Bromoform and dibromomethane in the tropics: a 3-D model study of chemistry and transport

. We have developed a detailed chemical scheme for the degradation of the short-lived source gases bromoform (CHBr 3 ) and dibromomethane (CH 2 Br 2 ) and imple-mented it in the TOMCAT/SLIMCAT three-dimensional (3-D) chemical transport model (CTM). The CTM has been used to predict the distribution of the two source gases (SGs) and 11 of their organic product gases (PGs). These ﬁrst global calculations of the organic PGs show that their abundance is small. The longest lived organic PGs are CBr 2 O and CHBrO, but their peak tropospheric abundance relative to the surface volume mixing ratio (vmr) of the SGs is less than 5%. We calculate their mean local tropospheric lifetimes in the tropics to be ∼ 7 and ∼ 2 days (due to photolysis), re-spectively. Therefore, the assumption in previous modelling studies that SG degradation leads immediately to inorganic bromine seems reasonable.


Introduction
Bromine-containing very short-lived species (VSLS) are expected to provide an additional supply of inorganic bromine (Br y ) to the stratosphere (e.g. WMO, 2007). Emissions of such species are predominately of natural oceanic origin and have been shown to exhibit large variability, particularly in tropical coastal regions that harbour substantial amounts of macro-algae (e.g. Quack and Wallace, 2003;Carpenter et al., 2009). It is also at tropical latitudes that deep convection allows the rapid ascent of such species from the marine boundary layer to the tropical tropopause layer (TTL). The rate of transport of these species to, and through, the TTL is currently under discussion (e.g. Krueger et al., 2008;Fueglistaler et al., 2009).
Current estimates of the contribution of inorganic bromine derived from VSLS (Br VSLS y ) to the stratospheric Br y budget range from 3 to 8 pptv (Law and Sturges et al., 2007), with a more recent value of approximately 5 pptv derived by Dorf et al. (2008). Quantifying this additional source of bromine is important due to its role in catalytic ozone depletion in the stratosphere. In addition, it is expected that reactive Br VSLS y will impact tropospheric composition (e.g. Von Glasow et al., 2004).
Two distinct pathways have been identified leading to the arrival of Br VSLS y in the stratosphere; namely source gas injection (SGI) and product gas injection (PGI) (e.g. Ko and Poulet et al., 2003). SGI refers to the transport of a source gas (SG, e.g. bromoform, CHBr 3 ) to the stratosphere, where upon degradation it will provide an in-situ source of Br y . In contrast, PGI is the cross-tropopause transport of bromoorganic intermediates (e.g. CBr 2 O) and also inorganic products (e.g. HBr, BrO, Br), produced from SG degradation in the troposphere. The efficiency of both SGI and PGI depends largely upon the photochemical loss of source gases (mainly via reaction with OH or photolysis) and rate of removal of degradation products (via wet deposition) versus the timescale for troposphere-stratosphere transport. For this reason, it is of interest to investigate the impact of convection on this additional source of bromine, Br VSLS y , in model experiments.
Previous model work has concentrated on quantifying Br VSLS y from the more abundant very short-lived source gases such as CHBr 3 . From a 2-D model study, Dvortsov et al. (1999) concluded that CHBr 3 contributes around 1 pptv additional Br y to the lower stratosphere (LS). Similarly, Nielsen and Douglass (2001) also derived a value of 1 pptv from 3-D simulations with ∼50% of this from SGI. In both of these studies the lifetime of Br y following source gas degradation was assumed to be 10 days. More recently, Sinnhuber and Folkins (2006) used a 2-D mechanistic model of the tropical atmosphere to estimate that CHBr 3 contributes between 0.8-2.1 pptv bromine via both SGI and PGI to the lower stratosphere with assumed Br y lifetimes of 10-100 days. From this study, the contribution from SGI was approximately 0.5 pptv. The most recent model study, Kerkweg et al. (2008), confirmed earlier suggestions that CHBr 3 contributes "substantial amounts" of Br y to the lower stratosphere and that Br VSLS y should not be neglected in stratospheric modelling.
The impact of Br VSLS y in the stratosphere has been studied with multi-annual simulations by Feng et al. (2007) using the SLIMCAT 3-D chemical transport model (CTM). Salawitch et al. (2005) also performed a similar 2-D model study. Both studies report an ∼10 DU decrease in the ozone column with an additional 6 pptv of Br VSLS y in the lower stratosphere. The impact of this additional bromine depends on the aerosol loading; the bromine causes a larger decrease in ozone when the aerosol loading is high and ClO is elevated.
The model studies discussed above have not directly considered the bromo-organic products (i.e. product gases, PGs) formed following source gas degradation. This omission is addressed in this study, which evaluates the contribution of CHBr 3 and CH 2 Br 2 to the stratospheric Br budget, along with the relative contribution of SGI and PGI. Furthermore, results include novel estimates of the major and minor PGs formed following CHBr 3 and CH 2 Br 2 removal. Vertical source gas profiles are compared to measurements made during several aircraft campaigns in the tropical troposphere and near-tropopause regions. To date, there have been no measurements of organic PGs in the tropical atmosphere. We provide the first model estimates of the local lifetimes and abundances of these species in the tropical atmosphere.
Section 2 describes the derived chemical scheme for degradation of CHBr 3 and CH 2 Br 2 . Kinetic and mechanistic assumptions are also discussed. Section 3 contains a description of the basic CTM setup, along with details of sensitivity runs carried out. Section 4 presents the model results. Conclusions and recommendations for future research are discussed in Sect. 5.

