The sunspot cycle, the QBO, and the total ozone over Northeastern Europe: a connection through the dynamics of stratospheric circulation

The interaction between the factors of the quasi-biennial oscillation (QBO) and the 11-year solar cycle is considered as an separate factor influencing the interannual January–March variations of total ozone over Northeastern Europe. Linear correlation analysis and the running correlation method are used to examine possible connections between ozone and solar activity at simultaneous moment the QBO phase. Statistically significant correlations between the variations of total ozone in February and, partially, in March, and the sunspot numbers during the different phases of QBO are found. The running correlation method between the ozone and the equatorial zonal wind demonstrates a clear modulation of 11-y solar signal for February and March. Modulation is clearer if the QBO phases are defined at the level of 50 hPa rather than at 30 hPa. The same statistical analyses are conducted also for possible connections between the index of stratospheric circulation C1 and sunspot numbers considering the QBO phase. Statistically significant connections are found for February. The running correlations between the index C1 and the equatorial zonal wind show the clear modulation of 11-y solar signal for February and March. Based on the obtained correlations between the interannual variations of ozone and index C1, it may be concluded that a connection between solar cycle – QBO – ozone occurs through the dynamics of stratospheric circulation.


Introduction
It is well known that the quasi-biennial oscillation (QBO) of zonal winds in the equatorial stratosphere has an in¯uence on the variability of many meteorological parameters in the stratosphere and troposphere (Holton andTan, 1980, 1982;Labitzke, 1982;Wallace and Chang, 1982;Tarasenko, 1988;Holton and Austin, 1991;Dunkerton and Baldwin, 1991;Fedorov et al., 1994;Holton, 1994). However, the relation between the equatorial QBO and the variations of meteorological parameters, as a rule, shows only a tendency and not an obvious connection. However, Labitzke and van Loon (1992) have established that the northern winter stratospheric polar vortex is colder during the westerly ( ) phases than during the easterly (i) phases of QBO only in 64% of winters.
Many researchers (Angell, 1988;Stolarsky et al., 1991;Aleksandrov et al., 1992;Karol et al., 1992;Khrgian, 1992;Bekorukov et al., 1994) have studied a possible connection between the QBO and interannual variations of ozone. They have showed that, on average, at the i phase of QBO the ozone concentration in the polar latitudes is higher than at the phase. According to these authors, the in¯uence of the QBO in¯uences total ozone variations by 1±2%.
Some researchers (Kondratiyev, 1990;Stolarsky et al., 1991;Chandra and McPeters, 1994) have tried to connect the observed interannual anomalies of ozone with solar activity. It was found that the ozone variations caused by the 11-y solar cycle are within the limits of 1±3%. According to the WMO Ozone Assessment (1994), the global average total ozone changes are correlated with solar UV¯ux variations in the 11-y solar cycle, changing by about 1.5% from solar maximum to solar minimum.
Thus, considered separately, the factors of the QBO and the 11-y solar cycle cannot explain the observed interannual variations of ozone, because the role of each of these factors is small. In order to explain the observed anomalies of ozone it is necessary to account for possible interaction between both factors. This interaction may be a third factor and sometimes the role of this factor may be great. Following this assumption, the possibility of connection between the wintertime (January to March) variations of total ozone and the sunspot numbers at simultaneous moment in the QBO phase is investigated in the present study. This research is based on the results of Labitzke (1987), Labitzke and van Loon (1988, 1993, Kodera (1991Kodera ( , 1993, Kodera et al. (1991), van Loon andLabitzke (1994), Soukharev (1997a) which were obtained by analysing a possible connection between the meteorological parameters in the winter stratosphere and the sunspot numbers when the data are grouped according to the phases of QBO. Thus, these authors have noted that during the phase of QBO there is positive correlation between the variations of temperature and geopotential heights in the winter polar stratosphere, and the solar activities. On the contrary, during the i phase of QBO, there is negative correlation. Taking into account also the known close relation between the temperature in the lower stratosphere and the total ozone (Angell, 1987(Angell, , 1988Bojkov, 1988;Kondratiyev, 1990;Shalamyansky et al., 1992;Taalas and Kyro, 1992), in the present study the connection between the ozone and the solar activity is studied with ozone data grouped according to the phases of QBO. In addition a correlation analysis between the indexes of stratospheric circulation and the sunspot numbers bearing in mind the QBO phase is also conducted. Such an analysis allows us to suggest the dynamical mechanism of the connection between ozone and solar activity.
In the present research, to obtain more objective results on possible connection between the solar cycle ± QBO ± ozone, the analysis is conducted separately for two cases: (1) for the de®nition of the QBO phase at 50 hPa, and (2) for the same but at 30 hPa.
