Article pubs.acs.org/est

Development and Evaluation of a New Sorption Model for Organic Cations in Soil: Contributions from Organic Matter and Clay Minerals Steven T. J. Droge*,†,‡ and Kai-Uwe Goss†,§ †

Department of Analytical Environmental Chemistry, Helmholtz Centre for Environmental Research−UFZ, Permoserstrasse 15, 04318 Leipzig, Germany ‡ Institute for Risk Assessment Sciences, Utrecht University, Yalelaan 104, 3508 TD Utrecht, The Netherlands § Institute of Chemistry, University of Halle-Wittenberg, Kurt Mothes Str. 2, 06120 Halle, Germany S Supporting Information *

ABSTRACT: This study evaluates a newly proposed cation-exchange model that defines the sorption of organic cations to soil as a summed contribution of sorption to organic matter (OM) and sorption to phyllosilicate clay minerals. Sorption to OM is normalized to the fraction organic carbon ( f OC), and sorption to clay is normalized to the estimated cation-exchange capacity attributed to clay minerals (CECCLAY). Sorption affinity is specified to a fixed medium composition, with correction factors for other electrolyte concentrations. The model applies measured sorption coefficients to one reference OM material and one clay mineral. If measured values are absent, then empirical relationships are available on the basis of molecular volume and amine type in combination with corrective increments for specific polar moieties. The model is tested using new sorption data generated at pH 6 for two Eurosoils, one enriched in clay and the other, OM, using 29 strong bases (pKa > 8). Using experimental data on reference materials for all tested compounds, model predictions for the two soils differed on average by only −0.1 ± 0.4 log units from measured sorption affinities. Within the chemical applicability domain, the model can also be applied successfully to various reported soil sorption data for organic cations. Particularly for clayish soils, the model shows that sorption of organic cations to clay minerals accounts for more than 90% of the overall affinity.

■

INTRODUCTION Many drugs, household products, and industrial substances occur beyond wastewater treatment plants in trace levels in receiving environments.1,2 Many specific classes of emerging contaminants in environmental systems, such as pharmaceuticals and personal care products,3,4 involve bases with a pKa above 7, which occur largely as cationic species in the environment. For example, about 45% of all pharmaceutical drugs contain a single base moiety. More than 70% of these bases have a pKa above 7, especially compounds that target the central nervous system5 and illicit drugs.6 About 14% of the chemicals preregistered for REACH are bases,7 of which about half are structures with a pKa above 7. Correspondingly, about 5% of the chemicals on the Canadian Domestic Substances List8 are bases with a pKa above 7. To assess properly the environmental fate properties for these strong bases, such as groundwater transport and chemical bioavailability, sorption affinities to environmental substrates are required.9 In the absence of adequate experimentally derived soil sorption affinities, risk assessement will rely heavily on predictive sorption models. The sorption models within current environmental risk assessment approaches, however, are mostly focused on sorption of neutral contaminants by partitioning into natural organic matter (NOM).10 Such sorption models © 2013 American Chemical Society

have been shown to be incapable of adequately predicting sorption coefficients to environmental substrates for bases that are largely protonated under environmentally relevant pH.11−15 For such compounds, the sorption affinity to soils may be strongly underestimated if only the fraction of neutral species is accounted for because the sorption affinity of cationic species can be in the same range as that of the corresponding neutral species.16−19 Neutral partitioning also differs both quantitatively and qualitatively from processes governed largely by electrostatic interactions.16,17 Furthermore, soil phases other than NOM are not considered as a relevant sorption phase in most sorption models. Organic cations are well-known to sorb to negatively charged environmental substrates like NOM16−22 as well as to phyllosilicate clay minerals.23−26 Finally, using neutral species as reference values in sorption models hampers the assessment of quaternary ammonium compounds (QACs), which are permanently charged organic cations that are commonly used in industrial products (biocides, corrosion Received: Revised: Accepted: Published: 14233

July 22, 2013 November 15, 2013 November 25, 2013 November 25, 2013 dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241

Environmental Science & Technology

Article

has to account for the influence of the molecular structure of the solute on the contribution from both electrostatic interactions and hydrophobic effects.17,23 Obviously, the cation-exchange capacity (CEC) is the soil parameter most closely related to the sorption affinity of a protonated base, and several studies have indeed applied sorption models for organic cations that normalize sorption to soil CEC.33,39−41 However, both NOM and clay minerals contribute to the CEC. In addition, one has to consider that CEC depends on the pH and on the specific clay mineral composition and that NOM may cover mineral phases such that some ion-exchange sites are blocked for organic cations. 30,31 Using cationic probe compounds may provide implicit measures of the abundance of cation-exchange sites under certain conditions,30,31 but this still does not distinguish between cation-exchange sites on NOM and clay.23 We propose a cation-exchange sorption model that estimates the CEC contribution from clay minerals using soil parameters CEC and f OC. This likely still oversimplifies the actual detailed sorption mechanisms, but it facilitates comparisons with the limited amount of reported soil sorption data for which these parameters are routinely provided. In addition, we have to acknowledge that the actual molecular interactions underlying the sorption processes for organic cations in soils are still not fully understood. As a starting point, we recognized that the CEC of peat samples42 and different humic acids43 occurs in a narrow range of ∼1.3−4.7 mol charge/kg dry weight (molC/kgdw), whereas different types of phyllosilicate clay minerals44 can vary in CEC between 0.010−1.5 molC/kgdw. Between pH 4 and 7, the difference in CEC for NOM (because of proton binding) is less than a factor of 343 and is most likely negligible for clay minerals Illite and bentonite. Below pH 4, the carboxylic acids in NOM become increasingly protonated, and specific differences between NOM may arise. Therefore, we assume a fixed NOM-based CEC (CECNOM) for pH conditions above 4. The CEC for Pahokee peat was reported to be 2 molC/kgdw,42 and because Pahokee peat was our reference NOM material in our previous work, we applied this value in our model. When applying a common conversion factor of 1.7 from f OC to dry weight of NOM ( f OC,NOM of 0.59), this results in a fixed NOMbased CEC of 3.4 molC/kgOC. (Note that these are crude assumptions for all possible soil types regarding origin, conditions, depths, and land uses.) The CEC contribution from clay minerals (CECCLAY) is then estimated as the total soil CEC (CECSOIL) minus the estimated CECNOM. As a result, the estimated CECCLAY in a soil (molC/kgdw,soil) equals (CECSOIL − 3.4f OC). The overall NOM-clay cation-exchange-based sorption affinity of an organic cation on a certain soil (Kd) is then modeled as

