Volume 107, Issue D17 p. ACH 9-1-ACH 9-16
Composition and Chemistry
Free Access

Modeling halogen chemistry in the marine boundary layer 1. Cloud-free MBL

Roland von Glasow

Roland von Glasow

Atmospheric Chemistry Division, Max-Planck-Institut für Chemie, Mainz, Germany

Search for more papers by this author
Rolf Sander

Rolf Sander

Atmospheric Chemistry Division, Max-Planck-Institut für Chemie, Mainz, Germany

Now at Scripps Institution of Oceanography, University of California, San Diego, California, USA.

Search for more papers by this author
Andreas Bott

Andreas Bott

Meteorologisches Institut, Universität Bonn, Bonn, Germany

Search for more papers by this author
Paul J. Crutzen

Paul J. Crutzen

Atmospheric Chemistry Division, Max-Planck-Institut für Chemie, Mainz, Germany

Also at Scripps Institution of Oceanography, University of California, San Diego, California, USA.

Search for more papers by this author
First published: 14 September 2002

This is part of DOI:10.1029/2001JD000943.

Abstract

[1] A numerical one-dimensional model of the marine boundary layer (MBL) is presented. It includes chemical reactions in the gas phase and aerosol particles, focusing on the reaction cycles of halogen compounds. Results of earlier box model studies were confirmed. They showed the acid-catalyzed activation of bromine from sea salt aerosol, and the role of halogen radicals in the destruction of O3. A distinct diurnal variation in BrO mixing ratios with maxima at sunrise and sunset was found which might be the cause of the recently published “sunrise ozone destruction.” Maxima of BrO and sea salt acidity are predicted at the top of the MBL and not close to the sea surface where sea salt spray is produced. The presence of sulfate aerosol was found to be important for the recycling of less reactive to photolyzable bromine species. Day/night and seasonal differences in halogen chemistry are shown.

1. Introduction

[2] Chemical reactions on and inside atmospheric particles can have a significant influence on the chemistry of the gas phase (see Jacob [2000] for a review). In the case of the Antarctic “ozone hole” where rapid O3 destruction by halogen radicals takes place, interactions between the gas and the particulate phase (polar stratospheric clouds) are very important (see e.g., Brasseur et al. [1999] for an overview). A sudden O3 destruction is also observed in the Arctic boundary layer during polar sunrise. It is mainly due to reactions of bromine species that probably originate from deposits on the snow pack. Reactions on aerosols are also important here [e.g., Barrie et al., 1988; Bottenheim et al., 1990; Barrie and Platt, 1997].

[3] In this study we focus on the interactions between gas and particulate chemistry in the marine boundary layer (MBL). Sea salt aerosol contains the halogens chlorine and bromine. Provided that an efficient path for the release of halogens from the aerosol exists, sea salt aerosol is a potential source for gas phase halogen species in the MBL. As about 70% of the earth's surface is covered by oceans, reactions involving halogens released from sea salt aerosol are of potential global importance.

[4] Mozurkewich [1995] proposed an acid-catalyzed mechanism for the release of bromine from Arctic sea salt aerosol. In several laboratory studies release of Br2 and BrCl from sea salt solutions could be shown [Abbatt and Waschewsky, 1998; Hirokawa et al., 1998; Disselkamp et al., 1999; Behnke et al., 1999]. Fickert et al. [1999] could show in the laboratory that the bromine activation depends on the pH of the solution. Contrary to HCl, HBr cannot be released significantly by acid displacement due to its high solubility. Recently, several authors [Ghosal et al., 2000; Zangmeister et al., 2001] found that surface segregation of bromide and chloride ions occurs in salt crystals. This could facilitate the release of bromine and chlorine from the surfaces. Numerical simulations by Knipping et al. [2000] and Jungwirth and Tobias [2001] suggest that in salt solutions the halides are also present at the surface and Br and I have actually higher concentrations at the surface than in the bulk.

[5] Data from kinetic studies have been incorporated into numerical models showing the activation of reactive bromine in the MBL [Sander and Crutzen, 1996; Vogt et al., 1996; Keene et al., 1998; Sander et al., 1999; Dickerson et al., 1999].

[6] Field measurements of sea salt aerosol composition have been made for a long time, which often show a decrease of the bromine to sodium ratio of aged sea salt aerosol compared to sea water [e.g., Duce et al., 1965; Kritz and Rancher, 1980; Ayers et al., 1999].

[7] Ayers et al. [1999] presented measurements from Cape Grim, Tasmania, showing bromine deficits of 30% to 50% on an annual average and maximum monthly mean values of more than 80%. They also show that Br deficits were linked to the availability of sulfate acidity in the aerosol, pointing to the importance of acid catalysis in the dehalogenation process. The measurements of Rancher and Kritz [1980] and Kritz and Rancher [1980] show that total inorganic gas phase bromine and particulate bromine are of the same order of magnitude.

[8] Individual reactive gas phase halogen species are difficult to detect because of their relatively low mixing ratios. Most available measurements of single bromine species are for BrO using the differential optical absorption spectroscopy (DOAS) technique. Rather large BrO mixing ratios of 20–30 pmol/mol were found in the Arctic during polar sunrise [Hausmann and Platt, 1994; Tuckermann et al., 1997] which were always associated with dramatic loss of ozone. Satellite measurements show large plumes of BrO in the Arctic spring [Richter et al., 1998; Wagner and Platt, 1998]. McElroy et al. [1999] provided evidence for the presence of BrO in the free troposphere during Arctic polar sunrise. At the Dead Sea, BrO mixing ratios of more than 90 pmol/mol [Hebestreit et al., 1999] were found, which are due to the special location of the measurement site downwind of large salt pans. These extreme events show that the processes involved in bromine related ozone loss do not necessarily depend on the availability of ice surfaces as was proposed for the Arctic case. Several groups tried to measure BrO in the MBL with the DOAS technique. During these campaigns the detection limit for BrO was between 1 and 10 pmol/mol. Whenever BrO was measured, it was near the detection limit (personal communication with D. Perner, B. Allan, and K. Hebestreit). Therefore these measurements provide us only with upper limits for BrO that are consistent with our model results (see below). Measurements of total inorganic bromine, however, show that there is significant bromine in the gas phase.

[9] Iodine compounds may also be important for the chemistry of the MBL. Iodocarbons are emitted at the sea surface and are photolyzed producing chemically reactive species like I, IO, and OIO. IO and OIO have been measured in the pmol/mol range [Alicke et al., 1999; Allan et al., 2000; Hebestreit et al., 2000; Allan et al., 2001]. These studies highlight the importance of halogen chemistry in the MBL.

[10] In this work we present a modeling study of the multiphase chemistry in the MBL focusing on the influence of halogens. For this purpose, we extended a one-dimensional boundary layer model to include chemical reactions in the gas phase as well as in and on aerosol (sea salt and sulfate particles). This model (MISTRA-MPIC) is designed for detailed process studies.

[11] From some measurement campaigns, field data are available that we use to discuss the model results. Our findings might also help to design future field campaigns or laboratory studies.

2. Description of the MBL Model

[12] We used the one-dimensional model MISTRA-MPIC, where MPIC stands for Max-Planck-Institut für Chemie version. The meteorological and microphysical part of the boundary layer model is described in detail by Bott et al. [1996] and Bott [1997]. In addition to a description of the dynamics and thermodynamics, it includes a detailed microphysical module that calculates particle growth explicitly and treats interactions between radiation and particles. Chemical reactions in the gas phase are considered in all model layers, aerosol chemistry only in layers where the relative humidity is greater than the crystallization humidity (see section 2.2). Fluxes of heat, moisture, sea salt aerosol particles, and gases from the ocean are included. Figure 1 shows schematically the most important processes that are included in the model for the cloud-free MBL. More details are given by von Glasow [2000].

