Feasibility study on retrieval of tropospheric/stratospheric columns of BrO and NO 2

Transcription

1 Feasibility study on retrieval of tropospheric/stratospheric columns of BrO and NO 2 Final report May 31, 2006 Piia Post 1, Leif Backman 2 1 Institute of Environmental Physics University of Tartu Ülikooli Tartu Estonia 2 Finnish Meteorological Institute Erik Palmen aukio 1 FI Helsinki Finland

2 Table of Contents 1. Introduction BrO in the atmosphere NO 2 in the atmosphere Methods to separate stratospheric and tropospheric columns Uncertainties in retrieval NO 2 retrieval uncertainties BrO retrieval sensitivity studies Air mass factors AMF sensitivity to BrO profile AMF sensitivity to clouds Conclusion about AMF sensitivity investigation Description of the residual method applied Data BrO and NO 2 total column data from GOME ERS UV albedos FRESCO clouds CTM FinROSE output CTM SLIMCAT output BrO profiles from SAOZ measurements NO 2 profiles from HALOE Stratospheric NO 2 climatology FinROSE-CTM validation Vertical column densities of stratospheric BrO Monthly mean zonal profiles of BrO concentrations Comparison with measured profiles NO 2 vertical column densities Monthly mean zonal profiles of NO 2 concentrations Comparison with measured profiles Conclusions about FinROSE-CTM validation Tropospheric columns: first results Conclusions and further studies Acknowledgements References List of Figures

3 1. Introduction From several nadir-measuring satellites GOME, SCIAMACHY, OMI it is possible to detect in addition to ozone total columns also columns of NO 2, BrO, OClO, HCHO, SO 2, O 4 and H 2 O. The column information on BrO and NO 2 has severe limitations from endusers point of view, since these gases exist both in the stratosphere and in the troposphere as related to different chemical mechanisms. Therefore, it is very important to be able to distinguish between the tropospheric and stratospheric parts of the total column. NO 2 and BrO are both photochemically active and their concentrations have a remarkable daily cycle. NO 2 columns order of magnitude is about 100 times higher than of BrO, and molecule/cm 2 respectively. Already Crutzen (1970) has described the importance of NO 2 in the atmospheric chemistry and since then long-term research has been performed to understand its differing role in the determination of the earth s ozone distribution. In the stratosphere it plays a role both as an ozone depleting substance and as a terminator in the halogen oxide ozone depleting cycles. In the troposphere, NO 2 participates in ozone formation via photochemical smog. While the main sources and source regions in the troposphere are known: thunderstorms and pollution from transportation, power plants and industrial sources, large uncertainties remain on the individual source strengths and their latitudinal and seasonal variations. Inorganic bromine is the second most important halogen affecting stratospheric ozone. During daylight the most abundant stratospheric bromine species, which accounts for 60-70% of total inorganic bromine is BrO (Lary 1996). BrO has an ozone-depleting role also in the troposphere (Platt and Hönniger 1998). But our knowledge of the mechanisms related to bromine in the atmosphere is incomplete, what is mostly caused by sparse or nonexistent measurements of most bromine compounds. The most feasible inorganic bromine species for detection is BrO, therefore experimental studies essentially rely on observations of it. There has not been made enough measurements to make a global BrO climatology: Bruns et al (2003) have compiled a climatology of BrO based on 3D chemistry transport model (CTM) SLIMCAT simulations, and only for stratosphere. Now, more than two years later the situation is improving: for validation of SCIAMACHY retrieval, stratospheric BrO profile measurement campaigns have been organised and the first results already published (Dorf et al 2005). 3

4 The main aim of this study is to develop a relevant methodology for operational retrieval of stratospheric/tropospheric BrO and NO 2 columns. The approach is to estimate the tropospheric columns from the residual of the total columns retrieved by satellite (GOME) and the stratospheric part given by a chemical-transport model. 2. BrO in the atmosphere Bromine monoxide (BrO) is an important trace species in the ozone chemistry because of its large efficiency as catalyst of the ozone destruction. Its role in the polar stratosphere has been studied for a long time after McElroy et al (1986) suggested the BrO-ClO cycle. Recent findings indicate that the ClO-BrO cycle is responsible for approximately 50% of the seasonal spring O 3 destruction (Chipperfield and Pyle 1998). Stratospheric bromine itself currently contributes about 25% to global ozone loss (Dorf et al 2005). The surface bromo-organic measurements suggest that stratospheric bromine is likely to have peaked around 2001 with inorganic Br close to ppt (Dorf et al 2005). It has long been assumed that reactive halogen species were confined to the stratosphere, playing a significant role only in polar region during spring. However, during the last few years, significant amounts of BrO were also observed in the troposphere at first by ground-based instruments and more recently from space by the Global Ozone Monitoring Experiment (GOME). The most common measurement technique is groundbased zenith sky UV-visible spectroscopy (Carroll et al 1989; Arpag et al 1994; Fish et al 1995; Aliwell et al 1997; Eisinger et al 1997; Kreher et al 1997; Otten et al 1998; Frieß et al 1999; Richter et al 1999; Hönniger and Platt 2002). The measured spectra of scattered light from zenith sky are analysed for BrO by the technique of differential optical absorption spectroscopy and the slant column (or apparent column) densities are received. In case of high solar zenith angles the predominant part of the total atmospheric column density is located in the stratosphere, therefore these measurements usually give evidence about stratospheric BrO, where mixing ratios have been measured in the range of 5-20 pptv. The stratospheric BrO profiles have been measured also by balloon-borne UV-visible spectroscopy (Harder et al 1998, 2000; Fitzenberger et al 2000; Pfeilstcker 2000; Pundt et al 2002) and in-situ resonance fluorescence spectroscopy both from aircraft (Brune et al 1989; Toohey et al 1990; Avallone et al 1995) and balloon (McKinney et al 1997). The tropospheric profiles have been measured the first time by LPMA/DOAS balloon gondola (Fitzenberger et al 2000). 4

5 Global Ozone Monitoring Experiment (GOME) measurements from the ESA ERS- 2 satellite have enabled for the first time to estimate globally the BrO vertical columns in the atmosphere (Wagner and Platt 1998; Richter et al 1998). From GOME measurements also enhanced amounts of BrO in the boundary layer (BL) over the polar regions in spring have been detected (Wagner and Platt 1998; Richter et al 1998, 1999b; Wagner et al 2000). These enhancements have been assigned to the BL because of complete ozone depletion in the BL at the same time. Later ground measurements have been carried out to investigate the boundary layer BrO during polar ozone depletion events (Avallone et al 2003, Frieß et al 2004; Hönniger et al 2004b), but also in mid-latitudes marine boundary layer (Leser et al 2003; Sainz-Lopez et al 2004). BrO has been found also in volcanic plumes (Bobrowski et al 2003) and over salt lakes (Hebestreit et al 1999; Stutz et al 2002; Hönniger et al 2004a). Conversely collocated ground-based and satellite column measurements show significantly more total atmospheric BrO (50-100%) than the integrated stratospheric balloon profiles can account for. Analysis of ground-based measurements for GOME validation gives that GOME measures more BrO than zenith-sky experiments do (Richter et al 1999, 2002; Van Roozendael et al 2002). This indicates a global tropospheric background estimated at 1-2 ppt (Platt and Hönninger 2003). The presence of significant amounts of BrO not only in the boundary layer, but also in the free troposphere was suggested also by McElroy et al (1999) relaying on the composition and photodissociative flux measurements instrument, which flew on board the NASA ER-2 high-altitude aircraft. There have been also first measurements confirming the assumption about ubiquitous BrO tropospheric column. Schofield et al (2004) have estimated from their measurements over Lauder, New Zealand, BrO tropospheric columns to be from 0.2 to 0.9 pptv. The mechanisms responsible for the production of reactive bromine in both the boundary layer and in the free-troposphere are not well understood at the moment, although the role of sea-ice and sea-salt aerosol has been clearly identified (Frieß et al 2004; Kaleschke et al 2004). Nevertheless, at the levels (up to 100 ppt) produced in the polar boundary layer during the so-called "polar spring bromine explosion events", it is clear that BrO has a strong impact on the tropospheric chemistry, being responsible for complete removal of the ozone within hours or days. Furthermore, the accumulating evidence for the presence of BrO in the free-troposphere of polar regions but also at mid-latitude, raises the question of the possible impact of reactive halogens on the tropospheric chemistry at the regional scale or even more widespread. Knowledge of spatial and temporal distribution of 5

6 stratospheric and tropospheric BrO enables to estimate contribution of different sources of bromine to the whole halogen chemistry in the atmosphere. First results from SCIAMACHY BrO measurements have been published by Afe et al (2004) and Sinnhuber et al (2005), from OMI by Kurosu et al (2004). 3. NO 2 in the atmosphere NO 2 plays a key role in both stratospheric and tropospheric chemistry and there are several reasons why an improved knowledge of the global distribution of NOx (NO+NO 2 ) is important: 1) In the troposphere the photolysis of the NO 2 results in the formation of O 3. NO 2 can be regenerated by catalytic cycles involving both organic and peroxy radicals (RO 2 ), the hydroperoxyradical (HO 2 ), the hydroxyl radical (OH) and both volatile organic compounds (VOC) and carbon monoxide (CO). Thus NO 2 plays a significant role in determining the oxidizing capacity of the troposphere. 2) The chemical budget of ozone in the troposphere is largely determined by the concentration of NOx. The knowledge of the ozone distribution and its budget is strongly limited by a severe lack of observations of NO and NO 2 in the troposphere. As tropospheric O 3 is also a significant greenhouse gas, NO 2 also contributes indirectly to the radiative forcing. 3) NOx and volatile organic compounds are emitted in large quantities due to human activities such as traffic, industrial combustion and biomass burning (Lee et al 1996). In the summer months this mixture produces photochemical smog. NO 2 is known to impact on human health and the environment both directly and through the production of O 3 (EPA 2000). The variability of NOx concentrations in the lower troposphere in industrialised areas and near biomass burning sites is very large. The few available point observations of NOx, on the ground or from aircraft measurements, are therefore difficult to translate to regional scale concentrations. 4) NOx is produced in significant amounts by the natural sources as lightning and emissions from soils. The residence time of NOx in the lower troposphere is short. Therefore observations of boundary layer NOx contain important information on the emissions of nitric oxide, and the trends in these emissions. Overall it is necessary to monitor and understand the global impact of this pollutant on the physics and chemistry of the atmosphere (Lauer 2002). 6