Chemistry scheme
Here we outline our chemical scheme to describe the tropospheric degradation of bromoform and dibromomethane. Reasonable mechanistic and kinetic assumptions have been made and are discussed. Kinetic data either is taken from Sander et al. (2006) (hereafter "JPL") or the Leeds Master Chemical Mechanism (hereafter "MCM", see http://mcm. leeds.ac.uk/MCM/). A summary of reactions and kinetic data used within the scheme is given in Tables 1 and 2 for CHBr 3 and CH 2 Br 2 , respectively.

Bromoform
The degradation of CHBr 3 has been examined in previous theoretical studies (e.g. McGivern et al., 2002McGivern et al., , 2004. Its local tropospheric lifetime is ∼26 days with photolysis being the dominant loss process (e.g. Ko and Poulet et al., 2003). Our bromoform scheme considers 7 organic species: CHBr 3 , CBr 3 O 2 , CHBr 2 O 2 , CBr 3 OOH, CHBr 2 OOH, CBr 2 O and CHBrO and is summarised in Fig. 1. The following subsections discuss the details of this scheme.

Removal of CHBr 3 source gas
The scheme assumes removal of CHBr 3 occurs via reaction with OH/Cl radicals and also, more rapidly, by photolysis Reactions (R1-R3). The rates of reaction with OH and Cl (k 1 , k 2 ) are calculated using the JPL recommended temperature-dependent expressions. The rate of photolysis (j 3 ) is calculated using JPL absorption cross section data along with a parameterisation for their temperaturedependence (Moortgat et al., 1993). The quantum yield for Br atoms following R3 is assumed to be unity. It is also assumed that the immediate products of Reactions (R1-R3) (CBr 3 and CHBr 2 ) will be rapidly oxidised under tropospheric conditions.

Removal of peroxy species
The two peroxy radicals formed in Reactions (R1-R3) are assumed to be removed via reaction with NO and HO 2 Reactions (R4-R9). Self-reaction of these species is deemed slow and is therefore not considered here. The CBr 3 O 2 +NO Reaction (R4) is assumed to produce CBr 2 O, an expected major product of bromoform degradation (Ko and Poulet et al., 2003). Excited intermediates, such as CBr 3 OONO* (not considered here), are expected to fragment rapidly to form CBr 3 O, which itself would undergo a rapid decomposition to CBr 2 O (e.g. McGivern et al., 2002). The rate constant for Reaction (R4) (k 4 ) is calculated using the recommended JPL expression for the analogous species CCl 3 O 2 . Similarly, rate constants for the CBr 3 O 2 +HO 2 reactions (k 5 , k 6 ) are assumed equal (i.e. equal branching ratio of products) and taken from the MCM. These reactions produce CBr 2 O and the minor hydroperoxide product CBr 3 OOH, respectively.
The reaction of CHBr 2 O 2 +NO Reaction (R7) is assumed to produce a second major product of bromoform degradation, namely formyl bromide (CHBrO). As for Reaction (R4) it is likely that Reaction (R7) would proceed via an excited intermediate (not considered here as sufficiently short-lived) such as CHBr 2 OOH*. The rate constant for this reaction (k 7 ) is assumed equal to the analogous species CHCl 2 O 2 and taken from the MCM. This is also the case for the CHBr 2 O 2 +HO 2 Reactions (R8, R9) which produce CHBrO and the minor hydroperoxide, CHBr 2 OOH respectively. A further reaction pathway for these peroxy radicals (not considered here) is that with NO 2 . This would likely lead to the formation of Br-containing peroxynitrates (e.g. Ko and Poulet et al., 2003). In future versions of the degredation scheme these reactions will be considered.

Removal of minor end products
Removal of the two hydroperoxide species produced in Reactions (R6) and (R9) is assumed to be achieved via reaction with OH (Reactions R10, R12) and also by photolysis (Reactions R11, R13). Rate constants for the OH reactions are assumed equal to that of the analogous chlorine-containing species, CCl 3 OOH and CHCl 2 OOH, from the MCM. The photolysis rates are calculated using the absorption cross sections of methylhydroperoxide (CH 3 OOH). Reaction of these hydroperoxides with OH leads to the reformation of the respective peroxy radical (initially formed in Reactions R1 and R3).

Removal of major end products and Br y
Removal of the major products of bromoform degradation, CBr 2 O and CHBrO, is assumed to occur via photolysis (Reactions R14, R15). For CBr 2 O+hν we assume a yield of two Br atoms, with other photolysis pathways, such as that leading to HBr production, not considered. Similarly, for CHBrO+hν it is assumed the quantum yield for Br atoms is unity. Photolysis rates for both reactions are calculated using the recommended JPL cross section data.
In the current scheme all inorganic bromine species produced are grouped together as Br y without any further partitioning. Depending on the model run (see Sect. 3.2), Br y is removed in the troposphere by washout given a specified assumed lifetime.

