The use of SMILES data to study ozone loss in the Arctic winter 2009 / 2010 and comparison with Odin / SMR data using assimilation techniques

The Superconducting Submillimeter-Wave LimbEmission Sounder (SMILES) on board the International Space Station observed ozone in the stratosphere with high precision from October 2009 to April 2010. Although SMILES measurements only cover latitudes from 38 S to 65 N, the combination of data assimilation methods and an isentropic advection model allows us to quantify the ozone depletion in the 2009/2010 Arctic polar winter by making use of the instability of the polar vortex in the northern hemisphere. Ozone data from both SMILES and Odin/SMR (SubMillimetre Radiometer) for the winter were assimilated into the Dynamical Isentropic Assimilation Model for OdiN Data (DIAMOND). DIAMOND is an off-line wind-driven transport model on isentropic surfaces. Wind data from the operational analyses of the European Centre for MediumRange Weather Forecasts (ECMWF) were used to drive the model. In this study, particular attention is paid to the cross isentropic transport of the tracer in order to accurately assess the ozone loss. The assimilated SMILES ozone fields agree well with the limitation of noise induced variability within the SMR fields despite the limited latitude coverage of the SMILES observations. Ozone depletion has been derived by comparing the ozone field acquired by sequential assimilation with a passively transported ozone field initialized on 1 December 2009. Significant ozone loss was found in different periods and altitudes from using both SMILES and SMR data: The initial depletion occurred at the end of January below 550K with an accumulated loss of 0.6–1.0 ppmv (approximately 20 %) by 1 April. The ensuing loss started from the end of February between 575 K and 650 K. Our estimation shows that 0.8–1.3 ppmv (20–25 %) of O3 has been removed at the 600 K isentropic level by 1 April in volume mixing ratio (VMR).


