Geochemical Multisurface Modeling of Reactive Zinc Speciation in Compost as Influenced by Extraction Conditions.

Knowledge on organic matter (OM) concentration and composition is of major importance for predicting Zn speciation and bioavailability in soils, especially for low-Zn soils. However, comprehensive knowledge on the effect of soil-like organic amendments such as compost on metal speciation is limited. For the first time, multisurface modeling is applied on compost to study the effect of solid and dissolved OM composition on the speciation of reactive Zn as influenced by conditions applied in frequently used extractions to estimate Zn bioavailability. First, compost OM composition was determined by fractionation in operationally defined humic, fulvic, and hydrophilic acid pools under various extraction conditions, and subsequently, Zn speciation was modeled using the generic non-ideal competitive adsorption-Donnan (NICA-Donnan) model in addition to adsorption to hydrous ferric oxide (HFO) and clay. The results show a strong effect of extraction conditions on OM concentration and composition and related dissolved Zn speciation. Model predictions show that Zn in solution is mainly bound to dissolved humic acids. Analysis of deviations between measured and modeled Zn concentrations reveal specific limitations of the current generic model parameters, particularly with regard to Zn binding to OM at low concentrations and Ca-Zn competition, that is, typical conditions that occur in low-Zn soils.

Introduction

Multisurface models are important in predicting and understanding metal behavior in soils.[1−3] These models assume the soil solid phase to consist of independent reactive surfaces such as clay minerals, iron oxides, and organic matter[4] and can be used to predict metal speciation over the reactive surfaces and solution species under equilibrium conditions. Previously, these models have successfully been applied to predict metal speciation in contaminated soils[5,6] and recently in metal deficient soils also, in particular for zinc (Zn).[7] Since more than 30% of the soils worldwide contain suboptimal levels of plant available Zn,[8] improving our understanding of the mechanisms that control soil Zn speciation in such low-Zn soils is of major importance to optimize management practices. A multisurface modeling study has recently shown that the organic matter (OM) plays a key role in controlling Zn speciation in these soils.[7] In addition, several studies have shown that complexation of Zn with organic ligands in soil solution can increase plant uptake of Zn significantly.[9,10] Especially at high pH levels and low Zn concentrations, which are typical characteristics of low-Zn soils, Zn complexation by the dissolved organic matter (DOM) is important and generally assumed to be controlled by the humic and fulvic acid fractions (HA and FA).[11] This observation suggests that in a practical agricultural setting, organic amendments can be an effective means to improve Zn availability to crops when such amendments can supply adequate DOM to solubilize native soil Zn.[12,13] Especially, compost deserves attention as a broadly available organic soil amendment.[14] The capacity of these amendments to increase Zn concentrations in soil solution is highly influenced by the solid/liquid partitioning of the organic matter (OM) and DOM composition,[15] particularly the contribution of the HA and FA fractions.

To determine OM concentrations and composition, chemical extractions are a commonly used approach. Due to the complex and highly dynamic nature of OM in heterogeneous materials such as soil and compost, the sample preparation and extraction method will have a large influence on recovered DOM concentration and composition and the related complexation of metals, including Zn.[16,17] When OM is used as a model input variable, these compositional differences subsequently influence model output, that is, predicted Zn speciation and bioavailability. Drying of soil samples is an often applied technique to store samples over time, although recovered DOM concentration and composition have generally been observed to change substantially compared to the fresh material.[17,18] The subsequent estimation of DOM concentrations is commonly performed by using weak salt extracts. A frequently applied method is extraction with 0.01 M CaCl2 (soil/solution ratio 1:10),[19] but various studies have shown a suppressing effect of Ca on total DOM concentrations at increasing Ca2+ concentrations.[20,21] Despite these studies, comprehensive knowledge on the effect of sample pretreatment and extraction conditions on DOM composition and related metal speciation and bioavailability is limited for soils and particularly for soil-like organic amendments such as compost.

This study focuses on modeling of reactive Zn speciation in compost as affected by the solid/liquid partitioning of OM and its individual fractions, in particular HA and FA. These humic substances have been shown to be functional proxies for the metal binding properties of natural OM, despite the current debate on their actual existence.[22] For the first time, multisurface modeling is applied to compost to obtain a quantitative, mechanistic understanding of Zn speciation as affected by OM composition. The effects of drying, type of extractant and equilibration time on OM composition and Zn speciation are studied in particular. Modeled Zn speciation is related to compost characteristics, and observed deviations between predicted and measured dissolved Zn concentrations are evaluated in the light of uncertainties in model parameters.

