An assessment of changing ozone loss rates at South Pole: Twenty five years of ozonesonde measurements

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116,, doi: /2011jd016353, 2011 An assessment of changing ozone loss rates at South Pole: Twenty five years of ozonesonde measurements B. Hassler, 1,2 J. S. Daniel, 2 B. J. Johnson, 3 S. Solomon, 2,4 and S. J. Oltmans 3 Received 2 June 2011; revised 19 September 2011; accepted 21 September 2011; published 18 November [1] In 2010, 25 years of regular, year round ozone soundings at South Pole station, Antarctica, were completed. These measurements provide unique information about the seasonality, trends, and variability of ozone depletion in the polar stratosphere at high vertical resolution. Here, we focus on the observed loss rates, and their changes since the measurement series began. The fastest loss rates occur between the end of August and end of September between 50 hpa and 30 hpa. Loss rates at these pressure levels increased by approximately 40% from the late 1980s to the late 1990s and have remained stable within estimated uncertainties since then. To estimate the time frame when a reduction in ozone loss rates will be observable outside the range of dynamical variability at the South Pole, we scale the estimated loss rates to the future projected concentrations of equivalent effective stratospheric chlorine (EESC). If a linear relationship between ozone loss rates and EESC is assumed, we project that a change in lower stratospheric ozone loss rates at South Pole station will be first detectable in the time period. Citation: Hassler, B., J. S. Daniel, B. J. Johnson, S. Solomon, and S. J. Oltmans (2011), An assessment of changing ozone loss rates at South Pole: Twenty five years of ozonesonde measurements, J. Geophys. Res., 116,, doi: /2011jd Introduction [2] The year 2010 marked the 25th anniversary of regular, year round ozone soundings at South Pole station, Antarctica. This time series is among the longest records of vertically resolved ozone measurements and provides information for one of the most remote places on Earth, in the heart of the ozone hole region. It offers a unique basis for monitoring the evolution of ozone depletion and is expected to play an important role in identifying future ozone layer recovery. [3] Here we provide an update to previous South Pole ozone sounding analyses [Hofmann et al., 1997; Solomon et al., 2005; Hofmann et al., 2009]. We have analyzed the temporal development of ozone loss rates as a function of altitude. Our analyses further our understanding that chemical effects strongly dominate the observed decreases in Antarctic ozone in the key season of late August/September, when most of the loss occurs. Our study also finds that lateral mixing is necessary within the polar vortex as discussed by Sato et al. [2009]. 1 Cooperative Institute for Research in Environmental Sciences, University of Colorado at Boulder, Boulder, Colorado, USA. 2 Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA. 3 Global Monitoring Division, NOAA Earth System Research Laboratory, Boulder, Colorado, USA. 4 Department of Atmospheric and Oceanic Science, University of Colorado at Boulder, Boulder, Colorado, USA. Copyright 2011 by the American Geophysical Union /11/2011JD [4] We update the time series of South Pole ozonesonde measurements to reflect the 25 year record and compare this time series with the combined Georg Forster and Neumayer time series (located at approximately 70 S) in Section 2. We then explore differences in the onset of ozone loss and loss rates at those stations with simple box model calculations (Section 3). In Section 4 we calculate ozone loss rates for different pressure levels over five different time periods, and the differences in loss rates are discussed. To obtain an estimate of the time when future ozone loss rates will be lower than the peak loss rates by an amount large enough to be observable outside the range of dynamical variability when averaged over five years, we analyze the observed variability and project future chemistry loss rates using EESC (Section 5). A summary and discussion follow in Section Twenty Five Years of Ozone Soundings at South Pole: Overview [5] Two and a half decades of ozone soundings from South Pole station (latitude: S, longitude: W) have been obtained from an updated version of the Binary Database of Profiles (BDBP) [Hassler et al., 2008]. Only soundings reaching pressure levels of at least 30 hpa, and being made by identifiable ozonesondes (e.g., Regener sondes, Electrochemical sondes (ECC)) are included in that database. The earliest measurements go back to the mid 1960s and early 1970s, but the number of soundings from this period is small, on average about 15 to 16 soundings per year. After a gap of 15 years, regular ozone soundings have been made at South Pole station from 1986 through the 1of12

2 present. In the last 25 years, about 60 soundings per year are available, allowing for detailed analyses of stratospheric ozone. South Pole ozone soundings are submitted regularly to the databases from the World Ozone and Ultraviolet Radiation Data Centre (WOUDC, index_e.html) and the Network for the Detection of Atmospheric Composition Change (NDACC, ncep.noaa.gov/). Both databases are used as sources for the BDBP. [6] In addition, ozone soundings from Georg Forster station (latitude: S, longitude: E) and Neumayer station (latitude: S, longitude: 8.25 W) are analyzed here for comparing to South Pole station data. Data are available from 1985 to 1992 for Georg Forster station (from Brewer GRD sondes), and from 1992 to 2010 for Neumayer station (from ECC sondes). Sounding frequency for Georg Forster station is less than for South Pole with about 32 soundings per year, whereas the yearly average number of soundings available for Neumayer station is comparable to South Pole (66 soundings per year). [7] The accuracy and precision of electrochemical concentration cell (ECC) ozonesondes, which are used in this analysis, is reduced in humid conditions at very low ozone values when the sensor current approaches the measured background current. Vömel and Diaz [2010] demonstrated that ozone