Introduction
According to many studies of stratospheric ozone (O 3 ), major ozone depletion inside the isolated polar vortex is caused by the formation of polar stratospheric clouds (PSCs) and the associated heterogeneous release of active chlorine species (Cl, ClO) (e.g.Solomon, 1999).However, in comparison with the Antarctic polar vortex, the Arctic vortex is less stable due to the propagation of planetary waves from the troposphere.Therefore, the periods during which the temperature inside the vortex go below the threshold for PSC formation are limited (WMO, 2011).These effects make the quantification of chemical ozone depletion in the Arctic generally more difficult.
The winter of 2009-2010 was colder than other winters in the last decade during January (e.g.Dörnbrack et al., 2012).Figure 1 indicates the minimum temperature (T min ) as a function of day of the year (DOY) derived from the European Centre for Medium-range Weather Forecasts (ECMWF) operational forecasts on the 500 K potential temperature (PT) surface at equivalent latitudes (EQL) greater than 70 • N. The T min for the winter period of 2009/2010 was lower than 185 K from 1 January and became as low as 180 K on 7 Published by Copernicus Publications on behalf of the European Geosciences Union.January (see also Dörnbrack et al., 2012).Khosrawi et al. (2011) reported that strong denitrification caused by the formation of PSCs was observed during the synoptic cooling event in mid-January 2010.However, a sudden stratospheric warming (SSW) ended the coldest period after 19 January (Dörnbrack et al., 2012;Kuttippurath and Nikulin, 2012).SSWs are wintertime phenomena that are characterized by suddenly increasing temperatures and a reversal of the zonal wind (Scherhag, 1952).The planetary wave disturbance of the vortex with the occurrence of the SSW event makes this winter dynamically complicated.
SMILES (Superconducting Submillimeter-wave Limb-Emission Sounder), a passive atmospheric sensor attached to the Japanese Experiment Module (JEM) on board the International Space Station (ISS), was developed by the Japan Aerospace Exploration Agency (JAXA) and the National Institute of Information and Communications Technology (NICT).SMILES used 4 K superconducting detector technology to measure high-precision vertical profiles of stratospheric and mesospheric species related to ozone chemistry.The instrument was operated from October 2009 until April 2010 when the local oscillator failed and provided atmospheric composition data typically within the latitude range of 38 • S-65 • N (Kikuchi et al., 2010).
The aim of this paper is to demonstrate the use of the high-sensitivity observations by SMILES to quantify polar ozone loss.However, it is still a challenge to use SMILES data to analyse the polar regions because of its limited latitude coverage.The Odin/SMR ozone has been also analysed in this study for comparison with SMILES ozone because this instrument also uses the limb sounding technique and has a long record of stratospheric ozone measurements starting in 2001.Figure 2 shows a typical observation map of SMILES and SMR and Table 1 summarizes     N on a potential temperature surface of 500 K against day of year (0 refers to 1 January).Note that only measurements with measurement response above a threshold of 0.85 are considered.
specification of both observations.The higher vertical scan rate of SMILES compared to SMR explains the larger number of measurements.In addition, the dynamical instability of this winter season, displacing the vortex to latitudes below 65 • N, permitted a considerable number of SMILES observations within the vortex.Thus SMILES made more measurements than SMR even in the vortex (EQL ≥ 70 • N) (see Fig. 3).On the other hand, there are periods when SMILES measurements inside the vortex were not available due to ISS manoeuvres.In the first half of December, the field of view of the SMILES antenna was blocked by the ISS solar paddles, resulting in only a few useable measurements.Another period without measurements, in the middle of February, is due to the rotation of the ISS to dock with the space shuttle Endeavour.When the space shuttle was docked, the ISS was rotated by 180  In this paper, we employ the data assimilation technique developed for other Arctic winters by Rösevall et al. (2007aRösevall et al. ( , b, 2008) ) to investigate the ozone depletion in the 2009/2010 winter using SMILES ozone data.Other similar studies have used various models and assimilation methods (El Amraoui et al., 2008;Jackson and Orsolini, 2008;Søvde et al., 2011).One advantage of data assimilation is that it allows us to optimally use all measurements and is useful for interpolating or extrapolating the ozone distributions when and where no measurements are available.In this study we have used the DIAMOND assimilation model developed by Rösevall et al. (2007b) to produce active and passive tracer fields.However, because it is a two-dimensional model, Rösevall et al. (2007aRösevall et al. ( , b, 2008) ) needed to account for the effect of the diabatic descent inside the vortex a posteriori.Thus we have implemented a new vertical transport scheme that continuously accounts for the descent rather than an a posteriori correction.Ozone observed by SMR is analysed for comparison.This paper is structured as follows.Sections 2 and 3 describe the measurements and the model, respectively.Section 4 tests the effectiveness of the new vertical transport scheme using the long-lived species N 2 O measured by SMR and then shows the results of the ozone analyses.Finally, we conclude the study in Sect. 5.

Measurement descriptions
Profiles of ozone were obtained from the SMILES and SMR instruments.Nitrous oxide (N 2 O) from SMR was used for this study as a tracer of transport in the stratosphere due to its long lifetime for checking the dynamics in the model.

SMILES
SMILES observed atmospheric limb emissions from the ISS flying at an altitude of ∼ 340-360 km.It vertically scanned the tangent heights of ∼ −20-120 km with an antenna fieldof-view of ∼ 3 km.A single spectrum was obtained with a data integration time of 0.47 s, and one vertical scan took 53 s including the calibration data acquisition.About 1630 scans were obtained per day.Because the ISS has a non-sunsynchronous orbit, the local time of SMILES measurement location evolved over 24 h after approximately 1 month.
SMILES operated in three frequency bands: 624.32-625.52GHz (band A), 625.12-626.32GHz (band B) and 649.12-650.32GHz (band C).Bands A and B contain the emission line of ozone at 625.371 GHz.The measurements are spectrally resolved with an Acousto-Optical Spectrometer (AOS) which has a bandwidth of 1.2 GHz and a resolution of 1.2 MHz.Since SMILES only had two AOSs, the bands were observed on a time-sharing basis.The measurement noise of SMILES is as low as < 0.7 K (for a single AOS channel and a single spectrum) due to the low noise performance of the superconductor-insulator-superconductor (SIS) mixers.See Kikuchi et al. (2010) for further details about the SMILES instrumentation.
We used the ozone data processed by the NICT level-2 chain version 2.1.5.This level-2 algorithm employs a leastsquares method with a priori regularization (e.g.Rodgers, 2000) as described by Baron et al. (2011).The SMILES NICT ozone data were validated in Kasai et al. (2013).The SMILES ozone profile covers altitudes from 16 to 90 km with a resolution of ∼ 3-4 km in the stratosphere (see Fig. 2 in Kasai et al., 2013).Based on the error analysis and comparison studies, Kasai et al. (2013) reported a systematic error of better than 0.3 ppmv in the stratosphere (∼ 60-8 hPa).The random error for a single ozone profile is as low as 1 % for this altitude region.It is also reported that the data quality of ozone profiles from band B is better than that from band A. Ozone data from both bands are used in this study since no bias exists between ozone from bands A and B (see Kasai et al., 2013).Band A ozone data are only used when data from band B are not available.

