development and breaking. An indication for such a separation is also found by Hoskins and Hodges [2002], who determine the genesis region of cyclones ending over western North America. The genesis region they find is collocated with the

Figure 12. Frequency of strong positive anomalies (shaded area) and monthly averaged PV as isolines (0.2, 0.8, 2, 4, 6, and 8 PVU) for (a) January and (b) July from 1979 to 1993. 9 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

region identified as secondary maximum of variability over the North Pacific in the present study. [50] For the Atlantic storm track the spatial subdivision is not so apparent. The trough over eastern America is more pronounced and exhibits stronger meridional PV gradients. Downstream over the central North Atlantic the ridge is less strong than its Pacific counterpart. The maximum in variability extends from the American east coast across the Atlantic and into eastern Europe and even Siberia. It is significantly broader within the climatological stratosphere than its Pacific counterpart. On the tropospheric side the maximum (of positive variability) is located slightly downstream toward Europe and the Mediterranean. [51] The July hemisphere features shorter storm tracks with a weaker meridional and zonal variability within the stratosphere. The minimum of variability along the tropopause is not located over eastern Asia but over central Siberia. Also, the Pacific maximum is shifted significantly from western North America to the central North Pacific, and both maxima are now collocated over the oceans. The stronger baroclinicity due to the land-sea contrast in winter is said to have a particularly strong impact on winter atmospheric variability [Hoskins and Valdes, 1990]. However, the anomaly frequency distribution exhibits a stronger asymmetry between the land and ocean domains in July at upper levels. The weaker meridional temperature gradient at the surface and the corresponding weaker thermal wind at upper levels could result in more localized wave activity. The westward shift of high variability compared to winter and the related localization of strongest wave activity over the oceans are also reflected in the contour length stretching rates in these sectors [Scott and Cammas, 2002]. This indicates that the observed anomaly frequency maxima are directly linked to increased isentropic mixing across the tropopause. [52] Within the climatological troposphere, regions of high PA frequency and weaker PV mean and standard deviation extend into the subtropics. As indicated in Figure 1b, the selected isentropes are located above 500 hPa in the zonal mean. Indeed, an inspection of individual events reveals that these PV anomalies are not located in the lower parts of the troposphere but at upper levels around 200 hPa and are thus not related to diabatic effects but rather to possible subtropical Rossby wave breaking. [53] The origin and path of the anomalies cannot be captured by the method employed in this study. For the Pacific a strong monsoon anticyclone could contribute to the advection of anomalies from the western Pacific toward the subtropical central Pacific. Over the Atlantic, there is no similar anticyclone documented in summer, but the PV mean and the PA frequency indicate an analogous structure of Rossby wave breaking. [54] Subtropical Rossby wave breaking over the oceans was also identified along the tropopause at 350 K by Postel and Hitchman [1999]. They find that the maxima are located slightly more to the west in the regions of the subtropical Atlantic and Pacific during summer, and their Pacific maximum is twice as strong as the Atlantic one. This difference could be related to their method: Meridionally aligned structures in the form of streamers would not be captured.

D17102

[55] A small ridge in the PV mean north of China is accompanied by a weak local maximum in NA, SPA, and variance. Therefore the overlap of tropospheric and stratospheric PV values in this region could be a signature of enhanced Rossby wave breaking. Negative PV cutoffs have been observed in this region and were attributed to the Asian summer monsoon [Popovic and Plumb, 2001]. However, the center of monsoon activity is located significantly more to the south. The observed feature could also be a signature of variability on higher isentropes. [56] The local PDFs of PV exhibit a wide heterogeneity. The strong interseasonal variability within the stratosphere and the fundamental difference between the tropospheric and stratospheric PDF structure complicate the derivation and use of physically meaningful quantities for statistics. For example, the bimodal structure in the storm track regions is not symmetric. This results in negative anomalies dominating positive anomalies. Nevertheless, the stratospheric variability structure is represented very well by the standard deviation. On the tropospheric side, however, features such as the subtropical extension over the oceans are not clearly discernible. It is shown that the local PDFs are strongly non-Gaussian, in particular within the lowermost stratosphere. Thus the interpretation of local fields of PV standard deviation, skewness, or variation has to be carried out very carefully by also considering the corresponding PDF structure directly or employing additional measures. [57] The results also have implications for the determination of a background PV mean state, as it is necessary for PV inversion [Kleinschmidt, 1950; Hoskins et al., 1985]. A strong zonal asymmetry is found in January (that is more pronounced than the conventional December – February depiction) and a significantly different structure in July. Therefore it is of advantage to assess a spatially and seasonally confined background mean state. Also, verification metrics based on PV must be handled with care in the tropopause region. [58] Acknowledgments. The authors would like to thank Cornelia B. Schwierz for fruitful discussions and for motivating several aspects of this study. Credit also goes to Heini Wernli for helpful contributions of both a technical and a scientific nature.