Dibromomethane
The degradation of CH 2 Br 2 has also been examined in previous theoretical studies (e.g. McGivern et al., 2002McGivern et al., , 2004. The local tropospheric lifetime is quoted as ∼120 days with reaction with OH being the dominant loss process (Ko and Poulet et al., 2003). Our scheme considers six organic species, CH 2 Br 2 , CHBr 2 O 2 , CH 2 BrO 2 , CHBrO, CHBr 2 OOH and CH 2 BrOOH. The major products of CH 2 Br 2 degradation are expected to be CHBrO and Br y with CHBr 2 OOH being a minor product.

Removal of CH 2 Br 2 source gas
Our scheme assumes removal of CH 2 Br 2 is achieved via reaction with OH/Cl radicals and also, less rapidly, by photolysis (Reactions R17-R19). The rates of reaction with OH and Cl (k 17 , k 18 ) are calculated using the JPL temperaturedependent expressions. The rate of photolysis (j 19 ) is calculated using JPL absorption cross section data at 295-298 K. The quantum yield for Br atoms is assumed to be unity. As for the CHBr 3 scheme, it is assumed that following H abstraction and photolysis, the immediate products of source gas degradation (CHBr 2 , CH 2 Br) will be rapidly oxidised under tropospheric conditions forming associated peroxy radicals.

Removal of peroxy species
The two peroxy radicals formed in Reactions (R17-R19) are assumed to be removed via reaction with NO and HO 2 . Loss of CHBr 2 O 2 via these Reactions (R20-R22) is treated as that described in Sect. 2.1.2 for the bromoform scheme.
The CH 2 BrO 2 +NO Reaction (R23) is assumed to produce NO 2 and Br y . The rate constant for this reaction (k 23 ) is calculated using the recommended JPL expression. For reactions with HO 2 Reaction (R24, R25), rate constants are taken from the MCM. The products of these reactions are Br y and the hydroperoxide CH 2 BrOOH, respectively.

Removal of end products
Removal of CHBrO Reaction (R26), produced in Reactions (R20) and (R21), is achieved via photolysis as discussed in Sect. 2.1.3. Similarly removal of CHBr 2 OOH (Reactions R27, R28) produced in Reaction (R22) is analogous to removal in Reactions (R12) and (R13) (i.e. by reaction with OH and by photolysis respectively). Removal of CH 2 BrOOH, produced in Reaction (R25), has yet to be considered and is also assumed to be removed via reaction with OH (Reaction R29) and also by photolysis (Reaction R30). The rate constant for the OH reaction (k 29 ) is taken from the MCM and the reaction products are expected to be the peroxy species CH 2 BrO 2 and water. The photolysis rate (j 30 ) is calculated using the cross sections of CH 3 OOH.

TOMCAT/SLIMCAT 3-D CTM
TOMCAT is an off-line 3-D CTM described in detail by Chipperfield (2006). The model has performed well in previous tropospheric studies and has been shown to simulate key chemistry and transport reasonably (e.g. Arnold et al., 2005). The model uses the Prather (1986) conservation of secondorder moments advection scheme, a parameterisation of convection (Stockwell and Chipperfield, 1998) and also a parameterisation of boundary layer mixing (Holtslag and Boville, 1993). The CTM in TOMCAT mode uses a terrain-following hybrid σ -p vertical coordinate and diagnoses the large-scale vertical motion from divergence. The CTM has an option ("SLIMCAT") for running with isentropic (θ) levels in the upper troposphere and stratosphere with the vertical motion calculated from heating rates. The SLIMCAT model only considers transport by large-scale advection; there is no parameterisation of convection and boundary layer mixing. The CTM includes a scheme to calculate trajectories (e.g. Monge-Sanz et al., 2007).

