HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Larsbo, M. Articles by Stamm, C. Search for Related Content PubMed Articles by Larsbo, M. Articles by Stamm, C. Agricola Articles by Larsbo, M. Articles by Stamm, C. Related Collections Pharmaceuticals Uncertainty Analysis Solute Transport Models Vadose Zone Processes and Chemical Transport

TECHNICAL REPORTS

Ecological Risk Assessment

Simulating Sulfadimidine Transport in Surface Runoff and Soil at the Microplot and Field Scale

^a Swiss Federal Inst. for Aquatic Science and Technology (Eawag), Ueberlandstrasse 133, CH-8600 Duebendorf, Switzerland
^b current address: Dep. of Soil Sciences, Swedish Univ. of Agricultural Sciences (SLU), P.O. Box 7014, 750 07 Uppsala, Sweden
^c Institute of Biogeochemistry and Pollutant Dynamics, Swiss Federal Inst. of Technology (ETH), Universitätsstrasse 16, CH-8092 Zürich, Switzerland
^d RCC, Ltd., Zelgliweg 1, 4452 Itigen, Switzerland

^* Corresponding author (Mats.Larsbo{at}mv.slu.se).

Received for publication August 15, 2007.

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION Materials and Methods Results and Discussion Conclusions REFERENCES

 
To prevent residues of veterinary medicinal products (VMPs) from contaminating surface waters and ground water, an environmental impact assessment is required before a new product is allowed on the market. Physically based simulation models are advocated for the calculation of predicted environmental concentrations at higher tiers of the assessment process. However, the validation status of potentially useful models is poor for VMP transport. The objective of this study was to evaluate the dual-permeability model MACRO for simulation of transport of sulfonamide antibiotics in surface runoff and soil. Special focus was on effects of solute application in liquid manure, which may alter the hydraulic properties at the soil surface. To this end we used data from a microplot runoff experiment and a field experiment, both conducted on the same clay loam soil prone to preferential flow. Results showed that the model could accurately simulate concentrations of sulfadimidine and the nonreactive tracer bromide in runoff and in soil from the microplot experiments. The use of posterior parameter distributions from calibrations using the microplot data resulted in poor simulations for the field data of total sulfadimidine losses. The poor results may be due to surface runoff being instantly transferred off the field in the model, whereas in reality re-infiltration may occur. The effects of the manure application were reflected in smaller total and micropore hydraulic conductivities compared with the application in aqueous solution. These effects could easily be accounted for in regulatory modeling.

Abbreviations: EIA, environmental impact assessment • PEC, predicted environmental concentration • PNEC, predicted no-effect concentration • SDM, sulfadimidine • VMP, veterinary medicinal product

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION Materials and Methods Results and Discussion Conclusions REFERENCES

 
Large quantities of veterinary medicinal products (VMPs) are used in livestock production. After administration, these substances may be excreted as the parent compound and/or as metabolites. When manure containing residues of VMPs is used as fertilizer on agricultural land, there is a risk that these substances might leach to ground water or reach surface waters through field drains or surface runoff. There they pose a potential risk, primarily to water-living organisms (Boxall et al., 2003).

To reduce the risk of soil and water contamination, the European Union has since 1997 requested environmental impact assessment (EIA) of new VMPs before they are allowed on the market. Guidance on how to carry out this assessment has been developed by the VICH (International Cooperation on Harmonisation of Technical Requirements for Registration of Veterinary Medicinal Products) committee, a trilateral program between the EU, Japan, and the USA (VICH, 2000 and 2006). In Europe, these guidelines have recently been amended by a technical guideline on how to carry out the different evaluation steps (EMEA, 2007). These documents include guidance on EIAs for VMPs used in aquaculture, pasture animals, and intensively reared animals. In this study we focus on intensively reared animals. The assessment should be performed in a phased manner in which the first phase involves the estimation of concentrations to which different environmental compartments might be exposed (calculations of predicted environmental concentrations [PECs]). Relevant for our case is the PEC in soil (PEC_soil). For the estimation of PEC_soil, the Uniform Approach suggested by Spaepen et al. (1997) is recommended. A first estimate of PEC_soil is calculated assuming a worst case scenario where the total yearly amount administered to livestock is applied to agricultural fields. If needed, PECs can then be refined (e.g., by inclusion of more realistic agricultural practices, metabolism in animals, and degradation in manure and soil). If the PEC exceeds trigger values defined for relevant environmental compartments (100 µg kg^–1 for soil), the second phase of the EIA has to be performed. In this phase, predicted no-effect concentrations (PNECs) are compared with the PECs in, for example, ground water (PEC_gw) and surface water (PEC_sw). These PECs should, in a first step, be calculated using a simple, steady-state compartment approach. If a PEC exceeds the corresponding PNEC, further refinement of the PEC is required, or risk mitigation measures have to be linked to the authorization. At this phase of PEC_sw and PEC_gw refinements, more advanced modeling tools may be used.

