Measurements of the vertical profile of ozone concentration using balloon-borne ECC ozonesondes have been made weekly since 1980 at several sites in Canada (Edmonton, Goose Bay, Churchill, and Resolute), since 1987 at Alert, and since 1992 at Eureka. Previous analyses of ozone trends over Canada have shown strong negative trends in tropospheric ozone. We present here a new analysis of trends in the vertical distribution of ozone with data up to the end of 2001. In addition, more detailed attention is paid to some potential sources of bias: total ozone correction, background current correction, and time-of-launch (diurnal) variation. For the 1980–2001 period the overall linear trends are primarily negative, both in tropospheric and stratospheric ozone. However, when the data for 1991–2001 only are considered, the trends are positive, even in the lower stratosphere. When the time series are compared with previously reported trends (to 1993), it is evident that ozone has rebounded at all levels below about 63 hPa. These differences do not appear to be related to changes in tropopause height, as the average height of the tropopause (as measured over the ozonesonde stations) has not changed over either the 22-year or the 11-year period. Nevertheless, comparison with another dynamical indicator, the wintertime frequency of occurrence of laminae in the ozone profile, suggests that this rebound may be partly a result of small changes in the atmospheric circulation, rather than a recovery of the ozone layer from halocarbon-induced depletion. The long-term trends in average tropospheric ozone concentrations over Canada are similar to corresponding lower stratospheric trends, and tropospheric ozone levels show significant correlation with lower stratospheric ozone amounts.
 In addition to the information they provide on the vertical distribution of ozone in the lower stratosphere, ozone soundings are the major source, worldwide, of information on ozone amounts in the free troposphere. Ozone plays a major role in the chemical and thermal balance of the troposphere, controlling the oxidizing capacity of the lower atmosphere and also acting as an important greenhouse gas, while ozone changes in the stratosphere, as well as strongly affecting surface UV radiation, may also affect future climate [e.g., Gillett and Thompson, 2003, and references therein]. Questions of changes in the vertical distribution of ozone are therefore of considerable importance and have been extensively analyzed and reviewed [Bojkov and Fioletov, 1997; Harris et al., 1997; Randel et al., 1999; World Climate Research Programme, 1998; Staehelin et al., 2001].
 The time series of ozone soundings from Canadian stations comprises some of the longest records of ozone measurement that exist, as well as the only time series of measurements in the free troposphere over Canada. Previous studies of ozonesonde data from Canadian stations have indicated declining concentrations of ozone in the troposphere over Canada [Tarasick et al., 1995, 1996; Oltmans et al., 1998; Logan et al., 1999]. Recent assessments of total ozone time series indicate a rebound in column ozone since about 1993 in the Northern Hemisphere [Staehelin et al., 2001; World Meteorological Organization (WMO), 2003]. Here we use Canadian data up to the end of 2001 to show that this rebound has occurred not just in the lower stratosphere, but at all levels below 63 hPa (20 km).
 All measurements analyzed here were made with ECC sondes; the data record is from 1980–2001, except at Alert, where sondes have been flown only since the end of 1987, and Eureka, which was added to the network in 1992. From 1966–1979, sondes of the Brewer-Mast type were flown at Canadian stations. There are evident systematic differences in sensitivity to tropospheric ozone between the two sonde types [Tarasick et al., 2002] which introduce discontinuities in the time series for each station at 1980, when the Canadian network switched from the Brewer-Mast to the ECC sonde. Because of this, combining the two records will in general produce spurious positive trends in the troposphere, even when both data sets individually have negative trends. The Brewer-Mast data have not yet been reevaluated and are not reliable for trend assessment in their current state. Table 1 describes the locations of Canadian ozonesonde stations and their data records.
