Sixteen Years of Measurements of Ozone over Athens, Greece with a Brewer Spectrophotometer

: Sixteen years (July 2003–July 2019) of ground-based measurements of total ozone in the urban environment of Athens, Greece, are analyzed in this work. Measurements were acquired with a single Brewer monochromator operating on the roof of the Biomedical Research Foundation of the Academy of Athens since July 2003. We estimate a 16-year climatological mean of total ozone in Athens of about 322 DU, with no signiﬁcant change since 2003. Ozone data from the Brewer spectrophotometer were compared with TOMS, OMI, and GOME-2A satellite retrievals. The results reveal excellent correlations between the ground-based and satellite ozone measurements greater than 0.9. The variability of total ozone over Athens related to the seasonal cycle, the quasi biennial oscillation (QBO), the El Nino Southern Oscillation (ENSO), the North Atlantic Oscillation (NAO), the 11-year solar cycle, and tropopause pressure variability is presented.


Introduction
Ozone is a minor natural component of the clean atmosphere, found primarily in two regions. Approximately 10% of the Earth's atmospheric ozone resides in the troposphere, while 90% is found in the stratosphere (commonly referred to as the "ozone layer") [1].
Year-to-year fluctuations in total ozone are determined by the balance between chemical processes that produce and destroy ozone and the effects of atmospheric motions that transport ozone [2]. Certain industrial processes and human activities are the root cause of the release of ozone-depleting substances (ODSs) into the atmosphere. ODSs are manufactured halogen source gases that are controlled worldwide by the Montreal Protocol.
Concern about changes in ozone abundance is an important subject, not only for the scientific community, but the general public and governments as well. The importance of observational and modeling results about ozone trends lies in its tremendous importance for the life and ecosystems at the location under investigation [3]. Changes in stratospheric ozone can change the large-scale atmospheric state, influencing the climate, both directly through radiative effects, and indirectly by affecting stratospheric and tropospheric circulation [4].
Total ozone measurements have been conducted in Athens, Greece, since 1989, with a Dobson spectrophotometer No 118, which is part of the World Ozone and UV Data Centre (WOUDC) of the WMO [5]. The authors found a correlation coefficient of 0.96 with Total Ozone Mapping Spectrometer (TOMS) data, although the TOMS values were slightly lower than the Dobson ones. The Dobson measurements were also compared with TOMS (version 6) and solar backscatter ultraviolet radiometer (SBUV) measurements, and better correlations were obtained on sunny days [6]. Long-term measurements of stratospheric ozone in Greece have also been conducted in Thessaloniki since 1982, with a MKII Brewer spectrophotometer #005 [7].
The aim of this study was the estimation of the variability and trends of total ozone over Athens, Greece, from a Brewer spectrophotometer operating in Athens since July 2003. This research contributes to developing understanding of the processes that control ozone abundance. The ozone data set from the Brewer spectrophotometer is compared with TOMS, Ozone Monitoring Instrument (OMI), and Global Ozone Monitoring Experiment 2 (GOME-2A and GOME-2B) satellite retrievals. This is the first time we have analyzed long-term ground-based measurements of total ozone in Athens with the Brewer spectrophotometer. The measurements cover the period 2003-2019; i.e., after the ozone decline of the 1980s and 1990s [8]. Detailed information on the data sources and methods are provided in Section 2. In Section 3, daily values, correlations, and monthly mean total ozone time series, as well as the ozone variability, are presented and described in detail. Finally, Section 4 provides concluding remarks on the main findings of this study.