7 Only recently has been found, that NO 2 can build up a significant radiative forcing during periods with extremely elevated NO 2 levels in the troposphere (Solomon et al 1999). Such pollution events have been underestimated in duration and horizontal extension in the past. It was shown only recently that NO 2 -rich pollution hot spots can spread over large areas for several days (Leue et al 2001), and that NO 2 transport over large distances is possible (Stohl et al 2003; Schaub et al 2005). The quantitative contributions of both natural and anthropogenic tropospheric sources of nitrogen dioxide to the total NO 2 budget are not very well known, due to sparse ground-based measurements. Here, satellite-based measurements may contribute significantly to provide tropospheric nitrogen dioxide levels on a global scale (Lauer et al 2002; Velders et al 2001; Martin et al 2002). Recent airborne measurements substantiated that considerable contributions originate from lightning in thunderstorms and vertical transport of polluted air masses (ie NO 2 -rich air) from the planetary boundary layer to higher atmospheric levels (Huntrieser et al 2002; Brunner et al 2001). At the tropopause level further contributions from air traffic have also be taken into account (Schumann, 1994). Nitrogen oxide plays a number of important roles in the chemistry of the stratosphere. It is not only involved in catalytic cycles leading to ozone destruction, but also in the processes buffering active chlorine and oxides of hydrogen through the formation of temporary reservoirs, such as chlorine nitrate (Brasseur and Solomon 1986). Reservoir species like HO 2 NO 2 (peroxynitric acid) and N 2 O 5 (dinitrogen pentoxide) have the property to remove reactive species like NO 2 for a certain time from fast reactions. Both reservoirs can be photolysed to form NO 2 and NO 2 itself photolyses to form NO. This results in significant diurnal variation of NO 2 amount, with a minimum after sunrise and maximum shortly after sunset (when NO is rapidly converted to NO 2 ). The partitioning of N 2 O 5 also leads to a seasonal variation in NO 2 densities. During polar summer, the near-constant sunlight prevents build-up of NO 3 and hence precludes formation of N 2 O 5. Thus, NO 2 densities are higher in the polar summer than in the winter, at which time more NOx is sequestered in the N 2 O 5 reservoir (Solomon and Keys 1992). The life-cycle of NO 2 is determined by a fast exchange between NO and NO 2 and the slow formation of N 2 O 5. During the night, N 2 O 5 is in thermal equilibrium with NO 2 and NO 3, and also reacts on liquid or solid surfaces to form HNO 3. During the day, the partitioning between NO, NO 2 and N 2 O 5 depends strongly on solar zenith angle due to the rapid photolysis of NO 2 and the slower photolysis of N 2 O 5. This dependency on solar 7

8 zenith angle makes validation of NO 2 measurements difficult, as measurements seldom coincide both in location and local time. In the upper stratosphere NO 2 is mainly produced via atomic oxygen from nitrous oxide (N 2 O) and is subsequently transported poleward into the lower stratosphere. There, denitrification and transport into the troposphere contribute to the main loss processes. The daily averaged stratospheric NO 2 column is dominated by available solar insolation and shows a pronounced latitudinal dependency, while tropospheric NO 2 levels are highly variable in time and space. The short-lived radical NO 2 has been observed in the atmosphere since the 1970s by means of passive remote sensing in the ultraviolet and visible spectral regions from instrumentation either at the ground, or flown on aircraft, balloons, and spacecraft e.g. HALOE measurements. More details can be read about it in Bracher et al (2005). 4. Methods to separate stratospheric and tropospheric columns BrO and NO 2 are present in detectable amounts both in the troposphere and stratosphere. From satellite nadir measurements (e.g. GOME, SCHIAMACHY, OMI) the total slant column densities (SCD total ) of BrO and NO 2 can be retrieved using DOAS. Residual technique enables to process this total column product and retrieve the tropospheric part of the vertical column density (VCD trop ). The difference between total (SCD total ) and the stratospheric BrO slant column densities (SCD strat ) is attributed to tropospheric BrO. The tracer s tropospheric VCDs are then computed: VCD = ( SCD SCD ) / AMF. (1) trop total strat trop The air mass factor AMF - is the ratio of the slant column density (SCD) of the absorber (ie that viewed by the satellite in the measured radiance spectrum) to the vertical column density (VCD): SCD AMF =.. (2) VCD In Eq. (1) the AMF of troposphere is used. Investigators have made different assumptions about the stratospheric columns that are based on the observation that stratospheric tracers have a smooth spatial behaviour: 1. Leue et al (2001) have used Image Processing Technique (IPT) to get tropospheric NO 2 columns. Implicit assumption is that over oceanic, cloudy pixels, the retrieved 8

9 column is in fact the stratospheric column. This is the potential weakness of the method since tropospheric NO 2 may still be present above a cloud. 2. Reference Sector Method (Richter and Burrows 2002; Martin et al 2002; Randall et al 2002; Velders, et al 2001). The main tropospheric sources of NO 2 are either anthropogenic or thunderstorms, therefore the total column over a remote Pacific region is assumed as composed of only stratospheric NO 2. The tropospheric loading at any location can then be obtained as the residual of the total column and a latitude dependent stratospheric column of a remote Pacific region. The drawback of the method is that one needs to assume longitudinally homogeneous NO 2 and actually a small amount of tropospheric NO 2 may still be present in the reference pixels themselves. However, longitudinal variations can clearly not be neglected close to the Polar Vortex or during major changes in stratospheric dynamics, introducing some artifacts at high latitudes in winter and spring. 3. CTM stratosphere (Van Roozendael et al 2003; Theys et al 2005). The stratospheric column of BrO or NO 2 can be obtained from chemistry-transport model (CTM) simulations. The main advantage is that the model accounts for dynamical features in stratospheric NO 2 or BrO, but the drawback is that the retrieval will depend quantitatively on model. 4. Data assimilation (Eskes et al 2003). A CTM stratosphere is made consistent with the observations by assimilating the GOME NO 2 data. The advantage of this method over CTM stratosphere is that the dynamical features in stratospheric NO 2 are still predicted by the model, but that the model stratosphere is driven by the actual GOME observations. 5. As the variability of BrO is much lower in the stratosphere than in the troposphere, the short-term variability is ascribed to the troposphere while the annual component is taken as stratospheric variability (Hollwedel et al 2004). The method s weakness is that tropospheric BrO may have also annual cycle. The methods 1 and 2 do not suit for BrO because of the ubiquitous BrO tropospheric column. 5. Uncertainties in retrieval A major challenge is the derivation of good quality quantitative tropospheric and stratospheric NO 2 and BrO column amounts for individual ground pixels based on the 9

10 satellite data. The retrieval of tropospheric trace gas species is characterized by large uncertainties, related to clouds, the surface albedo, the trace gas profile, the stratospheric column of trace gas, and aerosols. Retrieval uncertainty estimates for vertical tropospheric NO 2 columns based on theoretical error source discussions combined with actual GOME observations are presented by Boersma et al (2004). They have done it in very detailed way, therefore on NO 2 error analysis we relay mostly on them. The retrieval of tropospheric BrO is very similar to retrieval of NO 2 : error sources are similar and of the same magnitude. The text about error sources for NO 2 retrieval is valid for both tracers. For GOME BrO retrieval we have made sensitivity calculations, what compliment error estimates NO 2 retrieval uncertainties Contributions to the total retrieval uncertainty are divided into three categories: (1) errors caused by measurement noise and spectral fitting, affected the slant column density; (2) errors related to the separation of stratopheric and tropospheric NO 2 affecting the estimate of the stratospheric slant column, and (3) errors due to uncertainty in model parameters such as clouds, surface albedo and a priori profile shape, affecting the tropospheric air mass factor. In brief these errors may be called slant column, stratospheric column and tropospheric air mass errors. These three components of errors contribute in different way to total error depending on geographical region and time of the year. NO 2 plumes are in a great deal of anthropogenic origin, this causes that over polluted areas, where tropospheric columns of NO 2 are high, the relative uncertainty in the retrieved column reduces to 35-60% and is dominated by the uncertainty in the tropospheric air mass factor. Over the oceans and remote continental regions, the overall tropospheric retrieval uncertainty is dominated by errors in the spectral fitting and the stratospheric column estimate. Residual method s relative uncertainty is very high for regions where tropospheric slant columns excess is small. Especially in this case very exact total and stratospheric slant columns are needed. A standard approach applied to GOME is based on the assumption that stratospheric NO 2 is zonally uniform, or at least has only a small longitudinal variation. Such simplification makes the retrieval of small tropospheric NO 2 columns practically impossible. As presented in Table 1 of Boersma et al (2004) for regular pixels systematic slant column errors are about 3%, for enhanced NO 2 areas these errors can increase up to 20%. Uncertainties in the estimation of the stratospheric vertical column density depend on 10

