Height resolved ozone hole structure as observed by the Global Ozone Monitoring Experiment–2



[1] We present Global Ozone Monitoring Experiment-2 (GOME-2) ozone profiles that were operationally retrieved with the KNMI Ozone ProfilE Retrieval Algorithm (OPERA) algorithm for the period September–December 2008. It is shown that it is possible to accurately measure the vertical distribution of stratospheric ozone for Antarctic ozone hole conditions from spectra measured at ultraviolet wavelengths from a nadir viewing instrument. Comparisons with ozone sonde observations from the Neumayer station at the Antarctic coast show a good agreement for various ozone profile shapes representing different phases of the annual recurring ozone hole cycle. A preliminary analysis of the three-dimensional structure of the ozone hole shows for example that at the vortex edges ozone rich mid-latitude middle and upper stratospheric layers can be found over ozone depleted lower stratospheric ‘ozone hole’ layers. These Antarctic ozone profile observations combined with the daily global coverage of GOME-2 enables the monitoring of the three-dimensional structure of the ozone hole on a daily basis.

1. Introduction

[2] The launch of the Global Ozone Monitoring Experiment (GOME) on board of the European Remote Sensing 2 satellite in April 1995 started a new era of measuring earth-reflected solar radiation with relatively high spectral resolution for UltraViolet and VISible wavelengths (UV-VIS). A number of studies have shown that it is possible to retrieve ozone profiles from these UV-VIS spectral observations [e.g., Chance et al., 1997; Munro et al., 1998]. Over the years, various research groups have developed ozone profile retrieval algorithms for spectral UV-VIS measurements, and identified and improved upon various errors due to calibration issues, retrieval methodology and parameter uncertainties [e.g., Hoogen et al., 1999; Hasekamp and Landgraf, 2001; van der A et al., 2002; Liu et al., 2005]. Meijer et al. [2006] provide an evaluation of nine different GOME ozone profile retrieval algorithms available at the time. They concluded that stratospheric ozone profiles can be determined quite well, but that accurately measuring tropospheric ozone remains a challenge due to inadequate instrument calibration and the weak signal for tropospheric ozone information. These retrievals are expected to improve considerably for instruments like the Ozone Monitoring Instrument (OMI) and Global Ozone Monitoring Experiment-2 (GOME-2) due to smaller ground pixels, a better characterization of the instrument spectral response (slit) function and an improved polarization correction, which were major limitations for GOME [van der A et al., 2002; Schutgens and Stammes, 2003]. Liu et al. [2005] showed nevertheless that GOME ozone retrievals can provide realistic and valuable tropospheric ozone information for tropical and mid-latitude locations.

[3] A common feature of UV-VIS ozone profiles is their limited vertical resolution, estimated typically at 7–15 km [e.g., Hoogen et al., 1999; Liu et al., 2005], with a considerable amount of vertical smoothing. As a consequence, small scale vertical ozone features cannot be observed. This has important implications for both tropospheric and lower stratospheric ozone observations, which frequently show large ozone variations with a small vertical extent and sharp ozone gradients which cannot be resolved in UV-VIS ozone profiles. In particular, the Antarctic ozone hole provides a challenge as the thickness of ozone depleted layers in the ozone hole is typically about 5 km - the ozone hole is located between 12 to 20 km altitude - and sharp ozone changes occur at the upper and lower edges of the ozone hole. An additional problem for Antarctic ozone profile retrievals is the difficulty in discriminating snow/ice surfaces and middle/high level clouds, which can cause large retrieval errors in the retrieved ozone profiles. Therefore, until now little emphasis has been put on assessing the quality of Arctic and especially Antarctic UV-VIS ozone profiles [Meijer et al., 2006; de Clercq and Lambert, 2007]. With the launch of various instruments with improved instrumentation and spatial resolution since 2002 (SCanning Imaging Absorption spectroMeter for Atmospheric CartograpHY, OMI, GOME-2), it is expected that the retrievals from these instruments will result in improved knowledge about the vertical distribution for the Antarctic ozone hole.