Table 1. Canadian Ozonesonde Stations
Start of Sonde Record
Brewer-Mast (1970); ECC (1979)
Brewer-Mast (1969); ECC (1980)
Brewer-Mast (1973); ECC (1979)
Brewer-Mast (1966); ECC (1979)
 Ozonesondes are flown weekly at Canadian stations, normally on Wednesdays at 12:00 or 0:00 GMT. Additional flights are made during times of special interest, like the Arctic spring. Unsuccessful flights and rejection of suspect data reduce the average data frequency to approximately three profiles per month. Sonde preparation procedures have followed Komhyr , with the following minor exceptions: deionized water, rather than distilled, has been used to prepare sensing solutions; sondes are typically charged with solution one week to one month before launch, with 1.5 mL of saturated buffered KI solution in the anode cell and 5.0 mL of 1% buffered KI solution in the cathode cell; solutions are topped up weekly, where necessary, with cathode solution; sensing solution is changed on the day of launch, with 2.5 ml of 1% buffered KI solution used in the cathode cell; procedures such as the decay test and pump flow rate measurement are performed the day before launch, while background current is measured immediately before launch; reozonization of the pump and conditioning of the sensor on low ozone are not performed. Other details of sonde preparation are as given by Komhyr . Sondes are rejected if they show excessive background current (>0.2 μa), or fail the ozone decay test.
 These details regarding sonde preparation procedures are presented here as a matter of record, but do not affect trends, since procedures, sensing solution volumes and concentrations have not been changed at any time in the Canadian ECC record. Operational constraints have at times, however, required some variation in launch times (see below) and station personnel. The construction of the ECC sonde has also undergone minor changes over the period 1980–2001, and the associated radiosonde has changed (Table 2). These differences and their possible effects are discussed below.
Table 2. Changes in the ECC Ozonesonde and Associated Radiosonde
ECC 4A introduced
redesigned pump; maximum change <1%, at 50–20 hPa.
ECC 5A introduced
new pump correction; maximum change ∼1%, at 100 hPa.
Vaisala RS-80, RSA-11 introduced
may introduce altitude shift in profile above 25 hPa (25 km)
 The 3A sondes were used in Canada until about 1985, when a change was made to the newer 4A type. The major difference from the 3A was a redesigned pump. Differences, as evidenced by the difference in pump corrections supplied by the manufacturer, are at most about 1%, in the 50–20 hPa region. A few years later the pump motor was changed in the 4A sonde. Neither of these alterations should affect tropospheric results, nor is any effect in stratospheric values evident. More recently the 5A and 6A digital sondes have been introduced, as well as the ENSCI 1Z. No systematic differences in these models have been observed below about 20–25 km [Smit et al., 2000]. The type of radiosonde has also changed, from the older VIZ 1680 MHz to the Vaisala RS-80 with the RSA-11 interface. Changes in radiosonde may influence the ozone profile by introducing altitude shifts, primarily above 25 hPa (25 km). Lemoine and De Backer  suggest that this change could increase ozone values by 10% or more above 25 hPa. No such large changes are evident in the time series presented below (Figures 8 and 9), but this could influence the trends we show here for the three uppermost layers.
 The efficiency of the ozonesonde pump decreases at low pressures, and a correction for this is part of normal data reduction. Standard pump corrections recommended by the manufacturer (which necessarily differ between 3A and 4A models) have been used for all sondes. The correction curve supplied in 1983 has been used for all 4A flights; this was revised slightly by Komhyr  and, as the Vaisala software uses the newer correction curve, this has caused a change with the 5A sondes. The maximum difference is about 1%, which will introduce a bias in the trends for the 158–100 hPa and 100–63 hPa layers of about +0.3% per decade. We have not made a correction for this insignificant bias. The pump flow rate measured in the preflight procedure is used in the data reduction. Data reduction and quality control are performed individually on each profile.
 Although the sonde measurement is in principle absolute, in practice average errors of 3–6% are typical [Smit et al., 1996]. For this reason Canadian ozone soundings are scaled linearly to the total ozone measurement, where available. This process (correction) introduces a degree of uncertainty because the amount of ozone above the balloon burst height can only be estimated. For Canadian soundings the residual ozone is estimated by assuming a constant mixing ratio above the balloon burst altitude. The assumed mixing ratio is the average of either all the data points above 17 hPa, or (if there are more than three) of the last three data points in the profile. No correction is applied for flights that fail to reach 17 hPa.