11 method used. Image processing techniques have errors in range 10% (Velders et al 2001) to 20% (Leue et al 2001). Reference sector method errors have been estimated to be in the range of 15% (Richter and Burrows 2002) and molecule cm -2 (Martin et al 2002). For chemical transport model atmosphere no error estimates have been given and for data assimilation methods the errors stays under molecule cm -2 (Boersma et al 2004). Now about uncertainties of the tropospheric air mass factor, what is the most critical source to total error. Air mass factor depends on conditions in the atmosphere: the a priori assumed profile shape, cloudiness properties, surface albedo, aerosol loading. Following estimates are all taken from Boersma et al (2004) for GOME NO 2 retrieval. 1. The largest uncertainties are due to clouds, as they will shield near-surface NO 2 from the view of the satellite. The retrieval depends very sensitively on the presence of clouds, and even small cloud fractions (between 5 to 20%) have a major impact. Mean uncertainty of tropospheric air mass factor due to cloud fraction may reach to 15% for heavily polluted areas, for clean regions it is less: 5%. Due to cloud top height uncertainties in tropospheric AMF are 2-3%. 2. The surface albedo directly influences the sensitivity of GOME for boundary layer NO 2. For heavily polluted areas mean uncertainty due to inaccurate surface albedo is 15%. The cleaner the atmosphere, the smaller is the uncertainty: for clean areas only 5%. 3. Profiles of NO 2 are characterized by a large range of variability. At emission areas the NO 2 concentration will peak at the surface, while downstream of such areas the pollution plume will peak at higher altitudes. Because of profile inexactness the uncertainties in tropospheric AMF are estimated to be 9 to 11 %. 4. A large source of uncertainty is aerosol. Thick aerosol layers influence the radiation field and the sensitivity of GOME for near-surface NO 2. Presence of aerosols modifies the top of atmosphere reflectance input to cloud retrieval schemes. Cloud algorithms typically do not account for aerosols, and overestimate cloud fraction and underestimate cloud top altitude values when a low atmospheric aerosol layer is added to a pure Rayleigh atmosphere. The increased cloud fraction and decreased cloud height tend to increase the light path through the troposphere and hence the tropospheric air mass factor. The neglect of aerosols in the air mass factor calculations is therefore thought to be partly compensated by the indirect effect of aerosols on the cloud retrieval scheme. Boersma et al (2004) have controlled this 11

12 assumption about compensation and conclude that even in case of large aerosol optical paths the radiative transfer aerosol correction factor and the actual retrieval cloud correction factor agree to within 10%. Martin et al (2003) suggested that NO 2 retrievals can be improved upon by directly accounting for aerosols, but they did not take into account the sensitivity of their cloud retrieval algorithm to aerosol. These errors are assumed to be mutually uncorrelated, since they arise from nearly independent retrieval steps. If to add up them all, then the total uncertainty may be higher than 100%. First comparisons of satellite derived tropospheric NO 2 with chemical models and ground-based UV/visible measurements have shown discrepancies up to 300% (Lauer et al 2002; Petritoli et al 2004). Obviously, the current satellite retrievals of tropospheric NO 2 are still of limited accuracy. Therefore intensive further joint investigations together with ground-based measurements are required targeting at both validation and synergistic use, i.e. an improved error assessment as well as an improvement of the satellite measurements themselves, by synergistically combining them with the complementary information attainable from ground-based measurements. One important improvement of SCIAMACHY as compared to GOME is the smaller ground pixel size. In this way the variability of NO 2 can be better resolved, and the fraction of cloud-free pixels will be larger, improving the quality of the retrieval BrO retrieval sensitivity studies Air mass factors The air mass factor is the ratio of the slant column density (SCD) of the absorber (ie that viewed by the satellite in the measured radiance spectrum) to the vertical column density (VCD): AMF SCD = VCD = c( z) w( z) dz c( z) dz, (3) where c(z) is the absorber concentration and w(z) weighting function, that both depend on the altitude z. Air mass factor describes the weighted path of the light through the atmosphere compared to the vertical path. is an integral measure of the sensitivity of the measurement, larger values indicating higher sensitivity. AMF computation is a pure radiative transfer simulation, as it represents a calculation of absorption paths in the 12

13 atmosphere and requires viewing geometry information and atmospheric extinction properties. In case of GOME measurements we can not consider simple geometric light path in the atmosphere as there is not only one, but multiple paths in this case. Therefore air mass factors (AMFs) were calculated using the radiative transfer model package libradtran developed by Mayer and Kylling (2005). This package enables to calculate radiances and irradiances applying the pseudospherical discrete ordinate model (DISORT) with refraction included and multiple scattering. The data from the US Standard Atmosphere 1976 have been used for the profiles of air pressure and temperature, O 3 and NO 2. All computations have been performed at a wavelength of 352 nm, representative for the wavelength region used in the BrO analysis AMF sensitivity to BrO profile AMFs have been computed for a number of different scenarios and are used to characterize the dependence of the measurement sensitivity on solar zenith angle, surface albedo, BrO profile and cloud parameters (height, thickness and type). The sensitivity of the BrO AMFs to other parameters like the profiles of air pressure and temperature as well as other trace species is much smaller (Richter et al 1999b; for NO 2 Hild et al 2002) and has not been considered in our calculations. Neither has been taken into account the variability of BrO concentrations in the atmosphere depending on the solar zenith angle. BrO profiles 80 Height /km/ TOTAL STRA TROP 0 0.E+00 5.E+06 1.E+07 2.E+07 2.E+07 3.E+07 3.E+07 4.E+07 4.E+07 BrO concentration /mol/cm3/ Figure 1. BrO profiles used in AMF sensitivity calculations. In the case of boundary layer BrO (BL), the layer with concentration of molecule/cm 3 locates at 0-1 km. TOTAL profile s tropospheric part is exactly the same as the TROP one. 13

14 In order to simplify the problem and due to current limitations in our knowledge of the BrO vertical distribution, we have considered the atmosphere as being made of three main successive layers: the stratosphere (10-70 km altitude), the free troposphere (2-10 km) and the planetary boundary layer (0-2 km). To evaluate the sensitivity of the BrO AMFs in each of these layers, sensitivity tests have been performed separately for three different profiles (Fig 1), one for the boundary layer (BL) one for the free troposphere (TROP), and one for the stratosphere (STRA). In the discussion we will assume, following Palmer et al (2001), that the total BrO AMF can be obtained from an average of the three individual AMFs weighted by the BrO column in each layer. It depends on the viewing conditions how deep we can see in the atmosphere from space. The lower is the sun the longer is its path in the atmosphere. As shown in Fig 2, the measurement sensitivity to stratospheric BrO increases with the solar zenith angle, with AMF reaching up to 6º at 80 SZA. In case of low sun, radiation can not penetrate low into the atmosphere, it reaches only to the stratosphere, therefore the existence of stratospheric BrO increases so much the amount of radiation that is scattered back to the satellite instrument. This is also the reason why the surface albedo and cloudiness conditions influence only slightly the stratospheric AMF (the signal from the lower atmosphere does not reach the instrument). When limiting the analysis to SZAs lower than 80, the different scenarios cause only some percent (3-6%) variances from the clear sky case. Given the other error sources in presence, the stratospheric AMFs can therefore be treated as a simple function of the SZA. 14

15 BrO profle STRA 6 BrO AMF at 352 nm km alb= km alb= km alb=0.8 Ci h=7km alb=0.8 clear alb= km alb= km alb= km alb=0.05 Ci h=7km alb=0.05 clear alb= SZA /degrees/ Figure 2. Statospheric BrO AMF-s depending on solar zenith angle, surface albedo and cloudiness situation. In the cases of free tropospheric BrO the AMF-s are much more sensitive to the surface albedo values and to cloudiness conditions. For the free tropospheric BrO case (Fig 3), a maximum of AMF is obtained at 74 degrees of solar zenith angle in clear sky condition and without thick clouds. The measurements are more sensitive to BrO if there is a highly reflecting surface under the BrO layer: in case of 0.8 albedo the AMF is 2 times higher than in case of albedo 0.05 (typical sea albedo). Cirrus clouds have nearly no influence on the AMF. Low (1 km) and not very thick (1 km) clouds increased the AMF: in case of albedo 0.8 about 5%, in case of albedo 0.05 nearly 2 times. Thick (5 km) and middle height (base at 3 km) clouds tend to decrease the radiative flux and in case of high surface albedo the AMF can be 2 times lower than in clear sky conditions. 15

16 BrO profile TROP 5 1-2km alb= km alb=0.8 BrO AMF at 352 nm SZA /degrees/ 3-8km alb=0.8 Ci h=7km alb=0.8 clear alb= km alb= km alb= km alb=0.05 Ci h=7km alb=0.05 clear alb=0.05 Figure 3. BrO AMF-s for the free troposphere depending on solar zenith angle, surface albedo and cloudiness situation. As to be expected, the calculations also show that the sensitivity towards the boundary layer BrO is very small if the surface albedo is low or if the pixel is covered by thick clouds. Then the AMF-s stay under unity (Fig 4). The situation improves significantly in case of high surface albedo. Then the AMF can rise up to 3.5. Calculations are carried out also for combined profiles in the case of what BrO locates through the atmosphere (Fig 5). High solar zenith angle cases have higher AMF-s what is still caused by the stratospheric part of the BrO profile. In case of 0.05 surface albedo the AMF is very sensitive to the tropospheric BrO profile: boundary layer BrO causes always about 2 times higher AMF- s than free tropospheric BrO. As most of the UV surface albedos are close to this 0.05 value, then it means that using wrong BrO profile will give us up to 2.5 times different BrO VCD-s. In case of high surface albedos, the AMF differs from 2.5 more than 20% in case of SZA-s higher than 65º and if there are opaque clouds over the boundary layer BrO. What means that for clear skies and high sun conditions over the high albedo areas (polar regions with snow and ice) is not very important to know exactly tropospheric BrO profile. But it means also that if we get high tropospheric BrO VCD-s over snow and ice regions, we can t say if BrO is in the free troposphere or in the BL. 16

