Provided for non-commercial research and education use. Not for reproduction, distribution or commercial use. Volume 42 Issue 3 January 21X18 ISSN 1.152-2311) ATMOSPHERIC ENVIRONMENT This article was published in an Elsevier journal. The attached copy is furnished to the author for non-commercial research and education use, including for instruction at the author's institution, sharing with colleagues and providing to institution administration. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier's archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright 53 Available online at www.sciencedirect.com ScienceDirect ATMOSPHERIC ^cienceuirect ENVIRONMENT ELSEVIER Atmospheric Environment 42 (2008) 554-567 www.elsevier.com/locate/atmosenv Regional differences in gas-particle partitioning and deposition of semivolatile organic compounds on a global scale Christian W. Götz, Martin Scheringer*, Matthew MacLeod, Fabio Wegmann, Konrad Hungerbühler Safety and Environmental Technology Group, Institute for Chemical and Bioengineering, Swiss Federal Institute of Technology, ETH Zurich, HCl Gl27, CH-8093 Zurich, Switzerland Received 1 June 2007; received in revised form 10 August 2007; accepted 13 August 2007 Abstract Variability in gas-particle partitioning of semivolatile organic compounds (SOCs) and related atmospheric processes (particle-associated deposition, rain washout and degradation) are investigated on a global scale. Two different sorption approaches (one using the octanol-air partition coefficient, Aoa, and one based on poly-parameter linear free energy relationships, ppLFER) and two different atmospheric box models (unit-world and highly spatially and temporally resolved) are applied. In the unit-world model, the overall deposition and atmospheric fate of SOCs calculated with the ^oa-based sorption approach are similar to the ones calculated with the ppLFER approach. Rain washout dominates the atmospheric removal of polar chemicals in the unit-world model while non-polar chemicals are removed mainly through degradation or particle-associated deposition. In contrast, big differences and a high sensitivity to the selected sorption approach are found in the spatially and temporally resolved model. The highly resolved geographic variability cannot be represented using the A"0a-based approach if aerosol components other than OM are of importance for sorption. In particular, aerosols in dry regions (desert) and regions with low OM aerosols (arctic, some oceanic regions) are more appropriately described by the ppLFER approach. With the ppLFER approach, good agreement between modeled deposition fluxes and measurement data are found for higher chlorinated PCBs and TCDD/Fs. In general, we recommend the ppLFER approach for highly resolved environmental fate models. © 2007 Elsevier Ltd. All rights reserved. Keywords: Atmospheric deposition; Gas-particle partitioning; Semivolatile organic compounds; POPs; Remote sensing; GCM 1. Introduction Atmospheric processes play an important role in the fate of semivolatile organic compounds (SOCs, Mackay and Paterson, 1991; Bennett et al., 2001; * Corresponding author. Fax: +41 44 632 11 89. E-mail address: martin.scheringer@chem.ethz.ch (M. Scheringer). Scheringer et al., 2003). In the atmosphere, SOCs are simultaneously present in the gas phase and associated with aerosol particles (Bidleman, 1988). In contaminant fate models, partitioning of SOCs to aerosol particles and removal of aerosol particles through dry deposition or rainfall are key factors determining long-range transport potential and overall persistence (Scheringer, 1997; Scheringer et al., 2003; Lohmann and Lammel, 2004). Nevertheless, 1352-2310/$-see front matter © 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2007.08.033 C. W. Götz et al. / Atmospheric Environment 42 (2008) 554-567 555 most multimedia box models treat gas-particle partitioning and particle-associated deposition processes in a rather simplistic manner. In chemical fate modeling, two different types of models are common: multimedia box models with relatively low spatial and temporal resolution (Mackay and Paterson, 1991; Scheringer et al., 2003); and models with high spatial and temporal resolution, based on general circulation models (GCM) (Gong et al., 2002; Semeena and Lammel, 2003; Leip and Lammel, 2004). In both types of models, gas-particle partitioning is generally described with single-parameter linear free energy relationships (spLFERs) that estimate the aerosol-air partition coefficient from vapor pressure or ocatanol-air partition coefficient (koa). These relationships have two important limitations. First, they are only valid within the compound class from which they have been