Influence of Temperature, Relative Humidity, and Soil Properties on

Article pubs.acs.org/est

Influence of Temperature, Relative Humidity, and Soil Properties on the Soil−Air Partitioning of Semivolatile Pesticides: Laboratory Measurements and Predictive Models Cleo L. Davie-Martin,†,∥ Kimberly J. Hageman,*,† Yu-Ping Chin,‡ Valentin Rougé,†,⊥ and Yuki Fujita§,∇ †

Department of Chemistry, University of Otago, Dunedin 9016, New Zealand School of Earth Sciences, The Ohio State University, Columbus, Ohio 43210, United States § Department of Mathematics and Statistics, University of Otago, Dunedin 9016, New Zealand Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on August 28, 2015 | http://pubs.acs.org Publication Date (Web): August 21, 2015 | doi: 10.1021/acs.est.5b02525

‡

S Supporting Information *

ABSTRACT: Soil−air partition coefficient (Ksoil‑air) values are often employed to investigate the fate of organic contaminants in soils; however, these values have not been measured for many compounds of interest, including semivolatile current-use pesticides. Moreover, predictive equations for estimating Ksoil‑air values for pesticides (other than the organochlorine pesticides) have not been robustly developed, due to a lack of measured data. In this work, a solid-phase fugacity meter was used to measure the Ksoil‑air values of 22 semivolatile currentand historic-use pesticides and their degradation products. Ksoil‑air values were determined for two soils (semiarid and volcanic) under a range of environmentally relevant temperature (10−30 °C) and relative humidity (30−100%) conditions, such that 943 Ksoil‑air measurements were made. Measured values were used to derive a predictive equation for pesticide Ksoil‑air values based on temperature, relative humidity, soil organic carbon content, and pesticide-specific octanol−air partition coefficients. Pesticide volatilization losses from soil, calculated with the newly derived Ksoil‑air predictive equation and a previously described pesticide volatilization model, were compared to previous results and showed that the choice of Ksoil‑air predictive equation mainly affected the more-volatile pesticides and that the way in which relative humidity was accounted for was the most critical difference.

■

INTRODUCTION Soil is an important medium governing the fate of pesticides in the environment. Even decades after pesticide application and/ or production has ceased, soils remain a prominent storage reservoir and can become a source of semivolatile pesticides to the atmosphere.1 Consequently, pesticide volatilization from soil can be an important exposure route for humans, particularly agricultural workers, and nontarget organisms. Moreover, short- to long-range atmospheric transport of volatilized pesticides can lead to their widespread dispersal in the environment.2 The soil−air partition coefficient (Ksoil‑air) describes the equilibrium distribution of chemicals between soil and air and is therefore used in chemical fate models to predict or understand the behavior of chemicals in multiphase environments involving soils (e.g., pesticide volatilization models3,4). Ksoil‑air values have previously been measured under both laboratory-controlled5−8 and field conditions9−12 for several compound classes, including polychlorinated biphenyls (PCBs),5,6,10,11 chlorobenzenes,5,6 polycyclic aromatic hydrocarbons (PAHs),5,8,10,12 and organochlorine pesticides (OCPs) and their degradation products.7−9 However, measured Ksoil‑air values have not been previously reported for any pesticide class other than the OCPs. Whereas field experiments provide the © XXXX American Chemical Society

most realistic conditions, they are often hindered by nonequilibrium processes caused by fluctuations in the meteorological conditions across the experimental period and are often designed to measure concentration gradients of pesticides in the air above the soil surface, rather than equilibrium partition coefficients directly.13 In laboratory experiments, environmental parameters can be more tightly constrained, thus allowing for a systematic investigation into the individual and combined effects of varying environmental conditions. Laboratory measurements of Ksoil‑air values have previously been made using a solid-phase fugacity meter.14 With this approach, air flows through a contaminated soil at a rate slow enough for semivolatile contaminants to establish equilibrium distribution between the soil matrix and air. The air exiting the soil passes through a sorbent trap to which the gas-phase contaminants sorb. The ratio of contaminant concentrations in air and soil are combined to obtain their Ksoil‑air values. Due to the time and expense involved in measuring Ksoil‑air values, efforts have been made to use measured values to Received: May 22, 2015 Revised: July 25, 2015 Accepted: August 10, 2015

