Atmospheric test of the J ( BrONO 2 ) / k BrO + NO 2 ratio : implications for total stratospheric Br y and bromine-mediated ozone loss

Introduction Conclusions References


Introduction
The effect reactive bromine has on stratospheric ozone is largely dominated by the Reactions (1), (2a), and (2b) (Spencer and Rowland, 1977): The brackets give the recommended quantum yields for λ > 300 nm.They determine the amount of reactive bromine (BrO) and thus the bromine-mediated ozone loss in almost the whole global lower stratosphere during daytime, except Published by Copernicus Publications on behalf of the European Geosciences Union.S. Kreycy et al.: In situ test of the J(BrONO 2 )/k BrO+NO 2 ratio in the chlorine-activated polar ozone hole regions.Sander et al. (2011) report for the termolecular Reaction (1) a 1σ uncertainty factor of 1.47 (at 220 K) and for the BrONO 2 absorption cross-section, σ (BrONO 2 ) (Reactions 2a and 2b) and hence for J(BrONO 2 ) an overall uncertainty factor of at least 1.4 (e.g.taken from Table 4.2 in JPL-2011).The former uncertainty factor mostly arises from the extrapolation of the laboratory measurements of k BrO+NO 2 from high to low temperatures.The uncertainty of σ (BrONO 2 ) is due to its large decrease by 3.5 orders of magnitude with wavelength, when going from the extreme UV (λ = 200 nm) to λ > 300 nm, where the actinic fluxes, and thus the spectral contribution to J(BrONO 2 ), strongly increase.

Methods
We report on spectroscopic measurements taken during a balloon flight of the LPMA/DOAS (Limb Profile Monitor of the Atmosphere/Differential Optical Absorption Spectroscopy) payload at Kiruna, Sweden (67.9 • N, 22.1 • E), on 7 and 8 September 2009.The payload accommodated three spectrometers: (a) a near-IR (LPMA) spectrometer that is suitable for the detection of O 3 , NO 2 , CH 4 , N 2 O, HNO 3 , and other trace-gases (e.g.Camy-Peyret et al., 1995;Payan et al., 1998); (b) a UV/vis spectrometer for the high-precision detection of O 3 , NO 2 , BrO, IO, O 4 , etc. in direct sunlight (e.g.Harder et al., 1998;Ferlemann et al., 2000); and (c) a UV/vis mini-DOAS instrument primarily for the detection of O 3 , NO 2 , and BrO in limb-scattered skylight (e.g.Weidner et al., 2005;Kritten et al., 2010).While spectrometers (a) and (b) measured direct sunlight during balloon ascent, solar occultation at sunset and sunrise, the mini-DOAS instrument recorded the atmosphere in limb geometry, with the azimuth angle being clockwise perpendicular (α = 90 • ) to the sun's azimuth direction.Viewing elevation angles were held constant (+0.05 • ) during balloon ascent but subsequently changed from +0.6 • to −4.88 • elevation angle in steps of 0.39 • for the limb observations at balloon float altitude.
The balloon was launched at 14:50 UT at a solar zenith angle (SZA) of 75 • on 7 September 2009, and balloon float altitude (≈ 33.5 km) was reached around 16:45 UT (SZA = 86 • ).Please note that here and in the following all SZAs refer to local SZA (i.e. for the balloon gondola position).On 7 September 2009 the solar occultation and limb observations during sunset lasted until 18:15 UT (SZA = 94 • ), and were resumed at 02:30 UT during sunrise on 8 September 2009 (SZA = 94 • ).However, the scanning motor of the limb observation spectrometer malfunctioned, and only spectra recorded after 03:50 UT (SZA = 86 • ) are used for analysis.They lasted until 06:00 UT (SZA = 75 • ), when the payload was separated from the balloon.Due to the low stratospheric winds at high latitudes during summer/winter circulation turn-over, the balloon payload gently drifted from Kiruna to the Finnish-Russian border (at around 350 km distance) within the 16 h long flight.Accordingly, due to the low shear winds the azimuth stabilisation of the balloon gondola, and therefore the sun and limb pointing, was extremely stable as compared to previous balloon flights (e.g.see Table 1 in Dorf et al., 2006a;Kritten et al., 2010).Further, the ambient temperature (T = 220 K) did not change by more than ±5 K within the altitude range from 11.5 to 31 km.
Here we primarily report on the DOAS data obtained from the limb observations during sunset and sunrise and data recorded during sunset using the direct sunlight spectrometer.
For both instruments the spectral retrieval is based on the DOAS method (Platt and Stutz, 2008).Since in previous studies the spectral retrieval has been described at length (e.g.Weidner et al., 2005;Dorf et al., 2006a;Butz et al., 2006;Kritten et al., 2010), here only those details are described which depart from our previous work.The retrieval of O 3 , NO 2 , and BrO from the solar occultation and the mini-DOAS measurements is performed along the parameters as given in Butz et al. (2006) and Aliwell et al. (2002), with updates as recently described in Dorf et al. (2008) and Kritten et al. (2010).Also, since the errors and uncertainties of the DOAS retrievals have already been discussed in length in previous studies (e.g.Harder et al., 1998;Aliwell et al., 2002;Weidner et al., 2005;Dorf et al., 2006a;Butz et al., 2006), they are only referred to when necessary.The results of the DOAS analysis are differential slant column densities (dSCDs) with respect to a reference spectrum, typically recorded directly after reaching balloon float altitude at low SZA (see Fig. 1).These values are then corrected for the amount of the targeted trace gas in the reference spectrum resulting in absolute slant column densities (SCDs).
The limb radiances are modelled using version 2.1 of the Monte Carlo radiative transfer (RT) model McArtim (Deutschmann et al., 2011).The model's input is chosen according to measured atmospheric temperatures and pressures, including a climatological high-latitude summer aerosol profile inferred from SAGE III (https://eosweb.larc.nasa.gov/project/sage3/sage3_table) and confirmed with the direct sun measurement of spectrometer (b) (Gurlit et al., 2005), the balloon altitude and the geolocation, SZAs as encountered during each measurement, the azimuth and elevation angles, as well as the field of view (FOV) of the mini-DOAS telescopes.Since the mini-DOAS spectrometer is not absolutely radio-metrically calibrated, all simulations are performed relative to the first limb spectrum (elevation  For the interpretation of the direct sun observations, our group's ray-tracing model (DAMF) is used that was extensively tested in the profile retrievals of past balloon flights (e.g.Harder et al., 2000, see Figure 1