derived (Goss and Schwarzenbach, 2002). Most spLFERs are based on sorption data for non-polar chemicals such as PCBs and PAHs, for which total sorption is assumed to be dominated by absorption into organic matter (OM, Cousins and Mackay, 2001). Other possible chemi-cal-sorbate interactions are not explicitly considered. Second, the effect of variability in composition and properties of aerosol particles is not taken into account. In most cases the sorptive capacity of the aerosol is described in terms of OM content only. In multimedia box models, a single aerosol particle size is often assumed and, thus, different atmospheric lifetimes of aerosol particles are not distinguished. However, the atmospheric lifetime of fine and coarse aerosol particles can differ considerably (Seinfeld and Pandis, 1998). Furthermore, average particle deposition velocities assumed for single-size aerosol particles in multimedia models often represent the deposition velocity of coarse aerosol particles, which have up to 100 times higher deposition velocities than fine particles (Seinfeld and Pandis, 1998; Mackay, 2001; Scheringer, 2002). Accordingly, particle-associated deposition may be overestimated in contemporary multimedia box models. Recently, it has been shown that gas-particle partitioning to aerosols can be described in a more comprehensive way with polyparameter linear free energy relationships (ppLFERs), which provide a more detailed description of different sorption processes by describing aerosols in more detail (Götz et al., 2007). In particular, different aerosol sizes, components, and surface characteristics can be specified separately. Here, we use this approach and distinguish two size fractions of atmospheric aerosols: fine (PM2.5) and coarse (PM 10-2.5) particles (Putaud et al., 2004). This differentiation allows us to specify particle-size dependent deposition velocities and aerosol compositions. The objectives of this paper are (i) to investigate the influence of gas-particle partitioning on atmospheric removal processes in a multimedia box model, and (ii) to investigate the global variability of gas-particle partitioning and atmospheric removal processes in a spatially highly resolved model that is representative of GCM-based models. On this basis, we make recommendations for the parameterization of sorption and aerosol related processes for different kinds of environmental fate models. Concerning objective (i), we investigate two different sorption models (XoA-based spLFER and ppLFER) and compare the results for the gas-particle partition coefficient and analyze the influence of the selected sorption model on gas-particle partitioning and deposition in a unit-world model. We use a two particle-size aerosol model, globally averaged aerosol composition, and globally averaged values for rain intensity and frequency, OH radical concentration, temperature, and relative humidity. To address objective (ii) we explore the global variability of gas-particle partitioning and deposition velocity using a set of highly resolved box models covering the earth at a spatial resolution of 180 x 360 cells and a temporal resolution of 1 month. We include geo-referenced data from remote sensing for temperature, precipitation, number of rain events, relative humidity, OH-radical concentration, and data from a GCM to characterize aerosol composition and concentration. Using these data, we calculate geo-referenced gas-particle partitioning (with XoA-based and ppLFER sorption models separately), deposition velocities, and rate constants for reaction with OH radicals. Furthermore, we compare modeled deposition fluxes with data from field measurements. 2. Chemicals and data sources We investigate four different groups of SOCs: (1) polychlorinated biphenyls (PCBs), (2) poly chlorinated dibenzo-dioxins and furans (PCDD/Fs), (3) DDT and DDE (DDTs), and (4) various polar current-use pesticides (CUP) such as phenylurea C.W. Götz et al. / Atmospheric Environment 42 (2008) 554-567 556 herbicides, amides, and triazines. Abraham solvation parameter and XoA-values of the chemicals were taken from Götz et al. (2007) and ABSOLV (Pharma-Algorithms, 2004), see Supplementary data. Monthly rainfall rate (f7rain [md-1]), average length of rain events (7wet [d]), and the average time between rain events (7dry [d]) were obtained from publicly available databases (GCPC, 2007; IPCC, 2007). Data for twst and tAry were available for continents only. Therefore, estimates of twet and tAry over the oceans and Antarctica were made by averaging data in latitudinal bands. Temperature and concentration of OH are based on data from Spivakovsky et al. (2000). Aerosol component concentration and relative humidity (RH) were