Convective transport of very short lived bromocarbons to the stratosphere

. We use the NASA Goddard Earth Observing System (GEOS) Chemistry Climate Model (GEOSCCM) to quantify the contribution of the two most important brominated very short lived substances (VSLSs), bromoform (CHBr 3 ) and dibromomethane (CH 2 Br 2 ), to stratospheric bromine and its sensitivity to convection strength. Model simulations suggest that the most active transport of VSLSs from the marine boundary layer through the tropopause occurs over the tropical Indian Ocean, the tropical western Paciﬁc, and off the Paciﬁc coast of Mexico. Together, convective lofting of CHBr 3 and CH 2 Br 2 and their degradation products supplies ∼ 8 ppt total bromine to the base of the tropical

The most important VSLS pathway to the stratosphere is via convective lofting through the tropical tropopause layer Published by Copernicus Publications on behalf of the European Geosciences Union.
(TTL). Atmospheric chemistry and transport of brominated VSLSs involves the coupling of various complex processes, e.g., highly un-uniform ocean emissions, convective transport of source gases and product gases, and wet scavenging of soluble product gases. An accurate modeling representation of these processes remains a challenging task. Modeling of these processes in general requires assumptions and/or parameterizations that have to some extent been simplified and may differ greatly from one model study to another. These differences have lead to significant differences in the modeled Br VSLS y estimates. For example, a recent model study by Hossaini et al. (2013) found that modeled Br VSLS y could vary by a factor of 2 when using four recently published emission inventories Pyle et al., 2011;Ordóñez et al., 2012;Ziska et al., 2013). In general, model studies agree relatively well on source gas injection (SGI), suggesting that about ∼50 % of bromoform (CHBr 3 ) and ∼90 % of dibromomethane (CH 2 Br 2 ) can reach the stratosphere through SGI (e.g., Dvortsov et al., 1999;Nielsen and Douglass, 2001;Aschmann et al., 2011;Hossaini et al., 2010Hossaini et al., , 2012a. However, modeled contribution on product gas injection (PGI) is highly uncertain depending on how wet scavenging is implemented. Many early model studies assume a uniform washout lifetime against wet scavenging (e.g., Dvortsov et al., 1999;Nielsen and Douglass, 2001;Hossaini et al., 2010;Aschmann et al., 2009), which was inadequate and led to an underestimate in modeled Br VSLS y (Hossaini et al., 2012a;Aschmann et al., 2011). A recent study by Aschmann and Sinnhuber (2013) shows that treating Br VSLS y as a single soluble tracer (e.g., Liang et al., 2010) can also lead to an underestimate. They found that Br VSLS y increased from 3.4 ppt to 5 ppt when switching from an idealized setup with a single soluble inorganic bromine tracer to a full chemistry scheme. Thus, an accurate modeling of transport and wet scavenging of PGI seems to be the key in narrowing the uncertainty of model estimate of Br VSLS y .
Understanding how the contribution of VSLSs to stratospheric bromine varies with convection strength has significant climate implications. A recent modeling analysis by Hossaini et al. (2012b) suggests that VSLS SGI will increase from ∼1.7 ppt in 2000 to ∼2.0-2.7 ppt in 2100 using the Intergovernmental Panel on Climate Change (IPCC) representative concentration pathways (RCPs) scenarios, as the future simulations feature stronger tropical deep convection transport to the lower stratosphere. On the other hand, the overall response of Br VSLS y to convection strength is somewhat murky. Earlier studies deploying a uniform washout lifetime found different washout rates result in a significant range in the contribution of CHBr 3 to Br y , 0.5-3 ppt in Sinnhuber and Folkins (2006) and 1.6-3 ppt in Aschmann et al., (2011). Liang et al. (2010), which deploys explicit wet scavenging in convective updrafts, found that convective scavenging only accounts for ∼0.2 ppt difference in modeled Br VSLS y . Similar results were reported in Aschmann et al. (2011), who showed that, while SGI is highly correlated with convective activity, the impact on total stratospheric bromine in a full chemistry scheme is nearly insensitive to dehydration, likely due to convection dilution and increased scavenging.