therein).
For the photochemical modelling, the output from the most recent simulations of the 3-D CTM SLIMCAT (Chipperfield, 1999) at Kiruna for 6 September 2009 is used to initialise our 1-D facsimile code Labmos (e.g.Bösch et al., 2003).This approach is necessary here since the output from a global SLIMCAT run is only available every 48 h.This time resolution is too coarse to be used for comparisons with measurements.On the other hand using a 1-D photochemical model for the model versus measurement intercomparison appears justified, since during the balloon flight stratospheric winds were low, and thus very likely the same air masses were probed throughout our observations.However, both photochemical models use the most recent version of the Jet Propulsion Laboratory (JPL) kinetics and thermochemical data for all relevant gas-phase and heterogeneous reactions (Sander et al., 2011).Finally, the Labmos simulations are constrained to the measured N 2 O and CH 4 from spectrometer (a) to correct for small mismatches in the profiles of the source gases due to a small bias in the diabatic heating rate of SLIMCAT.Total stratospheric bromine (Br y ) is set to 20.3 ppt, and the Br y mixing ratio profile is accordingly vertically shifted (about 2 km) until the modelled and measured N 2 O and CH 4 profiles match.For this flight [Br y ] was (20.3±2.5)ppt (presumably in 5 yr-old air), which is determined from the direct sun measurements at balloon float using the so-called Langley method (e.g.Dorf et al., 2006bDorf et al., , 2008)).Also note that, for the assessment of Br y using Langley's method, stratospheric BrO is probed above balloon float, where the BrO/Br y partitioning is mostly due to Br atoms and BrO, and thus is insensitive to J(BrONO 2 ) and/or k NO 2 obtained from the direct sun observations of spectrometer (b) (see Fig. 4).
As an example of the simulations, Fig. 3 shows the simulated 2-D fields of BrO, BrONO 2 , and HOBr over Kiruna for 7 and 8 September 2009.Here, the simulation indicates that balloon soundings are well suited to study the Reactions (1), (2a), and (2b) at northern high latitudes during the summer to winter circulation turn-over, mostly because NO x concentrations are large and the profiles of both targeted gases (NO 2 , and BrO) nicely overlap as well, thus providing a good sensitivity for Reaction (1) during sunset.
At early dawn, unlike at dusk the Br y partitioning (the period of the red dashed lines in Fig. 3), is largely given by the efficiency of the heterogenous reaction of BrONO 2 + H 2 O| aqueous −→ HOBr + HNO 3 (R4) at night.This is mostly because (a) Reaction ( 4) is much slower than Reaction (1) in the volcanically quiet stratosphere, and (b) the photolysis of HOBr is similarly rapid as for BrONO 2 (Erle et al., 1999) (see also Fig. 3).Therefore, HOBr does not interfere with our measurements, except for the solar occultation measurements during early sunrise, which are not considered any further here.They will be discussed in a separate study addressing the stratospheric HOBr photochemistry.
In order to support an inter-comparison of measured and modelled SCDs of O 3 , NO 2 and BrO, the simulated photochemical fields are fed into the RT models McArtim and DAMF, where path integrals through the simulated photochemical fields are calculated and then compared with the measured SCDs.

