Chemosphere 85 (2011) 1353–1359

Contents lists available at SciVerse ScienceDirect

Chemosphere

journal homepage: www.elsevier .com/locate /chemosphere

Atmospheric fate of non-volatile and ionizable compounds

Antonio Franco a,⇑, Michael Hauschild b, Olivier Jolliet c, Stefan Trapp a

a Department of Environmental Engineering, Technical University of Denmark, Miljøvej, Building 113, DK-2800 Kgs. Lyngby, Denmarkb Department of Management Engineering, Technical University of Denmark, Produktionstorvet, Building 426, DK-2800 Kgs. Lyngby, Denmarkc Center for Risk Science and Communication, University of Michigan, 109 South Observatory, Ann Arbor, MI 4819-2029, United States

a r t i c l e i n f o a b s t r a c t

Article history:Received 16 May 2011Received in revised form 25 July 2011Accepted 26 July 2011Available online 31 August 2011

Keywords:Activity approachAtmospheric fateLong-range transportIntermittent rainInterface partitioningpH

0045-6535/$ - see front matter � 2011 Elsevier Ltd. Adoi:10.1016/j.chemosphere.2011.07.056

⇑ Corresponding author. Tel.: +44 1234264939.E-mail address: antonio.franco@unilever.com (A. F

A modified version of the Multimedia Activity Model for Ionics MAMI, including two-layered atmosphere,air–water interface partitioning, intermittent rainfall and variable cloud coverage was developed to sim-ulate the atmospheric fate of ten low volatility or ionizable organic chemicals. Probabilistic simulationsdescribing the uncertainty of substance and environmental input properties were run to evaluate theimpact of atmospheric parameters, ionization and air–water (or air–ice) interface enrichment.

The rate of degradation and the concentration of OH radicals, the duration of dry and wet periods, andthe parameters describing air–water partitioning (KAW and temperature) and ionization (pKa and pH) arethe key parameters determining the potential for long range transport. Wet deposition is an importantremoval process, but its efficiency is limited, primarily by the duration of the dry period between precip-itation events.

Given the underlying model assumptions, the presence of clouds contributes to the higher persistencein the troposphere because of the capacity of cloud water to accumulate and transport non-volatile (e.g.2,4-D) and surface-active chemicals (e.g. PFOA). This limits the efficiency of wet deposition from the tro-posphere enhancing long-range transport.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

Partitioning and atmospheric transport of organic pollutants areinfluenced by the presence of water (e.g. clouds, fogs) and its depo-sition rates (e.g. rainfall, snow). Hence, modeling the concentrationand the deposition of atmospheric water is essential to predict theatmospheric fate, especially for non-volatile chemicals. In this con-text, two critical issues have been identified: the adsorption to theair–water interfaces, affecting the phase distribution of chemicals(Valsaraj, 1988; Hoff et al., 1993; Donaldson and Valsaraj, 2010)and the dynamic nature of the water mass balance in air, influenc-ing atmospheric phase composition and deposition rates (Jollietand Hauschild, 2005).

Chemicals equilibrium partitioning is typically described byphase partition coefficients. However, many chemicals, especiallyhydrophobic and surface-active substances, can be enriched atthe air–water interface. In the atmosphere, the sorption capacityof the interface can easily be larger than the capacity of the bulkwater, because of the high specific surface area of clouds, aerosolor fog droplets (Hoff et al., 1993). The importance of interface par-titioning was recognized early (Valsaraj, 1988), and a mathemati-

ll rights reserved.

ranco).

cal framework integrating interface partitioning in multimediafugacity models was soon developed (Hoff et al., 1993). Yet, theimplementation in multimedia fate models is poor (Valsaraj andThibodeaux, 2010). One exception is the consideration of the sur-factant nature of perfluorooctanoic acid, though an empirical ap-proach to model surface enrichment was used (Webster et al.,2010).

Until recently, little attention has been paid to consider the dy-namic nature of atmospheric phase composition and mass trans-port velocities, above all intermittent rainfall (Jolliet andHauschild, 2005). The mass balance of atmospheric water is extre-mely variable in time and space. Water is abundant in the atmo-sphere as water vapor. After condensation, liquid water becomesa phase to which chemicals can partition (e.g. clouds, rain droplets,and snow). Cloud droplets grow from condensation nuclei andwhen exceeding a critical size, they reach the ground asprecipitation.

In the past, conventional multimedia fate models typically de-scribed rainfall as a continuous loss process. Rainfall is an intermit-tent process with variable intensity and frequency. Jolliet andHauschild (2005) demonstrated that the assumption of continuousrain can lead to a severe overestimation of wet deposition ratesand subsequent underestimation of travel distances by up to threeorders of magnitude and proposed an approximation for the incor-poration of intermittent rain in steady-state multimedia models.

1354 A. Franco et al. / Chemosphere 85 (2011) 1353–1359

The mass balance of atmospheric water is particularly relevantfor low volatility (KAW < 10�4) and ionizable chemicals because oftheir strong tendency to associate to liquid water. Polar and ioniz-able compounds are increasingly frequent among chemicals underregulatory scrutiny (Franco et al., 2010). We recently showed howcloud water can increase the activity capacity of the atmospherefor non-volatile ionizable substances (Franco and Trapp, 2010)and hypothesized that clouds may store and transport them. Toour knowledge, current multimedia fate models do not includecloud water, although several describe mixing between the atmo-spheric boundary layer and the troposphere (Wania et al., 1999;MacLeod et al., 2001; Webster et al., 2010).