inhibitors, and ionic liquids), household chemicals (fabric softners),27,28 and some drugs. Sorption of organic cations to isolated NOM and clay mainly involves a cation-exchange process at negatively charged surface groups, as evidenced by the competitive interaction with both inorganic cations16,18,19,29 and organic cations.22,30 Reversible sorption to cation-exchange sites is also generally regarded as the main sorption process of organic cations to whole soils.31 NOM and clay are both expected to contribute to the sorption affinity of organic cations to soils,31−35 but the distinction between the relative contributions from both phases has, to our knowledge, never been clearly assessed. New sorption models are therefore needed for organic cations that focus on the ionexchange process, but consistent sorption data for organic cations have been too limited to derive mechanistically appropriate models.31,36 In our previous work, we applied a HPLC system to measure the sorption affinity of a wide variety of largely protonated amines (>95%) and permanently charged QACs to the NOM material Pahokee peat16,17 and the phyllosilicate clay minerals kaolinite, Illite, and bentonite.23 Those sets of data showed that the CEC-normalized sorption affinity of many compounds are in the same range for NOM and various clay minerals, indicating that clay sorption should be included in soil sorption models. Furthermore, specific differences between NOM and clay were observed with regard to how several polar moieties contribute to the overall affinity,23 indicating that new soil sorption models should address sorption to NOM and clay individually to improve accuracy. Interestingly, CEC-normalized sorption affinities to kaolinite, Illite, and bentonite (Kd,dry‑weight/CECclay) were within a factor of 3 for most tested compounds.23 These clay minerals are structurally quite different and together make up for most of the mineral fraction in soils.37 With sorption normalized to 1 mol of charged sorbent sites (molC), data on each of these minerals could function as a reference sorption coefficient (KCEC,CLAYS, in liters per molC) for the clay fraction in soils. Batch sorption studies with a cationic surfactant showed that the ion-exchange-based sorption coefficient to Pahokee peat (DOC,IE, in liters per kilogramOC) applied in the HPLC systems16,17 was comparable within a factor of 3 to other NOM materials.38 The first aim of the current study was to propose a cationexchange-based soil sorption model for organic cations based on the individual sorption models proposed for NOM17 and clay.23 To evaluate this model, the second aim was to obtain a consitent data set of sorption affinities to two natural reference soils (Eurosoils 1 and 5) for a wide selection of the same set of organic cations used earlier. This should elucidate whether measured soil affinities can be approximated by the summed contribution of measured reference values for isolated NOM (DOC,IE) and clay (KCEC,CLAYS). It may thereby also provide insight into, and allow for semiquantitative predictions of, the contribution of soil organic matter and mineral surfaces (phyllosilicates) to the overall soil sorption affinity for organic cations. Third, we compared sorption data for organic cations reported in the literature to the model predictions. Cation-Exchange-Based Soil Sorption Model. Defining Ion-Exchange Sorption Sites in a Refined Sorption Model for Organic Cations. Three main issues need to be resolved when developing a model based on the ion-exchange affinity of organic cations to soils. First, sorption sites should be identified quantitatively and qualitatively.31 Second, the role of aqueous medium composition has to be elucidated.16 Third, the model

Kd = K CEC,CLAYSCECCLAY + fOC DOC,IE = K CEC,CLAYS (CECSOIL − 3.4fOC ) + fOC DOC,IE

(1)

with all sorption coefficients derived at constant medium composition (e.g., pH 6 and 5 mM CaCl2). For sandy soils with very low fractions of clay and NOM, the crude assumption of a fixed CECNOM of 2 molC/kgdw,NOM may result in a negative CECCLAY. In such cases, the model approach considers that the NOM fraction fully dominates the soil CEC capacity, but evidently the uncertainty of the model outcome will be increased. Other uncertainties that limit the applicability domain for soil and medium properties in this model approach are the unknown sorbent properties of other soil phases, such 14234

dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241

Environmental Science & Technology

Article

as carbonaceous material (soot) and the presence of competing organic and inorganic cations. Correction Factors for the Medium Composition. Sorption of organic cations to NOM and clay is strongly impacted by dissolved cationic electrolytes, for example, it is more than a factor of 100 higher for the same organic cation at low NaCl (0.15 mM) compared to high CaCl2 (50 mM).16 Our previous work listed sorption data to NOM16,17 and clay23 at a single aqueous medium composition and provided empirical conversion factors for different medium compositions (e.g., approximately a log unit higher sorption in 15 mM NaCl compared to 5 mM CaCl2). In general, dissolved calcium concentrations dominate the impact of natural electrolytes, and sorption increases for NOM at most by a factor of 3 with a 10 times higher Ca2+ concentration16 as well as for clay (unpublished results). Because of the current absence of welldefined correction factors for electrolyte cations other than Na+ and Ca2+, the sorption model is thus currently limited to deal with soils with f OC above ∼0.5%, in medium with pH above 4, and electrolyte composition dominated by Ca2+. Calculations on the basis of modified Gouy−Chapman theories45 and related Donnan models46,47 may provide more theory behind these empirical observations, but they equally rely on crude assumptions and are likely too detailed to be included in common environmental fate models. Accounting for Solute Properties in a Refined Sorption Model for Organic Cations. Regarding the effect of the molecular structure on the ion-exchange affinity, our previous work lists ion-exchange-based sorption coefficients DOC,IE for NOM and KCEC,clay for clay for about 60 strong bases (pKa > 7) and QACs.16,17,23 These data may serve as reference values that can be applied directly to soils or used to align data from previous and future studies on reference materials to the current data set. Our previous studies could not yet convincingly relate observed sorption affinities for bases to single or multiple molecular descriptors.16,17,23 Still, several trends stood out in data sets for both NOM and clay. These allowed for a constructionist approach for molecular fragments, similar to the commonly used logP approach by Leo and Hansch48 only with a very limited data set. At a fixed medium composition, primary amine structures sorb stronger to NOM than QACs with the same molecular formula,16,17 whereas most QACs tend to sorb stronger to clay than to equally sized primary amines.23 The contribution of electrostatic interactions in the overall ion-exchange sorption affinity can therefore be considered as being partly due to competition with other dissolved cations and partly due to specific interactions between charged amine moieties and negatively charged surfaces. At a fixed medium composition, sorption to both NOM and clay increased with larger nonpolar alkyl chains and was, for example, decreased by the presence of polar amide groups (R1− C(O)NC−R2), both revealing specific hydrophobic effects. For simple organic cation structures lacking polar groups other than a benzene ring and charged nitrogen with molecular formula CXHYN, the sorption affinity of different amine types was modeled by a model based on volume (McGowan’s Vx; the calculation method is shown in the Supporting Information above Figure S7) and charged surface area (simplified by the number of hydrogens bound by the charged nitrogen, NAi):