Results
Figure 4 displays the inter-comparison of the measured and modelled limb SCDs of O 3 , NO 2 and BrO.While for O 3 and NO 2 the agreement is naturally close to perfect for all elevation angles due to the scaling of the profiles, measured limb BrO is in general larger than obtained from the simulations for the standard run (i.e.[Br y ] = 20.3ppt, σ (BrONO 2 ) and k BrO+NO 2 from JPL-2011).This is in particular true for the large BrO SCDs, which are obtained for large negative elevation angles (tangent heights down to 14 km), where the bulk of BrO and BrONO 2 resides.In short, our observations indicate that during dusk BrO tends to react later (or at larger SZAs) into its major night-time reservoir gas BrONO 2 , while at dawn (SZA < 86 • ) limb BrO tends to appear more rapidly than the standard simulation suggests.
A similar finding is obtained from the solar occultation measurements during sunset (see Fig. 5, upper panel) using the direct sun instrument (b) (Ferlemann et al., 2000) even though they are less sensitive to Reactions (2a) and (2b).This is mostly since the tangent height is by definition at SZA = 90 • where most of the absorption signal comes from.Hence, our solar occultation observations mostly probe the atmosphere for a more-or-less constant J(BrONO 2 ), but at the same time the effectively probed air masses (i.e.tangent points) move more and more away from the payload (up to 1200 km), towards the northwest during sunset.
Noteworthy in the comparison of measured and modelled BrO SCDs is the kink observed in the measurements for SZAs > 92.5 • , or tangent heights lower than 25.5 km.It is suspected that this kink is due to a dynamically different lowermost stratosphere close to and distant to the balloon sounding, rather than due to deficits in modelling the photochemistry of BrO.Two pieces of evidence support this hypothesis.First, a similar kink for the same SZA occurs in the measured O 3 SCDs (see Fig. 5, lower panel), which requires the inferred profiles of O 3 close to and distant to the sounding to be distinctly different for tangent heights lower than 25.5 km (see the insets in the lower panel of Fig. 5).Accordingly, since in the lowermost stratosphere at high latitudes the profile shape of O 3 is mostly due to dynamical processes (i.e.quasi-isentropic exchange of air masses with mid-latitudes, or the height of the tropopause), it is also suspected that the profile shape of Br indicates a negligible PV gradient at the upper level (950 K), and a small west/east PV gradient at the lower levels (at 475 K and 550 K), again indicative of different origins of the air masses close to and distant to the balloon sounding.Therefore, it is rather likely that the kink of the measured BrO SCD at around SZA = 92.5 • is not due to photochemistry, but due to transport.Therefore, the solar occultation measurements during dusk for SZAs > 92.5 • are disregarded for the following discussion.Accordingly, the findings inferred from the comparisons of the limb and solar occultation measurements with the photochemical modelling can be taken as evidence that either the ratio J(BrONO 2 )/k BrO+NO 2 is larger than indicated by the JPL-2011 compilations, or that Br y is incorrectly assumed in the model.These possibilities are investigated in the following section.