 The sets of correction factors for each station are summarized in Figure 1. Only one station, Edmonton, shows a trend that is more than marginally significant: −3.1 ± 1.2% (95% confidence) per decade. This trend is puzzling, because it is both large (nearly 20 DU over the period 1980–1995) and robust: we find approximately the same trend when only correction factors derived from direct sun measurements are considered, or only those from 1980–1990 or from 1985–1995. The residual ozone for this site does show a relatively large trend (10.8 ± 8.4% per decade), which is caused by a small negative trend in the average balloon burst height. This would contribute to the correction factor trend if in fact the constant mixing ratio assumption were overestimating the residual ozone, as suggested by McPeters et al. . However, the possible contribution is small: since the average amount of calculated residual ozone is 35.6 DU, even if this were 50% too high it would account for only 3 DU of the 20 DU change over the period 1980–1995. Another possible contributor is a change in the gradient of mixing ratio above burst height. Observations by SAGE for this period indicate a decrease of nearly 10% per decade at 40 km, relative to the average balloon burst height of 30 km [WMO, 1999]. However, this would account for only another 2–3 DU. This implies that either the Edmonton Brewer or the average sonde response (at Edmonton only) has changed by between 4 and 6%, both possibilities that seem unlikely. Comparison of the Brewer measurements with TOMS overpasses shows the Edmonton Brewer to be as much as 2–3% lower than TOMS in the 1992–1994 period. However, the TOMS record during this period is less reliable, spanning as it does the last years of Nimbus-7 TOMS and Meteor-3 TOMS. In any case, Goose Bay, at the same latitude, shows a similar negative bias relative to TOMS over this period, but a positive trend in correction factors.
 In Figure 1 both Resolute and Churchill also show negative trends in correction factors, but these are for a relatively small number of points (for most flights at these sites no correction factor is available). The lack of consistency between the trends at these four sites argues against there being a common cause, such a change of sonde response, or altitude measurement, since the same effect should have been seen at all sites. However, if the Edmonton trend in correction factors does indeed indicate a real change of effective sonde response, then the ozone trends for the 1980–2001 period should be more negative than those we present here.
 The small background current that sondes produce in the absence of ozone is measured just before launch by exposing the sonde to ozone-free air, and subtracted from the raw ozone measurement. Standard practice is to subtract an amount proportional to atmospheric pressure, on the assumption [Komhyr, 1969] that the background current is due to reaction with oxygen rather than ozone. This pressure-dependent adjustment for background current is used throughout the Canadian record. Laboratory studies [Thornton and Niazy, 1982, 1983; Smit et al., 1994], however, indicate that background current in the ECC sonde shows no oxygen dependence and is probably due to residual tri-iodide in solution. It should therefore be constant, at least in the troposphere. An analysis of background current data (Figure 2) indicates that at four stations there are significant trends in background current, some of which are large enough to affect derived ozone trends in the upper troposphere (where the ozone partial pressure is low), if the background current were in fact constant with altitude. The causes of these trends appear to be several: the negative trends at Churchill and Edmonton result from a greater frequency of outliers (sondes with high background current) in the early part of the record, while the positive trends at Resolute and particularly at Goose Bay indicate a real, continuous change in measured background, possibly caused by a deteriorating ozone filter or some other instrumental drift. There are no significant differences related to sonde type or manufacturer.
 Total ozone correction, background current subtraction, and adjustments for quasi-biennial oscillation (QBO) and solar cycle effects, are all part of the postmeasurement analysis, and so it is possible to conduct a sensitivity analysis of the effect of these different corrections. We therefore calculated trends for four cases: (1) using the total ozone corrected data, with pressure-dependent background subtraction and no QBO or solar cycle adjustment; (2) using the uncorrected data, but otherwise as for the previous case; (3) using the corrected data, with constant background, no QBO or solar cycle; and (4) using the corrected data, with pressure-dependent background subtraction, and using a statistical model [World Climate Research Programme, 1998] that includes both quasi-biennial oscillation (QBO) and solar cycle terms. The uncorrected data set includes all flights, without scaling. The corrected data set includes all flights with correction factors between 0.8 and 1.2, as well as all flights (necessarily unscaled) for which no total ozone measurement was available. Comparison of the trends for the four cases shows mostly minor differences resulting from these correction and calculation methods. The large trend in correction factors at Edmonton, discussed above, causes a much smaller difference (∼1% per decade) in the trends for this station, in part because some 40% of soundings are uncorrected in either data set (i.e., no total ozone measurement is available). The background current subtraction, as anticipated, has an important effect only on trends in the upper troposphere. At Goose Bay the effect is as large as −4% per decade, and at Edmonton +3%. These (maximum) differences are of the same order as the trends at this level, however. Trends calculated by the statistical model are somewhat more positive in the stratosphere at Resolute and Churchill (by 2–3% per decade) and at Alert (by 3–6% per decade, although because of the shorter time series at Alert these are all nonsignificant). These differences are important; nevertheless the overall pattern of trends is similar for the four cases.