In this study, we use a three-dimensional chemistry climate model with fully interactive CHBr 3 and CH 2 Br 2 , the two most important brominated VSLSs, to better understand how VSLSs and their degradation products enter the stratosphere. We will also test the uncertainty in modeled Br VSLS y due to varying strength in convection and scavenging.

Model description
We conduct model simulations using the NASA Goddard Earth Observing System (GEOS) Chemistry Climate Model (GEOSCCM) Version 2, which couples the GEOS-5 General Circulation Model (GCM) (Reinecker et al., 2008) with a stratospheric chemistry module. Model simulations have a horizontal resolution of 2.5 • × 2 • (longitude by latitude) with 72 vertical layers from surface to 0.01 hPa. The GEOS-5 model uses a flux-form semi-Lagrangian dynamical core (Lin, 2004) and the Relaxed Arakawa Schubert (RAS) parameterization for convection (Moorthi and Suarez, 1992). The GEOS-5 moist processes are represented using a convective parameterization and prognostic cloud scheme.
The GEOSCCM V2 stratospheric chemistry module includes all important gas phase stratospheric reactions as described in Douglass and Kawa (1999), and chemical calculations are carried out above the 27th eta layer (approximately 350 hPa on average, globally). All chemical kinetics and photolysis rates are calculated following Jet Propulsion Laboratory (JPL) 2010 (Sander et al., 2011). We modify the standard V2 chemistry scheme to include CHBr 3 and CH 2 Br 2 ; both are interactive with the full stratospheric chemistry scheme. The two organic source gases are released at the ocean surface following the geographically resolved emission distribution described in Liang et al. (2010) and are destroyed in the atmosphere via photolysis and reaction with the hydroxyl radical (OH). OH above ∼350 hPa is calculated online in the stratospheric chemistry module. OH in the lowest 26 layers (troposphere) is relaxed to monthly mean climatological fields documented in Spivakovsky et al. (2000). Global annual mean OH from Spivakovsky et al. (2000) is 1.16 × 10 6 molecules cm −3 , yielding an atmospheric methyl chloroform (CH 3 CCl 3 ) lifetime of ∼5.5 years. In most of the troposphere, only simple bromine chemistry is considered; 80 % of inorganic bromine products are classified as HBr and the remaining 20 % as HOBr -ratios adopted from Yang et al. (2005). Above the 27th eta layer (∼350 hPa), bromine from CHBr 3 and CH 2 Br 2 degradation is released as Br and interacts fully with stratospheric chemistry. Compare this to the idealized case in Liang et al. (2010) (Br, BrO, and BrCl) in the fully interactive stratospheric chemistry scheme in the TTL and stratosphere. Although BrONO 2 is not produced in the troposphere in the current simple chemistry scheme, BrONO 2 produced in the stratosphere is allowed to transport to lower altitudes via large-scale descent. A previous detailed tropospheric chemistry model study using the Harvard GEOS-Chem model and VSLS emissions from Liang et al. (2010) shows a small fraction of Br y exists as BrO (0.1-0.2 ppt) below 10 km in the tropics (Parrella et al., 2012). However, the impact of the absence of this tropospheric BrO on stratospheric bromine is likely small. No heterogeneous chemistry for VSLS is included in our simulations, but results from Aschmann and Sinnhuber (2013) suggest that the impact of heterogeneous chemistry is minor and that the inclusion of heterogeneous activation prevents loss in scavenging and can increase Br VSLS y by 10 %. While heterogeneous chemistry can shift Br y partition and increase HBr up to four times between 12 and 18 km in the tropics (Aschmann et al., 2011), the absence of heterogeneous chemistry in GEOSCCM is likely to have only a small impact on Br VSLS y since the model HBr is present in very low quantities during daytime when the majority of convective scavenging takes place (see Sect. 3.1 and Fig. 4) and the overall convective scavenging in the TTL is rather inefficient (see Sect. 3.2).