A

DOI: 10.1021/acs.est.5b02525 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology develop equations to predict Ksoil‑air values for chemicals and conditions under which measurements have not been made.5−8,15 Because a chemical’s affinity for the organic matter in soil is commonly assumed to be proportional to its affinity for octanol (as expressed by the equilibrium octanol−air partition coefficient, Koctanol‑air), initially developed predictive equations for Ksoil‑air values were in the form of eq 1.5

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on August 28, 2015 | http://pubs.acs.org Publication Date (Web): August 21, 2015 | doi: 10.1021/acs.est.5b02525

K soil − air = 0.411·fOC ·ρsoil ·Koctanol − air

predicted pesticide volatilization losses from soil, as calculated with a previously described pesticide volatilization model.3

■

MATERIALS AND METHODS Pesticide Selection. Semivolatile pesticides were selected with the aim of obtaining a suite that (a) exhibited a range of vapor pressures and log Koctanol‑air properties, (b) included some OCPs previously investigated,7−9 and (c) contained pesticides that are currently registered for use in New Zealand and/or the United States. Ultimately, nine legacy pesticides and degradation products (cis-chlordane, dichlorodiphenyldichloroethane (o,p′-DDD and p,p′-DDD), dichlorodiphenyldichloroethylene (o,p′-DDE and p,p′-DDE), dichlorodiphenyltrichloroethane (o,p′-DDT and p,p′-DDT), dieldrin, and methoxychlor), 11 current-use pesticides (atrazine, chlorpyrifos, cyprodinil, dacthal, eptam, metribuzin, molinate, pyrimethanil, terbuthylazine, triallate, and trifluralin), and the recently deregistered components of technical endosulfan (I and II) were chosen for measurements of Ksoil‑air. Soil Preparation and Characterization. Two natural soils, classified as semiarid and volcanic, were selected. A key criterion was that they have significantly different organic carbon contents; however, they varied in other significant ways as shown in Table 1, which lists the important physicochemical

(1)

where f OC is the fraction of organic carbon in soil and ρsoil is the particle density of the soil. However, soil−air partitioning is a complex process that involves both absorption into soil organic carbon and adsorption to mineral surfaces.16 These processes are governed by the soil composition as well as temperature and relative humidity. Thus, using measured Ksoil‑air values for PCBs and chlorobenzenes, Hippelein and McLachlan developed a predictive equation for Ksoil‑air using a semiempirical approach based on eq 1 (describing absorptive interactions) and the combined influence of temperature (T, K) and relative humidity (RH, %), which are primarily associated with adsorptive interactions eq 2.6 ⎛ g m −3 ⎞ ⎟ = [2·fOC ·Koctanol − air(298.15K)]· K soil − air ⎜ −3 ⎝gm ⎠ [(

e

Table 1. Selected Physicochemical Properties of the Semiarid and Volcanic Soils

Δsoil − air U 1 1 )( − ) − 0.0437(RH − 100)] R T 298.15K

(2) −1

−1

where R is the ideal gas constant (0.008314 kJ K mol ) and Δsoil‑airU is the internal energy of soil to air phase transfer (kJ mol−1). Using the terminology defined by Hippelein and McLachlan,6 the normalization constant and RH-sensitivity factor (B value) in this equation are 2 and −0.0437, respectively. In a previous publication,3,4 we developed an expression analogous to eq 2 that was fitted to measured values of Ksoil‑air reported specifically for the OCPs.7−9 This fitting exercise resulted in a normalization constant of 0.411 and a B value of −0.20.3 In addition to using this equation with OCPs, we used it with several classes of current-use pesticides due to the lack of alternative predictive equations and measured values for these pesticides. However, as a class of compounds, the current-use pesticides are structurally diverse, possess various functional groups, and cover a broad range of polarity. This suggests that their soil−air partitioning behavior could vary significantly from that of the PCBs, chlorobenzenes, and even the OCPs. Thus, it is imperative that measurements of Ksoil‑air be made for semivolatile current-use pesticides to expand the training sets needed to improve the accuracy and predictive capability of models used to assess their environmental fate. The primary objective of this work was to measure the Ksoil‑air values of 22 semivolatile current-use and legacy pesticides and degradation products in two soils under a variety of environmentally relevant temperature (10−30 °C) and relative humidity (30−100%) conditions using a solid-phase fugacity meter. We then used these values as a training set to derive an equation for predicting pesticide log Ksoil‑air values based on the combined influence of pesticide-specif ic log Koctanol‑air values and temperature, relative humidity, and f OC. Finally, we investigated the effect that selection of Ksoil‑air predictive equation (i.e., the formerly used semiempirical equation versus the newly developed multiple linear regression equation) had on

