Analysis of satellite-derived Arctic tropospheric BrO columns in conjunction with aircraft measurements during ARCTAS and ARCPAC

Abstract. We derive tropospheric column BrO during the ARCTAS and ARCPAC field campaigns in spring 2008 using retrievals of total column BrO from the satellite UV nadir sensors OMI and GOME-2 using a radiative transfer model and stratospheric column BrO from a photochemical simulation. We conduct a comprehensive comparison of satellite-derived tropospheric BrO column to aircraft in-situ observations of BrO and related species. The aircraft profiles reveal that tropospheric BrO, when present during April 2008, was distributed over a broad range of altitudes rather than being confined to the planetary boundary layer (PBL). Perturbations to the total column resulting from tropospheric BrO are the same magnitude as perturbations due to longitudinal variations in the stratospheric component, so proper accounting of the stratospheric signal is essential for accurate determination of satellite-derived tropospheric BrO. We find reasonably good agreement between satellite-derived tropospheric BrO and columns found using aircraft in-situ BrO profiles, particularly when satellite radiances were obtained over bright surfaces (albedo >0.7), for solar zenith angle

Abstract. We derive tropospheric column BrO during the ARCTAS and ARCPAC field campaigns in spring 2008 using retrievals of total column BrO from the satellite UV nadir sensors OMI and GOME-2 using a radiative transfer model and stratospheric column BrO from a photochemical simulation. We conduct a comprehensive comparison of satellite-derived tro-5 pospheric BrO column to aircraft in-situ observations ofBrO and related species. The aircraft profiles reveal that tropospheric BrO, when present during April 2008, was distributed over a broad range of altitudes rather than being confined to the planetary boundary layer (PBL).
Perturbations to the total column resulting from tropospheric BrO are the same magnitude as perturbations due to longitudinal variations in the stratospheric component, so proper ac-10 counting of the stratospheric signal is essential for accurate determination of satellite-derived tropospheric BrO. We find reasonably good agreement between satellite-derived tropospheric BrO and columns found using aircraft in-situ BrO profiles, particularly when satellite radiances were obtained over bright surfaces (albedo> 0.7), for solar zenith angle < 80° and clear sky conditions. The rapid activation of BrO due to surface processes (the bromine explosion) is 15 apparent in both the OMI and GOME-2 based tropospheric columns. The wide orbital swath of OMI allows examination of the evolution of tropospheric BrO on about hourly time intervals near the pole. Low surface pressure, strong wind, and high PBL height are associated with an observed BrO activation event, supporting the notion of bromine activation by high winds over snow.