Discussion
In order to investigate potential causes for the deviation of the measured versus modelled BrO SCDs, a sensitivity test for the magnitude of the parameters J(BrONO 2 ), k BrO+NO 2 , and Br y is performed for limb and solar occultation measurements (Figs. 6 and 7).In both cases the best agreement between measurements and simulations is found by increasing J(BrONO 2 ) and decreasing k BrO+NO 2 , when forcing the regression line of measured versus modelled BrO SCDs through 0. Figure 8 illustrates the situation when varying J(BrONO 2 ) and k BrO+NO 2 for both the limb (dusk and dawn) and the solar occultation measurements (dusk), whereby the colour coding denotes the slope of the regression of measured versus modelled BrO SCDs.The figure indicates that for the limb measurement the best combination of J(BrONO 2 ), and k BrO+NO 2 is obtained when increasing J(BrONO 2 ) to a range of 1.1 to 1.4 and decreasing k BrO+NO 2 to a range of 0.65 to 0.80.RT calculations show that the limb samples are taken 30-70 km on the right-hand side of the payload, while the probed air masses of the solar occultation measurements move further and further away (up to 1200 km) as the sun sets.
For the solar occultation observations, the comparison between measured and modelled SCDs fully supports this conclusion, however, only for SZA < 92.5 • .For the solar occultation observations, the best agreement is obtained for a ratio of J(BrONO 2 )/k BrO+NO 2 | obs = 1.7(+0.2−0.1).Here the uncertainties are smaller compared to the limb observations, when varying the J(BrONO 2 )/k BrO+NO 2 ratio within the errors of the slope due to the higher accuracy of calculating the RT.However, since the direct sunlight observations are less sensitive to Reactions (2a) and (2b), these results are only used to confirm the results of the limb observations.Sensitivity runs for [Br y ] within the given uncertainty range (corresponding to ±2.5 ppt of Br y ) are also performed (not shown).Proportionally increasing/decreasing the modelled BrO SCDs by ±12.5 % (±2.5 ppt) slightly decreases/increases the slope of the data, but not far enough to obtain a 1 : 1 agreement between the modelled and measured BrO SCDs (Figs. 6 and 7).
Next, we address potential causes for the uncertainty in the J(BrONO 2 )/k BrO+NO 2 ratio.Incorrectly modelled actinic fluxes by the required amount are rather unlikely, since our RT model nicely explains the measured limb radiances at different wavelengths, different elevation angles and SZAs (e.g.see Figure 2 in Bösch et al., 2001 andFigure 5 in Deutschmann et al., 2011).Also the largely dominant contribution of the direct solar irradiance to the actinic flux seems to be well understood (e.g.Bösch et al., 2001;Gurlit et al., 2005).Moreover, since our measurement is insensitive to the quantum yields ( ) of Reactions (2a) and ( 2b to form BrO), an incorrect J(BrONO 2 ) points to an incorrect σ (BrONO 2 ).For J(BrONO 2 ) JPL-2011 states an overall uncertainty factor of 1.4 (Table 4.2) most likely due to uncertainties of σ (BrONO 2 ) in the UV-A and visible, and its temperature dependence.Since at T = 298 K σ (BrONO 2 ) agrees fairly well among the different studies, one may speculate whether the recommended temperature correction for T = 220 K at which our measurements are made is in fact too strong.Furthermore, JPL-2011 states a 1σ uncertainty factor of 1.47 for k BrO+NO 2 at 220 K. Here, the major uncertainty arises from the T dependence of the high pressure limit of the reaction, which is found to be rather large (m = 2.9).Attempts to fit the data with the JPL master equation were found to be insufficient at low pressures (Sander et al., 2011).Therefore, one may again speculate as to whether k BrO+NO 2 at 220 K is somewhat lower than recommended by JPL-2011.We note while our finding requires that J(BrONO 2 ) is larger and k BrO+NO 2 is lower than recommended by JPL-2011, such that J(BrONO 2 )/k BrO+NO 2 =1.7(+0.4−0.2),we cannot individually assess the magnitude of each of the involved parameters.However, it is likely that, with respect to the given error bars for each parameter, both parameters need to be changed from recommended values at low temperatures.
The finding has also implications for total stratospheric bromine.Using the inorganic method to assess stratospheric Br y relies on a photochemical correction: Br y is calculated from measured BrO according to . .)where ". .." indicates contributions from minor bromine species in the stratosphere (e.g.HOBr, Br, BrCl, and HBr).In our case, taking the revised J(BrONO 2 )/k BrO+NO 2 ratio (i.e.1.7(+0.4− 0.2)), stratospheric Br y may decrease by as much as 1.4 ppt.In fact, a smaller Br y assessed using the inorganic method would tend to close the existing gap to total stratospheric bromine assessed using measurements of organic source gases.Also the potential contribution of so-called very short-lived substances (VSLSs) to stratospheric bromine would accordingly decrease.For example, while our assessment of Br y for 4.5 year old air probed over Brazil in 2005 indicated a VSLS contribution of [VSLS] inorg = (5.2± 2.5) ppt (Dorf et al., 2008), the organic method resulted in only [VSLS] org = (1.25±0.08)ppt (Laube et al., 2008).A more recent comparison indicated [VSLS] inorg = (3.5 ± 2.5) ppt (inferred from our own BrO measurement using the Langley method) and about [VSLS] org = 2.25 ppt (Brinckmann et al., 2012), for the air masses jointly probed by both methods over Brazil in 2008.
The implications of our finding for stratospheric ozone loss are small or even negligible, since the ozone loss by the BrONO 2 photolysis has a small contribution to the total ozone loss by bromine (dominated by the reaction BrO + ClO), which is presently assessed to amount to about 30-35 % on a global average (Sinnhuber et al., 2009).Furthermore, even though BrONO 2 mostly photolyses into Br+NO 3 (Reaction 2b), only about 12 % of the produced NO 3 photolyses into the channel NO + O 2 , which in fact may cause some ozone loss (via reformation of BrO and NO 2 and the consumption of two ozone molecules).A possibly larger impact would be through the altered partitioning of bromine between BrO and its reservoir BrONO 2 , as BrO participates in a number of ozone loss cycles.
To quantify the impact of our findings, two runs with the SLIMCAT off-line 3-D CTM are performed (e.g.Chipperfield, 2006;Feng et al., 2007) integrated for 14 months using ECMWF meteorology.The model run has a horizontal resolution of 2.8 • × 2.8 • and included a detailed stratospheric chemistry scheme (Chipperfield, 1999).One model run is performed with standard JPL kinetics.In the other run J(BrONO 2 ) is scaled by 1.27 and k BrO+NO 2 is scaled by 0.75, thereby scaling the ratio J(BrONO 2 )/k BrO+NO 2 by 1.7. Figure 9 shows the percentage ozone difference between these runs for 2009 as an annual and zonal mean and as a zonal mean at 18 km altitude.Overall the impact of these kinetic changes on stratospheric ozone is small and confined to altitudes below about 30 km where BrONO 2 is a reservoir for bromine.The largest decrease in ozone is around 0.8 % at the edge of the Antarctic ozone region in September/October.The kinetics changes lead to less BrONO 2 (and HOBr) and more bromine in the form of BrO which catalyses ozone loss.The small effect itself is more important at the edge of the polar vortex and late in the season when NO y is more readily available to form BrONO 2 .Smaller changes are seen in the Arctic and at mid-latitudes.
In consequence, an increase in the photolysis rate and a decrease in the formation rate of BrONO 2 as determined here would eventually imply only a small change in the brominemediated ozone loss in the stratosphere.