The goal of this paper is to investigate the effects of ionization,interface partitioning, intermittent rain and variable cloud contenton the transport of non-volatile and ionizable chemicals in the atmo-sphere and to identify the key physicochemical and environmentalparameters. Probabilistic simulations at steady-state were run fora selection of ten non-volatile or ionisable substances, using a mod-ified version of the Multimedia Activity Model for Ionics (MAMI).

2. Methods

2.1. Multimedia Activity Model MAMI

MAMI is based on the activity approach, which is analogous tothe fugacity approach with activity and activity capacities replac-ing fugacity and fugacity capacities (Trapp et al., 2010). It mirrorsthe regional scale of Simplebox, which is the exposure model forchemical safety assessment in the EU (Brandes et al., 1996). Addi-tionally, MAMI features pH-dependency and species-specific phasepartitioning. The total concentration (C, mol m�3) of a monovalentionizable chemical in air with gas, water and aerosol is the productof the total activity (at, mol m�3) and the bulk apparent activitycapacity (B, m3 m�3), which includes the fractions of the neutral(Un) and ionic species (Ui) in gas, water and solids (G, W and S inm3 m�3) (Trapp et al., 2010).

C ¼ at GUnKAW;n þWUn

cnþUi

ci

� �þ S

UnKd;n

cnþUiKd;i

ci

� �� �¼ atðBG þ BW þ BSÞ ð1Þ

where KAW,n, Kd,n and Kd,i (m3 m�3) are the species-specific air–water and solid-water partition coefficients, cn and ci are thedimensionless activity coefficients of the neutral and ionic speciesand BG, BW and BS (m3 m�3) are the activity capacities of gas, waterand solids. At equilibrium, the concentration ratio between twoenvironmental compartments (e.g. the lower middle troposphere(LMT = 1) and the atmospheric boundary layer (ABL = 2) is the prod-uct of the activity ratio (the ratio of total activities in each compart-ment, e.g. R12 = at,1/at,2) and the ratio of activity capacities in eachcompartment (Trapp et al., 2010):

C1

C2

� �eq¼ R12

B1

B2ð2Þ

Mass transport and removal processes are quantified by T-values(m3 h�1) and include horizontal advection, vertical eddy diffusion,intermittent wet deposition, dry particle deposition and degrada-tion by OH radicals. Mass fluxes (mol h�1) are the product of the to-tal activity of one compartment (mol m�3) and the T-value of aspecific process (Franco and Trapp, 2010):

dmdt¼ at � T ð3Þ

Interface partitioning, atmospheric layering, vertical transport,intermittent rain and temperature dependency of air water parti-tioning and of atmospheric degradation (Supplementary material,

Section S1) are new features compared to the previous MAMI ver-sion. OH radicals are assumed to react with the fractions in gasphase and with the fraction adsorbed at the air–water interface,while no degradation is considered for the fraction absorbed inaqueous and solid phase (Goss, 2004).

2.2. Bulk and interface partitioning

The total concentration of a substance in water, CW,t, is the sumof the concentration in the bulk phase and the excess concentra-tion at the surface. Interface partitioning can be incorporated inthe activity capacity of water (BW in Eq. (1)) as a function of theinterface-water partition coefficient KIW, in (mol m�2) (mol m�3)�1

or (m), and the effective depth dW, which is equal to the inverse ofthe specific surface area of a water particle in air (m3 m�2) or (m).For spherical particles, to which cloud and rain droplets can beapproximated, this is equal to 1/6th of the diameter (Hoff et al.,1993). Physically, KIW can be interpreted as the depth of the watercontaining the same amount of substance as the interface. AtT > 0 �C, BW for a monovalent ionizing compound includes the con-tribution of the bulk aqueous phase and of the surface for neutraland ionic species:

BW ¼W Un

1þ K IW;ndW

cnþUi

1þ KIW;idW

ci

!; T > 0 �C ð4Þ

The activity coefficients can be approximated to 1 because the ionicstrength of clouds and rainwater is negligible. Species-specific mea-surements of KIW are not available because only apparent values arereported for ionizable substances and no estimation method existsfor ions. Using an apparent KIW, that covers both species, Eq. (4) canbe simplified:

BW ¼W 1þ K IW

dW

� �; T � 0 �C ð5Þ

At T < 0 �C, BW only includes the contribution of ice or snow surface.Ice crystals are characterized by the effective thickness dS (m), i.e.the inverse of the specific surface area.

BW ¼WKSW

dS; T < 0 �C ð6Þ

Species fractions in Eqs. (1) and (4) refer to the bulk. The surface pHof water may be different than the bulk pH, possibly altering speci-ation and sorption equilibriums, but the extent and direction of thepH shift is still controversial (Winter et al., 2009).