[4] The GOME-2 instrument [Callies et al., 2000], launched on 19 October 2006 onboard Metop-A, flies in a sun-synchronous polar orbit with an equator crossing time of 09:30 hrs (local solar time). GOME-2 is a nadir looking cross-track scanning spectrometer. The instrument measures backscattered solar light from the Earth's atmosphere between 240–790 nm in four channels with a relatively high spectral resolution (0.2–0.4 nm). In its normal mode, the instrument has an almost global daily coverage with a cross-track swath width of 1920 km which is split up in ground pixels with a horizontal resolution of 80 × 40 km between 307–790 nm but with 640 × 40 km below 307 nm due to much weaker signals. The measurements from GOME-2 are especially suitable for retrieval of the total column and vertical profiles of atmospheric ozone (which is one of the essential climate variables [Mason, 2003]), and other key atmospheric trace gases, such as NO2, SO2, BrO and formaldehyde. GOME-2 has been measuring ozone profiles from January 2007 onwards, providing daily coverage of the Antarctic ozone hole area during the entire 2008 ozone hole season. The operational retrieval algorithm for GOME-2 ozone profiles is the Ozone ProfilE Retrieval Algorithm (OPERA) algorithm [van der A et al., 2002; van Oss and de Haan, 2004; B. Mijling et al., “Preparing for GOME-2 ozone profile retrievals: improving the profile retrieval algorithm using GOME data,” manuscript in preparation, 2009], which will be described in the next section.

[5] We explore one season of GOME-2 ozone profiles under Antarctic ozone hole conditions to investigate the current status and quality of GOME-2 ozone profile observations, and demonstrate the effect of improvements that have been made to the OPERA algorithm over the last couple of years. We investigate how well the ozone profile retrievals from GOME-2 UV-VIS measurements can capture the vertical distribution of ozone during Antarctic ozone hole conditions and discuss the use of these observations for monitoring the 3-D structure of the Antarctic ozone hole on a daily basis.


[6] Ozone profile retrieval algorithms from UV spectra use the fact that the absorption cross section of ozone decreases steeply with wavelengths between 270–340 nm. Scattered sunlight detected by the satellite at short wavelengths experiences strong absorption by ozone and therefore has only traveled the top layers of the atmosphere: it only carries information on the ozone distribution in these layers. With increasing wavelengths, photons also carry ozone information from lower layers. Ozone absorption structures in the Huggins bands are temperature dependent, providing additional information for tropospheric ozone retrievals [Chance et al., 1997]. Above 340 nm the spectrum is more transparent and used to extract information on surface reflection and cloud parameters. Retrieving ozone profiles from this information is an underconstrained inverse problem: there are more parameters describing the profile than there are independent pieces of information available in the spectrum.

[7] OPERA is an algorithm that solves this problem using the optimal estimation method [e.g., Rodgers, 2000], which includes a priori information to stabilize the inversion. The state vector consists of the atmospheric parameters that are fitted, in this case 40 layers of ozone and the albedo. The albedo can be the cloud albedo or the surface albedo; when the cloud fraction is larger than 0.20, the cloud albedo is used. Optimal estimation requires a priori information (including error covariance) for the state vector elements and the derivatives of the measurement to the state. The a priori information is taken from the ozone climatology of McPeters et al. [2007]. The derivatives of the measurement to the state vector (i.e., weighting functions) are calculated with the radiative transfer model LidortA [van Oss and Spurr, 2002]. Since this model does not include polarisation, the derivatives are corrected afterwards, for the neglect of polarisation, using pre-calculated look-up tables. The radiative transfer model uses the cross sections from Daumont et al. [1992], Malicet et al. [1995], and Brion et al. [1998] and meteorological information from ECMWF. The optimal estimation method is applied iteratively until convergence is reached. The convergence criteria for the retrieval are based on the magnitude of the state update and the deviation between measured and simulated radiances. The retrieval of ozone profiles is currently done using ground pixels of 640 × 40 km.

3. Intercomparison of Ozone Profiles

[8] For validation of the GOME-2 ozone profiles we use data from the ozone balloon sounding program conducted at the Antarctic research station Neumayer (8.26°W and 70.65°S) which has been operational since 1992, with ozone profiles obtained weekly. During the development of the ozone hole, the sounding frequency is increased to three times a week. All measurements were made using an ECC 6A type ozone sonde. These measurements are available at the World Ozone and Ultraviolet Radiation Data Centre database (http://www.woudc.org/index_e.html).