Log K CEC,CLAYS = 1.22(± 0.15)Vx − 0.22(± 0.05)NAi + 1.09(± 0.05);

sy.x = 0.22, df = 29

(3)

For Log DOC,IE, the average sorption values were used for each compound for data at pH 4.5 and 7, both in 5 mM CaCl2. Log KCEC,clays data are from the average of all available CECnormalized clay data on kaolinite, Illite, and bentonite at pH 6, extrapolated to 5 mM CaCl2. In a next step, the models in eqs 2 and 3 for CXHYN amines, which are based on descriptors Vx and NAi, were used to derive corrective increments for specific polar moieties for organic cations. Analogous compounds with the same polar moieties were clustered according to their position relative to the VxNAi models (eqs 2 and 3). The derived empirical factors were mostly based on data for only four or less related structures and should be considered as preliminary values. The additivity of corrective increments for polar moieties as well as their position in relation to other moieties remains to be tested. The chemical applicability domain is therefore limited to molecular volumes (Vx) between 0.96 (benzylamine) and 3.79 (verapamil) and to those moieties that were represented in the data sets for NOM and clay, as outlined in ref and Supporting Information Table S6.

■

MATERIALS AND METHODS Soils, Chemicals, and Eluents. Standardized Eurosoils 1 and 5 were selected for testing in a HPLC column set up. Eurosoils are commonly used for soil sorption studies and soil leachate tests,33 and Eurosoils have been used in the same HPLC column set up in studies with neutral chemicals.49,50 Eurosoil 1 (ES-1) is the Eurosoil type with the lowest f OC (1.3%) and the highest clay content (75%), and 5 (ES-5) is the Eurosoil type with the highest f OC (9.3) and the lowest clay content (6%) (Supporting Information Table S1). Prolonged conditioning in the flow-through HPLC system should equilibrate pH and inorganic ions at the soil matrix according to the tested eluents. Several organic cations were selected from the set that was tested as used in our previous sorption studies with isolated NOM and clay. Most compounds were easy to detect by UV and were available as hydrochloride salts. Abbreviation codes, pKa (all >7.7), and ionic fractions at pH 6 (all >95%) are listed in Supporting Information Table S2. We applied the same coding for the chemicals for seven primary (1°) amines, eight secondary (2°) amines, 12 tertiary (3°) amines, and nine QACs (4°). 2-Methylbenzofuran was used as a polar neutral reference compound. Tables with CAS numbers and suppliers are provided in our previous NOM sorption study.17 Supporting Information Figure S1 shows structures of all cationic species. Injected test solutions with bases were prepared in eluent (pH 6, 5 mM CaCl2) from >100× diluted methanol stocks. For test compounds purchased as free bases, the pH was lowered by addition of HCl. Column Packing and Liquid Chromatography Set Up. Silicium carbide powder (SiC, diameter 3 ± 0.5 μm, ESKSiC118 Frechen, Germany) was used to dilute the soils so that all tested compounds could be eluted with fully aqueous eluent (for ES-1, 51:49 soil/SiC w/w, 9.8:90.2 and 4.9:95.1; for ES-5, 100% soil (∼100 mg) and 23:77 soil/SiC). SiC also served to stabilize the packing of clay fractions in the HPLC columns. A similar column (10 mm stainless steel with 2 mm i.d.) packed only with SiC was used to measure background retention apart from the soil. NaNO3 was used as nonretained tracer to

Log DOC,IE = 1.53(± 0.10)Vx + 0.32(± 0.04)NAi − 0.27(± 0.21);

sy.x = 0.26, df = 20

(2) 14235

dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241

Environmental Science & Technology

Article

peat NOM (Log KOC of 2.6, ref 17). As listed in Table S3, the Kd for some primary amines on ES-5 is considerably higher than to ES-1, whereas for many tertiary and quaternary amines, the sorption affinity to ES-1 is more than an order of magnitude higher than to ES-5. Applying eq 1, the estimated contribution of clay minerals to the soil CEC (CECCLAY) is 85% for ES-1 and only 4% for ES-5 (Table S1). These soils therefore present fairly extreme cases where sorption of organic cations to the soil is expected to be either dominated by NOM or by clay fraction. Table S4 again lists all Kd values for ES-1 and ES-5 along with sorption coefficients for each of the tested amines to NOM (logDOC,IE, ref 17) and to average phyllosilicate clay (log KCEC,CLAYS, ref 23). Figure 1 plots the observed soil Kd for ES-1

measure the void volume. Test solutions were injected onto the columns via a 0.2 or 0.02 mL loop by a Rheodyne 7125 injection valve. The aqueous mobile phase was pumped by a Jasco PU-980 HPLC pump at a flow rate of 0.1 mL/min to the injection valve and column. Pilot experiments showed no significantly higher retention times at lower flow rates. Eluting peaks of the tested compounds were detected by UV (Jasco 870-UV or Bischoff Lambda 1010). Single external calibration points on a spectrophotometer were used to convert UVabsorption signals to concentrations. Liquid Chromatography Conditions. The eluent was 5 mM CaCl2 prepared using MilliPore water (Milli-Q). The eluent was not actively buffered, but all were left standing for more than 1 day to reach stable pH of 6.1 ± 0.1. Soil sorption coefficients (Kd) were derived from retention times using the net eluent volume required to elute organic cations (in liters) divided by the dry weight of soil packed in the column (in kilograms). The minimum required net retention volume was set to be at least 150% of the sum of the tracer volume and SiC retention. Measurements were performed for at least three different concentrations, and peak maxima were used to estimate aqueous concentrations and corresponding sorbed concentrations. This allowed us to test for sorption linearity and the calculation of error margins for the derived soil sorption coefficients, as described in our earlier work.16,17