The models used in the regulatory process have recently been evaluated. In a test using field data on three VMPs, the Uniform Approach resulted in up to two orders of magnitude higher environmental concentrations compared with measured data (Blackwell et al., 2005). Montforts (2006) tested different screening models used to calculate PECs for soil, surface water, and ground water against available field data. He concluded that the models generally performed poorly and questioned the usefulness of such simple models in the registration process. Advanced physically based solute transport models have been included since the mid-1990s in the pesticide registration process within the FOCUS (forum for the coordination of pesticide fate models and their use) framework. The use of these models is supposed to lead to a more science-based regulatory process, thereby resulting in a better protection of the environment while avoiding unnecessary restrictions on harmless (or less harmful) substances (FOCUS, 2001). Even though it is too early to say to what degree these goals have been achieved, a similar approach for VMPs seems attractive because VMPs and their possible routes of contamination are in many respects similar to pesticides (Kay et al., 2004; Montforts, 2006). The use of more advanced models for refined calculations of PECs of VMPs in ground water and surface water is also recommended in the new technical guideline developed by the European Medicines Agency (EMEA, 2007). These guidelines specifically recommend the use of the FOCUS models. However, other studies have shown that differences between pesticides and VMPs, e.g., the manure application matrix (Burkhardt et al., 2005) and the fact that manure in many cases is incorporated into the top soil after application (Kay et al., 2004) are important for the transport processes. This implies that pesticide-leaching models should not be uncritically used for simulations of VMPs. To build up confidence in the models, they need to be evaluated against field data for VMPs.

Sulfadimidine belongs to the sulfonamide group of antibiotics, which are mainly used for prophylactic or therapeutic purposes in pig production. Sulfonamides account for a large percentage of the total use of veterinary antibiotics in Europe, and there is a risk of contamination of ground water and surface waters due to their moderate sorption strength in soil compared with other groups of VMPs (Thiele-Bruhn, 2003). Sulfonamides have been detected in tile drainage water (Campagnolo et al., 2002), in surface waters (Campagnolo et al., 2002; Christian et al., 2003; Perret et al., 2006), and in ground water (Hamscher et al., 2005; Perret et al., 2006). Most of these findings were from areas with intensive livestock production.

The objective of this work was to evaluate the MACRO 5.1 model (Larsbo et al., 2005), one of the models included in the FOCUS suite of models, using data from one short-time irrigation microplot study of sulfadimidine (SDM) transport (Burkhardt et al., 2005; Burkhardt and Stamm, 2007) and one 3-mo field study (Stoob et al., 2007). MACRO was chosen because it includes preferential flow and transport through the soil, artificial field drainage, surface runoff of water and solutes, and a full canopy water and solute balance, processes that were considered important based on the experimental data. Many model parameters (e.g., those governing the degree of preferential flow) are impossible to derive from direct measurements. It is therefore important to design simple experimental flow and transport experiments that can provide information on these parameters. Therefore, we tested to what degree the data from the microplot study could be used to constrain parameters that could not be derived from direct measurements.

	Materials and Methods

TOP NOTES ABSTRACT INTRODUCTION Materials and Methods Results and Discussion Conclusions REFERENCES

 
Description of the Experiments
The data used in this study are from two experiments conducted in the same subcatchment of the lake Greifensee catchment, southeast of Zürich, Switzerland. The subcatchment, which has a size of about 60 ha, is dominated by grasslands used as pasture, which is also the land use of the investigated fields. The catchment is drained by a small brook. The soil in the studied fields was classified as loamy Eutric Cambisol (FAO). Soil properties are summarized in Table 1 .

The Field Study
Sulfonamides were applied to two fields, each approximately 0.4 ha. Manure applications (616 g SDM ha^–1 for application 1 and 614 g SDM ha^–1 for application 2, both in approximately 3 L m^–2 liquid manure) were done using a tractor-mounted band spreader on 24 Mar. and 8 May 2003. Hourly values of total SDM loads in the brook were calculated from high time resolution measurements of brook discharge and concentrations in the brook at the outlet of the catchment. Because SDM was applied only to the studied fields, the loads in the brook could be used as validation data for the simulated losses from the fields. However, it was not possible to discriminate between losses through surface runoff and drain discharge. Pore water concentrations in the top 5 cm soil measured nine times during the experimental period were also used as validation data. The fields are underlain by tile drains, but the depth, spatial arrangement, and efficiency of the drainage system are not known. During the experimental period, parts of the fields were water logged. The field study is described in detail by Stoob et al. (2007).

Model Description
MACRO is a one-dimensional, process-oriented, dual-permeability model for water flow and reactive solute transport in soil. A detailed description of the model can be found in Larsbo and Jarvis (2003). Here we focus on the processes that are most important for this study.

It is assumed that all precipitation is intercepted by the crop canopy. Intercepted water is stored in the canopy until the storage capacity, S_can (mm), is exceeded. Any water exceeding S_can enters the soil surface as drip throughfall. The canopy pool is depleted by canopy evaporation. The rate of change of solutes stored in the canopy is given by:

[1]

where Q_can (g m^–2) is the solute stored in the canopy, t (d) is time, w_i (mm d^–1) is the canopy interception rate, c_p (g m^–3) is the concentration in the intercepted water, w_d (mm d^–1) is the drip throughfall rate, f_e (mm^–1) is the foliar washoff coefficient introduced to account for solubility and sorption on the canopy, and µ_can (d^–1) is the canopy degradation rate, accounting for volatilization losses and degradation. For nonreactive solutes, f_e is set to S_can^–1, and µ_can is set to zero.

