Impacts of bromine and iodine chemistry on tropospheric OH and HO2: comparing observations with box and global model perspectives

Abstract

The chemistry of the halogen species bromine and iodine has a range of impacts on tropospheric composition, and can affect oxidising capacity in a number of ways. However, recent studies disagree on the overall sign of the impacts of halogens on the oxidising capacity of the troposphere. We present simulations of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ radicals for comparison with observations made in the remote tropical ocean boundary layer during the Seasonal Oxidant Study at the Cape Verde Atmospheric Observatory in 2009. We use both a constrained box model, using detailed chemistry derived from the Master Chemical Mechanism (v3.2), and the three-dimensional global chemistry transport model GEOS-Chem. Both model approaches reproduce the diurnal trends in $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$. Absolute observed concentrations are well reproduced by the box model but are overpredicted by the global model, potentially owing to incomplete consideration of oceanic sourced radical sinks. The two models, however, differ in the impacts of halogen chemistry. In the box model, halogen chemistry acts to increase $\mathrm{OH}$ concentrations (by 9.8 % at midday at the Cape Verde Atmospheric Observatory), while the global model exhibits a small increase in $\mathrm{OH}$ at the Cape Verde Atmospheric Observatory (by 0.6 % at midday) but overall shows a decrease in the global annual mass-weighted mean$\mathrm{OH}$ of 4.5 %. These differences reflect the variety of timescales through which the halogens impact the chemical system. On short timescales, photolysis of $\mathrm{HOBr}$ and $\mathrm{HOI}$, produced by reactions of ${\mathrm{HO}}_{\mathrm{2}}$ with$\mathrm{BrO}$ and IO, respectively, increases the $\mathrm{OH}$ concentration. On longer timescales, halogen-catalysed ozone destruction cycles lead to lower primary production of $\mathrm{OH}$ radicals through ozone photolysis, and thus to lower$\mathrm{OH}$ concentrations. The global model includes more of the longer timescale responses than the constrained box model, and overall the global impact of the longer timescale response (reduced primary production due to lower ${\mathrm{O}}_{\mathrm{3}}$ concentrations) overwhelms the shorter timescale response (enhanced cycling from ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$), and thus the global $\mathrm{OH}$ concentration decreases. The Earth system contains many such responses on a large range of timescales. This work highlights the care that needs to be taken to understand the full impact of any one process on the system as a whole.

Introduction

Halogen chemistry in the troposphere influences budgets of ${\mathrm{O}}_{\mathrm{3}}$, $\mathrm{HO}$${}_x$ ($\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$) and $\mathrm{NO}$${}_x$ ($\mathrm{NO}$ and ${\mathrm{NO}}_{\mathrm{2}}$) (von Glasow et al., 2004; Saiz-Lopez and von Glasow, 2012; Simpson et al., 2015; Schmidt et al., 2016; Sherwen et al., 2016a, b), affects the oxidation state of atmospheric mercury (Holmes et al., 2006, 2010), and impacts aerosol formation (Hoffmann et al., 2001; O'Dowd and Hoffmann, 2005; McFiggans et al., 2004, 2010; Mahajan et al., 2011; Sherwen et al., 2016c).

The production of bromine and iodine atoms in the marine boundary layer (MBL) following emissions of organohalogen compounds and the inorganic compounds ${\mathrm{I}}_{\mathrm{2}}$ and $\mathrm{HOI}$ has been shown to result in considerable destruction of tropospheric ozone (Read et al., 2008), leading to the production of bromine monoxide ($\mathrm{BrO}$) and iodine monoxide ($\mathrm{IO}$) radicals. Observations of $\mathrm{BrO}$ and $\mathrm{IO}$ radicals within the MBL have demonstrated widespread impacts on atmospheric composition and chemistry (Alicke et al., 1999; Sander et al., 2003; Leser et al., 2003; Saiz-Lopez and Plane, 2004; Saiz-Lopez et al., 2004, 2006; Peters et al., 2005; Whalley et al., 2007; Mahajan et al., 2010a; Commane et al., 2011; Dix et al., 2013; Gomez Martin et al., 2013; Le Breton et al., 2017), including significant effects on $\mathrm{HO}$${}_x$ concentrations and on the ${\mathrm{HO}}_{\mathrm{2}}$ : $\mathrm{OH}$ ratio in coastal and marine locations (Bloss et al., 2005a, 2007, 2010; Sommariva et al., 2006; Kanaya et al., 2007; Whalley et al., 2010).

The role of halogens in $\mathrm{HO}$${}_x$ chemistry was demonstrated during the NAMBLEX campaign in Mace Head, Ireland (Heard et al., 2006), following several studies which attributed box model overestimates of ${\mathrm{HO}}_{\mathrm{2}}$ observations in marine environments to unmeasured halogen monoxides (Carslaw et al., 1999, 2002; Kanaya et al., 2001, 2002, 2007). Simultaneous measurements of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ by laser-induced fluorescence (LIF) (Bloss et al., 2005a; Smith et al., 2006) and halogen species by a combination of DOAS (for $\mathrm{BrO}$ or $\mathrm{IO}$, $\mathrm{OIO}$ and ${\mathrm{I}}_{\mathrm{2}}$) (Saiz-Lopez et al., 2006) and broadband cavity ring-down spectroscopy (BBCRDS) (for $\mathrm{OIO}$ and ${\mathrm{I}}_{\mathrm{2}}$) (Bitter et al., 2005) during NAMBLEX enabled box model calculations to fully explore the impacts of halogens on the local composition. A box model without halogen chemistry was able to reproduce the NAMBLEX $\mathrm{OH}$ observations to within 25 %, but ${\mathrm{HO}}_{\mathrm{2}}$ observations were overestimated by up to a factor of 2 (Sommariva et al., 2006). The introduction of halogen chemistry, using differential optical absorption spectroscopy (DOAS) measurements of $\mathrm{BrO}$ and $\mathrm{IO}$ (Saiz-Lopez et al., 2006) to constrain the model, increased the modelled $\mathrm{OH}$ concentrations by up to 15 % and decreased ${\mathrm{HO}}_{\mathrm{2}}$ by up to 30 % owing to reactions of ${\mathrm{HO}}_{\mathrm{2}}$ with $\mathrm{XO}$ radicals to form HOX which subsequently photolysed to $\mathrm{X}$ $+$ $\mathrm{OH}$ (Sommariva et al., 2006). Bloss et al. (2005a) indicated that up to 40 % of the instantaneous ${\mathrm{HO}}_{\mathrm{2}}$ loss could be attributed to ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{IO}$, and that photolysis of $\mathrm{HOI}$ was responsible for 15 % of the noontime $\mathrm{OH}$ production.

The impacts of halogen chemistry on $\mathrm{HO}$${}_x$ radicals at a site representative of the open ocean have been investigated at the Cape Verde Atmospheric Observatory (CVAO). Measurements of halogen monoxides (Mahajan et al., 2010a) at the site have been shown to have significant impacts on local ozone concentrations, notably in the magnitude of the daily cycle (Read et al., 2008) and have been used to constrain box model calculations used to explore observations of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ made during the RHaMBLe campaign in 2007 (Whalley et al., 2010). The model calculations showed generally good comparisons with the observed $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ concentrations, apart from a period characterised by unusually high concentrations of $\mathrm{HCHO}$. Compared to a model run in which halogen chemistry was absent, bromine and iodine chemistry led to a 9 % increase in the modelled $\mathrm{OH}$ concentration (Whalley et al., 2010). Owing to the dominance of the tropics in global methane oxidation (Bloss et al., 2005b), such an impact of halogens on $\mathrm{OH}$ could have significant consequences for estimates of global methane lifetimes, as well as on our understanding of the impacts of halogen chemistry on climate change.

In general, observationally constrained box model simulations suggest that halogens in the troposphere will increase $\mathrm{OH}$ concentrations, primarily because of a change in the ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ ratio occurring as a result of reactions of halogen oxides ($\mathrm{XO}$) with ${\mathrm{HO}}_{\mathrm{2}}$ to produce a hypohalous acid (HOX) which photolyses to give an $\mathrm{OH}$ radical and a halogen atom (Kanaya et al., 2002, 2007; Bloss et al., 2005a; Sommariva et al., 2006, 2007; Whalley et al., 2010). Other impacts on the $\mathrm{HO}$${}_x$ photochemical system are observed (impacts from changes to $\mathrm{NO}$${}_x$ chemistry etc.), but these are minor and overall the general conclusion is that the halogen chemistry tends to increase the $\mathrm{OH}$ concentration and thus the oxidising capacity of the atmosphere.