17 BrO profile BL BrO AMF at 352 nm km alb= km alb= km alb=0.8 Ci h=7km alb=0.8 clear alb= km alb= km alb=0.05 Ci h=7km alb=0.05 clear alb= SZA /degrees/ Figure 4. AMF-s for the boundary layer BrO depending on solar zenith angle, surface albedo and cloudiness situation. BrO profile TOTAL 6 BrO AMF at 352 nm km alb= km alb= km alb=0.8 Ci h=7km alb=0.8 clear alb= km alb= km alb= km alb=0.05 Ci h=7km alb=0.05 clear alb= SZA /degrees/ Figure 5. AMF-s for combined total BrO profile depending on solar zenith angle, surface albedo and cloudiness situation. 17

18 AMF sensitivity to clouds In the case of GOME measurements the most important sources to the signal are single scattered and reflected photons. Single scattering takes place in the whole underlying atmosphere and reflection at the ground surface or at the clouds tops. In cloudy case there is added also multiple scattering inside the cloud and it is important to know how large is the signal coming through the clouds and how much is the light path length enhanced in the cloud by multiple scattering. The clouds influence on AMF was investigated calculating weighting functions. Surface albedo is 5% if it is not written otherwise. BrO profile is the TOTAL one from Fig 1, having both stratospheric and free tropospheric BrO. Following cloud properties are chosen: water content 0.15 [g/m 3 ], droplet radius 10 µm and the step inside the cloud km. In case of 50ºsolar zenith angle for lower levels in the stratosphere the difference in the weighting functions with and without cloud is about 10%, not depending on cloud height and thickness (Fig 6). For the high SZA (85º), the thickness of the cloud is important neither (Fig 7). But at the lower levels in stratosphere it is important if there is cloud or not: difference is more than 15%. The results of calculations for the same weighting functions profiles in case when cloud was considered as an opaque surface at some altitude with surface albedo 0.8 are presented in Fig 8. And there are very little differences between these two figures (7 and 8), what means that as our knowledge about daily cloud properties in global scale is usually not very precise, then it is better for calculations to use instead of clouds just highly reflecting elevated surfaces. This assumption does not certainly suit for cirrus clouds, but our calculations show very small influence of cirrus clouds on weighting functions. 18

19 BrO sensitivity to clouds (SZA 50) Height /km/ no cloud albedo 0.05 cloud 1-3 km cloud 1-2 km cloud 2-3 km cloud 1-5 no cloud albedo BrO AMF Figure 6. BrO AMF-s calculated for different cloudy scenarios with different cloud base height and thickness for the solar zenith angle 50º. Height /km/ BrO AMF sensitivity to clouds (SZA 85) no cloud albedo 0.05 cloud 1-3 km cloud 1-2 km cloud 2-3 km cloud 1-5 km no cloud albedo BrO AMF Figure 7. BrO AMF-s calculated for different cloudy scenarios with different cloud base height and thickness for the solar zenith angle 85º. 19

20 BrO AMF sensitivity to clouds (albedo 0.8) SZA Height /km/ km 1 km 2 km 3 km 4 km 5 km BrO AMF Figure 8. BrO AMF-s calculated for clouds that are taken as a reflecting surface at the hheight of 0-5 km from the surface for solar zenith angle 50º. The cloud effect to the integral AMF-s is shown in Table 1. If the surface albedo is small (0.05), then the cloud effect is also remarkable for the stratospheric AMF (stratosphere begins with 12 km). In the case of large albedo (0.8), the effect of clouds (even to total AMF) is not important, what means actually that the effect of clouds is actually the effect of boundary layer albedo and it is not very important at what altitude this highly scattering (or refracting) surface is. This conclusion may depend on properties of clouds and also on model with what the AMF is calculated. The cirrus clouds had very small effect also with this model. Table 1. Comparison of BrO air mass factors for different solar zenith angles, boundary surface albedos and BrO profiles in case of clear sky and 2 km thick low cloud. SZA=55º SZA=85º total stratospheric ratio total stratospheric ratio no cloud cloud 1-3km % effect of albedo 0.05 clouds no cloud cloud 1-3km albedo 0.8 % effect of clouds

21 Conclusion about AMF sensitivity investigation From what we can draw the following conclusions towards the needs for AMF calculations: 1) For stratospheric AMF calculation, we need to vary only the SZA. The ground albedo is taken the same everywhere (0.05). The BrO AMF variability caused by varing stratospheric profiles has been found to be in the range of 2%, what influences very little tropospheric BrO amounts, therefore the stratospheric BrO profile shown in Fig 1 is used for all cases. 2) The BrO AMF has been found to be very sensitive to ground albedo if there are large concentrations of BrO in the lower atmosphere. Therefore the tropospheric AMFs are calculated depending on both SZA and ground albedo. Tropospheric AMFs are also very sensitive to BrO profile. In case of large albedo, BL and free-tropospheric AMFs are similar in size, while BL AMFs are strongly reduced in conditions of low albedo. Simply said, this means that GOME is generally poorly sensitive to surface BrO except above regions of high albedo. As the BrO profile in the troposphere is not well known, the free troposphere profile (maximum concentration of BrO at 5 km) has been used for further calculations, meaning that BL layer BrO contents if present above regions of low albedo will be underestimated (by factor of 2 approximately). The influence of clouds on AMF can actually be reduced to the albedo effect of clouds. The effect is large for cases of low surface albedo, when instead of low reflector a highly reflecting surface is included below the BrO layer. Then the AMF of the area will be several times larger than before, what brings large differences in BrO VCD-s over all areas except snow and ice regions. Using data about cloudiness offers also possibility to locate the altitude of the BrO layer in the troposphere. 6. Description of the residual method applied We shall use residual method to separate from total column product stratospheric and tropospheric contributions. The BrO and NO 2 total slant column densities are retrieved from the GOME product of BIRA-IASB and KNMI. The slant column densities for the stratosphere are calculated from the FinROSE model VCD-s multiplying them by stratospheric AMF-s. Using on this level SCD-s helps to avoid one source of errors by using the same surface albedo for AMF calculations over the whole globe. This enables to divide the calculation of AMF-s into two: the calculation of the tropospheric AMF-s what are sensitive to several parameters in the atmosphere and stratospheric AMFs what can be 21

22 calculated more simply. The tropospheric BrO and NO 2 abundances will be calculated from the difference between measured total and stratospheric SCD-s. Finally, using tropospheric AMF-s the tropospheric SCD-s are converted into VCD-s. Details of data and calculations are given in following sections. 7. Data 7.1. BrO and NO 2 total column data from GOME ERS-2 GOME is a 4-channel spectrometer covering the spectral region from 230 to 800 nm with a spectral resolution of nm, with a main objective of global measurement of ozone columns (Burrows et al 1999). Other trace gases that have structured absorptions in the visible and UV like NO 2, OClO, SO 2, H 2 CO and BrO can also be retrieved from the spectra. There exist several GOME products of total columns of NO 2 and BrO, what differ in details. We used the BIRA-IASB GOME product for BrO and KNMI GOME product NO 2 slant columns. GOME data is available from Jul 1, 1995 to Jun 21, For NO 2, the data is available from TEMIS homepage ( for BrO from BIRA-IASB homepage ( For BrO, the analysis is conducted in the nm wavelength region where this molecule presents large vibrational structures. The spectral analysis of GOME spectra is performed using WinDOAS, a multi-purpose DOAS analysis software developed over the nineties at BIRA-IASB. This software initially developed for ground-based applications has been thoroughly validated through participation at various intercomparison exercises (Hofmann et al 1995; Roscoe et al 1999; Aliwell et al 2002). In the current state of the BrO algorithm, the inversion is performed in the nm spectral range and basically follows the recommendations issued in Aliwell et al (2002). The BrO absorption crosssections used are those from Wilmouth et al (1999), convolved to the GOME resolution, which is derived as part of the retrieval algorithm. The DOAS procedure accounts for the GOME undersampling (Chance, 1998) and slant columns are corrected for the GOME diffusor plate spectral artefact according to Richter et al (2002). More details on the DOAS procedure applied can be found in Van Roozendael et al (1999). BrO vertical columns are derived using Air Mass Factors (AMFs) calculated by means of a discrete-ordinate multiple-scattering radiative transfer model, under the assumption that BrO is located in the stratosphere and accounting for the line-of-sight angles of the GOME pixels. Tropospheric AMFs are generally smaller than corresponding stratospheric AMFs (except in case of 22