measurements in the tropical upper troposphere were sometimes underestimated because the background current of the ozonesonde cell was overestimated. Humid tropical conditions can affect the filters used to supply ozone free air during the sonde preparation leading to a slightly higher background. However, the cold and dry conditions at South Pole and the use of a dry purified air source optimizes conditions for a low, stable sensor background current. Within the heart of the maximum ozone loss layer the sensor readings can drop, nonetheless, below the limit of detection of 0.02 ppmv at 50 hpa ambient pressure. But errors in measuring these trace levels will not affect the overall analysis presented in this paper. [8] Figure 1 provides an update to Figure 7 of Solomon et al. [2005] with six years ( ) of additional data added. It shows individual ozone measurements, given in ozone mixing ratio, near 50 hpa at South Pole (Figure 1, top) and for the combined Georg Forster and Neumayer time series (Figure 1, bottom; from here on called Georg Forster/Neumayer ) as a function of day number within a year. Since measurements are stored in the BDBP on predefined pressure levels (to facilitate comparisons between different measurements), the closest available database level to 50 hpa (i.e., hpa) is displayed rather than measurements taken at 50 hpa directly. Four different time periods are displayed if data are available (1960 to 1979, 1980 to 1989, 1990 to 1999 and 2000 to 2009), to show the impact of changing concentrations of man made ozonedepleting substances (ODSs) in the atmosphere on the seasonal evolution of depletion. [9] As discussed by Solomon et al. [2005], this graph shows the well known seasonal changes of ozone concentrations in the Antarctic stratosphere, in particular the rapid ozone decrease in spring caused by heterogeneous chemical reactions [e.g., Farman et al., 1985]. Over the decades shown in Figure 1 the destruction became more and more severe. At South Pole, up to 99% of the ozone at 50 hpa is destroyed in some years within the span of a bit more than a month. This occurred not just in the 1990s, but also in the 2000s. Ozone destruction seen at Georg Forster/ Neumayer station is not as severe at this pressure level, but it starts earlier (see Figure 1, right). This earlier start is due to the earlier exposure to sunlight at Georg Forster/ Neumayer at its lower latitude position [see Lee et al., 2000; Kuttippurath et al., 2010]. [10] Figure 2 shows decadally averaged ozone mixing ratios at South Pole and Georg Forster/Neumayer stations. A10 day sliding averaging window was applied to calculate the mean of all available measurements within each 10 day period at the selected pressure level for the respective decade. The earlier onset of ozone destruction at the lowerlatitude station Georg Forster/Neumayer appears consistently throughout the 1980s, the 1990s, and the 2000s. In addition, the difference in minimum ozone mixing ratios for the two stations around day 270 to 280 (end of September to beginning of October) can be seen. With an earlier onset and less total ozone destruction, the average ozone loss rates experienced at Georg Forster/Neumayer station are smaller than those obtained at South Pole. Differences in the minimum ozone levels between the two stations are significant at the 92% level for the 1990s, and at the 82% level for the 2000s. The difference between the minima in the 1980s are not statistically significant because of the small number of available independent data points for this decade. The geographical location of the ozone measurements is not the only factor important for the ozone losses at each site; the histories of the air masses sampled at the two stations are also important, as will be explained in more detail in the next section. 3. Modeling Antarctic Ozone Destruction [11] We have used a box model to probe the seasonal development of ozone losses within the Antarctic polar vortex. This model uses the same chemical mechanism and solving approach of Solomon et al. [1990] and uses updated kinetic parameters and photolysis cross sections based upon the JPL recommendations from 2010 [Sander et al., 2010]. Our goal is to qualitatively probe processes that contribute to ozone depletion rather than fully quantify them. It is assumed that total Cl y and Br y are 3.5 ppbv and 22 pptv, respectively, and that all of the available chlorine and bromine is activated during the winter season before the return of sunlight for the purpose of these illustrative calculations. These Cl y and Br y concentrations are chosen because they represent approximate represent peak halogen loadings in the stratosphere [World Meteorological Organization (WMO), 2011]. Deactivation of chlorine and bromine is not considered. Thus the calculation illustrates only the upper limit to photochemical ozone loss rates under the assumption of complete activation of available chlorine and bromine. Transport terms are also not considered. [12] The key photochemical reactions considered in the model are: ClO þ ClO þ M! Cl 2 O 2 þ M Cl 2 O 2 þ M! 2ClO þ M ð1þ ð2þ 2of12

3 Figure 1. Ozone mixing ratio at 50 hpa at (top) South Pole station and (bottom) Georg Forster/ Neumayer station (lower panel) for four time periods: , , , The gray shaded areas in the graphs of the left column are expanded in the right column. Note that measurements for the first period are only available for some years at South Pole, and no data are available then at Georg Forster/Neumayer station. Cl 2 O 2 þ h! Cl þ ClO 2 BrO þ ClO! OClO þ Br BrO þ ClO! BrCl þ O 2 BrO þ ClO! Br þ ClO 2 ClO 2! Cl þ O 2 Full spherical geometry is considered for the photolysis rates. Key to this study is the photolysis rate for Cl 2 O 2 and its dependence on solar zenith angle in polar twilight (at large angles). A comparison with detailed calculations of Cl 2 O 2 photolysis rates shown from Figure 5 of Anderson et al. [1995] shows very good agreement (as can be seen in Figure S1 of the auxiliary material). 