Cloud-scale modelling of the impact of deep convection on the fate of oceanic bromoform in the troposphere: a case study over the west coast of Borneo

Coastal oceans emit bromoform (CHBr3) that can be transported rapidly to the upper troposphere by deep convection. In the troposphere, the spatial and vertical distribution of CHBr3 and its product gases (PGs) depend on emissions, chemical processing, transport by large scale flow, convection, and associated washout. This paper presents a modelling study on the fate of CHBr3 and its PGs in the troposphere. A case study at cloud scale was conducted along the west coast of Borneo, 30 when several deep convective systems triggered in the afternoon and early evening of November 19 2011. These systems were sampled by the Falcon aircraft during the field campaign of the SHIVA project. We analyse these systems using a simulation with the cloud-resolving meteorological model C-CATT-BRAMS at 2 × 2 km resolution that describes transport, photochemistry, and washout of CHBr3. We find that simulated CHBr3 mixing ratios and the observed values in the boundary layer and the outflow of the convective systems agree. However, the model underestimates the background CHBr3 mixing 35 ratios in the upper troposphere, which suggests a missing source. An analysis of the simulated chemical speciation of bromine within and around each simulated convective system during the mature convective stage reveals that >85% of the bromine https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c © Author(s) 2020. CC BY 4.0 License.

undergoes net radiative heating and eventual buoyancy-driven slow ascent to the stratosphere. For clear sky conditions, the 70 LZRH is approximately 15 km in the tropics, and can be as low as 11 km for air masses within clouds resulting from convective outflow (Corti et al., 2005(Corti et al., , 2006. Tropical deep convection can also loft air masses directly into the stratosphere through a process called convective over-shooting, but this process is less frequent (e.g. Zipser 2005, Luo et al. 2008).
After emission, CHBr3 undergoes oxidation in the troposphere during transport either via reaction with the hydroxyl radical (OH) or via photolysis (approximate lifetime of 16 days, Burkholder et al., 2018). The oxidation products are organic and 75 inorganic product gases (PGs) (Hossaini et al., 2010;Krysztofiak et al., 2012). The most important organic PGs are the brominated organic peroxides, CBr3O2H, CHBr2O2H; and the brominated carbonyl species, CBr2O and CHBrO. The inorganic PGs consist of the bromine radical (Br), molecular bromine (Br2), bromine oxide (BrO), hypobromous acid (HOBr), hydrogen bromine (HBr), and bromine nitrate (BrONO2). These PGs have a range of solubilities, and thus washout within convective systems is expected to exert a strong control on the vertical transport of bromine to the upper troposphere (Hossaini et al., 80 2010).
CHBr3 is important at the global scale, but deep convection occurs at the local scale. Convective transport and the associated chemistry and washout of all bromine containing species (Bry) cannot be simulated in detail with global 3-dimensional models because of their coarse resolution, and because of the complexity of the chemical processes (e.g. Hossaini et al., 2010). In global models, deep convection is a sub-grid scale process taken into account by parameterizations. These parameterizations 85 are a known source of uncertainty for tracer transport, including CHBr3, from the boundary layer to the upper troposphere (e.g. Hoyle et al. 2011, Liang et al., 2014Hossaini et al., 2016;Butler et al., 2018). Regarding chemical processes and their interactions with liquid and ice hydrometeors, global models have made progress (Hossaini et al. 2012, Aschmann and Sinnhuber, 2013, Liang et al. 2014), but they still need to compromise between complexity and computing resources. The analysis of the detailed transport and chemical processes occurring within deep convection and in its vicinity needs fine scale 90 modelling at the kilometre-resolution with detailed chemistry (e.g., Barth et al., 2001;Marécal et al., 2006). This knowledge, gained from studies at the convective scale, may then improve the representation of the fate of chemical species in global models.
Only two detailed model studies examining VSLS degradation chemistry (both gas and aqueous phase), which were idealised cases, have been carried out at the convective scale Marécal et al., 2012). Krysztofiak et al. (2012) 95 focused on developing and optimising a photochemical mechanism for CHBr3 degradation for use within models, and they estimated Henry's Law coefficients for some of the organic bromine species within the optimised mechanism. Marécal et al. (2012) implemented the CHBr3 photochemical scheme of Hossaini et al. (2010) in addition to aqueous phase uptake and chemistry based on the Henry's Law coefficients from Krysztofiak et al. (2012). Using idealised simulations of a tropical convective cloud, Marécal et al. (2012) explored the CHBr3 chemistry at the cloud scale and highlighted the importance of 100 aqueous phase processes for understanding source gas and PG chemistry and transport.
Given the limitations of the previous studies at the convective scale, we wish to expand upon that previous work, e.g., Marécal et al. (2012), by carrying out a real-world case study. In this paper we present a study based on a high-resolution cloud resolving https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License. modelling of the transport, chemistry, and washout of CHBr3 and its derivatives within convective clouds along the west coast of Borneo. Specifically, we aim to look at the bromine compound speciation in the areas of convective entrainment and 105 detrainment and both within and outside of several convective systems. The modelling is supported using aircraft observations of CHBr3, which we use to establish the credibility of the simulated chemical processes.
Section 2 gives a description of the measurement campaign over the west coast of Borneo, it provides an overview of the case study, and it includes a description of the meteorological situation based on observations. Section 3 describes the model used, its new developments and the simulation setup. Section 4 presents our modelling results, which includes: i) an evaluation of 110 the simulated meteorology with respect to the situation described in Sect. 2; ii) a detailed analysis of the transport, chemistry and washout of bromoform and its PGs; and iii) a discussion of the chemistry. Section 5 discusses the limitations, Sect. 6 presents a summary and our conclusions, and Sect. 7 presents a brief outlook.