1 Introduction
Bromine plays an important role in tropospheric ozone chemistry and the resulting oxidation capacity of the polar boundary layer. Bromine radicals catalytically destroy ozone, leading to nearly complete removal near the surface that is termed an ozone (0 3 ) depletion event (ODE). Once 0: 3 is depleted, high levels of reactive halogen species including atomic bromine (e.g. Br) become the 25 primary oxidants for many species, including methane (CH 4 ) and mercury (Hg) (e.g. Simpson et aI., 2007b;Schroeder et aI., 1998). During mercury depletion events (MDEs), Bf and BrO are thought to oxidize elemental mercury to more reactive gaseous mercury that deposit to the polar ecosystem (Schroeder et aI., 1998;Lu et aI., 200 I;Ariya et aI., 2004;Douglas et aI., 2005;Holmes et aI., 2010).
Despite the many measurements of BrO and related species obtained by various techniques, significant uncertainties remain regarding the importance of very short lived source compounds on the stratospheric bromine budget (e.g. Sect. 1.3.3.3 of WMO, 2011) as well as the magnitude of the global, ubiquitous, background level of tropospheric BrO (e.g. Sect. 5.3 of Theys et aI., 2011). gradients with respect to longitude at high latitude during spring and Salawitch et al. (2010) questioned pllor estimates of residual tropospheric BrO found assuming that the stratospheric burden was zonally symmetric. Theys et al. (2011) derived tropospheric BrO columns from GOME-2 spectra with a model-based stratospheric BrO climatology (Theys et aI., 2009) similar to that used here 110 and described below.
Many studies related to bromine and ozone chemistry have been conducted using satellite-derived tropospheric BrO columns. Wagner and Platt (1998) reported elevated regions of BrO vertical column density in the Arctic and Antarctic regions. They noted these enhancements werc likely due to increased abundance of tropospheric BrO, rather than a stratospheric disturbance, based on a vari-115 ety of factors including the correlation between enhanced columns of BrO and the O 2 -0 2 collision complex. Wagner et a!. (2001) showed that elevated BrO column amounts observed by GOME were correlated with low ozonc in the boundary layer observed in-situ at Ny-Alesund (Spitsbergen), Norway. In other studies, spatial and temporal features of ODEs have been simulated using 3-dimensional regional chemical transport models and GOME-derived tropospheric BrO columns 120 (Zeng et a!., 2003(Zeng et a!., , 2006. Connections between BrO-rich air masses and first-year sea ice have been indicated with back-trajectory analyses using SCIAMACHY data . Transport of a large BrO plume near the North Pole is also reported by Begoin et a!. (20 I 0). A back trajectory study using satellite-derived tropospheric BrO columns indicated that ODEs can be ditferentiated into locally activated and transport driven events (Koo et aI., 2011). 125 Despite the numerous studies of tropospheric polar bromine chemistry using satellite BrO observations, estimation and interpretation of tropospheric BrO infonnation from space presents ongoing challenges. To properly estimate tropospheric BrO column amounts, the stratospheric contribution to the satellite-derived total column must be accurately represented (e.g. Theys et aI., 2009;Salawitch et aI., 2010). The global, ubiquitous background tropospheric level of BrO inferred from the 130 satellite record is sensitive to the amount of Bry delivered to the stratosphere by VSL bromocarbons (Salawitch et aI., 2005). Furthennore, low solar elevation angles in the early polar spring lead to large uncertainties in satellite total BrO column retrievals (see below). The presence of clouds further complicates the retrieval of tropospheric BrO from satellite observations. Theys et al. (2011) have recently addressed many of these issues. They showed maps of tropospheric BrO columns 135 derived using a method similar to that described below, and evaluated these columns using groundbased measurements of the tropospheric and stratospheric contributions to the total column. Our study builds upon the work of Theys et al. (2011)  The stratospheric BrO burden found by Salawitch et al. (2010) for their "best case" simulation of 7 ppt from POI and 2 ppt from SOl (see their Figure S7) is considerably larger (",27 % overall difference, with quite a bit of geographic variability) than the stratospheric BrO burden used by Theys 150 et al. (20 II) as shown in Figure S1.  (2003) for a detailed discussion of these pathways). Salawitch et al. (2010) showed agreement, to within uncertainties, between total column BrO measured by OMI and the sum of modeled stratospheric and aircraft-measured tropospheric BrO partial columns. However, their treatment of the tropospheric column did not explicitly account for tropospheric air mass factors (AMFs) and thus must 165 be treated with caution. Our study builds on this prior work by using a radiative transfer model to calculate tropospheric AMFs and also by considering the effects of clouds, surface reflectivity, and viewing geometry on the evaluation of the bromine budget for Arctic spring 2008. were acquired once every 2 s. Measurement uncertainties for BrO are ±40 % with a detection limit of 3 pptv for WP-3D data and ±40 % and a detection limit of 2-5 pptv for DC-8 data (Neuman et aI., 2010;Liao et aI., 2011b). In this study, we use I min averaged data for both DC-8 and WP-3D measurements. The spatial resolution of the I min averaged aircraft data is approximately 10 km.