Conclusions
We performed an atmospheric test of the J(BrONO 2 )/ k BrO+NO 2 ratio assisted by photochemical and radiative transfer modelling.It is found that, under stratospheric conditions (T = 220 ± 5 K and p ≈ 65 mbar), the ratio J(BrONO 2 )/k BrO+NO 2 is 1.7(+0.4− 0.2) larger than given in the JPL-2011 compilation.Our sensitivity study indicates that very likely both σ (BrONO 2 ) and k BrO+NO 2 differ from the JPL-2011 recommendation.
The major consequences of our study are threefold.(1) Recent assessments of total stratospheric bromine using the inorganic method during high stratospheric NO x loadings may have overestimated the necessary correction for the BrO to Br y ratio.As a consequence, stratospheric [Br y ] should be 1.4 ppt lower, which amounts to 6.8 % of the total stratospheric bromine.(2) A larger J(BrONO 2 )/k BrO+NO 2 ratio may also cause a small increase (maximum −0.8 %) in the bromine-mediated ozone loss in the stratosphere, because ozone loss by BrONO 2 and its products is anyhow small.Also an overestimated stratospheric Br y due to an incorrect J(BrONO 2 )/k BrO+NO 2 ratio would be compensated in the photochemical models, when reactive bromine is calculated using the inorganic method.(3) In the cold troposphere, a diminished formation of BrONO 2 where high NO x meets reactive bromine released from the degradation of organic bromine compounds, or bromine being heterogeneously released from salty aerosols or salt lakes, may lead to a longer lifetime of ozone-destroying BrO.In consequence, the revised J(BrONO 2 )/k BrO+NO 2 ratio may cause more ozone destruction and a more efficient degradation of organic molecules by their reaction with Br atoms on the one hand.On the other hand, it may hinder the activation of reactive bromine tied to the aerosol or in bulk salt at low temperatures (e.g.von Glasow et al., 2004;Salawitch, 2006).Accordingly, the consequences of our finding for ozone, and the oxidation capacity, in the troposphere may largely depend on the specific conditions.
cooperation.Finally we are grateful for the hospitality and support given by the personnel of Esrange/Kiruna to successfully perform the balloon flight.
Edited by: W. T. Sturges Fig. 1.Measured differential slant column densities of O 3 (upper panel), NO 2 (middle panel), and BrO (lower panel) in limb geometry during the balloon flight from Kiruna on 7 and 8 September 2009.During balloon ascent from 14:50 UT until 17:10 UT, the limb radiation was observed for an elevation angle of 0.05 • .During dusk (17:10 to 18:30 UT), 3.5 limb scans were performed at 33 km altitude.During early dawn, the scanning telescope of the limb spectrometer malfunctioned (the reddish area), and the limb scans (in total 7.25) started at 03:50 UT and continued until 06:00 UT, when the balloon was floating at 31 km altitude.Each limb scan consists of limb observations from