23 large surface albedo), which means that BrO product generally underestimates the BrO columns where contribution from the troposphere is significant. This effect is largest at low sun near the terminator. Details on the BIRA-IASB product can be found in Van Roozendael et al (1999). The retrieval of NO 2 is generally similar to retrieval of BrO. Difference is in the spectral region: the wavelength range nm is used. Good signal-to-noise ratio's (of about 20) are obtained for NO 2 with the DOAS retrieval technique. This is related to the absence of strong other absorbers (e.g. ozone) in this spectral interval. Total nitrogen dioxide columns derived from GOME, other satellite instruments (HALOE, POAM) and ground-based measurements agree in general within 5-20%, but within ± molecule cm -2 in areas of low NO 2 in the planetary boundary layer and within ± molecule cm -2 in areas of very low slant columns of nitrogen dioxide. Recent studies (Richter and Burrows, 2002; Martin et al 2002) showed that GOME nitrogen dioxide measurements partially suffer from instrumental artifacts that may cause a temporally varying offset in the slant columns up to molecule cm -2 depending on the time of the year. These structures, however, appear to be similar from year to year. For November 2001, a slight underestimation of retrieved NO 2 slant columns around molecule cm -2 can be estimated using the offset determined for 1998/1999 by Martin et al (2002) UV albedos The BrO air mass factors for the troposphere were very sensitive to the surface albedo, therefore we have used daily UV albedos. There are no large differences in the monthly mean tropospheric BrO VCD-s if to use climatological albedos or daily ones, but the advantage of using daily albedos comes out on daily bases if to take also clouds into account. One possible source for climatological UV albedos is the Earth surface reflectivity climatology from Nimbus-7/TOMS at the 380 nm (Herman and Celarier 1997). This climatology is obtained from the 14.5 year (November 1978 to May 1993) minimum Lambertian equivalent reflectivity (LER) values for 1 X 1.25 (latitude X longitude) pixel, with outlier values removed. In general the reflectivity is lower over the land (2-4%) than over the oceans (5-7%). Over land, permanent features with greater LER than average are associated with permanent glaciers, high desert, and regions with large dry lake and stream beds. There are some regions of the Earth where persistent cloud cover is present 23

24 throughout a month, no attempt has been made to remove those features. The reflectivities are nearly independent of wavelength (the difference in the region 340 to 380 nm is less than 0.2 %), therefore they could be used also for 352 nm region without restrictions. The using of climatological albedos instead of real daily ones can cause large errors in tropospheric BrO VCD-s over the regions of varying snow and ice cover. Daily UV albedos are computed by Tanskanen et al (2003) using a moving timewindow technique to analyse the TOMS Lambertian equivalent reflectivity data for the years and The method treats the measured reflectivity data as a sample from a reflectivity distribution whose lower tail corresponds to surface albedo. The basic method assumes that the distribution is homogeneous, ie. surface albedo is constant within the window. Adequate statistics is achieved only by using a wide time-window what, unfortunately, leads to underestimation of the surface albedo during spring and autumn transitions. Therefore, the method was developed further to account for transitions. However, the method assumes no absorbing aerosols and fails in regions with persistent cloud cover. We have used the daily albedo values for tropospheric AMF-calculations, only in case the daily values missing, the monthly ones from Herman and Celarier (1997) were used FRESCO clouds FRESCO effective cloud fractions and cloud top pressures from (Koelemeijer et al 2001) are used to weight the AMF-s for partly cloudy pixels. This algorithm, called Fast Retrieval Scheme for Clouds from the Oxygen A band, makes use of reflectivities as measured by GOME or SCIAMACHY inside and outside the oxygen A band ( nm). Three onenanometer wide parts of the oxygen A-band spectrum are used in the FRESCO near-real time retrieval, both inside and outside the oxygen A-band, namely at 758 nm (no absorption), 761 nm (strong absorption), and 765 nm (moderate absorption). The reflectivity outside the oxygen A-band is almost independent of cloud top pressure, but depends mainly on cloud fraction, cloud optical thickness, and surface albedo. The reflectivities inside the band depend on cloud top pressure as well, and are used to derive cloud top pressure. Using non-linear least-squares fitting of a measured spectrum to a simulated spectrum for each GOME pixel an effective cloud fraction and cloud top pressure are derived. 24

25 The effective cloud fractions are derived by assuming that the clouds have an albedo of 0.8, and must therefore be interpreted as effective cloud fractions. Note that the derived cloud top pressures are rather insensitive to the assumed cloud albedo. For areas with effective cloud fractions smaller than 0.05, cloud top pressures cannot be derived reliably. There are two operation modes, depending whether the surface is assumed to be free of snow/ice cover or not. The snow/ice mode is used if according to TOMS UV minimum surface reflectivity climatology (Herman and Celarier 1997) the surface is brighter than 0.2 for a given month. No attempt is made to account for the presence of snow, ice, or sun-glint. Thus if cloud-free land or ocean is covered by snow or ice shelves, or if a pixel is affected by sunglint, these areas will show up as having low-altitude clouds with high cloud coverage. The AMF of clouds is taken at the level of effective cloud top pressure. In case of large surface albedo (0.8), the effect of clouds (even to total AMF) is not important. The effect of clouds is actually the effect of highly scattering boundary surface and in case of high boundary layer albedo it is not very important at what altitude this surface is situating CTM FinROSE output The spatial distribution and the temporal evolution of stratospheric NO 2 and BrO are simulated using the 3D global chemical-transport model FinROSE (version 1.2) (Damski 2005). The FinROSE-CTM is a global 3d grid point model based on the NCAR ROSE model (e.g. Brasseur et al 1997). The model covers the relevant gas-phase stratospheric chemical processes. Heterogeneous processes on polar-stratospheric clouds and in sulphate aerosols are also included in the model. In total it accounts for almost 200 reactions, including oxygen, hydrogen, carbon, nitrogen, chlorine and bromine species. The chemical rate constants and cross-sections are taken from Sander et al (2000, 2002). Photolysis rates are derived from a look-up table depending on solar zenith angle, ozone column and altitude. The look-up tables have been compiled using PHODIS-radiative transfer model (e.g. Kylling et al 1997). The chemical rate equations are solved by considering a chemical equilibrium state for short-lived species (e.g. ClO, NO 2, OH, BrO). A semi-implicit scheme is used for the integration of the more stable reactants (e.g. HNO 3, N 2 O): All short-lived species are grouped and integrated using families e.g. ClOx=Cl+ClO+Cl 2 O 2, NOx=NO+NO 2. All long-lived species and families are transported using the semi- Lagrangian flux-form scheme of Lin and Rood (1996). The bromine chemistry included in 25

26 the model is shown in Fig 9. ECMWF meteorological data are used as input for simulating the stratospheric chemistry and dynamics. The output for GOME overflow time (10:30 LT) is available for 37 latitudes and 36 longitudes for the time period 1996 to The model output was interpolated onto a vertical grid of 43 log-pressure levels between 0 and 60 km altitude, resulting in a vertical step size of 1.3 km. In the model s troposphere all chemical species are given as boundary conditions. Special output was made for balloon measurement comparison with output every 30 minutes at the grid point where the measurement site is located. A special output was also made for HALOE measurement comparison, the closest grid point was chosen and at the same time the solar zenith angle was checked as the HALOE measurements are carried out during the sun rise and sun set. The BrO and NO 2 output has not been validated in detail prior to this work. Heterogeneous chemistry - LBA, STS, NAT, NAT- HOBr (g) (s) HCl rocks and ICE- particles - Sedimentation hν H 2 O (s) HET (g) O( 3 P) HO 2 (>98%) (g) HET hν (29%) BrONO 2 hν (71%) NO OH/O( 3 P)/O( 1 2 D) O 3 /OClO ClO HBr Br BrO BrCl HO 2 /CH 2 O O( 3 P)/NO/ClO/BrO/OH/hν O( 3 P) hν OH/HO 2 (<2%) Cl (s) (g) HOCl HET (g) ClONO 2 Figure 9. A schematic of the bromine chemistry described in FinROSE-CTM (HET= heterogeneous reactions, (s)=from solid phase, (g)= from gas phase) CTM SLIMCAT output At our disposal were global distributions of stratospheric BrO and NO 2 loadings simulated using 3D Chemical Transport Model SLIMCAT. SLIMCAT is a widely used threedimensional off-line chemical transport model that is described in detail by Chipperfield 26

27 (1999). The model temperatures and horizontal winds are specified from UKMO meteorological analyses and the vertical transport in the stratosphere is diagnosed from radiative heating rates. In the stratosphere the model uses an isentropic coordinate extended down to the surface using hybrid sigma-theta levels. The troposphere is assumed to be wellmixed. The model contains a detailed chemical scheme and simulates the distribution of all species involved in stratospheric ozone depletion. BrO and NO 2 VCD-s are calculated for each day for 36 latitudes and 48 longitudes at satellite "overpass" times. We have used two versions of the model output. For the first model version time period is Oct 25, 1996 to Feb 16, The model uses isentropic levels, ranging from 335 to 2700 K, corresponding to an altitude range of approximately 10 to 55 km. The second version of model also has levels in the troposphere and the available time period was four full years from 1996 till Sinnhuber et al (2002) have done validation of SLIMCAT model (our version 1) calculations against ground-based UV-visible measurements of BrO from a network of observing stations. The measurement sites span from the Arctic over northern and southern hemisphere (SH) mid-latitudes to Antarctica. Seasonal and, to some extent, interannual variation of BrO during a period of two and a half years from January 1998 until June 2000 were investigated. In order to compare directly observations and model calculations, a radiative transfer model was coupled to the chemical model to calculate simulated slant column densities. The conclusion was that the model reproduces the observations in general very well. Taking into account the estimated accuracy of the measurements of about 20% as well as the uncertainties in the model, the absolute amount of the BrO slant columns was consistent with total stratospheric bromine loading of 20±4 ppt for the period The seasonal and latitudinal variations of BrO are well reproduced by the model. In particular the good agreement between the observed and modelled diurnal variation provides strong evidence that with SLIMCAT the BrO-related bromine chemistry is correctly modelled BrO profiles from SAOZ measurements The SAOZ-BrO Balloon experiment is a lightweight UV optical sonde specifically designed for the measurement of BrO by solar occultation from small and relatively inexpensive balloons. The data are spectrally analysed by the DOAS method and vertical profiles are retrieved by an onion peeling procedure (Pundt et al 2002). There exist more 27