[9] The Neumayer station was selected since it is located near the edge of the Antarctic polar vortex and here we expect to observe some interesting variability in the ozone concentration associated with the vortex dynamics. In addition, the Neumayer station is the only Antarctic station that makes its measurements publicly available within a few days after observation.

[10] We only compare ozone sonde data and satellite data when the following three criteria have been fulfilled. First, the Neumayer Station should be located inside the satellite footprint. Second, within a footprint, the distance between Neumayer and the pixel centre should not exceed 300 km, the typical length scale of lower stratospheric ozone variations [Sparling et al., 2006]. Third, the launch of the sonde and the overpass of the satellite should be within 12 hours of each other. In total 37 collocations were found for the period September–December 2008, some of which are collocations of one sonde with multiple GOME-2 overpasses.

[11] The sonde profile is convolved with the averaging kernel (A, defined as the sensitivity of the retrieval equation image to the true state x) and the a priori profile (xa) of the satellite observation, according to the equation equation image = xa + A(xxa) + ε [e.g., Rodgers, 2000, chapter 3], where ε is the measurement error. Replacing x by the sonde observation, equation image gives us the retrieved sonde profile, which is smoothed in the same way as if it would have been observed by GOME-2, and can therefore be compared to the actual retrieved ozone profile. Both original and convolved sondes are used in the intercomparison with the GOME-2 profiles (see Figure 1).

Figure 1.

Ozone profiles (number density n in molecules cm−3) for the Neumayer Station using original, non-gridded data. The black line is the GOME-2 profile, the solid red line is the sonde profile convolved with the a priori profile and the averaging kernel (see section 3), the dotted red line is the original sonde profile and the blue line is the a priori profile. Above the sonde burst level (indicated by the horizontal dotted line) the sonde profile is set equal to the a priori. The date and solar zenith angle (SZA) are indicated above each plot. The bottom right plot shows the mean difference (GOME-2 - sonde) of all collocations in the period September–December 2008 (black line) and the 1σ error (grey area).

[12] With the collocation criteria described above it is possible for multiple collocations to occur on the same day if the GOME-2 pixels overlap. We therefore have gridded the data to 1° × 1° using an area weighted averaging. The gridded dataset is used in constructing plots of time series (see Figure 2) and cross-cutting views of the atmosphere (see Figure 3).

Figure 2.

Time series of the ozone concentration over the Neumayer Station. Arrows along the top indicate the locations of the profiles in Figure 1. Grey areas indicate missing data. Concentrations are given in 1012 molecules cm−3.

Figure 3.

(top) North–South and (middle) East–West cross sections of the atmosphere (in molecules cm−3) for 13 October 2008. Grey areas indicate missing data. In the top plot we show the ozone cross section for constant longitude (8.26°W) with the South Pole to the left and the North Pole to the right. The middle plot shows the cross section for constant latitude (70.65°S) with 180°W to the left and 180°E to the right. In both plots, the location of Neumayer is indicated by the arrows and the thermal tropopause (based on the climatology) has been indicated by a white line for reference. (bottom) GOME-2 total ozone columns integrated from the ozone profile retrievals (in DU) for (left) the Southern Hemisphere and (right) the Northern Hemisphere and the exact location of the cross sections.

4. Results

[13] In Figure 1 we show a comparison between sonde and satellite profiles representing various stages in the ozone hole life cycle with very different ozone profile shapes. As an indication of the quality of the retrievals we include the mean and standard deviation of the differences between all retrieved GOME-2 profiles (in the period September–December 2008) and the convolved sonde profiles. We see that the ozone profiles compare very well, even when the a priori information does not resemble ozone hole conditions. Note that the retrieved profiles are also in good agreement with the sonde profiles at high solar zenith angles.

