IODINE DYNAMICS IN SOILS

We investigated changes in iodine ( 129 I) solubility and speciation in nine soils with contrasting properties (pH, 13 Fe/Mn oxides, organic carbon and iodine contents), incubated for nine months at 10 o C and 20 o C. Loss of I - 14 from solution was extremely rapid, apparently reaching completion over minutes-hours; IO 3- loss from solution 15 was slower, typically occurring over time periods of hours-days. For both I - and IO 3- losses were faster in soils 16 with greater soil organic carbon contents (%SOC) and low pH and at higher temperatures (10 o C cf. 20 o C). 17 Instantaneous sorption of IO 3- was identified in all soils and was greatest in a soil with high Fe/Mn oxide, low 18 pH and low SOC content. Evidence for immediate sorption of I - was less clear as reaction rates were faster. Phosphate extraction (0.15 M KH 2 PO 4 ) of soils, ~100 hr after 129 I spike addition, indicated that concentrations 20 of sorbed inorganic iodine ( 129 I) were very low in all soils suggesting that, even if IO 3- is initially adsorbed onto 21 oxide phases, this has little impact on the rate of iodine assimilation into humus. The transformation of dissolved inorganic 129 IO 3- and 129 I - to sorbed organic forms was modelled using a range of reaction- and diffusion-based approaches. Irreversible and reversible first order kinetic models, and a spherical


ABSTRACT 12
We investigated changes in iodine ( 129 I) solubility and speciation in nine soils with contrasting properties (pH, 13 Fe/Mn oxides, organic carbon and iodine contents), incubated for nine months at Iodine is an essential trace element for human and animal health. It is used by the thyroid gland in the 3 production of hormones which control a range of physiological processes. Insufficient thyroid hormone levels 4 are associated with a range of health issues including problems of growth and development in children, and 5 goitre in adults (Trotter, 1960;Underwood, 1977). Collectively, iodine deficiency diseases (IDDs) are a serious 6 worldwide health problem, estimated to affect~35% of the world's population, and a significant social and 7 economic stress on developing countries (WHO, 2004). 8 9 Rocks contain little iodine and most soil iodine is derived from volatilization of methylated forms from 10 seawater which then enter the soil-plant system via rainfall and dry deposition. IDDs are prevalent in regions 11 where people have limited access to food that is naturally rich in iodine (e.g. seafood) or iodized food products 12 (Underwood, 1977;Johnson et al., 2002). Availability of iodine in such regions depends largely on transfer 13 from soil to food or fodder crops but local produce may not be able to supply the recommended daily intake of 14 dietary iodine (Johnson, 2003). There is therefore a need to increase our knowledge of iodine behaviour in soil, 15 in particular how added iodine (in rainfall or fertilizers) reacts with soil and the mechanisms by which iodine 16 becomes available to plants. Furthermore, understanding the environmental behaviour of long lived iodine 17 isotopes ( 129 I t ½ = 1.6 x 10

Soil sampling and preparation 2
Topsoil and subsoil were sampled from two areas in the East Midlands of England, chosen to represent 3 contrasting land-uses, soil pH values and concentrations of Fe/Mn oxides, organic matter, carbonate and 4 iodine. Wick series (sandy loam) soil samples were taken from an arable field, a permanent grassland strip and Nottingham farm, Sutton Bonington, Leicestershire (UK). Topsoil (0-20 cm depth) and subsoil (30-50 cm 7 depth) samples were taken from the arable and woodland sites; only topsoil was sampled from the grassland 8 as its associated subsoil was thought to be similar to the arable subsoil. Iodine concentrations in these soils 9 were known to be low (2 -4 mg kg stainless steel spades, augers and trowels and sealed in plastic bags for transport. Soils were air dried until 17 they could be sieved to <4 mm but were not allowed to dry completely so as to maintain microbial activity; 18 they were then kept unsealed in a cold room (at 10 1 after reaction samples were centrifuged (20 min at 3000 g), filtered (<0.22 µm) and supernatant solutions 2 retained for analysis. Total soil iodine was extracted with tetra methyl ammonium hydroxide (TMAH) from 3 finely ground soil samples according to the method developed by Watts and Mitchell (2009). 