References Ambaum, M. (1997), Isentropic formation of the tropopause, J. Atmos. Sci., 54, 555 – 568. Appenzeller, C., and H. C. Davies (1992), Structure of stratospheric intrusions into the troposphere, Nature, 358(6387), 570 – 572. Appenzeller, C., H. C. Davies, and W. A. Norton (1996a), Fragmentation of stratospheric intrusions, J. Geophys. Res., 101(D1), 1435 – 1456. Appenzeller, C., J. R. Holton, and K. H. Rosenlof (1996b), Seasonal variation of mass transport across the tropopause, J. Geophys. Res., 101(D10), 15,071 – 15,078. Beekmann, M., G. Ancellet, and G. Me´gie (1994), Climatology of tropospheric ozone in southern Europe and its relation to potential vorticity, J. Geophys. Res., 99(D6), 12,841 – 12,853. Blackmon, M. L. (1976), A climatological spectral study of the 500 mb geopotential height of the Northern Hemisphere, J. Atmos. Sci., 33, 1607 – 1623. Brunet, G., R. Vautard, B. Legras, and S. Edouard (1995), Potential vorticity on isentropic surfaces: Climatology and diagnostics, Mon. Weather Rev., 123, 1037 – 1058. Danielsen, E. F. (1968), Stratospheric-tropospheric exchange based on radioactivity, ozone and potential vorticity, Q. J. R. Meteorol. Soc., 25, 502 – 518. Danielsen, E. F., R. S. Hipskind, S. E. Gaines, G. W. Sachse, G. L. Gregory, and G. F. Hill (1987), Three-dimensional analysis of potential vorticity

10 of 11

D17102

LINIGER AND DAVIES: PV VARIABILITY AT TROPOPAUSE LEVELS

associated with tropopause folds and observed variations of ozone and carbon-monoxide, J. Geophys. Res., 92(D2), 2103 – 2111. Davis, C. A., and K. A. Emanuel (1991), Potential vorticity diagnostics of cyclogenesis, Mon. Weather Rev., 119, 1929 – 1953. Derome, J., G. Brunet, and J. H. Wang (2001), On the potential vorticity balance on an isentropic surface during normal and anomalous winters, Mon. Weather Rev., 129, 1208 – 1220. Edouard, S., R. Vautard, and G. Brunet (1997), On the maintenance of potential vorticity in isentropic coordinates, Q. J. R. Meteorol. Soc., 123, 2069 – 2094. Gibson, J. K., P. Kallberg, S. Uppala, A. Hernandez, A. Nomura, and E. Serrano (1999), ERA-15 description (version 2), ECMWF Re-Anal. Proj. Rep. Ser. 1, 77 pp., Eur. Cent. for Med. Range Weather Forecasts, Reading, U.K. Holton, J. R., P. H. Haynes, M. E. McIntyre, A. R. Douglass, R. B. Rood, and L. Pfister (1995), Stratosphere-troposphere exchange, Rev. Geophys., 33(4), 403 – 439. Hoskins, B. J., and K. I. Hodges (2002), New perspectives on the Northern Hemisphere winter storm tracks, J. Atmos. Sci., 59, 1041 – 1061. Hoskins, B. J., and P. J. Valdes (1990), On the existence of storm-tracks, J. Atmos. Sci., 47, 1854 – 1864. Hoskins, B. J., M. E. McIntyre, and A. W. Robertson (1985), On the use and significance of isentropic potential vorticity, Q. J. R. Meteorol. Soc., 111, 877 – 946. Kiladis, G. N., K. H. Straub, G. C. Reid, and K. S. Gage (2001), Aspects of interannual and intraseasonal variability of the tropopause and lower stratosphere, Q. J. R. Meteorol. Soc., 127, 1961 – 1983. ¨ ber Aufbau und Entstehung von Zyklonen. 1., Kleinschmidt, E. (1950), U Meteorol. Rundsch., 3(1), 1 – 6. Ko¨ppen, W. (1881), Die Zugstrassen der barometrischen Minima in Europa und auf dem nordatlantischen Ocean und ihr Einfluss auf Wind und Wetter bei uns, Mitt. Geogr. Ges. Hamb., 4, 76 – 97. Liniger, M. A., and H. C. Davies (2003), Sub-structure of a MAP streamer, Q. J. R. Meteorol. Soc., 129, 633 – 651. Massacand, A. C. (1999), Linkages between upper-tropospheric flow and European seasonal weather, Ph.D. thesis, Swiss Fed. Inst. of Technol., Zu¨rich, Switzerland. Massacand, A. C., and H. C. Davies (2001a), Interannual variability of European winter weather: The potential vorticity insight, Atmos. Sci. Lett., 2, doi:10.1006/asle.2001.0026. Massacand, A. C., and H. C. Davies (2001b), Interannual variability of the extratropical northern hemisphere and the potential vorticity wave guide, Atmos. Sci. Lett., 2, doi:10.1006/asle.2001.0027. Massacand, A. C., H. Wernli, and H. C. Davies (1998), Heavy precipitation on the Alpine southside: An upper-level precursor, Geophys. Res. Lett., 25(9), 1435 – 1438.