1 The box model 1 Auxiliary materials are available in the HTML. doi: / 2011JD ð3þ ð4þ ð5þ ð6þ ð7þ analysis provides qualitative estimates of the importance of air mass excursions to lower latitudes for the timing of the onset of springtime ozone depletion. It is not the intention to provide a comprehensive evaluation of the latest findings in polar heterogeneous chemistry cross section research (for more detail about these see WMO [2011] and Sander et al. [2011]), but to strengthen findings described later in the manuscript. [13] Figure 3 shows the model results compared with the measured ozone mixing ratios from South Pole ozone soundings between the years 2000 and 2010 (without 2002) as a function of day number. It shows upper limits to the ozone losses that could be expected for heterogeneously processed air in the region of 80 S and 90 S if the polar vortex were zonally symmetric, i.e., all geographical longitudes for a specific latitude value see the sun coming back in spring at exactly the same day, allowing photochemistry to begin only at that time. The actual position of the Antarctic vortex is not generally centered above the South Pole, and the true position changes over the course of the vortex s existence. The test case for the ozone evolution at 80 S is intended to illustrate the dependence of ozone losses on the timing of the return to sunlit conditions. The differences 3of12

4 Figure 2. Mean ozone mixing ratios near 50 hpa for the four different time periods shown in Figure 1. A 10 day sliding averaging window was applied to all available data for the respective time periods to obtain the mean values. South Pole data is shown as solid lines, Georg Forster/Neumayer station is shown as dotted lines. obtained between the simple model runs for 80 S versus 90 S show that earlier exposure to sunlight is critical to an earlier start of the reactions destroying ozone (e.g., ClO dimer reactions, shown in equations (1) (3) and (7)). However, the calculated onset of ozone destruction for zonally symmetric conditions at neither 90 S or 80 S agrees with the lower ranges of South Pole observations, with calculated depletion starting much too late in spring; thus, more photochemical exposure is required to match the data. [14] Figure 3 shows an additional illustrative test case in which the position of air parcels were at 80 S for 4 days out of 5, but experienced excursions to 65 S for 1 day in 5. This case illustrates that vortex displacements (even if limited in duration) would allow a substantially earlier onset of ozone depletion that would agree with some South Pole observations more closely than the zonally symmetric cases we considered. The accurate simulation of the onset and slope of ozone destruction at South Pole may require consideration of both local air masses with complex photochemical histories and mixing of air masses from lower latitudes exposed to sunlight earlier. Isaksen et al. [1990] used a wave like structure for the airflow in their model to simulate the effects of heterogeneous chemistry on ozone depletion. They showed that strong ozone depletion exists for days after air masses had passed through areas with PSCs into sunlight regions. Several other studies showed that within the well mixed Antarctic vortex core, ozonedepleted air (and activated chlorine) from lower latitude regions with early season exposure to sunlight is transported quickly poleward [Sanders et al., 1993; Roscoe et al., 1997; Lee et al., 2001]. Such processing could explain why ozone loss at South Pole can be observed even before the return of sunlight to the stratosphere above South Pole. Similarly, Tilmes et al. [2006b] showed with backward trajectories that air parcels within the vortex core (poleward of 70 S equivalent latitude) in spring 2003 saw sunlight before the sun was actually above the horizon at those latitudes. These processes may also explain a high variability in ozone concentrations within the same latitude band. Besides the geographical location of an observing site within the vortex, the air mass history is clearly very important for explaining the timing of the onset of ozone depletion. The effects of vortex displacement and mixing on the timing of the onset of ozone depletion are expected to be smaller at Georg Forster/Neumayer station, because of its greater exposure to sunlight, than at South Pole station. 4. Loss Rates Derived From South Pole Ozone Soundings [15] During Southern Hemisphere spring, the frequency of soundings at South Pole station is approximately every three days. This would be enough data to accurately estimate loss rates for each individual year if chemistry were the only factor in determining ozone loss. However, as pointed out by Hofmann et al. [1997, 2009] ozone values measured at South Pole are influenced by the prevailing phase of the quasi biennial oscillation (QBO) in the tropical stratosphere. Because of the variability caused by the QBO and other stratospheric processes (e.g., differences in sunlight exposure due to vortex displacements or ozone concentration changes due to diabatic descent), for this study, five year blocks of measurements were analyzed. In determining the Antarctic ozone loss rate representative for a 5 year period, the impact of year to year variability is reduced. Further, Sato et al. [2009] and Sonkaew et al. [2011] showed that the diabatic descent rate is small compared to the rate of Antarctic photochemical ozone destruction in August and 4of12