4 5 Elemental concentrations were assayed using a Thermo-Fisher Scientific X-Series II ICP-MS in standard mode 6 (for iodine) and employing a 'hexapole collision cell' (7% H 2 in He) prior to the analytical quadrupole for Fe, Al, 7 and Mn analysis. Samples were introduced from an autosampler (Cetac ASX-520 with 4 x 60-place sample 8 racks) through a concentric glass venturi nebuliser (Thermo-Fisher Scientific; 1 mL min -1 ) and Peltier-cooled ) for I by oxidation using sodium chlorite as described by Yntema and Fleming, (1939). Soil 24 samples for incubation were prepared by mixing moist sieved soil (< 4 mm) in a food mixer with Milli-Q water 25 added to each soil was simply judged from the friability of the aggregated soil rather than being based on a 28 fixed proportion of water holding capacity or a specific soil moisture tension. We considered the need to maintain moist but aerobic soils capable of free gas exchange and able to be sub-sampled for periodic analysis; 1 the final water contents of the incubated soils are shown in Table 1. Spiked soils were distributed between 2 triplicate 500 mL Duran bottles (~180 g dry wt of soil per replicate) with a hole drilled in the lid to allow gas 3 exchange, and incubated in the dark at 10 o C or 20 o C (± 2 o C). Moisture loss was monitored regularly and 4 restored when necessary by re-mixing the soil in a food mixer with the required volume of Milli-Q water 5 before returning the soil to the microcosm bottle and incubator. 6 7

Iodine extraction and analysis 8
After incubation for 114, 306, 810 and 3975 hours, samples (~4.5 g) of moist soil were equilibrated with 20 mL 9 of 0.01 M KNO 3 , followed by extraction with 0.15 M KH 2 PO 4 and then 10 % TMAH, in 40 mL polycarbonate 10 centrifuge tubes. At each stage soil suspensions were shaken for 16 hours on a reciprocal shaker, centrifuged 11 (25 min at 3500 rpm), and filtered through 0.22 m PTFE syringe filters. Calculation of phosphate-extractable 12 iodine accounted for carry over from the previous KNO 3 equilibration gravimetrically. To follow shorter term 13 iodine dynamics (< 72 hours), samples equivalent to~3.5 g dry soil were taken from control microcosms and 14 were assayed by ICP-MS following in-line chromatographic separation using a Dionex ICS-3000 ion 23 chromatography system operated in isocratic mode with a Hamilton PRP-X100 anion exchange column (250 x 24 4.6 mm; 5 µm particle size). The mobile-phase (flow rate 1.3 mL min ensure complete recovery of iodine, the extraction was repeated three times, followed by a further two 16 of a partition coefficient (kd). The addition of the coefficient kd allows for instantaneous adsorption of 1 inorganic iodine, possibly on metal oxide sites; the value of kd was optimized alongside the kinetic parameters. 2 The empirical Elovich equation has been shown to describe the reaction kinetics of a wide range of inorganic 3 compounds with soils and soil components (Atkinson et al., 1970;Chien and Clayton, 1980;Martin and Sparks, 4 1983). It is characterised by a greater ability to describe kinetics over a wide range of timescales, in contrast to 5 other models, because it includes both a constant term which effectively describes instantaneous adsorption 6 and an exponential term. Echevarria et al., (1998) andSinaj et al., (1999) applied an equation based on an 7 infinite series of exponential terms to describe the progressive mixing of metal isotopes with the native soil 8 metal pool -described here as the 'ISE' model. 9 Where diffusion or transport-controlled processes are the rate-limiting steps a parabolic diffusion expression 10 (Par-diffn model) has been used previously (Chute and Quirk, 1967, Jardine and Sparks, 1984, Havlin et al., 11 1985. Application of the spherical diffusion equation (Sph-diffn model; e.g. Brown et al., 1971) assumes that 12 reactions are controlled by diffusion into uniform spherical aggregates of adsorption surfaces (e.g. humic acid). 