Materials and Methods

Collection of Composts

Four composts with low reactive Zn contents derived from a broad range of plant/wood-based input materials were selected for this study. Two composts were obtained from the Van Iersel composting facility, Biezenmortel, The Netherlands: a three-month-old compost of 100% forest litter (FL) and a three-month-old green waste compost (GWC), comprised of 55% shredded wood, 25% grass litter, and 20% leaf litter. The third compost (wood chips, WC) was obtained from the Orgapower/Van Berkel Groep composting facility, Wintelre, The Netherlands. This compost, from a mixture of both coniferous and deciduous wood chips, had been composted for 2 years. All composts were taken from a minimum depth of 30 cm below the surface of the compost heap, and samples from several points in one heap were combined and homogenized to a mixed sample of 10–15 kg. All fresh composts were collected at the same day and stored moist in closed plastic bags in the dark at 4 °C until use. The fourth compost was a reference material (MARSEP 235) used in the MARSEP (Manure and Refuse Sample Exchange Program) program of WEPAL (Wageningen Evaluating Programs for Analytical Laboratories, Wageningen, the Netherlands). MARSEP 235 (MAR; certificate of analysis available from WEPAL upon request) is a compost that originated from De Ceuster Meststoffen (DCM) N.V., Grobbendonk, Belgium, which was previously dried at 40 °C and milled to pass a 0.5 mm sieve. The analyses of the composts as described below were all performed within a few weeks to limit changes in OM properties over time.

Preprocessing and Characterization of Composts

The FL, GWC, and WC composts were sieved moist over a plastic 2 mm sieve to homogenize the samples. The moisture content of all composts was determined by drying for 24 h at 105 °C. The total OM content was determined by the loss on ignition from 105 to 550 °C. Reactive concentrations of Cu and Zn were determined with 0.43 M HNO3 extraction with a 1:10 solid/solution ratio based on the compost dry weight (4 h of equilibration).[23] Ammonium-oxalate extractable Al, Fe, and Mn were determined according to Schwertmann.[24] The clay content (<2 μm) was determined by laser diffraction. Volume percentages of clay were converted to a mass fraction by assuming a clay particle density of 2.6 g cm–3, and fresh compost density was assumed to be equal to dry compost density. All these characteristics were determined for both moist and dried composts. Total elemental concentrations of the four composts dried at 40 °C were determined by X-ray fluorescence spectroscopy (XRF).

Extraction Conditions To Measure Total Dissolved Zn and Organic Carbon (DOC)

To determine dissolved Zn and total DOC concentrations of the composts, the following default extraction conditions were used: equilibration of fresh compost for 2 h with 0.001 M CaCl2 with a 1:10 solid/solution ratio based on the compost dry weight. Samples were equilibrated by horizontal shaking with 180 bpm at 20 °C. A 0.001 M CaCl2 concentration was chosen as the default extraction for this study since strong coagulation of HA has been shown to occur around 0.01 M Ca2+.[21] MAR was only extracted with 0.001 M CaCl2.

To investigate the effect of extraction conditions on the compost OM concentration and composition and associated Zn speciation, three extraction conditions were altered for the GWC, WC, and FL composts. Either the moisture content ((original) moist material, a material dried for 48 h at room temperature under forced air ventilation or a material dried for 48 h at 40 °C under forced air ventilation), extractant (ultrapure water (UPW), 0.001 M CaCl2 or 0.01 M CaCl2), or equilibration time (2 or 168 h) was modified from the default extraction conditions, while the other conditions were kept at the default setting. Given the high OM content and binding capacity of the composts, an equilibration time of 168 h was chosen to better approximate equilibrium. The bottles were wrapped in aluminum foil to prevent photodegradation of OM. For MARSEP, which was dry as received, a comparison was made between the dry material and material brought to a 50% water content and incubated in the dark for 8 days at 20 °C.