The wet scavenging of the soluble inorganic bromine is the same as that detailed in Liang et al. (2010). It includes scavenging in rainout (in-cloud precipitation) and washout (below-cloud precipitation) in both large-scale precipitation (Giorgi and Chameides, 1986) and deep convective updrafts (Balkanski et al., 1993). We assume high solubility for all three inorganic bromine reservoir species, HBr, HOBr, and BrONO 2 . At each time step, we follow the Giorgi and Chameides (1986) parameterization to compute the wet scavenging using the GEOS-5-model-calculated large-scale and convective precipitation rates and the parameterized fraction of grid square area, F , that actually experiences precipitation (Balkanski et al., 1993). Wet scavenging occurs only when air temperature is above 258K, following Giorgi and Chameides (1986). Previous studies using the Giorgi and Chameides (1986) parameterization showed that, for largescale precipitation, the global mean F is about 2.5 % for all grid boxes and the median value is 10 %, with values exceeding 40 % in ∼10 % of the grid boxes (Balkanski et al., 1993;Liu et al., 2001). The global mean F for convective precipitation is much smaller, only 0.4 % (Liu et al., 2001). When evaporation occurs during large-scale and convective transport, the corresponding fraction of the dissolved inorganic bromine is released back to the atmosphere. This wet scavenging scheme has been applied in many atmospheric modeling studies of soluble trace gases and aerosols, e.g., 210 Pb, 7 Be, sulfate, sea salt, and dust, and the simulated concentra-tions compare well with surface observations at many observation sites around the globe (e.g., Balkanski et al., 1993;Chin et al., 2000;Ginoux et al., 2001;Liu et al., 2001).
In this study, we present results from two 51-year simulations between 1960 and 2010, one with brominated VSLSs chemistry (R VSLS ) and one without (R BASE ), to examine troposphere to stratosphere transport (TST) of VSLS and their contribution to stratospheric bromine. It takes a considerably long time (∼15 years) for the full impact of VSLSs to reach the upper stratosphere; thus, the first 16 years are considered as spin-up. A full model evaluation and discussion of the impact of VSLSs on stratospheric ozone are presented in a separate paper, currently in preparation. For this study, we focus on model results of the two simulations from the last 10 years.

Convection sensitivity simulations
To quantify the uncertainty in Br VSLS y due to changes in convection strength, we conduct two sensitivity simulations from 1980 to 2010 to represent minimum convection (R MINCNV ) and maximum convection (R MAXCNV ) conditions by varying five convective parameters in RAS that impact the strength of deep convection, clouds, convective condensate, and re-evaporation (Ott et al., 2009(Ott et al., , 2011. These five convective parameters are RASAL1 and RASAL2 (regulates the strength and vertical profile of the relaxation time scale for deep convection), ACRITFAC (used to compute the critical value of the cloud work function which determines the initiation of convection), BASE_EVAP_FAC (regulates the amount of rain evaporated into the environment below the cloud base), and AUTOC_CN (used to calculate the autoconversion of convective condensate). These five are identified as the most strongly influencing parameters out of the total of 16 parameters examined using a large number of "Monte Carlo" type simulations and ensemble simulations in both a single-column model as well as the GEOS-5 GCM (Ott et al., 2011). In the two sensitivity simulations, the five parameters are varied to produce the strongest (MAXCNV) and the weakest (MINCNV) representations of convection considered reasonable. In general, compared to the minimum convection condition, the maximum convection condition yields significantly large increases in shallow convection below 5 km and ∼20-30 % stronger vertical mass flux and more horizontal divergence between 12.5 and 16.5 km (Ott et al., 2011). The values used for the five convective parameters for the minimum and maximum conditions have been tested extensively in the previous ensemble analysis. They produce reasonable precipitation patterns when compared with data from the Global Precipitation Climatology Project (GPCP) that were compiled from satellite and rain gauge observations (Ott et al., 2011). The correlation coefficients of the standard run, MAXCNV and MINCNV simulations, and the GPCP data sets are 0.65, 0.64, and 0.62, respectively (Ott et al., 2011).  The black dash-dotted line shows simulated BrO from the run without VSLS (R BASE ). Simulated BrO from the two convective sensitivity simulations is also shown (dotted line for maximum convection condition and dashed line for minimum convection condition). The model is sampled at the same month and latitude as the balloon measurements but only daytime profiles from all available longitude grid points are collected to calculate the one sigma variance of simulated BrO.