However, the observationally constrained studies are typically concerned with processes occurring at the surface, and in a single location. The role of halogen chemistry in the troposphere as a whole is more uncertain, particularly in the free troposphere and on a global scale (Saiz-Lopez and von Glasow, 2012; Simpson et al., 2015). Inclusion of bromine chemistry in the three-dimensional (3-D) chemistry transport model (CTM) MATCH-HPIC resulted in decreases in tropospheric ozone concentrations of $\sim$ 18 % over widespread areas, with regional decreases of up to 40 % (von Glasow et al., 2004). Increases of more than 20 % were found for $\mathrm{OH}$ in the free troposphere, but, globally, changes to $\mathrm{OH}$ were dominated by decreases in $\mathrm{OH}$ in the tropics owing to a reduction in primary production from ${\mathrm{O}}_{\mathrm{3}}$ photolysis, leading to a decrease of 1–2 % in the global mean $\mathrm{OH}$ concentration (von Glasow et al., 2004).

Significant decreases in tropospheric ozone (up to 30 % at high-latitude spring) were also reported for the pTOMCAT model on inclusion of bromine chemistry (Yang et al., 2005). The CAM-Chem global chemistry–climate model has shown an approximate 10 % decrease in global mean tropospheric ozone concentration on incorporation of lower bromine emissions (Saiz-Lopez et al., 2012), while the GEOS-Chem CTM displays a global decrease of 6.5 % (Parrella et al., 2012). The GEOS-Chem model indicated that bromine-catalysed loss of ozone is limited by the rate of production of $\mathrm{HOBr}$, and that ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{BrO}$ is responsible for over 95 % of the global tropospheric $\mathrm{HOBr}$ production. While $\mathrm{HOBr}$ can act as a source of $\mathrm{OH}$ on photolysis, the changes to ${\mathrm{O}}_{\mathrm{3}}$ and $\mathrm{NO}$${}_x$ resulting from the inclusion of bromine chemistry in GEOS-Chem led to a 4 % decrease overall in the global annual mean $\mathrm{OH}$ (Parrella et al., 2012).

Vertically resolved airborne measurements of $\mathrm{IO}$ radicals in the free troposphere over the Pacific Ocean have also demonstrated a role for iodine chemistry throughout the free troposphere, with $\mathrm{IO}$ observed at a mixing ratio of $\sim$ 0.1 $\mathrm{ppt}$ in the free troposphere and found to be present in both recent deep convective outflow and aged free-tropospheric air (Dix et al., 2013). Model simulations to investigate iodine-driven ozone destruction throughout the troposphere indicated that only 34 % of the total iodine-driven ozone loss occurs within the marine boundary layer, with 40 % occurring in a transition layer and 26 % in the free troposphere (Dix et al., 2013).

The CAM-Chem and GEOS-Chem models have also been updated to encompass iodine chemistry, with results from CAM-Chem showing iodine chemistry to be responsible for 17–27 % of the ozone loss in the tropical MBL and 11–27 % of the ozone loss in the marine upper troposphere (Saiz-Lopez et al., 2014). The GEOS-Chem model also showed iodine chemistry to be responsible for significant ozone destruction throughout the troposphere (Sherwen et al., 2016a, b, 2017). The GEOS-Chem simulations, which incorporate chlorine, bromine and iodine chemistry, show a reduction in global tropospheric ozone concentration of 18.6 %, compared to simulations with no halogen chemistry, a reduction in the global mean $\mathrm{OH}$ of 8.2 % to a concentration of 1.28 $\times$ 10${}^{\mathrm{6}}$ ${\mathrm{cm}}^{-\mathrm{3}}$ and a resulting increase in global methane lifetime of 10.8 % to 8.28 $\mathrm{years}$ (Sherwen et al., 2016b).

There is thus a discrepancy between box and global models as to the impact of halogen chemistry on $\mathrm{OH}$ concentrations in the troposphere. Box models suggest that $\mathrm{OH}$ radical concentrations should increase and thus that halogens tend to increase the oxidising capacity, whereas the global models tend to suggest the opposite.

In this work, we use both a detailed chemical box model approach and a global chemistry-transport model to investigate the local and global impacts of halogen chemistry on $\mathrm{HO}$${}_x$ radical concentrations. We focus on seasonal $\mathrm{HO}$${}_x$ observations available from the Cape Verde Atmospheric Observatory (Vaughan et al., 2012). We first provide a summary of the measurement site and the observations, followed by details of the two models used in this study. We then evaluate the impact of halogens on the concentrations of oxidants in the two modelling frameworks and consider the impact of halogen chemistry on global oxidising capacity.

The Cape Verde Atmospheric Observatory

The Cape Verde Atmospheric Observatory is situated on the north east coast of the island of São Vicente (16.848${}^{\circ }$ N, 24.871${}^{\circ }$ W), approximately 500 $\mathrm{km}$ off the west coast of Africa. The observatory is in a region of high marine biological production and, for 95 % of the time, receives the prevailing northeasterly trade wind directly off the ocean (Read et al., 2008; Carpenter et al., 2010). Measurements at the observatory are considered to be representative of the open ocean, and CO, ${\mathrm{O}}_{\mathrm{3}}$, VOCs, $\mathrm{NO}$${}_x$ and $\mathrm{NO}$${}_y$ have been measured near-continuously at the observatory since October 2006 (Lee et al., 2009; Carpenter et al., 2010).

In 2007, the observatory was host to the RHaMBLe intensive field campaign, during which a number of additional measurements were made to complement the long-term measurements at the site (Lee et al., 2010), including LP-DOAS measurements of halogen species (Read et al., 2008; Mahajan et al., 2010a) and formaldehyde (Mahajan et al., 2010b), and LIF-FAGE measurements of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ (Whalley et al., 2010). The halogen monoxide radicals $\mathrm{BrO}$ and $\mathrm{IO}$ exhibited a “top-hat” diurnal cycle (Vogt et al., 1999, 1996; Read et al., 2008; Mahajan et al., 2010a) with essentially zero concentration in the hours of darkness and generally constant values of approximately 2.5 $\mathrm{ppt}$ $\mathrm{BrO}$ and 1.4 $\mathrm{ppt}$ $\mathrm{IO}$ during the day.

The RHaMBLe campaign was followed by the Seasonal Oxidants Study (SOS) in 2009, during which measurements of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ were conducted over three periods (February–March (SOS1), May–June (SOS2), and September (SOS3)). These measurements are discussed in detail by Vaughan et al. (2012). We present here the results from a modelling study of the $\mathrm{HO}$${}_x$ measurements made during SOS1 and SOS2, when supporting measurements are available, using both box and global model approaches. SOS3 is not considered in this work owing to a lack of supporting measurements.

Measurements of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ during the Seasonal Oxidant Study were made by laser-induced fluorescence (LIF) spectroscopy at low pressure using the fluorescence assay by gas expansion (FAGE) technique, and they are described in detail by Vaughan et al. (2012). Briefly, ambient air is drawn into a fluorescence cell situated on the roof of a shipping container and maintained at pressures of $\sim$ 2 $\mathrm{Torr}$. The fluorescence cell has two excitation axes, with excess $\mathrm{NO}$ added at the second axis to titrate ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$, enabling simultaneous detection of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$. $\mathrm{OH}$ radicals in both excitation axes are excited by laser light at $\mathit{\lambda }$ $\sim$ 308 $\mathrm{nm}$, generated by frequency tripling the output of a solid-state $\mathrm{Nd}$ : $\mathrm{YAG}$ pumped $\mathrm{Ti}$ : Sapphire laser system (Bloss et al., 2003). Channel photomultiplier tubes coupled to gated photon counters are used to detect the $A^{\mathrm{2}}{\mathrm{\Sigma }}^{+}-X^{\mathrm{2}}{\mathrm{\Pi }}_i$ $\mathrm{OH}$ fluorescence signal at $\mathit{\lambda }$ $\sim$ 308 $\mathrm{nm}$.

Calibration of the instrument is achieved by measurement of the fluorescence signal from known concentrations of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$, produced by the photolysis of water vapour, and was performed over a range of conditions before, during and after the campaign. For $\mathrm{OH}$, the 1$\mathit{\sigma }$ limit of detection (LOD) was in the range (2–11) $\times$ 10${}^{\mathrm{5}}$ ${\mathrm{cm}}^{-\mathrm{3}}$ for a 5 $\min$ averaging period, while for ${\mathrm{HO}}_{\mathrm{2}}$ 1$\mathit{\sigma }$ LOD was in the range (6–13) $\times$ 10${}^{\mathrm{5}}$ ${\mathrm{cm}}^{-\mathrm{3}}$ for a 4 $\min$ averaging period. Uncertainties (2$\mathit{\sigma }$) in the measurements of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ are $\sim$ 32 % (Vaughan et al., 2012).