The SHIVA Campaign and Case Study Overview
The EU funded SHIVA ('Stratospheric Ozone: Halogen Impacts in a Varying Atmosphere') project (http://shiva.iup.uni-115 heidelberg.de/) was designed to address uncertainties in our understanding of VSLS, their contribution to stratospheric bromine, and their impact on stratospheric ozone. A measurement campaign within SHIVA was carried out in November 2011 that focused on the South East Asia Maritime Continent (SEA-MC) to better understand the emissions and the transport of oceanic VSLS, including CHBr3, to the upper troposphere and stratosphere. This region was selected for two reasons. First, it represents globally the most important region for deep convection (Liu and Zipser, 2005) and second, the SEA-MC was believed to be 120 an important region for VSLS emissions due to its many coastlines and its location in the tropics. The campaign primarily relied on measurements of chemical species onboard the Deutsches Zentrum für Luft-und Raumfahrt (DLR) Falcon aircraft based in Miri, Sarawak, Malaysia, which were complemented by ship and ground-based observations (Pfeilsticker et al., 2013).
We selected a case study from the SHIVA campaign on the afternoon of 19 November 2011. Figure 1 shows hourly maps of brightness temperature contours measured by the 11 μm channel IR108 on board the MTSAT-2 satellite. Note that the 125 brightness temperatures contours in Fig. 1 are chosen to highlight only cloud tops in the upper troposphere. The brightness temperature imagery illustrates that several deep convective systems initiated inland along the West coast of Borneo where CHBr3 emissions are expected to be strong . The Falcon aircraft sampled two of these convective systems, which we henceforth refer to as Obs_Conv1 and Obs_Conv2 and are shown in the blue box and pink box in Fig. 1, respectively.
The temporal evolution of both convective systems develops over several hours (05 UTC to 12 UTC) and is shown in Figs 1 130 (a) through (h), which indicates that the two follow a similar development scenario. They were both initiated early afternoon inland close to the west coast of Borneo from offshore low-level winds encountering the steep topography of the island. These low levels winds come from the north/north-westerly large scale flow as indicated by ECMWF analysis (not shown), possibly combined with local diurnal variations of the sea breeze (Johnson and Priegnitz, 1981). The initial convective cells developed vertically and then horizontally to form an anvil from its outflow (also named stratiform part of the convective system) in the 135 https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License. upper troposphere driven off the coast by the easterly/south-easterly upper tropospheric flow.
Obs_Conv1 was already well developed at 05 UTC (13h local time: 13 LT) and was located around 5.5°N and 116°E at this time. Obs_Conv1 produced a large anvil on its west flank that started weakening after 10 UTC (18 LT). The anvil of Obs_Conv1 was well sampled by the Falcon aircraft during its mature stage at altitudes between 11 and 13 km from 8.05 UTC to 9.35 UTC (16.05 LT to 17.35 LT). The trajectory of the aircraft is plotted in Fig. 2 shows the intersection of the Falcon and 140 the anvil cloud. The other convective cell, Obs_Conv2, initiated at 06 UTC (14 LT) and was located at about 4.3°N and 114.4°E. It later produced an anvil of convective outflow that developed and moved north-westward similarly to Obs_Conv1.
It lasted several hours and started to decay from 10 UTC (18 LT).
The Falcon flight on the afternoon of November 19, 2011, was aimed at sampling the outflow of the Obs_Conv1 system. Krysztofiak et al. (2018) identified from humidity data and webcam images the times when the aircraft flew within the 145 convective outflow (i.e., in cloudy conditions) and when it was in cloud-free conditions. This information has been used here to show in Fig. 2 where the flight sampled cloud-free air or cloudy air. It shows that the Falcon aircraft sampled the Obs_Conv1 system multiple times but also flew within the Obs_Conv2 system at ~12.5 km altitude around 9.20 UTC (visible at 5.2 N and 114.8 E) and in cloud-free conditions below Obs_Conv2 on its way back to Miri around 9.50 UTC at an altitude ~6 km.
Since CHBr3 emissions and its marine boundary layer (BL) mixing ratios are large close to Borneo's west coast (Ziska et al., 150 2013;Fuhlbrügge et al., 2016), it was expected that CHBr3 would be transported from the BL aloft by the Obs_Conv1 and Obs_Conv2 systems. This scenario was confirmed by observations of elevated CHBr3 mixing ratios relative to the background conditions during the flight (Sala et al., 2014;Krysztofiak et al., 2018). CHBr3 measurements were performed by the GHOST instrument onboard the Falcon (Sala et al. 2014).

Model Description
We use the Chemistry-Coupled Aerosol and Tracer Transport model to the Brazilian developments on the Regional Atmospheric Modeling System (C-CATT-BRAMS) (Longo et al. 2013), which is a version of the CATT-BRAMS model (Freitas et al., 2009) coupled on-line with a chemistry model. This system is capable of resolving meteorological processes and the resultant tracer transport and chemistry. C-CATT-BRAMS has its original heritage in the Regional Atmospheric 160 Modeling System version 6 (RAMS) (Walko et al., 2000). RAMS is a fully compressible non-hydrostatic model consistent with Tripoli and Cotton (1982). RAMS can run in a nested grid configuration and includes various physical parameterizations to simulate sub-grid scale meteorological processes for turbulence, shallow cumulus convection, deep convection, surface-air exchanges, cloud microphysics, and radiation. BRAMS builds upon RAMS with the inclusion of several modifications that serve to improve the model performance within the tropics. For example, BRAMS includes an ensemble implementation of 165 the deep and shallow cumulus convection schemes, a soil moisture initialisation using model prognostication combined with a remote sensing rainfall product, and more realistic surface characteristics for vegetation type derived from the MODIS NDVI https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License.
The model represents microphysical processes using the single-moment bulk parameterization (Walko et al., 1995) whereby rain, cloud, pristine ice, snow, aggregates, graupel and hail are considered. The radiation scheme used in the model calculates 170 the effects of clouds, hydrometeors, and aerosols upon radiation (Toon et al., 1989). The model considers turbulent mixing using the turbulent kinetic energy (mean kinetic energy per unit mass for eddies in turbulent flow) as a prognostic variable (Mellor and Yamada, 1982).
The chemistry scheme used in C-CATT-BRAMS simulates gas and aqueous phase chemistry, photochemistry, uptake described by Henry's Law, and hydrolysis. Marécal et al. (2012) and Krysztofiak et al. (2012) provide a detailed overview of 175 the equations describing the chemistry solved by the model. To summarise, the model calculates chemical loss and production rates, and it computes chemical species concentrations in the gas phase, in cloud particles, and in rain droplets. To this end, the chemistry scheme couples with the microphysical scheme which explicitly resolves cloud and precipitation processes . In practice, the model considers, within the bulk microphysical scheme, the effects on the chemical species of condensation, evaporation, water vapour deposition, and sedimentation. In addition, the reversible exchange of gases 180 between the gas and aqueous phases (cloud and liquid hydrometeors) is estimated using Henry's Law and accommodation constants. Once within the condensed phase, the model includes the transfer of chemical species from within cloud particles to the different types of hydrometeors during coalescence and riming, and within the individual types of hydrometeor. In brief, it assumes Henry's law with cloud water and uses the production of precipitation to determine how much of the soluble trace gas is removed by wet deposition. We use retention coefficients to describe the proportion of a chemical compound that is 185 retained in the condensed phase upon the transition from one type of hydrometeor to another. We simplify the treatment of retention for the formation of ice precipitates by assuming a retention coefficient of 1 (i.e., the entirety of the compound) for all chemical species dissolved in liquid precipitate that undergo freezing. This is a frequently used assumption within washout schemes in global and regional scale chemical models.
The photolysis rates are computed on-line in the model using the Fast-TUV radiative model (Tie et al., 2003). This is done in 190 such a way as to consider the effects of clouds on photolysis rate in an interactive way.
The BrONO2 hydrolysis reaction within cloud particles and rain droplets has been added to the chemical scheme. Its mathematical implementation is described by Marécal et al. (2012). The reaction scheme considers the reaction within cloud particles and rain droplets separately using the mean mass radius and mean mixing ratios for cloud particles and rain droplets from the bulk micro-physical scheme, the thermal velocity, the gas phase diffusivity, and the accommodation coefficient of 195 BrONO2.

