1. Introduction

Dobson et al. (1927) reported ground-based measurements of the total column ozone using a spectrometer that observed the solar ultraviolet irradiance. They noted that when an upper-tropospheric front passed over the instrument, the total ozone value either dropped or rose sharply. Shalamyanskiy and Romanshkina (1980) and later Karol et al. (1987) divided ground-based total ozone measurements into three regions, separated by the polar and subtropical jet streams. They found that total ozone and temperature profiles had small variability within each region but changed sharply at the polar and subtropical fronts. The same change in ozone across a frontal boundary can be seen in the data from the Total Ozone Mapping Spectrometer (TOMS; McPeters et al. 1996). Figure 1 shows an image of the total ozone measurements for the Northern Hemisphere, taken on 11 March 1990. Three distinct regions can be identified—one dark blue, one green/yellow, and the other red—separated by two boundaries in the total ozone field. These boundaries are shown as solid blue and yellow lines in Fig. 1, and it will later be shown that these lines correspond to the upper-tropospheric subtropical and polar fronts, respectively. The solid red line marks the position of the sharp gradient in the isentropic potential vorticity (IPV) contours on the 450-K isentropic surface, which traditionally is assumed to mark the edge of the polar vortex (Nash et al. 1996; Traub et al. 1995; Schoeberl et al. 1992). These three lines divide the total ozone field into four meteorological regimes, which we define as 1) the arctic regime, within the polar vortex; 2) the polar regime, between the polar front and the polar vortex; 3) the midlatitude regime, between the subtropical and polar fronts; and 4) the tropical regime, between the equator and the subtropical front.

The total ozone archived dataset used in this paper is the TOMS level-3 hierarchical data format product (for details of this dataset see McPeters et al. 1996), which is made up of 1° latitude by 1.25° longitude pixels. In all cases the mean total ozone shown in this paper is an area-weighted mean. Figure 2a shows the zonally averaged (over 1° latitude bands) total ozone amount for all of the data (open circles) and the same average within each of the four regimes. It should be noted that every pixel was used in this analysis. The average for all of the data slowly increases with latitude until the polar vortex is reached. On the other hand, the average for the tropical, midlatitude, and polar regimes are relatively constant over a wide range of overlapping latitudes. There is also a clear difference between the average total ozone amounts for each of these regimes. If those total ozone values close to and at the boundaries are excluded from the average, then the range of values within each regime, except for the Arctic regime, becomes narrower, as shown in Fig. 2b. The important point to be noticed in both Figs. 2a and 2b is that each regime, except for the Arctic regime, has a distinct range of ozone values, which do not overlap with the other regimes. The Arctic regime is only present in the winter months, and for the remainder of this paper the analysis will be restricted to the other three.

This paper is divided into six sections. In section 2, the method used to define the ozone boundaries of the regimes is explained. The meteorological significance of these boundaries is discussed in section 3. In sections 4 and 5 the ozonesonde and temperature profiles within each regime are analyzed. The summary and conclusions of the paper are given in section 6.

2. Derivation of the total ozone boundaries

Shalamyanskiy and Romanshkina (1980) and Karol et al. (1987) used the geopotential height on the 200-hPa level of the maximum wind speed in the subtropical jet to obtain the position of the subtropical front, and on the 300-hPa level for the polar front. They then used these boundaries to separate ground-based total ozone measurements into three regimes on a daily basis. They observed that over the course of a month the value of the geopotential height of these boundaries differed by less than 80 m. Similarly, when they analyzed the ozone value at the position of the boundary, they found that, over the course of a month the total ozone was constant to within a few Dobson units (1000 DU = 1 atm cm). This monthly mean varied little over the 7 yr of the study. The asterisks in Fig. 3 are the mean boundary values that they obtained. These boundary values will be referred to as the Karol data in the remainder of this paper. Under the assumption of small variability of total ozone within each regime, a simple schematic of the ozone field as a function of latitude can be drawn (Fig. 4). In this picture, the ozone value for the boundary of the front is defined as being numerically equal to the average of the ozone values on either side of the front. This definition of the boundary has been adopted in this paper.

In order to obtain objective values of the total ozone at the boundaries as a function of time, the following procedure has been developed. First, the Karol boundaries were used to generate an initial mask for each day. This mask delineates the fronts and identifies those pixels that are within each regime. Figure 5 shows the mask generated for 11 March 1990. The positions of the subtropical and polar fronts were obtained by using a contour program on the TOMS level-3 dataset. The position of the polar vortex was derived from the position of the 31.5–potential vorticity (PV) unit contour [1PV unit (PVU) = 10⁻⁶ K m² kg⁻¹ s⁻¹] on the 450-K potential temperature surface obtained from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996; Kistler et al. 2001).