If the infiltration capacity of the micropores is exceeded, infiltrating water is routed into the macropores. Any precipitation exceeding the total infiltration capacity of the soil becomes surface runoff. Solutes in the precipitation reaching the soil surface are assumed to mix completely with any solutes contained in a thin surface layer (the "mixing zone") of a specified thickness, z_mix (mm). A new sorption equilibrium is calculated before redistribution from the mixing zone into macropores and surface runoff.

The soil pore space is divided into a micropore domain and a macropore domain, characterized by different flow rates and solute concentrations. The division between pore domains is defined by a pressure potential, _b (cm), a water content, _b (-), and a hydraulic conductivity, K_b (cm h^–1). For the micropore domain, water flow is governed by Richards' equation, the water retention function is described by the van Genuchten (1980) model, and the unsaturated hydraulic conductivity is described by the Mualem (1976) model. Solute transport is governed by the convection–dispersion equation. In the macropores, water flow, assumed to be dominated by gravity, is described by a modified kinematic wave equation (Germann, 1985). Because water flow rates are large, solute transport in the macropores is assumed to be solely convective.

Lateral water flow from macropores to micropores is described as a first-order process:

[2]

where d (mm) is the diffusion path length related to aggregate size, _w (-) is a scaling factor (Gerke and van Genuchten, 1993), _m (-) is the micropore volumetric water content, and D_w (m² s^–1) is the effective water diffusivity. The mass transfer term for solutes accounts for convection and diffusion:

[3]

where D_e (m² s^–1) is an effective diffusion coefficient, c_mic (kg m^–3) is the solute concentration in the liquid phase in micropores, c_mac (kg m^–3) is the solute concentration in the liquid phase in macropores, and c' (kg m^–3) indicates the solute concentration in macropores or micropores, depending on the direction of water flow, S_w.

Even though the model allows for more advanced sorption models, we assumed linear equilibrium sorption. First-order solute degradation was calculated in liquid and solid phases for both pore domains. A first-order approach can also be used to model formation of non-extractable residues under the assumption that the formation rate is independent of the concentration in the non-extractable pool.

A primary drainage system located in the soil profile and a secondary drainage system surrounding the fields were considered. In both cases, flux rates from saturated soil layers above the drainage depths were calculated using potential seepage theory for layered soils (Leeds-Harrison et al., 1986).

Calibration Method
The MACRO 5.1 model was decoupled from the windows shell program, which is used for model setup in the original version. This was done to enable the use of MACRO with the UNCSIM software for statistical inference, sensitivity, identifiability, and uncertainty analysis (Reichert, 2005). MACRO and UNCSIM were loosely coupled via simple text files.

The MACRO model can be recast into a nonlinear regression model:

[4]

where Y is the measured data of group i at time j, f() is the model estimations, x is the set of model inputs (meteorological data), is the set of unknown model parameter values, and is the residual error term.

Bayesian techniques have been gaining in popularity in hydrological modeling (Beven and Binley, 1992; Campbell et al., 1999) because increasing computer power has made it possible to use these computationally demanding methods on personal computers. The desired outcome of Bayesian inference is a posterior probability distribution, P(|Y), describing what is known about the model parameters, , given measured data, Y, and prior information. Any knowledge about parameters before considering the measured data should be included in the prior distribution, P(). The updating of the prior is described by Bayes' theorem:

[5]

where P(Y) is a constant of proportionality, and P(Y|) is the likelihood function. If the residuals (Eq. [4]) are normally distributed and independent with homoscedastic SDs, the likelihood function can be written in multiplicative form:

[6]

where _i is the SD of the residuals.

Markov Chain Monte Carlo methods are numerical techniques for sampling the posterior parameter distributions. In UNCSIM, the simplest of these methods, the Metropolis algorithm is implemented. For a detailed description on Bayesian inference and Markov Chain Monte Carlo methods, see Gelman et al. (1995). To give proper weight to all data regardless of magnitude, we used the power transformation suggested by Box and Cox (1964). To represent limited prior knowledge, we used uniform distributions for all parameters included in the calibration.

We used the Heidelberger and Welch diagnostic as implemented in the CODA software (Best and Vines, 1995) to test for convergence of the Markov chains. Each calibration was run until all parameters passed the test, and any nonstationary warm-up phases were removed before further analyses. Plots of the Markov chains were visually inspected to ascertain convergence.

The posterior distributions can be summarized by statistical measures (e.g., percentiles). In this study, we also used the reduction coefficient, defined as 1 minus the posterior SD divided by the prior SD. A reduction coefficient close to 1 indicates a high degree of parameter conditioning, and a value of 0 means that the calibration had no effect on the parameter distribution. It is also of interest to look at the simulation resulting in the largest posterior probability. This simulation is hereafter referred to as the "best simulation."

