The magnitude of chemical loss of polar ozone induced by anthropogenic halogens depends on the extent of chlorine activation, which is controlled by polar stratospheric clouds (PSCs) and thus by temperature. We propose a new quantity, the PSC formation potential (PFP) of the polar vortex, suitable for comparing the amount of ozone depletion in the Arctic and Antarctic regions. PFP represents the fraction of the vortex, over an ozone loss season, exposed to PSC temperatures. Chemical ozone loss in the Arctic correlates well with PFP, for winters between 1991 and 2005. For Antarctic and cold Arctic winters, PFP has been increasing over the past 30 years. In winter 2005, PFP and ozone loss in the Arctic reached record highs, approaching Antarctic levels. Nevertheless, column ozone in spring in the Arctic is much larger than the Antarctic, because of larger dynamical resupply of ozone to the Arctic.
 Ozone-depleting substances are slowly decreasing in the atmosphere as a consequence of the Montreal Protocol so that a recovery of the ozone layer is expected over the coming decades [World Meteorological Organization, 2003]. Annual- and global-mean temperatures indicate that the stratosphere has been cooling in the last two decades [World Meteorological Organization, 2003]. Should this cooling extend to the polar region, this would enhance the potential for the existence of polar stratospheric clouds (PSC). Here we focus on the vortex average of chemical ozone loss in the polar lower stratosphere. A compact relation between chemical ozone loss and the volume of air below the temperature threshold for PSC existence (VPSC) was derived for Arctic winters [Rex et al., 2004; Tilmes et al., 2004]. VPSC is used as a measure of the potential for halogen activation in the polar vortex. Chemical ozone loss is usually reported as an average value over the entire vortex whereas VPSC is not normalized by the vortex volume. Here, we extend VPSC to create a measure of halogen activation that is normalized with the vortex volume and therefore suitable for comparison with chemical loss in column ozone, the PSC formation potential (PFP). Using PFP as well as VPSC is especially important if the volume of the various vortices considered differs greatly, for example, in comparing Arctic and Antarctic conditions. This allows us to discuss how climate change, i.e. changing temperatures in the stratosphere, may impact the potential for ozone destruction in the polar vortex. Moreover, we will discuss the ozone column that would occur in the absence of chemical changes, thus for assumed chemically unperturbed conditions (proxy ozone column [Tilmes et al., 2006]) during winter and spring. Smaller proxy ozone columns for colder Arctic winters indicate a different dynamical situation in the stratosphere, namely one where less ozone was transported towards the polar region during the winter.
2. Differences Between Arctic and Antarctic Winter Meteorology
 The calculation of VPSC (Figure 1, top) requires a given PSC threshold temperature (TPSC). TPSC is defined here as the threshold temperature for the existence of nitric acid trihydrate (NAT) [Hanson and Mauersberger, 1988]. It is calculated assuming seasonally dependent stratospheric profiles of HNO3 and H2O derived from ILAS-II observations for the Arctic and Antarctic, including the effect of denitrification and dehydration. The UK Met Office (MetO) and ECMWF reanalysis (ERA 40) [Randel et al., 2004, and references therein] are used to calculate VPSC between 400–550 K potential temperature. The ERA40 reanalysis is unreliable in the Antarctic before the satellite era in 1979 [Randel et al., 2004; Simmons et al., 2004] and will not be investigated. Further, Antarctic temperatures after 1979 in ERA40 show a cold bias and an unrealistic vertical structure [Manney et al., 2005a, 2005b; Randel et al., 2004]. Therefore, MetO VPSC values are slightly lower than the ERA40 data. Additionally, VPSC was derived using ECMWF operational analyses and data from the Free University of Berlin (FU-Berlin) [Rex et al., 2004], averaged between 360–550 K (Figure 1, top, red and green symbols). The impact of averages of different altitude intervals for the calculation of VPSC is small (see auxiliary materials). For the Arctic, the patterns of interannual variability from different data sets agree rather well between 1966 and 1998 [Manney et al., 2005a]. Before 1966, radiosonde instrumentation was limited in the Arctic stratosphere and temperatures are very uncertain. In summary, between 1966 and 1998, the conclusions from all long-term meteorological data sets about the general evolution of VPSC are consistent [Manney et al., 2005a].