200
We estimate tropospheric BrO columns from the aircraft BrO mixing ratio measurements for each ascent and descent. Figure 1 shows the flight tracks and locations of the 16 (29)   in layers at altitudes from 2 to 4 km. WP-3D measurements of BrO exhibit larger variability than 220 those from the DC-8 (Fig. 2d, e, and f).
The DC-8 and WP-3D instruments reported much lower mixing ratios of BrO near the surface than have been measured in the past by LP-DOAS instruments in the springtime Arctic boundary layer (Hausmann and Platt, 1994;Tuckermann et aI., 1997;Martinez et aI., 1999). To estimate the tropospheric BrO column from aircraft measurements, we must first make assumptions about mixing ratios between the surface and the lowest altitude sampled by the aircraft. 250 We only take profiles with aircraft minimum altitudes less than 500 m. Then, we assume that the BrO mixing ratio in the lowest bin (surfaee to 500 m) is the median BrO mixing ratio between the lowest aireraft altitude and 500 m. When the aireraft did not sample up to the tropopause, we made assumptions about mixing ratios between the highest aircraft altitude and the tropopause. Here, we use the upper part of the DC-8 eomposite profile to fill empty upper bins. However, the DC-8 com-255 posite profile only goes up to 7.5 km, the highest altitude where BrO is sampled by the aircraft. We assume that the BrO mixing ratio between 7.5 km and the tropopause is zero. Aircraft data suggest that BrO mixing ratios at these altitudes are very small for the sampled air masses (see Fig. 2). It is possible that stratospheric to tropospheric transport of air mass could supply BrO to the upper troposphere, particularly along the western flank of Arctic low pressure systems (i.e. after low alti-260 tude tropopause systems pass over a region) (Salawitch et aI., 2010 the satellite-derived tropospheric columns from OMI (COl.Ol'vII) and GOME-2 (Col.GOME-2) corresponding to the in-situ aircraft BrO profiles, and the ratios of satellite-derived tropospheric BrO columns to the in-situ BrO columns (Ratiool\H and RatioGoME-2) (see Section 2.2 and 3.2 for a description of the satellite data and related parameters in these tables).

275
The lowest altitudes sampled by the DC-8 and WP-3D during their descents over BrOenhanced regions were 75 and 77 m, respectively. In our analysis we use the composite DC-8 or WP-3D profile to extrapolate between the lowest sampled altitude and the surface, for each formulation of in-situ column BrO shown in Tables 1, 2a, and 2b. At times, surface BrO can reach mixing ratios as high as 40 pptv (Liao et aI., 2011a). We have assessed the impact of ele-280 vated surface BrO on our analysis of aircraft, satellite, and modeled stratospheric columns by conducting a probability distribution function for daytime surface BrO, observed at Barrow, Alaska. Two thirds of the time, surface BrO is below 8 pptv ( Figure S2). A uniform distribution of 8 pptv of BrO between the surface and 75 m altitude would contribute 0.18x 10 13 cm-2 to the column, an amount much smaller than the in situ and satellite-based columns discussed 285 throughout the paper (see caption, Figure S2). Levels of BrO reaching 40 pptv below the air-

Total BrO slant column density from OMI and GOME-2
OMI is a nadir-viewing ultraviolet and visible (UVNis) sensor (Levelt et aI., 2006) aboard the NASA 305 Aura satellite that is in a sun-synchronous orbit with an overpass of 01 :38 p.m. local time. The spectral resolution in the OMI UV-2 channel used to retrieve BrO columns is approximately 0.5 nm.
The OMI swath width is about 2600 km. The pixel size of OMI UV·2 channel is approximately 13 x 24 km 2 at the swath center and significantly larger at the swath edges. With its wide swath, OMI provides multiple daily observations at high latitudes in spring and daily global coverage at low and 310 middle latitudes. An obstruction outside the instrument that produces radiance errors (known as the "row anomaly") reduced the swath coverage mainly after May 2008; it does not significantly affect the obscrvations shown here (Claas et aI., 2010).
OMI BrO SCDs are retrieved by directly fitting backscattered UV radiances to absorption crosssections of BrO (the target gas), N0 2 , HCHO, and 80 2 as well as inelastic rotational-Raman scat-315 tering (also known as the Ring effect) using a non-linear least-squares approach (Chance, 1998 (Callies et aI., 2000). It has a local equator crossing time of 9:30 am in the descending node, a swath width of 1920 km, and a spatial resolution of 40 x 80 km 2 • For the BrO SCD retrieval used here (Begoin et aI., 2010), measurements in the window 336 to 347 nm are used, where GOME-2 has a resolution of about 0.3 urn. SCDs of BrO are retrieved the standard DOAS approach. Absorption due to BrO (Wahner et aI., 1988), 0;3, 1\0 2 , and the effects of rotational-325 Raman scattering are included. The uncertainty of the GOME-2 BrO total SCDs is 10 to 30 % depending on solar zenith angle (SZA) and surface albedo and the uncertainty has both random and systematic contributions from spectral interferences and the cross-sections. As a result of throughput loss of the UV channels, random errors have increased since launch. However, this etTect is not significant during the time period studied here (Dikty et aI., 2011). BrO column inferred from satellite observations of total column BrO, our approach of using a value for stratospheric Bry near the upper end is supported by the generally close quantitative agreement between inferences of tropospheric BrO column from the satellites and the in-situ data, described 385 below. As noted previously, the fields of stratospheric BrO used here are approximately 27 % larger than those reported by Theys et al. (201 I) (with percentage differences that vary considerably with respect to location).