Odin/SMR
Odin is a Swedish satellite mission in association with Canada, Finland and France, which was designed for rawww.atmos-chem-phys.net/14/12855/2014/Atmos.Chem.Phys., 14, 12855-12869, 2014 dio astronomy and limb sounding of the Earth's middle atmosphere (Murtagh et al., 2002).Odin was launched on 20 February 2001 into a sun-synchronous polar orbit with an inclination of 98 • , altitude of ∼ 600 km and descending and ascending nodes at 6 and 18 h local solar time, respectively.It carries two different limb sounding instruments, OSIRIS (Optical Spectrograph/InfraRed Imaging System) and SMR (Sub-Millimetre Radiometer).The SMR instrument, described by Frisk et al. (2003), consists of four tunable single-sideband Schottky-diode heterodyne microwave receivers.
The data sets for ozone and N 2 O from SMR used in this paper are products of the stratospheric mode that is operated every other day since April 2007 (every third day previous to this).In the stratospheric observation mode, two of the receivers, covering the bands centred at 501.8 GHz and 544.6 GHz, are used for detecting the spectral emission lines of ozone and N 2 O.The ozone and N 2 O profiles used in this study are retrieved from emission lines at 501.5 GHz and 502.3 GHz, respectively, using the Chalmers version 2.1 retrieval scheme.
The SMR ozone profiles cover the altitude range ∼ 17-50 km with an altitude resolution of 2.5-3.5 km and an estimated single-profile precision of ∼ 1.5 ppmv (Urban et al., 2005a).SMR v2.1 ozone data have been validated against balloon sonde measurements as described in detail by Jones et al. (2007).They show that SMR ozone in the 60-90 • N latitude band has mixing ratios that are 0.0-0.1 ppmv lower than sonde measurements below 23 km and a positive bias of SMR ozone 0.1-0.3ppmv in the 23 to 30 km range.The validation study (Kasai et al., 2013) shows that SMILES generally gives slightly lower ozone values than SMR at altitudes below 20 hPa.
The N 2 O profiles cover altitudes in the range 12-60 km with an altitude resolution of ∼ 1.5 km.The estimated systematic error is less than 12 ppbv (Urban et al., 2005a).The validation of the N 2 O is reported by Urban et al. (2005b).Further comparisons with the Fourier transform spectrometer (FTS) onboard the Atmospheric Chemistry Experiment (ACE) and the Microwave Limb Sounder (MLS) on the Earth Observing System (EOS) Aura satellite are shown by Strong et al. (2008) and Lambert et al. (2007), respectively.SMR N 2 O agrees with ACE/FTS N 2 O within 7 % between 15 and 30 km (Strong et al., 2008).And SMR N 2 O v2.2 is larger than MLS N 2 O by ∼ 5 % in the pressure range of 68-4.6 hPa and 10 % larger at 100 hPa (Lambert et al., 2007).