Modeling Strategy
We simultaneously calibrated hydrological parameters and substance properties using all data from the microplot experiment. Posterior parameter distributions from the microplot simulations were then used for the field data modeling exercise in predictive simulations using the stationary Markov Chains. The predictive simulations were run to study the value of the information gained from the microplot scale simulations when changing to the field scale. The bottom boundary conditions were largely unknown. Test simulations revealed that microplot results were insensitive to the bottom boundary conditions. For the field-scale simulations, we tested two scenarios representing the two possible major routes for SDM losses from the fields: (i) a drainage scenario (a 1-m-deep soil profile with tile drains at 0.8 m depth and 10 m spacing) and (ii) a runoff scenario (a 0.5-m-deep soil profile without tile drains). In both scenarios, the ground water table was allowed to rise in the profile. These two scenarios represent two extreme conditions: a fully functioning drainage system and impermeable subsoil at 0.5 m depth creating favorable conditions for saturation excess surface runoff. Neither of these scenarios is likely to be representative for the whole fields. These scenarios were chosen to get a better understanding of the major routes for SDM losses from the fields.

Parameterization
Ideally, all parameters that are to some extent uncertain and sensitive for the experimental conditions at hand should be included in the calibration. For the sake of clarity and to limit the computational demand, we limited the number of parameters. We included only the most sensitive parameters that were not well constrained by direct measurements. Because a high degree of preferential flow was reported for the experiments (Burkhardt and Stamm, 2007), we focused on parameters governing flow in macropores and exchange between pore domains. Processes at the soil surface are of major importance for the formation of runoff. Therefore, parameters regulating the water and solute balance on the canopy and in the soil mixing zone were included in the calibration. In addition to the organic carbon content of the soil, the type and content of clay minerals and the soil pH are important for the sorption of sulfonamides (Thiele-Bruhn et al., 2004; Kahle and Stamm, 2007a; Kahle and Stamm, 2007b). However, Kahle and Stamm (2007b) showed that sorption of the sulfonamide sulfathiazole to inorganic sorbents was more than an order of magnitude lower than sorption to organic sorbents. Therefore, SDM sorption was described using the organic carbon–water partition coefficient, K_oc (L kg^–1). The organic carbon–water partition coefficient and the first-order degradation rate coefficient in the topsoil, µ_top (d^–1), were included in the calibration. The degradation rate was assumed to decrease with depth according to the recommendations by FOCUS (2001). Guidance on initial parameter uncertainty was gathered from previous experience with the model (Dubus and Brown, 2002; Larsbo and Jarvis, 2006), literature (Thiele-Bruhn, 2003; Blackwell et al., 2005; Kreuzig et al., 2005), and measurements. Parameters included in the calibration and their initial uncertainty intervals are listed in Table 2 .

Soil water retention was measured on replicate undisturbed soil cores (volume 100 mL) sampled at three depths by the Institute of Agrosphere at the Forschungszentrum Jülich. These data were used to estimate the parameters in the van Genuchten (1980) water retention equation using the RETC program (van Genuchten et al., 1991). These parameters (Table 3 ) were not included in the calibration because the SDs for the parameter values were considered small. Crop parameters for grass were taken from FOCUS (2001), except for the crop height, which was set to the approximate average field value of 7 cm. Consistent with the canopy interception of rain (see Model Description), all irrigation water was assumed to be intercepted before reaching the soil surface as drip throughfall. We treated SDM as a nonreactive solute concerning canopy wash-off (Eq. [1]) because no information on the behavior of SDM on plant canopies was available. Additional parameters not included in the calibration are listed in Table 4 .

View this table: [in this window] [in a new window]   Table 3. Hydraulic parameters and SDs (in parentheses). The parameters in the van Genuchten equation were fitted to measurements using the RETC-program (van Genuchten et al., 1991).

The initial water contents for the microplot simulations were set to the average value of the measurements. Because the microplots were protected with plastic covers during the experimental period, the potential evapotranspiration was set to zero. For the field study, initial soil water contents were calculated assuming drainage equilibrium with a water table at the bottom of the soil profile. We used a 3-wk warm-up period before the first solute application to limit the effects of the initial soil moisture condition. Daily averages of air temperature, global radiation, air humidity, and wind speed measured at the site at 10-min resolution were used as driving data, and hourly averages of rainfall were used. For the microplot and the field study, initial solute concentrations were set to zero based on measurements and knowledge of earlier field management (Stoob et al., 2007).

	Results and Discussion

TOP NOTES ABSTRACT INTRODUCTION Materials and Methods Results and Discussion Conclusions REFERENCES

 
Microplot Runoff Simulations
Measurements and simulation results for runoff volumes from the control and the manure microplots are shown in Fig. 1 . The model could not simulate the measured approximately linear increase in runoff volumes with time for the control treatment (Fig. 1a). One possible explanation for this behavior in the measured values is the time lag between runoff generation at the top end of the microplots and runoff collection at the lower end of the plots. With time, a larger proportion of the plots might have contributed to the collected runoff volumes. In the model, where this effect is not accounted for, a steady-state runoff rate is reached when the topsoil approaches saturation. For the manured plots, runoff rates increased during the whole measurement period, but, unlike in the control plots, the increase got smaller with time (Fig. 1b).

View larger version (16K): [in this window] [in a new window]   Fig. 1. Simulated results compared with measured runoff volumes at 15-min intervals. (a) Control plots. (b) Manured plots. The best simulation is the simulation with the highest posterior probability.

View larger version (16K): [in this window] [in a new window]   Fig. 2. Simulated results compared with measured bromide concentrations in runoff. (a) Control plots. (b) Manured plots. The best simulation is the simulation with the highest posterior probability.