[14] The results of the profile comparison give confidence in the retrieval algorithm. Therefore, we made a time series of the gridded dataset over the Neumayer Station. It is plotted in Figure 2, starting at 1 September 2008 and continuing up to 31 December 2008. The time series shows the development and breakup of the ozone hole. The double peaked ozone profile that was visible in the collocations in Figure 1 is also visible in the time series. The region of maximum ozone depletion is located between 100 and 50 hPa between 16 September and 25 October 2008 and between 100 and 70 hPa afterwards. Around 26 October, air masses with elevated ozone concentrations start to appear between 50 and 20 hPa altitude and a few days later the concentrations suddenly drop by approximately a factor of two. This behavior recurs a couple of times in the rest of the time series. The air masses with elevated ozone concentration slowly descend during the latter half of the ozone hole period. This downward movement replenishes part of the ozone depleted air in the Antarctic ozone hole, eventually leading to its disappearance [Sato et al., 2009]. The variability in ozone concentrations can be explained by the dynamics of the ozone hole. Neumayer is located near the edge of Antarctica and usually also close to the edge of the ozone hole. The ozone hole is, at any given moment, generally not circular symmetric, and due to Rossby wave propagation the edges of the ozone hole slowly rotate around Antarctica. As a result, the Neumayer station is located either inside or outside the Antarctic vortex depending on the shape and rotation of the ozone hole edge.

[15] Note the persistence of the region with low ozone concentrations between 100 and 70 hPa after approximately 29 October 2008. Clearly the ozone depleted air from inside the vortex is flowing under ozone rich air at higher altitudes. This interesting dynamical feature might be obscured by the traditional ozone hole definition, which is based on the total ozone column. The ozone rich air compensates the ozone depleted air in the total column, thereby giving the impression that the ozone hole is smaller than it actually is.

[16] The maximum ozone depletion occurred around the beginning of October. For 13 October 2008, the middle left plot of Figure 1 gives the collocation of GOME-2 with a sonde launched from the Neumayer Station. To get a better impression of the global behavior of the ozone concentration around the time of maximum ozone depletion, we made North–South and East–West cross sections of the atmosphere (see Figure 3), centered on Neumayer.

[17] In the North–South cross section, the ozone hole can be seen very clearly at latitudes poleward of 60°S. The minimum ozone concentration is reached around 70 hPa with slightly higher concentrations above and below it. The East–West cross section shows a region of high ozone concentration from 90°E crossing the date line to 80°W. This coincides with a region of increased total ozone columns (that were integrated from the ozone profile retrievals) as can be seen in the bottom left plot. Below this ozone rich air, there is the region of maximum depletion, interrupted by a couple of ozone rich intrusions from above.

5. Discussion and Conclusions

[18] The first GOME-2 ozone profiles for Antarctic ozone hole conditions are presented. The algorithm OPERA is capable of retrieving the ozone profile in good agreement with ozone sonde measurements in this region. The time series shows that we are capable of monitoring the three dimensional structure of the ozone hole on a day-to-day basis. One interesting observation is the presence of ozone depleted laminae below the ozone maximum at the edge of the Antarctic vortex, possibly caused by a combination of lower stratospheric vortex dynamics and middle and upper stratospheric transport of ozone rich air from mid-latitudes. Since the ozone hole definition is based on total ozone columns, this might affect the derived size of the ozone hole. Future research will investigate in more detail how the standard ozone hole definition relates to an ozone hole area based on the GOME-2 ozone profiles.

[19] One of the Essential Climate Variables (ECV) that is defined in the Global Climate Observing System programme [Mason, 2003] on atmospheric composition is ozone. We have demonstrated that we are able to monitor the ozone concentration in 4D, that is in both space and time. The retrieved profiles can be used in an assimilation process, thereby satisfying the requirement of global coverage for the ECVs.

[20] Since the GOME-2 instrument series will continue flying until about 2020, the expected recovery of the ozone hole can be monitored. In addition, the daily three dimensional coverage of the ozone hole opens new exciting possibilities to study Antarctic vortex dynamics, transport processes and evaluating climate models that include stratospheric chemistry.


[21] The authors want to thank EUMETSAT for providing the GOME-2 data and the World Ozone and Ultraviolet Radiation Data Centre data archive (http://www.woudc.org/index_e.html) for providing access to the ozone sonde data.