DIAMOND model
The DIAMOND (Dynamic Isentropic Assimilation Model for OdiN Data) is an off-line wind-driven isentropic transport and assimilation model designed to simulate quasi-horizontal ozone transport in the lower stratosphere with low numerical diffusion.Isentropic off-line wind-driven advection has been implemented using the second-order momentum (SOM) advection scheme (Prather, 1986) which is a mass conservative Eulerian scheme.The idea of the Prather scheme is that by preserving the zero-to second-order moments of the subgrid-scale tracer distribution the quality of the transport is preserved.In this study, the wind fields from the operational analyses of the ECMWF have been used.Advection calculations are performed on separate layers with constant potential temperature (PT) ranging from 400 K to 1000 K in 25 K intervals.
The tracer profiles from SMILES or SMR are sequentially assimilated into the advection model.The assimilation scheme in DIAMOND is described as a variant of the Kalman filter.Details of the assimilation scheme can be found in Rösevall et al. (2007b).

Cross-isentropic transport
Under adiabatic conditions, PT is conservative in dry air and thus the air parcels normally move on a constant PT surface.The descent of air in the polar vortex caused by radiative cooling during polar night had not been taken into account in the previous model version.It is, however, necessary for a correct evaluation of ozone loss.
For a flow field (u, v, w), the advection equation of a (passive or active) tracer (x, y, , t) at given horizontal coordinates x and y, vertical coordinate in potential temperature and time t in an Eulerian coordinate system is Here, u and v are the horizontal wind speeds and w is the vertical component of air mass advection with units of K d −1 .As we mentioned previously, DIAMOND employs the SOM method to solve the first and second terms in the right-hand side of Eq. ( 1).On the other hand, to account for the descent we implemented a simple vertical transport scheme into DIAMOND.The following equation is the one-dimensional first-order upstream method implemented into the model: The first-order upstream method often produces numerical diffusion.In order to avoid this, it is at least necessary to satisfy the condition  in the model (= 2.5 K min −1 ) is much larger than the general descent rate inside the polar vortex (∼ 1 K d −1 ), and therefore the first-order upstream method can be used satisfactorily.It is also important to have a sufficiently small separation of the layers to obtain a good representation of the descent.
To quantify the vertical transport, we used the diabatic heating rate Q (K s −1 ) derived from SLIMCAT 3-D chemical transport model calculations (Chipperfield, 2006).The vertical velocity w was calculated as where T is the absolute temperature.