Potential interferences in ${\mathrm{HO}}_{\mathrm{2}}$ measurements arising from conversion of alkene- and aromatic-derived peroxy radicals to $\mathrm{OH}$ within the LIF detection cell, as described by Fuchs et al. (2011), are expected to be small for this work owing to relatively low concentrations of alkenes and aromatics at the Cape Verde Atmospheric Observatory (Carpenter et al., 2010; Vaughan et al., 2012). Speciation of the peroxy radicals in the box model output (see Supplement) shows that 87.4 % of the peroxy radicals are ${\mathrm{HO}}_{\mathrm{2}}$ and ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$, 6.5 % ${\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}$ and 1.1 % ${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}{\mathrm{O}}_{\mathrm{2}}$, all of which display no ${\mathrm{HO}}_{\mathrm{2}}$ interference in the laboratory (Whalley et al., 2013: Stone et al., 2014). Peroxy radicals derived from $\mathrm{OH}$-initiated oxidation of ethene and propene (${\mathrm{HOC}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{4}}{\mathrm{O}}_{\mathrm{2}}$ and ${\mathrm{HOC}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}{\mathrm{O}}_{\mathrm{2}}$, respectively) were found to result in an interference signal for ${\mathrm{HO}}_{\mathrm{2}}$ in the laboratory ($\sim$ 40 % for the experimental configuration in this work), but each radical comprises only $\sim$ 0.6 % of the total R${\mathrm{O}}_{\mathrm{2}}$ in this work. Thus, model calculations reported here do not include representation of potential ${\mathrm{HO}}_{\mathrm{2}}$ interferences, although such phenomena may be important in other environments (see for example, Whalley et al., 2013; Stone et al., 2014).

Model approaches

We interpret the observations using two different modelling frameworks. The first is an observationally constrained box model (DSMACC); the second is a global tropospheric chemistry transport model (GEOS-Chem).

Constrained box model

The Dynamically Simple Model of Atmospheric Chemical Complexity (DSMACC) is described in detail by Emmerson and Evans (2009) and Stone et al. (2010), and is a zero-dimensional model using the Kinetic Pre-Processor (KPP) (Sandu and Sander, 2006). In this work we use a chemistry scheme based on a subsection of the hydrocarbons (ethane, propane, iso-butane, $n$-butane, iso-pentane, $n$-pentane, hexane, ethene, propene, 1-butene, acetylene, isoprene, toluene, benzene, methanol, acetone, acetaldehyde and DMS) available from the Master Chemical Mechanism version 3.2 (MCM v3.2 http://mcm.leeds.ac.uk/MCM/home.htt) (Jenkin et al., 2003; Saunders et al., 2003), with a halogen chemistry scheme described by Saiz-Lopez et al. (2006), Whalley et al. (2010) and Edwards et al. (2011). We also include the reaction between $\mathrm{OH}$ and ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ (Bossolasco et al., 2014; Fittschen et al., 2014; Assaf et al., 2016; Yan et al., 2016), with a rate coefficient of 1.6 $\times$ 10${}^{-\mathrm{10}}$ ${\mathrm{cm}}^{\mathrm{3}}{\mathrm{s}}^{-\mathrm{1}}$ (Assaf et al., 2016) and products ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{CH}}_{\mathrm{3}}$O (Assaf et al., 2017), the impact of which on the ${\mathrm{HO}}_{\mathrm{2}}$ : $\mathrm{OH}$ ratio and ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ budget is described in the Supplement. The total number of species in the model is $\sim$ 1200, with $\sim$ 5000 reactions. The full chemistry scheme used in the model is given in the Supplement.

Summary of inputs to the model. Zero values indicate measurements below the limit of detection. Further details can be found in Vaughan et al. (2012) and Carpenter et al. (2010).

Species	Mean	Median	Range
${\mathrm{O}}_{\mathrm{3}}$/$\mathrm{ppb}$	33.8 $\pm$ 8.6	30.7	19.6–49.7
$\mathrm{CO}$/$\mathrm{ppb}$	102.3 $\pm$ 10.3	99.3	87.8–127.3
${\mathrm{H}}_{\mathrm{2}}\mathrm{O}$/$\mathrm{ppm}$	20 542.3 $\pm$ 2753.8	21 290.0	16 778.5–24 909.2
$\mathrm{NO}$/$\mathrm{ppt}$	11.2 $\pm$ 10.6	9.0	0.06–96.2
Ethane/$\mathrm{ppt}$	961.3 $\pm$ 289.4	864.0	625.4–1799.2
Propane/$\mathrm{ppt}$	136.1 $\pm$ 87.05	111.8	20.2–521.5
iso-Butane/$\mathrm{ppt}$	13.4 $\pm$ 9.8	11.1	0–62.7
n-Butane/$\mathrm{ppt}$	21.9 $\pm$ 17.6	17.8	0–112.9
Acetylene/$\mathrm{ppt}$	79.0 $\pm$ 27.8	70.4	45.0–180.5
Isoprene/$\mathrm{ppt}$	0.1 $\pm$ 0.4	0	0–2.6
iso-Pentane/$\mathrm{ppt}$	3.9 $\pm$ 3.2	3.3	0–22.9
n-Pentane/$\mathrm{ppt}$	4.3 $\pm$ 3.0	3.9	0–21.7
n-Hexane/$\mathrm{ppt}$	1.0 $\pm$ 0.7	0.9	0–4.4
Ethene/$\mathrm{ppt}$	43.6 $\pm$ 15.2	46.3	6.4–73.2
Propene/$\mathrm{ppt}$	13.5 $\pm$ 3.6	13.0	6.2–24.1
But-1-ene/$\mathrm{ppt}$	6.5 $\pm$ 1.4	6.3	3.5–10.6
Benzene/$\mathrm{ppt}$	13.0 $\pm$ 17.0	8.3	0–64.4
Toluene/$\mathrm{ppt}$	77.9 $\pm$ 388.8	0	0–2013.9
Acetaldehyde/$\mathrm{ppt}$	511.8 $\pm$ 526.0	599.3	0–2622.6
Methanol/$\mathrm{ppt}$	247.6 $\pm$ 336.2	173.3	0–3337.4
DMS/$\mathrm{ppt}$	8.3 $\pm$ 38.3	0	0–291.8

All measurements are merged onto a 10 min timebase for input to the model and the model is run with constraints applied as discussed in our previous work (Stone et al., 2010, 2011, 2014). Concentrations of ${\mathrm{CH}}_{\mathrm{4}}$ and ${\mathrm{H}}_{\mathrm{2}}$ are kept constant at values of 1770 $\mathrm{ppb}$ (NOAA CMDL flask analysis, ftp://ftp.cmdl.noaa.gov/ccgg/CT-CH4/) and 550 $\mathrm{ppb}$ (Ehhalt and Rohrer, 2009; Novelli et al., 1999), respectively. Formaldehyde measurements were not available during the SOS and we thus use $\mathrm{HCHO}$ concentrations generated by the chemistry in the model, with the modelled $\mathrm{HCHO}$ concentrations in broad agreement with previous measurements at the observatory (Mahajan et al., 2010b). Table 1 shows a summary of the input parameters to the model.

Deposition processes, including dry deposition and wet deposition, and diffusion are represented in the model by a first-order loss process, with the first-order rate coefficient equivalent to a lifetime of approximately 24 h. As discussed by Stone et al. (2010), variation of the deposition lifetime between 1 h and 5 days results in limited changes to the modelled concentrations of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$. Loss of reactive species to aerosol surfaces is represented in the model by parameterisation of a first-order loss process to the aerosol surface (Schwarz, 1986), as discussed by Stone et al. (2014).

A range of aerosol uptake coefficients for ${\mathrm{HO}}_{\mathrm{2}}$ have been reported in the literature, with recent measurements indicating values of ${\mathit{\gamma }}_{{\mathrm{HO}}_{\mathrm{2}}}$ between 0.003 and 0.02 on aqueous aerosols (George et al., 2013), while others have reported values of ${\mathit{\gamma }}_{{\mathrm{HO}}_{\mathrm{2}}}$ $\sim$ 0.1 (Taketani et al., 2008) and increased uptake coefficients in the presence of Cu and Fe ions (Thornton et al., 2008; Mao et al., 2013). In this work we use a value of ${\mathit{\gamma }}_{{\mathrm{HO}}_{\mathrm{2}}}$ $=$ 0.1 in order to maintain consistency with previous modelling studies at the site (Whalley et al., 2010) and to account for potential impacts of ions of copper and iron in aerosol particles influenced by mineral dust (Carpenter et al., 2010; Müller et al., 2010; Fomba et al., 2014; Matthews et al., 2014; Lakey et al., 2015).

The aerosol surface area in the model is constrained to previous measurements of dry aerosol surface area at the observatory, corrected for differences in sampling height between the aerosol and $\mathrm{HO}$${}_x$ measurements and for aerosol growth under humid conditions (Allan et al., 2009; Müller et al., 2010; Whalley et al., 2010).