 The results are not critically sensitive to the choice of analysis model, and we choose, for simplicity, as well as for consistency with a majority of previous analyses, to base the subsequent analysis on the corrected data set, using the standard pressure-dependent background current subtraction, and without subtraction of QBO or solar cycle terms. For this analysis the ozone profile was represented by a ground-level measurement (the ozone measured at sonde release) and 11 layers equally spaced in log pressure (layers ∼3 km in thickness). In addition, the troposphere and stratosphere have been explicitly separated: that is, integration for the 400–250 hPa layer is from 400 hPa to 250 hPa or the tropopause, whichever comes first. Similarly, integration of the 250–158 hPa layer starts either at 250 hPa or at the tropopause, if the latter is found above 250 hPa. Cases where the tropopause is below the 400 hPa height or above 158 hPa occur rarely but are dealt with similarly). This provides a cleaner separation of the time series of tropospheric and stratospheric air at the tropopause. We use the WMO definition of the tropopause, that is, the lowest height at which the temperature lapse rate falls to 2°C/km or less, provided that the average lapse rate for 2 km above this height is also not more than 2°C/km.
 The sonde data were integrated within these 11 layers to produce values for the partial ozone column in each layer; this was then divided by the pressure change across the layer to produce values for average ozone mixing ratio. The ground-level and layer mixing ratio values were deseasonalized by subtracting the smoothed average annual cycle as described in Tarasick et al. . The deseasonalized time series were then adjusted for the effects of diurnal variation in ozone concentration. Sondes are launched at either 12:00 or 0:00 GMT, which are early morning and midafternoon in Edmonton, and somewhat later at other stations. The effect of changing the time of launch is quite evident in the tropospheric ozone data for the mid-1980s at Edmonton, where mean values increased by about 42% at ground level and 14% below 700 hPa when the standard launch time was changed from morning to afternoon [Tarasick et al., 1995]. This diurnal difference could bias calculated long-term trends, if unevenly distributed through the record. The amount of diurnal shift (a single scalar value for each station at each level) was calculated as the average difference between values for the two launch times, where both were available in the same year and month. Although the effect was significant only at Edmonton and Alert, and only in the troposphere there, for consistency all stations were adjusted at all levels.
 The time series were then averaged by month, and a simple linear regression was used to derive trends from these final averages. Trends are expressed as per cent per decade, relative to the layer mean.
 The time series were also examined for autocorrelation. This analysis showed, in general, positive and significant correlations at all levels, at time lags of one month or less. Autocorrelation coefficients calculated on the time series of monthly means were also in general positive and significant, ranging between 0 and 0.6. The highest values were found for the 250–158 hPa layer at Alert and Eureka. Over all stations, average autocorrelation coefficient values were highest in the lower and middle stratosphere (0.2 to 0.35) and the lower and middle troposphere (0.2 to 0.24). A significant autocorrelation coefficient implies that the data are oversampled (i.e., not statistically independent) at the monthly level, so that the probable errors associated with trends so derived would be underestimates. Allowance is made for this in the confidence limits for trends by basing the confidence limit calculation on a (reduced) effective sample size
where ρ is the lag 1 autocorrelation coefficient, and the ozone variability is assumed to be an AR(1) process [Zwiers and von Storch, 1995; Thiébaux and Zwiers, 1984]. Values of neff/n ranged between 0.25 and 1, with average values of about 0.7. The lowest values were found for the 250–158 hPa layer at Alert and Eureka.
 Correlations between different ozone layers and dynamical indicators were calculated on annually averaged data, and correlation coefficients were tested for significance via the statistic [Anderson, 1958]
where r is the calculated correlation coefficient. In all cases correlation calculations were performed on the residuals, after removal of the linear trend from each time series.