Dynamics of the Arctic winter 2009-2010
In order to test the performance of the model and study the dynamics of this winter, we modelled stratospheric N 2 O fields by assimilation of SMR N 2 O.A summary of the calculations is given in Table 2. Initialization (i.e. the spin-up calculations with assimilation) for 1 month prior to the investigation period (from 1 December to 31 March) is required to ensure the accuracy of the initial model field.In order to remove contamination by erroneous observations, the SMR data are used only if the measurement response is larger than 0.85.The measurement response, the sum of the elements of the rows of the averaging kernel matrix, gives the contribution of the measurement to the retrieved information.To reduce any boundary condition problems realistic tracer fields are required.These are used as buffer layers to feed the vertical transport scheme.Note that the measurement response especially for SMR N 2 O is generally less than 0.7 at lower altitudes (< 450 K).Therefore, we relaxed the measurement response threshold to 0.7 for the boundary layers.In the results, we only show the output of the model from 450 K to 900 K to avoid boundary effects.The uncertainty of the DI-AMOND model due to imperfections in the transport scheme and/or unimplemented chemical processes has to be considered.We set the initial error fields to 30 % of the initial field of the species, which corresponds to the standard variation of the 40-day prediction without assimilations.The error field grows linearly to this value in 40 days if no measurements are available.
Figure 4 shows the model results for N 2 O and the corresponding error fields at 600 K.The polar vortex is clearly seen as the area where the volume mixing ratio of N 2 O is low.In the maps the EQL of 70 • N, which is used as the vortex boundary, is drawn as the black contour line.The white contour lines denote the Lait modified potential vorticity of 38 PVU (1 PVU =1 × 10 −6 K m 2 kg −1 s −1 ) (Lait, 1994).The modified potential vorticity refers to the reference level 475 K.The potential vorticity of 38 PVU is also used for the vortex boundary in other studies (Sonkaew et al., 2013;Hommel et al., 2014).The polar vortex was formed at the beginning of December and stayed at high latitudes for 2 weeks (DOY −31 to −15) then distorted and split into two parts caused by changes in the wind fields due to a minor SSW in the middle of December (∼ −15 DOY).The two separate vortices combined by 17 December.Kuttippurath and Nikulin (2012) have given a detailed analysis of these processes by using the potential vorticity.Our N 2 O results are consistent with their findings.After that, the vortex stayed cold and remained pole centred until the major SSW occurred at the end of January 2010 (e.g.Dörnbrack et al., 2012;Kuttippurath and Nikulin, 2012).This period shows the lowest temperatures of this winter at potential temperature of 500 K (see Fig. 1).The major SSW changed the wind field again: massive inflow of air from the Pacific forced the vortex to move to middle latitudes with flattening over Eurasia (see for example the Odin/SMR quick look at http://www.rss.chalmers.se/~jo/SMRquicklook/Qsmr-2-1/N2O_5018/gm/).Furthermore, the vortex again split after 10 February.The two parts are reunited on 1 March with entrainment of some extravortex air.The vortex was then relatively stable for some weeks.Finally, when the polar night ended, the vortex broke and the vortex air horizontally mixed with air from outside (Wohltmann et al., 2013).
To illustrate the advection in the DIAMOND model, we derived the vortex mean of N 2 O from the daily fields.Figure 5 shows the mean of the N 2 O volume mixing ratios inside the area where the EQL is equal to or greater than 70 • .The solid lines in the figure are calculated from the results of assimilation of SMR N 2 O.The two dashed lines, red and blue, are the vortex mean of the fields predicted by the advection model using the initial N 2 O distribution as of 1 December with and without vertical transport, respectively.If the vertical transport would be perfectly simulated in the model, the predicted and assimilated results should have the same values.Compared to the predictions from the 2-D advection scheme, the predictions with the vertical transport scheme shows good agreement with the vortex mean assimilated N 2 O field until the final break-up of the vortex.The uncertainty of the mean, plotted as the shaded areas in Fig. 5, is calculated as σ 2 + Ê2 .Here σ and Ê are the standard deviation of N 2 O inside the vortex and the vortex mean of the error field, respectively.More details of these compo-   nents can be seen in Fig. 6.Ê characterizes the error from the point of view of the instrument.However, the dominant factor in the uncertainties is the variability inside the vortex (σ ).The temporal evolution of σ allows us to assess the contribution of the (mostly horizontal) mixing.At the end of February (∼ 50 DOY), an exponential increase in σ occurs caused by the breaking of the vortex and associated mixing.This is particularly noticeable at 600 K.

Ozone inside the vortex
Figures 7 and 8 display maps of the results for ozone from the assimilations of data from SMILES and SMR.The results from the two instruments have similar patterns in the ozone maps although those from SMR exhibit more features and larger variations.The reasons for the differences are the number and quality of the measurements.Specifically, SMR has fewer measurements at lower latitudes because of its orbit and has a higher noise level.The SMILES ozone abundance, as expected due to known biases, was approximately 0.1 ppmv lower than SMR ozone below 700 K corresponding to 20 hPa in pressure (see Fig. 20 in Kasai et al., 2013).Another important point is the incomplete coverage of the centre of the vortex for the SMILES assimilation.As noted in the Introduction, SMILES did not observe at higher latitudes than 65 • N. As a result the measurement information on ozone in the polar region is transported from lower latitudes by the model.Thus, when the vortex is stable and well isolated, modelled ozone distributions may deviate from the true atmosphere.This is clearly seen in the SMILES ozone maps at the end of December where higher concentrations compared to earlier are seen inside the vortex due to the descent from higher levels and the lack of any chemical ozone loss processes in the model.
To avoid the effects of large local variations, we have chosen to use the average for the entire vortex for this study.The sampling issues described above are mitigated by employing a weighted average over the vortex as shown in Fig. 9.The weights are given by the estimated model error fields.Note the fact that the vortex mean of the SMILES assimilation thereby emphasizes the contribution near the vortex edge.Vortex averages of ozone from both instruments show similar patterns, especially before the major SSW event at     the end of January.Uncertainties in Fig. 9 are also calculated using the standard deviation σ and the mean of the error field Ê inside the vortex.Since SMR ozone is much noisier, information on the mixing from the vortex internal variation of the ozone fields σ are masked by the average error fields Ê, while for SMILES the total error reflects the variation inside the vortex especially when there are sufficient measurements available.