■

RESULTS AND DISCUSSION Sorption Data Quality. Sorption data obtained on columns with different dilutions of the same soil with SiC showed agreement within 0.1 log unit, as demonstrated in Figure S2 for 1° amine serotonin (P16) on Eurosoil 1 (ES-1) and for neutral reference compound 2-methylbenzofuran on Eurosoil 5 (ES-5). The sorption data used to derive soil sorption coefficients (Kd, in liters per kilogramdw) for 29 compounds on ES-1 are presented in Figure S3 and data for 31 compounds on ES-5, in Figure S4. Most compounds are tested with dissolved concentrations (estimated at peak maximum) ranging over 1 or 2 orders of magnitude (Table S3). All tested peak concentrations corresponded to sorbed concentrations well below 10%, and in most cases were below 1%, of the reported CEC for ES-1 and ES-5 (Figures S3 and S4), suggesting that the total available sorption sites as a whole were not saturated by the organic cations. The limited collected data for each compound indicate that linear sorption isotherms (Freundlich exponent nF of 1) describe the data for nearly all compounds adequately. This facilitates comparison among data obtained at different concentration ranges. Standard error margins for Kd are mostly below 0.1 log unit (Table S3), and standard deviation of the residuals of the fitted curve (sy.x) are mostly less than 0.2 log units except for some strongly sorbing compounds. This appears to be adequate for the goals of this study (i.e., to compare soil sorption coefficients with reference sorption values from NOM and clay and to develop a refined environmental risk assessment with meaningful descriptors and within reasonable uncertainty margins). Prediction of Cation Sorption to Eurosoils by the NOM Clay Cation-Exchange-Based Sorption Model (eq 1). Although soil CEC values are fairly equal for ES-1 and ES-5 (299 and 327 mmol/kg, respectively), the soil f OC in ES-5 is a factor of 7 higher than in ES-1. The polar neutral reference compound 2-methylbenzofuran is predicted to be within a factor of 2 of the observed sorption affinity to both soils by accounting only for measured sorption coefficient to Pahokee

Figure 1. Sorption data for Eurosoil 1 and 5 predicted by the cationexchange-based approach of eq 1 using measured sorption coefficients on reference materials for NOM and clay. Gray symbols are predictions on the basis of contributions from NOM reference material alone.

and ES-5 against the predictions from the cation-exchange model from eq 1 using only measured Pahokee peat DOC,IE values, average measured phyllosilicate KCEC,CLAYS values, f OC, and CECsoil. For the clayish soil ES-1, predicted Kd values are within a factor of 3, on average −0.1 log units (±0.3 SD, n = 29), from observed Kd values. In addition, for the organicmatter-rich soil ES-5, predicted Kd values show an excellent agreement (±0.4 SD) to observed Kd values. So, despite using only sorption coefficients from two reference materials and inevitable propagation of uncertainty margins on all measured coefficients, parameters, and empirical correction factors used, the use of eq 1 appears to be a promising starting point for a refined sorption model for organic cations. The improvement over models that apply a bulk soil CEC value as a parameter for 14236

dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241

Environmental Science & Technology

Article

under-represented in their calibration data set). Sorption to the clay-rich ES-1 scatters mostly within a factor of ±10. This may relate to the use of soil sorption data as calibration values, which include the contribution from clay minerals to overall soil sorption. The pKa-based approach aimed to include QACs, but this appears to be inadequate. The reference material-based ion-exchange approach of modeling sorption of organic cations appears to be superior to the currently available predictive models that are based on molecular descriptors. The model in eq 1 also applies a set of empirical, yet mechanistically meaningful, parameters to describe the sorption process. Prediction of Reported Soil Sorption Data by the NOM Clay Cation-Exchange-Based Sorption Model (eq 1). The functionality of the proposed cation-exchange-based sorption model is further tested by comparing reported soil sorption data for highly protonated bases (test pHsoil < pKa −1) with predicted values. The most accurate predictions are expected when (i) the reference sorption coefficients (DOC,IE and KCEC,CLAYS) are available or the chemical structures are at least within the chemical applicability domain for eqs 2 and 3, (ii) the medium conditions are readily extrapolated to those used for the reference values (pH 4−7, Ca2+ present as dominant cation), and (iii) the soil properties f OC and CECsoil are measured appropriately. The Footprint database, for example, which was extensively used by Franco and Trapp,11 does not list CEC values with the sorption data, whereas other soption studies report electrical conductivity instead of CEC51,52 or are only focused on f OC. Detailed sorption data are available for six pharmaceutical bases with a pKa > 8 onto 12 different Australian soils and sediments with pH < 7.5.53 All data for atenolol (pKa 9.6), metoprolol (pKa 9.7), propranolol (pKa 9.6), imipramine (pKa 9.5), and verapamil (pKa 8.9) could be predicted using reference DOC,IE and KCEC,CLAYS values. For propranolol, reference values were available, but the data for only one soil were reported. For chlorpheniramine (pKa 9.1) and promethazine (pKa 9.6), no measured reference values were available, but these were estimated using the VxNAi models in eqs 2 and 3 with corrective increments for specific polar moieties. The 12 soils varied in f OC (0.08−8.6%), and the CECCLAY, calculated according to eq 1, ranges from 19 to 90% for 10 soils. For the soil Mount Shank (f OC 7%), a negative CECCLAY was calculated, and for soil A51, a CEC of 0 was reported. A complicating factor in comparing reported with predicted values is that the sorption data for several of the cations from the Australian soil study showed considerable nonlinearity on some soils (Freundlich exponent 1.4), whereas, for example, for metoprolol all exponents were within 0.83−1.18. Because the Australian soil study applied limited test ranges with initial concentration of 0.5−5 mg/L but reported Kd values at 1 μg/L, we recalculated sorption coefficients at 1 mg/L. Figure 2 plots data for the compounds for which we could use both NOM and clay reference values to predict the sorption coefficients to each soil. Apart from a few outliers, nearly all predicted values are within a factor of 3 of the observed sorption coefficients: on average for atenolol, −0.07 log units; for metoprolol, −0.16; for imipramine, −0.57; and for verapamil, −0.1. We also plot predicted soil sorption affinities based on only NOM reference values, as was done in Figure 1, to estimate the contribution of sorption to clay. For most soils, these cations sorbed approximately 10 times stronger because of the estimated contribution of clay. Predictions for these four compounds based on estimated reference values, using eqs 2 and 3 and

