The Arctic stratospheric winter of 2010/2011 was one of the coldest on record with a large loss of stratospheric ozone. Observations of temperature, ozone, nitric acid, water vapor, nitrous oxide, chlorine nitrate and chlorine monoxide from the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) onboard ENVISAT are compared to calculations with a chemical transport model (CTM). There is overall excellent agreement between the model calculations and MIPAS observations, indicating that the processes of denitrification, chlorine activation and catalytic ozone depletion are sufficiently well represented. Polar vortex integrated ozone loss reaches 120 Dobson Units (DU) by early April 2011. Sensitivity calculations with the CTM give an additional ozone loss of about 25 DU at the end of the winter for a further cooling of the stratosphere by 1 K, showing locally near-complete ozone depletion (remaining ozone <200 ppbv) over a large vertical extent from 16 to 19 km altitude. In the CTM a 1 K cooling approximately counteracts a 10% reduction in stratospheric halogen loading, a halogen reduction that is expected to occur in about 13 years from now. These results indicate that severe ozone depletion like in 2010/2011 or even worse could appear for cold Arctic winters over the next decades if the observed tendency for cold Arctic winters to become colder continues into the future.
 Large losses of Arctic stratospheric ozone have been observed during winter 2010/2011, exceeding observed losses during cold winters over the past decades, characterized as the first Arctic Ozone Hole [Manney et al., 2011]. Although in general Arctic ozone is expected to recover because of the reductions in ozone depleting substances as a result of the Montreal Protocol and its amendments, the observation that apparently the cold Arctic winters in the stratosphere have been getting colder over the past decades [Rex et al., 2004; World Meteorological Organization (WMO), 2011] raises some concern that Arctic ozone depletion may worsen over the next decades if the cooling trend continues while concentrations of ozone depleting substances remain sufficiently high.
 Here we show observations and model calculations of the 2010/2011 ozone loss and investigate its sensitivity to a further cooling of the Arctic winter stratosphere. Our results are based on observations from the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) [Fischer et al., 2008] onboard ENVISAT together with calculations from an isentropic chemical transport model (CTM), driven with temperatures and wind fields from analyses of the European Centre for Medium-Range Weather Forecasts (ECMWF).
2. Model and Data
 In this study we use MIPAS data processed with the research processor of the Institute for Meteorology and Climate Research (IMK) [von Clarmann et al., 2003], adapted to the reduced spectral resolution measurement mode under which MIPAS has been operated since 2005 [von Clarmann et al., 2009]. The latest version of data used here is based on ESA's version 5 calibrated radiance spectra and version 220 IMK level-2 processing version for temperature, ozone (O3), nitric acid (HNO3), chlorine nitrate (ClONO2), water (H2O), nitrous oxide (N2O) and chlorine monoxide (ClO). The vertical resolution is for all species around 3 km (2.0 to 4 km) in the altitude range of interest here. For all parameters except ClO, the precision estimate for a single profile is in the order of few percent (3–8%), while spectroscopic uncertainties contribute with 5–15% to a systematic, i.e., constant-in-time, uncertainty. Thus, the random error of the vortex means of these species (consisting of about 80 samples per day) is negligible. The altitude range covered by these species is cloud top (or 6–12 km) to the stratopause. ClO has reduced precision of 170% (at 15 km) to 75% (at 25 km) for a single profile, i.e., the random error is 19–8% for the vortex means, and 15–20% systematic uncertainties due to spectroscopy. ClO from MIPAS reduced resolution data is known to be biased high at the upper stratospheric maximum above 30 km. N2O is known to have a high bias of about 10% below 20–25 km. An empirical correction with N2O′ = N2O/(1 + 0.1N2O/324ppbv) was applied to account for this bias. All MIPAS data were interpolated to isentropic levels corresponding to the model levels and vortex averages were calculated by an area weighted average of the data poleward of 70°N equivalent latitude according to ECMWF potential vorticity (PV).
