Ozone data from the solar occultation Polar Ozone and Aerosol Measurement (POAM) III instrument are included in the ozone assimilation system at NASA's Global Modeling and Assimilation Office, which uses Solar Backscatter UltraViolet/2 (SBUV/2) instrument data. Even though POAM data are available at only one latitude in the southern hemisphere on each day, their assimilation leads to more realistic ozone distribution throughout the Antarctic region, especially inside the polar vortex. Impacts of POAM data were evaluated by individual and statistical comparisons of assimilated ozone profiles with independent ozone sondes. Major improvements in ozone representation are seen in the Antarctic lower stratosphere during austral winter and spring in 1998. Limitations of assimilation of sparse occultation data are illustrated by an example.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Antarctic ozone loss is largely confined to a distinct lower stratospheric layer within a polar vortex because of the unique meteorology of this region [e.g., World Meteorological Organization (WMO), 2003; Solomon, 1999, and references therein]. Polar processes influence the ozone distribution outside the polar regions. After the breakup of the polar vortex, ozone-poor polar air dilutes ozone-richer air in mid-latitudes causing seasonal changes in mid-latitudes. Transport of vortex-processed air, which contains active chlorine radicals, also contributes to ozone loss at mid-latitudes. The amount and interannual variability of mid-latitudinal ozone loss is affected by polar processing and in particular by the strength and stability of the polar vortex [e.g., Millard et al., 2002; Hadjinicolaou et al., 1997].
 Nadir viewing satellite instruments measuring ultraviolet or infrared radiation have provided multi-decadal records of global ozone data that have been available in near-real time. Thus, many assimilation systems use ozone data from nadir instruments for real time estimation and forecasting or in multi-year analyses of historical data [e.g., Stajner et al., 2004]. However, the usability of data from nadir instruments for quantifying the vertical distribution of the polar ozone loss is limited by their coarse vertical resolution and, for solar backscatter ultraviolet instruments, by their inability to make measurements in the polar night.
 Limb viewing satellite instruments provide ozone profiles with a better vertical resolution. Measurements of microwave or infrared radiances provide daytime and nighttime coverage, typically over a wide range of latitudes. Consequently they have the potential to improve the quality of assimilated stratospheric ozone globally, and especially in the polar regions [e.g., Levelt et al., 1998; Struthers et al., 2002].
 Solar occultation satellite instruments provide sparse atmospheric profile measurements with even better vertical resolution. The Polar Ozone and Aerosol Measurement (POAM) III instrument provides polar ozone profiles from solar occultation measurements [Lucke et al., 1999]. The assimilation of occultation data is hampered by their sparseness. This was illustrated in the assimilation of Halogen Occultation Experiment (HALOE) data. Ménard et al.  assimilated HALOE methane data into a two-dimensional transport model on a single isentropic surface in the stratosphere. Using a Kalman-filtering approach, they found that oversimplified error covariance models led to unreliable results. Chipperfield et al.  assimilated HALOE data for four species with long photochemical lifetimes into a three-dimensional chemistry and transport model (CTM), also including a scheme to conserve tracer-tracer correlations. The assimilation had an impact on the global distribution of constituents. However, the largest reduction in the analysis errors, and the improvement in the agreement with independent data were seen close to the locations of HALOE measurements.
 A question arises: Can assimilation of the sparse POAM III data capture the three-dimensional stratospheric ozone evolution over the entire Antarctic? In this study we assimilate POAM III ozone data and evaluate their impact on the Antarctic ozone.
2. Assimilation System
 The ozone assimilation system used in this study is based on that of Stajner et al. . Satellite ozone data are assimilated into a CTM with a parameterization of the gas phase stratospheric chemistry. Data from the NOAA 14 Solar Backscatter UltraViolet/2 (SBUV/2) instrument constrain ozone in the sunlit stratosphere. Assimilated ozone agrees within 10% with HALOE data between 70 and 0.2 hPa [Stajner et al., 2001]. The resolution and accuracy of SBUV/2 data degrades in the lower stratosphere. At high solar zenith angles, near the polar night region, the SBUV/2 data are often not assimilated because the retrieval algorithm flags them as lower quality data. Comparisons with sonde profiles (below) show that the representation of lower stratospheric ozone profiles at high latitudes is inadequate in the SBUV-only assimilation during winter and spring. We investigate if additional constraint from POAM data provides assimilated ozone fields that are adequate for studies of polar ozone processes.