13 It has been applied successfully to describe diffusion-controlled kinetics in minerals and soils (Cliff et al. 2002;14 Altfelder and Streck, 2006;Iznaga et al., 2007). Altfelder and Streck (2006) demonstrated the greater 15 consistency of the spherical diffusion approach over a first order kinetic equation when parameterised for 16 short time periods and applied to longer reaction times (days-months) because the rate constants of the first 17 order approach are strongly time dependent unlike the diffusion approach. Thus predicting long-term 18 behaviour on the basis of parameters derived at a shorter timescale using a first-order approach is particularly 19 problematic (Altfeder & Streck, 2006). 20 All the models were optimised for individual soils by minimising the residual standard deviation (RSD) between 21 modeled and experimental data, while systematically changing the values of model parameters, using the 22 'Solver' function in the software package Excel 2007. In addition, an attempt was made to fit a single spherical 23 diffusion model to all soils simultaneously by relating model parameters to soil variables; this is described in 24 section 4.6.

Soil Characteristics 3
Measured soil characteristics are presented in Table 1. Soils from Sutton Bonington (SB) were typically lower 4 in pH (4-7) and total iodine concentration (I tot = 2-4 mg kg -1 ) than those from Stoke Rochford (ST) (pH~7, I tot = 5 7.5-12 mg kg -1 ). Woodland topsoils (SB-WT, ST-WT) and the Stoke Rochford grassland soil (ST-GT) had 6 relatively large organic carbon contents (6-10%), and loss on ignition (LOI), than the arable soils. Carbonate 7 content was greatest in soils from Stoke Rochford where the underlying geology is limestone. A value of 2.5% 8 carbonate in the SB arable topsoil (SB-AT) may reflect liming shortly before sampling occurred. Iron and Mn 9 oxide concentrations were typically higher in soils from the ST site. Within 800 hrs all of the 129 IO 3 added to grassland and woodland soils had been converted to organic forms 6 whereas in arable subsoils only 30% of the total 129 I was present in solution as organic complexes. Conversion 7 of inorganic to organic iodine was also high where either pH was low or organic matter content high, seen by 8 comparison of soils SB-GT (moderately organic and slightly acidic), ST-WT (highly organic and slightly alkaline), 9 and ST-GT (highly organic and slightly acidic) ( Figure 1 The rapid initial loss of 129 IO 3 -, may be attributable to a combination of volatilization, electrostatic sorption on 14 inorganic soil phases and rapid immobilization by reduction at sites on organic matter e.g. quinone groups. 15 Volatilization of 129 I from solution is considered unlikely as such losses have been shown to be small in previous 16 studies (e.g. Sheppard et al., 2004;Sheppard et al., 2006). Sorption of iodide and iodate to oxide phases is 17 weak at pH>6 where sorption to organic matter dominates (see e.g. Sheppard and Thibault, 1992 and 18 references therein) but has been reported up to pH 9.6 (Yoshida et al., 1992;Kaplan et al., 2000). Below pH 6 19 iodate sorption is predominantly to iron and aluminium oxides with iron oxides becoming increasingly 20 important as pH drops (Whitehead, 1974b). Iodate is non-reactive toward organic matter and studies have 21 shown that it is reduced to electrophilic species such HOI or I 2 before incorporation into the organic structure 22 of humus (Francois, 1987a & b;Bichsel andvon Gunten, 1999, 2000;Radlinger and Heumann, 2000;Warner et 23 al., 2000;Reiller et al., 2006;Schlegel et al., 2006;Steinberg et al., 2008c). The reduction of iodate has been 24 shown to be faster under acidic conditions (Brummer and Field, 1979); in soils, humic substances can reduce 25 iodate due to their electron-donor characteristic (Wilson and Weber, 1979). In the current study the rate of 26 loss of 129 IO 3 from solution was higher in