2.3. Atmospheric layering and vertical mass exchange

Compared to the existing version of MAMI (Franco and Trapp,2010), the lower/middle troposphere (LMT = 1), ranging between1000 and 5000 m height, was added over the atmospheric boundarylayer (ABL = 2) (Fig. 1). The LMT includes gas, solid aerosol and theliquid water of clouds. The base of low and middle level cumulusand stratus cloud is typically in this elevation range. The ABL con-tains gas, aerosol solids and an aerosol aqueous phase. Convectivecirculations ensure vertical mixing throughout the ABL, while a cap-ping inversion separates it from the troposphere (Seinfeld and Pan-dis, 1998).

In MAMI, the T-values for upward and downward vertical mix-ing were derived from the D-values for eddy diffusion of the fugac-ity model GloboPOP (Wania et al., 1999):

T21;EDDY ¼ AB2v12;EDDYq12

q2ð7Þ

Fig. 1. Conceptual model of the lower/middle troposphere (LMT) and the atmo-spheric boundary layer (ABL) in MAMI, with mass transport and removal processes.

A. Franco et al. / Chemosphere 85 (2011) 1353–1359 1355

T12;EDDY ¼ A BG;1 þ BS;1ð Þv12;EDDYq12

q1ð8Þ

where A is the area, v12,EDDY is the vertical mixing velocity (m h�1),q1 and q2 are the densities of the LMT and the ABL, and q12 is thedensity between the layers. A fixed value of 5 m h�1 was chosenfor v12,EDDY (MacLeod et al., 2001). If convective thermals reachthe condensation level above the ABL, water vapor condensates intocumulus clouds. Clouds float above the ABL and their liquid water(BW,1) does not participate in the vertical mixing.

2.4. Intermittent rain

Rain and snow transport chemicals downwards to the ABL andfurther to the ground. Most precipitating clouds are formed in thelower troposphere (Seinfeld and Pandis, 1998). The fraction of LMTexposed to precipitation, a1, is assumed 0.5, while precipitationfalls all through the ABL (a2 = 1). The T-values for wet depositionare proportional to the area, A, the precipitation intensity, vWETDEP,the fraction of layer exposed to precipitation (a), and the total B-value of precipitation, which is the sum of particles scavenging(BS�ecoll where ecoll is the particle collection efficiency), and gaswashout. Gas washout is determined by the equilibrium concen-tration in precipitating water, considering that the chemical maybe mostly in droplets during rainfall. Equilibrium, i.e. the ratio ofB-values, is then calculated between rainy air and rain droplets(Lei and Wania, 2004).

TWETDEP ¼ A vWETDEP a BS ecoll þ BBW;PREC

BþWPRECBW;PREC

� �ð9Þ

where WPREC, assumed 6 � 10�8 m3 m�3 from Jolliet and Hauschild(2005), is the volumetric fraction of precipitation in air, and BW,PREC

is the activity capacity of precipitation in air, calculated by Eq. (5) or(6). The particles collection efficiency in each layer, ecoll, is propor-tional to the layer’s thickness and is deduced assuming that the to-tal scavenging ratio, typically 2 � 105, can be calculated additivelyfrom the layer-specific scavenging ratios (Wania et al., 1999).

The algorithm for periodic intermittent rain proposed by Jollietand Hauschild (2005) for steady-state multimedia models wasconverted in terms of T-values. The overall removal T-value foreach layer is the minimum between two cases:

T ¼min TDRY þ 2BVtDRY

� �tDRY þ tWET

tDRY

� ���� TDRYtDRY þ TWETtWET

tDRY þ tWET

�ð10Þ

where TWET and TDRY are the total T-values including all removalprocesses during dry and wet periods and tDRY and tWET are thedurations of the periods. The overall residence time in each layeris determined by the ratio of the total mass (at�B�V) and the mass re-moval of that layer (at�T) excluding advection (at�TOUT) and quanti-fies the persistence in an atmospheric layer, regardless of themodel surface area:

tRES ¼BV

T � TOUTð11Þ

2.5. Parameterization of atmospheric compartments

The simulated scenario comprises the time and space variabilityof environmental conditions for a generic European region. Proba-bility density functions were defined for the pH, the liquid water(W) and the solids volumetric contents (S), the temperatures (T),the concentration of hydroxyl radicals (OH), the effective diameterof cloud droplets (dW,CLOUD), rain droplets (dW,RAIN), snow crystals(dW,SNOW) and wet aerosol (dW,AER), the duration of dry and wetperiods (tDRY and tWET), the precipitation intensity (vWETDEP), thevertical mixing velocity (v12,EDDY) and the advective residencetimes (s). Environmental parameters and their uncertainty rangesare reported in Table S1, Supplementary material, with references.

2.6. Test chemicals

The ten test chemicals comprise two neutral compounds: meth-omyl (MET) and propoxur (PROP); four acids: 2,4-dichlorophe-noxyacetic acid (2,4-D), perfluorooctanoic acid (PFOA),pentachlorophenol (PCP) and thiophenol (THP); and four bases:diazinon (DIA), 4-chloroaniline (CHA), quinoline (QIN), and tri-methylamine (TMA). The input properties together with the uncer-tainty parameters of the test substances are reported in Table 1.Experimental data were chosen whenever available, but KAW,n

and octanol–water partition coefficient of the neutral molecule(KOW,n) were calculated for ionizable substances, using ACD-Labs�.Details of the derivation of interface partition coefficients (KIW andKSW) are given in Section S3 of the Supplementary material.