3. POAM III Data
 The POAM II and POAM III instruments have provided ozone data from solar occultation measurements for several years (1993–1996 and from 1998 to present). This feasibility study focuses on the austral winter and spring in 1998. From a sun-synchronous satellite orbit, POAM III makes measurements of 14–15 sunsets and sunrises per day, 25.4° longitude apart, on two latitude circles [Lucke et al., 1999]. The northern circle changes from 54°N to 71°N, and the southern from 63°S to 88°S during the course of the year. Routine measurements are made down to the mid-troposphere in cloud free conditions. The vertical resolution of version-3 ozone data is about 1.1 km and they have random errors smaller than 5% above 15 km [Lumpe et al., 2002]. Comparison of POAM III with other satellite and ozone sonde data showed agreement to within 5% from 13 to 60 km [Randall et al., 2003]. POAM and other occultation data were used successfully to construct polar ozone maps on isentropic surfaces [Allen and Nakamura, 2003]. Randall et al.  also successfully use POAM data to reconstruct polar ozone fields. Thus the POAM data provide high quality information about polar ozone distributions.
 We assimilated version-3 POAM ozone data, starting on June 1, 1998, between 14 and 60 km altitude. This altitude range was motivated by the ease of implementation of the observation operator for our model, which has hybrid levels between the surface and 161 hPa, and pressure levels above. All the chosen POAM levels lie in the pressure level part of our model grid. The random error estimates provided in the POAM data files vary from ∼3% near 30 km to over 20% at altitudes below 20 km, especially in the presence of polar stratospheric clouds and low ozone values. A simple model for POAM errors was used in the assimilation: uncorrelated errors of 5%. This choice was made to illustrate the impact of POAM data in the existing assimilation system. In the assimilation experiments no change was made to the treatment of the SBUV data or forecast errors [Stajner et al., 2004]. In particular, forecast error correlations are modeled as by Stajner et al. , with an increased length scale L = 500 km.
4. Evolution of the Lower Stratospheric Ozone Over Antarctica
 The evolution of ozone in the polar vortex is illustrated by instantaneous maps from the POAM and SBUV assimilation at 70 hPa (Figures 1a, 1c, and 1e). In wintertime, before heterogeneous ozone loss begins, slow descent leads to an accumulation of ozone in the lower stratosphere within the polar vortex [e.g., WMO, 2003, chapter 3]. Thus, higher ozone values are seen over Antarctica than in the middle latitudes (Figure 1a). In springtime the sun starts to illuminate the air mass within the polar vortex. Activated chlorine and bromine compounds serve as catalysts for rapid ozone loss, which begins near the polar vortex edge (Figure 1c). A distinction between the weakly mixed ring of air near the Antarctic vortex edge and the well-mixed vortex core was pointed out by Lee et al. . Distortions of the vortex push parts of the vortex to lower latitudes where sunlight is stronger. As sunlight advances towards the pole almost complete ozone depletion is seen throughout the vortex (Figure 1e).
 The ozone fields from the SBUV-only assimilation fail to capture the main aspects of the ozone evolution over Antarctica: wintertime accumulation within the vortex (Figure 1b), early springtime progression of the loss from the vortex edge towards the inner core (Figure 1d), and the severity of the depletion later in the spring (Figure 1f). Figure 1 shows that there is a substantial impact of POAM data throughout the vortex, not just near the measurement locations, which are marked by black circles in Figure 1. Note that no heterogeneous chemical processes are included in the CTM, so the success of the assimilation depends on using high-quality data.
 A wintertime profile of independent Neumayer ozone sonde (near 70°S, 8°W) on July 8, 1998 is compared with assimilated ozone in Figure 2a. At this time POAM measurements are near 66°S. Assimilation of NOAA 14 SBUV/2 data does not capture the ozone accumulation in the lower stratosphere. When POAM data are assimilated the profile shape is changed: ozone amounts at pressures higher than 40 hPa increase, and the ozone amounts for pressures lower than 40 hPa decrease. Including POAM data leads to substantially better agreement with the Neumayer sonde, especially in the lower and middle stratosphere.
 An example of a partial ozone reduction in early spring is seen in the independent South Pole ozone sonde profile on September 22, 1998 (Figure 2b). Two maxima of ozone partial pressure near 15 and 150 hPa bracket the region in which a varying degree of loss is seen. There are two ozone minima at 40 and 100 hPa. The POAM assimilation reproduces this stratospheric profile shape. At this time POAM measurements are made near 88°S. The SBUV assimilation gives a profile with an improperly placed ozone maximum at 30 hPa, and strongly underestimates the maxima at 50 and 150 hPa.
 Nearly complete ozone destruction is seen in the South Pole ozone sonde profile in the layer between 40 and 100 hPa on October 20, 1998 (Figures 1e, 1f, and 2c). The assimilation with POAM data captures this depleted layer and very sharp ozone gradients from the edges of this layer towards well-captured maxima at 30 and 150 hPa. In contrast, the SBUV assimilation underestimates the extent of the ozone loss at pressures less than 100 hPa and the magnitude of the ozone maximum in the lowermost stratosphere. Large differences between two assimilations are seen even though POAM measurements are made near 78°S. The profiles in Figures 2a–2c were selected to show the impact of POAM data on analyses in different phases of the evolution of Antarctic ozone in winter and spring.