The choice of Northeastern Europe for the investigation is determined by the facts that over this region, during the winter-spring season, there is an area of climatological minimum of total ozone, and because of great probability of the formation of ozone mini-holes just over this region (Bojkov et al., 1994).

Data
In the present investigation we used the monthly means of total ozone, updated recently in the Central Aerological Observatory, Moscow and in the Main Geophysical Observatory, St.-Petersburg, Russia, derived by special techniques in months January to March between 1973±1995 on ®ve stations in Northeastern Europe. In all of these stations ®lter ozonometers (M-83 and M-124) are used. Detailed information about the total ozone measurements by ®lter ozonometers as well as about the procedure used for the calibration of the instruments is contained in the paper by Bojkov et al. (1994)  that despite the lack of observations at high-latitudinal stations in January related to diculties in making the ground-based ozone measurements in midwinter, nevertheless because of uniformity of the ozone ®eld inside the coldest part of circumpolar stratospheric vortex, it is possible to calculate the monthly ozone means for a whole region of Northeastern Europe just from the data for 2±3 stations. A minimum of 12 days of observation were required for the monthly ozone mean to be included in consideration. As a rule, during the winter-early spring period the coldest part of stratospheric circumpolar vortex is situated over the region including the all studied stations. According to Shalamyansky (1987), and Shalamyansky et al. (1992), the ozone ®eld inside this sector is uniform. From Table 2 it is clear that monthly long-term means of the total ozone for these stations are very similar. Coecients of correlation between ozone data on the studied stations are high (r b 0X75 always). This fact demonstrates that total ozone for all ®ve stations follows the same annual pattern which is typical for Northern Europe (minimum of ozone is observed in October±November, and maximum in March±April). In addition, correlation analysis between monthly anomalies (deviations from long-term 1973±1995 monthly average) of total ozone at these stations shows that the anomalies are synchronous and correlations are positive 0X4`r`0X7. It is obvious that monthly anomalies of total ozone over the studied stations, are caused by atmospheric processes common over the whole region of Northeastern Europe. The uniformity of ozone ®eld over this region allows the use of monthly ozone means for the whole region.
The long-term monthly means of total ozone for the whole Northeastern region (Table 2) were calculated by the following way. First the data from all ®ve stations for particular months were averaged for each year (available data from stations were summarized and then divided by the number of items), allowing us to obtain the monthly ozone mean for the whole region for the given year. Then the obtained values were divided by number of years for which the monthly means for the whole region were calculated. Thus, in calculating longterm means of total ozone for the whole region the lengths of time series for each station were taken into account.
For analysis of the connection between the total ozone and the dynamics of stratospheric circulation the monthly mean index of stratospheric circulation g 1 , oered by Wallace and Chang (1982), is used. Index g 1 represents a dierence in zonal-mean geopotential height ®eld between 40 and 70°N at the 30 hPa level. It is a characteristic of the intensity of zonal circulation inside the circumpolar vortex (the more g 1 , the stronger zonal transport). The use of index g 1 permits consideration of the dynamics of the winter lower stratosphere. The monthly mean indexes g 1 for the period 1958±1996 were calculated using monthly mean data of geopotential at the 30 hPa in grid points (10 Â 10) for the Northern Hemisphere. These data were received from the Stratospheric Research Group at Berlin Free University.
It should be noted that there is no standard criterion for de®ning the QBO phase. It is well known (Naujokat, 1986;Fedorov et al., 1994;Matveev et al., 1994) that the opposite zonal winds are often observed on the near isobaric levels of equatorial stratosphere (e.g. at 50 and 30 hPa). Holton and Tan (1980) took into account the QBO phase at 50 hPa. van Loon (1988, 1992) for de®nition of the QBO phase used the averaged zonal wind at 50 and 40 hPa. Bugaev and Kats (1971), and Ugryumov (1971) have de®ned the QBO phase at 30 hPa. For the results to be more objective, in the present research the QBO phases were de®ned separately at 30 and 50 hPa. Because of frequent changes of the zonal wind direction in midwinter, in the present work the QBO phase has been de®ned as the averaged zonal winds in January and February (January plus February winds divided by two). Data of zonal wind were also received from Stratospheric Research Group at Berlin Free University.