Cloud parameters
Optically thick clouds shield the underlying atmosphere from satellite sensors. The aMI rotational 390 Raman (RR) cloud product (Vasilkov et aI., 2008) is used to infer information about the shielding effects of clouds over snow and ice (Vasilkov et aI.,20 I 0). This product provides an estimate of the scene (combined cloud and surface) pressure over snow and ice surfaces. The Near-real-time SSM/I EASE-grid daily global Ice and snow concentration and Snow Extent (NISE) data set (Nolin et aI., 1998) is used to identify snow and ice-covered pixels. When the difference between scene tropospheric BrO from the satellite sensor et aI., 2010), as discussed below.

Meteorological data sets
Tropopause heights are inferred along the flight tracks using tropopause pressure and geopotential height profiles from MERRA, a reanalysis based on GEOS-5 system (Rienecker et aI., 2007) The SCD of a given absorber seen by a satellite sensor is defined as the amount of the absorber along an average light path taken by photons as they travel from the sun, through the atmosphere, and back to the sensor. The SCD is affected by scattering and absorption within the atmosphere as well as reflection off the surface and clouds.
For a given altitude range (denoted by a subscript z), The air mass factor (AMF z ) is used to 425 convert SCD" to VCDz as follows: The sensitivity of UV radiance measurements to the BrO layer amounts varies with altitude. This variation on geometry solar zenith angle, SZA, and view zenith VZA), surface albedo, cloud effects, and the vertical BrO profile. We must account for this varying 430 sensitivity in the AMF. Using the optically thin absorber assumption, the AMFz can be formulated as (Palmer et aI., 2001;Theys et aI., 2011), where z is altitude, N(z) is the number density profile of the absorber, and W (z) is the weighting function profile that represents all the parameters influencing 435 the AMF except the vertical profile of the absorber.
Here, we use estimates of AMFs trat provided in the OMI and GOME-2 total BrO column products. AMFStrat for GOME-2 is computed using the SCIATRAN radiative transfer model (Rozanov et aI., 2005 The calculation of AMFTrop uses the derived OMI reflectivity at 331 nm for each pixel, from the OMI Total Ozone Mapping Spectrometer (TOMS) total 0 3 product, as a proxy for surface albedo. 455 We thus avoid using a surface albedo climatology, which can lead to AMF errors when the climatology differs from the actual surface albedo. Such errors can occur when the snow or sea ice distribution differs from the climatological mean, which is of particular concern at high latitudes during spring.
3.2 Sensitivity of the derived tropospheric column to SZA, surface albedo, and clouds 460 Here, we investigate the sensitivity of the satellite-derived tropospheric BrO retrieval to SZA, surface albedo, and clouds. Theys et al. (201l) analyzed the dependence of the weighting function on surface albedo and clouds for SZA = 45°. Here, we extend this analysis to a wider SZA range (40°relevant to polar observing conditions; at high latitudes (>60° N) during the day in the early SZA is >50 0 < 3a shows that when the sun is in the (SZA 465 aMI and GOME-2 should have good sensitivity to tropospheric Bra for surface albedos >0.5. Sensitivity is significantly reduced for darker surfaces (e.g. albedo = 0.1). Sensitivity to tropospheric BrO decreases with increasing SZA; there is significantly lower sensitivity at 80° even for a surface albedo of 0.5. Figure 3b shows that at high surface albedo (0.9), there is increased sensitivity to BrO near the surface in addition to good overall tropospheric sensitivity for SZA up to 80°.