Simulations
In all TOMCAT simulations described here (see Table 3) the resolution of the model was 5.6 • ×5.6 • with 38 (or 31) vertical levels (∼1 km deep in mid troposphere) extending from the surface to ∼35 km. The model was forced using the European Centre for Medium-Range Weather Forecasts (ECMWF) 6-hourly analyses. The model was initialised on 1 January 2006 and run for 2 years. Year 1 was treated as model spin-up and year 2 output (2007) was saved every 3.75 days for analysis. SLIMCAT simulations differed in that the model was spun up for 7 years, due to the slower circulation, prior to analysis of 2007 output.
For the "base run" (run B), the TOMCAT model included specified oxidant fields along with the CHBr 3 and CH 2 Br 2 degradation schemes described in Sect. 2. Monthly mean diurnal mean fields of the concentration of fixed species (e.g. OH, NO, HO 2 ) were read from a previous TOMCAT full chemistry run for 2005. The background concentration of atomic chlorine, which was not calculated in the tropospheric full chemistry run, was set to 1×10 4 molecules cm −3 . The model chemical scheme used a climatological tropical ozone profile for photolysis calculations (Chipperfield, 1999). Figure 3 shows example tropical zonal mean profiles of temperature and the primary oxidant OH. The mixing ratio of CHBr 3 and CH 2 Br 2 source gases were fixed uniformly in space and time at 1.2 pptv in the bottom two layers of the model in the tropical regions (±20 • ). This value is typical for both mean CHBr 3 and CH 2 Br 2 in the marine boundary layer (MBL). Quack and Wallace (2003) report background MBL CHBr 3 in the range 0.5-1.5 pptv. Similarly for CH 2 Br 2 , Butler et al. (2007) report CH 2 Br 2 in the range 0.6-1.3 pptv for the tropical MBL. Furthermore, the use of a SG volume mixing ratio (vmr) of 1.2 pptv provides the best fit to observed profiles (see Sect. 4). In this study we are interested in the relative mixing ratios of bromine species in the TTL compared to the surface and we do not need to introduce the complication of specifying uncertain emission fluxes. All other advected tracers were initialised at zero at the start of the simulation. The lifetime of Br y was assumed to be infinite for run B. A number of sensitivity runs were also performed. Run S NOCONV differed from run B in that model convection was switched off (note, mixing in the PBL remained switched on). Runs S 10 , S 20 and S 40 differed from run B in that the lifetime of Br y was set to 10, 20 and 40 days below the cold-point tropopause (CPT), respectively. Run S 2OH differed from the base in that model [OH] was doubled. Run S L31 differed in that the model employed coarser vertical resolution (factor of 2) above ∼300 hPa and therefore had only 31 levels. Finally, SLIMCAT simulations were also carried out in which artificial mixing in the tropical troposphere was assumed by fixing both CHBr 3 and CH 2 Br 2 SGs in the lower 8 levels of the SLIMCAT σ -θ model (surface to ∼10 km). For run S SLIMCAT the assumed lifetime of Br y due to washout was infinite. For runs S SLIMCAT10 , S SLIMCAT20 and S SLIMCAT40 the assumed lifetime of Br y was 10, 20 and 40 days respectively. It should be noted that for SLIMCAT simulations (due to the artificial mixing in the troposphere) the assumed washout is only "switched on" between ∼10-17 km. All other aspects of chemistry were consistent between the two models. 4 Results and discussion Figure 4 shows the mean modelled loss rates (due to reaction with OH and photolysis) and the local photochemical lifetimes for CHBr 3 and CH 2 Br 2 in the tropics. Results from this analysis are also summarised in Table 4. For CHBr 3 , the dominant loss process is photolysis. The calculated local lifetime of CHBr 3 (τ local ) ranges between ∼25-30 days in the TTL and has a surface value of ∼15 days. This is generally consistent with previous model calculations (e.g. Warwick et al., 2006;Sinnhuber and Folkins, 2006). For CH 2 Br 2 , in the mid-troposphere loss is dominated by reaction with OH with photolysis being slow. At the CPT, the two loss channels are roughly equal. The local lifetime of CH 2 Br 2 ranges from ∼50 days at the surface to a maximum of ∼520 days in the TTL. This is somewhat large given the working definition of a VSLS as a species whose lifetime is less than 6 months (Law and Sturges et al., 2007).  Figure 5 shows the tropical zonal mean profiles for the source gases CHBr 3 and CH 2 Br 2 from the base run B and sensitivity runs S NOCONV (no convection), S 2OH (2×[OH]) and S SLIMCAT . The location of the CPT and the approximate base of the TTL is shown for reference. Note that we define the base of the TTL as the level of maximum convective outflow (approximately 12 km) and the top of the TTL as the cold-point (e.g. Law and Sturges et al., 2007). The results here show that with the full TOMCAT model transport (run B) the mean CHBr 3 mixing ratio at the CPT (∼17 km) is ∼0.126 pptv resulting in an SGI contribution of ∼0.38 pptv of Br y to the lower stratosphere. This is in general agreement with the SGI value of 0.5 pptv predicted by Sinnhuber and Folkins (2006  reduce to 0.32 pptv and 1.57 pptv, i.e. ∼84% and ∼98% of run B values for CHBr 3 and CH 2 Br 2 , respectively. These results suggest that SGI via both species is not overly sensitive to model parameterised convection, particularly at the CPT. For CHBr 3 , this apparently contradicts the findings of Nielsen and Douglass (2001) who report a treatment of convection is required in their model simulations in order for CHBr 3 to reach the tropical lower stratosphere. Similarly, Warwick et al. (2006) report from a 3-D model study CHBr 3 to be highly dependent on convection in the tropical upper troposphere. Without further details on the experiments performed in these other studies we cannot comment further on the differences. In our experiments, although we switch off convection we still include the parameterisation of mixing in the PBL which causes mixing of surface-emitted tracers in the bottom few km. If we also switch off PBL mixing then we see CHBr 3 especially largely confined to the lowest model level (∼100 m) and a large decrease in upper troposphere (UT) values. It may be that the studies of Nielsen and Douglass (2001) and Warwick et al. (2006) also included transport due to this process in their definition of convection. Given that turbulent mixing in the PBL can transport tracers to the lower free troposphere, then the lifetimes of CHBr 3 and CH 2 Br 2 would indicate that resolved vertical advection by the analysed winds could still cause some transport to the UT. There is uncertainty in the modelled OH profile in the UT and so a sensitivity run was performed to investigate the impact of a large (×2) change in [OH] on the modelled SG profiles. From the run S 2OH profile in Fig. 5 it is clear the abundance of CH 2 Br 2 and associated SGI is more sensitive to changes in OH concentration than that of CHBr 3 . This is expected given the dominance of the CH 2 Br 2 +OH reaction over photolysis, relative to that of CHBr 3 +OH. For the SLIMCAT run S SLIMCAT , SG profiles have been scaled to approximately mimic that of the base run in the midtroposphere. It can be seen that SLIMCAT predicts a lower abundance of both SGs in the TTL and near-tropopause regions.