soil properties

semiarid soil

volcanic soil

f OC sand, 0.05−2.0 mm silt, 0.002−0.05 mm clay, < 0.002 mm soil texture particle density pH (dry soil) dominant minerals

0.92% 60% 24% 16% sandy loam 2.69 g cm−3 5.8 mica, chlorite, feldspar, clay-vermiculite

7.5% 33% 34% 32% clay loam 2.17 g cm−3 5.9 halloysite, kaolin, crystobalite

properties of the soils. Both soils were dried and sieved to less than 1 mm diameter particle size and using a procedure described elsewhere (Section 1 of the Supporting Information, SI),5,17 then contaminated with the 22 pesticides and degradation products to a concentration of ∼1−2 mg kg−1. Measurement of Ksoil‑air Values. Soil−air partition coefficients were measured using a solid-phase fugacity meter (Figure S1) similar to that described by Hippelein and McLachlan.5 In brief, a glass sample chamber, with an aluminum foil covering used to prevent photodegradation, was filled with 200−500 g of contaminated soil and placed inside a temperature-controlled chamber. Nitrogen, which was precleaned by passage through a 10 g Florisil sorbent trap, was used as a proxy for “air” and was delivered from a pressurized cylinder. The relative humidity of the nitrogen stream was controlled by passage through two temperature-controlled bubblers, containing distilled water, that were set to the dew point temperature. The humidified nitrogen stream then passed through the soil-filled sample chamber, a particle trap (i.e., 30 mm glass fiber filter), and finally, a sorbent trap filled with ∼13 g XAD-2 (a styrene/divinylbenzene copolymer that was used to collect gas-phase pesticides). The particle trap prevented pesticides bound to soil particles from entering the B


Article


Environmental Science & Technology

Figure 1. Linear regressions of log Ksoil‑air with reciprocal temperature (panels A and B, where RH = 70%) and relative humidity (panels C and D, where T = 15 °C for semiarid soil and 25 °C for volcanic soil) for selected pesticides using semiarid (panels A and C) and volcanic (panels B and D) soils. In panel C, the dotted line is an example of the “hockey stick”-type trend observed with the semiarid soil (not fitted).

mass of pesticide collected in the XAD-2 divided by the total volume of nitrogen sampled). Pesticide concentrations were also measured in three 1.2 g subsamples of soil removed from the chamber. The dimensionless Ksoil‑air of each pesticide was calculated from the ratio of its mean concentration in the three soil subsamples to its concentration in the nitrogen stream. Measurements of Ksoil‑air were made for various combinations of environmentally relevant temperatures (10−30 °C) and relative humidity (30−100%) conditions. The details of each experiment (including temperature, relative humidity, f OC, flow rate, and the resulting log Ksoil‑air values for all pesticides) are reported in Table S1. Extraction and Quantification of Pesticides. Pesticides were extracted from XAD-2 and soil subsamples with pressurized liquid extraction (PLE) using an accelerated solvent extractor system (Dionex, Sunnyvale, CA). Samples were packed into individual stainless steel extraction cells (Figure S2) and spiked with a mixture of isotopically labeled pesticide surrogates (containing d5-atrazine, d10-chlorpyrifos, d8-p,p′DDE, d8-p,p′-DDT, d6-alpha-HCH, and d14-trifluralin) prior to extraction to account for potential loss of target analytes during sample workup. The sample extracts were solvent exchanged to ethyl acetate and concentrated for analysis using an Agilent 6890N gas chromatograph coupled to an Agilent