2.7. Probabilistic simulations

Steady-state Monte Carlo probabilistic simulations were run fora constant emission of 100 mol h�1 into the ABL, using the softwareCrystal Ball�. Internal correlation between input parameters wasonly taken into consideration for the temperatures in the two com-partments. The total number of runs was set to 5000. The variabil-ity of results was quantified by the 5- and 95%-ile of predictedvalues for activity capacities (B-values), activity ratio R12, residencetimes (tRES,1 and tRES,2), total activities (at,1 and at,2), concentrations(Ct,1 and Ct,2) and mass transport and removal fluxes. To identifythe most sensitive parameters, the contribution of each inputparameter to the variance of the estimate was measured by rankcorrelation.

3. Results

3.1. Phase distribution within compartments

The median of the fractions of chemicals in vapor, water and so-lid phase are reported in Table 2. Substances with very low appar-ent KAW (i.e. KAW = Un KAW,n < 10�6) (MET, PRO, 2,4-D, TMA) or withhigh interface adsorption coefficients (PFOA, DIA, PCP, 2,4-D) areinfluenced by the water content in air, which is significant in thecloudy LMT (W1 = 10�8 to 3 � 10�7 m3 m�3). In the LMT, median

Table 1Physicochemical properties and uncertainty distributions of the test chemicals. Substance parameters are normally distributed around the reported value (mean) with thestandard deviation (r). Uniform distributions (U) were chosen for highly uncertain parameters. References different from the one indicated on top of the table are specified asfootnotes.

pKa log KOW,n log KAW,n log KIW log KSW log kDEG

Ref ACD/Labs EPI Suite (285 K) Roth et al. (2002) Roth et al. (2004) EPI Suite (293 K) in cm3 (h molec)�1

r 0.20a 0.23a 0.67a 0.87 0.87 0.22

MET – 0.60 �9.45 �8.08b �8.48b �7.62PRO – 1.54 �7.46 �7.41b �7.91b �6.942,4-D 2.98 2.81 �7.30c �5.91b �6.10b �7.62PFOA U 0/3.8d 5.38, r = 0.87e �3.19d U �5.05/�4.23f U �5.05/�4.23f �8.73PCP 4.78 5.11 �4.14c �5.45b �4.91b �8.70THP 6.61 2.33 �2.21c �7.12 �6.36 �7.39DIA 1.21 3.81 �5.79 U �5.65/�3.35g U �5.65/�3.35g �6.46CHA 3.97 1.95 �4.74 �7.15h �6.74h �6.95QIN 4.97 2.13 �4.46 �6.65h �6.61h �7.38TMA 9.75 0.06 �2.65i �7.52l �7.52m �6.60

a Standard deviation based on ECHA (2008).b Molecular descriptors calculated by ChemProp�, v.5.2.c Calculated with ACD/Labs for the neutral structure.d From Li et al. (2007).e From Webster et al. (2010).f Min/max range from Ju et al. (2008), McMurdo et al. (2008), and Psillakis et al. (2009).g Selected range from Kelly et al. (2004).h Molecular descriptors from Poole and Poole (1999).i From Christie and Crisp (1967).l Calculated from surface tension vs. concentrations in Mmereki and Hicks (2000).

m Assumed equal to KIW.

Table 2Median phase distribution (%) of the test chemicals in the lower/middle troposphere(LMT) and in the atmospheric boundary layer (ABL).

LMT ABL

Gas Water Solids Gas Water Solids

MET 17 82 1 92 5 2PRO 83 16 <1 100 <1 <12,4-D <1 99 <1 96 <1 4PFOA 16 84 <1 100 <1 <1PCP 88 10 2 100 <1 <1THP 100 <1 <1 100 <1 <1DIA 19 80 1 99 <1 1CHA 100 <1 <1 100 <1 <1QIN 99 1 <1 96 <1 4TMA 63 35 2 16 5 79

1356 A. Franco et al. / Chemosphere 85 (2011) 1353–1359

temperatures are below freezing and surface–air partitioningdetermines the median activity capacity of clouds. At T > 0 �C, li-quid water contributes to the activity capacity of the LMT fornon-volatile compounds (e.g. MET, 2,4-D). In the ABL, all chemicalsare predominantly in the gas phase, except TMA, which is mostlyabsorbed as the charged species (cation).

Fig. 2. Calculated median and variability range (5- and 95%-ile) of the residencetime in the lower/middle troposphere (�) and in the atmospheric boundary layer(j).

3.2. Activity and concentration ratios at equilibrium

The ratio of total activities (R12 = at,1/at,2) refers to the waterphase of each compartment and is calculated assuming equilib-rium between the gas phase of the two layers, i.e. equal activityof the neutral species (an,1 = an,2). Equilibrium at unequal totalactivities (R12 – 1) occurs for ionizing compounds because onlythe neutral molecule diffuses in gas phase. The higher pH of clouddroplets in the LMT compared to the pH of aerosol aqueous phasein the ABL (Table S1, Supplementary material) results in activity ra-tios R12 > 1 for acids because of increasing ionization at high pH,while the opposite occurs for bases. The calculated activity ratiosand the activity capacities B of the LMT and the ABL and of precip-itation in LMT and ABL are presented in Fig. S1, Supplementarymaterial.