5 Figure 3. Individual ozone measurements (red crosses) from 2000 to 2010 (excluding 2002) averaged over 65 to 75 hpa as a function of day number within a year. The green (blue) line shows the modeled ozone destruction at 90 S (80 S) if a zonally symmetric vortex is prescribed. The black line shows the modeled ozone destruction at 80 S if air parcels experience illumination conditions characteristic of 65 S for one day out of five (see text). September. The impact of the dynamical variability that remains after the 5 year averaging is reflected in our uncertainty analysis. [16] For the determination of the ozone loss rate on one selected pressure level at South Pole station, all available measurements at that level between a defined start and end date were considered. We chose nominal start and end dates of day 235 and day 270, respectively, since from around day 230 to about day 270 ozone mixing ratios decrease nearly linearly. Hoppel et al. [2005a] and Sonkaew et al. [2011] showed that ozone loss rates are small from July to early August, and increase quickly thereafter as soon as the sun comes back. The exact onset of the rapid ozone depletion is somewhat variable from year to year so that linear behavior cannot be assumed each year before roughly day 230 (see, for example, Figure 4, 33 hpa, days 220 to 235). After approximately day 270 the ozone decline ceases, either because most of the available ozone is already destroyed (saturation) or because the activated chlorine compounds are deactivated again. A linear description of the temporal evolution of ozone mixing ratios is not applicable after saturation occurs (see, for example, Figure 4, 67 hpa, day 265 to 280, measurements from 2006 to 2010; Figure 4, 33 hpa, day 265 to 280, measurements from 1986 to 1990). In addition, the onset of ozone destruction and the date when the minimum ozone concentrations are reached has changed over the decades, as appears in Figure 2. Thus, the choice of the nominal day limits used in this study, 235 as a start day and 270 as an end day, are chosen in order to be applicable throughout the analyzed time periods. The linear fit was performed on five blocks of data of 31 consecutive days. The first block spanned days 235 to 266, and each subsequent block was then shifted later by one day. The greatest loss rate determined from these five blocks was then used in the analysis. However, it should be noted that ozone loss saturation sometimes occurs before day 270 on some pressure levels (e.g., Figure 4, 67 hpa, measurements for 2006 to 2010). Therefore, a threshold of minimum ozone was chosen (0.1 ppmv) over which the fit was performed. If that threshold was reached twice before the end day of a 31 day fitting period, measurements taken after the day where the threshold was reached for the second time were not considered. With this cut off, a linear fit can still be used to describe the ozone decrease in cases in which ozone loss saturation occurs before day 270. [17] Figure 4 presents ozone changes for three different pressure levels and two different five year periods: and Significant decreases in ozone can be detected at all levels of the two time periods. While the loss rate for the lowest level (119 hpa) changes only slightly within the covered 25 years of observations, the change in loss rates for the two higher levels (67 hpa and 33 hpa) is substantial (up to 40%). For example, during the late 1980s no measurement showed ozone mixing ratios below 0.1 ppmv at 67 hpa, however, in the late 2000s there were many measurements below that threshold by day 270. At 33 hpa and 119 hpa, saturation also nearly occurred in the later time period. However, some ozone remained after rapid chemical ozone destruction ended. 5of12

6 Ozone mixing ratio [ppmv] Ozone mixing ratio [ppmv] Ozone mixing ratio [ppmv] Date 6 Aug 23 Aug 10 Sep 28 Sep 15 Oct LR: ± ppmv/day LR: ± ppmv/day 33 hpa Day of year Date 6 Aug 23 Aug 10 Sep 28 Sep 15 Oct LR: ± ppmv/day LR: ± ppmv/day 67 hpa Day of year Date 6 Aug 23 Aug 10 Sep 28 Sep 15 Oct LR: ± ppmv/day LR: ± ppmv/day 119 hpa Day of year [18] Profiles of the estimated ozone loss rates at South Pole and their associated 1 s fitting uncertainties are shown for five five year periods in Figure 5a. On some pressure levels the defined threshold of 0.1 ppmv ozone was reached before day 270, as mentioned above. Figure 5b shows loss rate profiles on potential temperature levels, determined with the same method as already described. As can be seen, the magnitude of the loss rates for the different regions of the atmosphere and the different time periods are very similar whether observed as a function of pressure or potential temperature; this supports the assumption that diabatic descent for the analyzed 35 days is small enough to be disregarded. [19] The vertical profile of the ozone loss in Figure 5a shows the same structure in all five 5 year periods analyzed, with loss rates increasing from around 150 hpa up to 50 hpa, and decreasing from around 40 hpa up to 20 hpa, as noted by Hofmann et al. [1997, 2009]. The data of the late 1980s show significantly lower loss rates than the other four profiles from about 100 hpa to about 20 hpa, due largely to the lower chlorine and bromine concentrations during that period (so that chemical ozone loss rates were therefore smaller). Loss rates for the period 1991 to 1995 are high on the topside of the profile near 50 to 20 hpa compared to the other profiles (Note that these loss rates fall within the combined range of the respective error bars). The eruption of Mount Pinatubo in June 1991 increased sulfate aerosol concentrations in the stratosphere globally, enhancing heterogeneous chemistry in polar regions, and thereby increasing ozone depletion [Portmann et al., 1996]. Although in 1992 enhanced aerosol surface area density was observable up to potential temperature level of 650 K ( 40 to 30 hpa) [Thomason and Poole, 1993], enhanced ozone depletion was found to be confined to the lower stratosphere [Hofmann et al., 1997; Solomon et al., 2005] rather than near 50 to 20 hpa. Kinnison et al. [1994] and Rosenfield et al. [1997] described additional dynamical effects to the chemical changes after the eruption of Mount Pinatubo. During the years following the eruption enhanced upwelling in the tropics was simulated due to perturbation in heating rates, with resulting stronger downwelling in high latitude areas. This might have decreased the loss rates during those years. On the other hand, some authors have argued that transport of enhanced NO x could increase ozone losses at these levels [e.g., Konopka et al., 2007; Vogel et al., 2008]. [20] The maximum loss rates for the last three 5 year periods are very similar for all pressure levels up to 30 hpa. Note that measurements for 2002 were excluded from the analysis of the period 2001 to 2005 because of the unusual ozone development due to the Antarctic vortex split in Figure 4. Ozone measurements at South Pole at three different pressure levels and during two different 5 year periods, 1986 to 1990 (black circles) and 2006 to 2010 (blue triangles), as a function of day of the year. A linear fit (black and blue solid lines) was performed for measurements taken between day 235 and day 270 of each time period (red dashed lines). For all three shown pressure levels, the differences in slopes are significant. The minimum ozone threshold for the fitting procedure was set to 0.1 ppmv (dark gray line and gray box). 6of12