View larger version (19K): [in this window] [in a new window]   Fig. 3. Simulated results compared with measured bromide concentrations in soil pore water. (a) Control plots. (b) Manured plots. The best simulation is the simulation with the highest posterior probability.

View larger version (15K): [in this window] [in a new window]   Fig. 4. Simulated results compared with measured sulfadimidine concentrations on manured plots. (a) Runoff. (b) Soil pore water. The best simulation is the simulation with the highest posterior probability.

Posterior Parameter Distributions and Parameter Identifiability
Statistics describing the posterior parameter distributions from the microplot scale simulations are summarized in Table 2. Not surprisingly, K_tot was well constrained by the measured data. The simulated infiltration rate at steady state and hence the runoff rate (during constant excessive irrigation) is solely controlled by K_tot at the soil surface (assuming the water table does not reach the soil surface). The micropore hydraulic conductivity was also fairly well constrained. It determines how deep into the soil matrix solutes are transported and has an effect on when runoff starts and thereby on the concentrations in runoff. Apart from K_tot, the water flow in the macropores is determined by the macroporosity, _mac (-), and the kinematic exponent, n* (-), which describes the continuity and tortuosity of the macropores. The latter was not conditioned by the data (Table 2). The exchange of water and solutes between pore domains (Eq. [2] and [3]) is determined by the diffusion path length, d, which reflects the geometry of the macropore system. The diffusion path length has previously been proven highly sensitive for solute leaching from soils prone to preferential flow (Dubus and Brown, 2002). In this study, d_top was constrained to values between 7.47 and 80.2 mm, whereas d_sub was between 94.2 and 186 mm. This relatively large posterior uncertainty is probably due to the lack of leaching data. The mixing depth has been shown to be relatively insensitive for pesticide leaching (Dubus and Brown, 2002) and for resident and effluent tracer concentrations in lysimeter experiments (Larsbo and Jarvis, 2006). We expected z_mix to be more sensitive to concentrations in runoff because there is a direct connection between z_mix and the rate of the depletion of solutes from the mixing zone. The mixing depth was highly conditioned toward the lower limit of the initial uncertainty.

The mean value of K_oc was in the lower end of the interval of reported literature values ranging from 80 to 208 L kg^–1 (Thiele-Bruhn, 2003). This may at least partly be explained by the short contact time between SDM and the soil material because sorption strongly increases with time (Kahle and Stamm, 2007a). The mean value for µ_top was much larger than measured values for other sulfonamides in soil (Blackwell et al., 2005; Kreuzig et al., 2005). The time for sorption equilibration has proven to exert a strong influence on the extraction recovery (Hamscher et al., 2005; Wehrhan et al., 2007; Kahle and Stamm, 2007a). The reason for this is considered to be formation of non-extractable residues (Kreuzig et al., 2005). In our case, µ_top includes possible formation of non-extractable residues. The posterior distribution was highly skewed toward large values. This indicates that an even larger value of µ_top might have led to a better fit to the measured data. However, based on the available literature data, such fast dissipation seems unrealistic.

The mean value of the posterior distribution for S_can was 1.59 mm (Table 2), which means that approximately 53% of the applied solutes were intercepted by the canopy at application. Interception of pesticides by the crop canopy may lead to increased losses through runoff and higher concentrations in the soil after irrigation because the time for the compound to react with the soil before the onset of runoff is decreased (Reddy and Locke, 1996). This effect was very clear in the SDM simulations. Runoff concentrations were up to 50% lower in test simulations without the crop cover. The canopy degradation rate was well conditioned to values much smaller than µ_top. To our knowledge, there are no studies on the dissipation of SDM in crop canopies available, but Thiele-Bruhn and Peters (2007) reported a photodegradation rate of 0.1 d^–1 for SDM in pig slurry. This indicates that the posterior distribution for µ_can may be reasonable.

The larger measured runoff from manured plots (Fig. 1) was attributed to surface sealing by manure solids decreasing the infiltration capacity of the soil (Burkhardt et al., 2005). Of the three parameters describing the effects of the manure application, _tot,man was highly conditioned by the data (Table 2). The mean value of the posterior distribution for _tot,man reflects the lower steady-state infiltration rate for the manured plots. The measured bromide concentrations in runoff (Fig. 2) were higher for the manured plots. This was probably an effect of the decreased total and matrix infiltration rate due to surface sealing by the manure application. For the control plots, a larger fraction of the solutes was transported down into the profile below the mixing depth before the onset of runoff. There, the solutes were less available when runoff started. Similar reasoning has been proposed by Potter et al. (2003). The above hypothesized effect on the micropore hydraulic conductivity was reflected in the posterior distribution for _b,man. In contrast to the posterior distribution for z_mix, z_mix,man was highly conditioned toward the higher limit of the initial uncertainty. It is not clear why the manure application increased z_mix. The different posterior distributions for z_mix and z_mix,man may have compensated for other processes affected by the manure that were not handled by the model rather than reflecting differences in solute mixing.