Ozone loss quantification
Arctic ozone depletion is estimated by subtracting ozone fields passively transported in the DIAMOND model from the fields with assimilated data.The time evolution of the ozone losses derived from SMILES and SMR are presented in Fig. 10a and b.Ozone losses inferred from the two instruments have similar patterns.The major differences are the loss during the period between 13 January (12 DOY) and 30 January (29 DOY) at 650 K in the SMILES result and the lower loss value from 17 March (75 DOY) below 500 K in the SMR result.Because the SMILES results reflect not the pole centre but lower latitudes near the vortex edge, the apparent loss (12-29 DOY, 650 K) in SMILES is due to an overweighting of the losses near the vortex edge.The reason why SMR loss is lower than SMILES below 500 K is because SMR ozone measurements tend to overestimate ozone at these altitudes due to lower sensitivity/measurement response.
The first significant depletion occurred below 550 K from 25 January to 7 February (24-37 DOY): this corresponds to the period when the vortex moved towards lower latitudes and becomes exposed to sunlight.The loss rate is approximately 0.06 ppmv d −1 and accumulated loss of 0.8 ppmv can be seen on 7 February (37 DOY) at 500 K.The second ozone loss period took place from the end of February at the heights between 575 K and 650 K.This continued at a rate of 0.03 ∼ 0.04 ppmv d −1 .Figure 11 presents the accumulated ozone loss as of 28 February 2010 just before the beginning of the second loss.The first loss occurring below about 550 K is clearly seen in the profiles of ozone loss as peaks at 500 K.The maximum losses derived by SMILES and SMR are 1.0 ppmv and 0.7 ppmv at 525 K, respectively.Figure 12 shows on the other hand accumulated loss profiles as of 31 March.The loss values below 550 K for both instruments are almost the same as those on 28 February.However, the loss profiles are higher than on 28 February at most altitudes with the maximum values of 1.1 ppmv (SMILES) and 1.3 ppmv (SMR) at 600 K. Ozone loss as estimated from SMILES is slightly larger at most levels.It is proposed that the difference in ozone loss between the two instruments is a result of sampling differences.SMILES captures ozone changes near the vortex edge where the area has been more exposed to sunlight while SMR on the other hand represents the loss at the centre of the vortex because of its orbit.
The initial loss from 25 January below 550 K can be explained using the classical mechanism related to heterogeneous reactions on PSCs and the chlorine catalytic cycle (Solomon, 1988).From mid-December until the middle of January, the temperature inside the vortex was cold enough to form PSC (Figs. 10f, g and h).Vortex average ClO in daytime and nighttime is also presented in the figure panels (Fig. 10c and d).The upper-level (575 K to 650 K) ozone loss from the end of February was not due to chlorine-related destruction.The second loss is correlated with the sun exposure time inside the vortex (shown in Fig. 10e).Nevertheless, the vortex average ClO is still low at around 600 K (Fig. 10c).Similar losses were also found in other warm winters (Konopka et al., 2007;Grooß and Müller, 2007;Jackson and Orsolini, 2008;Søvde et al., 2011).Konopka et al. (2007)   indicates that the NO-NO 2 cycle is dominant in a PSC-free polar stratosphere in the PT range of 600-850 K.