We aim to use these minimum and maximum conditions to bound the range of uncertainty that can be introduced due to variations in the strength of convection and wet scavenging.

The contribution of VSLSs to stratospheric bromine
The simulated CHBr 3 and CH 2 Br 2 have been evaluated extensively in Liang et al. (2010) and compare well with aircraft and surface observations in their atmospheric distri-bution, geographically and vertically, as well as the associated seasonality. Here we present a summary comparison of observed and modeled CHBr 3 in the troposphere for 30-60 • S, 30 • S-30 • N, 30-60 • N, and 60-90 • N latitude bands (Fig. 1). The observed profiles are compiled using whole air sampler canister measurements Blake et al., 2003) from eight NASA aircraft missions, as detailed in Liang et al. (2010), and averaged within each latitude band at 1 km vertical intervals. The comparison clearly shows that the model captures well the source gas concentrations and vertical gradients in the tropics and mid and high latitudes in both hemispheres. Figure 2 shows a comparison of our simulated BrO with balloon measurements from the LPMA (Limb Profile Monitor of the Atmosphere) / DOAS (differential optical absorption spectroscopy) at Teresina (5.1 • S), Aire sur l'Adour (43.7 • N), and Kiruna (67.9 • N) collected between 2003 and 2005 (Dorf et al., 2006a(Dorf et al., , b, 2008Rozanov et al., 2011). The lower levels of BrO from R BASE in comparison to the measurements indicate the essential role of VSLSs in completing the stratospheric bromine budget. With VSLSs, the GEOSCCM model simulates the observed BrO well at all sampled locations in the tropics and mid and high latitudes. Currently the GEOSCCM does not include the remaining three of the five major brominated VSLSs, CH 2 BrCl, CHBr 2 Cl, and CHBrCl 2 (lifetimes between 70 and 150 days). Assuming a total tropospheric abundance of 1 ppt of these three VSLSs and assuming that ∼90 % (value for CH 2 Br 2 ) of the source gases survive the TST to the stratosphere, this adds ∼0.9 ppt to the current model estimate of stratospheric Br y of ∼24 ppt. With the current model BrO/Br y ratio, such an increase will lead to an increase in BrO of up to 0.6 ppt (not shown) in the stratosphere. This results in a better agreement with the DOAS BrO measurements in Teresina and Aire sur l'Adour and reasonable agreement at Kiruna, considering the large spatial variability in BrO in the high latitudes and rather localized balloon measurements. Since the model runs freely, with only surface source gas emissions being specified, the fact that the model simulates the observed concentrations for both the source gases and BrO well suggests that the model presents a credible representation of stratospheric bromine chemistry.