The measured data indicated that the manure application matrix strongly influenced runoff volumes and bromide concentrations under the microplot experimental conditions. Because these conditions were designed to induce surface runoff, the effects under natural field conditions may be very different. In the microplot experiments, the accumulated runoff was 2.6 and 14.7% of the applied irrigation for the control and manured plots, respectively. It seems likely that the effects of the application matrix on solute losses through runoff should be, in a relative sense, smaller when a larger proportion of the rain is becoming runoff. A decrease in the total hydraulic conductivity at the soil surface from 19.2 to 15.6 mm h^–1, which was indicated by the calibration, would probably have limited effect on the runoff losses under natural field conditions because rain intensities of that magnitude are rare. However, the changes in K_b may have large effects on the generation of macropore flow and thereby on leaching to ground water or to field drains.

Field-Scale Simulations
The results of the two scenarios from the field-scale simulations are presented in Fig. 5 , which shows the uncertainty intervals of the simulations due to parameter uncertainty. The two scenarios, which represent different lower boundary conditions, yielded very different SDM load predictions. For the runoff scenario, the timing of the major events was well captured; however, total losses were largely overestimated (Fig. 5a and 5c). The accumulated simulated losses were 9.9% of the applied amount, which should be compared with the accumulated measured losses of 0.33%. In contrast, the total losses of SDM were largely underestimated for the drainage scenario (Fig. 5a and 5b). Here, the simulated accumulated losses were only 3.1 x 10^–5%. This scenario also revealed a large temporal mismatch between the measured and simulated losses.

View larger version (20K): [in this window] [in a new window]   Fig. 5. Measured and simulated losses from the fields. (a) Measured losses and precipitation. (b) Uncertainty interval for simulated sulfadimidine losses for the drainage scenario. (c) Uncertainty interval for simulated sulfadimidine losses for the runoff scenario. Note the different scales and units on the y axes. The uncertainty intervals are defined by the 2.5th and the 97.5th percentiles.

Due to these lateral heterogeneities, a more empirically based approach, such as Soil Conservation Service runoff curve numbers applied (e.g., in the Pesticide Root Zone Model [PRZM]; Carsel et al., 1998), might have been more appropriate for the field-scale simulations. However, the PRZM runs on a daily time step, and it would not have been applicable to the simulation of the microplot experiments. Furthermore, the PRZM does not account for the combined effects of macropores and tile drains that may have been relevant in this case.

The information gained in the microplot experiments regarding the soil properties had limited value for the field-scale simulations. By applying different lower boundary conditions, the same soil properties resulted in a vastly different temporal SDM load pattern. The microplot experiments did not help to constrain these boundary conditions. Because the microplots were protected with plastic covers, these experiments could not provide information on the upper boundary conditions for the fields. However, test simulations assuming a bare soil surface showed that the large uncertainties in the field evapotranspiration did not significantly affect simulated loads.

Pore water concentrations in the top 5 cm soil are presented in Fig. 6 for the runoff scenario only because differences between scenarios were minor. The data were fairly well reproduced by the model even though the peak concentration for the second application was overestimated. The fairly good fit to measured data indicates that the degradation rate in the topsoil from the microplot calibrations was reasonable also for the field data. However, the persistence of SDM was underestimated, which can be seen in the poor fit to the measurement just before the second application and the final measurement. This is probably because the model did not account for the observed bi-phasic nature of the dissipation of SDM (Stoob et al., 2007). The simulated increase in concentrations before the third measurement is due to the wash-off of solutes stored on the canopy during the first rainfall after application.

View larger version (15K): [in this window] [in a new window]   Fig. 6. Measured values and simulated uncertainty intervals for pore water concentrations. The uncertainty intervals are defined by the 2.5th and the 97.5th percentiles.

	Conclusions

TOP NOTES ABSTRACT INTRODUCTION Materials and Methods Results and Discussion Conclusions REFERENCES

The major problems we encountered were related to applying parameters obtained at the microplot scale to the field scale. This issue is not specific to VMPs but represents a general problem of solute transport modeling. There are at least two aspects to this issue. First, the spatial information that is needed to correctly parameterize the models is often lacking. In this case, the lower boundary conditions had a major influence on the load dynamics. Second, one-dimensional models like MACRO have inherent problems in accounting for two- or three-dimensional processes such as surface runoff, retention, and re-infiltration. Such limitations have to be overcome to improve predictions for VMPs and for pesticides and other agrochemicals.

	ACKNOWLEDGMENTS

 
The model evaluation was funded by grants from the European Union (ERAPharm, project no. 511135). The experimental work took place within the framework of the Swiss National Research Program NRP 49 on Antibiotic Resistance. We thank the Institute of Agrosphere at Forschungszentrum Jülich, Germany, for providing the soil water retention data and Dr. Jing Yang at the Swiss Federal Inst. for Aquatic Science and Technology (Eawag) for advice on the Bayesian methods.