28 than 20 measured profiles of which 9 were available to us. The dates and the locations of these flights are presented in Table 2. Table 2. Date and location of the available SAOZ-BrO flights. The location with respect to the polar vortex (In, Out) and the chlorine activation as indicated by OClO is also given. The SZA is given at 15 km during the balloon ascent, when reaching float altitude and at noontime. Date Location Latitude Longitude Vortex SZA at 15 km SZA at float SZA at noon Andoya N E In, not activated Kiruna N E Kiruna N E Aire 43.7 N 0.25 W Kiruna N E Kiruna N E In, not activated In, not activated Kiruna N E Out Kiruna N E Kiruna N E In, not activated In, not activated NO 2 profiles from HALOE The Halogen Occultation Experiment (HALOE) was launched on the Upper Atmosphere Research Satellite (UARS) spacecraft September 12, 1991, and routine observations began science October 11, The experiment uses solar occultation to measure vertical profiles of O 3, HCl, HF, CH 4, H 2 O, NO, NO 2, aerosol extinction at 4 infrared wavelengths, and temperature versus pressure with an instantaneous vertical field of view of 1.6 km at the Earth's limb. Latitudinal coverage is from 80 S to 80 N over the course of 1 year and includes extensive observations of the Antarctic region during spring. The altitude range of the measurements extends from about 15 km to km, depending on the species. HALOE measurements are performed during local sunrise or sunset and it corresponds to a good approximation to a solar zenith angle of 90. We have used version 28

29 19 (v19) of HALOE NO 2 data product. Gordley et al (1996) have extensively validated the v17 and the quality of v18 data is characterized on the HALOE web page ( v18 data agree with correlative observations from 25 to 45 km within the ±10 to ±15%level with no obvious bias. The NO 2 data are described as excellent from the tropopause to 25 km in clear air conditions, but exhibit a low bias in the presence of aerosols. The aerosol correction in the lower stratosphere below about 20 km is large, being more than 100%. However the data from last five years should be more accurate because the aerosol loading is at its lowest since 1978, because of the lack of volcanic intrusions into the stratosphere. The vertical resolution of Haloe NO 2 data is given with 2 km (Gordley et al 1996) Stratospheric NO 2 climatology This database was generated by J.-C. Lambert and J. Granville (Lambert et al 1999) at IASB-BIRA. There are two files "HarmonicNO2_Sunrise.dat" and "HarmonicNO2_Sunset.dat", that contain a climatological parameterisation of stratospheric NO 2 number density at sunrise and at sunset, representative of volcanism-free conditions. They were generated using a Fourier harmonic decomposition of UARS HALOE v19 and SPOT-4 POAM-III v2 NO 2 profile data records and complementary information provided by NDSC and ERS-2 GOME NO 2 column data records. Fourier coefficients have been derived using a least-square analysis after removal of data of questionable quality. This climatology is intended primarily as an input to scattered-light air mass factor and retrieval studies in which the actual profile shape plays a crucial role. Therefore the NO 2 vertical distribution has not been smoothed. Based mainly on solar occultation and scattered-light nadir measurements, the validity domain of the parameterisation is obviously limited to the illuminated part of the atmosphere. Any attempt to retrieve NO 2 density during polar night might result in negative or aberrant values. Data at the lowest and highest altitudes might lack of accuracy due to increasing errors in satellite data. We have used this climatology for NO 2 profile comparisons. In general, harmonised profiles were good approximations of HALOE measurements. 8. FinROSE-CTM validation Uncertainties in retrieval of tropospheric columns of BrO and NO 2 depend highly on exactness of retrieval of stratospheric columns especially in regions where tropospheric 29

30 slant columns excess is small. Therefore it is very important to validate the stratospheric columns that are calculated by CTM FinROSE. For FinROSE output validation we compare it with output of SLIMACAT model, but also with measured NO 2 and BrO columns. We also compare absorber s profiles of model with measured ones, to see where the largest differences are. There are three issues what have to be taken into account before a direct comparison is possible: i) The stratospheric NO 2 and BrO diurnal cycle hinders a direct comparison, therefore model output should be taken out at the time of satellite overpass, ii) model output is for quite large (5ºx10º) grid-areas compared to in-situ profile measurements, it means that we compare ground-based data taken at one location to model data within a certain grid-area around that site, and, iii) model does not give only stratospheric column, but has levels also in the troposphere. There are remarkable concentrations of BrO and also NO 2 in the tropopause region, therefore it is important to determine the tropopause height as exactly as possible Vertical column densities of stratospheric BrO The easiest way for model output validation is by looking at the total vertical abundances of species. The chemistry transport model output at 24 levels for FinROSE-CTM (or 18 levels for SLIMCAT) in the atmosphere was integrated over pressure to get the vertical column densities of BrO in molecule/cm 2. Zonal and monthly averaged maps of stratospheric BrO VCD-s for 4 successive years beginning with 1996 are presented in Fig 10. SLIMCAT version 2 is exposed. Not a remarkable interannual variability is expected here, as the stratospheric bromine chemistry is mostly depending on solar radiation variability. For the tropical zone in all years the FinROSE BrO VCD is around molecule/cm 2. For polar summer the values are nearly the same molecule/cm 2, but for polar and midlatitude winter and spring the values are higher: up to molecule/cm 2. For the Arctic and Antarctic the highest values are for year The first year (1996) results may be affected by model spin-up (the model run was started from 1995). The VDC-s from FinROSE-CTM are generally lower compared to SLIMCAT (the scale of the colour bars is different for two models). There is a sharp transition from tropics to mid-latitudes at the 30th latitude in the FinROSE-CTM model output. The annual cycle of BrO VCD over the higher latitudes of the Northern Hemisphere (NH) is similar for two models: the highest values are for March (Fig 11). For the Southern Hemisphere two models give quite different results. For SLIMCAT the highest values of BrO VCD-s are in 30

31 the 50º S - 70º S latitudinal zone with small annual cycle amplitude, while for FinROSE the maximum VCD-s are around 70º S - 80º S in September-October. The difference in BrO monthly and zonal mean VCD-s between two model outputs exceeds 50%. Data in Figures 10 and 11 cover only stratosphere. As both CTM-s have also some levels in the troposphere, therefore the tropopause altitude has been detected. The SLMICAT BrO profiles (Fig 12) show remarkable abundancies of BrO in Upper troposphere/lower stratosphere region, what makes the tropopause detection very crucial. It is not a large problem for FinROSE BrO as the abundancies in the troposphere are low. FinROSE s BrO mixing ratios in the troposphere are not realistic values, but some climatological mean. In SLIMCAT output the maximum BrO area extends just over the UT/LS region. A quite rough 2 PV SI unit was chosen to separate tropospheric and stratospheric columns of BrO and also NO 2. In Fig 11 stratospheric BrO VCD-s from both SLIMCAT versions are shown (version 1 has no troposphere and from version 2 we have eliminated tropospheric column) and on monthly mean level no difference can be detected Monthly mean zonal profiles of BrO concentrations To detect the reason for differences in BrO VCD-s of FinROSE-CTM and SLIMCAT output, monthly zonal mean vertical profiles of BrO concentrations were compared. Four months representing different seasons for the year 1997 are presented in Fig 12: March, June, September and December. SLIMCAT: The maximum BrO concentration of around molecule/cm 3 is seen in March and September for the spring hemisphere polar latitudes at the altitude around 15 km. The concentration increased already during the winter at higher latitudes outside the area within the polar night. The lowest values for the stratospheric maximum are for tropical latitudes, where the annual cycle is weak and the concentration at 20 km altitude is around molecule/cm 3. For polar regions the lowest values are seen in the summer hemisphere, with a maximum concentration of molecule/cm 3. A dipole is observed with higher values mostly over the polar regions and all year round lower values over tropics. The height of the maxima is shifted in accord with the altitude of tropopause. Comparing the FinROSE-CTM and SLIMCAT output the following differences can be seen: The maximum in the profile over polar latitudes is located at higher altitudes in FinROSE-CTM, and the peak concentrations are lower. Both deviations result in a lower total column of BrO VCD over high latitudes for FinROSE-CTM than for SLIMCAT. 31

32 Figure 10. Stratospheric BrO vertical column densities from CTM output in molecule/cm2. 32

33 Figure 11. Annual cycle of BrO VCD-s in different latitudinal zones from FinROSE and SLIMCAT model output. Both versions (1 and 2) of SLIMCAT model are presented. 33

34 Figure 12. Comparison of monthly and zonal mean vertical profiles of BrO concentrations (in 10 7 molecule/cm3) for two models: SLIMCAT and FinROSE-CTM for 4 months in year

35 Figure 13. Comparison between zonal and monthly mean BrO concentrations (in 10 7 molecule/cm3) for different altitudes from FinROSE and SLIMCAT models for September