Details are in the caption following the image
Schematic depiction of the most important processes included in the one-dimensional boundary layer model MISTRA-MPIC. The free troposphere is denoted as FT, the marine boundary layer as MBL. A typical vertical profile of the relative humidity (rh) is shown as the dashed line. Gas phase chemistry is calculated in all layers.

2.1. Meteorology, Microphysics, and Thermodynamics

[13] The set of prognostic variables comprises the horizontal components of the wind speed u, v, the specific humidity q, and the potential temperature Θ as listed by Bott et al. [1996]. For the cloud-free runs that we discuss here, no subsidence was applied.

[14] The microphysics is treated using a two-dimensional particle size distribution function f(a, r) with the total particle radius r and the dry aerosol radius a the particles would have, when no water were present in the particles. The two-dimensional particle grid is divided into 70 logarithmically equidistant spaced dry aerosol classes, with the minimum dry aerosol radius being 0.01 μm and the maximum radius 15 μm. Choosing these values allows to account for all accumulation mode particles and most of the coarse particles. Each of the 70 dry aerosol classes is associated with 70 total particle radius classes, ranging from the actual dry aerosol radius up to 60 μm. Particle growth is calculated explicitly for each bin of the 2D particle spectrum including effects of radiation on the particle growth [see also Bott et al., 1996].

[15] For the calculation of the radiative fluxes a δ-two stream approach is used [Zdunkowski et al., 1982; Loughlin et al., 1997]. The radiative fluxes are used for calculating heating rates and the effect of radiation on particle growth. The radiation field is calculated with the aerosol particle data from the microphysical part of the model, so feedbacks between radiation and particle growth are fully implemented.

[16] Sea salt particles are emitted by bursting bubbles at the sea surface [e.g., Pruppacher and Klett, 1997]. The parameterization of Monahan et al. [1986] is used to estimate the flux F of particles with radius r80 (in μm) at a relative humidity of 80% per unit area of sea surface (in particles m−2 s−1 μm−1):
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0001
where B = (0.380 − log r80)/0.65 and u10 is the wind speed at 10 m height in m/s. For each dry aerosol radius bin of the 2D microphysical spectrum the equilibrium radius at the actual relative humidity is calculated and the appropriate number of particles according to equation (1) is added in the lowermost model layer. The Monahan et al. [1986] estimate of the sea salt aerosol flux based on wind tunnel experiments is believed to yield good results for small particles [Andreas, 1998]. Monahan et al. [1986] also parameterize the emission of particles with radii larger than r = 10 μm by spume production. However, only the bubble burst mechanism is included because as pointed out by, e.g., Gong et al. [1997] and Andreas [1998] the additional terms lead to an overestimation of the particle flux compared to measurements. For higher wind speeds the resulting mass flux of sea salt particles is less realistic, so for the model runs with high wind speed the parameterization of Smith et al. [1993] that is based on measurements off the Scottish coast is used.

2.2. Chemistry

[17] The multiphase chemistry module comprises chemical reactions in the gas phase as well as in aerosol particles. Transfer between gas and aqueous phase and surface reactions on particles are also included. A complete listing of the reactions is available as electronic supplement1 and on the web (http://www.rolandvonglasow.de). The reaction set is an updated version of Sander and Crutzen [1996] (see also http://www.mpch-mainz.mpg.de/~sander/mocca) plus some organic reactions from Lurmann et al. [1986].

[18] The prognostic equation for the concentration of a gas phase chemical species cg (in mol/mair3) including turbulent exchange, deposition on the ocean surface, chemical production and destruction, emission and exchange with the aqueous phases is
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0002
Kh is the turbulent exchange coefficient for heat that is usually used for the treatment of turbulent transport of gases as well. P and Lcg are chemical production and loss terms, respectively. The emission E as well as dry deposition D are effective only in the lowermost model layer. The dry deposition velocity for gases at the sea surface, vgdry, that is needed for the determination of D, is calculated using the resistance model described by Wesely [1989]. The last term in equation (2) describes the transport from the gas phase into the aqueous phases according to the formulation by Schwartz [1986] [see also Sander, 1999], nkc is the number of the aqueous classes as explained below and kHcc the dimensionless Henry constant. It is obtained by kHcc = kHRT, where kH is in mol/(m3 Pa), wl, i is the dimensionless liquid water content (maq3/mair3) of bin i. For a single particle, the mass transfer coefficient kt is defined as
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0003
with the particle radius r, the mean molecular speed equation image (M is the molar mass), the accommodation coefficient α (see Tables A1–A6 in the electronic supplement and on the web: http://www.rolandvonglasow.de), and the gas phase diffusion coefficient Dg. Dg is approximated as Dg = λequation image/3 [e.g., Gombosi, 1994, p. 125] using the mean free path length λ.
[19] When aqueous chemistry is treated with size bins, one has to account for particles with different radii within each bin. The mean transfer coefficient equation image for a particle population is given by the integral:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0007
where the size distribution function ∂N/∂lgr is defined in Table 3.
[20] Aqueous chemistry is calculated in two bins: deliquescent aerosol particles with a dry radius less than 0.5 μm are included in the “sulfate aerosol” bin, whereas deliquescent particles with a dry aerosol radius greater than 0.5 μm are in the “sea salt aerosol” bin. Using these values we ensure that most aerosol mass, which is the important property for the aerosol chemistry, is included in each class. Although the composition of the particles changes over time the terms “sulfate” and “sea salt” aerosol are used to describe the origin of the particles. In each of these classes the following prognostic equation is solved for each chemical species ca, i (in mol/mair3), where the index i stands for the i-th aqueous class:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0008
The individual terms have similar meanings as in equation (2). The dry deposition velocity of particles, va, idry, that is needed for the determination of D, is calculated as explained by Seinfeld and Pandis [1998].

[21] The concentration of H+ ions is calculated like any other species, i.e., no further assumptions are made. The charge balance (in other models sometimes used to derive the pH) is satisfied implicitly.

[22] It is assumed that water is associated with sea salt aerosol particles above their crystallization point rather than their deliquescence point because the particles are produced as droplets at the sea surface. Therefore the hysteresis effect ensures that, upon shrinking, the particles remain in a metastable highly concentrated solution state above their crystallization point, which is at about 45% relative humidity for NaCl [e.g., Tang, 1997; Pruppacher and Klett, 1997]. As long as the relative humidity in the MBL is well above the crystallization point, sea salt aerosol is not present in crystalline form. The crystallization humidity for many mixed aerosol particles containing sulfate or nitrate is below 40% relative humidity [Seinfeld and Pandis, 1998, and references therein], implying that aerosol particles that already had been involved in cloud cycles will also be in an aqueous metastable state. Therefore many soluble aerosol particles will be present in the atmosphere as metastable aqueous particles below their deliquescence humidity. Based on this we assume that sulfate aerosols are also liquified above their crystallization humidity. We assume a bulk crystallization humidity for the sulfate aerosol particles and do not include the formation of solid phases explicitly in the model. In the runs that are presented, this assumption has no impliciations, because the relative humidity never drops below the crystallization humidity. It does, however, drop below the deliquescence humidity.