Halogen species are constrained to “top-hat” profiles for $\mathrm{BrO}$ and $\mathrm{IO}$ (Vogt et al., 1999, 1996; Read et al., 2008), as observed during the RHaMBLe campaign in 2007 (Read et al., 2008; Mahajan et al., 2010a). The observations indicate that while there is day to day variation in $\mathrm{BrO}$ and $\mathrm{IO}$ concentrations, there is little seasonal variation (Mahajan et al., 2010a). $\mathrm{BrO}$ and $\mathrm{IO}$ are thus constrained to the mean observed mixing ratios of 2.5 and 1.4 $\mathrm{ppt}$, respectively, for time points between 09:30 and 18:30 GMT and zero for all other times. Sensitivity to these mixing ratios is discussed in the Supplement. In a similar way to $\mathrm{NO}$${}_x$ (see Stone et al., 2010, 2011), concentrations of all bromine or iodine species, including $\mathrm{BrO}$ and $\mathrm{IO}$, are permitted to vary according to the photochemistry as the model runs forwards. At the end of each 24 h period in the model, the calculated concentrations of $\mathrm{BrO}$ and $\mathrm{IO}$ are compared to the constrained value, and the concentrations of all bromine (Br, ${\mathrm{Br}}_{\mathrm{2}}$, $\mathrm{BrO}$, $\mathrm{HBr}$, $\mathrm{HOBr}$, ${\mathrm{BrONO}}_{\mathrm{2}}$, ${\mathrm{BrNO}}_{\mathrm{2}}$, $\mathrm{BrNO}$) and iodine ($\mathrm{I}$, ${\mathrm{I}}_{\mathrm{2}}$, $\mathrm{IO}$, $\mathrm{HI}$, $\mathrm{HOI}$, $\mathrm{INO}$, ${\mathrm{INO}}_{\mathrm{2}}$, ${\mathrm{IONO}}_{\mathrm{2}}$, $\mathrm{OIO}$, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}$, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}$, ${\mathrm{I}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}$, ${\mathrm{HOIO}}_{\mathrm{2}}$) species are fractionally increased or decreased such that the calculated and constrained concentrations of $\mathrm{BrO}$ and $\mathrm{IO}$ are the same. The model is run forwards in time with diurnally varying photolysis rates until a diurnal steady state is reached, typically requiring between 5 and 10 days.

Observed and modelled concentrations of (a) $\mathrm{OH}$ during SOS1 (February–March 2009, days 58–68); (b) $\mathrm{OH}$ during SOS2 (May–June 2009, days 157–168); (c) ${\mathrm{HO}}_{\mathrm{2}}$ during SOS1; (d) ${\mathrm{HO}}_{\mathrm{2}}$ during SOS2. Observed data are shown in black; box model concentrations with halogen chemistry are shown by filled red circles; box model concentrations without halogen chemistry are shown by open yellow triangles; global model concentrations with halogen chemistry are shown by solid dark blue lines; global model concentrations without halogen chemistry are shown by broken blue lines.

[Figure omitted. See PDF]

Global model

We use the 3-D global chemistry transport model GEOS-Chem (v10-01, www.geos-chem.org). The model has been extensively evaluated against observations (Bey et al., 2001; Evans and Jacob, 2005; Nassar et al., 2009; Mao et al., 2010; Zhang et al., 2010; Parrella et al., 2012; Hu et al., 2017). The model is driven by assimilated winds calculated by the Goddard Earth Observing System at a horizontal resolution of 4${}^{\circ }$ $\times$ 5${}^{\circ }$, with 47 vertical levels from the surface to 50 $\mathrm{hPa}$. Anthropogenic emissions of $\mathrm{CO}$, $\mathrm{NO}$${}_x$ and ${\mathrm{SO}}_{\mathrm{2}}$ are described by the EDGAR 3.2 monthly global inventory (Olivier et al., 2005). Emissions of volatile organic compounds (VOCs) are described by the RETRO monthly global inventory (van het Bolscher, 2008) for anthropogenic sources, with ethane emissions described by Xiao et al. (2008), and the MEGAN v2.1 inventory (Guenther et al., 2006; Barkley et al., 2011) for biogenic sources.

The $\mathrm{HO}$${}_x$–$\mathrm{NO}$${}_x$–$\mathrm{VOC}$–${\mathrm{O}}_{\mathrm{3}}$ chemistry scheme in the model is described in detail by Bey et al. (2001) and Mao et al. (2013), with the isoprene oxidation mechanism described by Paulot et al. (2009). Photolysis rates use the FAST-JX scheme (Bian and Prather, 2002; Mao et al., 2010), with acetone photolysis rates updated by Fischer et al. (2012). Stratospheric chemistry is based on LINOZ McLinden et al. (2000) for ${\mathrm{O}}_{\mathrm{3}}$ and a linearised mechanism for other species as described by Murray et al. (2012).

The model framework includes gas–aerosol partitioning of semi-volatile organic compounds (Liao et al., 2007; Henze et al., 2007, 2009; Fu et al., 2008; Heald et al., 2011; Wang et al., 2011), and heterogeneous chemistry (Jacob, 2000). Coupling between gas-phase chemistry and sulfate–ammonium–nitrate aerosol is described by Park et al. (2004) and Pye et al. (2009). A description of dust aerosol in the model is given by Fairlie et al. (2007). Treatment of sea-salt aerosol is described by Jaegle et al. (2011). The uptake coefficient for ${\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}$ uses the parameterisation by Evans and Jacob (2005), while that for ${\mathrm{HO}}_{\mathrm{2}}$ uses the parameterisation of Thornton et al. (2008). A full description of the organic aerosol chemistry in the model is given by Heald et al. (2011).

The model includes recent updates to the chemistry scheme to include bromine chemistry (Parella et al., 2012; Schmidt et al., 2016) and iodine chemistry (Sherwen et al., 2016a, b). Sources of tropospheric bromine in the model include emissions of ${\mathrm{CHBr}}_{\mathrm{3}}$, ${\mathrm{CH}}_{\mathrm{2}}{\mathrm{Br}}_{\mathrm{2}}$ and ${\mathrm{CH}}_{\mathrm{3}}\mathrm{Br}$, and transport of reactive bromine from the stratosphere. Debromination of sea-salt aerosol is not included in the model following the work of Schmidt et al. (2016), which showed better agreement with observations of $\mathrm{BrO}$ made by the GOME-2 satellite (Theys et al., 2011) and in the free troposphere and the tropical eastern Pacific MBL (Gomez Martin et al., 2013; Volkamer et al., 2015; Wang et al., 2015). Emission rates and bromine chemistry included in the model are described in detail by Parella et al. (2012), with the bromine chemistry scheme described by 19 bimolecular reactions, 2 three-body reactions and 2 heterogeneous reactions using rate coefficients, heterogeneous reaction coefficients and photolysis cross sections recommended by Sander et al. (2011).

Iodine sources include emissions of ${\mathrm{CH}}_{\mathrm{3}}\mathrm{I}$, ${\mathrm{CH}}_{\mathrm{2}}{\mathrm{I}}_{\mathrm{2}}$, ${\mathrm{CH}}_{\mathrm{2}}\mathrm{ICl}$, C${\mathrm{H}}_{\mathrm{2}}\mathrm{IBr}$, ${\mathrm{I}}_{\mathrm{2}}$ and $\mathrm{HOI}$. Emissions for ${\mathrm{CH}}_{\mathrm{3}}\mathrm{I}$ follow Bell et al. (2002), while those of other organic iodine species use parameterisations based on chlorophyll $a$ in the tropics and constant oceanic and coastal fluxes in extratropical regions (Ordonez et al., 2012). Emissions of inorganic iodine species ($\mathrm{HOI}$ and ${\mathrm{I}}_{\mathrm{2}}$) use the results of Carpenter et al. (2013), with oceanic iodide concentrations parameterised by MacDonald et al. (2014). The iodine chemistry scheme includes 26 unimolecular and bimolecular reactions, 3 three-body reactions, 21 photolysis reactions and 7 heterogeneous reactions, using recommendations by Atkinson et al. (2007) and Sander et al. (2011) where available. Full details are given by Sherwen et al. (2016a, b).

Photolysis rates of bromine and iodine compounds are calculated using the FAST-J radiative transfer model (Wild et al., 2000; Bian and Prather, 2002; Mao et al., 2010). Wet and dry deposition are determined as for the standard GEOS-Chem model (Liu et al., 2001; Wesely, 1989; Wang et al., 1998; Amos et al., 2012).

The tropospheric chemistry scheme is integrated using the SMVGEAR solver (Jacobson and Turco, 1994; Bey et al., 2001). The model provides hourly output at the site of the Cape Verde Atmospheric Observatory. Model simulations have been performed in the absence of halogens, with bromine chemistry, with iodine chemistry and with bromine and iodine chemistry combined. Each model simulation is run for 2 years, with the analysis performed on the second year (2009) and the first year discarded as model spin-up to enable evolution of long-lived species.