Source gas injection
TOMCAT model runs with ECMWF winds have been reported previously to exhibit too rapid vertical motion in the lower stratosphere region (e.g. Chipperfield, 2006;Monge-Sanz et al., 2007). The problem is more obvious with ERA-40 winds than that with the more recent ECMWF data sets and is mainly related to the noisy analysed wind fields. This is known to affect all CTMs using wind velocities or divergence to obtain the vertical motion (e.g. Scheele et al., 2005;Wohltman and Rex, 2008). SLIMCAT runs are not affected by the same problem as, in this case, above 350 K potential temperature vertical motion is computed from diagnosed heating rates. For this reason, the spurious vertical transport present in TOMCAT runs due to analysis noise is eliminated from SLIMCAT runs. In addition, SLIMCAT uses isentropic levels in the stratosphere which helps to separate vertical and horizontal motion and has also proven to provide more realistic transport in the lower stratosphere (LS) than TOMCAT (e.g. Chipperfield, 2006). Krueger et al. (2008) performed a Lagrangian model study in the TTL region using ECMWF operational winds and found that the use of the ECMWF vertical wind field resulted in significantly faster motion than the use of computed heating rates. A study by Wolthman and Rex (2008) with ECMWF winds has also shown improvements in the vertical velocities when obtained from diagnosed heating rates with respect to the vertical velocity field from the analyses (although we do not use the vertical velocity from the analyses).
The trajectory calculation inside the TOMCAT/SLIMCAT CTM has been used to estimate the mean tracer vertical transport in the TTL in runs B and S SLIMCAT . Trajectories were initialised at 80 hPa (run B) and 380 K (run S SLIMCAT ) and advected backwards in time using the vertical winds from the analysed divergence field and diagnosed heating rates respectively. The mean vertical motion, calculated from the trajectory displacement, was 0.64 mm/s in run B and 0.324 mm/s in run S SLIMCAT . Analysis of CO 2 data (Park et al., 2009) shows a range in vertical velocity, between the lower boundary of the TTL and the tropopause, of 0.5-0.14 mm/s. Krueger et al. (2008) find residence times in the winter TTL (2001/2002) of 36 days using operational ECMWF with diagnosed heating rates, and only 20 days when using the corresponding ECMWF vertical winds. Our calculations show a residence time in the 360-380 K region of 20 days based on the TOMCAT run and 52 days based on the SLIM-CAT run for the period November-December 2005. The residence times in the TTL calculated here are within the range of 20-80 days as published by WMO (2007). Our results show slower vertical transport through the TTL when using the SLIMCAT θ-coordinate model with vertical transport calculated from heating rates. Given this version of the model provides overall better agreement with observations in the TTL (see below), we conclude the vertical transport from the θ-coordinate model is more relastic than that of the TOMCAT p-level model. More recently, analysis of hydrochlorofluorocarbon (HCFC) and hydrofluorocarbon (HFCs) data from the WB57 aircraft during the NASA TC4 campaign indicates a transit time from 360 K to 380 K of about 3-4 months (Elliot Atlas, unpublished data). This estimate is somewhat larger than the 80 day upper limit quoted in WMO (2007) and may, in part, be due to the calculation being performed for summer months.
Overall, given the predicted mixing ratios of CHBr 3 and CH 2 Br 2 in the near-tropopause region, it seems the latter species may be deemed more significant. Wamsley et al. (1994) reported CH 2 Br 2 to have an atmospheric lifetime long enough to reach the stratosphere and CHBr 3 (with shorter lifetime) to contribute negligible amounts to stratospheric Br. The results here are also consistent with Laube et al. (2008) who, based on observations, deduced CH 2 Br 2 to be the "dominant" very short-lived SG. Their results, along with Schauffler et al. (1998), find CH 2 Br 2 to be present up to ∼18.5 km (∼0.15 pptv). The results here confirm CH 2 Br 2 at this level with run B predicting ∼0.5 pptv and run S SLIMCAT ∼0.16 pptv. The latter seems more reasonable given the observed values discussed above, along with the overestimation of SG in the near-tropopause region by TOMCAT (see below).
The simulated model profiles of CHBr 3 and CH 2 Br 2 have also been compared with a variety of in-situ aircraft measurements. For the campaigns discussed herein, whole air samples were collected using evacuated canisters onboard both the DC8 and WB57 aircraft prior to analysis in the laboratory. Post flight analysis was carried out using a 5column 5-detector gas chromatography (GC) system along with two flame ionization detectors (FID) and two electron capture detectors coupled to a quadrupole mass spectrometer (MS). A discussion of the analytical techniques employed along with the accuracy and precision on measurements can be found in Colman et al. (2001). For the PEM TROPICS-B campaign, precision of CHBr 3 and CH 2 Br 2 measurements are reported at 1.6%. The accuracy is estimated to be between 1-10% at 1σ . Figure 6 shows the modelled profiles of CHBr 3 and CH 2 Br 2 versus tropospheric observations from the PEM TROPICS-B (e.g. Colman et al., 2001; http://www-gte.larc. nasa.gov/pem/pemtb obj.htm) and INTEX-B (http://www. espo.nasa.gov/intex-b/) campaigns. The model profiles shown here are averaged over the same spatial domain and for the same months (but for 2007) as the observations. This shows that the model profiles from run B, constrained with a surface mixing ratio of 1.2 pptv for both species, fit the observations in the mid troposphere well. Particularly, for CH 2 Br 2 , the model is able to reproduce observed mixing ratios and profile shape. Note that the difference between runs B and S NOCONV here is small, showing that modelled convection is only playing a small role in this region. Figure 7 shows the modelled profiles of CHBr 3 and CH 2 Br 2 against tropical observations which extend into the TTL. These are the 2007 NASA TC-4 (http://www.espo. nasa.gov/tc4/), the NASA PRE-AVE (http://espoarchive. nasa.gov/archive/arcs/pre ave/) campaign and the 2006 NASA CR-AVE (http://www.espo.nasa.gov/ave-costarica2/) campaigns. For TC-4 DC8 flights targeted recent convective outflow while only a single WB57 flight targeted convection. Again the model profiles are averaged over the same spatial domain and for the same months as the observations. For CHBr 3 , the base model B performs reasonably well against the observations in the lower troposphere and near the tropopause. The model is able to reproduce the gradient seen from the surface to ∼500 hPa and the modelled profile lies within the min-max variability of observations at most levels. Of importance is the model's ability to simulate CHBr 3 in the near-tropopause region. A number of previous model studies have significantly overestimated CHBr 3 in this region when compared with observations (e.g. Warwick et al., 2006;Nielsen and Douglass, 2001). It can be seen that TOMCAT performs well in this regard. However, it seems that TOM-CAT may not capture convection effectively shown by the lack of a signature "C-shape" in the profile. Results from TC4, which targeted active convective outflow during most flights, may not be representative of the region as a whole. Similarly for CH 2 Br 2 , the modelled profile seems reasonable against observations in the lower troposphere. There is less variability seen here than for CHBr 3 , due to the longer lifetime of CH 2 Br 2 . However, in the near-tropopause region, TOMCAT overestimates the abundance of source gas. This will be in part due to the too fast modelled vertical transport through the TTL region in run B (see above). Note that the overestimation of the SGs in the TTL in TOMCAT is not due to the model vertical resolution. The SG profiles are nearly identical in run B and S L31 which has the higher vertical resolution (not shown). The vertical transport is controlled by the vertical winds and not by numerical diffusion.
From Fig. 7 it is apparent that run S SLIMCAT , with slower vertical transport in the TTL, reproduces observed CH 2 Br 2 in the near-tropopause region fairly well and better than the TOMCAT runs. In contrast, for PRE-AVE flights, the run B profile fits CHBr 3 observations well in the important upper TTL and tropopause regions. A larger difference is seen here between runs with and without convection along with more of a "C-shape" in the modelled profile. This is due to the model being sampled in the months January and Febru- Fig. 8. Correlation plot of observed CH 2 Br 2 versus observed CHBr 3 between 350 and 80 hPa from the TC-4 campaign (Fig. 7a). Also shown are model results from runs B, S 2OH , and S SLIMCAT in the same region. Power lines of best fit are included on all datasets of the form, Y =(aX) b . For an air parcel with an initial composition of 1 pptv CHBr 3 and 1 pptv CH 2 Br 2 the change in source gas values following (i) passive (no chemical loss) mixing with background air containing zero SGs (1:1 line) and (ii) chemical loss without mixing (assuming mean chemical lifetimes in ratio 30:300 days (CHBr 3 :CH 2 Br 2 )) are also shown (dashed grey lines). ary when convection is stronger. The θ-level model in this case seems to underestimate observed CHBr 3 in the TTL, although there are few observations at this level. For CH 2 Br 2 , both runs B and S SLIMCAT seem reasonable in the neartropopause region, with the former perhaps fitting the observations better. Unlike comparisons with TC4 and CR-AVE data, run B does not show a significant overestimation of CH 2 Br 2 . This could be explained by variation in the strength of deep convective uplift between campaigns. Finally, for CR-AVE data, modelled CHBr 3 is overestimated in runs B and S NOCONV in the TTL and near-tropopause regions. In these regions run S SLIMCAT performs well. This is the also the case for CH 2 Br 2 , where the TOMCAT overestimation is greater. Figure 8 shows a tracer-tracer plot of CHBr 3 vs CH 2 Br 2 in the 350-80 hPa region from runs B, S NOCONV , S 2OH and S SLIMCAT . Mean observations from the TC-4 data set (also in this region, see Fig. 7a) are included on this figure along with a power line of best fit of all data sets. The origins of the model lines (high SG mixing ratios) are arbitrary and the plot tests the ability of the the different model runs to fit both SG profiles simultaneously. The model lines here indicate that with the current model setup (chemistry and transport), both tracers cannot be simulated correctly at the same time. The S SLIMCAT run performs the best as can be seen in the gradient relative to that of the observations. This gradient of this line lies between that expected from the two extreme cases of chemical loss in an isolated air parcel ascending in the TTL and passive mixing (i.e. without chemical loss) with air, e.g. in the extra-tropical lowermost stratosphere, containing zero CHBr 3 and CH 2 Br 2 . This indicates that in the SLIMCAT run S SLIMCAT mixing between the TTL and extra-tropical lowermost stratosphere has a larger effect on the TTL composition (see also discussion of Br y in Sect. 4.3). In the TOMCAT runs the gradient of the lines is much closer to that expected from the relative chemical loss. In these model runs the rapid vertical transport dominates over horizontal mixing. Figure 9 shows the annual tropical zonal mean abundance of the product gases arising from CHBr 3 and CH 2 Br 2 degradation. For bromoform the major degradation products are CBr 2 O and CHBrO with CBr 3 OOH and CHBr 2 OOH being minor products. This is consistent with the suggestions of Ko and Poulet et al. (2003). The mixing ratios of these species in the TTL are very low (<0.03 pptv) for the assumed SG surface vmrs. From this we infer that the contribution of these species to PGI and thus total Br from CHBr 3 is negligible. The mixing ratios of the peroxy radicals in the scheme, CBr 3 O 2 and CHBr 2 O 2 , were found to be near zero throughout the profile (not shown).