[23] Aerosols are usually highly concentrated solutions. Laboratory measurements show that NaCl molalities can be in excess of 10 mol/kg [Tang, 1997] implying also very high ionic strengths. Therefore it is necessary to account for deviations from ideal behaviour, i.e., to include activity coefficients. We used the Pitzer formalism [Pitzer, 1991] to calculate the activity coefficients for the actual composition of each aqueous size bin. The Pitzer model in its original formulation is applicable to ionic strengths of 6 mol/kg. According to Luo et al. [1995], the Pitzer model can be extended to even stronger electrolyte solutions. The parameters used in our model were based on data over a wide range of particle composition. For example, for (NH4)2SO4 and H2SO4 the model is valid up to 40 mol/kg.

[24] In the atmosphere each aerosol or cloud particle is a closed “reaction chamber” with its own pH and concentrations of the species. In models the particles have to be lumped in bins where the individual properties of the particles vanish. In the model freshly emitted, alkaline sea salt particles are put into the same chemistry bin as aged (possibly acidic) particles leading to mean conditions and reaction paths that can be distinctly different from what is happening in the atmosphere. Especially in the case of high wind speeds, when sea salt aerosol production and therefore also sea salt aerosol loadings are high, the model-calculated pH is not representative for all particle sizes. This is due to the fact that the buffer capacity would then be determined by the large particles and reactions in the smaller, more acidic particles that involve acidity might be underestimated. In future versions of this model a finer resolution of the aqueous chemistry bins will be achieved.

[25] Photolysis rates are calculated online using the method of Landgraf and Crutzen [1998]. For the calculation of the actinic fluxes a four stream radiation code is used in addition to the two stream radiation code used for the determination of the net radiative flux density En because different spectral resolutions and accuracies are needed for these different purposes. Based on the findings of Ruggaber et al. [1997] a factor of 2 is applied to photolysis rates inside aqueous particles to account for the actinic flux enhancement inside the particles due to multiple scattering.

[26] Constant emission fluxes for the gases DMS and NH3 from the sea surface are applied with 2 × 109 molec/(cm2s) [Quinn et al., 1990] and 4 × 108 molec/ (cm2s) [Quinn et al., 1990], scaled to model their measured gas phase mixing ratios of 19 nmol/mol in clean air masses), respectively.

2.3. Model Resolution and Integration Scheme

[27] The atmosphere between the sea surface and 2000 m is divided into 150 layers. The lowest 100 layers have a constant grid height of 10 m, the layers above 1000 m are spaced logarithmically. For the “high boundary layer” run the grid height in the lowest 100 layers was increased to 20 m and the model domain was extended to 3500 m. The dynamical time step is 10 s and the chemical time step is chosen to differ between 1 ms for layers with low LWC associated to the particles or layers including freshly activated particles and 60 s for gas phase only layers. All chemical reactions in gas and aqueous phases, equilibria and phase transfer reactions are calculated as one coupled system using the kinetic preprocessor KPP [Damian-Iordache, 1996] (http://www.cs.mtu.edu/~asandu) which allows rapid change of the chemical mechanism without major changes in the source code.

[28] The very stiff chemical differential equation system is solved with a stable, mass-conserving second order Rosenbrock method (ROS2 by Verwer et al. [1997]) which was developed for atmospheric questions. We compared this solver off-line with the commercial integration system Facsimile Curtis and Sweetenham [1987] that uses a Gear solver. Using the complete set of reactions we yielded very good agreement between the different solvers.

3. Sensitivity Studies

3.1. Overview of the Runs

[29] Here we study only the cloud-free MBL whereas the cloudy MBL is discussed in a companion paper [von Glasow et al., 2002]. With the base run it was not intended to mimic a specific situation that has been observed in a field campaign, it is rather intended to present detailed process studies of the multiphase chemistry of the MBL under idealized conditions. The latitude chosen for this run is φ = 30°N at the end of July. The boundary layer height is roughly 700 m, moisture and heat fluxes from the sea surface are adjusted to yield a stable boundary layer. The relative humidity at the sea surface is roughly 65%, increasing with height to around 90% below the inversion that caps the MBL. As a result of the prescribed heat fluxes from the sea surface, the potential temperature in the well mixed MBL is slowly decreasing during the runs from Θ = 14.5 °C to 13.5 °C. The relative humidity in the different layers stays constant apart from a minor diurnal variation (±3%). After a spin-up of the dynamical part of the model for 2 days the complete model was integrated for 3 days (modelstart is at midnight).

[30] Many different sensitivity studies were performed and some of them are described here. See Table 1 for an overview of the runs.

Table 1. Overview of the Sensitivity Runsa
Name Gas Phase Initialization (See Table 2) MBL Height SST Differences From Base Run Main Results
base run remote 700 m 15 °C BrO vertical profile/diurnal variation, vertical profile of sea salt pH, recycling of HBr and HOBr by sulfate, “overall” effects of halogen chemistry
halogen off remote 700 m 15 °C halogen chemistry switched off overall importance of halogen chemistry shown
aerosol off remote 700 m 15 °C halogen and aerosol chemistry switched off overall importance of aerosol chemistry shown
low O3 low SO2 remote, but O3 = 12 nmol/mol, SO2 = 20 pmol/mol 700 m 15 °C gas phase initialization low O3 reduces Br activation, initial mixing ratios of SO2 unimportant for steady state conditions
low O3 low SO2 large sulfate surface same as “low O3 low SO2 700 m 15 °C sulfate aerosol mode radius increased by factor 1.5 recycling of HBr, HOBr by sulfate aerosol is important
continental influence continentally influenced 700 m 15 °C gas phase initialization quicker Br activation
winter remote 750 m 5 °C solar declination −20° instead of 20° higher BrO mixing ratios than in summer
spring remote 700 m 10 °C solar declination 0° instead of 20° similar to winter
winter Cape Grim remote, see text 550 m 9 °C conditions for “typical” winter at Cape Grim importance of acidity for release of Br
high BL remote 1300 m 15 °C inversion at 1300 m instead of 750 m, vertical resolution 20 m importance of recycling by sulfate aerosol, less strong Brx activation
strong wind remote 700 m 15 °C u10 ≈ 9 m/s (instead of 6 m/s), sea salt aerosol flux param. by Smith et al. [1993] strong increase in S(IV) oxidation in sea salt particles
carbonate remote 700 m 15 °C carbonate buffer in sea salt increased by 50% [Sievering et al., 1999] changes in importance of S(IV) oxidation paths
iodine (iod1/iod2) remote 700 m 15 °C iodine chemistry included quicker Br activation and greater O3 loss
  • a SST is the sea surface temperature.

[31] Tables 2 and 3 list the initial gas phase mixing ratios and parameters for the initial lognormal aerosol size distribution. Aerosol particles with dry radii less than r = 0.5 μm are assumed to be a mixture of 32% (NH4)2SO4, 64% NH4HSO4 and 4% NH4NO3 [Kim et al., 1995]. The resulting pH of the sulfate aerosols is between 0.5 and 1 (and even lower for the “continentally influenced” run) which is in the range estimated by Fridlind and Jacobson [2000] for sulfate aerosol sampled during the ACE-1 campaign. A run where the initial composition of the sulfate aerosol was assumed to be pure (NH4)2SO4 showed a pH between 2.5 shortly after model start and 1 after 3 days due to the uptake of acids from the gas phase. Consequences for the gas phase were very small, NH3 increased because of reduced uptake by the sulfate aerosol whereas HCl decreased due to uptake by the sulfate aerosol. In most acid dependent reactions changes in the sulfate pH in this very acidic range have no effect. Based on this we conclude, that for the most relevant situations, the results do not depend significantly on the initial chemical composition of the sulfate aerosol.

