Pesticide Nonextractable Residue Formation in Soil: Insights from

Publication Date (Web): August 15, 2012 ... It was found that only half of 73 pesticide degradation time series from a homogeneous soil source allowed...
1 downloads 0 Views 1MB Size
Download PDF
Policy Analysis pubs.acs.org/est

Pesticide Nonextractable Residue Formation in Soil: Insights from Inverse Modeling of Degradation Time Series Martin Loos,† Martin Krauss,‡ and Kathrin Fenner*,†,§ †

Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dübendorf, Switzerland Department of Effect-Directed Analysis, UFZ - Helmholtz Centre for Environmental Research, 04318 Leipzig, Germany § ETH Zurich, Institute of Biogeochemistry and Pollutant Dynamics, 8092 Zurich, Switzerland ‡

S Supporting Information *

ABSTRACT: Formation of soil nonextractable residues (NER) is central to the fate and persistence of pesticides. To investigate pools and extent of NER formation, an established inverse modeling approach for pesticide soil degradation time series was evaluated with a Monte Carlo Markov Chain (MCMC) sampling procedure. It was found that only half of 73 pesticide degradation time series from a homogeneous soil source allowed for well-behaved identification of kinetic parameters with a four-pool model containing a parent compound, a metabolite, a volatile, and a NER pool. A subsequent simulation indeed confirmed distinct parameter combinations of low identifiability. Taking the resulting uncertainties into account, several conclusions regarding NER formation and its impact on persistence assessment could nonetheless be drawn. First, rate constants for transformation of parent compounds to metabolites were correlated to those for transformation of parent compounds to NER, leading to degradation half-lives (DegT50) typically not being larger than disappearance half-lives (DT50) by more than a factor of 2. Second, estimated rate constants were used to evaluate NER formation over time. This showed that NER formation, particularly through the metabolite pool, may be grossly underestimated when using standard incubation periods. It further showed that amounts and uncertainties in (i) total NER, (ii) NER formed from the parent pool, and (iii) NER formed from the metabolite pool vary considerably among data sets at t→∞, with no clear dominance between (ii) and (iii). However, compounds containing aromatic amine moieties were found to form significantly more total NER when extrapolating to t→∞ than the other compounds studied. Overall, our study stresses the general need for assessing uncertainties, identifiability issues, and resulting biases when using inverse modeling of degradation time series for evaluating persistence and NER formation.

■

INTRODUCTION Upon their application to agricultural soils, pesticides undergo a range of transformation and sequestration processes, as do other organic contaminants.1 Specifically, it has been shown that in most cases a fraction of the applied compound or its metabolites becomes unextractable with simulated soil solutions or organic solvents and remains so over periods of years.2 This fraction is referred to as nonextractable residues (NER), and is operationally defined as compounds that persist in the soil matrix in the form of the parent substance or its metabolite(s), upon application of extraction procedures that do not substantially change the compounds themselves or the structure of the matrix.3 The significance of NERs and whether and how their formation should be considered in chemical risk assessment is still under debate. This debate centers on whether NERs can be considered inaccessible and hence ineffective toward biota, or whether NERs are partially bioavailable or can become so over time with soil organic matter (SOM) turnover.4,5 Any clarification in this debate must address the questions of (i) what molecular entity is forming NER, i.e., the parent compound or its structurally modified metabolite(s), (ii) which mechanisms of NER formation apply, and (iii) on © 2012 American Chemical Society

which time scales. Known interaction mechanisms between organic contaminants and soil with different implications for release include different sorption mechanisms, physical entrapment following intraorganic matter or pore diffusion,1 covalent bond formation with SOM,6 and microbial metabolism, effectively assimilating carbon derived from the contaminant into biomolecules.7 General answers to the above questions are hard to derive. One of the reasons for this is that studies on NER formation have mainly been carried out for individual compounds in a regulatory context, but not for larger sets of systematically varied compounds under the same soil incubation conditions or using the same extraction procedures. Moreover, in the majority of investigations on NER formation, 14C-labeled compounds are spiked into soil and the extractable, nonextractable, and mineralized radioactivity are determined over time. Unfortunately, this procedure does not allow for direct identification of the molecular entity forming NER.8 A Received: Revised: Accepted: Published: 9830