Fig. 2 .
Fig.2.Inter-comparison of measured (black)  and modelled (red) relative radiances at λ = 350, 450, and 495 nm for the limb scans as described in the legend of Fig.1.The RT simulations are normalised for the balloon ascent to the observations at 17:00 UT, and for each limb scan to the observation for the largest elevation angle (0.6 • ), i.e. for the lowest radiances.

Fig. 3 .
Fig. 3. Standard simulation of the key stratospheric bromine species for 7 September 2009 using the Labmos photochemical model: diurnal variation of BrO (upper panel), BrONO 2 (middle panel), and HOBr (lower panel).The red lines indicate the periods of the limb measurements (local SZA at the balloon gondola = 86-75 • at a.m. and 85-95 • at p.m.) and the dashed red lines the period of the direct sunlight measurements (local SZA = 95-90 • at a.m., and SZA = 90-95 • at p.m.).Solar zenith angles shown in the middle panel refer to the SZA at the balloon location, and the colour coding indicates the respective concentrations in units of 10 7 molec cm −3 .

Fig. 4 .
Fig. 4. Inter-comparison of limb measured versus modelled slant column concentrations of O 3 (upper panel), NO 2 (middle panel)and BrO (lower panel) for the standard JPL-2011 kinetics in units as given in the panels.Due to the scaling the slant column densities of O 3 and NO 2 excellently agree, but only fair agreement is obtained for BrO with standard model parameters.In the lower panel the JPL-2011 uncertainty of the J(BrONO 2 )/k BrO+NO 2 ratio is indicated (dashed lines).

Fig. 5 .
Fig. 5. Inter-comparison of measured versus modelled slant column densities of BrO (upper panel) and O 3 (lower panel) for the sunset solar occultation measurements on 7 September 2009.The data points shown with filled black dots are included, while the data points with open dots are disregarded in the analysis.The coloured lines in the upper panel show simulations for different pairs of J(BrONO 2 ) and k BrO+NO 2 as indicated by the legend.Local solar zenith angles (e.g. 90 • , 91 • , 92.5 • , and 94 • ) of some measurements are also indicated.The lower panel shows the measured O 3 SCDs in solar occultation, and the lines in red and light blue show the modelled O 3 SCDs for the two ozone profiles shown in the inset, respectively.The ozone profile (red line) is inferred from direct sunlight measurements in the visible spectral range during the balloon ascent (i.e.close to balloon sounding) and the ozone profile in light blue is a tuned profile to fit best the O 3 SCDs for SZA > 92.5 • .

Fig. 6 .
Fig. 6.Inter-comparison of limb measured versus modelled slant column densities of BrO for different scaling factors of J(BrONO 2 ) and k BrO+NO 2 as indicated in the individual panels.The open dots are for sunset and the full dots for sunrise observations.The s value indicates the slope and its uncertainty between modelled and measured BrO SCDs.

Fig. 7 .Fig. 8 .
Fig. 7. Inter-comparison of solar occultation measured versus modelled slant column densities of BrO for different scaling factors of J(BrONO 2 ) and k BrO+NO 2 as indicated in the individual panels.The closed dots are for SZA < 92.5 • and the open dots for SZA > 92.5 • , the latter being disregarded in the analysis (see text).The s value indicates the slope and its uncertainty between modelled and measured BrO SCDs.

Fig. 9 .
Fig. 9. Difference in ozone from a simulation of the SLIMCAT 3-D CTM with scaled J(BrONO 2 ) (×1.27) and k BrO+NO 2 (× 0.75) compared to a run with standard JPL kinetics.(a) Difference (%) in annually and zonally averaged ozone and (b) difference (%) in zonal mean ozone at 18 km as a function of time.Note the different colour scales in the two panels.