7 Figure 5. (a) Profile of loss rates for five time periods ( , , , without 2002, ) at South Pole, as determined by a linear fit to all available data for each pressure level between day 235 and day 270. Loss rates are given in [ppmv/day]. Error bars represent 1 s uncertainties. (b) Profile of loss rates for the same five time periods at South Pole as a function of potential temperature. (c) Profile of maximum ozone loss for the same five time periods at South Pole, given in [%]. Black dash dotted lines in both graphs denote the pressure levels shown in Figure 4. spring [Hoppel et al., 2003]. At the higher altitudes, loss rates for the period 1996 to 2000 exceed the loss rates of the most recent two 5 year periods slightly, however, the 1 s error bars for all three most recent periods still overlap. Loss rates for the period from 2001 to 2005 and 2006 to 2010 show slightly lower loss rates again, still within the range of the 1 s error bars, on the topside of the profile although the chlorine concentrations during the last 15 years did not change more than 4.5%. [21] Figure 5c shows profiles of maximum observed ozone loss for the same five 5 year periods. Error bars shown represent the 1 s uncertainties on these losses. A time series of mean ozone mixing ratios for each pressure level and each of the 5 year periods was calculated with a 10 day sliding averaging window as described above for Figure 2. Measurements obtained for the year 2002 were again excluded. From each of these time series the minimum ozone mixing ratio was determined between day 210 and day 290. The mean mixing ratio between day 160 and 200 (40 days) was calculated to represent ozone concentrations before the onset of depletion. This value and the minimum value were then used to calculate the percentage ozone loss. Day 210 and day 290 were chosen as search boundaries for the minimum value, instead of day 235 and day 270 as for the determination of the loss rates for Figure 5a, to capture not only the part of the fast ozone depletion that can be described by a linear fit, but also the slower ongoing depletion that can occur after late September (day 270) and before middle of August (day 235). Since the ozone mixing ratio time series for each pressure level describes the mean of five years, and the applied 10 day sliding averaging window and the 40 day mean value smoothes some of the short term variability, the actual maximum percentage ozone loss in some years is larger than that shown. [22] Figure 5c shows that the shapes of the five profiles of the percentage ozone loss at South Pole are qualitatively similar in each of the 5 periods analyzed. More than 90% of ozone is destroyed in the pressure region from around 100 hpa to about 40 hpa in the 1990s and 2000s. Lower and higher pressure levels experienced less severe ozone loss, as reported earlier by Solomon et al. [2005] and Hofmann et al. [2009]. Losses for the period 1986 to 1990 are smaller than for the other four periods, but nonetheless display a decrease of up to 86% in the pressure region with most ozone loss. Kawa et al. [2009] pointed out that Antarctic ozone loss was relatively consistent during the late 1990s and 2000s, with the exception of the years 2002 and They found that ozone mixing ratios in the lower stratosphere within the polar vortex get close to zero each year by mid to late September. The findings presented in Figure 5c agree well with this, as the maximum ozone loss profiles for the periods , and are very similar and demonstrate peak ozone losses of greater than 95% in the lower stratosphere. [23] Comparing the maximum ozone losses for the three pressure levels shown in Figure 4 reveals that although the highest loss rates are detected near 33 hpa, the highest total percentage ozone loss occurs near 67 hpa. Figure 6 (left) shows mean profiles of ozone mixing ratios for several days throughout late winter and spring for the period 2006 to All profiles taken within two days of the indicated date were averaged to produce these mean profiles. It can be 7of12

8 Figure 6. (left) Mean ozone profiles at South Pole for the time period 2006 to Different line styles denote six dates during Southern Hemisphere winter and early spring. Dates in red show the time period for which ozone loss was described by a linear fit. Measurements taken on those dates ±2 days were averaged. Black horizontal lines represent the three pressure levels shown in Figure 4. (right) Differences in ozone mixing ratio between day 235 and day 270 (red lines in Figure 6, left) for the time period 2006 to seen that ozone mixing ratios at 33 hpa are significantly larger than those at 67 hpa before depletion begins. Although minimum ozone mixing ratios at 33 hpa are not as low as at 67 hpa, the overall loss in the 35 day period is slightly greater at 33 hpa, which results in higher loss rates at this pressure level (see Figure 6, right). [24] The linear fit loss rates derived in this study and in studies by Hofmann et al. [1997, 2009] agree very well when compared on the same ozone and altitude unit scales, even though slightly different time periods were used in their study and individual years were analyzed as opposed to five year averages. Comparisons of our loss rates with other studies using different ozone data and methods are more difficult. Sonkaew et al. [2011], for example, determined the day to day change in ozone mixing ratios for several isentropic levels by taking an average of all available observations from the satellite instrument SCIAMACHY from inside the polar vortex for several individual years. They fit a straight line to these changes from mid August to mid November, covering a much broader seasonal range than considered here. Their loss rates (around 0.03 ppmv/day as an average for the