After equilibration of the samples, a subsample was taken for pH determination and the rest of the solutions were centrifuged at 3000 rpm for 10 min (Beckman Coulter Allegra X-12R). Subsequently, the supernatants were centrifuged at 10,000 rpm for 10 min by ultraspeed centrifugation (Thermo Scientific Sorvall RC6+) to facilitate filtration over 0.45 μm membrane filters. From the filtered samples, subsamples were taken for determination of Al, Ca, Fe, Mg, P, and S concentrations by inductively coupled plasma-atomic emission spectroscopy (ICP-AES; Varian Vista Pro or Thermo Scientific iCAP6500) and Cd, Cu, Mn, and Zn concentrations were determined by high-resolution inductively coupled plasma-mass spectrometry (HR-ICP-MS; Thermo Scientific Element 2). Total DOC concentrations were calculated from the difference between total carbon (TC) and inorganic carbon concentrations (IC) in the extracts as determined by a segmented flow analyzer (SFA-TOC; SKALAR San++). The remainder of the filtered solutions was used for fractionation of DOC.

DOC Fractionation

After extraction of the total DOC, the DOC solutions were fractionated in four operationally defined fractions: humic acids (HA), fulvic acids (FA), hydrophilics (Hy), and hydrophobic neutrals (HON).[25] FA was desorbed with 0.1 M KOH from the DAX-8 resin in two subsequent steps of 1 h as preliminary experiments showed that FA desorption was >95% after two steps.

Fractionation of NaOH-Extractable Organic Carbon (OC)

Besides determination of dissolved HA, FA, Hy, and HON concentrations, NaOH-extractable HA, FA, Hy, and HON concentrations in the composts were determined.[25] Both the moist and dried composts were examined. The supernatant of the acid extraction (FAHyHON1) and the supernatant of the acidified base extract (FAHyHON2) were combined and jointly equilibrated with DAX-8 resin. Subsequent determination of FA, Hy, and HON fractions was similar to the DOC fractionation procedure.[25]

Multisurface Modeling

Metal speciation was modeled by using the modeling framework ORCHESTRA.[26] Metal adsorption to the organic matter, clay, and hydrous ferric oxides was included in the multisurface model according to the setup of Dijkstra et al.[5] Both Zn and Cu speciation was modeled. Cu was included in this study since Cu binding parameters to HA and FA are relatively accurate and defined based on a much larger dataset.[27] This greater accuracy is also reflected in model predictions of Cu speciation in a similar multisurface model setup as used in this study.[5] If not stated otherwise, thermodynamic data from the MINTEQ.v4 database were used for speciation calculations.[28]

Metal adsorption to the organic matter was modeled by means of the NICA-Donnan model,[29] using the generic binding parameters of Milne et al.[27,30] For Fe binding to FA, binding parameters of Hiemstra et al.[31] were used (Table S1) as those of Milne et al. are derived from linear free energy relations and seem to strongly overestimate Fe binding. Solid phase HA and FA concentrations were calculated as the difference between the NaOH-extractable concentrations and the concentrations determined by extraction with either UPW or CaCl2. HA and FA concentrations derived from DOC fractionation were used as input for dissolved HA and FA. HA and FA were assumed to consist 50% wt of C. The Hy and HON fractions were not considered in the model because of their unknown composition and assumed much lower importance for metal speciation than HA and FA.

Metal adsorption to iron (hydr)oxides was modeled using the generalized two-layer model and generic parameters of Dzombak and Morel.[32] The amount of amorphous hydrous ferric oxides (HFO) was calculated by first allowing Al and Fe precipitation as Al(OH)3 and ferrihydrite based on the measured oxalate-extractable Fe and Al concentrations. Subsequently, the combined amounts of precipitated Al(OH)3 and ferrihydrite were converted to HFO by assuming a molecular mass of 89 g mol–1 HFO and an equivalent molar mass for both Al(OH)3 and ferrihydrite.[33] These values were subsequently used as model input. Only for the samples with pH > 6.6 HFO formation was predicted (Tables S2 and S3). A specific surface area of 600 m2 g–1 was used for HFO.[32]

Metal adsorption to clay minerals was modeled using the measured clay fraction (w/w) as input and a Donnan model assuming a charge density of 0.25 eq kg–1 and a fixed Donnan volume of 1 L kg–1, which are considered average values for illitic clay minerals.[34]

Model input files (Tables S2 and S3) consisted of: (1) pH of the samples, fixed to the value measured in the extracts, (2) measured total dissolved concentrations of Ca, Mg, Mn, P, and S in the CaCl2 or UPW extract, assuming that all measured P is PO43– and all measured S is SO43– and estimated total dissolved H2CO3 (based on measured IC concentrations) and Cl concentrations (based on the CaCl2 concentration of the extractant), (3) total Al and Fe concentrations as determined by oxalate extraction, (4) reactive Cu and Zn concentrations as determined by 0.43 M HNO3 extraction, and (5) concentrations of the solid and dissolved reactive surfaces (organic, HFO and clay) as described above.