organic cations is that our model includes the specific affinity of individual (types of) organic cations for organic matter or clay. Figure 1 also plots the observed soil Kd for ES-1 and ES-5 against the predicted Kd based on only sorption to NOM for the cations ( f OCDOC,IE), as is generally advised by risk assessment guidelines if measured sorption coefficients to NOM are available. For the organic-matter-rich soil ES-5 where the contribution to CEC sites by clay is almost negligible (4%), Figure 1 shows that this would have provided excellent predictions. It should be noted, however, that this would have required the inclusion of measured NOM data at the same medium composition as that used in the test with soil because NOM sorption data at, for example, 15 mM NaCl can be more than an order of magnitude higher than those obtained at 5 mM CaCl2.16 For the clayey soil ES-1 (CECCLAY of 85%), neglecting clay as a sorption phase would have strongly underestimated soil sorption affinities by a factor of 10−1000 for most compounds even if the NOM sorption data for the correct CaCl2 concentration had been used. The results from ES-1 further demonstrate that it is inaccurate to use sorption data from clayey soils to extrapolate to KOC values for chemicals that exist largely as organic cations. Because sorption coefficients on natural soils have been the only data source to develop models for ionic chemicals in the past, the inclusion of data from clayey soils may strongly impair the quality of such models.11 Equation 1 may be included as a sorption model in future environmental risk assessment guidelines as a function for protonated bases and QACs, whereas the contribution of neutral base species can be calculated according to existing regressions between molecular descriptors and KOC and the soil property f OC. The cation-exchange model in eq 1 requires values for DOC,IE and KCEC,CLAYS. If such values are not available, then they can be estimated using the VxNAi models in eqs 2 and 3 for NOM and clay, respectively, with additional corrective increments for polar moieties (listed in ref 23 and Table S6) if the organic cation fits the chemical applicability domain. Prediction of Cation Sorption to Eurosoils by Other Sorption Models. Supporting Information Figures S5 and S6 plot the Eurosoil data against the predictions from several currently available sorption models, some of which specifically addressing sorption of organic cations. More details on the comparisons are presented along with Supporting Information Figures S5 and S6. Briefly, the KOW-based approach in the EUTGD (in appendix XI on ionizing substances in ref 10) strongly underestimate sorption to ES-1 and ES-5 because it neglects the contribution of ionized species as well as sorption to clay. In addition, EPISuite’s KocWIN values for neutral species (assuming KOC,neutral = KOC,cation) are poor predictors for both soils. The recently proposed ECETOC Log DOW approach for bases15 includes the contribution of ionic species by using Log DOW values estimated by algorithms from SPARC or ACD/ Laboratories, which calculate KOW values for ionic species as well. The DOW regression, however, was largely derived from sorption data obtained with sewage sludge and strongly underestimates sorption to clay-rich soil ES-1. Interestingly, sorption to OM-rich ES-5 is predicted within a factor of ±10. A pKa-based approach by Franco and Trapp,11 with a KOW-related correction factor, is largely calibrated with soil sorption data. For the OM-rich ES-5, sorption of most compounds is even overestimated by a factor 10−100, probably because of the strong dependency of the model on pKa (strong bases were 14237

dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241

Environmental Science & Technology

Article

Figure 3. Soil sorption data for several highly protonated bases from various studies compared to the cation-exchange model predicted values using eq 1 (mostly applying eqs 2 and 3 to first estimate sorption to reference materials). Gray and pink symbols are predictions based on contributions from NOM alone. Green symbols are predicted using estimated reference sorption coefficients for NOM and clay. Blue symbols are from a single study54 with only estimated reference values for ifenprodil.

Figure 2. Sorption data from ref 53 for several highly protonated bases on 12 Australian soils compared to the cation-exchange model predictions with eq 1 using measured sorption coefficients on reference materials. Gray symbols are predictions based on contributions from NOM reference material alone.

moiety, and decarboxylated enrofloxacin (pKa = 6.9, Vx = 2.4), for which appropriate corrective increments to deal with the multifunctional moieties were missing. Blue symbols are from a single study54 from which the reported sorption isotherms were linearized to obtain sorption coefficients. Reference values for NOM and clay were available for fluoxetine, propranolol, and atenolol, whereas they were estimated for ifenprodil using VxNAi eqs 2 and 3 and corrective increments. Figure 3 again suggests that sorption coefficients for organic cations in most studied soils are underestimated when accounting for only sorption to NOM. Outlook for Further Improvements and Application of Cation-Exchange-Based Model Approach. The empirical cation-exchange model based on eqs 1−3 seems to make sense mechanistically in terms of qualitatively and quantitatively covering the most relevant processes and interactions. Supporting Information section S3 provides an overview of the most relevant equations and rules of thumb for the proposed soil sorption model for organic cations. We have shown for a wide variety of structures and soils that by including specific differences in sorption affinities to organic matter and clay most model predictions are within acceptable error margins, especially regarding the underlying pragmatic assumptions. On the basis of two measured reference values, the currenly proposed model for organic cations provides a simple framework to update sorption models in risk assessment guidelines,10 in fate models such as SimpleTreat,15 and in environmental fate studies on ionizable compounds,55,56 which so far have applied models that may not have been appropriately parametrized for organic cations. Furthermore, following the equilibrium partitioning theory, predicted no effect concentration (PNEC) in soils may be derived form aquatic toxicity data if effective pore water concentrations are calculated from adequately predicted sorption coefficients.56−61 Clearly, the presented cation-exchange model is still far from a fully predictive sorption model based on molecular descriptors only and relies on crude assumptions about how to account for specific soil properties. Our attempts to create descriptor-based

appropriate corrective increments, are slightly less accurate but are still mostly within a factor of 10 of reported values (Supporting Information Figure S7). This, of course, is not an ideal independent set of organic cations to test the fragmentbased sorption model approach because these four compounds were also used to derive the applied corrective increments. Reference values for chlorpheniramine and promethazin were not measured in our previous work, and estimations are made using eqs 2 and 3. Again, predictions for these two largely protonated bases are scattered within a factor of 10 of observed sorption coefficients: on average, +0.23 log units for chlorpheniramine and −0.33 log units for promethazin (Supporting Information Figure S8). For these compounds, however, there were almost no corrective increments needed because the structures are almost according to the CXHYN formula and the predicted values are dominated by size (Vx) and amine type (NAi). Several other reported soil sorption data for highly protonated bases were checked, as shown in Figure 3. Details on the organic cations used in Figure 3 are presented in Supporting Information Table S5. The large single-base antibiotic tylosin is a good example of a chemical outside of the chemical applicability domain (pKa = 7.7 and Vx = 7.03, whereas the maximum Vx in calibration sets for eqs 2 and 3 was 3.7 and there is no corrective increment for the large ring structure). All green symbols plotted in Figure 3 are data obtained for compounds for which reference values for NOM and clay had to be estimated from studies performed in 10 mM CaCl2 and with CEC and f OC reported. Most sorption data for these compounds on various soils are predicted within a factor of 10 by the cation-exchange model. Exceptions of predicted values that differed by more than a factor of 10 are fenpropimorph (pKa = 7, Vx = 2.8), which should be within the chemical applicability domain with only a single ether 14238

dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241

Environmental Science & Technology

Article

pKa of 3−8 and less for the stronger bases discussed in the current work. There are substantial numbers of studies and consistent data sets relating molecular descriptors to the partitioning of neutral species toward NOM (see the Supporting Information section on equations and rules of thumb). Improved modeling for neutral base species may move from K OW -based approaches, or fragment contribution approaches, toward thermodynamically and mechanistically more refined polyparameter linear solvation-energy relationships (ppLFERs50,66) and contribution of specific alternative soil substrates (black carbon and, as shown for atrazine67 and nonionic surfactants,68 clay).