4. Trends in Ozone
Figure 3 shows calculated trends in ozone mixing ratio from ECC ozonesonde data from 1980–2001 (for Alert and Eureka from 1987 and 1992 respectively), for the ground level and 11 layers equally spaced in log pressure. For the four stations with records spanning the entire 22-year period, at all levels the trends are primarily negative. However, except for the ground level and the lower/middle stratosphere (100–25 hPa), most of the trends are negligible, and not statistically significant at the 95% (2σ) level. Interestingly, the data for 1980–1990 (Figure 4) present a different picture, with strong declines from the middle stratosphere right down to the ground, while the data for 1991–2001 (Figure 5) show almost the complete opposite, with strong increases from the middle stratosphere right down to the ground. In Figure 5, all the trends below 63 hPa are positive. It is also clear that the trends for Alert and Eureka, which appear anomalous in Figure 3, are similar to the other stations when compared over the same time period. Also in Figure 5, the three Arctic sites (Resolute, Eureka and Alert) all show the strongest increases in the lowermost stratosphere. The time series for this layer is dominated by the high spring values of 1998, 1999, and 2001.
 The reason for these very different results from earlier analyses [Tarasick et al., 1995, 1996; Oltmans et al., 1998; Logan et al., 1999] appears to be a strong increase of ozone in the troposphere and lower stratosphere since 1993. This is illustrated in Figures 6–9 which show the average ozone mixing ratio time series for the three midlatitude stations and the three Arctic stations. The trend line based on data up to the end of 1993 is also indicated for reference. It is evident that ozone has rebounded at all levels below about 63 hPa (as indeed indicated by Figure 5).
 Trends above 40 hPa are all nonsignificant at the Arctic sites, partly because of the much higher variability in winter and spring at these sites when compared with the midlatitude sites (Figures 8 and 9). This is probably due in part to the changing position of the Arctic vortex. Interestingly, the midlatitude time series for the 40–25 hPa layer shows the least variability of any in Figures 6–9, and also appears unaffected by both the dramatic losses of 1992 and 1993 and the subsequent rebound, seen clearly in lower layers; it fits well over the entire 22-year record to the trend calculated from data up to 1993. In contrast, the same layer in the Arctic, and indeed all stratospheric layers up to 15.8 hPa, show the springtime losses of 1996, 1997, and 2000.
 Perhaps surprisingly, the ground-level trends are much larger, in general, than those in the eleven free atmospheric layers. Trends for the three tropospheric layers at Edmonton are close to zero, while the ground-level trend is dramatically negative. Goose Bay and Churchill also show large negative trends. The reasons for these dramatic trends are rather mundane, however: ozone is destroyed on most physical surfaces (including buildings), with the notable exception of snow. This is a known problem for ground-level ozone monitoring sites and most take measurements at least 4 to 5m above the surface, in order to reduce the effects of variable surface deposition. Ground-level measurements are taken just before sonde release, with the sonde at a height of about 1 m or less above the ground surface. Release-level measurements from the sondes will consequently be especially affected by surface deposition processes, particularly in summer and fall. In addition, like other ground-level ozone monitoring methods, they will be affected by boundary layer pollution. This also serves to explain the very high variability in the ground-level time series (Figure 6). Indeed, ground-level values for Edmonton are in general anomalously low in the winter, including frequent events of less than 10 ppbv; these are probably due to local pollution events (NO titration). However, the trend in ground-level ozone at this site is largest (most negative) in summer, which suggests a change in surface deposition, (rather than increased boundary layer pollution) as the city of Edmonton has expanded over the past twenty years toward the Stony Plain sonde station, which is located about 30 km west of the city center. The low values do not appear to be due to interference of pollutants with the sonde itself, as comparison of the sonde release-level measurements with those from a recently installed TECO Model 49 ozone analyzer shows generally good agreement throughout the past year.
 Tropopause height changes do not appear to explain any part of the observed rebound in ozone below 63 hPa. The average tropopause height (Figure 10) shows no trend on either the 10- or 20-year timescale. Therefore a simple change in the temperature structure of the atmosphere does not appear, on average, to have affected Canadian trends. While this is in contrast to the findings of Steinbrecht et al.  (for a single station, Hohenpeissenberg); in fact, trends in tropopause height for individual Canadian stations (Figure 11), do appear to be at least qualitatively consistent with the Steinbrecht et al. result. The surprising positive trend in ozone in the lowest stratospheric layers (250–158 hPa and 158–100 hPa) at Goose Bay in 1980–1990 (Figure 4) may be partly accounted for by the large negative trend of about 7% (700 m) per decade in tropopause height at that station (for reference, the trend found by Steinbrecht et al. was about 14% per decade). Edmonton and Churchill have positive trends in tropopause height during this period and large negative trends in lower stratospheric ozone, while the trend in 250–100 hPa ozone at Resolute, with a negative trend in tropopause height, is smaller. Again, ozone trends in the lowest stratospheric layers in 1991–2001 are most positive at Edmonton, Resolute, Alert and Eureka, the stations with negative trends in tropopause height, and smallest at the other two stations. The sole station with a significant (positive) trend in tropopause height over the entire 22-year period, Churchill, is also the station with the largest lower stratospheric ozone decline over that period. Thus it appears that a statistical model like that of Appenzeller et al. , incorporating a term proportional to tropopause height, might reduce some of the differences between trends at different Canadian stations, while not changing the overall trend picture.