analyzed isentropic levels) are significantly smaller than loss rates between 77 hpa and 25 hpa presented in our study. Their time period for the straight line fit sometimes includes the time after ozone loss saturation, and the satellite instrument s technique might not be able to measure the lowest local ozone values, both leading to lower (in magnitude) estimated loss rates. A similar method to the one described by Sonkaew et al. [2011] was also used by Godin et al. [2001]. [25] The Match technique, sampling the same air parcel several times in a prescribed interval using trajectory calculations [WMO, 2007], has also been used to determine chemical loss rates. In these studies, the ozone loss is usually expressed in units of mixing ratio per sunlit hour [e.g., Hoppel et al., 2005a; Frieler et al., 2006; Tripathi et al., 2007]. We are not able to provide a useful comparison to any Match results, because, in this study, we relied on data from a single station, South Pole. 5. When Will Loss Rates at South Pole Be Measurably Lower in the Future? [26] As seen in Figure 5a the loss rates for the last fifteen years have not significantly changed at South Pole. This is not surprising since equivalent effective stratospheric chlorine (EESC) concentrations, a good indicator for chemical ozone depletion in the Antarctic stratosphere, have been near the expected peak during those years, varying by less than 4.5%. Figure 7 shows an EESC time series with a mean age of air of 5.5 years and an age distribution representative for polar conditions (the width of the age distribution assumed to be 2.75 years [see Newman et al., 2007]). Colored boxes in Figure 7 represent the range of EESC present during the respective 5 year periods (indicated at the top of the graph). Thick lines within those boxes depict the mean EESC concentration for the respective five years. [27] The results of Daniel et al. [2010] suggest a linear relationship between EESC and globally averaged total column ozone depletion (see, e.g., Figure 2 of that study). Such a relationship has not yet been verified to hold for polar ozone depletion, and in this study, there is not data over a wide enough range in EESC values to determine whether a linear or higher order relationship best describes the dependence of ozone loss on EESC (see auxiliary material Figure S2). If loss rates are assumed to depend 8of12

9 Figure 7. Equivalent effective stratospheric chlorine (EESC) concentration time series adopting a mean age of air of 5.5 years (polar conditions). Overlaid are the mean EESC values for the five 5 year periods analyzed, color coded as shown at the top of the graph. Each box represents the highest and lowest EESC concentration during the respective period; the thick line within the box represents the mean. The black box shows the 5 year period for which the loss rates are significantly lower than peak loss rate at 89 hpa, considering the uncertainties on the loss rate. See text for more details. linearly on EESC concentrations, it is possible to evaluate this dependence by fitting a straight line to the loss rates on each pressure level as a function of EESC concentration. In doing so, any deviations of loss rates from the determined straight line are ascribed to dynamical effects on ozone concentrations, or chemical influences other than linear EESC effects including those arising from variations in stratospheric aerosols [e.g., Portmann et al., 1996]. [28] The linear dependence of ozone loss rates on EESC can be described by: LR ¼ a þ b EESC where LR is the loss rate in ppmv/day, EESC is the EESC concentration in ppbv, and a and b are the fitting parameters. With these fitting parameters it is then possible to calculate an ozone loss rate for every given EESC concentration. Considering the exceptional nature of the ozone loss rates of the period (see Section 4, Figure 5a, and auxiliary material Figure S2), the loss rates of this period were excluded from the fitting procedure. [29] Weatherhead and Andersen [2006] stated that the leveling off of global ozone depletion during the early 2000s is generally consistent with decreases in ozone depleting substances. Assuming that the future dynamical variability of the Antarctic stratosphere will be comparable to the variability of the last 20 years, and that there will not be a major volcanic eruption, the loss rates in August September, before reaching ozone loss saturation, should decrease with decreasing EESC concentrations. With the indicated relationship between ozone loss rates at South Pole and the EESC concentrations, it should be possible to estimate when future loss rates should be expected to become significantly lower than the loss rates consistent with peak EESC levels. ð8þ [30] In order to estimate when a reduction in future ozone loss rates will be detectable outside the range of dynamical variability when 5 year averages are used, profiles representing loss rates at maximum EESC concentrations can be compared to loss rates expected in the future as EESC levels decline. Once EESC declines far enough from the peak, the future loss rate profile will become significantly lower. The difference is expected to be observable in a limited altitude range at first, and then as EESC continues to decline, it will be seen over a larger portion of the loss rate profile. To identify when this will occur, the uncertainty in the determination of the individual ozone loss rates and the variations in loss rates of the past five year intervals need to be considered. [31] We first define the peak loss rate profile to which we compare our future projected loss rate profile to evaluate when a statistically significant decline has occurred. By using the determined fitting parameters a and b for each pressure level (equation (8)), and the highest EESC concentrations of the past 5 year periods (period ; see Figure 7), we can calculate a peak loss rate profile to which future loss profiles can be compared. Using the 5 year averaging period for estimating peak EESC makes this a conservative choice since it leads to peak loss rate profile with slightly lower magnitudes than if EESC levels for individual years were used. This, in turn, implies that the future loss rate profile has to be slightly lower (and therefore a few years later). [32] Future mean EESC concentrations were estimated by sliding a 5 year averaging window year by year. The determined fitting parameters a and b for each pressure level and the future mean EESC concentrations were then used to determine the projected loss rates for each respective 5 year period. With this approach, projected loss rates decline with EESC concentrations in the future. The 9of12