Total ozone data were given by A. M. Shalamyansky (Main Geophysical Observatory, St.-Petersburg, Russia), who is responsible for the Russian ground-based Table 2. Long-term (1973±1995) monthly means of total ozone (in Dobson units) for the ®ve Northeastern Europe stations from January to March. Besides the long-term means, for the each station the following information is added: the number of years of data which are used to calculate the means, and the standard deviation (sigma) of total ozone for the given month and station 3 Running correlations between the total ozone over Northeastern Europe and the equatorial zonal wind In the recent paper by Soukharev (1997b), where the same ozone data as in this study were used, it was shown that there is a linear correlation between the monthly anomalies of total ozone over Northwestern Russia in wintertime and the sunspot numbers when ozone data are grouped according to the QBO phases. During the phases of QBO, positive correlations and during the i phases of QBO negative correlations are observed. For February, and partially, for March the results are statistically signi®cant (with 5% level of signi®cance). During the phases the coecients of correlation for these months are r 0X63 and r 0X57 respectively. During the i phases the coecients are r À0X89 and r À0X49 respectively. It should be noted that these results were obtained with the QBO phases de®ned according to Labitzke's known catalogue of QBO phases (where the equatorial zonal wind is de®ned as the averaged wind at 50 and 40 hPa). However, it is known that some authors (e.g. Teitelbaum and Bauer, 1990;Salby and Shea, 1991) doubted the statistical reliability of connections between the atmospheric parameters and the solar activity at the account of the QBO phase. In the present research to reveal the connections between the ozone and the solar activity, two methods are used: (1) method of running correlations between the ozone and the equatorial zonal wind; (2) linear correlation between the ozone and the sunspot numbers at and i phases of QBO (at the de®nition of phase at 50 and 30 hPa). The running correlation method was recently used by Kodera (1993) to illustrate the quasi-decadal modulation of the QBO in¯uence on polar stratospheric temperatures. In this work, this method is used to show how this modulation of the QBO aects ozone over Northeastern Europe. Figure 1 shows the January, February and March mean total ozone (Fig. 1a,b,c) and the January-February mean equatorial zonal wind at 50 and  (Fig. 1d,e). According to Kodera (1993), a change in the relationship between and is investigated by calculating the running correlation r for year i as follows: where w 2m 1 is the window width and dashed quantities are deviations from w year mean as The running correlations between the zonal wind and the total ozone for January do not show a clear 11-y signal. However the running correlations between the equatorial zonal wind at 50 hPa and 30 hPa, and the total ozone for February and March respectively (Fig. 2a,b,c,d) demonstrate that there is excellent correspondence between the February variations in r (when the zonal wind is de®ned at 50 hPa) and the 11-y solar cycle. The correspondence is also good in March, especially when zonal wind is de®ned at 50 hPa. A visual comparison of the Fig. 2a, 2b and Fig. 2c, 2d reveals that using the 50 hPa level for the de®nition of the zonal wind, the correspondence between the variations of coecient of running correlations and the sunspot numbers is much clearer than using the 30 hPa level. This fact is in accordance with well-known vertical distribution of ozone: most of the ozone is found in the lower stratosphere. On Fig. 2 the running correlations are calculated for window width of w 3. The experiments which have been made in the present investigation, showed that the changes in window width do not aect the results, apart from the fact that the amplitude of obtained 11-y solar signal is larger for w 3 than for w 5.
In this analysis, the clear 11-y signals in February and March were obtained without any suitable grouping of ozone data according to the QBO phases. This fact shows that there is close connection between the interannual variations of ozone over North-eastern Europe in February and March and the solar cycle with reference to the QBO phase. February and c, d March means of total ozone over Northeastern Europe and the equatorial zonal winds at 50 hPa and 30 hPa respectively. Running correlations are calculated using 3-year window width. e is time series of January mean sunspot numbers. All time series have been normalized (procedure in Fig. 1 caption)

Connection between the dynamics of stratospheric circulation and the solar activity
At the present time it is dicult to explain physically, how the transition from solar minimum to solar maximum can change the sign of the in¯uence of the QBO on the geopotential, temperature and ozone ®elds. Nevertheless, it is clear that the in¯uence of the solar activity cannot be direct, and probably, it occurs as a result of changes in atmospheric circulation. If this is true, then the indexes of stratospheric circulation should re¯ect this in¯uence. Table 3 shows the coecients of linear correlation between the total ozone monthly means over North-eastern Europe and the sunspot numbers as well as between the monthly mean indexes of stratospheric circulation g 1 and the sunspot numbers.
It should be noted that before this analysis the statistically signi®cant negative linear ozone trends were subtracted from initial time series (trends were determined used the standard least squares method). As can be seen from Table 3, there is a positive correlation between the wintertime ozone variations and the January mean sunspot numbers during the phase of QBO, and a negative correlation during the i phase of QBO. Correlation coecients for February are most statistically signi®cant (5% signi®cance level) during the i phase where the QBO phase is de®ned at 50 hPa r À0X89. For March the coecients are less, but are signi®cant during the phase of QBO (r 0X57 and r 0X53 when the QBO phase is de®ned at 50 and 30 hPa respectively). In Table 3 the 5% levels of signi®cance for the correlation coecients were calculated using the method presented by R. Fisher (Panofsky and Brier, 1972). Table 3 shows also that there is negative correlation between the indexes g 1 and the sunspot numbers during the phases and positive correlation during the i phases of QBO. The maximal and statistically signi®cant coecients are for February. It is interesting that, in the same way as the correlation between the ozone and the solar activity, the correlation between the indexes g 1 and the sunspot numbers is also revealed better when the QBO phases is de®ned at 50 hPa rather than at 30 hPa.