New Model Developments
For the model to simulate chemical and physical processes associated with CHBr3 degradation chemistry and its transport, several important changes had to be made to the model. A photochemical mechanism for the degradation of CHBr3 was developed, tested, and optimised for use in C-CATT-BRAMS . The development of the new 200 https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License. mechanism also included the estimation of the most favoured branching ratios for the halogenated peroxy reactions with the hydro-peroxy radical (XRO2 + HO2) (see Table 1 of Krysztofiak et al., 2012 for more details) using ab initio calculations of the standard reaction enthalpies. Reaction rates were either estimated from the analogous chlorine compounds or via a generalised expression. In addition, to properly simulate the uptake and washout of PGs into cloud particles and rain droplets, Henry's Law coefficients had to be estimated using predictive methods: the bond contribution method (Meylan and Howard, 205 1991) and the molecular connectivity index (Nirmalaklandan and Speece, 1988) for the brominated organic peroxides, CBr3O2H, CHBr2O2H, and the brominated carbonyl species, CBr2O and CHBrO. Krysztofiak et al. (2012) discusses the validity of these estimates. The Henry's Law constants of bromine species are shown in Table 1. Note that the information on BrONO2 is not included in Table 1 since it undergoes rapid hydrolysis in water and thus its removal is uptake limited (see Marécal et al. (2012) for details). 210  RCHO and RR'CO (Sander, 1999) These new developments coupled with a simple tropospheric chemistry scheme including carbon monoxide (CO), methane 215 hydrocarbon (NMHC) chemistry was added to the chemical mechanism in order to provide a more realistic description of the chemistry for the SHIVA real case study. The NMHC mechanism is a reduced version of the Regional Atmospheric Chemistry Mechanism (Stockwell et al., 1997) called Regional Lumped Atmospheric Chemical Scheme (ReLACS, Crassier et al., 2000). 220 We term this modified version of the ReLACS scheme RELASH. RELASH includes 60 chemical species, 166 chemical reactions, treats NMHCs with up to 8 carbon atoms via lumping scheme, and is designed to describe tropospheric chemistry only. The full chemical mechanism is described in S1 of the supplement.

Model Configuration
The model simulation was run for 3 days from 12:00 UTC November 17 to 12:00 UTC November 20 2011. We used a nested 225 grid configuration with three grids. The coarsest and largest grid covers from 90°E to 135°E and from 14°S to 23°N and uses a spatial resolution of 50 × 50 km; the next coarsest grid covers from 106°E to 123°E and from 2°S to 12°N at a resolution of 10 × 10 km; and the finest scale grid covers from 112.7°E to 117.4°E and from 3.3°N to 7.6°N and has a spatial resolution of 2 × 2 km. This horizontal spatial configuration allows the finest grid to completely include the two convective systems (Obs_Conv1 and Obs_Conv2) and the region covered by Falcon flight on the afternoon of November 19 th . The finer domain 230 and its associated model orography are plotted in Fig. 3. This illustrates well the abrupt topography on the west side of Borneo Island leading to the development of deep convection in sea-breeze conditions. The model has 53 vertical levels with varying vertical separation using finer resolution within the BL. The top of the model reaches to 26.6 km. The model meteorology was initialised and forced along the coarse grid boundaries of C-CATT-BRAMS using 6 hourly European Centre for Medium-Range Weather Forecasts (ECMWF) analysis fields for vector wind components, temperature, geopotential height, and specific 235 humidity. We used the ECMWF operational analysis at 0.5°× 0.5° resolution.
Within the coarsest two model grids we enabled the parameterisations for shallow cumulus convection and for deep convection.
We used the deep convection parameterisation from Grell and Dévényi (2002) as implemented in CATT-BRAMS in (Freitas et al., 2009). We allowed the model to resolve clouds and convective processes directly for the finest resolution grid. The topography used in the model has a 10 km resolution within the coarsest two grids and a 1 km resolution within the finest grid. 240 Sea surface temperatures (SSTs) were initialised using the satellite observed weekly average SSTs (Reynolds et al., 2002).
To describe CHBr3 emissions, we have implemented the emission inventory of . This is a bottom-up inventory based on the atmospheric and oceanic measurements of the HalOcAt (Halocarbons in the Ocean and Atmosphere) database project (https://halocat.geomar.de/). Using SHIVA flight measurements and the TOMCAT CTM, Hossaini et al. Various chemical species were initialised and forced along the coarse grid boundaries using 6 hourly output from the TOMCAT 250 https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License.

Results
Consistent with the objectives of this paper, the results shown and discussed in this section are only those of the finest resolution grid (2 ×2 km, see Fig. 3) since it gives a detailed description of the meteorology and chemical composition within and in the 260 vicinity of deep convection developing over the west coast of Borneo.