February 7, 2012 July 24, 2012 August 15, 2012 August 15, 2012 dx.doi.org/10.1021/es300505r | Environ. Sci. Technol. 2012, 46, 9830−9837

Environmental Science & Technology

Policy Analysis

Assessment Reports (DARs) publicly listed by the European Food Safety Authority for one homogeneous standard soil, namely Speyer 2.2 soil.17 A second set of Speyer 2.2 data was provided by BAYER Crop Science (Monheim, Germany). Using data from one test soil only, we aimed at achieving higher consistency in terms of experimental conditions among the different studies with respect to soil type, pH, organic matter content, soil humidity, and microbial activity. Altogether, 73 time series were analyzed for 55 pesticides, of which 7 had duplicate times series from independent experiments, and for 11 metabolites as initial compounds (compounds are listed in Table S1 of the Supporting Information (SI)). Six more time series for Speyer 2.2 were originally found, but excluded from further analysis due to either incomplete mass balances, poor data coverage, or clear deviation from first-order kinetics. Only three compounds are overlapping with those studied in ref 8, resulting in a nicely complementary data set. The time series used had mostly been generated in accordance with OECD guideline 307.18 In short, they resulted from spiking of a 14C-labeled compound to incubation flasks of each 50−200 g of sieved Speyer 2.2 soil under standardized conditions (darkness, incubation temperatures 20, 22, or 25 °C, soil moisture 2.0 or 2.5 pF, mean pH = 6.1 (SD = 0.5), mean % OC = 2.3 (SD = 0.3), mean CEC = 10.0 meq/100 g (SD = 2.7), applied concentrations as recommended by pesticide use instructions). At time intervals, the percentages of applied radioactivity (%AR) present as parent compound (P), sum of extractable metabolites (M), NER (N), and volatiles (V) were determined from these batches (Figure 1a). P and M were

more in-depth specification of NER through experimental “release and identify” techniques9 is time-consuming and difficult, and it seems unlikely that the characterization of NER will become a routine in risk assessment.4 Another reason for difficulties in interpreting data from soil degradation studies is their duration. The Uniform Principles in support of Regulation (EC) 1107/2009 concerning the placing of plant protection products on the market10 state that “no authorization of pesticides shall be granted if, in soil laboratory tests, non-extractable soil residues are formed at >70% of initial radioactivity after 100 days with mineralization to CO2 at 120 d) based on their DT50 values, whereas 7 compounds would be deemed persistent based on their DegT50 values. NER Formation Estimated from Model Extrapolation. Model extrapolation to different time points shows that only a small fraction (approximately 10%) of compounds have reached 99% of their final NER levels at t = 100 d (see Figure 3 and Figure S6). After 365 days, this fraction has risen to 43% of all data sets. Evaluating amounts of NER formation solely from such arbitrary study durations can therefore be highly misleading. Instead, the theoretical potential of a compound to form NER may be predicted from subsampling of Θ̂ and extrapolation of the model to t → ∞. Figure 3 lists amounts and uncertainties for total N(∞), N(∞) formed directly from the parent compound, NP(∞), and N(∞) formed through the metabolite pool, NM(∞). Thus, 64% of the data sets have amounts of N(∞) > 50%AR; 36% of the data sets even show N(∞) > 70%ARthe latter being one of two regulatory criteria for rejection of a pesticide as an active substance due to extensive NER formation when met within 100 d. It needs to be noted though that such extrapolations do not address NER dynamics at prolonged time spans, including potential NER remobilization, transformation, or mineralization through SOM turnover.43 Moreover, uncertainties in N(∞), NP(∞), and NM(∞) are high and vary considerably among data sets. Uncertainties in NM(∞) are higher than in NP(∞) for the well-behaved data sets, most likely because NM(∞) is afflicted by uncertainties in all, kPM, kMV, and kMN, and because it subsumes several different metabolite pools into one, i.e., M(t). For some well-behaved data sets (e.g., IDs 11, 27, 30, 57, 59, 66, 78, and 89 in Table S1), high uncertainties in N M (∞) propagate to high uncertainties in N(∞). These time series often have overall slow kinetics, with M(t) still increasing at the end of the

Figure 2. Log10-scaled scatterplot of the MCMC PDF medians for kPM versus kPN. Spans define the 95% interquantile ranges of the PDFs.