models for sorption to both NOM and clay, on the basis of data for >60 different structures, were so far not successful, but we welcome any new initiatives to re-evaluate our data, which are detailed in the Supporting Information sections of our previous work. The availability of only a small set of poorly validated corrective increments to estimate DOC,IE and KCEC,CLAYS reference values with eqs 2 and 3 implies that these need to be continuously updated and extended with new data to cover more complex multifunctional structures that are common in many ionizable pharmaceuticals and pesticides. Alternatively, direct measurement of the two reference values, DOC,IE and KCEC,CLAYS, should give suitable predictions on the soil sorption affinities for a specific organic cation structure in the whole range of soil properties that need to be included for environmental risk assessment. Measuring D OC,IE and KCEC,CLAYS for probe compounds for a certain class of cationic structures may serve to facilitate extrapolation to all analogue structures,31 again requiring that the remaining corrective increments are available. Regarding the contribution of different soil properties, the successful prediction of sorption data on various soil types indicates that the pragmatic approach of estimating CECCLAYS from CECSOIL and f OC is adequate. The cation-exchange process, however, is more complex than just summing up the sorption to two soil components, especially regarding to the influence of pH, medium electrolytes, and competing organic cations. Future modeling strategies could further explore the extension of the NICA− Donnan approach into the domain of organic cations and whole soils.46,47 NICA−Donnan quantifies those specific effects and qualifies the different interactions with, and properties of, NOM sorption sites to predict the sorption of inorganic metal cations. As briefly discussed in other work with cationic surfactants38 and pesticides,62 the NICA−Donnan approach explains sorption of cations as a combination of an electrostatic model and a competition model. The competition model is fully corrected for contributions of electrostatic attraction and therefore operates with intrinsic sorption coefficients of sorbates. This could solve the dependency of our model concept to work in constant medium composition and application of empirical correction factors between different media (e.g., ∼1 log unit lower sorption affinity for both NOM and clay in 5 mM CaCl2 solution compared to 15 mM NaCl). Still, as long as all reference sorption values are appropriately normalized to a single medium composition, the relative difference in sorption affinities between organic cations is similar to those for intrinsic affinities. For common environmental risk assessment guidelines, however, such a detailed approach of the sorption mechanisms requires too many input parameters. Each additional parameter will contribute to uncertainty margins, hampering adequate risk assessment more than elucidating the actual sorptive properties of contaminants. The current model has not been tested for multivalent organic cations or zwitterionic species, which require additional consistent data sets and detailed insight into the processes underlying the observed sorption affinities. Covalent binding of organic bases to soil and humic acid has been demonstrated and studied in detail35,39,63,64 but is most likely only specific for the neutral species of weakly basic aniline moieties65 and is of only minor concern or even nonexisting for organic cations in general.22,31,39 Finally, we did not discuss any improvement on the prediction of neutral base species, which may be specifically relevant for bases that have an intermediate

■

ASSOCIATED CONTENT

S Supporting Information *

Chemical structures and properties, details on data quality, fitted isotherm parameters and plotted sorption isotherms, and detailed model evaluations. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: +31-30-2535217; fax: +31-30-2535077; e-mail: steven. [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This research was supported by a grant from APAG (The European Oleochemicals & Allied Products Group), a sector group of the CEFIC (European Chemical Industry Council).

■

REFERENCES

(1) Tolls, J. Sorption of veterinary pharmaceuticals in soils: A review. Environ. Sci. Technol. 2001, 35, 3397−3406. (2) Thiele-Bruhn, S. Pharmaceutical antibiotic compounds in soils − a review. J. Plant Nutr. Soil Sci. 2003, 166, 145−167. (3) Ternes, T. A.; Joss, A.; Siegrist, H. Scrutinizing pharmaceuticals and personal care products in wastewater treatment. Environ. Sci. Technol. 2004, 38, 392−399. (4) Brooks, B. W.; Riley, T. M.; Taylor, R. D. Water quality of effluent-dominated ecosystems: Ecotoxicological, hydrological, and management considerations. Hydrobiologia 2006, 556, 365−379. (5) Manallack, D. T. The pKa distribution of drugs: Application to drug discovery. Perspect. Med. Chem. 2007, 1, 25−38. (6) Daughton, C. G. Illicit drugs: Contaminants in the environment and utility in forensic epidemiology. Rev. Environ. Contam. Toxicol. 2011, 210, 59−110. (7) Franco, A.; Ferranti, A.; Davidsen, C.; Trapp, S. An unexpected challenge: Ionizable compounds in the REACH chemical space. Int. J. Life Cycle Assessm. 2010, 15, 321−325. (8) Rayne, S.; Forest, K. Dow and Kaw,eff vs. Kow and Kaw: Acid/ base ionization effects on partitioning properties and screening commercial chemicals for long-range transport and bioaccumulation potential. J. Environ. Sci. Health, Part A: Toxic/Hazard. Subst. Environ. Eng. 2010, 45, 1550−1594. (9) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental Organic Chemistry; John Wiley & Sons: New York, 2003. (10) European Commission Technical Guidance Document on Risk Assessment: Part II Environmental Risk Assessment; Technical Report from the Institute for Health and Consumer Production, European Commission. Joint Research Centre, European Chemicals Bureau, 2006.