the low pH soils than in high pH ones where the organic carbon 27 content was comparable (e.g. SB-WT/ST-GT and SB-WS/SB-AS) consistent with a mechanism involving iodate matter (Gallard et al., 2009). In soils with similar pH the rate of 129 IO 3 loss from soil solution was higher in 1 those with greater organic carbon contents, e.g. SB-WS compared to SB-WT and SB-AS compared to SB-GT, 2 demonstrating the importance of organic carbon in reducing iodate to a species (e.g. HOI, I 2 ) whereby it can be 3 converted into org-I species in solution or in the solid phases. The complete mechanism of each of these 4 reactions (illustrated schematically in Figure 3) cannot be fully elucidated as no attempt was made to measure 5 intermediate species in the reaction, however the rates of sorption and formation of soluble org-I are both 6 rapid (< 1 hr). In order to interact with soil organic matter it has been shown that iodide must be oxidised to an intermediate 26 such as I 2 or HOI (Warner et al., 2000;Reiller et al., 2006;Schlegel et al., 2006). Metal (Fe, Mn, Al) oxide 27 phases and soil organic matter are both possible oxidising agents. Soil metal oxides have been shown to 28 oxidise iodide in amounts proportional to their concentration, and inversely proportional to pH, in a reaction that is thermodynamically favourable up to pH 7.5 (Allard et al., 2009;Fox et al., 2009;Gallard et al., 2009). 1 Humic substances, which contain some electron acceptor sites, also act as oxidising agents for iodide (Blodau 2 et al., 2009;Keller et al., 2009). Sheppard and Thibault (1992) described rapid loss of iodide from solution in 3 organic soils as a first order reaction. However, they observed no evidence for specific bonding of iodide as 4 the majority of iodide was found to be easily desorbable by water within a few days indicating weak retention 5 in organic soils.

TMAH Extraction 11
Tetra methyl ammonium hydroxide (TMAH) has recently been shown to extract quantitatively the total iodine 12 content from environmental samples e.g. soils, sediments, plants, and food (Watts & Mitchell, 2009). Alkaline 13 extractants such as TMAH mobilise humic acids (and org-I) by negative charge generation and may also cause 14 some degree of hydrolysis of org-I compounds. In addition TMAH releases iodate from specific sorption sites 15 on Fe/Al hydrous oxides by replacement with hydroxide ions and negative charge generation on the oxide 16 surface (Yamada et al., 1996). One advantage of TMAH over inorganic extractants such as NaOH or KOH is that 17 high pH values can be achieved without increasing the salt concentration of the extraction solution and hence 18 reducing the possibility of precipitation in the ICP torch and nebuliser during analysis. 19 20 A single TMAH (10%) extraction was used as a final step for some samples, following phosphate extraction. On 21 average, total-129 I extracted ranged from 0.109 -0.129 mg kg -1 (representing 73-86% recovery of the 0.15 mg 22 kg -1 spike). Recovery was generally slightly worse in organic rich soils (e.g. 75-80% in ST-WT) and better in 23 those with lower organic matter contents (e.g. arable subsoils, SB-AS and ST-AS, 85-90%). The amount of total 24 129 I extracted was unaffected by incubation temperature or the iodine species used for initial spiking. 25 Consequently an exhaustive extraction procedure using three sequential extraction steps with 10% TMAH was 26 undertaken on two soils (SB-WT and SB-AS) chosen to represent 'end members' in terms of soil properties (pH 27 and %SOC). This more rigorous extraction produced c. 100% recovery of 129 I spikes and confirms that loss of 28 Model parameters and residual standard deviations (RSD) for individual model fits are given in Tables 3 and 4  3 for iodate and iodide respectively. Comparisons of how well individual models fit for iodate and iodide across 4 all soils is shown in Figure 4. For iodate, models in which no instantaneous adsorption was allowed (i.e. 5 irreversible first order (IFO), infinite exponential (ISE), reversible first order (RFO) and parabolic diffusion) gave 6 a poorer fit, with a greater range of RSD values, than those that incorporated a kd value (Figure 4a). The 7 reversible first order + kd (RFO-kd), irreversible first order + kd (IFO-kd) and spherical diffusion + kd (Sph-Diffn- for iodate than iodide. 