470
The effect of optically thin clouds and aerosols on the sensitivity of aMI and GOME-2 to tropospheric Bra is dependent on surface reflectivity and viewing geometry. Vasilkov et at. (2010) showed that the sensitivity of UV satellite measurements to trace gas absorption near the surface in clear skies with moderately high surface albedo (70%) is approximately the same as for substantially cloudy conditions (optical thicknesses up to about 30) over a higher albedo 475 surface (90%); the cloud shielding effect is much reduced for high albedo surfaces. In addition, enhanced absorption takes place above a cloud of moderate to high optical thickness.
According to Vasilkov et al. (2010), the UV cloud shielding effect is generally reduced over bright  The sensitivity of AMFTrop to the tropospheric BrO profile is assessed in Fig. 4b. For a bright surface (albedo 0.9), no significant difference exists between tropospheric AMFs computed using the DC-8 composite profile and a profile where all tropospheric BrO is contained below 500 m for SZA ~rv60°. This is a consequence of good sensitivity at all tropospheric altitudes for bright 500 surfaces and low SZA as shown in Fig. 3. For 60° < SZA < 75°, the sensitivity to middle-and upper-tropospheric BrO is maintained while the near-surface sensitivity drops; this leads to a small sensitivity to the profile shape under these conditions. Profile sensitivity decreases for higher SZAs as the sensitivity to the entire troposphere drops. For a darker surface (albedo 0.4), the retrieval has lower sensitivity to BrO near the surface even when the sun is relatively high (SZA cv 40°).
505 In this case, the retrieval of total column Bra and our inference of the tropospheric column will be sensitive to the shape of the profile of Bra in the troposphere. If most of the tropospheric column happened to originate from Bra in the lowest 500 m of the atmosphere, then our inference of tropospheric column would be biased low over dark surfaces.
We next examine tropospheric column Bra estimated from aMI in the context of the different 510 sensitivities discussed above. Figure  an enhanced tropospheric burden (Fig. 5c, red). This region of enhanced tropospheric BrO occurs over a bright portion of the Barents Sea (Fig. 5d, crimson, indicating snow or ice).
520 Nearly zero tropospheric BrO column amounts are obtained over low surface albedo areas (OMI reflectivity < 0.5) of the Barents Sea (Fig. 5d, blue). Here, the retrieved total columns are generally less than retrieved columns over adjacent areas with higher reflectivity, leading to low tropospheric column BrO over parts of the Barent Sea. However, satellite-derived tropospheric BrO may not be reliable when the surface albedo is low. While it is possible 525 tropospheric column BrO was truly low on 22 April over this region of the Barents Sea due to the lack of snow or ice leads that may be needed for bromine activation, it is also possible that our inference of low tropospheric BrO could either be a result of limited sensitivity over dark surfaces or an over-estimation of the stratospheric burden. The complicated sensitivity of satellite-derived tropospheric BrO to surface reflectivity and stratospheric burden requires 530 concerted future study.
In Fig. 5 We take the fitting uncertainty derived from observed minus fitted radiances for each pixel as crSCDTotal. This assumes that the SCDTotal error has a zero mean. The average fitting uncertainty at latitudes greater than 60° N is about 18 %. Here, we do not consider systematic errors in 555 SCDTotal or VCDTotal' Systematic error will generally result in either a geographically uniform over-estimate or under-estimate of total column BrO (VCDTotal). There is synergy between systematic error in VCDTotal and our prescription of the contribution of VSL species to stratospheric Bry. If subsequent analysis shows the estimates of VCDTotal BrO used here are biased high by a considerable margin, then clearly we must use a smaller contribution 560 to stratospheric Bry to derive similar overall magnitude of tropospheric BrO. However, the geographic distribution of tropospheric BrO will not be strongly altered due to this synergy.
An exploration of the systematic error in VCDTotal and the implication for tropospheric BrO will occur following analysis of a ground-based, OMI BrO validation campaign conducted in Fairbanks, Alaska during April 2011.