In order to demonstrate connection between stratospheric dynamics and solar activity the running correlations method, described already was used again. In a similar way to the running correlations between the total ozone over Northeastern Europe and the equatorial zonal wind, the running correlations between the index g 1 and the zonal wind were calculated separately for two cases: i.e., for the QBO phase at 50 and 30 hPa, respectively. Running correlations for January did not show an 11-y solar signal. However, Fig. 3 shows that there is a good correspondence between the running correlations for February and March, and the sunspot numbers. The best correspondence is observed in February. It is easy to see that the running correlations on Fig. 3 are the same as those on Fig. 2 but only in the reverse sense: the running correlations between the ozone and the zonal wind are in phase with the 11-y solar cycle, but the running correlations between the index g 1 and the zonal wind are in the directly opposite phase with it.
A visual comparison of Figs. 2 and 3 reveals that using the zonal wind at 50 hPa, the correspondence between the running correlations and the sunspot numbers is clearer than using the zonal wind at 30 hPa. An exception is observed for running correlations between g 1 and sunspot numbers in March, when the use of the zonal wind at 30 hPa gives better correspondence.

Discussion
Both the running correlations and linear correlation methods con®rm the assumption that the joint eect of the QBO and of the 11-y solar cycle may be an important factor which in¯uences the interannual variations of total ozone over Northeastern Europe in later winter ± early spring through the changes in stratospheric circulation. The joint eect may be a major factor, causing the interannual February variations of total ozone. If the relation solar cycle ± QBO ± ozone occurs through stratospheric circulation, a connection between the variations of total ozone and index g 1 Table 3. Coecients of linear correlation between the monthly means of total ozone over Northeastern Europe and the sunspot numbers, as well as between the monthly mean indexes g 1 and the sunspot numbers, during the dierent QBO phases de®ned separately at 50 and 30 hPa. For the analyses the monthly ozone means for 1973±1995 with subtracted linear trend for this period and monthly means of index g 1 for 1958±1996 were used. In the ®rst case, total ozone data used for the correlation analysis, were 15 for westerly and 8 for easterly phases of QBO. In the second case, the used data for g 1 , were 25 for westerly and 14 for easterly phases of QBO Correlation between ozone and sunspots Correlation between indexes g 1 and sunspots Isobaric levels should exist. The calculated correlation coecients between the ozone over Northeastern Europe, and the index g 1 , have showed that despite the fact that the index g 1 is global (on hemispheric scale) and ozone variations are considered over small region the connection between these parameters is reasonably close: the correlation coecients are r À0X28; r À0X58; r À0X40 for January, February and March respectively. The largest and statistically signi®cant (5% level of signi®cance) coecient is for February. The connection between changes of ozone and stratospheric dynamics has a clear physical explanation: a weakening of zonal ow in the lower stratosphere leads to the intensi®cation of ozone transport from low to high latitudes, and therefore, to an increase of polar ozone.
Why do the running correlations between the variations of ozone and index g 1 from one side, and the equatorial zonal winds from other side, demonstrate the best modulation of 11-y solar signal just for February? And why the connections between the interannual changes of ozone over Northeastern Europe and the index g 1 are closest just for February too? Probably, the reason is that in February the intensity of meridional circulation is the highest. According to the recent research of Pawson et al. (1995),``although temperature less than 192 K occurs most often in January, the stratosphere may be colder in February, when the 185 K contour encloses a region extending over Europe, which is larger than the small regions near the Pole, which it covers in January''. In addition Kane (1992) noted that most major winter stratospheric warmings are observed just in February.
An approximate scheme showing the association between the solar cycle (considering the QBO phase) and the ozone over the studied region may be the following. The solar cycle during a certain phase of the QBO causes correspondent changes in stratospheric circulation (Fig. 3). Stratospheric dynamics are most closely related to the solar cycle in February (Table 3). As the correlations between the indexes of stratospheric circulation and the variations of ozone are greatest in February, the closest correlations between the ozone and the solar activity are observed also in February ( Fig. 2 and Table 3).
It is obvious that the results obtained in the present study need a physical explanation of the relation between solar activity and stratospheric circulation. In spite of some attempts (e.g., Kodera, 1991) to explain this relation, the real mechanism of this relation continues to be unclear and awaits further investiga- tions. Nevertheless, based on the close connection found between the interannual variations of ozone in February and the solar activity during the certain QBO phase, it is possible to predict the sign of monthly anomaly of total ozone in February over Northeastern Europe.