To compare the capacity of the two compartments to sorb achemical, the concentration ratio at equilibrium C1/C2 (Eq. (2)),which is different from the ratio of B-values for ionized substances,must be considered. The concentration ratio at equilibrium is closeto one for most compounds but is greater than one for 2,4-D(650 m3 m�3), PFOA (11 m3 m�3), DIA (4 m3 m�3) and MET(3 m3 m�3) because of the contribution of clouds to the activitycapacity.

3.3. Residence times

Fig. 2 shows the residence times of the test chemicals in the twoatmospheric layers, (Eq. (11)). The input parameters with the high-est contribution to variance are reported in Table 3.

The predicted tRES ranges between a few hours (DIA, PRO, CHAand TMA) to >10 d (PFOA, PCP). Median and variability of tRES inthe two compartments are in most cases in the same range. Reac-

Table 3Contributions of input parameters to the variance of the calculated residence time inthe LMT (tRES,1) and in the ABL (tRES,2), measured by rank correlation. Positive rankcorrelation coefficients indicate direct proportionality. The three most importantparameters are reported.

tRES,1 tRES,2

MET tDRY (+0.47), tWET (�0.36), kDEG

(�0.24)tDRY (+0.52), tWET (�0.30), KAW

(�0.25)PRO kDEG (�0.50), OH1 (�0.36), T1

(+0.23)kDEG (�0.70), tDRY (+0.41), OH2

(�0.27)2,4-

DtDRY (+0.42), tWET (�0.37), W1

(+0.21)tDRY (+0.77), tWET (�0.31), kDEG

(�0.24)PFOA W1 (+0.37), vWETDEP (�0.34), tWET

(�0.34)pKa (+0.72), KAW (+0.45), pH2

(�0.29)PCP KAW (+0.38), v12,EDDY (�0.29), tWET

(+0.27)KAW (+0.83), T2 (+0.32), vWETDEP

(�0.20)THP kDEG (�0.78), OH1 (�0.58), T1

(�0.17)kDEG (�0.90), OH2 (�0.36), T2

(�0.15)DIA kDEG (�0.77), OH1 (�0.57), T1

(�0.14)kDEG (�0.86), OH2 (�0.34), tDRY

(+0.19)CHA kDEG (�0.77), OH1 (�0.54), T1

(�0.19)kDEG (�0.67), tDRY (+0.35), OH2

(�0.25)QIN kDEG (�0.69), OH1 (�0.49), KAW

(+0.19)tDRY (+0.61), KAW (�0.35), tWET

(�0.31)TMA kDEG (�0.79), OH1 (�0.56), T1

(�0.19)kDEG (�0.81), OH2 (�0.30), pH2

(�0.23)

A. Franco et al. / Chemosphere 85 (2011) 1353–1359 1357

tive compounds (THP, DIA, CHA and TMA) are more persistent inthe ABL because of the lower concentration of OH radicals in thiscompartment. A slight opposite trend is calculated for MET, 2,4-D(at least the upper range) and PFOA. In all cases except for PFOAand PCP, tRES are influenced by the parameters determining degra-dation (kDEG and OH concentration) or by the duration of dry andwet periods (Table 3), the latter being crucial for non-volatile com-pounds, particularly if not rapidly degraded (MET, 2,4-D). Long res-idence times are influenced by several parameters. The parametersdescribing air–water bulk partitioning (KAW and T) are positivelycorrelated to the persistence of some compounds (e.g. PCP, PFOA)because higher volatility limits wet deposition. The parametersdescribing ionization (pKa and pH) impact air–water partitioningand so the residence time of some ionizing compounds (PFOA,TMA). The liquid water content in the LMT (W1) explains the highertRES,1 of PFOA and the higher 95%-ile tRES,1 (corresponding toT > 0 �C) of 2,4-D and methomyl. Removal of PFOA from the LMTis influenced by the precipitation intensity (vWETDEP) as well as bythe KSW.

Similar or higher residence times in the LMT compared to theABL are observed for those compounds that are preferably sorbedin the cloudy LMT (MET, 2,4-D, PFOA).

(a)

Fig. 3. Median total activity (at, mol m�3), concentration (Ct, mol m�3), mass transport andat steady-state for an emission of 100 mol h�1 into the ABL.

3.4. Mass transport and removal processes

Transport and removal processes can be compared by lookingat individual T-values during dry and wet periods (Table S2, Sup-plementary material).

On a regional scale, the mass balance in the ABL sufficientlydescribes the atmospheric fate of chemicals because the fractiontransported to the troposphere is small, i.e. maximum 7% of theemission, considering only vertical eddy diffusion and horizontaladvection. During wet periods, wet deposition is the dominant re-moval process for all test substances except THP. In a time scaleincluding dry and wet periods, however, degradation is the dom-inant removal process for THP, DIA and TMA, wet deposition forMET, 2,4-D and QIN, wet deposition and degradation for PROand CHA, gaseous deposition (diffusive exchange with surfacemedia) for PFOA, gaseous deposition and wet deposition for PCP.