Results and Discussion

Compost Characteristics

Chemical characteristics of the composts for default extraction conditions are shown in Table and include pH, OM content, metal contents, and the most important reactive surfaces with regard to Zn speciation. The four composts show a large variation in the OM content, reactive surface content, and Zn loading. For all composts, NaOH-extractable OC was dominated by HA, covering between 61% of the total for WC and 82% for GWC. Reactive Zn concentrations of the composts were low to medium compared to Zn concentrations in green waste composts measured across Europe (141–470 mg kg–1 dw).[35] Reactive Cu contents were comparable among the composts except for WC, which had a distinctly lower Cu concentration (1.0 mg kg–1 dw). Table shows that FL and GWC on the one hand and MAR and WC on the other hand shared similar characteristics. FL and GWC have similar and relatively high pH values and oxalate extractable Al and Fe concentrations, and lower total OM contents and NaOH-extractable OC levels compared to MAR and WC. MAR had a distinctly lower pH and reactive Zn content and relatively high HA, Hy, and related total OC concentrations compared to the other composts. The reactive surfaces of WC were almost fully determined by NaOH-extractable OC. Clay particles in the composts were likely associated with plant materials in the feedstock. Compared to organic waste materials such as biosolids, total Zn, Cu, P, and Fe concentrations of the composts were low.[36,37] XRF shows that the composts contain an additional fraction of nonexchangeable Zn (Table S4). Based on the nearly quantitative recovery of reactive Zn by the HNO3 extraction,[23] we assume that this nonextractable amount does not contribute to the Zn partitioning with the solution.

Table 1

Compost Characteristics under Default Extraction Conditionsa

			reactive surfaces
			NaOH-extractable OC								reactive metals		total metals
compost	pH	OM	HA	FA	Hy	HON	total	Alox	Feox	clayb	Cu	Zn	Cu	Zn
		(g kg–1 dw)	(g C kg–1 dw)					(g kg–1 dw)			(mg kg–1 dw)
FL	7.67	349	14.4	2.6	2.1	0.9	20.0	0.49	1.44	71 (1.1%)	6.7	95.0	27.6	249
GWC	7.32	251	21.5	2.1	2.1	0.6	26.2	0.56	2.56	84 (1.6%)	6.5	90.6	35	196
MAR	5.22	899	48.6	8.2	14.1	2.2	73.1	0.29	0.49	53 (0.8%)	5.0	25.4	17.9	50
WC	6.50	883	27.3	5.5	10.9	1.1	44.8	0.12	0.32	12 (0.1%)	1.0	166.6	17.3	470

pH was determined in 0.001 M CaCl2, OM by loss on ignition, NaOH-extractable OC fractions were determined by acid/base fractionation, Al and Fe determined by ammonium oxalate extraction, the clay content by laser diffraction, reactive Cu and Zn by 0.43 M HNO3 extraction, and total Cu and Zn by XRF.

The clay content was determined on volume basis. Numbers are an estimation based on estimated clay and compost densities. Numbers between brackets are clay percentages v/v.

Compost OC Concentration and Composition in Relation to Extraction Conditions

Effect of Extractant

Results of the DOC fractionation of the four composts for the different extraction conditions are shown in Figure . Measurement data are listed in Table S5. For all composts, reactive OM was predominantly located in the solid phase; at most, 11% of NaOH-extractable OC was extracted as DOC (for WC-UPW, Figure a). DOC concentrations of MAR and WC were higher than those of FL and GWC in accordance with the higher total OM and NaOH-extractable OC contents (Table ). Extraction at increasing Ca2+ concentrations showed a distinct decrease in recovered DOC concentrations for all composts, caused by particularly decreased HA, and to a lesser extent FA and Hy concentrations (Figure a). Similar trends have been shown for soils[17,20,21] in that lower recovered HA concentrations were observed and attributed to coagulation of HA by Ca2+ ions and subsequent flocculation. Especially DOC extraction with a 0.01 M CaCl2 solution resulted for all composts in a pronounced decrease of HA concentration compared to extraction with UPW or 0.001 M CaCl2. This finding is in good agreement with previous work using purified forest floor HA in which strong HA coagulation was observed starting from a background Ca concentration of approximately 10–2.5 M.[38] Despite its relatively high NaOH-extractable HA content, a low dissolved HA concentration was observed for MAR, which can be partly explained by the low pH of this compost (5.22) and the relatively slow dissolution kinetics of HA as described below.