calculated with the aerosol-climate model ECHAM5 HAM (Stier et al., 2005), in which the double-moment aerosol scheme HAM is coupled to the GCM ECHAM5. The aerosol population comprises the following major compounds: sulfate, elemental carbon, organic matter, sea salt, and dust. Its size distribution is described as the superposition of seven lognormal modes, whose evolution is dynamically controlled by aerosol microphysical processes. The data used in this paper has been computed with a special version of ECHAM5 HAM which includes a double-moment cloud microphysics scheme, and which showed good agreement with measurement data (Lohmann et al., 2007). This simulation has been performed for the year 2000, with a spatial resolution of T63 (96 x 192) on 31 levels, and a temporal output of 12 h. However, here we use monthly averaged values. The surface areas of the individual aerosol components such as silica or elemental carbon are not directly available, because ECHAM5 HAM calculates weight fraction of the different aerosol components. Hence, we estimated surface areas assuming that all components of the coarse fraction are individually accessible for the chemicals and that they consist of spherical particles with a mean radius of 5 |im. In Fig. 1 the yearly mean of these data is given for both aerosol size fractions, fine (PM2.5) and coarse (PM 10-2.5). 3. Methods and model description 3.1. Gas-particle partitioning We describe gas-particle partitioning with a ppLFER-based sorption model that includes ad- OM fine fraction (g/m3) 180° 90°W 0° 90°E 180' OM coarse fraction (g/m3) 180" 90°W 0° 90°E 180° EC coarse fraction (m2/m3) 180° 90°W 0° 90°E 180" mineral dust coarse fraction (m2/m3) 180° 90°W 0° 90°E 180" sea salt coarse fraction (m2/m3) 180° 90°W 0° 90°E 180° Fig. 1. Mass and surface concentration (logarithmic scale) of aerosol components in fine fraction (PM2.5) and coarse fraction (PM 10-2.5) that contribute to the sorptive capacity of the aerosol particles for SOCs. Annual averages of year 2000. OM: organic matter, EC: elemental carbon. C. W. Götz et al. / Atmospheric Environment 42 (2008) 554-567 sorption to different aerosol surfaces and absorption into aerosol particles, and with a XoA-based sorption model. Recently, we showed that in some cases the ppLFER-based model is preferable to XoA-based models, especially for polar chemicals and for aerosol particles with low OM content as in desert or arctic environments (Götz et al., 2007). Furthermore, the ppLFER model allows us to treat the fine and coarse aerosol particle fraction individually and to describe the deposition processes of chemicals sorbed to fine and coarse particles separately. For comparison against the ppLFER model, we use the XoA-based model of Finizio et al. (1997). Only interactions of the chemicals with OM are considered in the XoA-based model. The ratio of chemical concentration on particles [mol m~3 air] and in the gas phase [mol m~3 air] is described with the dimensionless partition coefficient kp (Götz et al., 2007). The distribution of a chemical between aerosol particle and gas phase can also be described by , which is the particle bound fraction of a chemical. 20(im) and extremely small ones, the scavenging efficiency approaches unity. However, for mid-sized particles (diameter ~ 1 \\m) raindrops collect only particles that are close to the center of the volume swept by the raindrop, resulting in scavenging efficiencies about 0.001 to 0.01 (Slinn, 1983; Seinfeld and Pandis 1998). In this work, we assume for the fine particle fraction iiScav = 0.01, and for the coarse fraction 0.5 (Seinfeld and Pandis, 1998). Q is 200,000, which means that each raindrop falls through a volume of air about 200,000 times its own volume prior to landing on the surface. This value represents a cloud base at 200 m and a raindrop volume of 1 mm3 (Mackay and Paterson, 1991; Mackay, 2001). The average rainfall intensity (^rainevent) depends on the geographic region. We derived {7raineVent from remote sensing data (see Section 2). If intermittent rainfall is considered, the Z)-value for the total wet deposition has an upper limit. We assume for the upper limit that the chemical and the particles are removed completely from the atmosphere during rain events, and that there are rainfall events with an average duration of twst and dry periods of duration tAry. Thus, we can calculate a minimum residence time (Eq. (6)) that is based on system properties only: tdi ry ry 2 t. (6) 'wet + ^dry With this assumption, the maximum D-value for the total wet deposition is given as nmax total, wet V 2 ?wet ~\~ tdi ry ry td ry (Z bulk,atmosphere)- (7) Therefore, the Z)-value of the total wet deposition, Z)total wet, is