Second, total ozone data close to the boundaries, where the value is changing from one regime to another, were excluded. This was done by enlarging the initial mask boundaries such that pixels within ±2° latitude and longitude of the central pixel were now included in the boundaries. The daily mean total ozone was then obtained for each regime. New boundary values were obtained from these mean values and then used to derive new masks, and the procedure repeated until convergence was obtained. The average seasonal component of the midpoints for the period from 1980 to 1992, obtained using TOMS level-3 data, are shown in Fig. 3 as dashed lines. There is good agreement between the Karol results and the new analysis for the polar and subtropical fronts. In practice, these average seasonal boundaries are not used in the analysis of the total ozone data, instead a 30-day running mean of the daily boundaries is used. The yellow and blue solid lines in Fig. 1 are those mean daily boundaries for 11 March 1990.

Figure 6a shows the boundaries for the subtropical front, polar front, and polar vortex plotted on the total ozone map for 11 March 1990 along with the potential vorticity contours on the 330-K isentropic surface taken from the NCEP–NCAR reanalysis. Figure 6b shows a similar plot to Fig. 6a for PV values on the 315-K isentropic surface. It is important to point out that the TOMS image shown in Figs. 6a and 6b is taken over a period of 24-h, the right-hand edge corresponding to the international date line (0000 UTC), the left-hand edge to 0000 UTC for the next day. However, the dynamical fields given in the NCEP–NCAR dataset are synoptic, and are produced at four specific times per day, 0000, 0600, 1200, and 1800 UTC. In the comparisons shown in Figs. 6a and 6b, and in those that follow, the NCEP–NCAR dataset has been interpolated with time across the TOMS image. When this is done there is a close agreement between the positions of the fronts outlined by both the total ozone and the gradient in the meteorological data, as shown in Figs. 6a and 6b.

Nash et al. (1996) determined the position of the polar vortex by calculating the maximum slope of the equivalent latitude for PV on the 450-K potential temperature surface, which is always within the stratosphere. This technique however is not accurate for the 330- and 315-K surface as, at their respective fronts, these isentropic surfaces move from the stratosphere to the troposphere, where the PV value becomes small and noisy. Karol et al. (1987) found that although the geopotential height at the center of the front was constant over a month, it had a seasonal dependence. It is quite likely therefore that either the value of the potential temperature, the associated value for the PV, or both at the center of the front will be a function of season. In addition to the seasonal dependence, the total ozone boundary values are also found to decline between 1978 and 1992. It is likely that the isentropic surface and PV values will also change with time. The choice of the 330- and 315-K isentropes for Figs. 6a and 6b is subjective. The choice was made principally because these are two of the potential temperatures for which PV is given in the NCEP–NCAR dataset, and these surfaces move across the subtropical and polar fronts (Shapiro et al. 1987).

A closer look at Fig. 1 shows that the ozone boundary is not necessarily sharp. The yellow color, at around 280 DU, corresponds closely to the transition zone between the tropical and midlatitude regimes. The width of this color band is therefore a good indication of the sharpness of the boundary. The width of the boundary at 10°E longitude is narrow, indicating a sharp boundary, while the boundary around 60°W is wide, indicating a more diffuse boundary. Figures 6a and 6b show the same diffuseness for the PV boundaries, indicating that the variability in the total ozone field is a good surrogate for dynamical variability.

It should be noted that the NCEP–NCAR dataset is obtained from the NCEP global spectral model that has been constrained using data assimilation techniques. As with any optimal estimation scheme, whenever the restraining measurements become sparse, the output of the assimilation scheme will revert to the initial guess. Kistler et al. (2001) note that, “while the NCEP analysis system efficiently assimilates upper air observations, it is only marginally influenced by surface observations.” They note further, in a discussion of the use of the reanalysis data for trend estimates, “in the absence of rawinsondes, the re-analysis results are not reliable, even if there are plenty of surface observations.” The latter statement is significant considering that the bulk of the rawinsonde measurements are restricted to the continents.