The calculated mass transport and removal fluxes of two sub-stances, methomyl and perfluorooctanoic acid, are shown inFig. 3. MET is an example of neutral, non-volatile molecule, pri-marily removed by wet deposition, whose persistence in both lay-ers is governed by the duration of dry and wet periods. PFOA is anionizable surfactant, slowly removed from the atmosphere by dif-fusive exchange and wet deposition, whose persistence is influ-enced by ionization, rain intensity and enrichment in clouds(Table 3).

4. Discussion

4.1. Governing processes and parameters

The parameters describing removal by degradation (kDEG andOH concentration) and by intermittent wet deposition (tDRY, tWET

and vWETDEP) are the main factors affecting the atmospheric fateof the test compounds. By modeling intermittent rain, processesother than wet deposition, determining removal during dry peri-ods, become relatively more important than by assuming continu-ous precipitation, as previously discussed (Jolliet and Hauschild,2005). Wet deposition is a key removal process for persistent,non-volatile substances, but its efficiency is limited by the inter-mittent pattern of precipitation. The duration of dry and wet peri-ods, and not the precipitation intensity, are the most importantparameters for persistent substances with very low KAW, such asMET, 2,4-D, PRO (Table 3). The tDRY, ranging between 2 to >10 dfor Europe (Table S1, Supplementary material), explains >40% ofthe variance of tRES,1 and >50% of the variance of tRES,2 for METand 2,4-D. The rain intensity is a sensitive parameter for persistentsubstances, slowly removed by wet deposition, such as PFOA andPCP (Table 3), in agreement with Jolliet and Hauschild (2005).

(b)

removal fluxes (mol h�1) of methomyl (a) and perfluorooctanoic acid (b) calculated

1358 A. Franco et al. / Chemosphere 85 (2011) 1353–1359

4.2. Effect of ionization

Ionization decreases the activity capacity of the gas phase (BG)because only neutral species volatilize and favors partitioning ofions into aerosol, cloud and rain droplets. Substances that aremostly ionized at atmospheric pH have a relatively high tendencyto the aqueous phase, although BW does not change with ionization(Eqs. (5) and (6)). Ionization increases absorption of cations and de-creases absorption of anions into organic matter (Franco andTrapp, 2008), but the effect is only significant for one of the testcompounds (TMA). Ionized compounds have generally high chem-ical activities in air, indicating a high tendency to escape thatcompartment.

Ionization favors dry and wet deposition processes but may de-crease the reactivity, assuming that the fraction absorbed to aero-sol is not oxidized by OH radicals. In some cases (e.g. PFOA)ionization decreases the residence time, in others (e.g. TMA) it in-creases it (Table 3).

The highest impact of pH and pKa is observed for PFOA (Table 3).As reported elsewhere (Armitage et al., 2009; Webster et al., 2010),the pKa is the most relevant (and uncertain) physicochemical prop-erty determining the atmospheric fate of this compound. PFOA ismore persistent at high pKa and low pH, i.e. when it is unionized.Ionization enhances both diffusive exchange with surface mediaand wet deposition but, compared to a previous assessment(Armitage et al., 2009), the wet deposition is relatively smaller.The dominant effect of ionization is to increase gaseous depositionby increasing the activity gradient between air (low pH) and ter-restrial compartments (high pH). In the ‘‘stickier’’ soil and watercompartments, as observed by Armitage et al. (2009), the moleculesubstantially dissolves and ionizes, if its pKa is below the pH.

Conversely, ionization increases the persistence of substancesprimarily removed by degradation. The basic compounds QINand TMA are increasingly associated with aerosol at lower pH,while the fraction in gas, subject to oxidation by OH radicals,decreases.

4.3. Effect of interface partitioning

Adsorption at the air–water interface contributes the activitycapacity of clouds, rain, snow and aqueous aerosol, depending onvolumetric content and on the thickness of water (Eqs. (5) and (6)).

In the ABL, W2 is very low and the contribution of the interfaceto the total B-value is small for all test compounds. Boundary layerclouds, marine aerosol suspensions (Qureshi et al., 2009), or fog(Goss, 1994) can contribute to the capacity of the ABL but arenot included in the modeled scenario.

In the LMT, clouds contain liquid water dispersed in microdro-plets (W1 = 10�8 to 3 � 10�7 m3 m�3), whose interface contributesto the total activity capacity. At T > 0 �C, partitioning to cloud drop-lets in the LMT becomes significant approximately when theapparent air–water partition coefficient log KAW < �6.5 (absorptionof very low volatility chemicals MET, PFOA, 2,4-D, TMA) and/orwhen log KIW > �6 (adsorption of interface enriched chemicalsDIA, PFOA, PCP, 2,4-D).

The contribution of interface adsorption to the capacity of precip-itation decreases with increasing effective thickness (or specific sur-face) of raindrops and snow. While the effective thickness of snowdoes not increase much from cloud ice droplets to precipitatingsnow flakes at ground level, the diameter of rain droplets is on aver-age two orders of magnitude larger than the diameter of cloud drop-lets (Table S1, Supplementary material). The higher capacity of snowfor interface-enriched compounds (PFOA, PCP, DIA) explains thehigher 95%-ile of the B-values of precipitation in the LMT comparedto the B-values in the ABL (Fig. S1, Supplementary material), inagreement with the calculations by Lei and Wania (2004) for PCP.