Table 2. Initial Mixing Ratios of Gas Phase Speciesa
Species Remote (MBL) Remote (FT) Cont. Infl. (MBL) Cont. Infl. (FT)
CO 70.0 150.0
NO2 0.02 0.03 0.5
HNO3 0.01 0.05 0.1
NH3 0.08 0.2
SO2 0.09 1.0
O3 20.0 50.0 50.0 70.0
CH4 1800.0 1800.0
C2H6 0.5 5.0
HCHO 0.3 0.3
H2O2 0.6 0.8
PAN 0.01 0.1 0.1 1.0
HCl 0.04 0.04
DMS 0.06 0.06
CH3I 0.002
C3H7I 0.001
  • a Units are in nmol/mol. Values are for the two scenarios “remote” and “continentally influenced.” A value for the free troposphere (FT) is given only when it is different from the MBL value. CH3I and C3H7I are accounted for only in the “iodine” run.
Table 3. Initial Size Distribution of the Aerosola
Mode i Ntot, i, 1/cm3 RN, i, μm σi
1 100 0.027 1.778
2 120 0.105 1.294
3 6 0.12 2.818

[32] Particles larger than r = 0.5 μm are assumed to be sea salt particles. The composition of sea salt aerosol in the model is simplified by assuming all (unreactive) cations to be Na+. The anions considered are Cl, Br, HCO3, and for the iodine runs also I and IO3. Sulfate is not considered as initial component of sea salt aerosol as it is also assumed to be chemically unreactive. Therefore all sulfate that is present in sea salt aerosol in the model runs stems from uptake from the gas phase and can be labeled “non-sea salt sulfate.” Upon model initialization “fresh” sea salt with a pH of about 8—which is roughly the pH of ocean water [Riley and Skirrow, 1965]—is present everywhere in the MBL. Uptake of acidic gases like HNO3 from the gas phase rapidly acidifies the sea salt aerosol particles. Fresh sea salt aerosol particles emitted from the sea surface also have a pH around 8.

3.2. Halogen Activation

[33] The model runs start at midnight. Figure 2 shows the evolution with time of major gas phase species for the base run, the “continental influence” run and the “low O3 low SO2” run. Shortly after model start the sea salt aerosol is acidified due to rapid uptake of acids from the gas phase with pH values ranging between 6 near the sea surface and less than 3.5 at the top of the MBL (see section 3.9 for a discussion of the vertical profile of the sea salt aerosol pH).

Details are in the caption following the image
Evolution with time of the main gas phase species for the base run (solid line), the “continental influence” run (dotted line) and the “low O3 low SO2” run (dashed line) (in 50 m). Brx is the sum of all gas phase bromine species except HBr. The time is given in hours from model start.

[34] The halogen chemistry is started by reactions that transform Br into species that can degas from the aerosol and initiate quick reaction cycles in the gas phase. These starter reactions are reactions of Br with OH, HSO5 and NO3.

[35] The reaction cycles responsible in the model for the degassing of the reactive bromine species (Br2 and BrCl) involve the following steps as discussed by Sander and Crutzen [1996] and Vogt et al. [1996].
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0012
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0013
Depending on the concentrations of the species involved in the aqueous phase equilibria, BrClaq either degasses (when significant Br depletion has occurred) and is then photolysed:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0014
or takes part in the following autocatalytic cycles (when sufficient Br is available) leading to the production of Br2:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0015
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0016
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0017
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0018
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0019
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0020
Or, instead of equilibria (I.2), (I.4) and (I.5):
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0021
leading in both cases to the following net reaction:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0022
This autocatalytic bromine activation cycle consumes H+ ions which implies the need for acidification of the sea salt aerosol as shown by Fickert et al. [1999] in laboratory experiments.

[36] Keene et al. [1998] studied the influence of the pH on halogen activation. They showed that significant sea salt dehalogenation is limited to acidified aerosol but that differences between a pH of 5.5 and 3 are not significant.

[37] Sander et al. [1999] showed that when sufficient NOx is available, reactions on the surface of aerosols involving BrNO3 can also lead to the production of Br2 and BrCl without the need for acid catalysis.

[38] These reaction cycles lead to gradual build-up of total reactive bromine Brx (the sum of all gas phase bromine species except HBr) in the gas phase with maximum values between 8 and 10 pmol/mol in the model runs. During daylight the main Br species are HOBr, HBr and BrO, during night this is shifted towards Br2 and BrCl which are photolyzed readily during daylight (see Figure 2). Model calculated bromide deficits are between 70 and 85% in 50 m height, rising to more than 90% in higher model layers due to the lower sea salt aerosol pH in that layers (see section 3.9).

[39] The value of the gas phase mixing ratio of O3 is important for these cycles because the production of BrO and therefore also the O3 destruction is catalyzed by O3 (reaction (I.8)). If O3 mixing ratios are small, halogen activation is slowed down (see run “low O3 low SO2”).

[40] The order of magnitude of total inorganic gas phase bromine (Brt = Brx + HBr) in the lowermost model layers is in the same range as measurements by Rancher and Kritz [1980] that were made aboard a ship in 8 m height during a cruise off the equatorial African coast.

[41] In an additional run we switched off all bromine related reactions. The development of the gas phase chemistry is very similar to the “halogen-off” run where both chlorine and bromine chemistry reactions are switched off. Furthermore, the mixing ratio of Clx is very small. This shows that in the model significant chlorine activation depends on interactions with bromine.

3.3. Active Recycling of Brx by the Sulfate Aerosol

[42] The presence of sulfate aerosol particles is important in the model for the halogen activation cycle. The large surface area of the sulfate aerosol of 40 to 60 μm2/cm3, compared to the sea salt surface area of 25 to 45 μm2/cm3, is the reason that HBr, HOBr and BrNO3 are also scavenged to a significant amount by the sulfate aerosol and are involved in reactions on and in sulfate particles. Due to the low pH of the sulfate aerosol the acid-catalyzed cycle (I) quickly transforms Br and Cl (from the scavenged HBr and HCl) and HOBr to BrCl and Br2 which rapidly degas. As can be seen from Figure 3 the magnitude of recycling of bromine species is higher than in the sea salt aerosol by roughly a factor of 5 to 10. Due to this recycling process the sulfate aerosol is not a significant sink for gas phase bromine in the model; some bromine, however, will always be found in the sulfate aerosol.

Details are in the caption following the image
Net exchange of Brx and HBr between sea salt and sulfate aerosol and gas phase (in 215 m). A negative sign implies loss for the aerosol. The flux of Br atoms is plotted, i.e. the flux for the molecule Br2 is multiplied with 2 to get the number of Br atoms. The solid line is the net flux of Br atoms. The time is given in hours from model start. The nighttime hours are shaded.

[43] Particle composition measurements [e.g., Moyers and Duce, 1972; Duce et al., 1983] often showed an increase of the bromine to sodium ratio in small particles compared to sea water. The interpretation of these data is difficult because for some samples the chemical state of bromine in the particles is unclear. Furthermore, sodium in the aerosol samples could be a consequence of the sampling of an external mixture of sulfate and sea salt particles or of internally mixed particles that originate from collision-coalescence or cloud processes (or both). Therefore the use of the Br/Na+ ratio could be misleading. The enriched bromine could be in the form of stable bromine-containing ions or organic molecules. Known reactions [Haag and Hoigné, 1983] for the production of stable ions like BrO3 in the aqueous phase are, however, too slow to lead to significant production in the aerosol.