Data Sources and Methods
In this study, we used measurements of total ozone column, made using a single Brewer MKIV spectrophotometer. This Brewer #001 monochromator has measured the columnar amount of ozone in Athens on a daily basis, since July 2003. The measurements are conducted on the roof of the Biomedical Research Foundation of the Academy of Athens (37.99 • N, 23.78 • E) at approximately 180 m a.s.l. [9]. The institute is located in a green area at about 4 km, away from the city center. On the east side of the station is mountain Hymettus, at a distance of about 1 km, and to the north and northeast of the station we find the large mountains of the county of Attica, Parnes, and Penteli, at distances of about 15 and 20 km from the station, respectively. Finally, to the south, the Saronic Gulf is about 10 km away [10].
The Brewer is an automated, diffraction-grating spectrometer that provides observations of the sun's intensity in the near UV range. The instrument measures the intensity of radiation in the UV absorption spectrum of ozone at five wavelengths (306.3, 310.1, 313.5, 316.8, and 320.1 nm) with a resolution of 0.5 nm. These data are used to derive columnar ozone and sulfur dioxide amounts and the aerosol optical depth [11]. The total ozone column (TOC) is calculated as follows [12]: where F is the weighted ratio of direct sun measurements at 4 spectral channels, i.e., F 0 , ∆β, and ∆α are the same linear combinations for logI 0(λ) , β λ , and α λ , i.e., β λ is the Rayleigh scattering coefficient at λ, m is the effective pathlength of direct radiation through air, α λ is the ozone absorption coefficient at λ, and µ is the ratio of the effective pathlength of direct radiation through ozone to the vertical path. The extra-terrestrial constants F 0 are determined from a long series of intercomparison measurements, as well as zero air mass (µ) extrapolations. The instrument is calibrated regularly by the travelling standard Brewer #017, which is operated by International Ozone Services Inc., Toronto, Ontario, Canada (www. September 2019. Information about the stability of the instrument obtained from the results of the calibrations is presented in the Supplementary Materials of this study. Internal standard lamp tests are performed on a daily basis to detect possible instrumental drifts. Ozone data are recalculated after standard lamp test corrections and are analyzed using the O3BREWER data management software [13]. We note here that the Brewer #001 ozone data have been used in the past to evaluate NILU-UV multi-channel radiometer ozone data [14] and ultraviolet multifilter radiometer (UV-MFR) ozone retrievals [15].
The effect of stray light [16] or the effect of temperature dependence [17] may result in errors in the Brewer UV measurements and, consequently, in ozone retrievals. It is known that ozone measurements from a single monochromator Brewer spectrophotometers suffer from non-linearity at large ozone slant column amounts, due to the presence of instrumental stray light caused by scattering within the optics of the instrument. As the light path (air mass) through ozone increases, the effect of stray light on the measurements also increases [18]. In our study, in order to avoid any possible erroneous measurements at large solar zenith angles, we processed ozone measurements up to 70 solar zenith angles. Regarding the temperature dependence effect, there is no stratospheric temperature correction of ozone absorption coefficients in the latest version of the O3Brewer software which we used. At this point, it is worth mentioning that only direct sun (DS) measurements were processed to retrieve the daily TOC values; hence, measurements in the zenith sky scattered mode have not been considered.
The Quasi Biennial Oscillation (QBO) component at 30 hPa on total ozone was examined by analyzing the monthly mean zonal winds at Singapore at 30 hPa (QBO30). For QBO at 50 hPa, we analyzed the monthly mean zonal winds at 50 hPa (QBO50). The data were provided by the Freie Universität Berlin (FU-Berlin) at http://www.geo.fu-berlin.de/met/ ag/strat/produkte/qbo/qbo.dat (accessed on 8 May 2021) [19]. The possible impact of El Nino Southern Oscillation (ENSO) was examined by using the Southern Oscillation Index (SOI) from the Bureau of Meteorology of the Australian Government (http://www.bom. gov.au/climate/current/soi2.shtml) (access on 8 May 2021). The effect of the 11-year solar cycle on total ozone was investigated by analyzing the monthly sunspot number series from the World Data Center/Sunspot Index and Long-term Solar Observations (WDC/SILSO) of the Royal Observatory of Belgium, Brussels (http://sidc.be/silso/datafiles) (access on 8 May 2021). The monthly North Atlantic Oscillation (NAO) index was provided from the Climate Data Guide of NCAR at https://climatedataguide.ucar.edu/climate-data/hurrellnorth-atlantic-oscillation-nao-index-pc-based (access on 8 May 2021).
Total ozone variability is also related to variability related to tropopause height, e.g., [20][21][22]. The impact of tropopause height variations on total ozone variability was examined by analyzing the tropopause pressure from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis 1 data set computed on a 2.5 • grid. The NCEP/NCAR reanalysis data were downloaded from the website https://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis. tropopause.html (access on 30 April 2021) [23].
The mean annual ozone cycle was calculated for the period 2004-2018, and then the ozone time series were deseasonalized by subtracting the long-term monthly mean (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) pertaining to the same calendar month; i.e., monthly value-long-term monthly mean. Next, the deseasonalized data were used in a multivariate linear regression (MLR) model to describe influences of dynamic origin on total ozone variability. The MLR statistical model includes the QBO, SOLAR, ENSO, NAO, and trend terms, as described by Zerefos et al. [24] and later adopted by Eleftheratos et al. [25], for further analyses. Those studies, however, had a slightly different approach, as they also included the effects of aerosol optical depth (AOD) and Antarctic oscillation. Those studies examined stations in the northern and southern mid-latitudes, which justified the inclusion of the Antarctic oscillation in the MLR model. Due to the fact that our station is located in the northern and not in the southern mid-latitudes, we did not include the Antarctic oscillation proxy here. The AOD proxy was used by Zerefos et al. to account for the volcanic injections of El Chichon (1982) and Mt Pinatubo (1991) into the stratosphere, which caused large stratospheric disturbances, increasing ozone depletion. However, the AOD proxy has not been considered in this study, since the mentioned volcanic eruptions occurred in the past and should not affect the period of our analysis. The same procedure, i.e., deseasonalization and MLR analysis, was also applied to the tropopause pressure data, in order to estimate the tropopause pressure residuals (not shown here). Then, the residuals of tropopause pressure from the MLR analysis were correlated with the respective residuals of ozone, in order to determine the effect of tropopause height variations on total ozone variations. The correlation coefficient between the ozone and tropopause pressure residuals was R = +0.448 (t-value = 6.533, p < 0.0001, N = 172). The correlation is presented in Section 3.3.