	NOTES

TOP NOTES ABSTRACT INTRODUCTION Materials and Methods Results and Discussion Conclusions REFERENCES

 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION Materials and Methods Results and Discussion Conclusions REFERENCES

 

• Best, N.G.C., and M.K. Vines. 1995. CODA manual version 0.30. MRC Biostatistics Unit, Cambridge, UK.
• Beven, K., and A. Binley. 1992. The future of distributed models: Model calibration and uncertainty prediction. Hydrol. Processes 6 :279–298.[CrossRef]
• Blackwell, P.A., A.B.A. Boxall, P. Kay, and H. Noble. 2005. Evaluation of a lower tier exposure assessment model for veterinary medicines. J. Agric. Food Chem. 53 :2192–2201.[CrossRef][Web of Science][Medline]
• Box, G.E.P., and D.R. Cox. 1964. An analysis of transformations. J. R. Stat. Soc. Ser. B 26 :211–252.
• Boxall, A.B.A., D.W. Kolpin, B. Halling-Sorensen, and J. Tolls. 2003. Are veterinary medicines causing environmental risks? Environ. Sci. Technol. 37 :286A–294A.[Medline]
• Burkhardt, M., C. Stamm, C. Waul, H. Singer, and S. Müller. 2005. Surface runoff and transport of sulfonamide antibiotics and tracers on manured grassland. J. Environ. Qual. 34 :1363–1371.[Abstract/Free Full Text]
• Burkhardt, M., and C. Stamm. 2007. Depth distribution of sulfonamide antibiotics in pore water of an undisturbed loamy grassland soil. J. Environ. Qual. 36 :588–596.[Abstract/Free Full Text]
• Campagnolo, E.R., K.R. Johnson, A. Karpati, C.S. Rubin, D.W. Kolpin, M.T. Meyer, J.E. Esteban, R.W. Currier, K. Smith, K.M. Thu, and M. McGeehin. 2002. Antimicrobial residues in animal waste and water resources proximal to large-scale swine and poultry feeding operations. Sci. Total Environ. 299 :89–95.[CrossRef][Medline]
• Campbell, E.P., D.R. Fox, and B.C. Bates. 1999. A Bayesian approach to parameter estimation and pooling in nonlinear flood event models. Water Resour. Res. 35 :211–220.[CrossRef]
• Carsel, R.F., J.C. Imhoff, P.R. Hummel, J.M. Cheplick, and A.S. Donigian, Jr. 1998. PRZM-3, a model for predicting pesticide and nitrogen fate in the crop root and unsaturated soil zones: User's manual for release 3.0. Natl. Exposure Res. Lab., Office of Res. and Development, USEPA, Athens, GA.
• Christian, T., R.J. Schneider, H.A. Farber, D. Skutlarek, M.T. Meyer, and H.E. Goldbach. 2003. Determination of antibiotic residues in manure, soil, and surface waters. Acta Hydrochim. Hydrobiol. 31 :36–44.[CrossRef]
• Dubus, I.G., and C.D. Brown. 2002. Sensitivity and first-step uncertainty analyses for the preferential flow model MACRO. J. Environ. Qual. 31 :227–240.[Abstract/Free Full Text]
• Edwards, D.R., and T.C. Daniel. 1993. Abstractions and runoff from fescue plots receiving poultry litter and swine manure. Trans. ASAE 36 :405–411.[Web of Science]
• EMEA. 2007. Guideline on Environmental Impact Assessment for Veterinary Medicinal Products in Support of the VICH Guidelines GL6 and GL38. Committee for Medicinal Products for Veterinary Use; European Medicines Agency, London.
• FOCUS. 2001. FOCUS surface water scenarios in the EU evaluation process under 91/414/EEC. Report of the FOCUS working group on surface water scenarios. European Commission Document Reference SANCO/4802/2001-rev.2.
• Gangbazo, G., A.R. Pesant, D. Cluis, and D. Couillard. 1992. Laboratory analysis of water runoff and infiltration following application of pig manure. Can. Agric. Eng. 34 :17–25.
• Gelman, A.C., J.B. Stern, and H.S. Rubin. 1995. Bayesian data analysis. Chapman & Hall, London.
• Gerke, H.H., and M.Th. van Genuchten. 1993. Evaluation of a first-order water transfer term for variably saturated dual-porosity flow models. Water Resour. Res. 29 :1225–1238.[CrossRef]
• Germann, P.F. 1985. Kinematic wave approach to infiltration and drainage into and from soil macropores. Trans. ASAE 28 :745–749.[Web of Science]
• Hamscher, G., H.T. Pawelzick, H. Hoper, and H. Nau. 2005. Different behavior of tetracyclines and sulfonamides in sandy soils after repeated fertilization with liquid manure. Environ. Toxicol. Chem. 24 :861–868.[CrossRef][Web of Science][Medline]
• Kahle, M., and C. Stamm. 2007a. Sorption of the veterinary antimicrobial sulfathiazole to organic materials of different origin. Environ. Sci. Technol. 41 :132–138.[Medline]
• Kahle, M., and C. Stamm. 2007b. Time and pH-dependent sorption of the veterinary antimicrobial sulfathiazole to clay minerals and ferrihydrite. Chemosphere 68 :1224–1231.[Medline]
• Kay, P., P.A. Blackwell, and A.B.A. Boxall. 2004. Fate of veterinary antibiotics in a macroporous tile drained clay soil. Environ. Toxicol. Chem. 23 :1136–1144.[CrossRef][Web of Science][Medline]
