Frost flowers grow on newly-formed sea ice from a saturated water vapour layer. They provide a large effective surface area and a reservoir of sea salt ions in the liquid phase with triple the ion concentration of sea water. Recently, frost flowers have been recognised as the dominant source of sea salt aerosol in the Antarctic, and it has been speculated that they could be involved in processes causing severe tropospheric ozone depletion events during the polar sunrise. These events can be explained by heterogeneous autocatalytic reactions taking place on salt-laden ice surfaces which exponentially increase the reactive gas phase bromine (“bromine explosion”). We analyzed tropospheric bromine monoxide (BrO) and the sea ice coverage both measured from satellite sensors. Our model based interpretation shows that young ice regions potentially covered with frost flowers seem to be the source of bromine found in bromine explosion events.
 The discovery of events of low ozone concentration in the atmospheric boundary layer at measurement stations located at high latitudes in the Northern Hemisphere has prompted much research into their origin. These events were found to be associated with enhanced amounts of inorganic bromine compounds [Barrie et al., 1988; McConnell et al., 1992; Fan and Jacob, 1992; Foster et al., 2001]. Similar episodes have been observed in the Southern Hemisphere at high latitudes [Wessel et al., 1998; Frieß et al., 2004]. The advent of the measurement of tropospheric trace gases from space by the Global Ozone Monitoring Experiment, GOME, led to the discovery of enhanced amounts of BrO close to regions of sea ice in the Northern and the Southern Hemisphere [Richter et al., 1998; Wagner and Platt, 1998]. GOME measures the light scattered from the atmosphere and reflected by the ground between 240 and 790 nm wavelength with a horizontal ground resolution of about 320 × 40 km2 [Burrows et al., 1999]. A retrieval algorithm, based on differential optical absorption spectroscopy (DOAS) and stratospheric BrO modeled by a three-dimensional radiative- dynamical-chemical model, yields the tropospheric fraction of the column density of BrO [Richter et al., 1998; Wagner and Platt, 1998; Chipperfield, 1999]. Bromine destroys ozone very efficiently in two interlinked catalytic cycles which produce BrO and HOBr in the gas-phase [Foster et al., 2001]. Gaseous HOBr reacts with Br ions in a slightly acidic sea salt solution and releases Br2 and BrCl into the gas phase [Fickert et al., 1999; Adams et al., 2002]. The photolabile Br2 molecule is subsequently photo-dissociated into atomic Br [Foster et al., 2001]. Therefore, every Br atom of the HOBr molecule entering the liquid phase has the potential to release two Br atoms to the gas phase. The above gives a simplified description of the heterogeneous autocatalytic reaction that causes an exponential increase of gaseous Br radicals, the so-called bromine explosion. The main source of bromine over the open oceans in the marine boundary layer outside the polar regions was identified to be sea salt aerosol generated by breaking waves on the ocean surface [Tang and McConnell, 1996; Vogt et al., 1996; Sander et al., 2003]. The processes and sources unique to the polar ocean surfaces still remained unidentified, though the highest BrO amounts have been observed over the sea ice during the polar sunrise [Ridley et al., 2003; Zeng et al., 2003; Frieß et al., 2004]. Recently, the potential role of frost flowers (Figure 1) in this processes has been raised [Rankin et al., 2002]. Frost flowers are ice crystals which grow on frozen leads (linear breaks in the sea ice cover) and polynyas (openings between drift ice and fast ice or the coast). Frost flowers exhibit enhanced salinities and bromide ion concentrations of about three times of that of bulk seawater [Perovich and Richter-Menge, 1994; Rankin et al., 2002]. Frost flowers only last for a few days until they are blown away by strong winds or covered by drifting snow [Perovich and Richter-Menge, 1994]. The aerosol produced by frost flowers was identified in Antarctic ice cores and at coastal stations due to its depleted sulfate to sodium ratio compared to the aerosol originating from the open ocean [Wagenbach et al., 1998; Rankin et al., 2002; Rankin and Wolff, 2003].
2. Methods and Data Sets
 It is assumed, that the open water area (which is given by one minus the sea ice concentration) will be covered soon with thin new ice on which frost flowers can grow. The sea ice concentration is determined from analyses of the thermal microwave emission measured by the Special Sensor Microwave Imager (SSM/I) aboard the Defense Meteorological Satellite Program (DMSP) platform using the ASI algorithm [Kaleschke et al., 2001; Maslanik and Stroeve, 2003]. Briefly this algorithm takes advantage of the higher polarisability of a specularly reflecting water surface compared to the more diffusely reflecting sea ice surface [Kaleschke et al., 2001; Kern et al., 2003].