Model results

We now investigate the impact of halogen chemistry on tropospheric oxidation at the Cape Verde Atmospheric Observatory within our two modelling approaches.

Constrained box model

Figure 1 shows the observed and modelled time series for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ during SOS1 (February, March 2009) and SOS2 (May, June 2009). Observed concentrations of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ were typically higher in SOS2 than SOS1, reaching maximum values in SOS2 of $\sim$ 9 $\times$ 10${}^{\mathrm{6}}$ ${\mathrm{cm}}^{-\mathrm{3}}$ $\mathrm{OH}$ and 4 $\times$ 10${}^{\mathrm{8}}$ ${\mathrm{cm}}^{-\mathrm{3}}$ ${\mathrm{HO}}_{\mathrm{2}}$ (Vaughan et al., 2012). Similar concentrations were observed in May and June 2007 during the RHaMBLe campaign (Whalley et al., 2010), with a $t$ test indicating no statistically significant difference between the $\mathrm{OH}$ concentrations measured in May–June 2009 to those measured in May–June 2007 at the 95 % confidence level (Vaughan et al., 2012). Concentrations of ${\mathrm{HO}}_{\mathrm{2}}$ measured in May–June 2009 were significantly higher than those measured in May–June 2007 at the 95 % confidence level, but were within the 1$\mathit{\sigma }$ day-to-day variability (Vaughan et al., 2012). Temperatures during SOS2 were typically higher than those during SOS1, with higher relative humidity during SOS2 (Vaughan et al., 2012). Air masses during SOS1 had strong contributions from Atlantic marine air and African coastal region, with polluted marine air and Saharan dust contributing $\sim$ 20 and 10 %, respectively, for the first half of the measurement period. Conditions during SOS2 were typically cleaner, with Atlantic marine air representing the major source, although coastal African air contributed $\sim$ 40 % on some days. There was little influence from polluted air, dust or continental air (Vaughan et al., 2012). Analysis of the variance of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ during SOS indicated that $\sim$ 70 % of the total variance could be explained by diurnal behaviour, with the remaining $\sim$ 30 % related to changes in air mass.

Figure 1 shows the observed and modelled time series for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$, for model simulations with and without halogen chemistry. For SOS1, the box model overpredicts the $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ concentrations at the start of the campaign (Julian days 59 and 61), but it performs better for day 63 and captures both the observed diurnal profile and the observed concentrations. For SOS2, the box model tends to agree better with the observations for both $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$. A day-by-day comparison between the models and the observations is shown in the Supplement for days for which box model calculations were possible, which were limited by the availability of supporting data.

Comparison of modelled and observed concentrations of (a) $\mathrm{OH}$ during SOS1 (February–March 2009) and SOS2 (May–June 2009); (b) $\mathrm{OH}$ during SOS1; (c) $\mathrm{OH}$ during SOS2; (d) ${\mathrm{HO}}_{\mathrm{2}}$ during SOS1 and SOS2; (e) ${\mathrm{HO}}_{\mathrm{2}}$ during SOS1; (f) ${\mathrm{HO}}_{\mathrm{2}}$ during SOS2. In each plot, the solid red line indicates the $\mathrm{1}:\mathrm{1}$ line, with 50 % limits given by the broken red lines. The best fit lines are shown in blue and are described by (a) $[\mathrm{OH}{]}_{\mathrm{mod}}$ $=$ (1.09 $\pm$ 0.11) $\times$ $[\mathrm{OH}{]}_{\mathrm{obs}}$ $+$ (0.13 $\pm$ 0.38) $\times$ 10${}^{\mathrm{6}}$ ($r^{\mathrm{2}}$ $=$ 0.49); (b) $[\mathrm{OH}{]}_{\mathrm{mod}}$ $=$ (1.82 $\pm$ 0.26) $\times$ $[\mathrm{OH}{]}_{\mathrm{obs}}$ $-$ (0.01 $\pm$ 0.51) $\times$ 10${}^{\mathrm{6}}$ ($r^{\mathrm{2}}$ $=$ 0.56); (c) $[\mathrm{OH}{]}_{\mathrm{mod}}$ $=$ (1.11 $\pm$ 0.15) $\times$ $[\mathrm{OH}{]}_{\mathrm{obs}}$ $-$ (0.95 $\pm$ 0.66) $\times$ 10${}^{\mathrm{6}}$ ($r^{\mathrm{2}}$ $=$ 0.64); (d) $[{\mathrm{HO}}_{\mathrm{2}}{]}_{\mathrm{mod}}$ $=$ (1.26 $\pm$ 0.10) $\times$ $[{\mathrm{HO}}_{\mathrm{2}}{]}_{\mathrm{obs}}$ $-$ (0.08 $\pm$ 0.22) $\times$ 10${}^{\mathrm{8}}$ ($r^{\mathrm{2}}$ $=$ 0.77); (e) $[{\mathrm{HO}}_{\mathrm{2}}{]}_{\mathrm{mod}}$ $=$ (1.66 $\pm$ 0.21) $\times$ $[{\mathrm{HO}}_{\mathrm{2}}{]}_{\mathrm{obs}}$ $-$ (0.17 $\pm$ 0.34) $\times$ 10${}^{\mathrm{8}}$ ($r^{\mathrm{2}}$ $=$ 0.78); (f) $[{\mathrm{HO}}_{\mathrm{2}}{]}_{\mathrm{mod}}$ $=$ (1.21 $\pm$ 0.12) $\times$ $[{\mathrm{HO}}_{\mathrm{2}}{]}_{\mathrm{obs}}$ $-$ (0.32 $\pm$ 0.30) $\times$ 10${}^{\mathrm{8}}$ ($r^{\mathrm{2}}$ $=$ 0.91).

[Figure omitted. See PDF]

Figure 2 shows the point-by-point model performance for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ for all data points combined, and for SOS1 and SOS2 separately, for the full box model run including halogen chemistry. There is a tendency for overprediction of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ during SOS1 (slopes of modelled vs. observed concentrations are (1.86 $\pm$ 0.26) for $\mathrm{OH}$ and (1.66 $\pm$ 0.21) for ${\mathrm{HO}}_{\mathrm{2}}$), which is dominated by the model overpredictions on days 59 and 61, with better agreement observed during SOS2 (slopes of modelled vs. observed concentrations are (1.11 $\pm$ 0.15) for $\mathrm{OH}$ and (1.21 $\pm$ 0.12) for ${\mathrm{HO}}_{\mathrm{2}}$).

Average diurnal profiles during the Seasonal Oxidant Study (SOS) at the Cape Verde Atmospheric Observatory for (a) $\mathrm{OH}$ during both measurement periods; (b) ${\mathrm{HO}}_{\mathrm{2}}$ during both measurement periods; (c) ${\mathrm{HO}}_{\mathrm{2}}$ : $\mathrm{OH}$ ratio during both measurement periods; (d) $\mathrm{OH}$ during SOS1 (February–March 2009); (e) ${\mathrm{HO}}_{\mathrm{2}}$ during SOS1; (f) ${\mathrm{HO}}_{\mathrm{2}}$ : $\mathrm{OH}$ during SOS1; (g) $\mathrm{OH}$ during SOS2 (May–June); (h) ${\mathrm{HO}}_{\mathrm{2}}$ during SOS2; (i) ${\mathrm{HO}}_{\mathrm{2}}$ : $\mathrm{OH}$ ratio during SOS2. Observed data are shown in black, with grey shading indicating the variability in the observations; box model output with halogen chemistry is shown by solid red lines; box model output without halogen chemistry is shown by broken orange lines; global model output with halogen chemistry is shown by solid dark blue lines; global model output without halogen chemistry is shown by broken blue lines.

[Figure omitted. See PDF]

The measured and modelled average diurnal profiles of $\mathrm{OH}$, ${\mathrm{HO}}_{\mathrm{2}}$ and the ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ ratios are shown in Fig. 3. At midday (11:00–13:00 local time), the full model including halogen chemistry overpredicts $\mathrm{OH}$ by a median factor of 1.52 and ${\mathrm{HO}}_{\mathrm{2}}$ by a median factor of 1.21. A model run containing bromine chemistry but no iodine chemistry gave median midday overpredictions of 1.40 and 1.30 for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$, respectively, while a run containing iodine but not bromine gave equivalent median overpredictions of 1.50 and 1.26, respectively. With no halogen chemistry included in the model, the modelled $\mathrm{OH}$ decreases, giving a median overprediction at midday by a factor of 1.37, while the modelled ${\mathrm{HO}}_{\mathrm{2}}$ increases, resulting in a median overprediction by a factor of 1.37 at midday.