14239

dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241

Environmental Science & Technology

Article

(32) Ter Laak, T. L.; Gebbink, W. A.; Tolls, J. The effect of pH and ionic strength on the sorption of sulfachloropyridazine, tylosin, and oxytetracycline to soil. Environ. Toxicol. Chem. 2006, 25, 904−911. (33) Thomas, P. C.; Velthoven, K.; Geurts, M.; van Wijk, D. Bioavailability and detoxification of cationics: II. Relationship between toxicity and CEC of cationic surfactants on Caenorhabditis elegans (Nematoda) in artificial and natural substrates. Chemosphere 2009, 75, 310−318. (34) Li, H.; Lee, L. S.; Fabrega, J. R.; Jafvert, C. T. Role of pH in partitioning and cation exchange of aromatic amines on watersaturated soils. Chemosphere 2001, 44, 627−635. (35) Fábrega, J. R.; Jafvert, C. T.; Li, H.; Lee, L. S. Modeling shortterm soil-water distribution of aromatic amines. Environ. Sci. Technol. 1998, 32, 2788−2794. (36) Franco, A.; Wenjing, F. U.; Trapp, S. Influence of soil pH on the sorption of ionizable chemicals: Modeling advances. Environ. Toxicol. Chem. 2009, 28, 458−464. (37) Essington, M. E. Soil and Water Chemistry: An Integrative Approach; CRC Press: Boca Raton, FL, 2004. (38) Chen, Y.; Hermens, J. L. M.; Droge, S. T. J. Influence of organic matter type and medium composition on the sorption affinity of C12benzalkonium cation. Environ. Pollut. 2013, 179, 153−159. (39) Lee, L. S.; Nyman, A. K.; Li, H.; Nyman, M. C.; Jafvert, C. Initial sorption of aromatic amines to surface soils. Environ. Toxicol. Chem. 1997, 16, 1575−1582. (40) Brown, D. S.; Combs, G. A modified Langmuir equation for predicting sorption of methylacridinium ion in soils and sediments. J. Environ. Qual. 1985, 14, 195−199. (41) Fábrega, J. R.; Jafvert, C. T.; Li, H.; Lee, L. S. Modeling competitive cation exchange of aromatic amines in water-saturated soils. Environ. Sci. Technol. 2001, 35, 2727−2733. (42) Lyon, W. G. Swelling of peats in liquid methyl, tetramethylene and propyl sulfoxides and in liquid propyl sulfone. Environ. Toxicol. Chem. 1995, 14, 229−236. (43) Milne, C. J.; Kinniburgh, D. G.; Tipping, E. Generic NICADonnan model parameters for proton binding by humic substances. Environ. Sci. Technol. 2001, 35, 2049−2059. (44) Hassellöv, M.; Lyvén, B.; Bengtsson, H.; Jansen, R.; Turner, D. R.; Beckett, R. Particle size distributions of clay-rich sediments and pure clay minerals: A comparison of grain size analysis with sedimentation field-flow fractionation. Aquat. Geochem. 2001, 7, 155−172. (45) Chang, F.-.C.; Sposito, G. The electrical double layer of a diskshaped clay mineral particle: Effect of particle size. J. Colloid Interface Sci. 1994, 163, 19−27. (46) Tertre, E.; Ferrage, E.; Bihannic, I.; Michot, L. J.; Prêt, D. Influence of the ionic strength and solid/solution ratio on Ca(II)-forNa+ exchange on montmorillonite. Part 2: Understanding the effect of the m/V ratio. Implications for pore water composition and element transport in natural media. J. Colloid Interface Sci. 2011, 363, 334−347. (47) Kinniburgh, D. G.; Van Riemsdijk, W. H.; Koopal, L. K.; Borkovec, M.; Benedetti, M. F.; Avena, M. J. Ion binding to natural organic matter: Competition, heterogeneity, stoichiometry and thermodynamic consistency. Colloids Surf. A 1999, 151, 147−166. (48) Leo, A. J.; Hansch, C. Substituent Constants for Correlation Analysis in Chemistry and Biology; John Wiley & Sons: New York, 1979. (49) Bronner, G.; Goss, K. U. Sorption of organic chemicals to soil organic matter: Influence of soil variability and pH dependence. Environ. Sci. Technol. 2011, 45, 1307−1312. (50) Bronner, G.; Goss, K. U. Predicting sorption of pesticides and other multifunctional organic chemicals to soil organic carbon. Environ. Sci. Technol. 2011, 45, 1313−1319. (51) Drillia, P.; Stamatelatou, K.; Lyberatos, G. Fate and mobility of pharmaceuticals in solid matrices. Chemosphere 2005, 60, 1034−1044. (52) Achtenhagen, J.; Kreuzig, R. Laboratory tests on the impact of superabsorbent polymers on transformation and sorption of xenobiotics in soil taking 14C-imazalil as an example. Sci. Total Environ. 2011, 409, 5454−5458.

(11) Franco, A.; Trapp, S. Estimation of the soil-water partition coefficient normalized to organic carbon for ionizable organic chemicals. Environ. Toxicol. Chem. 2008, 27, 1995−2004. (12) Hörsing, M.; Ledin, A.; Grabic, R.; Fick, J.; Tysklind, M.; Jansen, J. l. C.; Andersen, H. R. Determination of sorption of seventy-five pharmaceuticals in sewage sludge. Water Res. 2011, 45, 4470−4482. (13) Bi, E.; Schmidt, T. C.; Haderlein, S. B. Sorption of heterocyclic organic compounds to reference soils: Column studies for process identification. Environ. Sci. Technol. 2006, 40, 5962−5970. (14) Stein, K.; Ramil, M.; Fink, G.; Sander, M.; Ternes, T. A. Analysis and sorption of psychoactive drugs onto sediment. Environ. Sci. Technol. 2008, 42, 6415−6423. (15) Franco, A.; Struijs, J.; Gouin, T.; Price, O. R. Evolution of the sewage treatment plant model SimpleTreat: Applicability domain and data requirements. Integr. Environ. Assess. Manage. 2013, 9, 560−568. (16) Droge, S. T. J.; Goss, K. U. Effect of sodium and calcium cations on the ion-exchange affinity of organic cations for soil organic matter. Environ. Sci. Technol. 2012, 46, 5894−5901. (17) Droge, S. T. J.; Goss, K. U. Ion-exchange affinity of organic cations to natural organic matter: Influence of amine type and nonionic interactions at two different pHs. Environ. Sci. Technol. 2013, 47, 798−806. (18) Sibley, S. D.; Pedersen, J. A. Interaction of the macrolide antimicrobial clarithromycin with dissolved humic acid. Environ. Sci. Technol. 2008, 42, 422−428. (19) Karthikeyan, K. G.; Chorover, J. Humic acid complexation of basic and neutral polycyclic aromatic compounds. Chemosphere 2002, 48, 955−964. (20) Chen, Y.; Droge, S. T. J.; Hermens, J. L. M. Analyzing freely dissolved concentrations of cationic surfactant utilizing ion-exchange capability of polyacrylate coated solid-phase microextraction fibers. J. Chromatogr. A 2012, 1252, 15−22. (21) Ishiguro, M.; Tan, W.; Koopal, L. K. Binding of cationic surfactants to humic substances. Colloids Surf. A 2007, 306, 29−39. (22) Richter, M. K.; Sander, M.; Krauss, M.; Christl, I.; Dahinden, M. G.; Schneider, M. K.; Schwarzenbach, R. P. Cation binding of antimicrobial sulfathiazole to leonardite humic acid. Environ. Sci. Technol. 2009, 43, 6632−6638. (23) Droge, S. T. J.; Goss, K. U. Sorption of organic cations to phyllosilicate clay minerals: CEC-normalisation, salt dependency, and the role of electrostatic and hydrophobic effects. Environ. Sci. Technol., in press; DOI: 10.1021/es403187w. (24) Polubesova, T.; Nir, S. Modeling of organic and inorganic cation sorption by Illite. Clays Clay Miner. 1999, 47, 366−374. (25) Polubesova, T.; Rytwo, G.; Nir, S.; Serban, C.; Margulies, L. Adsorption of benzyltrimethylammonium and benzyltriethylammonium on montmorillonite: Experimental studies and model calculations. Clays Clay Miner. 1997, 45, 834−841. (26) Rytwo, G.; Nir, S.; Margulies, L. Interactions of monovalent organic cations with montmorillonite − adsorption studies and modelcalculations. Soil Sci. Soc. Am. J. 1995, 59, 554−564. (27) Li, X.; Brownawell, B. J. Quaternary ammonium compounds in urban estuarine sediment environments − a class of contaminants in need of increased attention? Environ. Sci. Technol. 2010, 44, 7561− 7568. (28) Lara-Martin, P. A.; Li, X.; Bopp, R. F.; Brownawell, B. J. Occurrence of alkyltrimethylammonium compounds in urban estuarine sediments: Behentrimonium as a new emerging contaminant. Environ. Sci. Technol. 2010, 44, 7569−7575. (29) Brownawell, B. J.; Chen, H.; Collier, J. M.; Westall, J. C. Adsorption of organic cations to natural materials. Environ. Sci. Technol. 1990, 24, 1234−1241. (30) MacKay, A. A.; Seremet, D. E. Probe compounds to quantify cation exchange and complexation interactions of ciprofloxacin with soils. Environ. Sci. Technol. 2008, 42, 8270−8276. (31) MacKay, A. A.; Vasudevan, D. Polyfunctional ionogenic compound sorption: Challenges and new approaches to advance predictive models. Environ. Sci. Technol. 2012, 46, 9209−9223. 14240

dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241

Environmental Science & Technology

Article

(53) Williams, M.; Ong, P. L.; Williams, D. B.; Kookana, R. S. Estimating the sorption of pharmaceuticals based on their pharmacological distribution. Environ. Toxicol. Chem. 2009, 28, 2572−2579. (54) Yamamoto, H.; Nakamura, Y.; Moriguchi, S.; Nakamura, Y.; Honda, Y.; Tamura, I.; Hirata, Y.; Hayashi, A.; Sekizawa, J. Persistence and partitioning of eight selected pharmaceuticals in the aquatic environment: Laboratory photolysis, biodegradation, and sorption experiments. Water Res. 2009, 43, 351−362. (55) van Zelm, R.; Stam, G.; Huijbregts, M. A. J.; van de Meent, D. Making fate and exposure models for freshwater ecotoxicity in life cycle assessment suitable for organic acids and bases. Chemosphere 2013, 90, 312−317. (56) Trapp, S.; Franco, A.; Mackay, D. Activity-based concept for transport and partitioning of ionizing organics. Environ. Sci. Technol. 2010, 44, 6123−6129. (57) Droge, S. T. J.; Postma, J. F.; Hermens, J. L. M. Sediment toxicity of a rapidly biodegrading nonionic surfactant: Comparing the equilibrium partitioning approach with measurements in pore water. Environ. Sci. Technol. 2008, 42, 4215−4221. (58) Rico-Rico, Á ; Temara, A.; Hermens, J. L. M. Equilibrium partitioning theory to predict the sediment toxicity of the anionic surfactant C12-2-LAS to Corophium volutator. Environ. Pollut. 2009, 157, 575−581. (59) Di Toro, D. M.; Zarba, C. S.; Hansen, D. J.; Berry, W. J.; Swartz, R. C.; Cowan, C. E.; Pavlou, S. P.; Allen, H. E.; Thomas, N. A.; Paquin, P. R. Technical basis for establishing sediment quality for nonionic organic-chemicals using equilibrium partitioning. Environ. Toxicol. Chem. 1991, 10, 1541−1583. (60) Jager, T. Modeling ingestion as an exposure route for organic chemicals in earthworms (Oligochaeta). Ecotoxicol. Environ. Saf. 2004, 57, 30−38. (61) Kraaij, R.; Mayer, P.; Busser, F. J. M.; Van Het Bolscher, M.; Seinen, W.; Tolls, J.; Belfroid, A. C. Measured pore-water concentrations make equilibrium partitioning work − a data analysis. Environ. Sci. Technol. 2003, 37, 268−274. (62) Iglesias, A.; Lopez, R.; Gondar, D.; Antelo, J.; Fiol, S.; Arce, F. Effect of pH and ionic strength on the binding of paraquat and MCPA by soil fulvic and humic acids. Chemosphere 2009, 76, 107−113. (63) Gulkowska, A.; Krauss, M.; Rentsch, D.; Hollender, J. Reactions of a sulfonamide antimicrobial with model humic constituents: Assessing pathways and stability of covalent bonding. Environ. Sci. Technol. 2012, 46, 2102−2111. (64) Parris, G. E. Covalent binding of aromatic amines to humates. 1. Reactions with carbonyls and quinones. Environ. Sci. Technol. 1980, 14, 1099−1106. (65) Weber, E. J.; Colón, D.; Baughman, G. L. Sediment-associated reactions of aromatic amines. 1. Elucidation of sorption mechanisms. Environ. Sci. Technol. 2001, 35, 2470−2475. (66) Goss, K.-U.; Schwarzenbach, R. P. Linear free energy relationships used to evaluate equilibrium partitioning of organic compounds. Environ. Sci. Technol. 2001, 35, 1−9. (67) Laird, D. A.; Fleming, P. D. Mechanisms for adsorption of organic bases on hydrated smectite surfaces. Environ. Toxicol. Chem. 1999, 18, 1668−1672. (68) Droge, S. T. J.; Yarza-Irusta, L.; Hermens, J. L. M. Modeling nonlinear sorption of alcohol ethoxylates to sediment: The influence of molecular structure and sediment properties. Environ. Sci. Technol. 2009, 43, 5712−5718.

14241

dx.doi.org/10.1021/es4031886 | Environ. Sci. Technol. 2013, 47, 14233−14241