36 The maximum concentration in the profile at tropical latitudes is also located at higher altitudes in FinROSE-CTM, with higher values seen in the FinROSE-CTM output. The vertical gradient of the BrO concentration is steeper in the FinROSE-CTM. In spite of higher maximum concentrations in the tropical stratosphere the VCD values are still lower in the FinROSE-CTM (Fig 10). In Fig 13 very high (up to molecule/cm 3 ) BrO concentrations can be seen at 21 km altitude for latitudes 30ºS to 30ºN. This feature is connected with the total inorganic Br profile, which is modified in the tropics. There was no tropopause definition and therefore the profile shape was unrealistic. The model outputs differ most in the upper troposphere - lower stratosphere region. At higher altitudes (25-28 km) where pure photochemical equilibrium settles the BrO concentrations the difference between FinROSE and SLIMCAT model outputs is below 10% Comparison with measured profiles The SAOZ profiles can be divided into 5 different seasons/zones: Arctic winter (AW), Arctic summer, mid-latitude winter, mid-latitude summer and tropical summer (Pundt et al 2002). The majority of the profiles that were at our disposal have been measured at Kiruna during the winter, i.e. the profiles assigned to the AW subdivision. In Fig 14a three of these profiles are presented. Profiles for the same location simulated with FinROSE-CTM are presented in Fig 14b. The FinROSE-CTM profiles were written every 30 minutes at the grid points of the measuring stations for dates coinciding with measurement dates. The northern latitude of Kiruna is 67.88, which is located between the grid points of the model (65 N and 70 N). The sounding profile has been compared to model output at both 65º and 70 N depending on the solar zenith angle and the location of the vortex (Fig 14b). The measured profiles in Fig 14a appear similar. The largest differences are about 4 ppt-s, but mostly the difference stays inside 2 ppt-s. Higher than 20 km the difference stays inside the limit of error bars. The profiles are simulated up to ~65 km, but as the measured profiles reach only around 30 km, only the most interesting upper troposphere-lower stratosphere layer is shown. The model underestimates the mixing ratio of BrO. A clear difference can be seen at the lower levels, up to 16 km the vertical gradient of BrO mixing ratio is about two times less. The total inorganic bromine profile and the lower boundary condition should be improved. Furthermore, the aerosol vertical distribution will be revised. A steep gradient is seen in the simulated profile and at around 18 km an intermediate maximum is reached. Some kilometres higher the local minimum can be found. In this lower stratospheric layer (18-24 km) the solar zenith angle plays important 36

37 role: the altitude of this minimum-maximum pair from 65º N and 70º N profiles differs some kilometres. In case of higher sun elevation the local maximum of BrO mixing ratio is lower in the atmosphere. Arctic winter SAOZ height (km) SZA= SZA= SZA= E E E E E E E E E E E-11 BrO mixing ratio Arctic winter FinRose height (km) UT SZA= UT 70N SZA= UT 65N UT 70N SZA= UT 65N 8 0.0E E E E E E E E E E E-11 BrO mixing ratio Fig 14. a) Measured (SAOZ) and b) simulated (FinROSE-CTM) BrO mixing ratio profiles during the Arctic winter To measured profiles the error bars are included. 37

38 Arctic winter FinRose height (km) SZA= SZA= SZA= SZA= SZA= SZA= ,0E+00 2,0E-12 4,0E-12 6,0E-12 8,0E-12 1,0E-11 1,2E-11 1,4E-11 1,6E-11 1,8E-11 2,0E-11 BrO mixing ratio Arctic winter SLIMCAT height (km) E E E E E E E E E E E-11 BrO mixing ratio Figure 15. Simulated profiles for the Arctic winter 1997 from two CTM-s, a) FinROSE- CTM and b) SLIMCAT. In Fig 15 the output from two models is compared for profiles given for GOME overpass time 10:30 LT at dates coinciding with the SAOZ measurements. For FinROSE-CTM the profiles are from the 70º N grid-point and for SLIMCAT from 67.36º N. The solar zenith angle (SZA) for SLIMCAT is not given, but as the difference in grid-point centres of the models is around 2.5º, the SZA for SLIMCAT is about 2.5º smaller than for FinROSE- CTM. The variability seen in the profiles is quite small, but also the variability in SZA is only 13 degrees. The FinROSE-CTM BrO profile for Jan 21, 1997 differs from the others due to a very low sun. In the SLIMCAT output the structure of the profile for Feb 6,

39 is more complex than the other, the reason can be connected to polar stratospheric clouds. The FinROSE-CTM output systematically shows minimum around km, which may be connected to the total inorganic bromine profile. In Fig 16 the BrO mixing ratio vertical profiles are given for three different seasons/zones: a) Arctic winter, b) Arctic summer, c) mid-latitudes winter are presented. The profiles are from 0 to 70 km altitudes. The SAOZ measurements are usually made at sunset, but the SLIMCAT model results are for 10:30 am local time. At polar latitudes in winter/spring the sun rises close to ERS-2 overpass time, which means that the solar zenith angles for SAOZ and GOME measurements are similar (Fig 16a). However, it is important to note that the BrO mixing ratio profiles do not depend only on the SZA, but also whether it is sunrise or sunset. Therefore, in this figure at least two different BrO profiles from FinROSE-CTM are shown. In case of FinROSE-CTM the sunset profile (legend pm ) is included and this profile should be compared with SAOZ measurements. FinROSE-CTM am profiles are presented for comparison with SLIMCAT output. As from the model run the SZA is not exactly the same as for SAOZ measurements, several pm profiles are shown. To look at the interannual variability in simulated profiles, the output for two separate years is presented if available. Profiles of SLIMCAT (and also of FinROSE-CTM for GOME overpass time) mostly agree well with SAOZ profiles. However, as the conditions of measurement and simulation are different: SZA-s for GOME overpass are higher than during SAOZ flights and SLIMCAT profiles are morning ones against SAOZ evening ones, no big conclusions can be made of it. There is evidence that the morning BrO abundances should be higher than the evening ones, what brings to the conclusion that also SLIMCAT CTM underestimates BrO abundances. Changing SZA in case of FinROSE-CTM changes the BrO mixing ratio profile, but the best coincidence of FinROSE-CTM BrO mixing ratio profiles with SAOZ measurements gives the simulation for GOME overpass time (Fig 16 a, b, c). The interannual variability of SLIMCAT profiles is not remarkable: in Fig 16c simulations of 1996 and 1997 look very similar. For FinROSE-CTM the variability is larger (Fig 16a). 39

40 Andoya 2 March 1998 height (km) SAOZ SZA=84.9 Fin98 pm SZA= Fin98 pm SZA= SLIMCAT 1997 Fin97 am SZA=80.4 Fin98 am SZA= E E E E E E E E E-11 BrO mixing ratio Kiruna 28 Aug SAOZ sza=87.2 Fin98 pm SZA= Fin98 pm SZA= SLIMCAT 1997 Fin97am SZA=63.4 height (km) E E E E E E E E E E-11 BrO mixing ratio Aire 2 December 1998 height (km) SAOZ SZA=83.8 Fin98 pm SZA= SLIMCAT 1996 SLIMCAT 1997 Fin97 am SZA= E E E E E E E E E-11 BrO mixing ratio Figure 16. Comparison of BrO mixing ratio vertical profiles for a) Arctic winter, b) Arctic summer and c) mid-latitudes winter. See text for details. 40

41 8.4. NO 2 vertical column densities Vertical column densities of NO 2 (in molecule/cm 2 ) integrated from CTM output of FinROSE and SLIMCAT in the stratosphere are presented in Fig 17. Zonal and monthly averaged maps of stratospheric NO 2 VCD-s for 4 successive years beginning with 1996 are exposited. Largest interannual variability is detected over the polar regions. The highest values there occur at the first year For wide low-latitude zone (40ºS-40ºN) the FinROSE NO 2 VCD is around molecule/cm 2, without remarkable annual cycle. For polar winters the values are nearly the same molecule/cm 2, but for polar and midlatitude warm season the values are higher. The annual cycle maximum is over the warm season: for northern hemisphere high latitudes in April to September the values are up to molecule/cm 2, over southern hemisphere warm season only up to molecule/cm 2. Only for the first year (1996) FinROSE SH results are nearly as high as over the NH, but the first year results may be affected by model spinup. The VDC-s from FinROSE-CTM compared to SLIMCAT are generally lower. The annual cycle of NO 2 VCD over the higher latitudes of both hemispheres is similar for two models (Fig 18). But for SLIMCAT the latitudinal zone where the annual cycle is detected is wider: beginning from 30º and going towards poles, compared to 45º- 90º for FinROSE. The only zone where FinROSE has higher values is the 10 degrees circle around North pole. NO 2 VCD-s of SLIMCAT have nearly the same magnitude of maximum during warm season over both polar regions molecule/cm 2. The difference in NO 2 monthly zonal mean VCD-s between two model outputs is in average molecule/cm 2. In Fig 18 results from both SLIMCAT version outputs are presented, and no remarkable difference can be detected, what means that the extraction of tropospheric column from version 2 values was successful. 41

42 Figure 17. Stratospheric NO2 vertical column densities (in molecule/cm2) from FinROSE and SLIMCAT CTM output. 42

43 Figure 18. Annual cycle of NO 2 VCD-s (10 15 molecule/cm2) in different latitudinal zones from FinROSE and SLIMCAT model output. Both versions (1 and 2) of SLIMCAT model are presented. 43