11, a clear correlation between the median values of kPN and kPM exists for the well-behaved cases, covering 3 orders of magnitude. For most of these 36 data sets, the ratio kPN/kPM lies between 1 and 0.1, with median, minimal, and maximal values of 0.43, 0.03, and 1.61, respectively. In other words, transformation of parent compounds to metabolites or NER proceeds on similar temporal scales, with transformation to metabolites typically being somewhat faster than transformation to NER. For the well-behaved data sets, uncertainties in parameter estimates as expressed by the 95% confidence intervals are always larger for kPN than for kPM. In contrast, large uncertainties accompany both parameter estimates in the case of the ill-behaved data sets. However, within these increased uncertainty ranges, and although parameter estimates of illbehaved data sets are typically biased toward certain regions of the parameter space as indicated above, a correlation between kPM and kPN similar to that of well-behaved data sets is still evident. Matthies et al.11 attributed variation and correlation in kPM and kPN to varying microbial activities of different test soils used. Since we only used data from one test soil, this explanation may not apply to our case. Indeed, although different Speyer 2.2 soil batches were used in the different studies, we found little difference in microbial activities, and these were not related to differences in observed rate constants (see soil microbial activity listed in Table S1). A second factor with potentially similar influence on formation of metabolites and NER is compound bioavailability, which for most compounds is related to the organic-carbon/water partition coefficient, Koc (see Koc values in Table S1).36 However, a previous study with the same data set revealed only a weak dependency of kPM on Koc, but a much stronger dependency on structural fragments known to typically enable or inhibit enzymatic transformations.37 This points toward compound reactivity, i.e., its general tendency to be biologically transformed, to explain our findings. The recalcitrance of a compound toward degradation to metabolic intermediates was also found to be a key factor in a study comparing the 9834

dx.doi.org/10.1021/es300505r | Environ. Sci. Technol. 2012, 46, 9830−9837

Environmental Science & Technology

Policy Analysis

in N(∞) can still be accompanied by high uncertainties in both NP(∞) and NM(∞) for ill-behaved data sets. In these cases, posteriors of kPM and kPN are negatively correlated and suspected to be collinear (e.g., IDs 34, 55, 58, 61, 64, 69, and 70 in Table S1). Finally, we tested whether compounds containing aromatic amine or phenolic substructures are distinct in their total NER formation potential, i.e., N(∞). These substructures have been widely reported to form covalent bonds with SOM functional groups by oxidative cross-coupling.44−46 For this purpose, compounds with confidence intervals >25%AR for total NER formation were excluded from interpretation, leaving 47 data sets containing 5 duplicate compounds for two-sample t testing (two-sided, 1% significance level). We found that compounds with phenolic substructures (n = 18) were not significantly different from the other compounds with respect to N(∞). A significant difference could, however, be found between the group of compounds that contained aromatic amines (marked in Table S1) (n = 20) (N(∞) = 73.2 ± 13.9% AR) and those that did not (N(∞) = 46.2 ± 18.0% AR). The group of compounds with increased NER formation potential covers a range of different compound classes, including phenylureas, aromatic carbamates, pyrimidinyl sulfonylureas, sulfonamides, sulfonanilides, anilinopyrimidines, and organophosphates. NER Formation from Parent Compound and Metabolite Pools. Figure 4a shows how the relative importance of the two possible NER forming pools, i.e., NP(t) versus NM(t), evolves with time. It shows that the contributions of the two pools to N(t) vary greatly over time and among data sets. Most notably, Np(t)/NM(t) stabilizes (i.e., change in Np(t)/NM(t) < 0.01 at any consecutive time point with Δt ≥ 105 d) previous to the 100 and 365 d end point for only 49 and 66% of the data sets, respectively. The histogram given in Figure 4b reveals that only 53% of the data sets have a fraction of Np(∞) > 0.5. This means that, across our diverse set of compounds, the molecular entity dominating NER formation is about as equally often the parent compound as it is a metabolite. Thus, although kPM often exceeds kPN (Figure 2), this obviously does not imply that N(∞) is mostly formed through the metabolite pool. Some