Meteorology
Simulations with limited area models with horizontal resolutions of the order of 1 km, as in the present study, are largely used to study in detail the development of convective systems since they provide an explicit representation of their dynamical and thermodynamic processes. At this resolution, there is no need to use sub-grid-scale parameterizations for convection. But even 265 at this fine resolution, modelling tropical deep convective systems remains a challenge when one wants to reproduce the exact time, intensity and structure (extent of convective cloud component versus stratiform cloud component) compared to observations. This is particularly true in maritime conditions because of the uncertainties in the representation of hydrometeor properties and processes and because of their sensitivity to large scale meteorological conditions (e.g., Varble et al. 2014  To show the evolution of the modelled convective systems, we plot in Fig. 5 the simulated cloud top height derived by the simulation from 05 UTC to 12 UTC on November 19 th , 2011. For this, we use the total condensed water field (the sum of liquid plus ice) to discriminate between cloud and cloud-free grid points with a threshold of 0.01 g kg -1 to filter out noise in 275 the cloud ice field. Further, deep convection is very active at this time of the year, high thin cirrus clouds associated with earlier distant convection are also present in the simulation. Since cirrus can hide the development of deep convection occurring below, we add the condition that the model level below a grid box with cloud present must have a total condensed water mixing ratio > 0.025 g kg -1 . This threshold may lead to a slight underestimation of the cloud top height and horizontal extent of the convective systems. Therefore, Fig. 5 is mainly used here to show the general temporal and spatial development of the 280 simulated convective systems but does not provide a precise measure of the cloud top height and spread of the anvil. Figure 5 https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License.
shows that the model simulates three deep convective systems that develop during the afternoon. These systems (called hereafter Mod_Conv1, Mod_Conv2 and Mod_Conv3) can be identified and followed in time by coloured rectangles.
Consistent with the observations, the analysis of the simulated meteorological fields show that all three systems are triggered inland close to the coast in the early afternoon from the large-scale low-level winds (north/north-westerlies) enhanced by local 285 sea-breeze that encounters the fairly steep topography of west Borneo (Fig. 3). This process is illustrated in Fig. 6a, which shows the simulated low-level wind direction and intensity at 06 UTC (14 LT, a time when convection is at an early stage) and the associated temperature field that exhibits a sea-land positive gradient. After 06 UTC the deep convective systems move offshore towards the west/north-west driven by the upper tropospheric winds, and they develop an anvil from their outflow.
Upper tropospheric winds at 09 UTC (17 LT) are shown in Fig. 6b. Compared to the evolution of the observed brightness 290 temperatures ( Fig. 1), the vertical extension of the convective part of the systems during the mature stage tends to decrease a bit too rapidly in the model (Fig. 5), likely due to a too rapid removal of precipitation related to uncertainties in the microphysical parameters. The quantitative characteristics of the observed (Obs_Conv1, Obs_Conv2) and modelled (Mod_Conv1, Mod_Conv2, 295 the uncertainty is ~0.5 km or lower. For the model, we use a set of cross sections from the 3D-fields to estimate the model cloud tops. Table 2 shows a general agreement on times and altitudes between the observations and the model.
Overall, from collating the information from Fig.1, Fig. 5, and Table 2, we find that the simulation represents the general 300 characteristics of the observed convective systems well, in particular: -the origin of their development from the interaction of the large-scale flow and the steep orography of the west coast of Borneo combined to local effects, -the location of the initial convective cell about 30 km inland on the west coast of Borneo, -the development of an anvil (the stratiform part) from the convective outflow during the afternoon moves off-shore, 305 -the cloud top height of the outflow, -the transport of the convective systems north-westwards and westwards, -the duration of the system of several hours and decay during early evening.
The main discrepancy is that the condensed water in the simulated convective part of the systems tends to precipitate a bit too efficiently compared to observations. Nevertheless, in its early stages, the convective part of the system, as evidenced by 310 condensed water, reaches altitudes greater than 14.5 km. Achieving this altitude gives a strong indication that the intensity of the main updraft transporting bromoform into the upper troposphere is predicted well by the model.
In conclusion, the model is able to simulate the general meteorology of the observed convective systems, at least within the constraints and uncertainties of km-scale modelling of convection (Varble et al. 2014).
The three convective systems we examine detrain into altitudes ranging between 11 and 15 km. According to Corti et al. (2005) 315 and Corti et al. (2006), the LZRH can be as low as 11 km for air masses within ice clouds due to the effects of their radiative properties. Ice clouds are present in the anvils of all three of the simulated systems, and they could cause a shift in the radiative balance and sufficient heating to lower the altitude of the LZRH. This would imply that the simulated air masses could gain positive buoyancy sufficient to reach the stratosphere over long enough time and large enough spatial scales. Thus, the study of the chemistry and washout within these systems could have relevance for the transport of CHBr3 and its PGs to the 320 stratosphere.