Comparing results from the R VSLS and R BASE runs (Fig. 3), we find that the inclusion of CHBr 3 and CH 2 Br 2 adds a uniform ∼7.7 ppt Br y throughout most of the stratosphere. The current Br VSLS y estimate is 55 % higher than our previous estimate of ∼5 ppt  and ∼6 ppt estimate in Hossaini et al. (2013), though the latter two are driven by the same VSLS emissions. Compared to the idealized case in Liang et al. (2010) that tracks Br VSLS y in a single highly soluble tracer, Br VSLS y in this study is present in the fully interactive stratospheric chemistry scheme in both soluble forms (HBr, HOBr, and BrONO 2 ) and insoluble forms (Br, BrO, and BrCl) in the TTL. The ratio of insoluble and soluble Br y varies with the time of day as well as altitude and location. Figure 4 shows the contrast of model inorganic bromine speciation in the tropics between daytime and nighttime. In the TTL region, at daytime, when most convective lofting occurs, the majority of Br y exists as insoluble Br and BrO (Fig. 4a), therefore greatly increasing the amount of Br VSLS y that survives wet scavenging during the TST and ultimately reaches the stratosphere. The increase from ∼5 ppt to ∼7.7 ppt when switching from an idealized case to fully interactive stratospheric chemistry is consistent with results from Aschmann and Sinnhuber (2013), who found that Br VSLS y increased from 3.4 ppt to 5 ppt when switching from an idealized setup with a single soluble inorganic bromine tracer to a full chemistry scheme. The different estimates between Hossaini et al. (2013) and this work are likely due to how the ratio of soluble and insoluble inorganic bromine is determined in the two models. Compared to this work which partitions inorganic bromine in its various forms based on chemical reaction rates, Hossaini et al. (2013) used a mean altitude-dependent HBr:Br y ratio. This likely leads to an excessive washout of Br VSLS y , as, during daytime, when most of the convective lofting occurs, the majority of the inorganic bromine exists as insoluble Br and BrO. This suggests that a close approximation of soluble and insoluble inorganic product gases and the associated diurnal variation are critical to accurately quantify the contribution of VSLSs to stratospheric halogen in VSLS modeling.
A recent study by Kreycy et al. (2013), using balloonborne DOAS BrO measurements over Kiruna (67.9 • N, 22.1 • E), suggests that the ratio of J (BrONO 2 ) / k BrO+NO 2 should be larger than the JPL 2010 recommendations by a factor of 1.7 and is likely to shift more BrONO 2 into BrO. However, this result will only have a small impact on in the TTL during daytime. In addition, convective scavenging of the inorganic bromine in the TTL is not an efficient removal process (section 3.2). Increasing the ratio of J (BrONO 2 ) / k BrO+NO 2 in the model will shift Br VSLS y partition from BrONO 2 to BrO, which in turn will result in an increase in Br VSLS y due to less BrONO 2 scavenging. However, this increase will be small (at most a few tenths ppt).

Troposphere-to-stratosphere transport of VSLSs
We use the simulated monthly mean CHBr 3 distribution on the 355 K potential temperature surface (just below the TTL) to show the important tropical regions where active TST initiates (Fig. 5). Theoretically, trace gas distribution on the 365 K surface should be more indicative of TST as 365 K marks zero radiative heating, and air mass elevated above this level can enter the lower stratosphere through slow radiative ascent (e.g., Gettelman and Forster, 2002;Fueglistaler et al., 2009). However, it is difficult to identify active TST regions on the 365 K map (not shown) as a significant portion of CHBr 3 is converted to product gases. This is not surprising for a short-lived compound with a lifetime of ∼26 days (WMO 2011), while it takes about 10 days, on average, for air to travel ± 10 K (Fueglistaler et al., 2004;Levine et al., 2007). Compared to CHBr 3 , it is much more difficult to identify active TST regions on a CH 2 Br 2 map, as CH 2 Br 2 distribution appears more zonal with a smaller meridional gradient due to a longer lifetime of ∼120 days (WMO 2011) and up to ∼450 days locally in the TTL (Hossaini et al., 2010) and thus more mixing with the surrounding background air (not shown). The 355K CHBr 3 map shows three active regions that can efficiently deliver VSLSs to the base of the TTL: (1) the tropical Indian Ocean, (2) the tropical western Pacific, and (3) the area off the Pacific coast of Mexico (Fig. 5). There is significant seasonality associated with each entry re-gion. The Indian Ocean appears as the most active region for the TST of CHBr 3 and occurs all year long with a maximum in boreal winter (DJF). Lofting in the tropical western Pacific reaches its maximum in boreal summer (JJA), while the TST off the Mexico coast occurs mostly in boreal summer (JJA) and fall (SON). The importance of convective lofting in the western Pacific warm pool in TST has been noted in many previous studies (e.g., Hatsushika and Yamazaki, 2003;Fueglistaler et al., 2004;Aschmann et al., 2009;Hossaini et al., 2012a;Ashfold et al., 2012), with several others suggesting that the Indian Ocean is also an important region in the TST of VSLSs (Levine et al., 2007(Levine et al., , 2008Brioude et al., 2010;Hoyle et al., 2011).