Figure 1

DOC composition of composts under different extraction conditions, expressed as the percentage of NaOH-extractable OC. Figures show variation in (a) extraction solution, (b) equilibration time, and (c) drying. NaOH-extractable OC concentrations are presented in Table for fresh composts; data on dried composts can be found in the Supporting Information (Table S5). Numbers above the bars are absolute total DOC concentrations in g C kg–1 dw. Default extraction conditions are marked with an asterisk (*).

Effect of Equilibration Time

Increasing the equilibration time of the composts from 2 to 168 h resulted in increased solubilization of OC (Figure b). The increase in DOC concentrations can predominantly be attributed to an increased HA concentration in solution, irrespective of compost origin and composition. The observed slower dissolution of HA versus FA may be related to the lower diffusivity of HA in relation to the larger size of its molecular associations.[39] MAR displays the largest increase in HA concentration in solution: from 1.7 to 14.8% of DOC after 1 week of equilibration. The slow solubilization but strong increase between 2 and 168 h of the HA fraction of MAR is consistent with the relatively low pH of the sample together with its high NaOH-extractable HA content.

Effect of Drying

Drying the composts at increasing temperatures resulted in increased DOC concentrations for all composts (Figure c). This increase was related to increased concentrations of the Hy fraction, which is generally assumed to be caused by microbial cell lysis upon drying.[18] However, next to increased Hy concentrations, FA concentrations also increased in both absolute and relative terms in all dried composts. This effect could either point to the presence of microbial components in the FA fraction or competitive sorption of hydrophilic and fulvic acids on aluminum and iron (hydr)oxides in the composts. Although this assumption could not be validated, this interpretation is supported by increased FA concentrations for FL and GWC, which contained the highest amount of reactive mineral surfaces. For the MAR compost, the total DOC concentration of the incubated sample dropped by 35% and absolute FA and Hy concentrations by 37 and 43%, respectively, relative to the immediately extracted sample. The incubation of this originally dry standard sample has likely favored microbial decomposition, resulting in a preferential decomposition of the Hy fraction.[40]

The above described generic multisurface model is used to relate reactive Zn speciation to OM speciation and other characteristics of the individual compost samples as affected by the different extraction conditions. Modeling is thus used to analyze the major underlying processes that control Zn speciation in compost. The accuracy of the predicted dissolved Zn concentrations is very similar to that found for soils.[5]

Effect of Extractant

Model predictions of Zn speciation in both solution and solid phase under the different extraction conditions are shown in Figure . Except for WC, >95% of the total reactive Zn is bound to the solid phase, irrespective of extraction conditions. The majority of Zn present in the composts is bound to humic acids in the solid phase (SHA), accounting for 51–98% of total reactive Zn, consistent with previous studies in soils.[11,41,42] The extractant strongly influenced Zn speciation in solution. In general, increasing Ca2+ concentrations in solution caused a decrease of Zn complexation by dissolved humic acids (DHA), related to decreased DHA concentrations (Figure a). In addition, Zn electrostatically bound to HA and FA decreased with increasing Ca2+ concentrations (Table S6) as a result of the increased ionic strength and the related decreased Donnan volume. In addition to increased free Zn concentrations as a result of decreased Zn-DHA complexation, a strong increase in free Zn concentration was predicted for the samples extracted with 0.01 M CaCl2 compared to UPW or 0.001 M CaCl2, suggesting strong competition of Ca2+ with Zn2+ for binding sites. This effect will be addressed more extensively below. For the HA-rich WC, the fraction HA in solution in the 0.01 M CaCl2 extract remains relatively large compared to the FL and GWC composts (Figure a), resulting in a relatively large part of dissolved Zn complexed by DHA compared to the FL and GWC samples extracted with 0.01 M CaCl2.