6 7 IFO-kd and RFO-kd approaches described iodate and iodide reaction kinetics well, with the most important 8 factor in achieving a good fit for iodate being the inclusion of a kd value to allow for instantaneous adsorption 9 occurring at t=0. By contrast, for iodide, these models were only slightly better than those in which 10 instantaneous adsorption was not included. An Elovich modeling approach, typically used to describe soil 11 processes occurring across a range of timescales, generated a good fit for iodide but was less successful for 12 iodate. Overall the best model fits to both iodate and iodide were achieved using a spherical diffusion 13 approach. The success of the spherical diffusion model (Sph-Diffn-kd) for iodate appears to confirm its ability 14 to describe processes over a relatively wide range of times (Altfelder and Streck 2006). That it also worked 15 well for iodide suggests that it is useful for describing faster reaction kinetics as well. A comparison of 16 The diffusion parameter, p(D/r 2 ) was therefore expressed as a linear function of the three soil variables and 4 the distribution coefficient (kd) as an exponential function of pH in which the value of kd at pH = 0 (ko) was a 5 linear function of %SOC and %Ox: 6  (Table 5). In Table 5  does produce a potential instability in that it is possible to derive negative values for distribution coefficient at 22 very large soil humus contents. Also, it was found that k O applied to kd was reduced to zero when both %SOC 23 and %Ox were included as variables to give 14, rather than 16, as the number of model coefficients required to 24 give the best fit (Table 5). Table 6 shows the values of the soil coefficients used to derive the model 25 parameters p(D/r 2 ) and kd for each incubation temperature (Equations 11 and 12). For prediction of kd value, 26 the soil coefficients are broadly in line with expectation in that kd declined with pH (k pH is a negative initial adsorption of iodate is as an inorganic species on hydrous oxides. Similarly, the negative values of k c (for 1 calculation of kd values) in Table 6 suggest that humus restricts the initial adsorption of iodate -possibly 2 through competition for oxide sites and electrostatic repulsion. This agrees with the observation of Dai et al., 3 (2004) who observed iodate adsorption to be positively correlated with free iron oxide content of soils and 4 negatively correlated with soil organic matter content. 5 6 Figure 7 shows the fit of the soil-parameterised spherical diffusion model to iodate sorption. The overall 7 simulation was reasonable across the range of soils examined with most soils falling wholly within ± 1 RSD of 8 the 1:1 relation. However, some individual soils produced systematic deviation from the model trend. Thus, 9 iodate persisted in solution in the SB-AS, a sandy arable subsoil with low soil organic carbon (%SOC) content, 10 for longer than predicted by the model (at low iodate concentrations). The grassland topsoil from the same 11 site showed the reverse trend with more rapid sorption from solution than predicted. The main source of iodine to a soil is rainfall. The extent to which iodine in rainfall is retained by a soil will 28 therefore depend not only on soil properties but also on factors including (i) distance from the ocean and therefore iodine concentration in the rain, (ii) the speciation of iodine in the rainfall, (iii) the timing, duration 1 and intensity of the rainfall, (iv) whether the soil is dry or wet before a rainfall event, (v) the extent to which 2 the rainfall infiltrates or drains from a soil, which is dependent upon both the soil texture and its management 3 and soil temperature. Uptake by plant roots and microbial processing of the iodine may also be factors (see 4 e.g. Whitehead, 1975). Iodine concentration in rainfall is reported