Product gas injection
Similarly, for dibromomethane the model predicts the major degradation product to be CHBrO and a minor product to be CHBr 2 OOH. Again, this is consistent with the suggestions of Ko and Poulet et al. (2003) and the mixing ratios of these species are also near-zero throughout the profile. As for CHBr 3 , it is apparent that the contribution of organic products arising from CH 2 Br 2 SG degradation to PGI and thus total bromine, is negligible. Furthermore, although poorly quantified, organic PGs would be expected to be removed from the atmosphere by washout processes due to their solubility. The model work described in this paper has not accounted for this and thus the extremely low near-tropopause mixing ratios reported here for CBr 2 O, CHBrO and other PGs could indeed be overestimates. The results here suggest that PGI is dominated by the transport of inorganic products. Figure 10 shows the calculated loss rates due to photolysis and the resultant lifetime of CBr 2 O and CHBrO in the tropics. We find CBr 2 O to have a lifetime of ∼7 days and CHBrO of ∼1 day due to photolysis. Overall, these results show that the assumption made in previous model studies of instantaneous conversion between organic bromine product gases and Br y following SG degradation seems reasonable.   Table 5 and are quoted at the location of the CPT (i.e. the approximate contribution to the lower stratosphere −θ>380 K). Note, values within the TTL are somewhat larger where quasihorizontal mixing into the extra-tropical lowermost stratosphere is possible. All results here are calculated as an annual zonal mean in the tropics. Assuming a Br y lifetime of 10 days, our TOMCAT model predicts that CHBr 3 contributes ∼0.72 pptv additional bromine to the lower stratosphere. We find the fraction of this value delivered via SGI and PGI to be approximately equal. This is in general agreement with the work of Dvortsov et al. (1999) and Nielsen and Douglass (2001) who report similar values of ∼1 pptv. These studies also find the delivery via SGI and inorganic PGI to be approximately equal. Similarly, the results here are in good agreement with the work of Sinnhuber and Folkins (2006) who report total bromine from CHBr 3 reaching the cold point to be 0.8 pptv for the same assumed 10-day Br y lifetime. In addition, they report a PGI value of 0.3 pptv which is consistent with the 0.35 pptv reported here. From CH 2 Br 2 and with the same 10 day Br y lifetime, we find a delivery of ∼1.69 pptv of bromine to the lower stratosphere, with ∼94% from SGI. In this case, the contribution from PGI is small. Despite only two Br atoms per molecule (as opposed to 3 for CHBr 3 ), the dominance of the SGI pathway is due to the longer local lifetime of CH 2 Br 2 , allowing more SG to reach the upper troposphere. The results reported here for CH 2 Br 2 may constitute an upper limit given the observed overestimation of SG in the near-tropopause region in TOMCAT. Furthermore, we infer a total Br contribution from both SGs to be ∼2.4 pptv to the lower stratosphere. Increasing the assumed Br y lifetime below the tropopause to 20 and 40 days raises this value to ∼2.9 and ∼3.6 pptv. The mean lifetime of Br y in the troposphere and in particular the TTL region is, however, uncertain at present.