Figure 3. Sorted median amounts of total N(∞), total N(100 d), NP(∞) from the parent pool, and NM(∞) from the metabolite pool for each data set. Bars refer to 95% confidence intervals (CIs). Numbers to the left refer to the data set IDs listed in Table S1.

experiment, and thus lack information on subsequent transformation of the metabolite pool. In contrast, low uncertainties

Figure 4. (a) Median fraction q of N(t) at time t formed directly from the parent compound (metabolite fraction = 1 − q), modeled from subsampling of Θ̂. (b) Histogram of q for N(∞), (c) with ill-behaved data sets excluded, and (d) after removal of data sets with CIs of N(∞) > 10% AR. 9835

dx.doi.org/10.1021/es300505r | Environ. Sci. Technol. 2012, 46, 9830−9837

Environmental Science & Technology

Policy Analysis

posteriori parameter identifiability over much simpler standard Fisherian techniques for kinetic model fitting. The latter techniques may either not reveal ill-behaved distributions by making a priori assumptions about parameter distributions or in the best casemay fail from nonconvergence of the underlying algorithms.12 Even worse, any generalizations may be systematically biased toward parameter sets of low collinearity and high sensibility after omission of data sets failing the fitting procedure. The analysis of ill-behaved data sets, for instance, showed that in cases where downstream pools are much faster transformed than replenished, time series may not contain enough information to make parameters unambiguously estimable. Instead, compounds constituting these pools should be identified and individually spiked to soil batches to establish time series suitable for parameter estimation with a submodel. These findings also reveal limitations for model refinement: itemizing M(t) into several pools will not improve parameter identifiability for pools faster transforming than forming. Keeping these pitfalls in mind, a large number of times series from soil degradation studies with radiolabeled compounds are publicly available for inverse modeling.17 Doing so, however, requires a case-by-case verification of parameter estimability.

instances where the metabolite pool clearly dominates are easily hydrolyzable parent compounds such as esters (IDs 28, 43, 53, and 86) and nitriles (ID 78). Also, as can be inferred from Figure 3, extensive NER formation does not coincide with a dominance of either of the two pools forming NER. It needs to be noted that the above findings are only valid for metabolites that still contain the radioactive label; if the latter gets lost during degradation, metabolites cannot longer be assessed as % AR. Furthermore, interpretation of results from inverse modeling can be biased after exclusion of problematic data sets. For example, excluding ill-behaved data sets shows little effect on the relative importance of the two pools for N(∞) (Figure 4c). In contrast, a distinct shift seemingly indicating an increased importance of Np(∞) over NM(∞) results from excluding those data sets with large uncertainty in N(∞) (Figure 4d). The latter results from the propagation of large uncertainties in NM(∞) to large uncertainties in N(∞). Implications for Persistence Assessment. NER formation is central to the fate of pesticides applied to agricultural soils. The analysis of biodegradation data for a large number of 66 compounds in Speyer 2.2 soil indicated that variation in terms of total amounts of NER formed and the importance of the two NER forming pools, i.e., directly from the parent compound or from the metabolite pool, is remarkable. Aided by inverse modeling, however, it was possible to derive several general findings relevant for persistence assessment. First of all, the fairly constant ratio between kPN and kPM of around 0.1 to 1 across all compounds implies that DegT50 values are rarely more than twice as large as DT50 values, which helps to put into perspective the relevance of NER formation for persistence evaluation. The observed correlation between kPN and kPM further indicates that both NER and metabolite formation are in many cases a consequence of the parent compound’s tendency to undergo biologically mediated transformation reactions. Finally, it was found that formation of NER directly from the parent compound and through the metabolite pool are both important, and that the relative importance of the two pools needs to be established case-bycase. It is worth noting in this context that the four-pool model is a simplification of reality and does not consider the mechanistic details of the transformation processes. The kinetic analysis presented therefore cannot resolve the question of whether formation of NER from the parent pool proceeds indeed through direct reaction of the parent compound with SOM or through a highly reactive and hence transient intermediate. Our analysis further demonstrated that the amount of NER formed and the dominant pools of NER formation are both highly time-dependent. Arbitrary study durations in regulatory soil biodegradation studies might therefore be misleading and will most likely underestimate the contribution of metabolites to total NER formed. We therefore propose an MCMC inverse modeling approach, which allows extrapolating time series information to any specific time point, while taking into account associated uncertainties partly stemming from parameter combinations of low identifiability. In the present study, extrapolation to N(∞), for instance, allowed uncovering a correlation between the presence of aromatic amine moieties and high NER formation potential, which could not be recognized from 100 d end point data. We furthermore stress the general advantage of MCMC parameter estimation in combination with assessing the a