3.4. Nighttime Fluxes of Bromine

[44] During daylight, photolysis is the driving force for the net transport of bromine from the aqueous to the gas phase as formation of HOBr is dependent on photolytically produced HO2. As the photolysis slows down during sunset, Br2 and BrCl accumulate in the gas phase and the remaining HOBr is rapidly recycled by the aerosol to Br2 and BrCl. During night Br2 also degasses as evident from Figure 3. This is caused by a sudden rise in BrCl mixing ratios after sunset due to absent photolytical loss. Thus the gas-aqueous partitioning of these species is perturbed leading to the reaction sequence:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0023
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0024
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0025
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0026
where the net direction of the equilibria is indicated with the arrow.
[45] This cycle results in net transfer of Br from the aqueous to the gas phase:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0010
If the bromide depletion in the sea salt aerosol is close to complete, sequence (II) is reversed (cycle (−II), in agreement with laboratory experiments of Fickert et al. [1999]). The Br produced in this reversed reaction cycle can then take part in a reaction cycle that involves Cl2 and is similar to cycle (II), leading to a shift of halogen species:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0027
Neither cycle (−II) nor cycle (III) are a net sink for total gas phase bromine, they just shift gas phase bromine from Br2 to BrCl which is what is observed both in the model for the strongly depleted (e.g., “continental influence run”) cases and in the laboratory [Fickert et al., 1999].

3.5. Diurnal Variation of BrO

[46] The model predicts a distinct diurnal variation of BrO and BrNO3, showing a minimum around noon and maxima in the morning and evening (see Figure 2). This feature can be explained by differences in the photolysis spectra of O3 and Br2 and BrCl. Br2 and BrCl are more rapidly photolysed in twilight than O3 due to absorption at longer wavelengths.

[47] At sunrise (high solar zenith angle) the solar spectrum in the lower troposphere is shifted to longer wavelengths. This causes an earlier start of the reaction cycles that include Br2 and BrCl photolysis compared to cycles that rely on O3 photolysis (see Figure 4). It also causes a delay in the photolytically initiated production of OH and HO2 relative to the production of Br. Since HO2 is the main sink for BrO (reaction (I.9)) apart from photolysis, this leads to an increase in BrO mixing ratios (producing the morning peak). Later during the day HO2 mixing ratios are sufficiently high to reduce BrO mixing ratios leading to the noon minimum. In the late afternoon a similar mechanism as in the morning leads to the evening peak of BrO.

Details are in the caption following the image
Diurnal variation of the photolysis frequencies for O3 and Br2 (note the different units). The difference in the shape of the diurnal variation is the cause for the noon minimum in BrO mixing ratios (see text).

[48] For situations with very high BrO mixing ratios like the ones encountered at the Dead Sea [Hebestreit et al., 1999] the minimum around noon would not occur because then HO2 mixing ratios are too small throughout the day to reduce BrO significantly.

[49] The diurnal variation of BrO is accompanied by an early morning peak in O3 destruction by XO + YO and XO + NO2 reactions (X, Y = Br, Cl). Early morning O3 destruction in our base run is 27% of the maximum O3 destruction at noon, compared to 4% in the “halogen off” run.

[50] This could explain recently published measurements of Nagao et al. [1999], who found a “sunrise ozone destruction” (SOD) in the diurnal variation of O3 based on measurements over a 3-year period on an island in the sub-tropical Northwestern Pacific. SOD acts in addition to the conventional O3 destruction that has its maximum around noon. They already speculated that bromine spieces might play a role in SOD. A similar feature was found by Galbally et al. [2000] in a time series of 13 years of O3 observations at Cape Grim, Tasmania.

3.6. Effect on O3

[51] The net effect of photochemistry in the MBL is to destroy O3. Dynamics, especially the downmixing of O3-rich air from the free troposphere is thought to balance this destruction [e.g., Monks et al., 2000]. Here we show in agreement with Vogt et al. [1996] that halogen radicals can significantly contribute to photochemical O3 loss in the MBL.

[52] To properly asses the effect of chemical reactions on O3, the odd oxygen family Ox has been defined (based on Crutzen and Schmailzl [1983]): Ox = O3 + O + O(1D) + NO2 + 2 NO3 + 3 N2O5 + HNO4 + ClO + 2 Cl2O2 + 2 OClO + BrO + IO + 2 I2O2. Species in this family are readily transfered to O3 (e.g., NO2 which photolyses forming O which recombines with O2 to O3). Only if a reaction leads to a change in the mixing ratio of the odd oxygen family (ΔOx ≠ 0) an effective O3 destruction or production takes place.

[53] Chemical Ox production in the MBL is primarily by reaction of NO with HO2 and CH3OO, both of which form NO2. In the base run it sums up to 1.9 nmol/(mol day).

[54] Table 4 shows the contribution of the most important reactions to O3 destruction. The most important Ox loss reaction is O(1D) + H2O which accounts for 52% of total O3 destruction in the base run followed by O3 + HO2 with 13%. The main step in bromine-related photochemical O3 destruction is the production of BrO in reaction (I.8) followed by reactions of BrO with HO2, CH3OO, XO (X = Br, Cl) radicals, and NO2. In the base run the halogen-related O3 destruction reactions account for roughly 30% of total O3 destruction.

Table 4. Ox Destruction Ratesa
Run Net Ox Destruction, equation image Net Ox Destruction by Halogens, equation image O(1D) + H2O, % O3 + HO2, % O3 + OH, % ClO + HO2/NO2/CH3OO, % BrO + HO2, % BrO + ClO, % BrO + BrO, % BrO + CH3OO, % DMS + BrO, %
base run 1.9 0.6 52.0 12.6 5.3 3.7 12.4 2.7 1.3 6.6 2.5
halogen off 0.7 70.1 20.8 7.9
aerosol off 0.7 70.0 21.0 7.9
low O3 low SO2 0.6 0.3 56.4 14.0 4.7 1.7 11.0 0.9 0.8 4.0 3.2
low O3 low SO2 sulf 1.0 0.5 49.1 11.1 3.8 3.3 15.0 3.1 2.0 7.0 4.0
cont. influence 3.7 1.2 47.4 15.9 8.7 5.4 8.3 2.6 0.7 4.1 0.7
winter 0.8 0.6 18.0 10.5 2.9 7.3 20.0 18.0 7.1 10.0 10.2
spring 1.1 0.5 39.2 14.5 5.0 4.8 16.0 5.0 2.3 7.1 4.9
high BL (in 130 m) 2.9 0.4 67.0 13.4 7.2 1.7 5.4 0.4 0.3 3.4 0.5
strong wind 2.0 0.8 49.3 12.4 5.5 1.9 14.7 1.8 2.6 7.8 2.6
carbonate 1.9 0.6 52.1 12.7 5.4 3.7 12.3 2.6 1.3 6.5 2.5
iod 1 2.2 0.9 44.4 9.9 4.5 5.0 12.7 2.1 1.7 7.8 2.4
iod 2 2.7 1.5 34.8 7.2 3.6 4.1 9.2 1.6 1.3 6.7 1.9
  • a Net chemical Ox destruction and net chemical Ox destruction involving halogens (first two columns) are given in nmol/(mol day), the contribution of the different reactions to the total Ox destruction is expressed in %. For each reaction the appropriate ΔOx has been used as a factor to evaluate the contribution. The most important Ox destruction paths involving iodine species are IO + HO2: 4.6 (12.4) % and IO + BrO: 3.1 (11.8) % in the “iod 1” (“iod 2”) run. The data are given as averages of the second and third day in 65 m height. Only in the “high BL” run for technical reasons a height of 130 m was chosen.