4.4. Cloud trapping

During wet periods, precipitating droplets are essentially theonly water phase present in the ABL, whereas, in the LMT, non-pre-cipitating cloud microdroplets can hold a significant fraction ofnon-volatile and interface-enriched chemicals.

In the LMT, the surfactant PFOA is mostly adsorbed to cloudwater droplets, which have a higher activity capacity than precip-itating droplets, due to the smaller size. The fraction of PFOA trans-ported upwards to the LMT (5.85%) is almost entirely transportedaway (5.57%), while downward transport back to the ABL is small(0.15%) (Fig. 3), determining longer residence time of PFOA, as wellas of other interface-enriched substances (e.g. PCP Fig. 2), in theLMT. No significant reduction in wet deposition fluxes was calcu-lated for non-volatile compounds (e.g. MET, 2,4-D), indicating thatonly surface enhancement can decrease the efficiency of wet depo-sition. Cloud enrichment, driven by interface adsorption, was cal-culated by Lei and Wania (2004) for a-hexachlorocyclohexaneand some larger non-polar chemicals such as PCB-180 and ben-zo(a)pyrene, though this had only little impact on their overallatmospheric phase distribution because these two compoundswere predominantly found in particles. In the particular case ofPFOA, tRES,1 is proportional to the cloud water content and inver-sely proportional to the rain intensity (Table 3). Future modelingefforts with this compound would benefit from a refined parame-terization, differentiating the cloud water content during dry andwet periods. The introduction of a cloud water content specificfor wet periods, correlated to the rain intensity, would allow apply-ing a specific B-value for wet periods in the calculation of wetdeposition (Eq. (9)).

Absorption to clouds droplets decreases the fraction in the freegas phase available for oxidation by OH radicals, which explainsthe slightly longer tRES calculated for MET and 2,4-D in the LMTcompared to the ABL (Fig. 2), despite higher OH radicals concentra-tion in the LMT. It must be noticed that the assumption of phaseequilibrium between water (cloud water and precipitation) andsurrounding air may overestimate the fraction associated to theaqueous phase. Considering that cloud water has a mean life-timeof about 20 min (Pruppacher and Jaenicke, 1995), absorption tocloud water droplets may be too transitory to cause substantialshielding.

Interface partitioning and cloud enrichment depend on oftenuncertain adsorption coefficients and highly variable environmen-tal parameters (i.e. pH, cloud content, precipitation frequency),which increased the uncertainty range of estimated tRES for MET,2,4-D, PCP, and PFOA (Fig. 2). Comparing the effect of the cloudtrapping with the uncertainty in the underlying model assump-tions (e.g. phase equilibrium) and in the environmental and sub-stance parameters, the inclusion of cloud water in atmospherictransport models for high tier assessments of long range transportappears justified for very low volatility substances with logKAW < �7 (e.g. MET, 2,4-D) and, in particular, for persistent surfac-tants with log KIW > �6, such as PFOA. Despite the frequency ofionized substances among chemicals under regulatory scrutiny(Franco et al., 2010), the fraction of chemicals meeting these crite-ria and emitted to the atmosphere is probably small. The lack ofempirical interface adsorption coefficient for organic ions is a clearlimitation of the present modified version of MAMI. On the otherhand, the generation of such empirical data would not refine, inmost cases, the estimated atmospheric fate.

Acknowledgements

This work received financial support from the European Union6th Framework Program of Research, Thematic Priority 6 (Global

A. Franco et al. / Chemosphere 85 (2011) 1353–1359 1359

change and ecosystems), Contract Number GOCE 037017, ProjectOSIRIS.

The authors would like to thank Ralph Kühne for providing esti-mated molecular descriptors with ChemProp�, and Hans-ChristianLützhøft for valuable internal review.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.chemosphere.2011.07.056.

References

Armitage, J.M., MacLeod, M., Cousins, I.T., 2009. Modeling the global fate andtransport of perfluorooctanoic acid (PFOA) and perfluorooctanate (PFO) emittedfrom direct sources using a multispecies mass balance model. Environ. Sci.Technol. 43, 1134–1140.

Brandes, L.J., den Hollander, H., Van de Meent, D., 1996. Simplebox 2.0: a nestedmultimedia fate model for evaluating the environmental fate of chemicals.Report no. 719101029. National Institute of Public Health and the Environment,Bilthoven, the Netherlands.

Christie, A.O., Crisp, D.J., 1967. Activity coefficients on the n-primary, secondary andtertiary aliphatic amines in aqueous solution. J. Appl. Chem. 17, 11–14.

Donaldson, D.J., Valsaraj, K.T., 2010. Adsorption and reaction of trace gas-phaseorganic compounds on atmospheric water film surfaces: a critical review.Environ. Sci. Technol. 44, 865–873.

ECHA, 2008. REACH Guidance Documents: Guidance On Information RequirementsAnd Chemicals Safety Assessment. European Chemicals Agency, Helsinki,Finland.

Franco, A., Trapp, S., 2008. Estimation of the soil-water partition coefficientnormalized to organic carbon for ionizable organic chemicals. Environ.Toxicol. Chem. 27, 1995–2004.