Thus the inclusion of halogens (bromine and iodine) in the box model changes the mean noontime $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ concentrations by $+$9.8 and $-$9.9 %, respectively. This impact of halogen chemistry is consistent in sign and magnitude with previous studies (Kanaya et al., 2002, 2007; Bloss et al., 2005a; Sommariva et al., 2006, 2007; Whalley et al., 2010).

Processes controlling the (a) instantaneous $\mathrm{RO}$${}_x$ radical production (with $\mathrm{RO}$${}_x$ defined here as $\mathrm{OH}$ $+$ ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{HOBr}$ $+$ $\mathrm{HOI}$ $+$ $\mathrm{RO}$ $+$ ${\mathrm{RO}}_{\mathrm{2}}$ owing to the rapid processing between ${\mathrm{HO}}_{\mathrm{2}}$ and $\mathrm{HOBr}$/$\mathrm{HOI}$) around noon (11:00–13:00 LT) for box model simulations with halogen chemistry; (b) the instantaneous $\mathrm{RO}$${}_x$ radical production around noon for box model simulations without halogen chemistry; (c) the instantaneous $\mathrm{RO}$${}_x$ radical loss around noon for box model simulations with halogen chemistry; (d) the instantaneous $\mathrm{RO}$${}_x$ radical loss around noon for box model simulations without halogen chemistry. The main charts show the average results for SOS1 $+$ SOS2, with results for SOS1 and SOS2 shown separately in the smaller charts.

[Figure omitted. See PDF]

Figure 4 shows the mean midday total $\mathrm{RO}$${}_x$ budget (given the fast processing time between ${\mathrm{HO}}_{\mathrm{2}}$ and $\mathrm{HOBr}$ $/$ $\mathrm{HOI}$ we identify the $\mathrm{RO}$${}_x$ family as $\mathrm{OH}$, ${\mathrm{HO}}_{\mathrm{2}}$, $\mathrm{HOBr}$, $\mathrm{HOI}$, $\mathrm{RO}$ and ${\mathrm{RO}}_{\mathrm{2}}$) for the two measurement periods during SOS for model runs with and without halogens. These budgets are similar both between time periods (i.e. SOS1 vs. SOS2) and for box model calculations with and without halogen chemistry. Radical production dominated by photolysis of ozone ($\sim$ 83 %), with photolysis of $\mathrm{HCHO}$ ($\sim$ 10 %), ${\mathrm{CH}}_{\mathrm{3}}\mathrm{CHO}$ ($\sim$ 2 %) and ${\mathrm{H}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}$ ($\sim$ 2 %) playing a significantly smaller role. Radical termination reactions were dominated by ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ ($\sim$ 23 %), aerosol uptake of ${\mathrm{HO}}_{\mathrm{2}}$ ($\sim$ 21 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{HO}}_{\mathrm{2}}$ ($\sim$ 19 %), ${\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}$ $+$ ${\mathrm{HO}}_{\mathrm{2}}$ ($\sim$ 8 %), and $\mathrm{OH}$ $+$ ${\mathrm{HO}}_{\mathrm{2}}$ ($\sim$ 5 %). The inclusion of the reaction between $\mathrm{OH}$ and ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ reduces the importance of radical termination via ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ (from $\sim$ 26 % of the total to $\sim$ 23 % of the total), but otherwise has little impact on the total radical removal owing to the expected production of ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{CH}}_{\mathrm{3}}$O (Assaf et al., 2017). Further details regarding the impact of the reaction between $\mathrm{OH}$ and ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ on the ${\mathrm{HO}}_{\mathrm{2}}$ $/$ $\mathrm{OH}$ ratio and ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ budget are given in the Supplement.

The budget analyses for SOS are consistent with those determined for the RHaMBLe campaign (Whalley et al., 2010; Fittschen et al., 2014; Assaf et al., 2017), reflecting similarities in observed concentrations of long-lived species and the method of the model constraint with observed ${\mathrm{O}}_{\mathrm{3}}$ concentrations and photolysis rates. The primary source of radicals therefore remains fixed in all simulations, with the primary sinks for these species occurring through radical–radical reactions. Thus, the total radical concentration and budget is little impacted by the presence of halogens.

Mean ($\pm$1$\mathit{\sigma }$) midday (11:00–13:00 LT) ratios of ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ (SOS1 and SOS2 combined). Median values are given in parentheses.

	${\mathrm{HO}}_{\mathrm{2}}$ $/$ $\mathrm{OH}$ ratio
Observed	79.1 $\pm$ 34.1 (70.7)
Box model, no halogens	83.4 $\pm$ 15.4 (82.7)
Box model with $\mathrm{Br}$ chemistry	78.9 $\pm$ 15.6 (77.8)
Box model with $\mathrm{I}$ chemistry	71.5 $\pm$ 13.0 (70.4)
Box model with $\mathrm{Br}$ and $\mathrm{I}$ chemistry	68.3 $\pm$ 13.6 (66.9)
Global model, no halogens	80.8 $\pm$ 18.1 (78.9)
Global model with $\mathrm{Br}$ chemistry	81.9 $\pm$ 19.0 (79.7)
Global model with $\mathrm{I}$ chemistry	70.4 $\pm$ 12.5 (70.5)
Global model with $\mathrm{Br}$ and $\mathrm{I}$ chemistry	71.3 $\pm$ 13.2 (71.3)

However, the partitioning of the radicals is impacted by the halogens. Without halogens the average midday (11:00–13:00 LT) ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ ratio is (83.4 $\pm$ 15.4) (median $=$ 82.7), with the halogens this changes to (68.3 $\pm$ 13.6) (median $=$ 66.9) (Table 2). This change in partitioning is mainly due to the reaction of ${\mathrm{HO}}_{\mathrm{2}}$ with $\mathrm{BrO}$ and $\mathrm{IO}$ followed by the photolysis of $\mathrm{HOBr}$ and $\mathrm{HOI}$ to give $\mathrm{OH}$. In this way the halogens tend to reduce the concentration of ${\mathrm{HO}}_{\mathrm{2}}$ and increase the concentration of $\mathrm{OH}$.

Processes controlling the (a) instantaneous $\mathrm{OH}$ radical production around noon (11:00–13:00 LT) for box model simulations with halogen chemistry; (b) the instantaneous $\mathrm{OH}$ radical production around noon for box model simulations without halogen chemistry; (c) the instantaneous $\mathrm{OH}$ radical loss around noon for box model simulations with halogen chemistry; (d) the instantaneous $\mathrm{OH}$ radical loss around noon for box model simulations without halogen chemistry. The main charts show the average results for SOS1$+$SOS2, with results for SOS1 and SOS2 shown separately in the smaller charts.

[Figure omitted. See PDF]

Processes controlling the (a) instantaneous ${\mathrm{HO}}_{\mathrm{2}}$ radical production around noon (11:00–13:00 LT) for box model simulations with halogen chemistry; (b) the instantaneous ${\mathrm{HO}}_{\mathrm{2}}$ radical production around noon for box model simulations without halogen chemistry; (c) the instantaneous ${\mathrm{HO}}_{\mathrm{2}}$ radical loss around noon for box model simulations with halogen chemistry; (d) the instantaneous ${\mathrm{HO}}_{\mathrm{2}}$ radical loss around noon for box model simulations without halogen chemistry. The main charts show the average results for SOS1$+$SOS2, with results for SOS1 and SOS2 shown separately in the smaller charts.

[Figure omitted. See PDF]