 A one dimensional thermodynamic model has been developed to calculate the frost flower coverage (Figure 2). The model combines a frost flower growth parameterization obtained from laboratory experiments with the equations of sea ice heat balance which are described in more detail by Martin et al.  and Maykut . We assume that the growth of frost flowers depends only on two basic prerequisites, the existence of new ice which is formed in leads or polynyas and of a strong negative temperature gradient above the ice surface. The two model input parameters, the surface air temperature Ta and the open water (OW) fraction, are taken from numerical weather prediction reanalysis data (NCEP/NCAR) and from satellite passive microwave measurements, respectively. The ice thickness H = 1.33Θ0.53 [cm] is calculated from the cumulative freezing days Θ = ∫(Tf − Ta)dt with air temperature Ta and the freezing point of sea water Tf = −1.9°C. The sea ice surface temperature T0 = is approximated for thin ice using an averaged heat transfer coefficient which describes both sensible and latent heat exchange [Maykut, 1986]. Assuming that the influence of a varying wind field and the insulating effect of frost flowers on the heat flux are negligible, then the frost flower growth rate g = is readily calculated using coefficients from laboratory experiments [Martin et al., 1996; Maykut, 1986]. The area coverage Ft is calculated using the recursive expression Ft = Ft−δt + g(1 − Ft−δt)δt for a time step δt. Because the growth rate decreases rapidly as the ice thickness increases, the model yields a maximum percentage area Fmax(Ta) covered by frost flowers for a given surface air temperature. This is defined as the relative potential frost flower (PFF) area. The total PFF area is obtained by weighting the total area with the new ice fraction. The predicted relative PFF areas for different integration times and air temperatures are presented in Figure 2. The height and hence the volume of the frost flowers cannot be calculated as the parameterization of the growth rate was derived from video images which measured the area coverage. As there is currently insufficient microphysical understanding about the initial nucleation and growth of frost flowers as well as about the decay processes, a minimum frost flower area for a given set of conditions cannot be estimated [Perovich and Richter-Menge, 1994; Martin et al., 1995, 1996]. The original aim of the above described method was to identify dates and regions worth to be analysed using costly very high resolution satellite images for the future development of a more direct frost flower retrieval algorithm [Kaleschke and Heygster, 2004].
3. Comparison of Model Results and BrO Data
 One typical example of the resulting PFF and the BrO data for the Antarctic is shown in Figure 3. The overall mechanism requires the release of bromine atom precursors, either directly on the frost flower surface or within its aerosol. We calculated forward air trajectories starting at regions with a high probability of frost flowers from the NCEP/NCAR Reanalysis surface wind field in order to account for the atmospheric transport of the aerosol or BrO. Some regions with a high probability of frost flower occurrence for example at the Ronne-Filchner Ice Shelf in the Weddell Sea show no corresponding BrO plumes because sunlight is needed for the photochemical reactions. The polynya at the Ross Ice Shelf that occurs frequently due to strong katabatic winds is shown to be a strong source of frost flower aerosol responsible for the enhanced BrO concentration farther north in the illuminated area. The trajectories are only a rough approximation as errors could occur in convective regions which are not well represented in the atmospheric boundary layer of the NCEP/NCAR model over sea ice [Kaleschke et al., 2001]. Nevertheless, more than ninety percent of the trajectories hit the enhanced BrO areas in Figure 3. One key area of frequently occurring ozone depletion events in the Arctic is shown in Figure 4. This typical example shows the huge (almost 500 × 50 km2) recurring shore polynya in the northwest of the Hudson Bay which is potentially covered with frost flowers. This polynya frequently appears under offshore wind conditions.
 We investigated the entire 1996 to 2002 dataset and found commonly more than two third of the PFF trajectories hitting the enhanced BrO areas for the Arctic and Antarctic during polar sunrise. Almost all cases of enhanced BrO amounts were associated with a high probability of frost flowers on the previous days. Occasionally enhanced PFF values appear without enhanced BrO amounts. However, this does not reject the hypothesis of frost flowers causing BrO production as the BrO retrieval could be hampered by clouds [Richter et al., 1998; Wagner and Platt, 1998; Frieß et al., 2004]. Furthermore, the PFF is a potential theoretical upper limit. Specific meteorological conditions could have prevented the actual growth of frost flowers. The influence of the wind is ambiguous: the dynamical opening of leads and polynyas is a wind driven effect and a prerequisite for thin ice production, but persisting strong winds could prevent the growth of frost flowers [Perovich and Richter-Menge, 1994]. Changing wind fields such as passing cyclones probably support the growth of frost flowers.
 It can be summarized that the sea salt and associated halogen flux from the ocean to the atmosphere is governed by different processes inside and outside the polar regions: Outside the polar regions, the sea salt is injected into the atmosphere by breaking waves on the ocean surface dominated by the wind [Sander et al., 2003]. Whereas the process inside the sea ice covered regions is mainly modulated by the air temperature [Zeng et al., 2003; Frieß et al., 2004]. Previous work has been insufficient to localize the potential bromine sources [Richter et al., 1998; Wagner and Platt, 1998; Zeng et al., 2003; Ridley et al., 2003; Frieß et al., 2004]. We provided a method to localize the young ice regions potentially covered with frost flowers that seem to be a prerequisite for the bromine explosion. This provides a crucial step for a better understanding of the exchange processes between ocean, sea ice and atmosphere that are of great importance for the Earth's climate system [Shepson et al., 2003].
 L. K. and HW. J. gratefully acknowledge the German Research Foundation (DFG) for funding. L. K. thanks Gunnar Spreen, Stefan Kern, Roland von Glasow, Eric W. Wolff, Mark Drinkwater, Robert Ezraty, Wolfgang Dierking, Christof Lüpkes, Thomas Busche and Christian Haas for discussions.