■

ASSOCIATED CONTENT

S Supporting Information *

The SI contains methodological details for MCMC sampling and identifiability analysis, results from the identifiability analysis, and a table of all compounds studied including information on estimated parameters, Koc values, and soil microbial activity. This information is available free of charge via the Internet at http://pubs.acs.org/ .

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]; Phone: +41 58 765 50 85; Fax: +41 58 765 53 11. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Bayer Crop Sciences for providing access to unpublished data sets, and Dr. Gerhard Görlitz for critically reviewing an earlier version of this manuscript.

■

REFERENCES

(1) Gevao, B.; Semple, K. T.; Jones, K. C. Bound pesticide residues in soils: a review. Environ. Pollut. 2000, 108 (1), 3−14. (2) Burauel, P.; Führ, F. Formation and long-term fate of nonextractable residues in outdoor lysimeter studies. Environ. Pollut. 2000, 108 (1), 45−52. (3) Führ, F.; Ophoff, H.; Burauel, P.; Wanner, U.; Haider, K., Modification of the definition of bound residues. In Pesticide Bound Residues in Soil; Führ, F., Ophoff, H., Eds.; Wiley-VCH: Weinheim, Germany, 1998; pp 175−176. (4) Barraclough, D.; Kearney, T.; Croxford, A. Bound residues: environmental solution or future problem? Environ. Pollut. 2005, 133 (1), 85−90. (5) Craven, A.; Hoy, S. Pesticide persistence and bound residues in soil - regulatory significance. Environ. Pollut. 2005, 133 (1), 5−9. (6) Bollag, J.; Myers, C.; Minard, R. Biological and chemical interactions of pesticides with soil organic matter. Sci. Total Environ. 1992, 123−124, 205−217.