We show in Fig. 6 the vertical profiles of organic source gases (CH 2 Br 2 × 2, CHBr 3 × 3) and inorganic product gases in the three critical convective lofting regions to illustrate the transport and wet scavenging of the brominated VSLSs during the TST and the relative importance of SGI vs. PGI. Although tropospheric Br y is assumed to be highly soluble, only a fraction of the grid boxes and a fractional area of the precipitating grid boxes actually experience precipitation and scavenging. On average, of all inorganic bromine (∼4 ppt) produced from CHBr 3 and CH 2 Br 2 in the tropical troposphere, about 50 % (∼2 ppt) is removed below 500 hPa, mainly by large-scale precipitation , with an additional few tenths ppt scavenged in the upper troposphere. Together, ∼1.5 ppt of HBr and HOBr survive largescale and convective scavenging and remain in the tropical upper troposphere. Br, BrO, and BrONO 2 produced from CHBr 3 and CH 2 Br 2 degradation in the tropical upper troposphere adds an additional ∼2 ppt to Br VSLS y . Over the tropical Indian Ocean, on average annually, the mixing ratio of total bromine from VSLS (organic + inorganic) at 150 hPa (∼355 K) is ∼8.5 ppt, the same as its surface abundance (∼8.5 ppt). This implies that convective lofting in this region is so efficient that the amount of VSL bromine that enters the base of the TTL is set by the concentration at the marine boundary layer. As air ascends to higher altitudes, a small portion of Br VSLS y (∼0.8 ppt, 10 %) is gradually removed from the atmosphere via wet scavenging until it reaches a constant 7.7 ppt above 10 hPa. Results from the tropical western Pacific and off the coast of Mexico are similar, with a smaller surface abundance of organic bromine (7.8∼8.1 ppt) and slightly less washout around 100 hPa. The global averaged total bromine from the VSLS (Br VSLS ) profile shows a maximum value of ∼8 ppt at ∼100 hPa. To supply this amount of Br VSLS to the lower stratosphere, transport from the marine boundary layer has to initiate from tropical regions where active convective lofting is colocated with high surface concentrations where the collective CHBr 3 and CH 2 Br 2 abundance exceeds 8 ppt Br.
Mapping streamlines onto the VSLS organic bromine distribution in the tropics clearly illustrates the importance of the colocation of deep convection with high surface concentration regions (Fig. 7). Among the three ascending branches Atmos. Chem. Phys., 14, 5781-5792 of the Walker circulation (Webster, 1983), ascent in the tropical western Pacific penetrates deepest into the TTL, but ascent in the Indian Ocean is capable of delivering more VSLS bromine to the base of the TTL due to higher surface concentrations. The surface abundance of VSLS bromine is largely dependent on the emission distribution used.  assumed uniform zonal emission strength across all longitudes, but the rate of bromocarbon emissions for the coastal regions per unit area is much higher than that in the open oceans. In the tropics, the prevailing surface easterly trade winds tend to bring recent emissions of CHBr 3 from the adjacent Indonesian coastal regions while the tropical western Pacific Ocean sees recent emissions of CHBr 3 from the open ocean. Therefore, the surface concentrations are higher in the Indian Ocean than the tropical western Pacific. However, it is important to point out that only a few surface observation constraints from the tropical western Pacific and none from the tropical Indian Ocean were available to derive the Liang et al., (2010) emission scenario. Hossaini et al. (2013) conducted a recent model study comparing four independent brominated VSLS emission estimates. Of all four emission inventories examined Pyle et al., 2011;Ordóñez et al., 2012;Ziska et al., 2013), modeled CHBr 3 and CH 2 Br 2 concentrations using the emission estimates from Liang et al. (2010) compare very well against tropical observations from multiple years of National Oceanic and Atmospheric Administration/Earth System Research Laboratory (NOAA/ESRL) surface flask measurements. The global mean model biases (model -observations) are −0.25 ppt for CHBr 3 (ranging from +0.12 ppt in the tropics to −0.65 ppt in the northern high latitudes) and −0.02 ppt for CH 2 Br 2 (ranging from +0.14 ppt in the tropics to −0.11 ppt in the northern and southern high latitudes). Comparison with the High-performance Instrumented Airborne Platform