Figures 12–14 show trends calculated for 1-km layers within 5 km above or below the tropopause. These indicate, in general, little asymmetry about the tropopause, with strong declines before 1991 and strong increases after. The trends in the immediate vicinity of the tropopause thus show similar behavior to those in the rest of the troposphere and lower stratosphere (Figures 3–5). The sole exception is at Resolute in the upper troposphere, where significant losses appear for 3 km below the tropopause. (This is also apparent in Figure 3.) This decline is apparently local, as it is not seen at any of the other stations, and likely related to local meteorological changes, e.g., a change in horizontal transport, increased vertical mixing (reducing the vertical mixing ratio gradient), or a reduction in the frequency or magnitude of stratosphere- troposphere exchange (STE) events.
 The strong rebound of ozone at levels below about 63 hPa is surprising, and is also seen in total ozone records for the Northern Hemisphere [WMO, 2003]. One reason for it is evidently a recovery from the effects of the aerosol introduced by the 1991 eruption of Mt. Pinatubo. However, stratospheric aerosol loading had dropped to normal levels by 1996 [WMO, 2003], while the time series in Figures 6–9 show continued increases. Although total effective chlorine loading in the stratosphere probably peaked not long after this point, it would not have decreased by 2001 by more than 5% [WMO, 2003]. Thus it seems unlikely that the ozone increases in the last five years of Figures 6–9 are evidence of a recovery from halocarbon-induced depletion, like that recently considered by Newchurch et al.  and Steinbrecht et al.  for the 35–45 km region.
 Ozone profiles often show isolated layers of anomalous values. This laminar structure is a result of differential horizontal advection of air masses containing differing amounts of ozone, and so is a direct manifestation of the large-scale quasi-horizontal motions that dominate transport in the stratosphere [Orsolini, 1995]. Laminae in ozonesonde profiles have been used by several authors as an indicator of ozone transport [Reid and Vaughan, 1991; Mlch and Laštovička, 1996, 1997; Reid et al., 2000]. Following Reid and Vaughan , we defined a lamina as a sharp feature that deviated from the general shape of an ozone profile by at least 20 nanobars. A Fourier transform algorithm with a rectangular spectral window was used to pick out laminar features with depths between 240 m and 2500 m from the general shape of the ozone profile. A number of profiles were examined by hand to verify that the results of this method were very similar to those obtained with other algorithms based on comparing nearby points in the profile [e.g., Reid and Vaughan, 1991; Mlch and Laštovička, 1996].
 We compared the average frequency of occurrence of laminae in ozone profiles over Canada during the winter months (November–April), when most poleward transport of ozone takes place, with annual average ozone mixing ratio anomalies over Canada, for the lower stratospheric layers. Figure 15 shows an example of this for the three Arctic stations (Resolute, Eureka, and Alert). The annual values shown are averages of individual station averages, like the time series in Figures 6–9. Although the frequency of occurrence of laminae does not reproduce all of the features of the lower stratospheric ozone time series (notably not the large losses of 1993, at the midlatitude stations), it does show statistically significant (95% confidence) correlation with ozone interannual variations in most of the lower stratospheric layers, and it also shows a similar change of trend in the 1990s.
Table 3 shows the results of this correlation analysis: significant correlations with both positive and negative laminae are found for the three lowest stratospheric layers (250–63 hPa), at northern midlatitudes, and for the four lower stratospheric layers (250–40 hPa) in the Arctic. Correlation coefficients are as high as 0.59 for the midlatitude case and up to 0.75 in the Arctic. Correlations with negative laminae (representing excursions of Arctic, or vortex air) are somewhat higher than with positive laminae.