Daily Values and Correlations
The daily column ozone measurements made by the Brewer spectrophotometer at the Academy of Athens from July 2003 to July 2019 are presented in Figure 1. The respective ozone columns retrieved by TOMS, OMI, GOME-2A, and GOME-2B satellite instruments agree fairly well with the ground-based Brewer measurements. The satellite overpass data were selected to be within a 100 km radius from the Brewer site. The daily values span between 250 DU and 500 DU; in full agreement with Tzanis [26], who compared daily column ozone observations from the Dobson spectrophotometer with SCIAMACHY, TOMS, and OMI satellite data. A good agreement between satellite data and a Brewer spectrophotometer has been demonstrated in other studies, for instance in Kim et al. [27], who used a Brewer spectrophotometer to evaluate the quality of the total ozone column (TOC) produced by multiple polar-orbit satellite measurements at three stations in Antarctica. As a result of their study, high correlations between the TOC from the Brewer and the TOC from TROPOMI and OMI measurements were observed, contrary to the correlations from AIRS measurements. The study confirmed the high quality of OMI TOCs.  . Accordingly, we provide the mean biases and RMSE between the four satellite data pairs of OMI, TOMS, GOME-2A, and GOME-2B, as follows: −5.3 DU and 7.8 DU (OMI vs. GOME-2A), −4.8 DU and 7.6 DU (OMI vs. GOME-2B), −2.7 DU and 7.7 DU (OMI vs. TOMS), and −1.5 DU and 5.2 DU (GOME-2A vs. GOME-2B). All R were tested for significance using the t-test formula for the correlation coefficient with n−2 de grees of freedom [28] and were found to be statistically significant at a confidence leve greater than 99%. More detailed correlation statistics between the various data pairs are provided in Table 1. It is evident that all correlation coefficients pass the significance leve (p-values < 0.0001). We include here for the reader the statistical test of the correlation coefficient, which is: Table 1. Statistics of correlations between the Brewer and satellite ozone data pairs.  ). All R were tested for significance using the t-test formula for the correlation coefficient with n − 2 degrees of freedom [28] and were found to be statistically significant at a confidence level greater than 99%. More detailed correlation statistics between the various data pairs are provided in Table 1. It is evident that all correlation coefficients pass the significance level (p-values < 0.0001). We include here for the reader the statistical test of the correlation coefficient, which is:

Monthly Means and Annual Cycle
The monthly mean total ozone time series were computed from at least 14 daily averages and are shown in Figure 3 for the Brewer ground-based data in comparison to OMI, TOMS, GOME-2A, and GOME-2B satellite data. The monthly mean values range between 270 DU and 400 DU; again in agreement with results from Tzanis [18]. The long term mean ±2σ of total ozone over Athens is estimated to be 322 ± 53 DU, with no significant change since 2003. The respective estimates from OMI, TOMS, GOME-2A, and GOME-2B satellite data are 318 ± 51 DU, 316 ± 46 DU, 324 ± 53 DU, and 325 ± 51 DU, accordingly.