Bromoform Statistics and Convective Transport Efficiency
Before discussing the results of the simulated chemistry in detail (section 4.3), we evaluate if the simulation gives reasonable results for CHBr3 concentrations compared to the aircraft observations. We use statistical characteristics for this comparison, and we do not make a direct comparison of the chemical species' mixing ratios along the flight track. We choose this approach 325 for two reasons. First because of differences in location and timing between the observed and simulated convection events, and, second, because of spatial uncertainties in the emission inventory used in the simulation. This approach allows a clearer comparison of the observations and simulation by removing effects arising from inherent spatial uncertainties.
In order to diagnose the convective transport efficiency for both observed and simulated systems we follow the approach proposed by Bertram et al. (2007) that was recently applied to study the SHIVA aircraft data (Krysztofiak et al., 2018). In 330 order to estimate the air fraction, f, originating from the boundary layer (BL) and transported by convection we use the relationship: where the mean mixing ratios in the boundary layer, the upper troposphere within the convective systems, and the upper troposphere in the vicinity but outside the convective systems are represented by [X]BL, [X]UTconv, and [X]UTnoconv, respectively. 335 Table 3 shows the modelled and observed mixing ratios of CHBr3 that are used to calculate the f fraction. The mixing ratios are divided into three subsets corresponding to the boundary layer ([X]BL), the upper troposphere within the convective systems ([X]UTconv) and in the upper troposphere in the vicinity but outside the convective systems ([X]UTnoconv). The details on how the estimates from the observations and from the model were determined are given in Appendix A.   A higher f fraction is calculated for Mod_Conv2 meaning that this system was more efficient for transport of CHBr3 from the BL to the UT. However, this high fraction f is consistent with the average value calculated from all SHIVA aircraft data (0.29 ± 0.25) determined by (Krysztofiak et al., 2018) using carbon monoxide measurements from the SPIRIT instrument (Catoire 345 et al. 2017) and the GHOST CHBr3 measurements onboard the Falcon (Sala et al., 2014). Table 3  We now examine the magnitude of the CHBr3 mixing ratios in the BL and in the UT in both inside and outside of the convection 350 using Table 3 and Figure 7. The box and whisker plots in Figure 7 provide complementary statistical information on the observed and simulated CHBr3 mixing ratios.
In the BL, the Mod_Conv1 mean mixing ratio is higher compared (~22% higher in terms of the median, see Fig. 7; ~16% higher in terms of the mean, see Table 3) to the observations. Meanwhile, the mean mixing ratio in the BL below Mod_Conv3 is a bit lower than that observed, and Mod_Conv2 shows the lowest mean there (Table 3). This is likely related to two combined 355 factors: the emissions are weaker in the southern part of Borneo's west coast where Mod_Conv2 takes place (see Fig. 4); and Mod_Conv1 initiates in closer proximity to the coast compared to the other two systems where CHBr3 emissions and BL mixing ratios are higher. In the BL, there is a large spread in the observations because of the very large local variability of the emissions that the model cannot capture due the resolution of the CHBr3 emission inventory we used. Nevertheless, the median BL mixing ratios of all three lie within the 25 th and 75 th percentile of the observed BL mixing ratios, and the entire range of 360 simulated mixing ratios across all three simulated systems lie within the complete range of observed mixing ratios in the BL.
The mean background UT (UTnoconv) mixing ratios in the simulation, which largely depend on the chemistry initial conditions from the TOMCAT simulation, are underestimated, which leads to lower mixing ratios both within and out of the convection (differences of 0.17 to 0.22 pptv in UTnoconv, and of 0.11 to 0.17 pptv in UTconv). Hossaini et al. (2013) previously showed a comparable 0.08 pptv average negative bias in TOMCAT relative to the SHIVA aircraft measurements of CHBr3 throughout 365 the entire duration of the flight on the afternoon of November 19 th . If we consider the higher spatial resolution of our simulations and the smaller domain considered for the statistics compared to TOMCAT, these remaining differences appear consistent with one another.
In the UT, the underestimate in CHBr3 mixing ratio is more pronounced outside of the convection, which we suspect is linked to an underestimate in the emissions somewhere to the East of Borneo where these background UT air masses originate from 370 in the TOMCAT simulation. This finding is consistent with Fuhlbrügge et al. (2016) who showed that local sources alone cannot account for the observed CHBr3 levels in the UT. Furthermore, Keber et al. (2020) indicate that underestimates in the background tropical UT might arise due to underestimates in extratropical CHBr3 sources. Meanwhile, the simulated mixing ratios in the UT affected by convection (UTconv) are a much more direct measure of the emissions along the west coast of Borneo and are simulated by C-CATT-BRAMS. Background UT air mixes within the convective outflow during the 375 detrainment process in the UT. This mixing process could explain why mixing ratios simulated within the convective outflow are slightly lower than observed CHBr3.
In the background UT (UTnoconv), we see small differences in the mixing ratios in the vicinity of the three simulated convective systems with slightly higher values and variability for Mod_Conv2 and Mod_Conv3 because they developed in locations https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License.
previously affected by convective transport of bromoform. The three convective systems show similar results for [X]UTconv 380 though with a higher variability in Mod_Conv1 related to a more elongated horizontal extent of the anvil (see Fig. 5).
Note that the results presented in this section have little sensitivity to the threshold in ice concentration used to define the sampling of grid points within the convection and outside in its vicinity (see Supplement S2).

Bromoform 385
In this section, we analyse the cross sections of the simulated chemical fields of CHBr3 within the convective systems. As an illustration of how CHBr3 evolves during the convection, Fig. 8 shows a vertical cross section of the 5-hour time evolution of CHBr3 mixing ratios in the central part of the Mod_Conv3 system, together with contours of total condensed water from cloud and precipitation to locate the convective ascent and associated outflow. We selected Mod_Conv3 since it corresponds mostly close in space to Obs_Conv1. We can see the complete evolution of the convective system from a situation with elevated 390 CHBr3 concentrations in the boundary layer close to the location of the convection (of up to 2.1 pptv) at the very early stage of the system. Then the convective column ascends in a relatively vertical fashion and afterwards develops an anvil on its west side. The concentrations in the anvil are naturally highest closest to the point of convective detrainment and reach up to 0.9 pptv at 9 UTC (17 LT) and begin to decrease after one hour (up to 0.75 pptv) as the anvil is advected north westward by the high-altitude winds. 395 The analysis of the transport, chemical processes, and Br-atom speciation done hereafter is based on vertical cross sections chosen in the central part of each convective system at the time when the anvil is in its mature stage; the precipitation and vertical transport within the convective column are also near their maximum at this stage. This means that these cross sections are representative of the most intense convective activity which demonstrate the combined effects of intense vertical transport, washout, and development of the anvil. All of the numbers presented in the following sections correspond to those of the cross-400 sections presented in the figures. Also, to be able to compare the contribution of the different bromine species to the total Bratom mixing ratios all the figures hereafter are expressed as Br-atom mixing ratios (henceforth known as pptv Br).
We show the concentrations of CHBr3 in the three convective systems in Figs. 9 (a), (b), and (c). Each of the three simulated convective systems exhibit different CHBr3 mixing ratios within their convective columns and within their outflow anvils. This variability is because they each entrained different boundary layer mixing ratios of CHBr3, and they also detrained into 405 UT regions with slightly differing CHBr3 backgrounds and have different transport efficiencies (Table 3).
Despite these variations, we see a consistent result in the Br-atom speciation and mixing ratios of each convective system in Figs. 9 and 10 (expressed as the percentage of contribution). CHBr3 is elevated above background levels in all of the atmospheric regions dynamically linked to the boundary layer (surface to 600-800m height) on a timescale well below the lifetime of CHBr3 (i.e., the boundary layer, convective columns, and convective outflow). We can see that the CHBr3 in these 410 air masses has only undergone limited photochemical ageing because CHBr3 accounts for >85% of the total Br mixing ratio https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License. (Figs. 10 (a), (b), and (c)) in these atmospheric regions. However, we consistently see lower CHBr3 contributions to the total Br mixing ratio in atmospheric regions above the boundary layer not affected either directly or indirectly by convection. These regions include the low (from the top of the boundary layer to ~2 km height), mid (from ~2 km height to ~8 km) and uppertroposphere (from ~8 km height to ~13-14 km) where the model typically simulates 0.45-0.9 pptv Br of CHBr3 accounting for 415 60-70% of the total Br mixing ratio in the absence of convection to any vertical level. We can see some evidence of elevated