9836

dx.doi.org/10.1021/es300505r | Environ. Sci. Technol. 2012, 46, 9830−9837

Environmental Science & Technology

Policy Analysis

(7) Nowak, K. M.; Miltner, A.; Gehre, M.; Schäffer, A.; Kästner, M. Formation and fate of bound residues from microbial biomass during 2,4-D degradation in soil. Environ. Sci. Technol. 2011, 45 (3), 999− 1006. (8) Barriuso, E.; Benoit, P.; Dubus, I. G. Formation of pesticide nonextractable (bound) residues in soil: Magnitude, controlling factors and reversibility. Environ. Sci. Technol. 2008, 42 (6), 1845−1854. (9) Dec, J; Bollag, J. Determination of covalent and non-covalent binding interactions between xenobiotic chemicals and soil. Soil Sci. 1997, 162, 858−874. (10) EU Regulation (EC) No. 1107/2009 of the European Parliament and of the Council of 21 October 2009 concerning the placing of plant protection products on the market and repealing Council Directives 79/117/EEC and 91/414/EEC. Off. J. Eur. Union 2009, L 309 (52), 1−50. (11) Matthies, M.; Witt, J.; Klasmeier, J. Determination of soil biodegradation half-lives from simulation testing under aerobic laboratory conditions: A kinetic model approach. Environ. Pollut. 2008, 156 (1), 99−105. (12) Jacques, J.; Perry, T. Parameter estimation: local identifiability of parameters. Am. J. Physiol.-Endocr. M. 1990, 258 (4), E727−E736. (13) Pavan, A.; Thomaseth, K.; Valerio, A. Modeling population kinetics of free fatty acids in isolated rat hepatocytes using Markov Chain Monte Carlo. Ann. Biomed. Eng. 2003, 31 (7), 854−866. (14) Nong, A.; Tan, Y.; Krolski, M.; Wang, J.; Lunchick, C.; Conolly, R.; Clewell, H. Bayesian calibration of a physiologically based pharmacokinetic/ pharmacodynamic model of carbaryl cholinesterase inhibition. J. Toxicol. Environ. Health 2008, 71 (20), 1363−1381. (15) Rubach, M.; Ashauer, R.; Maund, S.; Baird, D.; van der Brink, P. Toxicokinetic variation in 15 freshwater arthropod species exposed to the inseticide chlorpyrifos. Environ. Toxicol. Chem. 2010, 29 (10), 2225−2234. (16) Görlitz, L.; Gao, Z.; Schmitt, W. Statistical analysis of chemical transformation kinetics using Markov-Chain Monte Carlo Methods. Environ. Sci. Technol. 2011, 45 (10), 4429−4437. (17) European Food Safety Authority. Rapporteur member state assessment reports submitted for the EU peer review of active substances used in plant protection products, 2010. http://dar.efsa. europa.eu/dar-web/provision. (18) OECD. OECD 307 Guideline for Testing of Chemicals - Aerobic and Anaerobic Transformation in Soil; 2002. (19) R Development Core Team. R: A Language and Environment for Statistical Computing; The R Foundation for Statistical Computing: Vienna, Austria. http://cran.r-project.org//. (20) Gelman, A.; Carlin, J.; Stern, H.; Rubin, D. Bayesian Data Analysis; Chapman & Hall, CRC Press: London, UK, 2003. (21) Metropolis, N.; Rosenbluth, A.; Rosenbluth, M.; Teller, A.; Teller, E. Equation of state calculations by fast computing machines. J. Chem. Phys. 1953, 21 (6), 1087−1092. (22) Hastings, W. Monte Carlo sampling methods using Markov chains and their applications. Biometrika 1970, 57, 97−109. (23) Geyer, C. MCMC: Functions for Markov Chain Monte Carlo (MCMC), 2010. http://www.stat.umn.edu/geyer/mcmc/. (24) Plummer, M.; Best, N.; Cowles, K.; Vines, K. CODA: Convergence Diagnosis and Output Analysis for MCMC. R News 2006, 6, 7−1. (25) Scott, D. Multivariate Density Estimation. Theory, Practice, and Visualization. John Wiley & Sons: New York, 1992. (26) Spiess, A.; Neumeyer, N. An evaluation of R2 as an inadequate measure for nonlinear models in pharmacological and biochemical research: a Monte Carlo approach. BMC Pharmacol. 2010, 10, 6. (27) Berk, R. A. Statistical Learning from a Regression Perspective; Springer: New York, 2008. (28) Lehmann, E. L.; Casella, G. Theory of Point Estimation, 2nd ed.; Springer: New York, 1998. (29) Pillonetto, G.; Sparacino, G.; Cobelli, C. Numerical nonidentifiability regions of the minimal model of glucose kinetics: superiority of Bayesian estimation. Math. Biosci. 2003, 184, 53−67.