for  Liang et al. (2010) yields the best CH 2 Br 2 (mean bias of +0.2 ppt compared with observations), compared to the other three emissions, but the simulated CHBr 3 is, on average, ∼+0.76 ppt too high on average. Note that the high bias in modeled CHBr 3 in the tropical western Pacific will likely lead to a high bias in the model Br VSLS y , although the lofting in the Indian Ocean seems to be more critical in determining the total amount of VSLS bromine in the lower stratosphere in this study. The modelsimulated high concentration of brominated VSLS over the Indian Ocean is yet to be evaluated when more surface observations become available, and the importance of the Indian Ocean in delivering higher amounts of VSLS bromine into the TTL needs to be assessed. The simulated high surface concentrations of CHBr 3 and CH 2 Br 2 near tropical Central America were validated with the NASA Tropical Composition, Clouds, and Climate Coupling (TC4) and Intercontinental Chemical Transport Experiment -Phase B (INTEX-B) measurements . This region, compared to the above two, is much less efficient in delivering VSLS bromine into the TTL due to the relatively weaker ascent. However, this ascending branch can be important in particu-lar seasons and in individual years as the Walker circulation moves in the east-west direction between different phases of the El Niño Southern Oscillation (ENSO). For example, using trajectory calculations, Levine et al. (2008) found a clear shift in the TTL air origin from the tropical western Pacific and Indonesia to those from the eastern Pacific and South America in El Niño years.
Of the total VSLS bromine that enters the base of the TTL, about half (∼4 ppt) is in the form of inorganic product gases (Fig. 6) despite our simple assumption that all inorganic bromine reservoir species are highly soluble and all inorganic bromine produced below 350 hPa is classified as the soluble forms HBr and HOBr. A commonly adopted approach in many previous modeling studies is to prescribe the VSLS concentration at the base of the TTL with observed organic source gas mixing ratios in the upper troposphere and track the subsequent chemistry and transport (e.g., Sinnhuber and Folkins, 2006;Aschmann et al., 2011;Aschmann and Sinnhuber, 2013). Our result implies that the Br VSLS y estimate from such approach is not complete as it lacks an important component associated with PGI, which is as large as SGI at the base of the TTL in our model simulation. A second implication of this result is that, if we were to use measurements from the TTL region to quantify the contribution of VSLSs to stratospheric bromine, it is necessary to make measurements of both organic and inorganic forms to fully account for the impact of VSLSs on the atmospheric bromine budget.

The Impact of Convection Strength
We examine the difference between the two convection sensitivity simulations, R MINCNV and R MAXRUN , to illustrate the Atmos. Chem. Phys., 14, 5781-5792 impact of deep convection on Br VSLS y (Fig. 8). The difference between these two simulations (R MINCNV -R MAXRUN ) suggests that in the regions that mattered most (the three tropical convection centers), surprisingly, the minimum convection condition is more favorable for the TST of VSLSs. While weaker convection slightly decreases SGI (by a few tenths ppt), PGI increases significantly (by 2-3 ppt) due to less scavenging under the minimum convection condition. The TST of the longer-lived CH 2 Br 2 is less sensitive to convection strength compared to that of CHBr 3 because of its longer lifetime. The significant increase in CHBr 3 and CH 2 Br 2 in the mid-latitude bands under minimum convection mainly reflects the compensating responses in largescale descent in a general circulation model as a result of changing convection strength (Ott et al., 2011). The decrease of the descent does not impact the TST of VSLSs. Aschmann et al. (2011) found a qualitatively similar behavior for changes of VSLS SGI and PGI under El Niño vs. La Niña conditions -while SGI increased slightly under enhanced deep convection, the sum of SGI and PGI decreased. However, the net decrease in the Br VSLS y (<0.5 ppt) due to changes in convection strength between El Niño vs. La Niña conditions reported in Aschmann et al. (2011) is much smaller than our results above, most likely due to a much higher Br VSLS y abundance in the TTL in this work.