As noted above, in general, the agreement between the position of the front determined by PV and ozone is good. However, the agreement is not exact, and in some regions can differ significantly. The ozone boundary is determined directly from the total ozone data using an objective technique. In addition, the total ozone values have a high precision (McPeters et al. 1996). On the other hand, the data from the NCEP–NCAR analysis is model dependent, especially in regions of sparse rawinsonde measurements. More importantly there seems to be no objective way to determine the exact isentropic temperature surfaces and PV values that correspond to the subtropical and polar frontal boundaries. For these reasons, the analysis presented in the remainder of this paper will rely largely on the determination of the frontal boundaries from the total ozone data.

There is one caveat to the last statement. It should be noted that the method described above for determining the boundaries from the ozone data, basically assumes that the tropospheric contribution to the total column ozone is constant over the regime. However, this is not true during the summer months when strong pollution events can occur. Over most of the year the difference in total ozone across a boundary is greater than 50 DU, but in the summer months, for example July, this difference for the subtropical front becomes as small as 30 DU. In general, when the tropospheric ozone does not deviate from the background level, the boundary can be delineated accurately. However, when the tropospheric ozone is much higher then the background, then the boundary becomes distorted. In these cases, geopotential height on the 200-hPa pressure surface from the NCEP–NCAR analysis has been used to define the subtropical front.

Figure 7 shows a histogram of the total ozone values for 19 July 1999 within the tropical regime, using the geopotential boundary for the subtropical front. The overall distribution shown in Fig. 7 (solid line) can be fit with two Gaussian functions displayed as dotted and dashed lines, with peak values of 280 and 310 DU, respectively. Hudson and Thompson (1998) determined from an analysis of TOMS and ozonesonde measurements that the background tropospheric level for column ozone in the tropical regime was constant throughout the year at a value of 24 ± 5 DU. However, when local area pollution occurred this value could be as large as 60 DU. The difference between the background and polluted values is about 30 DU, that is, the same as the difference between the two peaks in Fig. 7, and as large as the difference between the tropical and midlatitude mean values in July.

3. Meteorological significance of the total ozone boundaries

Several groups (Danielsen 1968; Shapiro 1978; Shalamyanskiy and Romanshkina 1980; Shapiro et al. 1987; Uccellini et al. 1985) have used aircraft measurements of ozone concentration, and total ozone measurements to study the upper-tropospheric fronts. Both Shapiro et al. (1987), and Uccellini et al. (1985), found a strong coincidence between large gradients in the total ozone measurements from TOMS and upper-level jet streams/frontal zone tropopause foldings. Shapiro et al. (1987), identified three fronts—the subtropical, polar, and arctic—each with an associated jet stream. It should be noted that the arctic front defined by Shapiro et al. is a tropopause phenomena and need not necessarily correspond to the polar vortex, which is a stratospheric phenomena. During the spring, summer, and fall seasons the arctic front is either weak or not formed.

Figure 8 shows a total ozone image over North America taken on 11 March 1990 for the latitude range 25°–60°N. Also shown in Fig. 8 are the positions of rawinsonde measurements for that day marked as white circles. The frontal boundaries derived from the ozone data are shown as continuous blue and yellow lines for the subtropical and polar fronts, respectively. The dotted lines correspond to 2.0 and 1.8 PVU on the 330- and 315-K potential surfaces, respectively. The white circles filled with red crosses are those measurements identified as being clearly within the tropical regime, green crosses are in the midlatitude regime, and blue crosses are in the polar regime. The white circles without a cross are measurements that fall within ±2° latitude and longitude of the frontal boundaries and were not used in this analysis. In Fig. 9, the temperature as a function of pressure altitude is plotted for each of the rawinsondes selected by regime.

The most significant point shown in Fig. 9 is that each temperature profile within a total ozone regime has the same, distinctive tropopause height. This supports the conclusion of Shalamyanskiy and Romanshkina (1980) and Shapiro et al. (1987), that the ozone boundary and the meteorological boundary are one and the same, a picture consistent with total ozone acting as a surrogate dynamical tracer.