[55] The most important bromine reaction (BrO + HO2) causes 12% loss. Chlorine related O3 destruction is less important than the bromine reactions, it accounts for roughly 4%.

[56] In the “continental influence” run the relative contribution of ClO and BrO changed strongly which is due to increased release of chlorine (see section 3.8). Furthermore, there is a strong increase in Ox destruction by HO2 and OH.

[57] The runs “halogen off” and “aerosol off” show the same net chemical Ox destruction (Table 4). Comparison with the base run shows a difference in net chemical Ox destruction of 1.2 nmol/(mol day) or 170%. This is twice as much as directly attributable to halogen related Ox destruction. The difference is due to indirect effects of halogen chemistry, namely, the reduction of NOx (see section 3.8) and thereby the reduction of chemical Ox production.

[58] Very interesting is the high relative contribution of halogen chemistry to O3 destruction in the winter run. The reason for this will be discussed in the next section.

3.7. Winter Run

[59] For the winter run, temperature and solar zenith angle are chosen to model the conditions at the end of January instead of July as in the base run (see Figures 5 and 6). This implies that the solar spectrum is shifted to greater wavelengths reducing the photolysis rates for O3 → O(1D) by 66% leading to a decreased formation of OH and HO2 compared to summer conditions. The reduction in the photolysis rates of Br2 and BrCl are only 7.5 and 15%, respectively, because they absorb at greater wavelengths.

Details are in the caption following the image
Evolution with time of the main gas phase species for the base run (solid line) and for the winter run (dotted line) (in 50 m). The time is given in hours from model start.
Details are in the caption following the image
Vertical profile of the main gas phase species for the base run (solid line) and for the winter run (dotted line) at noon of third day. Note that the height of the MBL is slightly different in the two runs.

[60] In the “winter” run net O3 destruction is only 32% of the destruction in the base (= summer) run (see Table 4) because of the smaller solar radiation intensity reducing OH to a third and HO2 to 50% of the summer values. Due to the reduction in HO2, which is the main sink for BrO and ClO, the mixing ratios of these radicals increase compared to the summer run. The absolute rate of halogen-related O3 loss is approximately the same as in the base run enhancing the relative halogen contribution to 67%. This is mainly due to an increase in the BrO and ClO mixing ratios, leading to an increased importance of the XO + YO (X, Y = Br, Cl) reactions, which have a ΔOx = −2 (see Table 4).

[61] The morning peak in the BrO mixing ratios calculated by the model has increased by approximately 20%, whereas the noon values increased by more than 90%. Activation of bromine from the aqueous phase is faster and total reactive bromine (Brx) increased by 12 to 25% compared to the summer run at the expense of HBr which decreased by 1 pmol/mol (roughly 50%). Also Clx (the sum of all gas phase chlorine species except HCl) increased by up to 30% (see Figure 5).

[62] Due to lower temperatures compared to the summer case, the solubility of the gases increases. A reduction in the temperature from 15 °C to 0 °C increases the solubility of SO2 by a factor of 1.8, HNO3 by a factor of 3.8, and HCl by a factor of 4. Enhanced uptake of acidic gases leads to a lower sea salt aerosol pH. Caused by the lower pH, bromine activation is more rapid and bromine depletion even stronger than in the base run.

[63] Comparison with the measurements of Ayers et al. [1999] shows a difference in the seasonality as they found a minimum in the Br deficit in winter. During winter the availability of acids (mainly due to reduced microbial activity which causes a smaller DMS flux) is reduced at their measurement site whereas sea salt loadings are increased due to higher wind speeds in winter. In the model run the wind speed (which controls the flux of sea salt aerosol), gas phase acid mixing ratios or the DMS flux were not changed in order to concentrate on differences in radiation and temperature only. Therefore these differences between measurements and model results were to be expected.

[64] To test the model under the conditions described by Ayers et al. [1999] another run for winter conditions was performed (“winter Cape Grim”), where the geostrophic wind speed was increased from 7 to 12 m/s yielding a higher sea salt flux. Furthermore the fluxes of DMS and NH3 were decreased by a factor of 4 to get typical mixing ratios of 20 pmol/mol for DMS and less than 5 pmol/mol for SO2 [Ayers et al., 1997a]. Other chemical initial values and the meteorology were adjusted to the conditions at Cape Grim [Ayers et al., 1997b]. Very little bromine activation was found. The depletion was even less than in the measurements of Ayers et al. [1999]. This possibly points to a source of acidity that was not considered in the model. Alternatively this can be caused by the fact that in the model only a mean aerosol pH is calculated, whereas in reality the pH strongly depends on the particle age, i.e., in the atmosphere particles of the same size but with different pH coexist. Due to the reduction of the DMS flux and the resulting decrease of SO2 mixing ratios, the major source for aqueous phase acidity was strongly reduced, whereas the aqueous phase buffer capacity was increased by the increase in sea salt particle mass (roughly a factor of 3). This lead to sea salt aerosol pH values greater than 8 (the value for sea water). When initial gas phase concentrations and fluxes were chosen for summer conditions at Cape Grim [Ayers et al., 1997a, 1997b, 1999], strong bromine depletion occurred as in the other runs that were presented here.

[65] The relative importance of these two opposing effects, i.e., increase in halogen chemistry relevance caused by a shift of the radiation spectrum to higher wavelengths vs. decrease in halogen chemistry relevance due to greater sea salt fluxes and lower gas phase acidity, will be different for every site and has to be assessed individually. The features for the “spring” run (not shown) are the same as for the “winter run” but the reduction in total net O3 destruction to 58% of the value in the base run is not as strong as in the “winter” run.

3.8. Influence of NOx

[66] Nitrogen oxides play a very important role in atmospheric chemistry by acting as catalyst in the photochemical production of O3. When NO2 reacts with OH, nitric acid (HNO3) is formed which is rapidly taken up by aerosol (mainly sea salt aerosol), providing acidity to the aerosol. If halogens are present, NOx (= NO + NO2) mixing ratios are further reduced by reactions of NO2 with ClO and BrO yielding XNO3 (X = Cl, Br). XNO3 can either be photolysed, thermally decomposed or taken up by the aerosol leading to a loss of reactive nitrogen from the gas phase and to an increase of NO3 in the particles and further activation of bromine [Sander et al., 1999]. The scavening of XNO3 by particles is a significant NOx loss process, so a NOx source was introduced in the model to avoid unrealistically small mixing ratios of less than 0.1 pmol/mol. The source strength is 10 pmol/(mol day) in the MBL which is equivalent to 2 × 108 molec/(cm2 s) as surface flux. Monks et al. [1998] report typical mixing ratios of 2 to 6 pmol/mol for NOx for clean maritime background conditions (Cape Grim) which are achieved in the model with the NOx source.

[67] This additional NOx could be a photolysis product from alkylnitrates that are emitted from the ocean (see e.g., Blake et al. [1999, and references therein] for latitudinal and vertical profiles during the ACE-1 campaign). Another possibility could be the direct emission of NO from the oceans (formed by photolysis of NO2) as suggested by Zafiriou et al. [1980] on the basis of NO measurements in surface waters of the central equatorial Pacific.