Table 3. Correlations Between the Frequency of Occurrence of Positive and Negative Laminae During the Winter Months (November–April) and Annual Average Ozone Mixing Ratio Anomalies in the Lower Stratosphere for Northern Midlatitude and Arctic Stationsa
Statistically significant (95% confidence) correlations are indicated in bold.
 The relative magnitude of the variations in frequency of lamina occurrence is much larger than that of the ozone variations themselves. The strong correlation between the two suggests that the interannual variability in the latter time series, as well as the change in trend direction, is driven by variations in the strength of the wintertime transport.
 The high correlation seen in Figure 15 permits us to calculate the ozone anomaly that is associated with a given change in the frequency of laminae, and therefore to subtract the part of the ozone time series in that is apparently due to dynamical changes. The result, shown in Figure 16, is a surprisingly linear curve. To the extent that the frequency of laminae is a valid indicator of dynamical influences on stratospheric ozone, this curve should more closely represent the “chemical” trend in ozone at this altitude.
 Also intriguing is the similarity between the tropospheric and lower stratospheric time series, most evident in the midlatitude time series (Figure 6) but also in the Arctic (Figure 7). This is also reflected in the shorter-term trends (Figures 4 and 5): The three free tropopheric layers show essentially the same behavior as the lower stratosphere. More significantly, the month-to-month variations in all layers up to 63 hPa appear to show some correlation. In particular the large losses in the lower stratosphere in 1993 [Kerr et al., 1993], thought to be caused by aerosol from the Mt. Pinatubo eruption, are reflected in losses in all three tropospheric layers.
Table 4 shows the results of a correlation analysis of annual average ozone anomalies for the ground level and the 11 free atmospheric layers. Statistically significant (95% confidence) correlations are found for the lowest stratospheric layer (250–158 hPa), right down to the ground, for both northern midlatitudes and the Arctic. At northern midlatitudes mixing ratio in the four lower stratospheric layers (250–40 hPa) is correlated with that in all three tropospheric layers, with coefficients as high as 0.69, while in the Arctic two lowest stratospheric layers (250–100 hPa) show significant correlations with all three tropospheric layers, with coefficients as high as 0.67.
Table 4. Correlations Between Annual Average Ozone Mixing Ratio Anomalies in the Troposphere and in the Lower Stratosphere for Northern Midlatitude and Arctic Stationsa
Ground to 630
Ground to 630
Statistically significant (95% confidence) correlations are indicated in bold.
Ground to 630 hPa
 A similar correlation in stratospheric and tropospheric ozone levels has been reported for Sodankylä (67N, 27E) and Marambio (64S, 57W) [Taalas et al., 1997]. Like the Canadian sites, these are relatively high latitude sites remote from major anthropogenic pollution sources.
 However, rates of both photochemical production and destruction of ozone in the troposphere are affected by levels of UV irradiance. In relatively clean areas, such as the six Canadian sites studied here, increased UV, from decreased lower stratospheric ozone, will cause enhanced photochemical destruction of ozone in the troposphere, and vice versa [Tang et al., 1998; Fuglestvedt et al., 1994; Liu and Trainer, 1988]. This mechanism would also explain the observed correlation in stratospheric and tropospheric ozone levels.
 An analysis of trends in the vertical distribution of ozone, employing ECC ozonesonde data for the period 1980–2001 from six Canadian sites, while showing overall decreases in concentration at all levels below 20hPa, indicates a rebound of tropospheric and lower stratospheric ozone since 1993. The overall pattern of these trends is relatively insensitive to several different methods of calculating them. In general the trends are not obviously related to changes in tropopause height, and the average height of the tropopause (as measured over the ozonesonde stations) has not changed over either the entire 22-year or the most recent 11-year period. Nevertheless, we find some success in relating these changes to another dynamical indicator, the wintertime frequency of occurrence of laminae in the ozone profile. This suggests that the aforementioned rebound may be partly a result of small changes in the atmospheric circulation, rather than a recovery from halocarbon-induced depletion, like that recently discussed by Newchurch et al.  and Steinbrecht et al.  for the 35–45 km region. Individual station trends also show notable differences, some of which may indeed be caused by relative changes in tropopause height at individual stations.
 The long-term trends in average tropospheric ozone concentrations over Canada are similar to corresponding lower stratospheric trends, and tropospheric ozone levels show significant correlation with lower stratospheric ozone amounts. This may be due to STE processes, or to the effect of stratospheric ozone changes on UV-induced photochemical destruction of tropospheric ozone.