In the box model without halogen chemistry, production of $\mathrm{OH}$ is dominated by ozone photolysis (76 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{NO}$ (12 %) and ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{O}}_{\mathrm{3}}$ (9 %), with $\mathrm{OH}$ loss controlled by $\mathrm{OH}$ $+$ $\mathrm{CO}$ (37 %), $\mathrm{OH}$ $+$ ${\mathrm{CH}}_{\mathrm{4}}$ (16 %) and $\mathrm{OH}$ $+$ ${\mathrm{CH}}_{\mathrm{3}}\mathrm{CHO}$ (15 %), as shown in Fig. 5. Production of ${\mathrm{HO}}_{\mathrm{2}}$ in the model excluding halogens is controlled by $\mathrm{OH}$ $+$ CO (45 %), ${\mathrm{CH}}_{\mathrm{3}}\mathrm{O}$ $+$ ${\mathrm{O}}_{\mathrm{2}}$ (19 %) and photolysis of $\mathrm{HCHO}$ (10 %), with loss governed by aerosol uptake (26 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{HO}}_{\mathrm{2}}$ (26 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{NO}$ (15 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ (12 %) and ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{O}}_{\mathrm{3}}$ (10 %). In the presence of halogens, the instantaneous budgets for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ are impacted by $\mathrm{BrO}$ and $\mathrm{IO}$, as shown in Figs. 5 and 6. For the model run including halogens, $\mathrm{OH}$ production is still dominated by ozone photolysis (68 %), but there are significant contributions from photolysis of $\mathrm{HOI}$ (10 %) and $\mathrm{HOBr}$ (3 %). Loss of ${\mathrm{HO}}_{\mathrm{2}}$ is also affected by the presence of the halogen species, with the dominant loss processes including aerosol uptake (20 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{HO}}_{\mathrm{2}}$ (19 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{IO}$ (14 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{NO}$ (12 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}$ (11 %), ${\mathrm{HO}}_{\mathrm{2}}$ $+$ ${\mathrm{O}}_{\mathrm{3}}$ (8 %) and ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{BrO}$ (6 %). The change in the relative importance of ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{NO}$ on inclusion of halogens in the model results from both the increase in the total ${\mathrm{HO}}_{\mathrm{2}}$ sink, owing to the additional losses through ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{BrO}$ and ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{IO}$, and the shift in $\mathrm{NO}$${}_x$ partitioning owing to the reactions $\mathrm{Br}$ $+$ $\mathrm{NO}$ $\to$ $\mathrm{Br}$ $+$ ${\mathrm{NO}}_{\mathrm{2}}$ and $\mathrm{IO}$ $+$ $\mathrm{NO}$ $\to$ $\mathrm{I}$ $+$ ${\mathrm{NO}}_{\mathrm{2}}$. The reactions of $\mathrm{BrO}$ and $\mathrm{IO}$ with $\mathrm{NO}$ result in a change in the ${\mathrm{NO}}_{\mathrm{2}}$ : $\mathrm{NO}$ ratio of approximately 10 %, on average, which reduces the impact of ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{NO}$ as both a sink for ${\mathrm{HO}}_{\mathrm{2}}$ and a source for $\mathrm{OH}$. As shown in Figs. 4–6 there is little difference in the radical budgets between SOS1 and SOS2.

This box modelling study is consistent with previous studies (Kanaya et al., 2002, 2007; Bloss et al., 2005a; Sommariva et al., 2007; Whalley et al., 2010; Mahajan et al., 2010a; Stone et al., 2012) in that it implies that halogen chemistry is likely to increase the $\mathrm{OH}$ concentration of the marine boundary layer (and potentially other regions of the troposphere) as it enhances the ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ conversion through the production of $\mathrm{HOBr}$ and $\mathrm{HOI}$. We now look at the impact of halogen chemistry on the concentrations of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ at the Cape Verde Atmospheric Observatory within the framework of a global atmospheric chemistry model.

Normalised probability distribution functions showing the fractional changes in $\mathrm{OH}$ (a, c, e) and ${\mathrm{HO}}_{\mathrm{2}}$ (b, d, f) in GEOS-Chem for all grid boxes on inclusion of bromine chemistry (a, b), iodine chemistry (c, d) and bromine and iodine chemistry combined (e, f).

[Figure omitted. See PDF]

Percentage changes to annual surface layer concentrations of $\mathrm{OH}$ (a, c, e) and ${\mathrm{HO}}_{\mathrm{2}}$ (b, d, f) in GEOS-Chem on inclusion of bromine chemistry (a, b), iodine chemistry (c, d) and bromine and iodine chemistry combined (e, f).

[Figure omitted. See PDF]

Annual surface layer mixing ratios (ppt) of BrO (a, c) and IO (b, d) radicals in GEOS-Chem for model runs with just bromine chemistry (a), just iodine chemistry (b) and bromine and iodine chemistry combined (c, d).

[Figure omitted. See PDF]

Global model

Figure 1 shows the time series for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ calculated by the global model GEOS-Chem, with the average diurnal profiles shown in Fig. 3. The global model displays a significant overprediction for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ during SOS1, but it exhibits reasonable skill at reproducing the observed concentrations during SOS2 and captures the ${\mathrm{HO}}_{\mathrm{2}}$ : $\mathrm{OH}$ ratio for both measurement periods. The overpredictions of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ in the global model likely result from a combination of missing $\mathrm{OH}$ sinks, particularly oxygenated volatile organic compounds (oVOCs), which are currently underestimated in the global model (Millet et al., 2015), and potential overprediction of the primary radical production rate owing to reductions in photolysis rates resulting from cloud cover that are not captured by the global model.

At midday (11:00–13:00 LT), the modelled to observed ratios for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ for the global model excluding halogen chemistry are (1.52 $\pm$ 1.02) and (1.72 $\pm$ 0.80), respectively, with a mean modelled ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ ratio of (80.8 $\pm$ 18.1) (compared to the observed ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ ratio of (79.1 $\pm$ 34.1)). For the global model run including bromine chemistry, but not iodine, the mean midday modelled to observed ratios for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ are (1.48 $\pm$ 1.05) and (1.69 $\pm$ 0.81), respectively, with a mean midday modelled ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ ratio of (81.9 $\pm$ 19.0). Bromine chemistry thus acts to decrease the concentrations of both $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$, in contrast to the box model results which show increased concentrations of $\mathrm{OH}$ and decreased concentrations of ${\mathrm{HO}}_{\mathrm{2}}$. For the model run including iodine, but not bromine, the midday modelled to observed ratios for $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ are (1.57 $\pm$ 1.00) and (1.59 $\pm$ 0.81), respectively, with a mean midday modelled ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ ratio of (70.4 $\pm$ 12.5). Iodine chemistry thus results in increased $\mathrm{OH}$ and decreased ${\mathrm{HO}}_{\mathrm{2}}$ for both the global and box model simulations at the Cape Verde Atmospheric Observatory. Inclusion of bromine and iodine chemistry combined leads to midday modelled to observed ratios of (1.53 $\pm$ 1.01) for $\mathrm{OH}$ and (1.57 $\pm$ 0.82) for ${\mathrm{HO}}_{\mathrm{2}}$, and a mean midday modelled ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ ratio of (71.3 $\pm$ 13.2). These results are shown in Table 2, alongside those for the box model.

The results from the global model at Cape Verde Atmospheric Observatory thus differ from those of the box model. For the box model, inclusion of bromine and iodine chemistry, whether separately or combined, leads to increased $\mathrm{OH}$ and decreased ${\mathrm{HO}}_{\mathrm{2}}$ through increased conversion of ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ through the production and subsequent photolysis of $\mathrm{HOBr}$ and $\mathrm{HOI}$. In the global model a more complex pattern emerges. In a similar way to the box model, the ${\mathrm{HO}}_{\mathrm{2}}$ concentrations in the global model are decreased on inclusion of bromine and/or iodine owing to the additional loss reactions ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{BrO}$ and ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{IO}$. When bromine chemistry is considered in the global model in isolation from iodine chemistry, the $\mathrm{OH}$ concentration decreases, despite the production and photolysis of $\mathrm{HOBr}$. This decrease occurs as a result of a reduction in the ${\mathrm{O}}_{\mathrm{3}}$ concentration in the model on inclusion of bromine chemistry owing to the reaction of $\mathrm{Br}$ with ${\mathrm{O}}_{\mathrm{3}}$, which leads to a decrease in the rate of primary radical production from ${\mathrm{O}}_{\mathrm{3}}$ photolysis and thus lower $\mathrm{OH}$ concentrations. The impact of the decreased radical production rate is greater than that leading to increased $\mathrm{OH}$ production through $\mathrm{HOBr}$ photolysis, and the net $\mathrm{OH}$ concentration is reduced in the global model. This effect is not observed in the box model calculations as the model runs are constrained to long-lived species – including ${\mathrm{O}}_{\mathrm{3}}$. The change in ${\mathrm{O}}_{\mathrm{3}}$ concentration on the inclusion of halogen chemistry is thus not considered in the box model simulations, and the subsequent impacts of halogens consider only those changes occurring on a more rapid timescale, which lead to increases in the $\mathrm{OH}$ concentration.

However, the inclusion of iodine chemistry in the global model does lead to increased $\mathrm{OH}$ concentrations at the Cape Verde Atmospheric Observatory. Direct emissions of $\mathrm{HOI}$ in the global model, in addition to chemical production through ${\mathrm{HO}}_{\mathrm{2}}$ $+$ $\mathrm{IO}$, result in increased $\mathrm{OH}$ production through $\mathrm{HOI}$ photolysis as well as the repartitioning of ${\mathrm{HO}}_{\mathrm{2}}$ and $\mathrm{OH}$ through $\mathrm{HOI}$ production in a similar manner to that for $\mathrm{HOBr}$. However, the more rapid cycling of ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$ through the more rapid production and photolysis of $\mathrm{HOI}$ compared to $\mathrm{HOBr}$ reduces the impact of iodine chemistry on the ${\mathrm{HO}}_{\mathrm{2}}$ : $\mathrm{OH}$ ratio compared to that for bromine chemistry. Iodine chemistry thus can reduce the $\mathrm{OH}$ concentration similarly to bromine chemistry, through the destruction of ${\mathrm{O}}_{\mathrm{3}}$ and subsequent reduction in primary production rate, but the impact is less than that for bromine, and can be offset by the direct emissions of $\mathrm{HOI}$, which increases the production rate of $\mathrm{OH}$ through photolysis.