Inorganic and Organic PGs
Figs. 9 (d), (e), and (f) show that there are relatively low levels of inorganic bromine (Br,Br2,BrO,HOBr,HBr,BrONO2) concentrations in the lowest areas of the boundary layer even in the areas not directly affected by convective precipitation with 425 values typically in the range of 0 to 0.4 pptv Br, i.e., <5 % contribution to the total Br mixing ratio in the boundary layer. The highest simulated inorganic bromine mixing ratios (0.3-0.4 pptv Br) in the boundary layer occur to the west of the Mod_Conv3 system still only contributes <10 % to the total boundary layer pptv Br (see Fig. 10 (f)). This spatial variability in the boundary layer inorganic bromine mixing ratios around each convective system arises due to differences in precipitation location and timing over the course of the simulation prior to November 19 th , 2011. Precipitation events occurring in the two preceding 430 days deplete the boundary layer of inorganic bromine due to washout (analysis not shown). Inorganic bromine concentrations are enhanced in the lower free troposphere between 1 and 4 km to the west of each convective system (Figs. 9 and 10 (d), (e), and (f)). There, total inorganic PGs peak up to 1 pptv Br, which constitutes a sizeable portion (up to 45%) of the total Br mixing ratio. Meanwhile, air masses in the convective column itself and convective outflow are almost entirely depleted of inorganic bromine with mixing ratios of <0.1 pptv Br and with contributions to the total Br mixing ratio well below 5%. 435 The highest bromine mixing ratios from organic PGs (CHBrO, CBr2O, CHBr2OOH, and CBr3COOH) are found mainly in the BL and lower troposphere (i.e., up to 0.2 pptv Br) where PGs are formed due to CHBr3 photochemical loss (Figs 9 and 10 (g), (h) and (i). Furthermore, we can see that within the convective systems, organic compounds are being driven from the low levels up to the upper troposphere in the main ascent and the outflow and show enhanced mixing ratios within the convective column compared to the free troposphere. Nevertheless, organic bromine exhibits much lower pptv Br compared to CHBr3 440 overall, and the contribution from the organic PGs is lower than the inorganic compounds within the lower troposphere above the BL away from convective systems. Among the three convective systems, Mod_Conv3 exhibits the highest concentrations of the organic PGs and relative contributions west of the convective system below 2.5 km in altitude. However, their relative https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License. contribution generally remains below 5% between the BL top and 5 km in altitude and around 1-2% within the convective systems and the outflow. 445

The Impact of PG Solubility
Given that washout is an important process within the convective systems, the relative solubilities of each component we look at is relevant for explaining the concentration levels of the inorganic and organic PGs. Table 1 shows the Henry's Law constants for the PGs. In order of increasing solubility, we first list the inorganic bromine PGs: BrO,Br,Br2,HOBr,HBr,BrONO2;and then the organic PGs: CHBrO, CBr2O, CHBr2OOH, and CBr3COOH. In discussions from this point on, we will classify the 450 inorganic PGs into two groups: soluble inorganic that includes HOBr, HBr, and BrONO2; and insoluble inorganic, comprising Br, Br2, and BrO. We also classify the bromo-carbonyls as insoluble organic and the bromo-methyl peroxides as soluble organic PGs. Note, that except for Br2, CHBr3 is insoluble relative to its PGs.
The gas phase mixing ratios of soluble inorganic bromine species are depleted to near zero in the convective columns and in the immediate area of detrainment for each system (Fig.11, a-c). Similarly, soluble inorganic species make almost no 455 contribution to the total gas phase Br within each convective system (Fig. 12, a-c). This depletion of soluble inorganic bromine species occurs even though their boundary layer mixing ratios range between 0.1 and 0.5 pptv Br and the soluble inorganic species form the bulk of the inorganic bromine at all levels in the troposphere outside of convective systems. This strongly implies that the model simulates the near total removal of the soluble inorganic bromine species within the convective columns.
The insoluble inorganic species (Figs. 11 and 12 (d), (e), and (f)) only make a negligible contribution to the total bromine 460 budget throughout the troposphere, only reaching peak mixing ratios of 0.1 pptv Br and 4% of the total bromine in areas of the lower troposphere not affected by convection. We also see no enhancement of the insoluble inorganic species (which include Br2) within the convection column or in the fresh convective outflow above the levels seen in the rest of the troposphere. There is also only a negligible contribution of insoluble inorganic bromine in the UT affected by previous convection as illustrated in Figs. 12 (e) and (f) (to the East of both Mod_Conv2 and Mod_Conv3). 465 The contribution to the total Br mixing ratio from the soluble organic bromine PGs is also negligible (at a maximum of ~1% in the low troposphere not affected by convection). The bulk of organic PG species are instead in the form of insoluble species.
The enhancements of the organic PGs we see in the convective system (up to 0.08 pptv compared to 0.02 pptv) in Figs. 9 (g), (h), and (i) are due to the insoluble bromo-carbonyls. These compounds contribute 86% of the PG bromine total.
In order to understand further the behaviour of inorganic bromine we need to examine its composition. HBr dominates the 470 inorganic Br-atom contribution in the regions of the UT outside of the convective systems (Figs. 13 (a) to (c)). It provides between 40-65% of the total Br mixing ratio in the regions of the low and mid-troposphere (1.5-4 km height) unaffected directly by precipitation washout (Fig. 14, a-c)