10 Figure 8. Estimated 5 year periods (red bars) in which differences between the peak loss rate profile and the future loss rate profile become significant. uncertainties associated with the loss rate of each pressure level of both profiles, peak loss rate profile and future loss rate profile, are identical. They were calculated to represent the mean fitting uncertainties in the loss rate estimates at each pressure level of each 5 year period, excluding the early 1990s period (the period that was also excluded from the straight line fit relating loss rates to EESC). Specifically, at each pressure level, the mean uncertainty is calculated as the square root of the squared sums of error bars shown in Figure 5a. [33] Finally, the projected loss rates and the minimum peak loss rate profile are compared, separately for each pressure level, for each future 5 year period. We use a onetailed test of significance to determine whether the future and the minimum peak loss rates are significantly different at the 90% confidence level. Using this approach, the first significant difference in loss rates is detected at the 89 hpa pressure level for mean EESC concentrations of 3563 pptv, reached in the 5 year period from January 2017 to December 2021, and denoted by the black box in Figure 7. By sliding the 5 year window further into the future differences on more and more pressure levels become significant, as can be seen in Figure 8. By the period January 2026 to December 2030, loss rates of all pressure levels between around 100 hpa and 20 hpa are projected to be significantly reduced. Loss rates on pressure levels above or below the mentioned range take at least 20 years longer to show significant differences. In the lower part of the profile, this is due to differences between peak loss rates and future loss rates being so small that changes in EESC do not convert into large changes in loss rates (see Figure 9). In the upper part of the profile, the size of the error bars on the determined loss rates is large so that even with large changes in EESC the differences in peak and the future loss rate profiles are not significant (see Figure 9). [34] Newman et al. [2006] concluded in their study that a statistically significant detection of a recovery of the Antarctic ozone hole would earliest occur in about Although they use the reduction in ozone hole area as a measure for ozone hole recovery rather than changes in ozone loss rates as employed here, their detection of recovery occurs at about the same time as our estimated observable loss rate changes. Vyushin et al. [2007] analyzed total column ozone data derived from ground based and satellite instruments to probe uncertainties in trend detection. They found that a statistically significant ozone increase attributable to EESC decreases is likely to be detectable late in the decade 2010 to 2020, similar to this study and that of Newman et al. [2006]. [35] Hofmann et al. [1997] and Hoppel et al. [2005b] suggest that the upper edge of the ozone hole (22 24 km) is expected to show reductions in total ozone loss first. However, our results suggest that significantly lower loss rates will first be detectable in the lower stratosphere (around 89 hpa). As mentioned before, our result does not only depend on the relationship between loss rate and EESC, but strongly depends on the differences in loss rate uncertainties in different pressure regions. Figure 5a shows that the uncertainties in loss rates are relatively small between roughly 77 hpa and 45 hpa, due to similar loss rates in the four time periods used to determine the mean uncertainty. Higher up in the stratosphere there is greater variability among the loss rate uncertainties of the four periods, which results in a larger mean uncertainty. Thus, a reduction of loss rates in the middle stratosphere is expected to take longer to become significantly different from the peak loss rates than in the lower stratosphere. Additionally, the timing of reaching a significant difference in loss rates is dependent on the chosen significance level of the differences (here: 90%). If a higher level of significance were chosen, the time when significantly different loss rates are observable would be delayed relative to the results presented in Figure 8. [36] Our approach to projecting future loss rates is based on the assumption that future dynamical conditions are similar to those present during the last 20 years. If changes Figure 9. Loss rate profile calculated for the 5 year mean EESC concentration of the EESC peak period (gray line, Peak loss rate profile ). The black profile represents the loss rate profile for , the 5 year period for which loss rates at one pressure level are significantly lower than the peak loss rates considering the variability of both profiles ( Future loss rate profile ). The red circle indicates the pressure levels for which the loss rates were different at the 90% confidence level. The blue profile illustrates the loss rates for the period of 12