The impacts of iodine chemistry in the global model are thus more complex than those for bromine chemistry. When bromine and iodine chemistry are combined in the global model there is a competition between the effects of the reduction in primary production of $\mathrm{OH}$, through depletion of ${\mathrm{O}}_{\mathrm{3}}$, and the production of $\mathrm{OH}$ from photolysis of $\mathrm{HOBr}$ and $\mathrm{HOI}$ and, for the model simulations at the Cape Verde Atmospheric Observatory, the impacts of direct $\mathrm{HOI}$ emissions dominate. The $\mathrm{OH}$ concentration is thus marginally increased compared to simulations containing no halogens, although the ${\mathrm{HO}}_{\mathrm{2}}$ concentrations are significantly decreased.

The impacts of halogens on $\mathrm{OH}$ radical concentrations in the global model thus display a complexity that is somewhat obscured in the box model simulations. Overall, the inclusion of halogens in the global model leads to a slight increase in $\mathrm{OH}$ at the Cape Verde Atmospheric Observatory, but, owing to the opposing effects of bromine and iodine, this result is subject to the modelled concentrations of bromine and iodine species. Observations at the Cape Verde Atmospheric Observatory made between November 2006 and June 2007 indicate “top-hat” profiles for $\mathrm{BrO}$ and $\mathrm{IO}$, with average daytime mixing ratios of 2.5 and 1.4 $\mathrm{ppt}$, respectively, and little variability over the entire campaign (Read et al., 2008; Mahajan et al., 2010a). The global model simulations reported here predict average mixing ratios of $\sim$ 0.5 $\mathrm{ppt}$ for $\mathrm{BrO}$ and $\sim$ 1 $\mathrm{ppt}$ for $\mathrm{IO}$ during SOS and thus underpredict $\mathrm{BrO}$ but perform well for $\mathrm{IO}$. The underprediction of $\mathrm{BrO}$ at the Cape Verde Atmospheric Observatory results from recent model updates which exclude emissions of bromine species from sea-salt debromination (Schmidt et al., 2016) in order to provide improved agreement with observations of $\mathrm{BrO}$ made by the GOME-2 satellite (Theys et al., 2011) and in the free troposphere and the tropical eastern Pacific MBL (Gomez Martin et al., 2013; Volkamer et al., 2015; Wang et al., 2015). If sea-salt debromination were included, daytime mixing ratios of $\mathrm{BrO}$ at the Cape Verde Atmospheric Observatory would be approximately 2 $\mathrm{ppt}$, as shown by Parella et al. (2012) and Schmidt et al. (2016), and thus in closer agreement to the observations. Increased modelled concentrations of $\mathrm{BrO}$ at the Cape Verde Atmospheric Observatory resulting from inclusion of sea-salt debromination would have a greater effect on $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$, leading to more significant decreases in $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ when bromine chemistry is included without iodine chemistry, with the larger decrease in $\mathrm{OH}$ potentially off-setting the increase in $\mathrm{OH}$ observed when bromine and iodine chemistry are combined. However, the current model simulations do not consider the coupling between bromine and sulfur chemistry, which may represent a significant sink for reactive bromine species in the troposphere and balance sources from sea-salt debromination (Chen et al., 2017). These results thus demonstrate the need for further investigation and constraint of sources and emission rates of bromine species, and of the coupling between sulfur chemistry and reactive bromine species. We now discuss the global impacts of halogen chemistry.

Global impacts of halogen chemistry on $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$

On the global scale, concentrations of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ are reduced on inclusion of bromine and iodine chemistry, both individually and combined. The global mass-weighted annual mean $\mathrm{OH}$ concentration decreases by 3.8 % on inclusion of bromine chemistry, but only 0.02 % on inclusion of iodine chemistry alone. When the chemistry of bromine and iodine is combined in the model, the global mass-weighted annual mean $\mathrm{OH}$ concentration decreases by 4.5 %. For ${\mathrm{HO}}_{\mathrm{2}}$, the global mass-weighted annual mean is decreased by 4.2 % by bromine, 5.6 % by iodine and 9.7 % by bromine and iodine combined. Figure 7 shows the probability distribution functions for the changes to $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ concentrations for the monthly mean values for all grid boxes within the troposphere for the year. For the majority of grid boxes, concentrations of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ are reduced on inclusion of bromine chemistry, with iodine also generally reducing ${\mathrm{HO}}_{\mathrm{2}}$ concentrations but leading to a wider spread of changes to the $\mathrm{OH}$ concentration, and similar numbers of grid boxes showing increased and decreased concentrations. When bromine and iodine chemistry are combined in the model, ${\mathrm{HO}}_{\mathrm{2}}$ shows a more significant decrease than for either halogen individually, and $\mathrm{OH}$, although exhibiting increased concentrations in a significant number of grid boxes, displays a greater tendency for decreased concentrations.

Figure 8 shows the changes to the annual modelled surface layer $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ concentrations on inclusion of halogen chemistry, with annual surface layer mixing ratios of $\mathrm{BrO}$ and $\mathrm{IO}$ shown in Fig. 9. The most significant changes to $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ occur over marine regions, particularly over the southern Pacific. Smaller changes are observed over land, and any increased concentrations, including those for $\mathrm{OH}$ over the Cape Verde Atmospheric Observatory, can be seen to occur in coastal regions where the impacts of direct $\mathrm{HOI}$ emissions dominate and concentrations of $\mathrm{IO}$ concentrations are typically elevated.

Thus, overall, halogens act to reduce the oxidising capacity of the troposphere through reductions to ${\mathrm{O}}_{\mathrm{3}}$ and subsequent reductions in the primary production rates of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$, despite the slight increase in $\mathrm{OH}$ concentration predicted by the global model at the Cape Verde Atmospheric Observatory. Consideration of the full extent of the impacts of halogens on the global oxidising capacity is hindered by uncertainties in the concentrations and distributions of halogen species, and model representations of halogen processes, particularly those relating to sea-salt debromination, ocean iodide emissions, parameterisations of iodine recycling in aerosols and photolysis of higher iodine oxides (Sherwen et al., 2016a).

Conclusions

Measurements of $\mathrm{OH}$ and ${\mathrm{HO}}_{\mathrm{2}}$ made by LIF-FAGE at the Cape Verde Atmospheric Observatory during the Seasonal Oxidants Study in 2009 have been simulated by a constrained box model and a three-dimensional global chemistry transport model. The observations are generally reproduced well by the box model but are overpredicted by the global model.

The oxidising capacity of the two models, as manifested by the $\mathrm{OH}$ concentration, shows opposing sensitivity to halogens. The constrained box model shows an increase in $\mathrm{OH}$ concentrations with the inclusion of halogens, whereas the global transport model shows a decrease in $\mathrm{OH}$ concentrations globally, despite a marginal decrease in the $\mathrm{OH}$ concentration at the Cape Verde Atmospheric Observatory. This difference between models reflects differing representation of chemical timescales by the models. The box model is constrained to concentrations of long-lived compounds, including ${\mathrm{O}}_{\mathrm{3}}$, and considers only impacts on short timescales, whereas the global model includes impacts occurring on longer timescales. Within this context, the box model includes the short timescale impact of halogens on the repartitioning of ${\mathrm{HO}}_{\mathrm{2}}$ to $\mathrm{OH}$, thus increasing $\mathrm{OH}$ and decreasing ${\mathrm{HO}}_{\mathrm{2}}$, but does not consider the longer timescale impact of halogen-mediated ozone destruction which impacts primary radical production. This highlights a general problem with understanding the complex interactions within atmospheric chemistry and the Earth system in general. Evaluating the impact of a small part of the system on the system as a whole can be difficult and the most significant processes may occur on timescales significantly longer than those of the perturbation.

Data deposited at the Centre for Environmental Data Analysis, CEDA: http://archive.ceda.ac.uk/.

The Supplement related to this article is available online at https://doi.org/10.5194/acp-18-3541-2018-supplement.

The authors declare that they have no conflict of interest.

Acknowledgements

This project was funded by the Natural Environment Research Council (NERC, NE/E011403/1), with support of the Cape Verde Atmospheric Observatory by the National Centre for Atmospheric Science (NCAS) and the SOLAS project. Daniel Stone would also like to thank NERC for the award of an Independent Research Fellowship (NE/L010798/1). Computational resources were provided by the NERC BACCHUS project (NE/L01291X/1). Edited by: Rolf Sander Reviewed by: two anonymous referees