Globally, the differences in convection strength and wet scavenging introduce a ∼2.6 ppt uncertainty into Br VSLS y (∼6.5 ppt in R MAXRUN and ∼9.1 ppt in R MINRUN ) (Fig. 3). Although these two simulations represent the two extreme parameterization conditions within a reasonable range, BrO from both simulations still falls within the uncertainty range of the DOAS balloon observations of ± 2.5 ppt ( Fig. 2; Dorf et al., 2006b).

Conclusions
We use the NASA Goddard 3-D Chemistry Climate Model, GEOSCCM, to quantify the contribution of brominated very short-lived substances (VSLSs) to reactive stratospheric bromine and to test its sensitivity to the strength of deep convection and the associated scavenging.
The inclusion of CHBr 3 and CH 2 Br 2 in a fully interactive stratospheric chemistry module deployed in the tropical tropopause layer (TTL) and the stratosphere adds ∼7.7 pptv to the present-day stratospheric inorganic bromine. The most active transport of VSLSs from the marine boundary layer through the TTL occurs where high surface concentrations of VSLSs are colocated with deep convection centers: (1) the tropical Indian Ocean, (2) the tropical western Pacific warm pool, and (3) off the Pacific coast of Mexico. On average, annually, almost all VSLS-originated bromine available in the marine boundary layer (∼7.8-8.4 ppt) at these active convective lofting regions enters the TTL, half of it in the form of source gas injection (SGI) and half of it in the form of product gas injection (PGI). After lofting above 150 hPa (∼355 K), the majority of the VSLS-originated bromine survives TST and reaches the lower stratosphere, with only a small portion (∼0.8 ppt, 10 %) removed by wet scavenging. Our model results point to a clear need for more surface measurements in the active VSLS lofting regions, in particular the tropical Indian Ocean and the tropical western Pacific, for an improved emissions estimate as well as a better quantification of how much VSLS bromine is available before it ultimately enters the stratosphere through the TTL.
Our current estimate of the contribution of CHBr 3 and CH 2 Br 2 to stratospheric bromine (Br VSLS y ) is higher than previous modeling estimates, mainly reflecting the differences in how transport and wet scavenging of product gases are treated in different models (Hossaini et al., 2012a;Aschmann and Sinnhuber, 2013). Interestingly, even though the models are driven with the same emission estimates Hossaini et al., 2013) and/or with similar source gas abundance at the base of the TTL (Aschmann et al. 2011;Aschmann and Sinnhuber, 2013), this study still yields higher Br VSLS y for two main reasons. Modeling VSLSs and their degradation products in a full stratospheric chemistry scheme in the TTL and stratosphere leads to a better representation of the partition between soluble and insoluble product gases and the associated diurnal variation; both are critical in an accurate Br VSLS y estimate. Secondly, tracking PGI from the free troposphere to the base of the TTL, which is as large as SGI at the base of the TTL in our model simulation, is also important. This was not included in many earlier model studies (e.g., Aschmann et al., 2011;Aschmann and Sinnhuber, 2013), thus implying a missing component in these earlier model VSLS contribution estimates.
Differences in model deep-convection strength can introduce ∼ 30 % (6.6-9.2 ppt) uncertainty in simulated Br VSLS y within the same convection scheme. It is important to point out that this uncertainty is likely larger among different models or different versions of the same model when different convective parameterizations are used.
Contrary to conventional wisdom, which holds that the VSLS impact on stratospheric bromine is larger under more intense deep convection, our simulations suggest that the minimum convection condition is favorable for TST of VSLS due to reduced scavenging of soluble product gases under weaker convection. The impact of convection strength on PGI greatly outweighs the impact on SGI with the change in PGI ∼10 times larger than that in SGI.