[68] Crutzen [1979] pointed to the importance of downward transport of PAN from the free troposphere (where it is long-lived) and the subsequent thermolysis in the MBL. This process is included in the model but exchange between the MBL and the free troposphere is weak as long as no clouds are present that strongly increase this exchange.

[69] On the other hand the need to include a NOx source for runs including aerosol/halogen chemistry might point to the fact, that processes exist that convert aqueous HNO3/NO3 back to NOx, that are not included in the model. The importance of the photolysis of NO3 in aerosol particles was also studied [Warneck and Wurzinger, 1988; Zellner et al., 1990]:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0028
Although this reaction was found to be negligible as a source of NO2, OHaq concentrations increased significantly speeding up the activation of bromine from the sea salt aerosol because OH is involved in one of the starter reactions (see also section 3.2). The resulting differences in the gas phase were only apparent on the first model day.

[70] The influence of NOx on halogen chemistry can be shown when comparing the base with the “continentally influenced” run. In the continentally influenced run the initially high mixing ratios of NO2 speed up the bromine activation and lower the sea salt aerosol pH through the uptake of HNO3. This causes a very quick depletion of bromide from the sea salt aerosol favoring the production of BrCl in the sea salt aerosol instead of Br2 which leads to a change in the speciation of gas phase Brx (Br2 mixing ratios are reduced and BrCl enhanced, see discussion above in section 3.4). Furthermore chlorine (both reactive chlorine, Clx and HCl) release is strongly enhanced. This results in increases of about 30–70% in ClO and the shift in Ox destruction rates as discussed above in section 3.6.

[71] If NOx mixing ratios are small (base run) the morning peak of BrO is quite large (up to 4.5 pmol/mol in 50 m height) because destruction of BrO by reaction with HO2 or NO2 is not efficient. In polluted air, however, there is enough NOx to make the reaction of BrO with NO2 important. Then a morning and afternoon peak of BrNO3 instead of BrO is predicted by the model. In the “continentally influenced” run this occurs only on the first day, because later on NOx mixing ratios are too small.

3.9. Variation of Sea Salt Aerosol pH and BrO With Height

[72] The calculated pH in the sea salt aerosol bin is highest at the sea surface where alkaline particles are emitted and it decreases towards the top of the MBL. The vertical decrease in aerosol pH is associated with increasing relative humidity and aerosol water content (see Figure 7). This feature is caused by the high salt loading of the aerosol and exchange of HCl between the aerosol and gas phases due to changes in LWC and the aqueous fraction of HCl. A detailed explanation is given by von Glasow and Sander [2001].

Details are in the caption following the image
Vertical profile of the sea salt aerosol pH and liquid water content (LWC) for the base run at noon of third day.

[73] An implication of the vertical gradient of the sea salt particle pH is that acid-catalyzed bromine activation is more efficient in higher layers of the MBL. The vertical profiles of both Brx and BrO show a maximum directly below the temperature inversion that caps the MBL (see Figure 6). In layers close to the sea surface with lower sea salt aerosol water content and high sea salt pH this leads to smaller BrO mixing ratios. The higher mixing ratios of, e.g., BrO at the top of the MBL might explain difficulties to detect BrO in surface measurements in the MBL. Based on this finding it would be desirable to modify measurement instruments for the detection of halogen radicals such that information on the vertical distribution in the MBL can be obtained. As a consequence of the maximum in BrO mixing ratios at the top of the MBL, O3 destruction rates by bromine radicals increase with height in the MBL. If significant detrainment of MBL air into the free troposphere occurs this could lead to enhanced BrO and Br mixing ratios in this region.

3.10. Iodine Chemistry

[74] The chemistry of iodine in the MBL has been the subject of some studies in the last decades (see Vogt et al. [1999] for an overview). In recent years it received even more attention due to the detection of IO in the MBL [Alicke et al., 1999; Allan et al., 2000]. Maximum mixing ratios measured by Alicke et al. [1999] were 6 pmol/mol at Mace Head, Ireland. Allan et al. [2000] measured IO at Tenerife and Cape Grim, Tasmania with a mean mixing ratio of about 1 pmol/mol. These data suggest a rather widespread presence of IO although it could not be measured on all days of the measurement campaigns. McFiggans et al. [2000] used a box model that was constrained with data from measurements and could show the importance of the cycling of iodine species on aerosol particles for the Mace Head and Tenerife measurements. They used a modified version of the mechanism suggested by Vogt et al. [1999].

[75] In recent field campaigns OIO was also measured in the MBL at Mace Head [Hebestreit et al., 2000] and Tasmania [Allan et al., 2001]. OIO was first detected in the laboratory by Himmelmann et al. [1996]. It is known to be formed in the self reaction of IO and in the reaction of IO with BrO [Bedjanian et al., 1998; Misra and Marshall, 1998]. Further reactions of OIO are highly uncertain. Ingham et al. [2000] studied the photolysis of OIO and found that OIO is photochemically very stable as already calculated by Misra and Marshall [1998]. Cox et al. [1999] and Hoffmann et al. [2001] proposed that OIO could be an intermediate in the formation of particulate iodate.

[76] Particulate iodine is found in marine aerosol samples 100 to 1000 times enriched compared to sea water but it is not clear in which chemical state the iodine is present. Some studies reported it to be IO3 but in others no IO3 was found (see references and discussion in the work of McFiggans et al. [2000]. Based on measurements of aerosol composition, Baker et al. [2000] state that “iodine is present in aerosol in varying proportions as soluble inorganic iodine, soluble organic iodine and insoluble, or unextractable, iodine.”

[77] In contrast to gas phase chlorine and bromine that is mainly released by sea salt particles, iodine compounds originate from biogenic alkyl iodides released by various types of macroalgae and phytoplankton in the sea. Photodissociation then produces I radicals that mainly react with O3 to IO. We assumed the following emission rates for: CH2I2, CH2ClI, CH3I, and iC3H7I of 3.0 × 107 cm−2 s−1, 6.0 × 107 cm−2 s−1, 0.6 × 107 cm−2 s−1, and 1.0 × 107 cm−2 s−1, respectively.

[78] OIO peaks were not measured simultaneously with IO but rather late in the afternoon and after sunset [Hebestreit et al., 2000; Allan et al., 2001]. Four different stages in the diurnal evolution can be distinguished: 1) During day the formation of OIO must not be too much at the expense of IO, because otherwise IO peaks would become too small. There should be strong destruction paths to keep OIO at a reduced level. 2) During dusk the production of OIO wins over its destruction, pointing to the need of photolytic processes for the destruction (e.g., reaction with OH, NO). 3) During night there is a slow sink, which might be reaction with a “nighttime species” like NO3, uptake on aerosol particles or deposition to the surface. 4) During dawn no OIO was observed, pointing to differences to the chemistry at dusk, when OIO increases.