 The model used in this study is an isentropic 3-D CTM with 29 levels between 330 and 2700 K (about 10 to 55 km) and a spatial resolution of 2.5° lat. × 3.75° lon. [e.g.,Sinnhuber et al., 2003; Aschmann et al., 2011]. It is driven here by temperatures and horizontal wind fields from ECMWF operational analyses. Vertical transport is calculated from interactively calculated diabatic heating rates using the MIDRAD scheme [Shine, 1987]. The model has a comprehensive stratospheric chemistry scheme with reaction rate constants and photolysis cross sections according to JPL 2006 [Sander et al., 2006]. The model runs use an equilibrium polar stratospheric cloud (PSC) scheme, assuming that PSCs consist of liquid particles down to the ice frost points. At colder temperatures it is assumed that PSCs consist of ice with a nitric acid tri-hydrate (NAT) coating. Ice particles sediment with a prescribed fall velocity, sedimenting simultaneously the condensed HNO3. There is no sedimentation of PSCs in the model above the ice frost-point as liquid particles are too small to have a significant fall velocity.
 The model is initialized on 1 October 2010 from a 22-year model integration [Aschmann et al., 2011], except for ozone, HNO3, N2O and ClONO2, which are initialized on 1 December 2010 from MIPAS observations. HCl is adjusted so that the sum of HCl and ClONO2is preserved during initialization. The model run includes an extra 5 ppt of bromine from the very short-lived source gases CHBr3 and CH2Br2 [Salawitch et al., 2005; Sinnhuber et al., 2009]. To diagnose polar ozone depletion, a passive ozone tracer was initialized on 1 December 2010. All model runs are integrated until 16 April 2011.
 To investigate the sensitivity of ozone loss to temperature, halogen loading and stratospheric water we performed a number of additional sensitivity calculations with the CTM. Two model runs were performed with temperatures in the stratosphere uniformly increased or decreased by 1 K, respectively. These temperature changes affect only the chemistry scheme, including PSC formation and sedimentation. There is a small indirect impact on the transport through the ozone dependent heating rates. One sensitivity run was performed with a 10% reduction in all chlorine and bromine containing gases and finally one sensitivity run was performed with stratospheric water uniformly increased by 1 ppmv.
Figure 1shows 50 hPa temperatures averaged northward of 60°N over the extended winter period December to March from ECMWF ERA-40 (1958/59–2001/02) and ERA-Interim (1979/80–2010/11) re-analyses. The winter 2010/2011 was about 4 K colder than the long-term average [see alsoHurwitz et al., 2011]. As noted before [Rex et al., 2004; WMO, 2011] the cold Arctic winters seem to become colder. A linear trend through the coldest winters in a given 5-year period gives a cooling of 1.0 ± 0.5 K/decade for ERA-Interim and 0.6 ± 0.3 K/decade for ERA-40. This cooling trend is even more pronounced when expressed in terms of the volume of air with temperatures below possible PSC existence in a given winter, VPSC, a quantity that is more directly related to ozone loss than polar cap temperatures. Rex et al.  have estimated an increase in VPSC over the period 1966–2003 that is consistent with a cooling of the cold Arctic winters of 1.0 ± 0.5 K/decade.
 The evolution of lower stratospheric temperatures between December 2010 and April 2011 is shown in Figure 2a for the 475 K isentropic level (about 18 km altitude or at about 40 hPa), averaged north of 70° equivalent latitude. Three periods of particularly low temperatures are observed at the end of December (around day 0), at the end of January (around day 25) and in second half of February (around day 50).
 The solid black line in Figure 2a shows ECMWF temperatures with the gray shading indicating a ±1 K temperature range. Temperatures retrieved from MIPAS observations are shown as red points. There is in general excellent agreement between MIPAS and ECMWF, except for the coldest periods where widespread PSC occurrence limited MIPAS observations, thus leading to a warm bias in the vortex average as the coldest regions are not observed. Sampling the ECMWF temperatures at the location of the MIPAS observations (small black points in Figure 2a) leads again to excellent agreement also for the coldest periods. In general ECMWF and MIPAS temperatures agree within ±1 K, with a tendency for ECMWF being slightly colder by about 0.5 K than MIPAS in early winter and slightly warmer by about 0.5 K in late winter.