 Antarctic temperatures are in general below the PSC threshold for most of the winter between mid-May and the end of September. The variation of VPSC between different winters using MetO data is small. ERA40 and MetO data indicate a slight increase of VPSC between 1991 and 2005 as well as a stronger variability during the last four years (Figure 1, top, colored triangles). VPSC in the Arctic is much smaller than in the Antarctic and shows a strong variation between warm and cold winters (between 0.1 × 107 km3 in 2001–2002 and 4.6 × 107 km3 in 2005, based on MetO analysis). An increase of both large and very small VPSC values, especially in the last decade, is obvious as described by Rex et al. .
 The volume of the vortex (VVortex) (Figure 1, middle) – which is the volume enclosed by the vortex edge, defined as the location of the maximum gradient of potential vorticity (PV) [Nash et al., 1996] – is significantly larger for the Antarctic than for the Arctic. The evolution of MetO and ERA40 values is in agreement, with differences of less than 10% in the overlapping time period. VPSC is dependent on temperature conditions in the vortex and, further, it is not normalized by the vortex volume. Therefore, the vortex volume has an influence on VPSC. A smaller vortex volume will result in less volume that can be activated and therefore in a smaller VPSC. These two factors can act against each other. The Arctic and Antarctic VVortex values indicate a decrease since 1979, which is most significant since 1998 for the Antarctic (Figure 1, middle). However, VPSC has not been decreased since 1990 because of the changing temperatures (Figure 1, top).
 The vortex volume should not have an direct impact on the loss in column ozone averaged over the vortex. Therefore we derive the ratio of VPSC over VVortex as a measure of the possible fraction of the vortex volume exposed to PSCs. A large value of this ratio is expected to correspond to large ozone loss rates independent of the lifetime of the vortex and the time period when halogen compounds are activated, the activation period. To obtain a measure of the ozone loss potential in each winter, which is also determined by the activation period of the vortex, we integrated the value VPSC/VVortex over all days when the vortex existed (using the Nash criterion for 3 potential temperature levels: 475 K, 550 K and 650 K) between mid-June and September for Antarctic winters and between mid-December and the end of March for Arctic winters and when temperatures were below the threshold temperature for PSC existence. This sum is then divided by the total number of days in the period considered. We are using the same length of time intervals for the Arctic and Antarctic to be consistent for both hemispheres. This quantity will be referred to as “PSC formation potential of the polar vortex” (PFP) and is suitable for correlation with the accumulated loss in column ozone in both hemispheres, because it is normalized with regard to VVortex.
 Antarctic PFP values for the MetO and ERA40 data are in agreement within 10% during the overlapping time period. Between 1991 and 2005, PFP indicates an increase of ≈25% for Antarctic and cold Arctic winters – i.e. the three winters with highest PFP values – (Figure 1, bottom, and Figure 2 in section 4). Further, the variation of the Arctic PFP has increased during the last 30 years with the largest value occurring in winter 2005 and very small values occurring in 2004 and 2006 (not shown).
 PFP for the coldest Arctic winter in 2005 is only ≈27% smaller compared to the smallest PFP value derived for the Antarctic. In 1995 and 1998 radiosonde instrumentation changed at many stations with a systematic bias between the two instrument types. For Alaska stations, an apparent warming of up to 2 K during the nighttime and a cooling during daytime at 50–100 hPa is reported [Elliott et al., 2002]. Underestimated nighttime temperatures – most of the PSCs are likely to occur during winter when there is no or only little day light – lead to an overestimation of VPSC before 1995/1998. Any conclusion of increasing VPSC and the resulting increasing PFP since 1995/1998 can therefore not be an artifact caused by changes in radiosonde instrumentation.