the minimum value of Eq. (7) (upper limit for total wet deposition) and the sum of Eqs. (3) and (5) (rain washout and wet particle-associated deposition): D total, wet min[-^wet%tal' ^g,wet + £>p,wet]- (8) 3.3. Mass transfer coefficients The chemical specific mass transfer coefficients (MTC, symbol kM, units of md_1) are calculated from the particle deposition velocities and rain rates, which are independent of the chemicals, and the partitioning properties of the chemical (^AW, (/)). In Eqs. (9) (11), the MTCs for dry particle-associated deposition (kMdry [md-1]), rain washout (/cM,rain [md"1]) and wet particle-associated deposition (fcM,wet [md-1]) are given: ^M,dry = ^d ryV; (9) ^M.rain = Umin[l/(KAW + VW/ VA){\ - )], (10) ^M,wet — U rain QE$, (11) C. W. Götz et al. / Atmospheric Environment 42 (2008) 554-567 3.4. Atmospheric box models We calculated atmospheric removal processes with a unit-world multimedia box model and with a global model composed of a set of spatially and temporally variable grid cells. Both models are solved as steady-state models with generic emission data. The spatially and temporally resolved model is a grid with 180 x 360 cells for which the environmental conditions are individually specified; in each cell, the environmental parameters change with a temporal resolution of one month. There are no horizontal transport processes in this model, i.e. atmospheric removal processes in each grid cell are calculated without outflow to or input from adjacent cells. In addition, in the grid cells there is no revolatilzation of chemicals from the surface media. Consequently, the global model is a model of atmospheric removal processes that can be used to analyze deposition and degradation processes, but not for prediction of phase partitioning and long-range transport. In the unit-world model, we applied globally averaged values of the remote sensing data and the data from the ECHAM5-HAM model (see Section 2) that are used in the spatially and temporally resolved model. The global averages of the concentrations of the aerosol components in the unit-world model are: cOM fine = 1.3 x 10~6gm~3, c0M,coarse = 2.3 x 10~7gm~3, CEQcoarse = 1.78 X 10~7m2m~3, CsilicajCOarse = 3.4 X 10~6m2m~3, cseasait;Coarse = 7 x 10~6m2m~3. The same generic emission scenario (lOOmold-1) was applied to the unit-world model and all cells of the spatially resolved model. 4. Results and discussion 4.1. Unit-world model Atmospheric removal processes considered are wet and dry particle-associated deposition, rain washout, and degradation. Results from the unit-world model are given in Fig. 2. We distinguish between three groups of chemicals according to their characteristic atmospheric removal processes: degradation, particle-associated deposition, and rain washout. The atmospheric removal of lighter PCBs is dominated by degradation, independent of the applied sorption model. However, if the XoA-based approach is used, particle-associated deposition is more important than for the ppLFER approach. 559 The XoA-based approach yields higher aerosol-air partition coefficients for non-polar compounds if the sorptive capacity of the aerosol particles is dominated by OM, which is the case in the unit-world model (Götz et al., 2007). Generally, the importance of particle-associated deposition processes increases with increasing degree of chlorina-tion of the PCBs. Rain washout is negligible for the lighter PCBs, which have relatively high Henry's law constants and do not partition to raindrops to a significant amount. The atmospheric fate of heavier PCBs, TCDD/ Fs, and DDTs is controlled by particle-associated deposition. Dry and wet particle-associated deposition processes are both important, with wet particle-associated deposition being higher than dry particle-associated deposition by about a factor of 2. Again, the ^OA-based approach tends to result in higher total particle-associated deposition than the ppLFER approach. In contrast to the general trend of increasing sorption with increasing degree of chlorination that occurs for the XoA-based sorption model, PCB 155 shows low sorption and, thus, particle-associated deposition if the ppLFER sorption model is used. This is due to the relatively low value of 8.72 for the logarithm of the hexadeca-ne-air partition coefficient of PCB 155 (Abraham and Al-Hussaini, 2005) that is used in the ppLFER approach. However, this estimated value appears to be too low compared with values of similar PCBs (e.g. PCB 153, log i^hexadecane-air = 9.59); there are no measurement results available. 