Figure 2b shows MIPAS observations and model calculations of ozone. At 475 K vortex averaged ozone decreases from about 3 ppmv in early December to about 1.5 ppmv in early April. There is excellent agreement between observed and modeled ozone. The gray shading shown in Figure 2b (and all other panels) indicates the range of the two sensitivity calculations with temperatures decreased or increased by 1 K. The dashed line in Figure 2b gives the model's passive ozone tracer that can be used to diagnose the chemical ozone depletion. About 60% of the vortex ozone at 475 K was destroyed by early April 2011. The column integrated ozone loss between 380 and 550 K, calculated with respect to the modeled passive ozone tracer, reaches 120 DU by early April (Figure 2c). The model shows slightly too much ozone depletion in December but agrees well with MIPAS at the end of the winter. A further 1 K cooling of the stratosphere results in an additional modeled ozone loss of up to 25 DU in early April. Interestingly, a 10% reduction of chlorine and bromine is almost exactly counterbalanced by a 1 K cooling (Figure 2c). Increasing stratospheric water by 1 ppmv has almost the same effect on ozone loss as a 1 K cooling (Figure 2c), in agreement with previous findings [Tabazadeh et al., 2000].
 Observed N2O (Figure 2d) decreases faster than modeled until end of February; during March observed N2O even increases again, while the model shows a continuing decrease until early April. Although this comparison at 475 K indicates that there are some discrepancies between modeled and observed rates of subsidence, the comparison of observed and modeled N2O profiles between 1 December and 31 March (Figure 3a) shows that overall the model reproduces the subsidence of vortex air over the winter quite well.
 Much of the temperature sensitivity of modeled ozone is due to the degree of denitrification, i.e., the removal of HNO3 in the model. Figure 2e compares modeled and observed HNO3. The dashed line in Figure 2e indicates a model calculation without sedimentation of HNO3. HNO3 decreases in three steps that coincide with the lowest temperatures at around day 0, day 25 and day 50. Denitrification occurs over an altitude range from about 16 to 24 km, with an observed maximum denitrification of about 50% at about 20 km. Below 16 km both MIPAS and the CTM show enhanced HNO3. In the model the denitrification is accompanied with some dehydration (removal of water), that is not observed by MIPAS (Figure 2f). The increase in water seen in the MIPAS observations is qualitatively consistent with the subsidence of air inside the vortex.
 The evolution of chlorine activation, as seen in ClONO2 and ClO, is generally well reproduced by the CTM. Because ClO has a strong diurnal cycle we have compared modeled ClO at the local time of the MIPAS observations (about 10 am). At high latitudes north of 80°N the local time of the measurements changes quickly from 10 am to 10 pm, so that the ClO comparisons are restricted here to latitudes southward of 80°N. The model with 1 K higher temperatures agrees better with MIPAS in early December, consistent with the slightly too cold ECMWF temperatures in early winter, but from late December until the end of March the model captures the development of ClONO2 and ClO well. Beginning in early April observed ClONO2 decreases faster than modeled, implying a faster conversion of ClONO2 into HCl than modeled.
Figures 3b and 3c compare the vortex mean modeled ozone loss and minimum ozone inside the vortex on 31 March 2011 with MIPAS observations. There is excellent agreement between model and observations for both mean ozone loss and minimum ozone in the 16 to 21 km region where most of the polar ozone loss occurred. Below 16 km modeled ozone loss is larger than the ozone loss diagnosed from MIPAS. Figure 3cshows that a 1 K cooling would result in a near-complete ozone destruction in parts of the vortex, with minimum ozone below 200 ppbv over a large vertical range from 16 to 19 km.