565
We estimate the uncertainty of the stratospheric slant column (crSCDSt,at) by multiplying the uncertainty of the stratospheric column BrO by the stratospheric AMF. The uncertainty of the stratospheric column BrO results from a root-sum-squares combination of three terms: the uncertainty in chemical kinetics that govern the BrO to Bry ratio, the uncertainty in the dynamics that govern CFC-12 (and hence Bry due to Source Gas Injection), and the uncertainly in VSL Bry. The uncer-570 tainty in chemical kinetics is evaluated by varying the rate constant of individual chemical reactions (including J values) by the I-sigma estimate of uncertainty given by Sander et al. (2006). The most important chemical term is BrO + N0 2 forming BrNO: 3 : this rate constant is uncertain by about a factor of 2 at 220 K. The uncertainty due to dynamics is found by repeating calculations for and minus 4 % variations in the abundance of CFC-12, because comparison to aircraft observations 575 showed CFC-12 from GEOS-5 was accurate to within ±4 % in the lower stratosphere (Salawitch et aI., 2010). Finally, the uncertainty due to VSL Bry is set at ±27 %, which represents the mean difference in stratospheric column BrO resulting from our approach to handling this term compared to the approach of Theys et al. (201l).
Here, we neglect the uncertainty in the tropospheric air mass factor as it is relatively small com-580 pared with the other error sources. For example, Theys et al. (201l) describe sensitivity tests that show use of a single wavelength for the weighting function profiles leads to an error of less than 5 %.
We find that the profile dependence produces error between 7 and 13 % for bright surfaces.   The OMI tropospheric BrO columns exhibit a magnitude and variability similar to that of the DC-8 in-situ columns. For example, profile #12 (See Figs. 1,2 and 6b) reports high BrO mixing ratios from the surface to ",4 km; the OMI tropospheric BrO column (as well as GOME-2, shown in Fig. 6a but not in Fig. 6b) is also relatively high at the same location.  Figure 7b shows that the derived tropospheric BrO eolumn from OMI is low over Eastern Canada, where the total column is high. Similar results 670 are found for GOME-2 (Fig. 7a). The OMI measurement of total column BrO shows very similar magnitude and structure as the calculated stratospheric column BrO. This suggests that the elevated total column BrO over Eastern Canada could be a consequence of high stratospheric columns.   Fig. 9b).

Cases of good agreement between satellite and in-situ data
The highest level of active bromine is observed from the aircraft sensor at the same location \vhere tropospheric column BrO from both satellite sensors show enhancements. 0 3 mixing ratios of rv 10 ppbv suggest the aircraft sampled a partial ozone depletion event (e.g. Ridley et aI., 2003). 720 Prior DOAS observations in the Arctic suggest that BrO should still be present for 0: 3 levels of rv 10 ppbv (Hausmann and Platt, 1994;Tuckermann et aI., 1997). The disagreement between satellite and aircraft BrO could be related to the different spatial scales, and/or vertical coverage spanned by the respective instruments: the aircraft does not observe the complete profile and the field of view encompassed by the satellite covers a larger area than that sampled by the airplane. The aircraft 725 may have missed an important part of the BrO profile as the minimum sampled height was 151 m above the surface. Timing may also be a possible explanation for the disagreement, as the aircraft flew near the location of the enhancement about five hours after the OMI overpass and about three hours after that of GOME-2.
Our analysis indicates that the large enhancement in total column BrO seen by OMI and 730 GOME-2 over the Chukchi Sea on 19 April 2008 was tropospheric in origin, rather than stratospheric. The WP-3D aircraft recorded highly elevated active bromine and partially depleted 0: 3 near the surface at this precise location, indicating recent association with elevated BrO. It is possible the satellite perturbation was caused by the presence of BrO at higher altitudes than those sampled by the aircraft at this location.  plots indicate that WP-3D in-situ columns #25, #26, and #27 (See Figs. 1,2 and lOb) are lower than 745 the OMI tropospheric column, but the differences are not significant given the uncertainties in OMI and in-situ tropospheric columns (i.e., error bars overlap). However, the absolute magnitude of the OMI tropospheric column is about a factor of 3 larger than the in-situ BrO column. There is, however, good agreement between in-situ column #28 and the satellite-based estimate.
WP-3D profiles #26, #27, and #28 (See Figs. 1,2 and 6b) were collected in close proximity (see 750 Fig. I), but show a large variation in the column amounts. As noted above, the profile #28 column agrees well with the satellite-based column whereas the other two in-situ based BrO columns are lower than the satellite-based estimates. As discussed above for other flights, this result may be explained by the fact that the aircraft captures small-scale spatial features while the satellite observes larger scales.