Total bromine
Results from the SLIMCAT sensitivity runs indicate lower values of total Br from both CHBr 3 and CH 2 Br 2 reaching the stratosphere compared with TOMCAT (see Table 6). The results here are derived assuming a fixed 0.5 pptv CHBr 3 and 1 pptv CH 2 Br 2 in the lower 8 levels of the model (surface to ∼10 km). These values are chosen as they mimic the mid-tropospheric values of TOMCAT profiles and also provide the best fit with aircraft observations (e.g. Fig. 7). From CHBr 3 , assuming a 10 day Br y lifetime, we find total Br reaching the stratosphere to be ∼0.2 pptv. Similarly for CH 2 Br 2 , we find a total Br contribution of ∼1.0 pptv. The lower values here result from the slower vertical motion in the SLIMCAT simulations (relative to TOMCAT) and hence a smaller SGI contribution. This slower transport would also cause a decrease in the contribution from PGI due to the increased time soluble Br y is available for washout (below the cold point tropopause). However, given the SLIMCAT simulations do not model all aspects of tropospheric tracer transport, these runs can only be seen as sensitivity tests. In addition, these SLIMCAT simulations have been performed over a period of 6 years which is not long enough for the model to spin up and reach equilibrium in the stratosphere. This can be seen in Table 6 for the run S SLIMCAT (infinite Br y lifetime) where total Br values have not equilibrated at 1.5 pptv and 2.0 pptv for CHBr 3 and CH 2 Br 2 respectively. This is likely due to the quasi-horizontal mixing of extra-tropical stratospheric air with low Br y content into the TTL. Note, this is not the case for the equivalent TOMCAT simulations where the model has been spun up sufficiently (see Table 5), so that total Br at the CPT is approximately the expected 3.6 pptv and 2.4 pptv for CHBr 3 and CH 2 Br 2 , respectively.