44 8.5. Monthly mean zonal profiles of NO 2 concentrations To detect the reason why FinROSE NO 2 VCD-s tend to be lower than SLIMCAT output, the monthly zonal mean vertical profiles of NO 2 concentrations for the year 1997 were compared. Four months representing different seasons are presented in Fig 19: March, June, September and December. SLIMCAT: The maximum NO 2 concentration of around molecule/cm 3 is seen in December and March for the SH polar latitudes at the altitude around km. The lowest values for the stratospheric maximum are for tropical latitudes, where the annual cycle is weak and the concentration at 28 km altitude is around molecule/cm 3. For polar regions the lowest values are seen at the cold season, with a maximum concentration of molecule/cm 3 for at 25 km of SH and molecule/cm 3 at 28 km for NH. The maximum NO 2 concentration layer during warm season extends from poles to 20º at both hemispheres. Comparing the FinROSE-CTM and SLIMCAT output the following differences may be seen: In the FinROSE data (Fig 19) unrealistically high (higher than molecule/cm 3 ) NO 2 concentrations can be seen from 13 to 20 km altitude in NH latitudinal belt 70º-90º in June and 80º-90º in September. At other latitudes the peak concentrations are too small. The maximum concentration in the NO 2 profile is located at lower altitudes in FinROSE-CTM (Fig 19). The lower position of FinROSE NO 2 concentration maximum can be seen also in Fig 20. At two first altitudes (28 km and 25 km) SLIMCAT NO 2 concentrations are overall higher than the FinROSE ones, at lower altitudes it is vice versa, but the differences at lower altitudes are smaller. This results in lower VCD-s of NO 2 for FinROSE CTM than for SLIMCAT. The outputs of two models differ mostly because of the different height of the maximum mixing ratio. In general, for FinROSE the vertical gradients of NO 2 are smaller than for SLIMCAT. For SLIMCAT the absolute maximum concentrations are in the SH warm season polar latitudes, but the FinROSE overall maximum is at NH high latitudes. 44

45 Figure 19. Comparison of monthly and zonal mean vertical profiles of NO 2 concentrations (in 10 9 molecule/cm3) for two models: SLIMCAT and FinROSE-CTM for 4 months in year

46 Figure 20. Comparison between zonal and monthly mean NO 2 concentrations (in 10 9 molecule/cm3) from FinROSE and SLIMCAT models for different altitudes for September

47 8.6. Comparison with measured profiles FinROSE NO 2 profiles have been compared to HALOE sunrise and sunset measured ones during the year Because of significant diurnal variability of stratospheric NO 2, the FinROSE modelled profiles are derived for the same time as HALOE measurements. Nevertheless, because of quite large model grid areas, the coordinate of the measured profile and the model grid point can differ up to several degrees. The coordinates of the measured profile (lat, lon) and of FinROSE grid area (finlat, finlon) are shown also in Figures The solar zenith angle is very important variable to determine the height of the NO 2 maximum and is therefore included in figures. The solar zenith angle in figure is of FinROSE. Unfortunately not all latitudes are represented in measurements equally, most of the measurements are from middle latitudes. In figures are included also the NO 2 profiles that are calculated using harmonic parameterisation of Lambert and Granville (1999). Figure 21. Comparison between FinROSE modeled and HALOE measured profiles of NO 2 concentrations (in 10 9 molecule/cm3) for winter. 47

48 Profiles comparison figures are compiled for four main seasons at two hemispheres and measurements from both modes: sunrise and sunset are presented. The same features as from monthly mean profile comparison with SLIMCAT model (what was done for GOME satellite overpass time 10:30 LT) come out also for this measurement comparison. Figure 22. Comparison between FinROSE modeled and HALOE measured profiles of NO 2 concentrations (in 10 9 molecule/cm3) for spring. 48

49 Figure 23. Comparison between FinROSE modeled and HALOE measured profiles of NO 2 concentrations (in 10 9 molecule/cm3) for summer. The first overall feature is that while the HALOE measured concentration maximum is in summer at about 25 km (Fig 23) and in winter at about 28 km (Fig 21), the maximum altitude of FinROSE tends to be lower from some, to 15 km. The last happens when some strange UT/LS features appear into modelled profiles. The second throughout feature is that the modelled profiles maximum area may extend over about 49

50 20 km in height, while the measured profile s one bounds with less altitude range. The value of the maximum NO 2 is in cold season around molecule/cm 3 for FinROSE output, against to molecule/cm 3 measured by HALOE (Fig 21). Spring NO 2 concentrations stay also low for FinROSE ( molecule/cm 3 ), while measured values extend already up to molecule/cm 3. (One exception is the 16 th of May, but this profile can be classified under summer ones). In summer high latitudes the values of the FinROSE NO 2 concentrations can even exceed measured ones by 10-20% (18 Feb, 24 Jan, before mentioned 16 May). One case when the measured maximum and modelled maximum are at the same altitude and of same magnitude is in Fig 23 the Aug the 1st. But in this case the model sun is under the horizon already (SZA=92.2º). The altitude of the profile maximum is obviously connected to the sun height (the higher the sun, the lower the maximum altitude). NO 2 maximum altitude for autumn profiles is simulated the best (Fig 24). Figure 24. Comparison between FinROSE modeled and HALOE measured profiles of NO 2 concentrations (in 10 9 molecule/cm3) for autumn. 50

51 8.7. Conclusions about FinROSE-CTM validation The FinROSE-CTM output of BrO and NO 2 abundances was compared with measured profiles and SLIMCAT output. The following conclusions can be made: The seasonal variation of BrO columns in both hemispheres is similar with maximum in late winter early spring and increase with latitude. This is in qualitative agreement with SLIMCAT model predictions. On the monthly mean level the FinROSE-CTM tends to underestimate the BrO VCD-s over the whole globe. This is caused by two reasons: the maximum value of BrO mixing ratio is usually smaller and it locates higher in the atmosphere than in SLIMCAT or in SAOZ profiles. The largest load to the BrO stratospheric VCD is from the lower stratosphere, and the differences between BrO mixing ratios are largest in this region, i.e. improvements should be directed especially to this region. The total inorganic bromine profile and the lower boundary condition should be improved. Furthermore, the aerosol vertical distribution will be revised. But we must consider that our knowledge about BrO profiles in the stratosphere is very local: There are too few sonde measurements to compile a climatology of BrO. When comparing model output with measured data it should be kept in mind that model values are some spatial means, while in situ measurements are local values with quite large measurement uncertainties. The model output is always smoothed compared with measurements, which can be seen also from our results. The same is true regarding the temporal variability: because of the low spatial variability several small scale phenomenon that should bring along temporal variability are missed and model output does not change a lot from year to year. SCIAMACHY limb-measured BrO profiles are maybe the best source for stratospheric BrO profiles. At the moment there are still a lot of uncertainties about the quality of these data, but it would provide an independent data source for CTM validation. Conclusions about NO 2 comparisons: 51

52 The seasonal variation of NO 2 columns in both hemispheres is similar with maximum in summer and increase with latitude. This is in qualitative agreement with SLIMCAT model predictions. On the monthly mean level the FinROSE-CTM tends to underestimate the NO 2 VCD-s over the whole globe (except the NH polar 10º circle). The maximum value of NO 2 concentration is usually smaller than in SLIMCAT or in HALOE profiles, but it locates 2-20 km lower in the atmosphere. NO 2 profiles by harmonic approximation followed HALOE profiles with good accuracy, but unfortunately they are valid only for sunset and sunrise times. The lower absolute concentration values and lower maximum location altitudes of FinROSE simulated BrO and NO 2 compared to SLIMCAT and measured profiles can be connected to inadequate reproduction of the Brewer-Dobson circulation by model. Damski (2005) has called it the most challenging problem for the future developments of FinROSE. The low concentrations of NO 2 and BrO may also be connected the photochemical rates. For residual method the absorber profile in the stratosphere is not very important as the stratospheric AMF is not sensitive to it. Important are the absolute values of stratospheric vertical columns especially in the regions where the tropospheric vertical column of absorber is low, it means for regions far from tracer sources. For NO 2 it means areas without large anthropogenic and lightning activity, for BrO low latitude regions. FinROSE NO 2 and BrO output is currently at the level where the total stratospheric columns uncertainty of these tracers may exceed 100%. 9. Tropospheric columns: first results The approach has been implemented and is applied to the GOME data. Nearly cloudfree pixels (cloud fraction less than 0.2 according to FRESCO clouds data set) have been used. For tropospheric BrO the free tropospheric profile has been used everywhere. Northern Hemispheric monthly-mean tropospheric BrO VCD maps for March 1997 are shown in Fig 25. The only difference in calculating the tropospheric VCD-s presented in figures 25a and 25b is the stratospheric column used. As was expected, the tropospheric columns calculated using FinROSE model data (Fig 25b) exceed the BrO VCD values calculated using SLIMCAT stratospheric columns (Fig 25a) over all latitudes except tropics. Areas of maximum and minimum are located 52

53 over same regions. In polar regions, large areas of enhanced boundary layer BrO associated to low ozone events are observed in spring, this is why March was chosen as an example month. Global distributions of tropospheric NO 2 VCD-s for June 1997 are presented in Fig 26. Tropospheric NO 2 columns in Fig 26a are derived from GOME observations based on BIRA-IASB slant column NO 2 retrievals with the DOAS technique, and the KNMI combined modelling/retrieval/assimilation approach. Profile estimates are from the TM3 chemistry-transport model and cloud fraction and cloud top pressures from the Fresco algorithm. This figure originates from TEMIS homepage 2.html. Scheme by what are calculated NO 2 columns in Fig 26b differs from previous one only that instead of CTM TM3 FinROSE output is used for stratospheric columns. In both figures tropospheric maximum columns locate over same highly populated areas. It is difficult to compare the values of maximums as the colour scales of two figures are different (one is linear and the other logarithmic). Therefore the TEMIS tropospheric NO 2 field looks much spotty. The white areas in Fig 26b over the SH ocean are caused by cloudy pixels that have been left out of examination. a) b) Figure 25. NH tropospheric BrO VCD-s (in molecule/cm 2 ) for March 1997 calculated from GOME SCD-s using residual technique. Stratospheric columns are taken from CTM a) SLIMCAT and b) FinROSE output for GOME overpass time. 53

54 a) b) Figure 26. Tropospheric NO 2 VCD-s (in molecule/cm 2 ) for June 1997 calculated from GOME SCD-s using residual technique. Stratospheric columns are taken from a) TM3 chemistry-transport model and b) FinROSE output for GOME overpass time. 54