Monthly Means and Annual Cycle
The monthly mean total ozone time series were computed from at least 14 daily averages and are shown in Figure 3 for the Brewer ground-based data in comparison to OMI, TOMS, GOME-2A, and GOME-2B satellite data. The monthly mean values range between 270 DU and 400 DU; again in agreement with results from Tzanis [18]. The long term mean ±2σ of total ozone over Athens is estimated to be 322 ± 53 DU, with no significant change since 2003. The respective estimates from OMI, TOMS, GOME-2A, and GOME-2B satellite data are 318 ± 51 DU, 316 ± 46 DU, 324 ± 53 DU, and 325 ± 51 DU, accordingly.  The highest values occurred in spring in March and April, while the lowest values occurred in autumn in October and November. This is a general and consistent feature seen in all three datasets. The explanation for the observed seasonal cycle is transport mechanisms. The spring maximums are a result of the increased transport of ozone from its source region in the tropics toward high latitudes during late autumn and winter. This poleward ozone transport is much weaker during the summer and early autumn periods and is weaker overall in the Southern Hemisphere [2]. Ozone transport from the tropics to the poles is caused by stratospheric wind patterns. In the mid-latitudes these patterns, known as the Brewer-Dobson circulation, make the ozone layer thickest in the spring and thinnest in the fall. Table 2 summarizes the monthly mean differences between the Brewer, OMI, and GOME-2A total ozone data. The Brewer-OMI differences are within ±1% in all months except June, July, and August, where they are within ±2%, but even these are considered small. We note here that a difference of 1% corresponds to about 3 DU. Differences larger  Figure 4 shows the seasonal cycle of total ozone over Athens for the period 2004-2018 from Brewer ground-based measurements and OMI and GOME-2A satellite retrievals. The highest values occurred in spring in March and April, while the lowest values occurred in autumn in October and November. This is a general and consistent feature seen in all three datasets. The explanation for the observed seasonal cycle is transport mechanisms. The spring maximums are a result of the increased transport of ozone from its source region in the tropics toward high latitudes during late autumn and winter. This poleward ozone transport is much weaker during the summer and early autumn periods and is weaker overall in the Southern Hemisphere [2]. Ozone transport from the tropics to the poles is caused by stratospheric wind patterns. In the mid-latitudes these patterns, known as the Brewer-Dobson circulation, make the ozone layer thickest in the spring and thinnest in the fall.
Oxygen 2021, 1, FOR PEER REVIEW than ±2% are found between Brewer and GOME-2A in the winter months (Novemb December, January, and February). Similar deviations, larger than 2%, are also found b tween GOME-2A and OMI satellite data.  With regard to the observed differences, we must keep in mind that the Brewer strument is operating at ground level, while the satellite instruments are measuring fro space using different retrieval algorithms than the ground based instrument. The Brew instrument is measuring continuously the ozone amount overhead, while the satellite struments provide few measurements during the day, sometimes one or two measu ments. In addition, the Brewer is a remote sensing instrument, while for the satellite da  Table 2 summarizes the monthly mean differences between the Brewer, OMI, and GOME-2A total ozone data. The Brewer-OMI differences are within ±1% in all months except June, July, and August, where they are within ±2%, but even these are considered small. We note here that a difference of 1% corresponds to about 3 DU. Differences larger than ±2% are found between Brewer and GOME-2A in the winter months (November, December, January, and February). Similar deviations, larger than 2%, are also found between GOME-2A and OMI satellite data. Table 2. Mean differences between Brewer and satellite total ozone data (1% ∼ = 3 DU).

Brewer-OMI
Brewer-GOME-2A GOME-2A-OMI With regard to the observed differences, we must keep in mind that the Brewer instrument is operating at ground level, while the satellite instruments are measuring from space using different retrieval algorithms than the ground based instrument. The Brewer instrument is measuring continuously the ozone amount overhead, while the satellite instruments provide few measurements during the day, sometimes one or two measurements. In addition, the Brewer is a remote sensing instrument, while for the satellite data, we processed measurements within a 100 km radius from the Brewer site. The aforementioned issues are known to cause differences between ground measurements and satellite overpasses. However, despite the different approaches of the ground and satellite instruments, the average long-term differences between the ground and satellite measurements are small, within ±1%, indicating the maturity of the measuring systems in achieving such small deviations in the long-term.