XAD-2 trap, thus ensuring the concentrations of pesticides measured in the XAD-2 trap were of gaseous origin. Flow meters (0.05−0.5 L min−1, Key Instruments, Trevose, PA) were placed at the inlet and outlet of the system to monitor the flow rate of the nitrogen stream and a gas meter (0.0−0.5 L min−1, Parkinson Cowan Industrial Products, England) measured the total volume of nitrogen sampled in each experiment. The temperature and relative humidity of the nitrogen stream immediately prior to the sample chamber were logged at 15 s intervals throughout all experiments with an Omega RH511 Meter (Stamford, CT). Before starting each experiment, contaminated soils were equilibrated with the humidified nitrogen stream by allowing it to flow through the apparatus overnight. Additionally, a second probe (HMP45A, Vaisala; Woburn, MA) was periodically installed at the outlet of the sample chamber and used to ensure that humidity conditions in the soil were stable. Typically, each experiment ran for 22 h with a nitrogen flow rate of ∼0.1 L min−1. In preliminary experiments, we verified that this flow rate was slow enough to allow equilibrium to establish between the pesticides bound to the soil matrix and the nitrogen stream as it passed through the sample chamber. At the completion of the experiment, gas-phase pesticide concentrations were measured in the XAD-2 sorbent trap (i.e., C


Article


Table 2. Regression Coefficients from Plots of Log Ksoil‑air = A/T + C for Pesticides Exhibiting Significant Correlations (p < 0.05). Δsoil‑airU is Equal to 2.303·A·R Where R Is the Ideal Gas Constanta


semiarid soil ( f OC 0.92%) pesticide

A

cis-chlordane chlorpyrifos dacthal o,p′-DDD p,p′-DDD o,p′-DDE p,p′-DDE o,p′-DDT p,p′-DDT endosulfan I eptam metribuzin molinate trifluralin

9978 17 145 18 094 12 525 11 215 11 660 10 928 12 276 8760 12 475 8186 9395 10 092 9275

pesticide

A

cis-chlordane chlorpyrifos dacthal o,p′-DDE p,p′-DDE o,p′L-DDT dieldrin endosulfan I endosulfan II eptam molinate triallate trifluralin

8097 7358 5071 10 204 8549 5707 6933 10 927 −3380 4816 7369 11 526 8852

literature ( f OC 1−44%) Δsoil‑airU

log Ksoil‑air

Δsoil‑airU

r2

(kJ mol−1)

(298.15 K, 70%)

(kJ mol−1)

0.970 0.968 0.939 0.930 0.843 0.963 0.969 0.963 0.844 0.952 0.964 0.468 0.961 0.972

191 328 347 240 215 223 209 235 168 239 157 180 193 178 Δsoil‑airU

log Ksoil‑air

Δsoil‑airU

C

r2

(kJ mol−1)

(298.15 K, 70%)

(kJ mol−1)

−20.44 −16.90 −8.89 −27.44 −21.46 −11.21 −15.86 −29.81 20.77 −10.62 −18.71 −31.99 −23.47

0.904 0.630 0.378 0.869 0.745 0.574 0.569 0.772 0.632 0.737 0.723 0.809 0.829

155 141 97 195 164 109 133 209 −65 92 141 221 170

6.72 7.78 8.12 6.78 7.21 7.93 7.40 6.84 9.43 5.54 6.01 6.67 6.22

93−104

C −26.82 −50.50 −53.59 −34.54 −29.74 −32.53 −29.76 −33.90 −21.64 −35.26 −22.10 −22.91 −28.05 −25.14 volcanic soil ( f OC 7.5%)

6.64 93−104 7.01 7.09 7.46 109−117 7.88 83−106 6.58 96−107 6.90 102−108 7.27 105−108 7.74 6.58 5.36 8.60 5.80 5.97 literature ( f OC 1−44%)

96−107 102−108 105−108 96−97

a Data was collected at 70% RH across a range of temperatures for semi-arid and volcanic soils. Calculated log Ksoil‑air values at 298.15 K are listed along with the range of Δsoil‑airU values reported in the literature for measurements at 100% RH with different soils (soybean, muck crop, Hawaiian, and silty clay loam).