D17102

Morgan, M. C., and J. W. Nielsen-Gammon (1998), Using tropopause maps to diagnose midlatitude weather systems, Mon. Weather Rev., 126, 2555 – 2579. Pelly, J. L., and B. J. Hoskins (2003), A new perspective on blocking, J. Atmos. Sci., 60, 743 – 755. Petterssen, S. (1956), Weather Analysis and Forecasting, vol. 1, Motion and Motion Systems, 2nd ed., 428 pp., McGraw-Hill, New York. Popovic, J. M., and R. A. Plumb (2001), Eddy shedding from the uppertropospheric Asian monsoon anticyclone, J. Atmos. Sci., 58, 93 – 104. Postel, G. A., and M. H. Hitchman (1999), A climatology of Rossby wave breaking along the subtropical tropopause, J. Atmos. Sci., 56, 359 – 373. Rao, T. N., S. Kirkwood, J. Arvelius, P. von der Gathen, and R. Kivi (2003), Climatology of UTLS ozone and the ratio of ozone and potential vorticity over northern Europe, J. Geophys. Res., 108(D22), 4703, doi:10.1029/2003JD003860. Santer, B. D., et al. (2003), Behavior of tropopause height and atmospheric temperature in models, reanalyses, and observations: Decadal changes, J. Geophys. Res., 108(D1), 4002, doi:10.1029/2002JD002258. Schwierz, C. B., M. Croci-Maspoli, and H. C. Davies (2004), Perspicacious indicators of atmospheric blocking, Geophys. Res. Lett., 31, L06125, doi:10.1029/2003GL019341. Scott, R. K., and J.-P. Cammas (2002), Wave breaking and mixing at the subtropical tropopause, J. Atmos. Sci., 59, 2347 – 2361. Sprenger, M., M. Croci-Maspoli, and H. Wernli (2003), Tropopause folds and cross-tropopause exchange: A global investigation based upon ECMWF analyses for the time period March 2000 to February 2001, J. Geophys. Res., 108(D12), 8518, doi:10.1029/2002JD002587. Swanson, K. L. (2001), Upper-tropospheric potential vorticity fluctuations and the dynamical relevance of the time mean, J. Atmos. Sci., 58, 1815 – 1826. Vaughan, G., and J. D. Price (1991), On the relation between total ozone and meteorology, Q. J. R. Meteorol. Soc., 117, 1281 – 1298. Vaughan, G., H. Gouget, F. M. O’Connor, and D. Wier (2001), Fine-scale layering on the edge of a stratospheric intrusion, Atmos. Environ., 35(12), 2215 – 2221. Wallace, J. M., G.-H. Lim, and M. L. Blackmon (1988), Relationship between cyclone tracks, anticyclone tracks and baroclinic waveguides, J. Atmos. Sci., 45, 439 – 462. 

H. C. Davies, Institute for Atmospheric and Climate Science, ETH, ETH-Hoenggerberg HPP, CH-8093 Zu¨rich, Switzerland. (huw.davies@ env. ethz.ch) M. A. Liniger, Climate Services, Federal Office of Meteorology and Climatology, Kra¨hbu¨hlstr. 58, CH-8044 Zu¨rich, Switzerland. (mark.liniger@ meteoswiss.ch)

11 of 11