3. Calculation of Chemical Loss in Column Ozone
 Chemical ozone loss for the Antarctic is derived from HALOE satellite measurements [Russell et al., 1993] using the tracer-tracer correlation method [e.g., Proffitt et al., 1990; Tilmes et al., 2004; Müller et al., 2005]. A detailed description of the technique and a discussion about uncertainties due to mixing processes is given by Tilmes et al. , Engel et al.  and Müller et al. . It is shown that this technique is a reliable tool for calculating chemical ozone loss. Antarctic ozone loss is derived here using tracer-tracer correlations in a similar manner to that used for the Arctic. Using this technique, the early winter reference function has to be derived carefully to calculate chemical ozone loss [Tilmes et al., 2004; Müller et al., 2005]. For the Antarctic, no HALOE observations are available in the early vortex. Therefore, ILAS and ILAS-II O3/N2O relations for winters 1997 and 2003 are converted to an O3/CH4 relation for application to HALOE measurements. For this purpose, we use a CH4/N2O relation derived (see auxiliary material) using 11 flights from whole air sampler measurements between 1995 and 2002, at different seasons, and measured at high northern latitudes [Engel et al., 2002]. ATMOS measurements [Michelsen et al., 1998] show that very similar relations between CH4/N2O in the Arctic and Antarctic (Plate 2 of that paper), thus, we assume the derived relation to be also valid for Antarctic conditions. Further, we use CH4/HF relations derived from HALOE measurements to calculate the O3/HF reference relation as described in detail by Tilmes et al. .
 Antarctic chemical ozone loss in the column between 350–550 K was derived for all HALOE measurements available inside the vortex core – poleward of the poleward edge of the vortex [Nash et al., 1996] – using HF as the long-lived tracer (auxiliary material Figure S2). The amount of ozone loss differences between different winters above 550 K [Hoppel et al., 2005] is insignificant if loss in column ozone is considered. Owing to the orbit of the HALOE instrument and the location of the polar vortex, in some winters there are no measurements available in the second part of September/October (5 out of 13) for the Antarctic and in March/April (3 out of 14) for the Arctic and therefore no ozone loss values are available for the analysis in Section 4, in Figure 2.
 Arctic chemical ozone loss was derived in a previous study [Tilmes et al., 2004] for winters between 1991–92 and 2002–03 between 380–550 K and for winter 2005 by von Hobe et al.  using tracer-tracer correlations. As shown by Tilmes et al. , no significant ozone loss was observed below 380 K before winter 2005. Chemical ozone loss values derived from ozone soundings are available between 1992 and 2004 [Rex et al., 2004] and for winter 2005 [Rex et al., 2006].
4. Ozone Loss and PSC Formation Potential
 Chemical ozone loss, as described in Section 3, is shown in relation to PFP in Figure 2. Here, PFP is averaged between 350–550 K based on the MetO analysis. The relation between ozone loss and PFP in the Arctic is more compact compared to the previously deduced relation between ozone loss and VPSC [Tilmes et al., 2004; Rex et al., 2004]. In winter 2005, chemical loss in column ozone reached Antarctic values, whereas the PFP value is significantly smaller than Antarctic values (Figure 2), as explained below. The study by Manney et al.  showed that Arctic local chemical ozone loss – maximum loss in mixing ratio in the profile – in 2005 did not reach the very large values observed in winter 2000. However, in considering chemical loss in column ozone, winter 2005 reached maximum values. This winter is characterized by significant ozone destruction at very low altitudes below 460 K [von Hobe et al., 2006; Rex et al., 2006].
 Although Arctic chemical ozone loss reached Antarctic values in 2005, the dynamical supply of ozone in the Arctic is much stronger than in the Antarctic. Therefore, the column of proxy ozone in the Arctic is substantially larger than the destroyed amount of column ozone (Figure 2). For cold Arctic winters – that show slightly smaller column of proxy ozone than warm winters – less than half the amount of the proxy ozone is destroyed at the altitudes considered. Thus, dramatic ozone hole values do not occur in the Arctic. However, the potential for larger amounts of chemical ozone loss is present for the Arctic winters in the near future until the halogen loading of the stratosphere has significantly decreased.