■

RESULTS AND DISCUSSION All measured pesticide Ksoil‑air values and the conditions under which they were measured are shown in Table S1. Ksoil‑air values for 14 of the 22 pesticides investigated have not been reported previously. Influence of Temperature. Values of Ksoil‑air were highly dependent on temperature, as shown for five representative pesticides in Figure 1 (panels A and B). Ksoil‑air decreased by approximately one order of magnitude for each 10 °C increase in temperature. A significant correlation (p < 0.05) was observed between log Ksoil‑air and reciprocal temperature for 14 of the 22 pesticides investigated when using the semiarid soil and for 13 of the pesticides when using the volcanic soil. Regression plots for all 22 investigated pesticides are shown in Figure S3. For cases in which correlations were significant, the slope (A), intercept (C), Δsoil‑airU, and calculated log Ksoil‑air at 298.15 K are displayed in Table 2. Lack of correlation occurred for pesticides whose concentrations in the XAD-2 trap were below detection limits at several of the experimental temperatures. This tended to occur for the less volatile pesticides (such as methoxychlor, several of the fungicides, and the triazines) under lower experimental temperatures (Figure S3).

5957B mass selective detector (GC-MS) (Santa Clara, CA) using selective ion monitoring. Soil extracts were analyzed with electron impact ionization (EI) mode. For the XAD-2 extracts, electron capture negative ionization (ECNI) mode was used to achieve greater sensitivity for those pesticides that were active in ECNI mode; otherwise EI mode was used. Pesticides were quantified using a 12-point calibration curve based on the ratios of the target peak areas to their corresponding surrogate peak areas (i.e., the isotope dilution method). Recovery analysis, performed by spiking the target pesticides into a precleaned or uncontaminated matrix, gave mean pesticide recoveries of 86, 97, and 84% for XAD-2, semiarid soil, and volcanic soil matrices, respectively. Further details about the chemicals, extraction methods, quantification methods, and quality control procedures are provided in the SI (Section 1, Tables S2 and S3). Statistical Methods. Correlation analyses were conducted with Microsoft Excel 2010. Multiple linear regressions were performed with R version 3.1.1; details regarding the development and validation of the predictive equation are provided in Section 2 and Table S5 of the SI. D


Article



Figure 2. Comparison of log Ksoil‑air values of select pesticides measured in this study under a range of relative humidity levels with those reported in the literature. Literature measurements include temperature regressions for different soils obtained from laboratory measurements with a solid-phase fugacity meter (black lines) and measurements carried out under field conditions for muck crop (44% f OC, ∼ 23.9 °C, ∼ 87% RH) and soybean soils (1.8% f OC, ∼ 16.5 °C, ∼ 81% RH) (black squares). All data were normalized to 1% f OC.

whether similar saturation effects occurred within the volcanic soil. Using a linear regression equation of the form, log Ksoil‑air = B/2.303·RH + A, the RH-sensitivity factor (B) and intercept (A) values were calculated for each pesticide and are displayed in Table S4 for the pesticides whose linear correlations were significant (p < 0.05). The average B value was −0.068 for the semiarid soil (ranging from −0.097 to −0.029) and −0.026 for the volcanic soil (ranging from −0.037 to −0.016). Influence of Soil Type. Pesticide Δsoil‑airU values (Table 2) and B values (Table S4) were both larger in absolute magnitude for measurements with the semiarid soil than with volcanic soil, suggesting that temperature and relative humidity had a greater influence on pesticide partitioning into semiarid soil. This may be explained by the semiarid soil having a lower organic carbon content (0.92%) than the volcanic soil (7.5%) and therefore, a larger inorganic component, resulting in more adsorptive interactions with pesticides and a higher activation energy threshold for volatilization. Additionally, the differences in experimental temperatures could also partially explain why smaller B values were observed for volcanic soil (i.e., experiments investigating the influence of relative humidity were carried out at 15 °C for semiarid soil and 25 °C for volcanic soil); under higher temperatures, transfer of pesticides from the soil to the air requires less energy, thus lowering the B value. Such an effect was previously observed for the PCBs.6 Comparisons to Soil−Air Partitioning Data for Organochlorine Pesticides in the Literature. Measurements of Ksoil‑air have previously been reported in the literature for eight of the pesticides and degradation products investigated in this study (cis-chlordane, o,p′-DDD, p,p′-DDD, o,p′-DDE, p,p′-DDE, o,p′-DDT, p,p′-DDT, and dieldrin).7−9 Figure 2 compares measured values of log Ksoil‑air made with semiarid and volcanic soils in the present study (containing