4. Ozonesonde and rawinsonde profiles within the boundaries

As noted in section 1, the total ozone values have small variability within each regime. This should restrict the values that the ozone profile can assume (Bojkov 1969). To test this hypothesis the ozonesonde profiles were separated by regime using broad masks (±2° latitude and longitude of the central pixel). In order to get enough profiles to obtain a reasonable statistical sample, all ozonesonde measurements in the Northern Hemisphere for the period from 1985 to 1990 were used. The stations that were used are listed in Table 1, and the total number of profiles within a regime that were selected within each month is given in Table 2. The data for each ozonesonde was obtained from the World Ozone and Ultraviolet Data Center (WOUDC). Figure 10 shows 20 profiles for each of the regimes, taken at random from the selected March profiles. As was the case for the rawinsonde measurements, the ozone profiles within a regime have a distinct tropopause height. Figure 11 compares the mean ozone and temperature profiles obtained for each regime averaged over the month of March (the mean profiles for the temperature are for March 1990 only). Also shown are the standard deviations of the profiles about the mean. From comparisons with mean ozone profiles from the Stratospheric Aerosol and Gas Experiment (SAGE) instrument, the increase in the standard deviation of the ozone profiles between 100 and 200 hPa represents atmospheric variability and not instrumental error.

The mean values for the tropopause height for each temperature measurement can be derived using the method described by the World Meteorological Organization (WMO 1957). The mean values obtained were 126 ± 26 hPa (480 profiles) for the tropical regime, 229 ± 46 hPa (660 profiles) for the midlatitude regime, and 338 ± 39 hPa (133 profiles) for the polar regime. These values are displayed in Fig. 11. It should be noted that, for each regime, the tropopause heights for both the temperature and ozone profiles are similar. These measurements, then, also lead to the conclusion that the stratospheric ozone field is divided into distinct regimes bounded by the upper-tropospheric meteorological fronts.

Many intercomparisons between ozone profile measurements from different instruments have been made using zonal averages (Harris et al. 1998). But, as shown above, the profile shape depends not only on latitude but also on regime. Figure 12 compares the profiles shown in Fig. 11, for the tropical and midlatitude regimes. Also plotted in Fig. 12 are the mean profiles for the 20 tropical and 20 midlatitude profiles and the mean profile of all 40 profiles. To be considered a physical quantity, the mean of a set of data must be a viable member of that dataset. The mean profiles obtained for the tropical and midlatitude regimes meet this criterion. The overall mean profile does not, and therefore is not a physical quantity. In addition, the average profiles obtained over a given latitude range will be a function of the relative mix of the regimes corresponding to each individual measurement, and will depend strongly on the area sampling for each instrument.

5. Total ozone values within the boundaries

In Fig. 13, the daily mean total ozone value for each regime is plotted for 1990. In general, the values vary quite smoothly from day to day, and even when there is a large variability, that is, in the polar regime from December to March, the mean value varies slowly with time, and variability is obviously not noise. The number of values included in the daily averages shown in Fig. 13 is at least 1000. The standard deviation (in percent of the total ozone value) for the points is also plotted in Fig. 13. These values can range from 1% up to 8%. However, the error of the mean is at least a factor of 30 less than these values, and this is reflected in the smoothness of the values from day to day.

The seasonal dependence of the daily mean is different for each regime. Total ozone in the tropical regime reaches a maximum in July and a minimum in December. The midlatitude regime reaches a maximum in April and a minimum in October; the polar regime reaches a maximum in February and a minimum in September. The seasonal dependence for a zonal average, such as was obtained by Dobson (1968), is a mix of the seasonal dependencies of the three regimes, and will depend on the relative number of total ozone values from each regime that are included in the overall average.

6. Discussion and conclusions

Clear evidence has been presented showing that the Northern Hemisphere total ozone field can be separated into distinct regimes, the boundaries of which are the subtropical and polar upper-tropospheric fronts, and, in the winter, the polar vortex. It has been shown, both from rawinsonde and ozonesonde measurements, that the tropical, midlatitude, and polar regimes are identified with a distinct tropopause height. The standard deviations about the daily mean total ozone value are less than 8% indicating that a unique total ozone value can be assigned to each of the three regimes. Each regime also has a distinct seasonal dependence.

As was noted in the abstract, studies of the variability of total ozone, ozone profiles, and dynamical variables in the lower stratosphere have been centered on zonal averages over specific latitude bands (Harris et al. 1998; World Meteorological Society 1999; Staehelin et al. 2001). However, the boundaries of the ozone regimes are the upper-tropospheric meteorological fronts. These fronts follow the Rossby waves in the atmosphere, and, on any day, can meander over a wide range of latitudes (see, e.g., Fig. 1). Thus, the midlatitude zonal average, whether ozone profile, total ozone, or dynamical variable, will depend on the relative mix of the respective values within each regime over the latitude range of the average. Furthermore, because each regime has such distinctive characteristics, these averages may not have physical significance.