 For the Antarctic, the column ozone loss does not change with changing PFP within the uncertainty of the results. This occurs because the column of proxy ozone in the Antarctic does not differ much from the chemical column ozone loss. Practically all of the ozone has been destroyed in the layer where chemical processing occurs and the chemical ozone loss is saturated. Although the potential for further ozone loss exists at other levels, no further ozone can be destroyed between 350 and 550 K in the Antarctic (as shown by e.g., Tilmes et al. ). The chemical signal of Antarctic ozone depletion averaged over the polar vortex does not show any evidence of a chemical recovery due to decreasing chlorine content [Engel et al., 2002] (Figure 2 and auxiliary material Text S1 Sect. 1.2 and Figure S2). The recovery due to decreasing stratospheric chlorine content will become visible if the saturation (that already occurred in the nineties) no longer occurs, which cannot be expected before ≈2015 [Newman et al., 2006].
5. Discussion and Conclusion
 We use 47 years of meteorological analyses and 14 years of satellite data to analyze the sensitivity of chemical ozone loss to temperature changes in both the Arctic and Antarctic and, extending the work of Rex et al. [2004, 2006], to introduce an improved diagnostic for chemical polar ozone loss. The meteorological conditions are different in the two hemispheres and both the column ozone at the beginning of the winter and the proxy ozone during the winter are much larger in the Arctic. For the cold Arctic and Antarctic winters, PFP has been increasing over the last 30 years because of decreasing temperatures and decreasing vortex volume. Correspondingly, chemical ozone loss values in the Arctic reached Antarctic values in 2005, whereas the remaining entire ozone column is much larger in the Arctic than in the Antarctic. The linear relation between column ozone loss and PFP indicates a potential for further increasing ozone loss values in the near future, because Arctic ozone loss is currently not saturated. Antarctic chemical ozone loss is almost saturated and therefore, increasing PFP cannot change ozone loss values. Increasing chemical ozone loss might occur if the recent tendency for greater variability of Arctic winter conditions is driven by increasing greenhouse gases. Further, stronger planetary wave activity may enhance the proxy ozone in early winter. This may change the volume of the vortex for both the Arctic and Antarctic and therefore impact the PFP and thus chemical ozone loss.
 Three-dimensional coupled chemistry climate models (CCMs), the main tool for predicting the future of polar ozone, show deficiencies in particular with regard to the prediction of temperatures and dynamics of the polar stratosphere [e.g., World Meteorological Organization, 2003]. It has been suggested that the relation between PSC formation potential and ozone loss could be used to evaluate the temperature sensitivity of accumulated polar chemical ozone loss in CCMs [Eyring et al., 2005]. Different vortex volumes simulated by different climate models may have a significant impact on the simulated ozone loss. Using the PFP instead of VPSC to describe the climate sensitivity of chemical ozone loss in CCMs, the impact of different vortex volume will be separated from the chemical signal. In this way, the chemical signal of ozone changes can be analyzed although models might miscalculate the volume of the vortex. In summary, the PFP is an improved diagnostic compared to those previously used; it is recommended for analyzing the sensitivity of polar ozone loss to climate change.
 We gratefully acknowledge all members of the HALOE team at NASA Langley and of the ILAS and ILAS-II team at NIES, Japan, for their work in producing a high-quality data set. Thanks are also due to ESA for preliminary MIPAS-ENVISAT spectra and the UK Meteorological Office and the European Centre for Medium-range Weather Forecasts for providing meteorological analyses. Finally, S. Tilmes thanks the Deutsche Akademie der Naturforscher Leopoldia and the Bundesministerium für Bildung und Forschung for supporting this study.