2,3,7,8-TCDD shows a slightly higher sorption and particle-associated deposition with the ppLFER approach, in contrast to 2,3,7,8-TCDF, which has a less pronounced electron-donor property, and PCBs with no electron-donor properties at all. This effect is attributable to the ether groups of 2,3,7,8-TCDD, which can participate in specific adsorption interactions with aerosol surfaces, which is only considered in the ppLFER approach. The atmospheric fate of CUPs is mainly controlled by rain washout and atmospheric degradation. Therefore, the influence of the aerosol model is less significant than for particle-associated deposition controlled chemicals. However, for some CUPs, dry particle-associated deposition plays an important role. Phenylurea herbicides (isoproturon, diuron, metoxuron), propazine, and metolachlor show a significant dry coarse particle-associated deposition in the ppLFER approach that does not show up in the XoA-based model. This is caused by f* f* ľ"\\# 560 C.W. Gôtz et al. / Atmospheric Environment 42 (2008) 554-567 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% contribution to removal from the atmosphere (ppLFER sorption approach) 3 F 1tF W 1 J : ■- - — : _ ■ - - - - - —_ - - - - _t 1 1 11 ' 100% 90% 80% 70% 60% •■ 50% j-40% 30% 20% 10% 0% contribution to removal from the atmosphere (Koa-based sorption approach) flj _ - I CO i- cm co CMor^t-ioooootoioo mtor^oo-t-cOLOLoeo □DCDCQcrjcocQ CLO-O.CLCLO.OOOOOOOO 0_0_Q_Q_D_D_D_Q_ Q U- -cl- CO CO Is- 5! S 8 § ---4-J ID Q) < E o o 2 o O g M CD E -Q O o CD -Q i_ 01 o O COH degradation Drain washout Hwet fine part. asso. deposition Hwet coarse part, asso deposition Odry fine part. asso. deposition Ddry coarse part, asso deposition Fig. 2. Relative importance of atmospheric removal processes of different chemicals for the ppLFER and ^OA-based sorption model. their strong sorption to mineral surfaces (silica, kaolinite, bentonite) in dry periods (average RH between rain events: 60%) that increases with increasing polarity of the chemicals (Goss and Eisenreich, 1996; Goss and Schwarzenbach, 2002). Mineral dusts are mainly coarse particles and, thus, settle much faster than fine particles, which contain a higher fraction of OM. Differences in deposition between the ppLFER and the XoA-based model are not large in the unit-world model, however they are more important in a spatially resolved model in regions where precipitation and RH are lower, as is the case in desert, arid, or arctic regions (see Section 4.2). Beside the deposition processes, degradation plays a major role for the removal of CUPs from the atmosphere. The highest degradation occurs for alachlor, which has a relatively high second order degradation rate constant with OH of 1.6 x 10_5d_1cm3. In summary, in a unit-world model the choice of the sorption model influences the relative importance of atmospheric removal processes significantly for particle-associated deposition controlled chemicals, and for some degradation-controlled chemicals. However, both sorption models yield similar overall atmospheric removal C. W. Götz et al. / Atmospheric Environment 42 (2008) 554-567 rates and both are appropriate to be applied. The specific values of the atmospheric removal processes depend strongly on system parameters, such as precipitation, RH, and aerosol composition, for which global average values are used in the unit-world model. 4.2. Spatially and temporally resolved global model To investigate the influence of sorption approach, particle deposition velocities, and rain rate on the atmospheric removal on a global scale, we use global remote sensing and GCM data as presented above, and focus on three chemicals representing the three groups defined above: PCB28 as a degradation-dominated chemical, PCB180 as particle deposition-dominated chemical, and terbuthylazine, for which atmospheric fate is controlled by rain washout. 561 4.2.1. Spatial and temporal variability of the particle-bound fraction, values of PCB28, PCB180, and terbuthylazine are given in Fig. 3. PCB28 is essentially in the gas-phase in the atmosphere all over the globe. However, from October to March, up to 30% can be bound to particles in the northern hemisphere in regions with very high particulate OM concentrations and low temperature. PCB28 is a non-polar chemical and thus sorbs mainly into OM (Götz et al., 2007). PCB 180 shows strong particle binding (0.9) in arctic and temperate regions, whereas calculated with the ppLFER approach for PCB28, PCB180, and terbuthylazine. C.W. Götz et al. / Atmospheric Environment 42 (2008) 554-567 562 the aerosol-air partition coefficient by several orders of magnitude compared to tropical regions. As an example, with a phase transition energy (AUOA) of -95.2k.Tmor1 (Schenker et al., 2005), the KGa of PCB 180 decreases approximately by a factor of 3 for every —10 °C. Therefore, we observe a high