Ozone Variability
We estimated the contribution of different explanatory variables to ozone fluctuations using MLR analysis, as explained in Section 2. The MLR statistical model is of the following form: where i denotes the month and j is the year of the deseasonalized total ozone column (desTOC) and its components; that is, the QBO at 30 and 50 hPa, the ENSO, the NAO, the solar cycle effect (SOLAR), a straight line to fit the long term trend (TREND), and finally a tropopause pressure related term (TROP). We remind here that TOC data were deseasonalized by subtracting the long-term monthly mean (2004-2018) pertaining to the same calendar month. The contribution of the individual proxy terms is shown in Figure 5.
The MLR analysis was applied to the deseasonalized ozone data, which are shown on the top panel of Figure 5 (black line). The two terms representing the QBO are shown by the lines with blue colors, followed by the ENSO term (red line), the NAO term (green line), the solar cycle term (orange line), and the trend term (brown line). The bottom panel of Figure 5 shows the residuals of ozone from the MLR model (grey color), together with the respective residuals of tropopause pressure from an MLR analysis that had been applied to the tropopause data in a previous step (magenta color). The ozone residuals are well correlated with the tropopause pressure residuals, indicating the dynamical influence on ozone induced by the tropopause movement. The graph shows that whenever the tropopause pressure decreases, i.e., tropopause height increases, the amount of ozone increases, and vice versa. We estimate that the correlation coefficient between ozone and tropopause height variations in Athens is +0.448 (slope = 0.376, error = 0.058, t-value = 6.533, p < 0.0001, N = 172). The MLR regression coefficients and their standard errors are presented in Table 3. It appears that the regression coefficient of the QBO50 proxy is significant at the 90% confidence level (coefficient = 0.111, error = 0.059, t-value = 1.874, p-value = 0.063). The regression coefficient of the solar proxy is more significant than the QBO50 proxy (coefficient = −0.070, error = 0.018, t-value = −3.853, p-value = 0.00017). The regression coefficient of the trend proxy is not statistically significant (coefficient = 0.003, error = 0.013, t-value = 0.249, p-value = 0.803). Finally, the regression coefficient of the tropopause proxy is 0.378 ± 0.059 (t-value = 6.459, p-value < 0.0001).  Oxygen 2021, 1, FOR PEER REVIEW 11   The contribution of all components to ozone fluctuations cumulatively is presented in Figure 7, which shows the observed versus the regressed ozone data. As can be seen, there is good agreement between the observed ozone data and the statistical model calculations obtained from Equation (7). The correlation coefficient between the observed and regressed ozone data is estimated as R = +0.941. The residuals (observed minus regressed data) are shown in the bottom panel. The contribution of all components to ozone fluctuations cumulatively is presented in Figure 7, which shows the observed versus the regressed ozone data. As can be seen, there is good agreement between the observed ozone data and the statistical model calculations obtained from Equation (7). The correlation coefficient between the observed and regressed ozone data is estimated as R = +0.941. The residuals (observed minus regressed data) are shown in the bottom panel.

Conclusions
We analyzed 16 years of total ozone measurements over Athens, Greece, with a Brewer spectrophotometer. The main findings can be summarized as follows:

•
There are strong correlations between total ozone from the Brewer spectrophotometer and total ozone from the OMI, TOMS, GOME-2A and GOME-2B satellite instruments greater than 0.9.

•
The main contribution to ozone variability comes from the seasonal cycle. We estimate that the seasonal variability explains about 64% of the variability in total ozone over Athens.

Conclusions
We analyzed 16 years of total ozone measurements over Athens, Greece, with a Brewer spectrophotometer. The main findings can be summarized as follows:

•
There are strong correlations between total ozone from the Brewer spectrophotometer and total ozone from the OMI, TOMS, GOME-2A and GOME-2B satellite instruments greater than 0.9.

•
The main contribution to ozone variability comes from the seasonal cycle. We estimate that the seasonal variability explains about 64% of the variability in total ozone over Athens.