Franco, A., Trapp, S., 2010. A multimedia activity model for ionizable compounds:validation study with 2,4-dichlorophenoxyacetic acid, aniline andtrimethoprim. Environ. Toxicol. Chem. 29, 789–799.

Franco, A., Ferranti, A., Davidsen, C., Trapp, S., 2010. An unexpected challenge:ionizable compounds in the REACH chemical space. Int. J. Life Cycle Assess. 15,321–325.

Goss, K.U., 1994. Predicting the enrichment of organic compounds in fog caused byadsorption on the water surface. Atmos. Environ. 28, 3513–23517.

Goss, K.U., 2004. The air/surface adsorption equilibrium of organic compoundsunder ambient conditions. Crit. Rev. Environ. Sci. Technol. 34, 229–389.

Hoff, J.T., Mackay, D., Gillham, R., Shiu, W.Y., 1993. Partitioning of organic chemicalsat the air–water interface in environmental systems. Environ. Sci. Technol. 27,2174–2180.

Jolliet, O., Hauschild, M., 2005. Modeling the influence of intermittent rain events onlong-term fate and transport of organic air pollutants. Environ. Sci. Technol. 39,4513–4522.

Ju, X., Jin, Y., Sasaki, K., Saito, N., 2008. Perfluorinated surfactants in surface,subsurface water and microlayer from Dalian coastal waters in China. Environ.Sci. Technol. 42, 3538–3542.

Kelly, C.P., Cramer, C.J., Truhlar, D.G., 2004. Predicting adsorption coefficients at air–water interfaces using universal solvation and surface area models. J. Phys.Chem. B 108, 12882–12897.

Lei, Y.D., Wania, F., 2004. In snow or rain a more efficient scavenger of organicchemicals? Atmos. Environ. 38, 3557–3571.

Li, H., Hellis, D., Mackay, D., 2007. Measurement of low air–water partitioncoefficients of organic acids by evaporation from a water surface. J. Chem. Eng.Data 52, 1580–1584.

MacLeod, M., Woodfine, D.G., Mackay, D., McKone, T., Bennet, D., Maddalena, R.,2001. BETR North America: a regionally segmented multimedia contaminantfate model for North America. Environ. Sci. Pollut. Res. 8, 156–163.

McMurdo, C.J., Ellis, D., Webster, E., Butler, J., Christensen, R.D., Reid, L.K., 2008.Aerosol enrichment of the surfactant PFO and mediation of the water–airtransport of gaseous PFOA. Environ. Sci. Technol. 42, 3969–3974.

Mmereki, B.T., Hicks, J.M., 2000. Adsorption of atmospheric gases at the air–waterinterface. 3: methylamines. J. Phys. Chem. 104, 10789–10793.

Poole, S.K., Poole, C.F., 1999. Chromatographic models for the sorption of neutralorganic compounds by soil from water and air. J. Chromatogr. A 845, 281–400.

Pruppacher, H.R., Jaenicke, R., 1995. The processing of water vapor and aerosols byatmospheric clouds, a global estimate. Atmos. Res. 38, 283–295.

Psillakis, E., Cheng, J., Hoffmann, M.R., Colussi, A.J., 2009. Enrichment factors ofperfluoroalkyl oxoanions at the air/water interface. J. Phys. Chem. A 113, 8826–8829.

Qureshi, A., MacLeod, M., Hungerbühler, K., 2009. Modeling aerosol suspensionfrom soils and oceans as sources of micropollutants to air. Chemosphere 77,495–500.

Roth, C.M., Goss, K.U., Schwarzenbach, R.P., 2002. Adsorption of a diverse set oforganic vapors on the bulk water surface. J. Colloid Interface Sci. 252, 21–30.

Roth, C.M., Goss, K.U., Schwarzenbach, R.P., 2004. Sorption of diverse organic vaporsto snow. Environ. Sci. Technol. 38, 4078–4084.

Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry and Physics. John Wiley,New York.

Trapp, S., Franco, A., Mackay, D., 2010. Activity-based concept for transport andpartitioning of ionizing organics. Environ. Sci. Technol. 44, 6123–6129.

Valsaraj, K.T., 1988. On the physico-chemical aspects of partitioning of non-polarhydrophobic organics at the air–water interface. Chemosphere 17, 875–887.

Valsaraj, K.T., Thibodeaux, L.J., 2010. On the physicochemical aspects of the globalfate and long-range transport of persistent organic pollutants. J. Phys. Chem.Lett. 1, 1694–1700.

Wania, F., Mackay, D., Li, Y.F., Bidleman, T.F., Strand, A., 1999. Global chemical fate ofa-hexachlorocyclohexane. 1. Evaluation of a global distribution model. Environ.Toxicol. Chem. 18, 1390–1399.

Webster, E., Ellis, D.A., Reid, L.K., 2010. Modeling the environmental fate ofperfluorooctanoic acid and perfluorooctanoate: an investigation of the role ofindividual species partitioning. Environ. Toxicol. Chem. 29, 1466–1475.

Winter, B., Faubel, M., Vacha, R., Jungwirth, P., 2009. Behaviour of hydroxide at thewater/vapor interface. J. Phys. Chem. Lett. 474, 241–247.