Figure 2

Zn speciation of the composts under the different extraction conditions, that is, variation in (a) extraction solution, (b) equilibration time, and (c) drying. Left column displays speciation in solution; right column displays speciation in the solid phase. Due to large differences in Zn concentrations, the Zn concentration of WC is shown at the secondary y-axis for clarity. Default extraction conditions are marked with an asterisk (*).

Effect of Equilibration Time

Extending equilibration times to 1 week resulted in a marked increase in Zn concentrations in solution for all composts (Figure b) in parallel with strongly increased DHA concentrations after 168 h equilibration (Figure b). For all composts, Zn electrostatically bound to HA and FA decreased as a result of increased Ca2+ concentrations (see further below) and the related decreased Donnan volume. Careful examination of Figure b shows a decrease in the relative importance of Zn-DHA complexes for FL and GWC in contrast to an increase for WC and MAR. Both the WC and MAR samples contained a high DHA concentration, and particularly, MAR showed a large increase after 1 week of equilibration (Figure b). For these composts, the increase in DHA is larger than the increase in dissolved Zn, resulting in increased Zn-DHA complexation especially to (phenolic) high-affinity binding sites of DHA (Table S6). For FL and GWC, the increase in dissolved Zn is higher than that of DHA, resulting in a predicted decrease in Zn binding to high-affinity sites of DHA. In addition, the FL and GWC composts experienced a relatively large increase in dissolved Ca concentrations after 168 h of equilibration compared to the MAR and WC composts (respectively 6.8× and 6.6× increase compared to 1.5× for MAR and 4.9× for WC, Figure S1), consistent with a decrease in pH (except for MAR). This increased Ca competition contributes to the relatively strong predicted increase of free Zn in solution.

Effect of Drying

Total predicted dissolved Zn concentrations are following fluctuations in measured DHA concentrations. For FL and GWC, the DHA concentration reduced about 2-fold upon drying, whereas the DHA concentration of WC remained virtually unchanged (Figure c). For FL and GWC, the contribution of Zn-DFA to total dissolved Zn doubled (from 6.3 to 12.2% and 4.7 to 11.2%, respectively, between moist and 40 °C dried) at the expense of Zn-DHA complexation (Figure c). For WC, a small increase in DFA is reflected in slightly increased total dissolved Zn concentrations. However, conclusions on the relative importance of Zn-FA complexation should be drawn with caution since the generic NICA-Donnan parameters for Zn-FA complexation are derived based on a very few data, with relatively high uncertainty margins.[27]

Model Performance and Limitations

The multisurface model consistently overestimated Zn concentrations in solution for the FL, GWC, and WC samples (Figure ). This overprediction suggests limitations in the current parameters for Zn binding to HA and/or FA resulting in a too weak binding as HA and FA dominate Zn binding and are predominantly present in the solid phase (Table ). If inaccuracies in Zn binding to HFO would have been the major cause for the overestimation, the prediction for samples without HFO (i.e., WC and MAR) would have been much better, which is not the case. In contrast to the data shown here, Duffner et al.[7] found a good agreement between measured and predicted Zn2+ concentrations for low-Zn soils. This observation can most likely be explained by a strong overestimation of SHA concentrations in their study. The estimated reactive HA concentrations in that study were 4 to 19 times higher than we have recently measured directly in four of their low-Zn soils (Klinkert et al., unpublished results). Duffner et al.[7] similarly overestimated DHA concentrations, but given the low DOC concentrations in their soils, the net result of their approach is a predicted stronger Zn binding to the solid phase at low Zn2+ concentrations.

Figure 3

Total measured Zn in solution plotted against the total predicted Zn in solution by the multisurface model. The data points within the dotted circles represent the composts extracted with 0.01 M CaCl2.

Concerning the samples where total Zn in solution was overpredicted compared to measured concentrations, especially the composts extracted with 0.01 M CaCl2 should be noted (Figure ). To study possible effects of Ca-Zn competition, we have also analyzed the Ca speciation in the same way as we have done for Zn (cf. Figure and Figure S1). Figure S1 shows that the Ca speciation in the solid phase resembles that of Zn and is dominated by binding to SHA, while the Ca in solution is for a greater proportion bound to DFA than Zn. As shown in Tables S6 and S7, Ca-Zn competition occurs predominantly in the Donnan volume and to a greater extent for FA than for HA. We also note that this electrostatic competition between Cu2+ and Ca2+ is negligible (Tables S7 and S8), while the model prediction of Cu in solution is much more accurate (Figure S2 and Table S9). As the current generic parameters imply a much greater relative importance of electrostatic binding for Zn than for Cu, Zn is much more susceptible to competitive electrostatic binding of divalent Ca2+ and thus to inaccuracies in the description of the electrostatic binding component. These observations suggest that the accuracy of the model predictions may be related to the electrostatic binding component of Zn to FA. This interpretation is consistent with the suggestion of Hiemstra et al.[31] (2006) that the Donnan volume of FA has been overestimated in the NICA-Donnan model, resulting in an overestimation of the electrostatic binding component and consequently the magnitude of the Ca-Zn competition on FA. As most of the FA is located in the solid phase, the net effect is an overestimation of Zn in solution. These findings thus indicate that the description of the electrostatic binding of cations by HA and especially FA needs further research.