Thus, the lack of correlation for certain pesticides was mainly attributable to analytical limitations rather than pesticide behavior. Influence of Relative Humidity. Relative humidity also exerted a strong influence on the soil−air partitioning of pesticides (Figure 1, panels C and D). For an increase in relative humidity from 60 to 90%, values of Ksoil‑air decreased by factors of 7−52 with semiarid soil and 2−4 with volcanic soil, depending on the specific pesticide. This trend is attributed to water molecules in the air competitively displacing (through hydrogen bonding) pesticide molecules bound (via adsorption) to the soil mineral surfaces.15 Significant correlations (p < 0.05) were observed between log Ksoil‑air values and relative humidity for 13 of the 22 pesticides when using the semiarid soil and for nine pesticides when using the volcanic soil (regression plots for all investigated pesticides are displayed in Figure S4). Interestingly, a “hockey stick”-type relationship was observed between relative humidity and log Ksoil‑air values with the semiarid soil (Figure 1, panel C); that is, relative humidity had minimal effect on soil−air partitioning above a certain level for this soil. Previous work has shown that above ∼98% relative humidity, mineral surfaces become saturated with water molecules, depleting the surface area available for pesticide binding and disrupting intermolecular interactions.15 Thus, at high relative humidity, the partitioning of organic chemicals into soil can become almost exclusively controlled by absorption into soil organic matter and adsorption of gasphase pesticides to mineral surfaces becomes negligible. In our study, it therefore appears that the mineral surfaces in the semiarid soil were more readily saturated under low relative humidity conditions than the mineral surfaces used in previous studies. In the absence of pesticide Ksoil‑air measurements with volcanic soil at 100% relative humidity, it is difficult to assess E


Article



it is important to be able to predict how different pesticides behave in soil. For this purpose, single physicochemical properties, such as log Koctanol‑air, are often used5,6,18,19 to account for pesticide-specific interactions with soil. Initially, we derived single-parameter equations (for temperatures between 10 and 30 °C) to predict values of log Ksoil‑air based on the subcooled liquid vapor pressure (pL*) and log Koctanol‑air (Figures S5 and S6). Although all regressions were statistically significant (p < 0.05), the resulting predictive equations are not particularly useful due to the additional dependence of Ksoil‑air on relative humidity and soil type. Multiple linear regression was therefore selected as a more appropriate approach and used to derive a complete predictive equation for pesticide log Ksoil‑air values using log Koctanol‑air (at 298.15 K), inverse temperature (1/T, K−1), relative humidity (RH, %), and the logarithm of the soil organic carbon content (log f OC, %) as the explanatory variables (eq 3, including β coefficients and their associated standard errors). Full details of the derivation and cross-validation procedures are provided in Section 2 of the SI.