 A limitation of the POAM assimilation is illustrated in ozone profiles at Neumayer on September 23, 1998 (Figure 2d). The sonde profile exhibits a shallow laminar feature near 40 hPa, not captured by either assimilation, embedded in a deep layer of depleted ozone between 20 and 100 hPa. Using contour advection with surgery, Moustaoui et al.  showed that this lamina is a part of a filament of ozone-rich air transported poleward from the inner edge of the polar vortex. The ability of the dynamics to capture such narrow and shallow filaments is often limited (G. L. Manney et al., Diagnostic comparison of meteorological analyses during the 2002 Antarctic winter, submitted to Monthly Weather Review, 2004). In this case the satellite ozone data that were assimilated did not capture this filament. The SBUV data lack vertical resolution needed to capture such shallow features. POAM was measuring near 88°S at this time, in a different air mass unaffected by the filament.
5. Statistical Comparisons
 Statistical comparisons of all available sondes launched from South Pole, Neumayer, and Syowa (69°S, 40°E) with collocated analyses are shown in Figure 3. The statistics for 52 profiles from June to August from all three stations are shown in Figures 3a and 3b. Figure 3a shows that mean analysis has more ozone at pressures higher than 50 hPa, less ozone at pressures lower than 50 hPa, and agrees better with sonde mean when POAM data are used (cf. Figure 2a). Assimilation of POAM data reduces the root-mean-square (RMS) difference between analyses and sondes in the layer between 50 and 200 hPa (Figure 3b). The largest impact of POAM data is seen at 100 hPa, where this RMS difference is reduced from ∼7 to 2.6 mPa. There is a major improvement in statistical comparisons of analyses against sondes from all three stations in wintertime when POAM data are assimilated.
 At the South Pole, 20 ozone sonde profiles were available during September and October. Inclusion of POAM data increases analyzed ozone between 50 and 200 hPa, decreases ozone at 30 and 40 hPa, and provides a better agreement with the mean South Pole sonde profile (Figure 3c). The inclusion of POAM reduces the RMS difference in September and October between analyses and South Pole ozone sonde profiles throughout the stratosphere, and especially at 30 and 70 hPa (Figure 3d). In September and October, the assimilation of POAM data results in a smaller, but still positive, impact at Neumayer and Syowa (not shown). Recall that these stations are near 70°S, where SBUV data become available in the spring, and the analyses may be drawn too strongly to the SBUV data. Statistical comparisons of analyses with sondes from three stations in springtime show that the improvements from assimilation of POAM data are the largest at the South Pole, which is not observed by the assimilated SBUV data.
 In this letter we present results from the first assimilation of POAM III ozone data. We used POAM III data to provide additional constraints on polar ozone within a global system that already assimilates NOAA 14 SBUV/2 data. Assimilation of POAM data significantly improved the representation of ozone profiles over Antarctica in wintertime and springtime of the year 1998. Wintertime accumulation of ozone in the lower stratosphere within the polar vortex was captured. Springtime ozone depletion was represented in a properly confined lower stratospheric layer. Ozone profiles from assimilations with and without POAM data were evaluated by comparison to independent ozone sonde profiles at Neumayer, Syowa, and the South Pole. The comparisons against selected individual profiles were followed by a statistical evaluation using all available sondes from the three stations.
 We found that assimilation of just 14 to 15 solar occultation profiles provided daily by POAM in one hemisphere significantly improves the representation of lower stratospheric ozone within the polar vortex, which is not adequately observed by SBUV data. This study shows that a successful assimilation of solar occultation data provided a sufficient constraint to capture a geophysical phenomenon on a regional scale. A case that illustrates a limitation of the dynamical model that was not overcome through POAM assimilation was presented. An ozone filament that originated at the vortex edge was missed while POAM was sampling the innermost vortex core.
 Assimilation of sparse POAM data constrains polar ozone profiles. A multi-year assimilation of POAM II and POAM III data could provide insight into interannual variability of the polar ozone distribution in early spring and potentially in qualifying the impacts of mixing into middle latitudes. Moreover, assimilation of POAM data, if they were available in real time, could significantly improve operational polar ozone analyses and forecasts.
 We thank S. Pawson and R. Rood for discussions and comments on the manuscript. We thank C. Randall and an anonymous reviewer for comments that improved the manuscript. This work was supported by NASA's Atmospheric Chemistry Modeling and Analysis Program.