755
The overestimation of the tropospheric column BrO by OM I, compared to in-situ columns for profile #25, #26, and #27 (See Figs. 1,2 and 6b) , could potentially be explained by a stratospheric column that is not fully removed from the total column. One factor not considered in our analysis is irreversible, cross-tropopause exchange of of air parcels with elevated levels of Bry from the stratosphere to the troposphere (STE). Such transport events occur on the western flank of Arctic 760 low pressure systems (Salawitch et aI.,20 I 0,and references therein). If the satellite signal were to originate from STE of Bry, the BrO signal associated with such air parcels may not have been sampled by the WP-3D because the maximum altitudes of the WP-3D profiles #25, #26 and #27  (Fig. 9a). Here we examine the event near the North Pole on 17 April 2008 and similar events in more detail using only OMI retrievals. The wide orbital swath and high spatial resolution of OMI, in addition to its frequent observations at high 775 latitudes, provide a unique view of the temporal evolution of these events. Figure II shows the evolution of a tropospheric BrO enhancement event ("BrO explosion") observed from 16 to 18 April 2008. Here, we only show observations when the following conditions are met, to provide reliable tropospheric BrO information as discussed in Sect. 3.2: SZA < 80°, retlectivity >0.7, and 6..P c < 250 hPa. The stratospheric column has been removed, as discussed in 780 Sect. 3.1, using photochemical model output for the VSL Bry = 7 ppt simulation, for the local solar time of each OMI pixel. The major activation of BrO starts at rv22:00 UTC on 16 April and lasts for rv30 h. We see activations near the North Pole and Canadian Archipelago. The spatial features of the elevated tropospheric column BrO change rapidly, with significant variations over the course of a day.

785
We have ruled out a stratospheric origin for the enhanced total column BrO observations near the North Pole and Canadian Archipelago on these days. Since stratospheric BrO and 0 3 columns exhibit a significant correlation (Theys et a!., 2009;Salawitch et aI., 2010), the OMI measurement of total O:i column can be used as a proxy for the spatial pattern of stratospheric column BrO. The comparison of the OMI tropospheric BrO column with the OMI total column shown in Fig. 12 790 indicates lack of correlation. Thus, the elevated region of total column BrO is not of stratospheric origin. Figure 12 also shows few shielding clouds near the areas with high BrO columns. 800 Figure 13 also shows that the spatial structure of high tropospheric column BrO is similar to that of the planetary boundary layer (PBL) height, although there is not always a precise alignment of these features. At high latitudes where the meteorological analysis is driven primarily by satellite data, the MERRA fields may contain displacement or other errors, particularly in near-surface fields.
The following discussion should therefore be considered somewhat speculative in light of these 805 uncertainties. We provide an animated visualization as a Supplement to depict evolution of enhanced tropospheric column BrO, sea level pressure, wind speed at 2 m and planetary boundary layer height from 16 to IS April200S. tropospheric BrO for SZA greater than rv80 degrees because, under these conditions satellite 875 radiances have decreased sensitivity to absorption by tropospheric BrO (Fig. 4).
Satellite and aircraft measurements of tropospheric column BrO do not always exhibit good agreement, at times for reasons that seem well understood and at other times for reasons that are unclear but may be related to differences in the timing or spatial coverage of the respective observations. For the severe ozone depletion event observed by the DC-8 near Alert on 8 April 2008, neither OMI to-880 tal column BrO column nor OMI tropospheric column BrO were elevated. Atmospheric conditions were favorable for the remote sensing from space of tropospheric BrO near Alert on this date (e.g. clear skies, high surface reflectivity). We believe the lack of a signal for OMI tropospheric column BrO near Alert on 8 April is consistent with our understanding of bromine chemistry: the production of BrO diminishes when ozone is severely depleted (e.g. Hausmann and Platt, 1994;Tuckermann et 885 aI., 1997). Hence, the association of elevated tropospheric column BrO and depleted 0 3 is expected to be much stronger for a partial ozone depletion event (ODE) than a major ODE. Tropospheric column BrO from OMI does show an enhancement, near Alert, 36 h prior to the major ODE. The aircraft may have been capturing very small scale variability (several nearby profiles showed large differences) compared to space-based observations. Finally, aircraft observations generally did not 890 sample the altitude region where stratosphere to troposphere transport of active bromine associated with air parcels with elevated levels of inorganic bromine could potentially be affecting the satellite measurement of column BrO.
We examined several events of rapid enhancement of tropospheric column BrO observed by OMI.