Comparison with other studies
Comparable studies of ozone loss in the winter 2009/2010 based on different analysis methods and measurements were done by Kuttippurath et al. (2010), Wohltmann et al. (2013) and Hommel et al. (2014).The inferred loss of SMR ozone agrees with loss simulated using the chemical transport model ATLAS by Wohltmann et al. (2013).They find cumulative ozone losses of 0.8-0.9ppmv at 500 K until 2010/03/30 based on a 68.5  Considering the potential temperature range of 600-650 K, our study finds cumulative ozone loss until 31 March 2010 of about 1.0 and 1.3 ppmv derived from SMILES and SMR, respectively.Wohltmann et al. ( 2013) find a loss of 1.3-1.4ppmv in their model result and a loss of 1.7 ppmv in MLS data.The analysis of SCIAMACHY data shows losses of 1-1.5 ppmv in the beginning of April (Hommel et al., 2014).
Plausible explanations for differences are instrument sampling differences and the vertical resolutions of the profile measurements.Other possible reasons are the different criteria for defining the vortex edge.We have tested two different vortex criteria.In addition to using the EQL of 70 • N, the modified PV criterion (38 PVU) applied in Hommel et al. (2014) has been used for quantifying ozone loss.Major differences between ozone loss derived with the two criteria can be seen around 800 K on 28 February in Fig. 11 and below 550 K on 31 March in Fig. 12.With the modified PV criterion we obtain roughly 0.2 ppmv higher loss above 800 K on 28 February in Fig. 11.In this height range, N 2 O has large variations and the standard deviation of ozone inside the vortex (not shown) is large for the period of the vortex separation.However, there are no differences in ozone loss below 700 K where ozone loss generally occurs.For example, Fig. 4 compares the two vortex edge criteria at the 600 K level.A good agreement of the two criteria can be seen until the end of February, but differences are found in March.Differences of ozone loss on 31 March at 500 K are approximately 0.3 ppmv.It is thus likely that the air near the vortex edge moderates the loss of ozone using the EQL criteria.The losses derived with the two criteria still agree with each other within 10 percentage units.

Conclusions
Data sets from SMILES and SMR have been used to quantify ozone loss inside the polar vortex for the Arctic winter 2009/2010.The investigation was performed using the Atmos.Chem.Phys., 14, 12855-12869, 2014 www.atmos-chem-phys.net/14/12855/2014/DIAMOND data assimilation framework.DIAMOND is an off-line wind-driven transport model advecting air on isentropic surfaces into which we introduced the vertical crossisentropic transport.Assimilation of SMR N 2 O was used to verify the cross-isentropic scheme when using SLIMCAT heating rates to calculate the diabatic descent.
We have demonstrated that indeed SMILES measurements combined with a data assimilation technique can be used to study ozone loss even at higher latitudes than 65 • N. Ozone fields from assimilation of SMILES and SMR show similar patterns, although the quality and coverage of measurements caused differences between the ozone fields, especially at lower latitudes.The agreement is also seen in the time evolutions of weighted vortex mean of ozone.
Ozone loss was derived by comparing fields acquired by sequential assimilation with passively transported fields.Significant losses are seen at different altitudes using both SMILES and SMR data and can be explained as follows: (I) Before the major SSW (∼ 21 January), the reasonably stable and isolated polar vortex remained centred around the north pole and cold temperatures allowed the formation of PSCs.(II) The SSW changed the wind field and the inflow of air from the Pacific pushed the vortex out towards middle latitudes.(III) The first rapid ozone depletion occurred below 550 K mostly close to the vortex edge where the polar night had ended (from 24 January to 7 February).The depletion of ∼ 1.0 ppmv is considered to be a result of ClO catalytic destruction.(IV) The second loss in the height range 575 K to 650 K started from the end of February and continued until the vortex break up.20-25 % of ozone was removed from this altitude.This loss was induced by the NO x reactions (Kuttippurath et al., 2010;Sonkaew et al., 2013).Further study is required to fully understand the mechanisms.The accumulated ozone loss as of 28 February derived from SMILES was higher than that from SMR by < 10 % and it can be attributed to loss occurring near the vortex edge.The initial peak of ozone loss at lower levels was 0.7 ppmv (15-20 %) at 475 K for SMR ozone and 1 ppmv (20-25 %) at 500 K for SMILES ozone, respectively.The second loss at 600 K was 0.8 ppmv (15-20 %) for both instruments.Ozone loss derived using the modified PV criterion agrees with the loss using the EQL criterion until the end of February below 700 K. Loss with the modified PV criterion on 31 March 2010 is approximately 0.3 ppmv higher at 500 K.Our estimation shows lower ozone loss than losses derived from other studies using MLS and SCIAMACHY measurements.