to be in the range of 0.5-5 g L -1 (e.g. 5 Truesdale and Jones, 1996, Neal et al., 2007, Hou et al., 2009 but there is little agreement on the mix of 6 species present with I -, IO 3 and organic iodine all reported as 'major species', the relative proportions of each 7 varying with location (e.g. Gilfedder et al., 2007, Yoshida et al., 2007. Low intensity rainfall will infiltrate the 8 soil more easily than high intensity rainfall which 'seals' the surface of the soil increasing run-off. Coarse 9 textured (e.g. sandy) soils will allow easier infiltration (> 50 mm hr -1 ) but will also drain completely within a few 10 hours whereas a fine textured (e.g. clayey) soil allows less infiltration (<15 mm hr -1 ) and will take 2-3 days to 11 drain. For a shallow sandy soil with low organic matter content and a saturated hydraulic conductivity (K sat ) of 12~1 0 cm hr -1 it is possible that during a period of intense rainfall over several hours a substantial proportion of 13 rainfall iodine may be lost from the topsoil. Under typical rainfall conditions however, the rate of iodine 14 reactions in the topsoil are sufficiently rapid for the majority of the iodine to be retained in this layer. Figure  15 10 demonstrates retention of the iodine in the topsoil for the sandy loam soils from the Sutton Bonington sites 16 where measured iodine:carbon (I:C) ratios in soil are plotted as a function of depth for the woodland and 17 arable soil profiles. The I:C ratio increases with depth for both soils demonstrating that whilst the majority of 18 iodine is retained in the top soil the smaller amounts of humus present at depth have a high iodine 19 concentration compared to the more abundant organic matter in the topsoil. Thus iodine moving beyond the 20 topsoil during rainfall or drainage events appears to be effectively retained in the deeper soil horizons by the 21 substantial adsorption capacity provided by relatively small amounts of humus. The capacity of topsoil and 22 subsoil to effectively scavenge iodine from drainage water is supported by the low concentrations of iodine 23 (typically <5 g L -1 ) reported in river waters and the observation that iodine speciation in freshwater tends to 24 be dominated by organic forms (e.g. Reifenhauser & Heumann, 1990). This study demonstrates that iodine added to soil is rapidly transformed from inorganic to organic forms. 3 Transformation of inorganic iodine into organic forms occurs rapidly in the soil solution and the rate of loss of 4 iodine from the soil solution is dependent upon its speciation, with iodide being lost more rapidly (minutes-5 hours) than iodate (hours-days) especially in high organic matter soils. The ultimate fate of iodine added to 6 soil appears to be incorporation into soil organic matter via formation of intermediates e.g. HOI or I 2 . Abiotic 7 reduction of IO 3 -, or oxidation of I by solid or aqueous organic matter are likely to be the main mechanisms by 8 which these intermediates are formed (although this work provides no specific evidence for this) as the 9 reaction rates observed appear to be too fast for biological processes to play a significant role. It appears that 10 inorganic adsorption of iodide and iodate plays only a minor, and probably transient, role in retention of iodine 11 in soils. Rates of iodine loss are greater at higher temperatures with the rate almost doubling as temperatures 12 increase from 10 to 20 o C. 13 14 Using a spherical diffusion modelling approach with instantaneous adsorption, that has been optimised across 15 all the studied soils for iodate and iodide, this work demonstrates that it is possible to predict iodine behaviour 16 as a function of pH, soil organic carbon, oxide content and temperature.

Model Equation Reference
Irreversible   Table 2. Quoted residual standard deviations are the average for both temperatures.   Table 2. Quoted residual standard deviations are the average for both temperatures.           Table 2). Model parameters (p(D/r 2 ) and kd) were estimated from the soil variables displacement of one residual standard deviation (RSD).