Discussion
A key finding is that inorganic bromine dominates the PG budget within the troposphere, yet despite this the inorganic PGs are almost entirely removed during convective transport by washout due to their solubility. We here argue that the regional 480 tropospheric composition present in our simulations is the underlying cause of this prevalence of HBr and in turn the washout of inorganic PGs that results from this.
Within the photochemical scheme of C-CATT-BRAMS, bromine atoms can react in the gas phase in one of three ways.
Br + O3 -> BrO + O2 (R1) Br + HCHO -> HBr (R2) 485 Br + HO2 -> HBr (R3) Probably of most significance in the troposphere, bromine atoms can react with ozone via R1 to form BrO. However, under relatively low ozone conditions, reactions R2 and R3 can begin to dominate the reaction budget for Br-atoms favouring the formation of HBr instead of BrO. This has important implications for the washout of Br within convective systems because HBr is extremely soluble but BrO is very insoluble. 490 The maritime continent has been noted for having low O3 levels throughout the depth of the troposphere during the winter monsoon resulting from westward transport of moist tropical air across the equatorial Pacific (Rex et al., 2014). The relatively low ozone levels are reproduced in our simulations (daytime values of as low as 10-15 ppbv at 2 km) and arise from the same mechanism across the equatorial Pacific represented in the TOMCAT global model forcing. This favours the formation of HBr and therefore this enhances the removal of inorganic bromine within the simulated convective clouds systems we study here. 495 The low levels of BrO have another implication. BrO is also an important precursor for other key inorganic bromine compounds in the troposphere.
BrO + HO2 → HOBr (R4) BrONO2(aq) → HOBr(g) + HNO3(g) (R6) 500 BrO is directly involved in the two chemical pathways leading to HOBr formation; either the direct formation pathway via R4 or via the two-step pathway involving the formation (R5) and then hydrolysis of BrONO2 (R6). Thus, the low levels of BrO can be implicated in the lack of significant quantities of HOBr (Figs. 13 and 14). This has another implication: HBr and HOBr react with one another in the aqueous phase to produce Br2 via reaction R7.
The atmospheric implication of R7 is that two soluble gases (HBr and HOBr) could potentially react rapidly within cloud and rain droplets to form the insoluble gas Br2. R7 is known to lead to the production and release of significant quantities of Br2 in other environments combining both the aqueous and gas phases, e.g., in the polar ice regions (McConnell et al., 1992) and 510 within volcanic plumes (Oppenheimer et al., 2006). In these examples the gas phase and aqueous phase chemistry combine to form a feedback loop leading to what is known as the "bromine explosion" (Wennberg, 1999). Thus, R7 could have the potential to significantly alter the Br-atom speciation within convective clouds in a short time in such a way as to reduce the PGs overall solubility; this would therefore promote the vertical transport of Br-atoms within convective systems even in the presence of abundant falling hydrometeors that would otherwise washout soluble gases like HBr very rapidly. In our simulation, 515 only negligible levels of Br2 form during the most active phase of each of the simulated convective systems (Fig. 15). At most we see up to 0.04 pptv of Br2 within Mod_Conv3, and this is only within the lowermost sections of the system and these levels do not propagate vertically into the UT.
From this analysis, we should expect more production of Br2 in convective systems in environments with higher levels of background ozone, which would consequently have more HOBr. A further reason is that the hydrometeors containing the 520 dissolved HBr and HOBr simply fall fast leading to a very short residence time that prevents their concentrations from accumulating to the point where R7 becomes important.
While other studies exist of convective scale modelling of ozone and aerosols (Crumeyrolle et al., 2008;Tulet et al., 2002), Marécal et al. (2012) is the only other study to have simulated the transport and photochemistry of bromoform at the convective scale and so we therefore compare our findings with that study. Marécal et al. (2012) showed in their analysis of simulations 525 of an idealized case study of a tropical deep convective system that most of the Br transported to the UT by convection is in the form of CHBr3, which is consistent with the results here. Nevertheless, in their case using an initial CHBr3 of 1.6 pptv in the BL (which is the most comparable to this study), we can see notable differences in the composition of the convective outflow. Both organic and inorganic PGs are in higher abundance within the upper reaches of each convective system in this study compared to Marécal et al. (2012). In particular, the mixing ratios of all the inorganic PGs are greatly increased by over 530 a factor of ten here compared to Marécal et al. (2012). As a result, Marécal et al. (2012) found a higher relative contribution to the total bromine PG mixing ratio in the UT for organic with respect to inorganic PGs, i.e., 86 % compared to 46% in this present study. The background tropospheric air contains elevated HBr levels in our simulation ( Fig. 13 and 14) resulting from photochemical ageing of convectively lofted CHBr3 represented both within the TOMCAT CTM initial and boundary conditions. In Marécal et al. (2012), the initial conditions are such that CHBr3 is mainly below 2 km and there is no CHBr3 535 above 5 km altitude. Furthermore, they used a PG background of zero in the mid to upper troposphere. Thus, we can conclude that the major differences in simulated inorganic and organic PG mixing ratios within the convective system arise as a result of the idealized PG background and initialization used within Marécal et al. (2012) compared to the more realistic simulation of the PG background (specifically HBr) in this study.
Another difference is that the simulations in Marécal et al. (2012) are set over Darwin as opposed to Borneo. The small 540 difference in latitude and atmospheric conditions is enough to cause longer CHBr3 lifetimes in our simulation with respect to https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License.
both photolysis (15.2 days compared to ~19 days here) and OH (22.6 days compared to 52.6 days here). Overall, both processes contribute to an increase in CHBr3 lifetime from 9 to ~14 days here (note both are fully consistent with the compiled ranges from Burkholder et al., 2018). The relative importance of photolysis has therefore increased in this study due to these changes, which could lead to an increase in the relative formation rates of inorganic bromine PGs compared to the organic PGs. This 545 occurs because the photolysis of bromine cleaves off single bromine radicals that then contribute directly to the inorganic PG budget. Furthermore, the remaining organic PGs contain one fewer bromine atoms as a result. Meanwhile, reaction with OH leads to the abstraction of a hydrogen atom, which creates organic PGs containing either two or three bromine atoms. Given the idealized initialization and experimental design within Marécal et al. (2012), it is not possible to properly assess these conjectures here, but it would be of interest to evaluate more formally differences in PG speciation arising in different 550 photochemical environments that display differences in the CHBr3 lifetime with respect to reaction with OH and to photolysis.
Also, Marécal et al. (2012) showed that significant Br2 can be released into the gas phase via aqueous processes in cloud droplets, but only when their idealized simulation was initialized with very high CHBr3 (40 pptv) in the BL. In the more realistic case study simulated here where boundary layer CHBr3 is ~2 pptv, we find that the production of Br2 via R7 also takes place but is a negligible contribution to the total inorganic bromine as shown in Fig. 15. This is consistent with the results of 555 Marécal et al. (2012) when using only 1.6 pptv as the initial CHBr3 in the boundary layer. This is because the simulations in Marécal et al. (2012) took place near Darwin where there was also still a relatively low O3 background (14 ppbv at 2 km).