Fig. 1 :
Fig. 1: Minimum ECMWF temperature, Tmin [K], at a potential temperature of 500K inside the area where the equivalent latitude (EQL) is greater than 70 • , corresponding to the area inside the Arctic polar vortex, as function of days of year (0 refers to 1 January).The black solid line shows the mean value for years 2001 to 2012.The red line is the Tmin temporal evolution from 1 October 2009 to 31 March 2010.The shaded area encompasses the minimum/maximum Tmin between 2001 and 2011.
the nominal the mean value for years 2001 to 2012.The red line is the T min temporal evolution from 1 October 2009 to 31 March 2010.The shaded area encompasses the minimum/maximum T min between 2001 and 2011.

Fig. 3 :Figure 3 .
Fig.3: The number of measurements inside the area where the equivalent latitude is greater than 70 • N on a potential temperature surface of 500K against day of year (0 refers to 1 January).Note that only measurements with measurement response above a threshold of 0.85 are considered.

Fig. 4 :
Fig. 4: Modeled N 2 O fields with assimilation of SMR data (top) and their error fields (bottom) on selected dates at 600K level.The number in the bracket indicates the days of year.The contour lines indicate the vortex edge described with two criteria.The black line is based on the equivalent latitude (=70 • N) and the white line is based on Lait's potential vorticity (= 38PVU = 3.8 × 10 −6 Km 2 kg −1 s −1 ).

Figure 4 .Fig. 5 :Figure 5 .
Figure 4. Modelled N 2 O fields with assimilation of SMR data (top) and their error fields (bottom) on selected dates at 600 K level.The numbers in parentheses indicate the day of year.The contour lines indicate the vortex edge described by two criteria: the black line is based on the equivalent latitude (= 70 • N) and the white line is based on Lait's potential vorticity (= 38 PVU = 3.8 × 10 −6 K m 2 kg −1 s −1 ).

Fig. 6 :Figure 6 .
Fig. 6: The estimated uncertainty of the vortex mean of N 2 O.The blue area shows the standa deviation (σ) inside the vortex (EQL ≥ 70).The yellow area shows the vortex mean of the err fields ( Ê).Finally, the red area indicates the total estimated error, which has been calculated σ 2 + Ê2 and is shown as uncertainties in Figure 5.All data are smoothed over 3 days.

Fig. 7 :
Fig. 7: Same as figure 4 but for ozone derived from SMILES.

Figure 8 .
Figure 8. Same as Fig. 4 but for ozone derived from SMR.

Fig. 9 :Figure 9 .
Fig. 9: (Left panels) Same as figure 5 but for ozone from SMR (blue) and SMILES (red).(Middle panels) Same as figure 6 but for ozone from SMR. (Right panels) Same as figure 6 but for ozone from SMILES.

Figure 10 .
Fig. 10: Vortex averages of several parameters in the area where equivalent latitudes are larger than 70 • as a function of time (days from 1 Jan 2010) and isentropic levels between 400K and 1000K.(a,b) Vortex mean ozone loss derived from SMR and SMILES, respectively.(c,d) Vortex mean ClO retrieved from SMR in daytime and nighttime, respectively.(e) Cumulative solar exposure time of the polar vortex.(f) Minimum air temperature inside the vortex derived from ECMWF.(g,h) Areas where the temperature is below T NAT and T ice , respectively.
discussed that the loss around 650 K in 2002/2003 was induced by the catalytic cycles of NO x transported from the mesosphere and lower latitudes.The available result byKuttippurath et al. (2010)

Fig. 11 :Figure 11 .
Fig. 11: Vertical profiles of accumulated ozone loss as of 28 February 2010 from 1 December 2009.Loss was derived by subtracting the passive ozone from the assimilated ozone.The error bars are given as the standard deviation of derived daily ozone loss inside the vortex for the period of ± 3 days from 28 February 2010.The left panel shows loss in VMR, and the right panel shows relative losses in percent.The solid lines indicate losses with using the EQL criteria for the vortex edge (≥ 70 • N) while dashed lines show losses obtained using the modified PV criteria (≥ 38PVU).

Table 1 .
Typical specification of the observations.

Table 2 .
Description of the calculations.