In contrast to FL, GWC, and WC, Zn concentrations in solution were underestimated for MAR (Figure ). This compost differs from the other three in that an even larger fraction of the HA is located in the solid phase, while a relatively large fraction of FA is present in solution (Table S5). As a result, overestimation of Ca-Zn competition on FA leads to an underestimation of Zn in solution. Moreover, MAR contained a high concentration of reactive OC and low concentration of reactive Zn (Table ). As a result of this low Zn loading of the reactive OC, Zn is likely to be more prominently associated with high-affinity binding sites, both within and between carboxylic and phenolic site groups. The predicted distribution of Zn over the carboxylic and phenolic binding sites of HA and FA did not show any relevant difference between the MAR and the other composts and thus cannot explain the underprediction of MAR in contrast to the overprediction of Zn in solution for the other composts (Table S6). However, the underprediction of Zn in solution for the MAR sample may also be related to the proportionally much higher association of Zn with high-affinity sites within the carboxylic and phenolic groups as a result of its much lower loading compared to the other composts. In the data used by Milne et al. (2003)[27] for derivation of the generic Zn parameters of the NICA-Donnan model, an increasing overprediction of Zn-HA complexation toward low Zn loadings has been shown to occur especially around pH 6,[43] that is, around the pH of the MAR sample (Table ). In that study, the lowest Zn loadings measured were 1.3 × 10–2 mole Zn kg–1 HA for pH 6,[43] which is still relatively high compared to the Zn loading of MAR (3.9 × 10–3 mole Zn kg–1 HA). This observation suggests that overestimation of Zn-HA complexation at such a low Zn loading may even be more pronounced for MAR and may thus contribute to the observed underestimation of Zn in solution for that compost (Figure ). Figure S2 and Table S9 suggest that these limitations are not associated with the NICA-Donnan model as such nor the input variables since Cu in solution is adequately predicted for all compost samples including MAR.

Finally, we note that the primary purpose of this paper was to base modeling predictions on generic parameters for identifiable OM in addition to HFO and clay. Thus, we have used a frequently applied multisurface model setup for soil systems in which binding to OM is assumed to be controlled only by the HA and FA fractions.[5,7,44] Consequently, possible complexation of Zn with hydrophilic organic acids was not considered. As the composition of the Hy fraction is unknown and variation of Hy properties between the different composts is likely to exist, metal complexation by Hy can currently only be predicted based on assumptions.[45] Since the Hy fraction is an important part of DOC and even of solid OM (representing 7–25% of SOC, Table S5), specific research on the nature and reactivity of Hy in soils and organic amendments may further improve the accuracy of multisurface model predictions. The same may hold for SOM that is not extracted with the current fractionation protocols for humic substances as well as for other than currently considered inorganic Zn species in the solid phase (e.g., by surface spectroscopic analysis).

Predicting Zn speciation by multisurface modeling has revealed the same underlying mechanisms in composts as in soils and that optimization of Zn binding parameters of the NICA-Donnan model is needed. Modeling showed that especially at low Zn concentrations, Zn solubility is controlled by HA. Deviations between measured and predicted dissolved Zn concentrations are likely associated with limitations in the description of Ca-Zn competition, particularly in their electrostatic binding to FA, as well as in the concentration-dependent binding affinity of Zn. These findings are especially relevant for application of multisurface modeling to predict Zn speciation in low-Zn soils as these soils often contain high levels of CaCO3 and low dissolved Zn concentrations. Improving model descriptions with regard to these features will widen the applicability of multisurface models in predicting Zn speciation and further our knowledge on mechanistic relationships between OM characteristics and Zn speciation.