• Kreuzig, R., S. Holtge, J. Brunotte, N. Berenzen, J. Wogram, and R. Schulz. 2005. Test-plot studies on runoff of sulfonamides from manured soils after sprinkler irrigation. Environ. Toxicol. Chem. 24 :777–781.[CrossRef][Web of Science][Medline]
• Larsbo, M., and N. Jarvis. 2003. MACRO 5.0. A model of water flow and solute transport in macroporous soil: Technical description. Emergo 2003:6. Swedish Univ. of Agric. Sci., Dep. of Soil Sci., Uppsala, Sweden.
• Larsbo, M., and N. Jarvis. 2006. Information content of measurements from tracer microlysimeter experiments designed for parameter identification in dual-permeability models. J. Hydrol. 325 :273–287.[CrossRef]
• Larsbo, M., S. Roulier, F. Stenemo, R. Kasteel, and N. Jarvis. 2005. An improved dual-permeability model of water flow and solute transport in the vadose zone. Vadose Zone J. 4 :398–406.[Abstract/Free Full Text]
• Leeds-Harrison, P.B.S., C.J.P. Jarvis, and N. Youngs. 1986. The influence of soil macroporosity on water retention, transmission and drainage in a clay soil. Soil Use Manage. 2 :47–50.[CrossRef]
• Montforts, M. 2006. Validation of the exposure assessment for veterinary medicinal products. Sci. Total Environ. 358 :121–136.[CrossRef][Medline]
• Mualem, Y. 1976. New model for predicting hydraulic conductivity of unsaturated porous-media. Water Resour. Res. 12 :513–522.[CrossRef]
• Perret, D., A. Gentili, S. Marchese, A. Greco, and R. Curini. 2006. Sulphonamide residues in Italian surface and drinking waters: A small scale reconnaissance. Chromatographia 63 :225–232.[CrossRef][Web of Science]
• Potter, T.L., C.C. Truman, D.D. Bosch, and C.W. Bednarz. 2003. Cotton defoliant runoff as a function of active ingredient and tillage. J. Environ. Qual. 32 :2180–2188.[Abstract/Free Full Text]
• Reddy, K.N., and M.A. Locke. 1996. Imazaquin spray retention, foliar washoff, and runoff losses under simulated rainfall. Pestic. Sci. 48 :179–187.[CrossRef]
• Reichert, P. 2005. UNCSIM: A computer programme for statistical inference and sensitivity, identifiability, and uncertainty analysis. p. 51–55. In J.M.F. Teixeira and A.E. Carvalho-Brito (ed.) Proceedings of the 2005 European Simulation and Modeling Conf. (ESM 2005), 24–26 Oct. 2005, EUROSIS-ETI, Porto, Portugal.
• Spaepen, K.R.I., L.J.J. van Leemput, P.G. Wislocki, and C. Verschueren. 1997. A uniform procedure to estimate the predicted environmental concentration of the residues of veterinary medicines in soil. Environ. Toxicol. Chem. 16 :1977–1982.[CrossRef][Web of Science]
• Stamm, C., R. Sermet, J. Leuenberger, H. Wunderli, H. Wydler, H. Fluhler, and M. Gehre. 2002. Multiple tracing of fast solute transport in a drained grassland soil. Geoderma 109 :245–268.[CrossRef][Web of Science]
• Stoob, K., H.P. Singer, S.R. Mueller, R.P. Schwarzenbach, and C.H. Stamm. 2007. Fate and transport of veterinary sulfonamides applied to grassland under typical agricultural practice in a small catchment. Environ. Sci. Technol. 41 :7349–7355.[Medline]
• Thiele-Bruhn, S. 2003. Pharmaceutical antibiotic compounds in soils: A review. J. Plant Nutr. Soil Sci. 166 :145–167.[CrossRef]
• Thiele-Bruhn, S., T. Seibicke, H.R. Schulten, and P. Leinweber. 2004. Sorption of sulfonamide pharmaceutical antibiotics on whole soils and particle-size fractions. J. Environ. Qual. 33 :1331–1342.[Abstract/Free Full Text]
• Thiele-Bruhn, S., and D. Peters. 2007. Photodegradation of pharmaceutical antibiotics on slurry and soil surfaces. Landbauforsch. Voelkenrode 57 :13–23.
• van Genuchten, M.Th. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44 :892–898.[Abstract/Free Full Text]
• van Genuchten, M.Th., F.J. Leij, and S.R. Yates. 1991. The RETC code for quantifying the hydraulic functions of unsaturated soils. EPA/600/2-91/065. USEPA, Office of Research and Development, Washington, DC.
• VICH. 2000. Environmental impact assessment of veterinary medicinal products—Phase I. London: CVMP/VICH; 2000. CVMP/VICH/592/98-final. Final Guidance, London, UK.
• VICH. 2006. Environmental impact assessment for veterinary medicinal products—Phase II VICH GL 38. Final Guidance, London, UK.
• Wehrhan, A., R. Kasteel, J. Simunek, J. Groeneweg, and H. Vereecken. 2007. Transport of sulfadiazine in soil columns: Experiments and modeling approaches. J. Contam. Hydrol. 89 :107–135.[CrossRef][Web of Science][Medline]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Larsbo, M. Articles by Stamm, C. Search for Related Content PubMed Articles by Larsbo, M. Articles by Stamm, C. Agricola Articles by Larsbo, M. Articles by Stamm, C. Related Collections Pharmaceuticals Uncertainty Analysis Solute Transport Models Vadose Zone Processes and Chemical Transport