4. Discussion and Conclusions
 Substantial loss of Arctic stratospheric ozone occurred during winter 2010/2011. Peak losses of about 2.3 ppmv at 450–475 K and about 120 DU for the ozone column between 380 and 550 K exceeded previous observed losses [Sinnhuber et al., 2000; Rex et al., 2006; Manney et al., 2011]. Peak losses during 2000 were similar to 2011 but restricted to a more limited altitude range [Manney et al., 2011], while 2005 had relatively large losses at lower altitudes, resulting in comparable column losses as 2011 [Rex et al., 2006]. While significant denitrification has been reported for previous cold Arctic winters [e.g., Waibel et al., 1999; Popp et al., 2001], the observed vortex mean denitrification of about 50% at 20 km in March 2011 exceeds previous observations [Manney et al., 2011]. Overall the MIPAS observations agree well with MLS, confirming the results of Manney et al. .
 The CTM reproduces observed ozone, HNO3, ClO and ClONO2 remarkably well, indicating that the model has an adequate representation of denitrification, chlorine activation and ozone loss. The equilibrium PSC scheme forming large sedimenting NAT particles in the presence of ice reproduces observed denitrification reasonably well within the temperature uncertainties. However, the accompanying dehydration in the model is not observed by MIPAS. In fact, it is well known from theoretical [Salawitch et al., 1989] and observational [Fahey et al., 2001] studies that denitrification in the Arctic stratosphere is possible without dehydration. Simultaneous agreement of ClO and ozone loss between the CTM and MIPAS implies that the main polar ozone loss mechanisms are sufficiently well represented by the JPL-06 reaction rates.
 Our calculated value of 25 DU additional ozone loss for a 1 K cooling in early April (or 18 DU when averaged over the last 10 days of March) agrees well with the value of 15 DU/K by Rex et al.  based on an empirical relation between Arctic winter temperatures, expressed in terms of VPSC, and diagnosed ozone loss by the end of March for previous winters.
 Model calculations with 1 K lower temperatures than 2010/2011 show locally near-complete ozone depletion with remaining ozone of less than 200 ppbv from late March onward over a large vertical extent from 16 to 19 km altitude. The calculated vortex mean ozone loss in this scenario of more than 140 DU even exceeds estimates of current Antarctic ozone loss of about 130 DU [Tilmes et al., 2006]. Hofmann et al. showed that near-complete removal of ozone over an extended range of altitude in the Antarctic did not occur until some ≈5 years after the initial discovery of the Antarctic ozone hole.
 Whether or not we will experience in the future comparable or even larger Arctic ozone loss than during 2010/2011 will critically depend on the future evolution of Arctic stratospheric temperatures. There is considerable uncertainty if and to what extent past cold Arctic winters have cooled and there is even more uncertainty as to how this will evolve in the future. In general one expects a further stratospheric cooling as a result of increased levels of greenhouse gases. However, most chemistry-climate models predict a simultaneous increase in wave activity that will lead to higher Arctic temperatures in the future [WMO, 2011]. Our calculations show that the effect of a 10% reduction in stratospheric halogen loading, which is expected to occur in about 13 years from now (about 2024) if the Montreal Protocol is followed [Newman et al., 2007; WMO, 2011], can be counterbalanced by a 1 K cooling, i.e., by a temperature tendency of about 1 K in 13 years or about 0.8 K/decade. Changes in stratospheric water, either due to the increase of methane, or due to changes in the water inflow through the tropopause, may contribute to enhanced Arctic ozone loss, with the effect on ozone loss of a 1 ppmv increase in water being comparable to a 1 K cooling.
 MIPAS level 1 data were provided by ESA. IMK data analysis was supported by DLR under contract 50EE0901. ECMWF data were made available through the special project SPDECDIO. We thank Gregor Kiesewetter for assistance in handling of the meteorological data.
 The Editor thanks Ross Salawitch and an anonymous reviewer for their assistance in evaluating this paper.