[79] Kinetic data on the formation and reactions of OIO are extremely scarce, however, so that we can only speculate on the chemistry that is occurring. Apart from the aforementioned formation reactions for OIO, we assumed reaction of OIO with NO to yield NO2 and IO. It was assumed to be 5 times faster than the reaction of OBrO with NO [Li and Tao, 1999], because the OBrO reaction is also 5 times faster than the OClO reaction [DeMore et al., 1997]. We further included the reaction of OIO with OH. In contrast to Plane et al. [2001] we assumed two branches of 50% yield each: HIO3 and HOI + O2. With this it is avoided that all OIO that reacts with OH ends up as IO3 in the particles, allowing some recycling of I via the particle phase. We assumed a rate constant of 2 × 10−10 cm−3/(molec s). The absolute value of the accommodation coefficient α(HIO3) is not important, as a smaller value only increases the gas phase mixing ratio of the unreactive HIO3. We chose α(HIO3) = 0.01. Knight and Crowley [2001] could not detect HIO3 as a product from the reaction IO + HO2. If, however, HIO3 would be a minor product of this reaction, than it might compete with OH + OIO due to the high HO2/OH ratio, which is in the order of 100–1000. To achieve a decrease of OIO during night uptake by the aqueous phase was included using α(OIO) = 0.01.

[80] It was recently found in the laboratory, that uptake of HOI on sea salt solution yields not only ICl but also IBr which degas [Holmes et al., 2001]. ICl is only released when bromide has been depleted in the aqueous phase. This points to interhalogen reactions/equilibria in the aqueous phase, that we implemented in analogy to the respective equilibria for bromine:
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0029
urn:x-wiley:01480227:media:jgrd9118:jgrd9118-math-0030
As no data are available on the above equilibria we assumed the ratio to be the same as measured for the BrCl, Br2, Br2Cl equilibria system [Wang et al., 1994] and varied it. The gas phase ratio of ICl to IBr depends on the aqueous phase ratio of Cl/Br and on the ratio of the equilibrium constants, as already pointed out by Fickert et al. [1999] for the BrCl, Br2, Br2Cl equilibria system. The remainder of the iodine reaction mechanism is as described by Vogt et al. [1999].

[81] We compare a model run where these assumptions are included (run iod1) with a run where the OIO self reaction (producing I2O2, k = 5 × 10−11 cm3/(molec s)) was added to yield higher IO mixing ratios (run iod2) and a base run without iodine chemistry.

[82] The amount of particulate iodine formed in the model, 12 (iod1) and 10 (iod2) pmol/mol in sulfate and 3.5 (iod1) and 3 (iod2) pmol/mol in sea salt particles after 3 days, is up to a factor of 10 higher than in measurements. This might imply that HIO3 formation by the reaction OIO + OH is still overestimated, because HIO3 is the main precursor for particulate iodine in the model.

[83] As already discussed by Vogt et al. [1999] the activation of chlorine and bromine chemistry is accelerated by iodine chemistry due to the reaction of HOI in the aqueous phase with Cl and Br producing ICl and IBr which escape to the gas phase, photolyze and produce halogen radicals that continue the reaction cycles discussed above. Figure 8 shows that the activation of chlorine and bromine is faster than in the base run and that the mixing ratios after 3 days are about 10 and 15% higher for Brx and Clx. Total inorganic iodine consists mainly of HOI, IO and OIO during day and of ICl and IBr during night. Maxima of ICl and IBr mixing ratios are below the inversion at the top of the MBL, whereas maxima of IO and OIO are found near the sea surface.

Details are in the caption following the image
Evolution with time of the main gas phase species for the base run (solid line), for the iod1 run (dotted line), and for the iod2 run (dashed line) (in 50 m). The time is given in hours from model start.

[84] The maximum mixing ratios of IO in the model are about 0.5 pmol/mol for the run without OIO self reaction and 1.0 to 1.5 pmol/mol with the self reaction. This and the above mentioned very high particulate iodine values suggest that rapid cycling of OIO and possibly also of the reaction products (HIO3?) are important. Inclusion of iodine chemistry increases chemical net O3 destruction by 15% to 40% compared to the base run.

[85] In the model, the reaction of IO with BrO is the major source for OIO. Contrary to field observations, OIO is already produced at sunrise. The morning peak is even greater in the model than the afternoon/night peak. It is due to early morning photolysis in the visible of iodine containing molecules that produce IO. The observed differences in chemical processes between dusk and dawn are not captured by the reaction mechanism that we used. The slow OIO destruction during night is captured by the model, but in the run including the OIO self reaction significant amounts of IO are produced during the night from the thermolysis of I2O2. This is also not observed in the field. It has to be concluded that the atmospheric chemistry of iodine and especially that of OIO is not yet understood and that our speculations based on preliminary data are not capable of reproducing the features of field measurements. The inclusion of reactions of OIO with IO, Br, or photolysis did not change these conclusions. It is obvious that more kinetic studies are urgently needed.

4. Conclusions

[86] Using a much more elaborate one-dimensional model of the MBL, results from earlier box model studies for cloud-free conditions could be confirmed. These are the activation of bromine from sea salt aerosol by an acid-catalyzed mechanism that follows starter reactions in the sea salt aerosol. These transform the rather unreactive Br to reactive substances like HOBraq and Braq. The importance of catalytic gas phase reaction cycles that involve bromine for the destruction of O3 was also confirmed. Roughly 30% of the total chemical O3 destruction in the model was due to halogen reactions. Furthermore the importance of iodine chemistry in the MBL for the activation of Br and Cl from the sea salt aerosol and in the destruction of O3 was confirmed and speculations on processes involving the OIO radical which has recently been detected in the MBL were made.

[87] Additionally it could be shown that under low NOx situations that prevail in the remote MBL the diurnal variation of BrO and ClO (= XO) show a maximum in the early morning and late afternoon and a local minimum around noon. This diurnal variation was also found in species that are produced from XO (e.g. XNO3). The reason for this behaviour is the different wavelength dependence of the photolytic source of the precursors of BrO and its most important reaction partner HO2. The early morning BrO peaks could be the reason for “sunrise ozone destruction” that has been observed in field measurements.

[88] The model further showed a distinct vertical profile of BrO with maxima below the inversion that caps the MBL. This is linked to the vertical profile of sea salt aerosol pH which has its minimum below the inversion where relative humidity is highest.

[89] In the simulations the presence of sulfate aerosol particles was important for the recycling of less reactive bromine species like HBr and HOBr to more reactive species like Br2 and BrCl. As this cycling is quick the sulfate aerosol was not found to be an important sink for bromine in the model.

[90] If the availability of acids is restricted, bromine activation is reduced in the model. Bromine activation is also slowed down in low O3 regions, because the reaction of Br with O3 is an important step in the bromine activation mechanism. On the other hand bromine activation can be accelerated in more polluted regions.

[91] As a consequence of the shift in the tropospheric radiation spectrum at high solar zenith angles the mixing ratios of BrO and ClO were predicted to be higher in winter than in summer and the relative importance of halogen reactions in O3 destruction and DMS oxidation increased significantly. This is true only for comparable availability of gas phase acidity and similar sea salt aerosol fluxes. From the model results large variations in halogen concentrations are expected for different situations.

[92] The model results showed the existence of vertical gradients of gas and aqueous phase chemical species. Many measurements of aerosol composition, etc. are done at the surface or from ships where different conditions prevail than in greater heights in the MBL. One should therefore be careful in extrapolating surface measurements to the complete MBL.

[93] Results from field campaigns with simultaneous measurements of both gas and particulate phase chemical composition are becoming more and more available. Comparisons of these results with model data should be made in the future. The availability of faster computers will make the introduction of more size bins for the description of aqueous phase chemistry possible.

Acknowledgments

[94] We thank John Crowley, Bill Keene, Dieter Perner, Kai Hebestreit, and Rainer Vogt for valuable discussions. The module for the calculation of the activity coefficients is courtesy of Beiping Luo, while the photolysis module was provided by Jochen Landgraf.