Conclusions
We have performed a 3-D model study using the TOM-CAT/SLIMCAT CTM in order to quantify, first, the contribution of CHBr 3 and CH 2 Br 2 to the stratospheric bromine budget and, second, the relative magnitude of SGI and PGI. A detailed chemical scheme describing the tropospheric degradation of both source gases, along with simplified product gas chemistry, has been developed. We have thus provided novel estimates of the organic products gases arising from CHBr 3 and CH 2 Br 2 degradation. The major degradation products have been found to be CBr 2 O and CHBrO whose local lifetimes are calculated at ∼7 and ∼2 days respectively. We find their contribution to total bromine negligible and thus infer that assumption of instantaneous production of Br y following CHBr 3 /CH 2 Br 2 degradation in model studies is reasonable. It is likely that this assumption will also be valid for other short-lived source gases (e.g. CHBr 2 Cl, CHBrCl 2 ) whose degradation products are comparably short-lived. However, attempt at measurements of species such as CBr 2 O and CHBrO in the troposphere and TTL would certainly be useful.
The TOMCAT/SLIMCAT CTM has been shown to perform reasonably against observations of these SGs in the tropical troposphere-lower stratosphere region. The σ -θ level model (SLIMCAT) tends to agree better than the σ -p level model (TOMCAT) due to the slower vertical transport in the TTL. The SLIMCAT run also gives the best simultaneous relative comparisons of the profiles of CHBr 3 and CH 2 Br 2 . We find a treatment of convection is not required in our TOMCAT simulations to transport significant quantities of SG to the TTL and lower stratosphere. In future work the convective transport parameterisation will be included in SLIMCAT to investigate the impact of this against the background of slower resolved advection.
The results presented here have shown CHBr 3 and CH 2 Br 2 together could contribute around 2.4 pptv of Br to the lower stratosphere when a Br y lifetime of 10 days is assumed along with mean surface mixing ratios of 1.2 pptv for both source gases. Assuming the Br VSLS y value of 5 pptv inferred by Dorf et al. (2008), then a shortfall of ∼2.6 pptv remains. This may, in part, be supplied from Br-containing SGs such as CH 2 BrCl, CHBr 2 Cl and CHBrCl 2 (local lifetimes of 150, 69 and 78 days respectively, Law and Sturges et al., 2007). However, these species are unlikely to explain a shortfall of 2.6 pptv given their relatively low abundance in the tropical near tropopause region (e.g. Kerkweg et al., 2008). Based on a compilation of field data, Law and Sturges et al. (2007) report mixing ratios of CH 2 BrCl, CHBr 2 Cl and CHBrCl 2 in the tropical upper troposphere of 0.32 (0.26-0.35) pptv, 0.08 (0.03-0.12) pptv and 0.12 (0.05-0.15) pptv respectively. Furthermore, although not quantified, it is possible that "additional bromine source gases" which remain unknown may contribute (Laube et al., 2008). One uncertainty within the current model work arises on assumption of a uniform prescribed lifetime of Br y in the troposphere. In future simulations we shall couple a detailed Br y scheme, that explicitly considers the partitioning and removal of soluble products, with the degradation schemes outlined in this paper (e.g. Breider et al., 2009).
Sensitivity simulations using the SLIMCAT σ -θ model have shown a smaller overall contribution from CHBr 3 and CH 2 Br 2 to stratospheric Br. From these runs we infer a value of ∼1.2 pptv (0.5× the TOMCAT estimate) when assuming a 10 day lifetime of Br y . Naturally this increases the discrepancy between the model and the value of 5 pptv discussed above. Our results are generally in agreement with previous model work in that the contribution of brominecontaining VSLS may supply a significant amount of Br y to the lower stratosphere. Therefore, in future stratospheric simulations it will be important to take this into account.