11 in dynamics occur in the future, loss rates would change differently than determined here. 6. Summary and Discussion [37] In 2010 the 25th year of regular ozone soundings at South Pole, Antarctica, was completed, marking a quarter of a century of indispensable and important monitoring activities of the stratospheric ozone layer. The study presented here provides an update and expansion of earlier South Pole ozonesonde studies [Hofmann et al., 1997; Solomon et al., 2005; Hofmann et al., 2009] and furthers the discussion of ozone loss rates observed at South Pole [Hofmann et al., 1997]. [38] In our study loss rates were determined over 5 year periods to reduce the effects of year to year changes in dynamics, and therefore obtain a clearer signal of chemical effects. Nevertheless, no significant reduction in Antarctic ozone loss was detected in the first decade of the 21st century at South Pole station, which is consistent with previous studies [Newman et al., 2006; Weatherhead and Andersen, 2006; Yang et al., 2008]. In a recent study, Salby et al. [2011] tried to remove the dynamical variability to improve the ability to see the influence of the slowly increasing halogen levels on polar ozone. They applied their method to the Antarctic ozone average over September October November, which includes not only the period of maximum ozone loss rates but also spans part of the resupply of ozone as the vortex breaks down. This is a fundamentally different approach from what was presented here. [39] Comparisons between ozone loss at South Pole and Georg Forster/Neumayer station show clear differences in spring ozone depletion: the onset of the rapid ozone loss starts earlier at Georg Forster/Neumayer station and the ozone loss is less severe, compared with South Pole observations. Experiments with a simple box model support previous studies showing that the earlier onset of ozone loss at lower latitude stations like Georg Forster/Neumayer is initiated by earlier exposure to sunlight. Further, the prominent temporal ozone loss development in spring at South Pole requires a history of some sunlight exposure earlier than the exposure for the station s geographical latitude. [40] With increasing EESC concentrations, loss rates increased over the last 25 years at the South Pole. However, after ozone loss saturation was reached on some pressure levels of the ozone profile, and because EESC concentrations changed slowly near the peak in early 2000 [Newman et al., 2007], loss rates at those levels exhibited little change during the last 15 years. [41] EESC scaled loss rate profiles were used to estimate the earliest time when loss rates can be expected to be measurably lower than at present. Considering the uncertainties in the estimated loss rates (including influences of stratospheric dynamics and measurement uncertainties, etc.) and further assuming that future stratospheric dynamics will not deviate from past conditions and that no major volcanic eruptions occur, South Pole ozone loss rates are predicted to begin to be significantly lower than they were at their peak during the period 2017 to [42] With increasing concentrations of greenhouse gases, however, it is possible that the assumption of unchanged future Antarctic stratospheric dynamics is not fully valid. As pointed out by Weatherhead and Andersen [2006], there is still considerable uncertainty about the rate of ozone recovery and future ozone levels as greenhouse gases abundances continue to rise. Shindell and Schmidt [2004] showed that both increasing greenhouse gas (GHG) concentrations and decreasing ozone concentrations have an impact on the Southern Hemisphere circulation. With a slow decrease in EESC, spring ozone concentrations are expected to recover. However, at the same time increasing GHG concentrations will cool the lower stratosphere and can lead to enhanced formation of polar stratospheric clouds (PSCs) [Hitchcock et al., 2009]. This could keep ozone loss rates high although EESC concentrations decrease. Tilmes et al. [2006a] pointed out in their study that future changes in Antarctic dynamics could change the volume of the Antarctic vortex, and therefore could reduce the potential to form PSCs. This could consequently lead to changes in chemical ozone loss. [43] The 25 years of ozone soundings at South Pole have provided an immensely valuable and vital contribution to measuring components of the Antarctic stratosphere and monitoring their changes. The soundings cover a remote region where other stratospheric ozone measurements are very limited, and are hence a widely used data set for stratospheric ozone studies, climate analyses, and model comparisons. This data set will continue to be of very high value in coming decades to help evaluate ozone recovery and the effects of the interplay between increasing GHGs and increasing ozone concentrations. [44] Acknowledgments. Funding for B. Hassler has been provided by the NOAA NCDC Climate Data Record Program. We thank three anonymous reviewers for helpful comments and suggestions. Understanding of ozone loss rates in Antarctica using the South Pole ozonesonde data series was a primary research interest and effort of David Hofmann. From his own early ozonesonde observations in Antarctica to his leadership of the laboratory responsible for the South Pole ozonesonde measurements, he understood the great value of continuous, ongoing observations. This work was initiated with Dave s involvement prior to his death. References Anderson,D.E.,R.DeMajistre,S.A.Lloyd,andP.K.Swaminathan (1995), Impact of aerosols and clouds on the troposphere and stratosphere radiation field with application to twilight photochemistry at 20 km, J. Geophys. Res., 100(D4), , doi: /95jd Daniel, J. S., E. L. Fleming, R. W. Portmann, G. J. M. Velders, C. H. Jackman, and A. R. Ravishankara (2010), Options to accelerate ozone recovery: Ozone and climate benefits, Atmos. Chem. Phys., 10, , doi: /acp Farman, J. C., B. G. Gardiner, and J. D. Shanklin (1985), Large losses of total ozone in Antarctica reveal seasonal ClO x /NO x interactions, Nature, 315, , doi: /315207a0. Frieler, K., M. Rex, R. J. Salawitch, T. Canty, M. Streibel, R. M. Stimpfle, K. Pfeilsticker, M. Dorf, D. K. Weisenstein, and S. Godin Beekmann (2006), Toward a better quantitative understanding of polar stratospheric ozone loss, Geophys. Res. Lett., 33, L10812, doi: / 2005GL Godin, S., V. Bergeret, S. Bekki, C. David, and G. Megie (2001), Study of the interannual ozone loss and the permeability of the Antarctic polar vortex front aerosol and ozone lidar measurements in Dumont d Urville (66.4 S, 140 E), J. Geophys. Res., 106(D1), , doi: / 2000JD Hassler, B., G. E. Bodeker, and M. Dameris (2008), Technical note: A new global database of trace gases and aerosols from multiple sources of high vertical resolution measurements, Atmos. Chem. Phys., 8, , doi: /acp of 12