data measured under different temperature and relative humidity conditions) with those reported previously, including temperature regressions reported for laboratory measurements of Ksoil‑air from different soils at 100% RH using a solid-phase fugacity meter7,8 and measurements carried out under field conditions9 (all normalized to 1% f OC by division of Ksoil‑air by f OC). At higher temperatures, there was significant overlap between the log Ksoil−air values measured for the six soils used in the present and previous studies. However, the results clearly demonstrated that the magnitude of soil−air partitioning was highly dependent on soil composition. While the slopes of the temperature regressions reported in the literature were consistent with one another, those measured with the semiarid and volcanic soils in this study were generally steeper, as confirmed by the Δsoil−airU values (Table 2). For semiarid soil, pesticide Δsoil−airU values were typically twice as high as the average literature values and a factor of ∼1.5 times as high for the volcanic soil. Variations in the soil characteristics (e.g., f OC, mineral composition, specific surface area, and density), and their influence on the soil’s adsorptive capacity could explain these observations. Interestingly, the field measurements of OCP Ksoil‑air values with muck crop and soybean soils showed better agreement with the semiarid and volcanic soils than with the temperature regressions based on laboratory measurements with the same muck crop and soybean soils (Figure 2). It is noteworthy that both the field measurements for muck crop and soybean soils and the laboratory-based measurements for semiarid and volcanic soils were carried out under lower relative humidity levels (70−88%)9 than the previous laboratory experiments with muck crop, soybean, Hawaiian, and silty clay loam soils (100% RH);7,8 this could explain the differences in Ksoil‑air values observed in Figure 2. At 70% RH, mineral surfaces are covered by less than 4−5 monolayers of adsorbed water, and as a result, interactions with adsorbed pesticides are still possible.15 In contrast, at 100% RH, the mineral surfaces are effectively covered by a “bulk” water layer and pesticide interaction with the soil is limited to absorption into the organic carbon alone.15 Thus, the higher Δsoil‑airU values measured for semiarid soil in this study (compared to those reported in the literature) were likely a combination of stronger pesticide interactions with the soil due to the combined influence of absorption into the organic carbon and adsorption to the mineral surfaces. Although one would expect f OC to largely control soil−air partitioning in cases where absorption dominates, this is not necessarily true where both adsorption and absorption contribute and this effect reveals itself here. Hippelein and McLachlan observed a similar phenomenon with PCBs.6 Taken together, these studies show that Δsoil‑airU values vary significantly with soil type and relative humidity, and that caution should be used when assumptions about the consistency of Δsoil‑airU values are made. B values have not been reported previously for any of the pesticides investigated in this study. However, an average B value of −0.0437 was reported for similar experiments performed with a suite of PCB congeners,6 which fell within the range of pesticide B values measured in the present study (Table S4). Derivation of a Predictive Model. The results presented herein have shown that for a given pesticide, soil−air partitioning is highly dependent on temperature, relative humidity, and soil characteristics. Despite these complexities,

log K soil − air = − 26.2(± 1.0) + 0.714(± 0.018)· log Koctanol − air + 8291( ±305) ·1/T − 0.0128( ±0.0014) · RH + 0.121(± 0.047)· log fOC (rmse = 0.57, n = 943, and adjusted r 2 = 0.705)

(3)

A plot of measured versus predicted log Ksoil‑air values, obtained using eq 3, revealed that 92% of the predictions fell within one order of magnitude of the 1:1 line (Figure 3). The

Figure 3. Comparison of measured versus predicted log Ksoil‑air values for all pesticides using eq 3. The solid line shows a 1:1 relationship and the two dashed lines represent a range of ±1 log unit from the 1:1 line.

root mean squared error (rmse) associated with eq 3 (0.57) was higher than rmse values previously reported for other equations used to predict partition coefficients for homogeneous matrices, such as organic solvents and water (rmse 0.1− 0.2),20 or for those involving a heterogeneous matrix with compounds from structurally similar compound classes (rmse typically 0.3 or higher).6,20 However, this was not surprising since eq 3 was derived from measurements of structurally diverse pesticides with two heterogeneous soils under a range of temperature and relative humidity conditions. Moreover, the rmse associated with eq 3 (based on experimentally determined Ksoil‑air measurements) was considerably lower than that F


Article



Figure 4. Chemical space diagrams illustrating the percent change in predicted 24-h volatilization losses calculated when using log Ksoil‑air values estimated using the multiple linear regression approach (eq 3) versus the semiempirical approach (eq 7) for 224 pesticides under three environmental scenarios. A negative percent change indicates the value was lower when calculated with eq 3. “No change” was defined as

Influence of Temperature, Relative Humidity, and Soil Properties on

Recommend Documents