Limitations
There are a few limitations to this study. First, it could be difficult to generalise these results to other tropical areas because we studied a region with a relatively low ozone background in the troposphere, which impacts the tropospheric bromine 560 chemistry. Furthermore, other tropical regions could have vastly different CHBr3 emissions, and in the case of much higher emissions, as was explored in Marécal et al., (2012), we could expect a larger role for Br2 formation. We also neglect other VSLS in our simulations and only focus on CHBr3, and as a result our findings are specific for the case where CHBr3 is the only VSLS. Including other VSLS would alter the relative contributions made by CHBr3 and its PGs to the vertical transport of bromine to the UT. 565 Another limitation is that our analysis is performed on cross sections of the most mature convective systems, and so our analyses offer a snapshot of the most intense convective activity and its effects.
Lastly, the performance of the various top-down and bottom-up CHBr3 inventories varies significantly by region and the divergent global emission estimates represent a significant source of uncertainty in estimates of strat-Bry VSLS (Hossaini et al., 2013). Despite the difference in simulated CHBr3 mixing ratios that we identify in the background UT we argue that the choice 570 in using the  emission inventory was the correct one. Here, we cite Hossaini et al. (2013) that determined that the  emissions gave the closest agreement to observations when evaluated in the TOMCAT CTM over the same region of Borneo. Indeed, our results show consistent CHBr3 mixing ratios compared to the simulations of Hossaini https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License. et al. (2013) for the November 19 th flight. Furthermore,  project a CHBr3 emission climatology ranging between 1200-1600 pmol m -2 hr -1 within the region of interest along the north west coast of Borneo while the emissions 575 measured locally (over several days) by Fuhlbrügge et al. (2016) range from 300-4300 pmol m -2 hr -1 . We conclude from this that  provides a good estimate of CHBr3 emissions over this specific area.

Summary and Conclusions
We used a convective scale model (2 × 2 km resolution) in order to gain a better understanding of the effects of tropical deep convection on the transport of CHBr3 and the speciation of its PGs in the troposphere. Until now, two modelling studies have 580 been carried out at the convective scale but only for idealised cases . Our objective was to go a step further by modelling a real case study of deep convection that occurred along the west coast of Borneo on the afternoon/early night of 19 November, 2011, during the SHIVA field campaign.
It was shown that the meteorological development of the convective systems in the model have general characteristics similar to those observed. We compared our modelled CHBr3 mean concentrations and convection transport efficiency to those derived 585 by Krysztofiak et al. (2018)  Despite variation in the timing, location, and availability of CHBr3 for entrainment in the BL below each convective system, the same general behaviour is observed across all three simulated convective systems. Most of the bromine (>85%) transported to the UT in each convective system is in the form of CHBr3. Within the convective systems, the remaining 1-2% of the total bromine present is mostly in the form of organic PGs, i.e., the insoluble brominated carbonyls CHBrO and CBr2O (86% as a 595 contribution to the total PG). This is despite the inorganic PGs making a larger contribution than the organic PGs to the total Br mixing ratios in the free troposphere (i.e., 45% versus only 5%). Falling hydrometeors within the convective column efficiently remove the inorganic PGs, whose tropospheric budget is dominated by the extremely soluble HBr gas. Overall, we conclude that organic PGs are more important than inorganic PGs for the vertical transport of bromine within the convective columns for the conditions that we study here. 600 The insoluble inorganic PGs, BrO and Br2, are only present at negligible mixing ratios and play no significant role in the vertical transport of bromine. Our interpretation is that the tropospheric inorganic PG budget is shifted heavily in favour of HBr formation due to the low background O3 mixing ratios simulated in this region. This limits the availability of HOBr leading to only very limited formation of Br2 within the cloud and rain droplets within the convective system resulting from the reaction between HBr and HOBr. More BrO and HOBr would form in cases with higher background O3, which would lead 605 https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License.
to enhanced Br2 formation within convective systems and a more important role of the inorganic PGs for the vertical transport of bromine.
Overall, these conclusions are valid in all parts of the convective system except for where the anvil detrains into residual convective outflow in the UT. In this case, the Br speciation within and around the anvil is perturbed with increases in the background UT CHBr3 and PGs produced by photochemical ageing of the CHBr3 within the UT. Thus, the exceptions to our 610 conclusions are due to factors external to the systems themselves.

Outlook
It would be of interest to evaluate findings from global CTM studies of VSLS transport to the UT and stratosphere in light of our findings based on modelling at the convection scale. This additional work, however, is beyond the scope of this study since it would require considerable additional technical work to reconcile the differences in spatial scale and conditions since global 615 CTMs derived grid scale mixing ratios representing the mean both within and outside of convective systems. We hope to reconcile these issues and test the hypotheses raised in this paper as part of future work. This future effort could provide insights into the processes represented within CTMs.

Appendix A
Here are the details on how we calculated the simulated and observed mixing ratios presented in Table 3. Following 620 Krysztofiak et al. (2018), the mean mixing ratio estimates based on Falcon measurements for the [X]UTconv and [X]UTnoconv (~11-13 km altitude range) were split according to humidity data and webcam images corresponding to cloudy or cloud-free conditions, respectively. [X]BL are data gathered during the flight ascent/descent from/to Miri between the surface and 400 m height combined with Sonne ship measurements  close to the convective system.
The estimates from the simulations are calculated similarly for each of the modelled convective systems and in a manner as 625 consistent as possible with Krysztofiak et al. (2018). For [X]UTconv and [X]UTnoconv, a box of 1°x1° and 11-13 km altitude is selected around the convective system and [X]UTconv are selected where cloud ice is present. We use two different ice concentration thresholds to define both the air masses within the anvil cloud and those outside of it: a cut-off of 0.01 g kg -1 to define air masses within the cloud, and 0.005 g kg -1 to define those outside of it that are part of the simulated upper tropospheric background. We use two thresholds to define these separate air masses to ensure we derive an adequate separation between 630 the two; air masses lying between the lower and upper cut-off show mixing and dilution of the convective outflow and neither represent background or convective outflow concentrations in a clear way. To demonstrate the sensitivity of our results to this choice, we perform a sensitivity test with two different single ice concentration thresholds to define air inside or outside the convective cloud: 0.01 g kg -1 and 0.005 g kg -1 . These results are presented in the supplement S2. For [X]BL, a box of 0.5° × 0.5° and 0-400 m altitude is selected around the area where the initial convective cell starts its development. For each system, 635 https://doi.org/10.5194/acp-2020-655 Preprint. Discussion started: 1 September 2020 c Author(s) 2020. CC BY 4.0 License. two model outputs one hour apart were selected during the main phase of the outflow development for [X]UTnoconv and [X]UTconv, and at the times when the initial convective cell starts developing for [X]BL. The times and boxes chosen are given in Table   A1. Author contributions. 640 PDH: co-designed the study, wrote the main text of the paper, ran the model, analysed the model results, and created many of the figures. VM: co-designed the study, helped develop this version of C-CATT-BRAMS, wrote some of the text, and created some of the figures.