(30) Shariati, M.; Korsgaard, I.; Sorensen, D. Identifiability of parameters and behaviour of MCMC chains: a case study using the reaction norm model. J. Anim. Breed. Genet. 2009, 126, 92−102. (31) Brun, R.; Kühni, M.; Siegrist, H.; Gujer, W.; Reichert, P. Practical identifiability of ASM2d parameters - systematic selection and tuning of parameter subsets. Water Res. 2002, 36, 4113−4127. (32) Brun, R.; Reichert, P.; Künsch, H. Practical identifiability analysis of large environmental simulation models. Water Resour. Res. 2001, 37, 1015−1030. (33) Brockmann, D.; Rosenwinkel, K.; Morgenroth, E. Practical identifiability of biokinetic parameters of a model describing two-step nitrification in biofilms. Biotechnol. Bioeng. 2008, 101, 497−514. (34) Turanyi, T. Sensitivity analysis of complex kinetic systems. Tools and applications. J. Math. Chem. 1990, 5, 203−248. (35) Boesten, J.; Aden, K.; Beigel, C.; Beulke, S.; Dust, M.; Dyson, J.; Fomsgaard, I.; Jones, R.; Karlsson, S.; van der Linden, A.; Richter, O.; Magrans, J.; Soulas, G. Guidance document on estimating persistence and degradation kinetics from environmental fate studies on pesticides in EU registration version 2.0. In Forum for the Co-Ordination of Pesticide Fate Models and Their Use (FOCUS), 2006. (36) Site, A. Factors affecting sorption of organic compounds in natural sorbent/water systems and sorption coefficients for selected pollutants. A review. J. Phys. Chem. Ref. Data 2001, 30 (1), 187−439. (37) Loos, M. Modeling of Pesticide Biodegradation in Soil; ETH Department of Environmental Sciences, ETH Zü rich: Zü rich, Switzerland, 2010. http://e-collection.library.ethz.ch/. (38) Burauel, P.; Führ, F. Formation and long-term fate of nonextractable residues in outdoor lysimeter studies. Environ. Pollut. 2000, 108, 45−52. (39) Bending, G. D.; Lincoln, S. D.; Sorensen, S. R.; Morgan, J. A. W.; Aamand, J.; Walker, A. In-field spatial variability in the degradation of the phenylurea herbicide isoproturon is the result of interactions between degradative Sphingomonas spp. and soil pH. Appl. Environ. Microb. 2003, 69, 827−834. (40) Bending, G. D.; Lincoln, S. D.; Edmondson, R. N. Spatial variation in the degradation rate of the pesticides isoproturon, azoxystrobin and diflufenican in soil and its relationship with chemical and microbial properties. Environ. Pollut. 2006, 139, 279−287. (41) Boethling, R.; Fenner, K.; Howard, P.; Klecka, G.; Madsen, T.; Snape, J. R.; Whelan, M. J. Environmental persistence of organic pollutants: Guidance for development and review of POP risk profiles. Integr. Environ. Assess. Manage. 2009, 5, 539−556. (42) EU Regulation (EC) no 1907/2006 of the European Parliament and of the Council of 18 December 2006 concerning the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH), establishing a European Chemicals Agency, amending Directive 1999/ 45/EC and repealing Council Regulation (EEC) no 793/93 and Commission Regulation (EC) no 1488/94 as well as Council Directive 76/769/EEC and Commission Directives 91/155/EEC, 93/67/EEC, 93/105/EC and 2000/21/EC. Off. J. Eur. Union 2006, L 396, 1−849. (43) Scheunert, I.; Schröder, P. Formation, characterization and release of non-extractable residues of [14C]-labeled organic xenobiotics in soils. Environ. Sci. Poll. Res. 1998, 5 (4), 238−244. (44) Bollag, J.; Myersa, C.; Minard, R. Biological and chemical interactions of pesticides with soil organic matter. Sci. Total Environ. 1992, 123−124, 205−217. (45) Bollag, J.; Liu, S.; Minard, R. Cross-coupling of phenolic humus constituents and 2,4-dichlorophenol. Soil Sci. Soc. Am. J. 1980, 44, 52− 56. (46) Senesi, N. Binding mechanisms of pesticides to soil humic substances. Sci. Total Environ. 1992, 123/124, 63−76.

9837

dx.doi.org/10.1021/es300505r | Environ. Sci. Technol. 2012, 46, 9830−9837