VZJ sign up for etocs
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 8 October 2007
Published in Vadose Zone J 6:793-804 (2007)
DOI: 10.2136/vzj2007.0067
© 2007 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (6)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Boesten, J. J. T. I.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Boesten, J. J. T. I.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Boesten, J. J. T. I.
Related Collections
Right arrow Agricultural Pesticides
Right arrow Vadose Zone Risk Assessment
Right arrow Ground Water Quality

ORIGINAL RESEARCH

Simulation of Pesticide Leaching in the Field and in Zero-Tension Lysimeters

Jos J. T. I. Boesten*

Alterra, Wageningen Univ. and Research Centre, P.O. Box 47, 6700 AA Wageningen, The Netherlands
* Corresponding author (jos.boesten{at}wur.nl).

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.


Received 3 April 2007.



   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 The PEARL Model
 Simulation Procedures
 Results and Discussion
 Discussion and Recommendations
 REFERENCES
 
Zero-tension lysimeters play an important role in groundwater risk assessments for pesticides in the European Union. In these assessments, measured lysimeter leachate concentrations are usually used directly for decision making. When doing so, one assumes (i) that the lysimeter bottom boundary condition itself did not lead to underestimating field leaching concentrations, (ii) that the number of application years and the duration of the lysimeter study were adequate to measure the maximum concentration in time, and (iii) that the pesticide-lysimeter system considered was sufficiently vulnerable with respect to leaching. These assumptions were tested using simulations with a Darcian water flow model combined with a chromatographic pesticide leaching model. The scenario consisted of a layered light-textured soil cropped with cereals and of multiyear weather data. The groundwater level in the field usually fluctuated between 0.7 and 2.5 m depth. The lysimeter bottom boundary condition resulted in pesticide leaching concentrations lower than those calculated for the field system. Simulations showed that a lysimeter study of 2 yr was too short to measure the maximum leaching concentration for pesticides with organic-matter/water distribution coefficient values exceeding 40 L kg–1. The probability that a lysimeter study results in a leaching concentration below 0.1 µg L–1 by a coincidental favorable combination of pesticide–soil parameters was assessed by Monte Carlo simulations. This probability exceeded 20% for pesticides that would give leaching concentrations of 1 µg L–1 under field conditions. Therefore, it is advisable that for each lysimeter study, modeling be used to assess the likelihood that the leaching concentration would be below 0.1 µg L–1, considering all relevant systematic and random factors.

Abbreviations: EU, European Union • LAI, leaf area index • OECD, Organization for Economic Cooperation and Development


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
The PEARL Model
 Simulation Procedures
 Results and Discussion
 Discussion and Recommendations
 REFERENCES
 
The risk assessment of pesticide leaching aims to assess whether normal agricultural use of the pesticide (i.e., application to hundred thousands of hectares over periods of decades with a certain frequency) will lead to unacceptable contamination of groundwater. Zero-tension lysimeters (lysimeters with free outflow of water) are considered to be a useful tool in this risk assessment because they almost exactly reproduce the environmental conditions that occur in the corresponding field soil (Bergström, 1990; Führ et al., 1998a). Since about 1990, such lysimeter studies have played an important role in the regulatory risk assessments for pesticide leaching to groundwater within the European Union (EU). The first guideline for performing lysimeter studies was established in Germany (BBA, 1990), followed a decade later by a guideline for lysimeter studies from the Organization for Economic Cooperation and Development (OECD, 2000). These guidelines do not include recommendations for how to use the results of a lysimeter study in the risk assessment procedure. In the German risk assessment procedure, annual average flux concentrations are derived from the lysimeter measurements, and the maximum of the annual averages is used directly for decision making (Nolting and Schinkel, 1998). In the EU risk assessment, the same approach has been followed (EFSA, 2005). Using measured lysimeter leachate concentrations directly for decision making is only justifiable if the lysimeter study represents the realistic worst-case conditions that need to be protected. Because lysimeters are experimental model systems, however, the direct use of measured lysimeter leachate concentrations for decision making can be questioned.

A first point of debate is the effect of the bottom boundary condition of the lysimeter on pesticide leaching concentrations (Jene et al., 1998; Flury et al., 1999; Abdou and Flury, 2004; Kasteel et al., 2007). Water will flow out of a zero-tension lysimeter only when the bottom layer is water saturated and capillary rise of water into such a lysimeter is impossible. This differs from the situation at the same depth in the field, where water can flow downward under unsaturated conditions and where water may also flow upward due to capillary rise from the groundwater table. This problem has been assessed in both modeling and experimental studies. Flury et al. (1999) calculated the effect of the bottom boundary condition for a conservative solute under steady-state unsaturated flow. For the lysimeter, they found a slower mean solute transport velocity and a larger spreading than for the field. Abdou and Flury (2004) included two-dimensional heterogeneous water flow in their analysis and also found a slower mean transport velocity for the lysimeter than for the field. They showed that the difference was larger for a vertical soil structure than for a horizontal or isotropic soil structure. Jene (1998) measured distinct differences between three lysimeters and a suction-based field system over 2 yr. Water percolation and amount of bromide leached were higher in the lysimeters (by 60 and 58%, respectively), but the average flux concentration of the pesticide benazolin was about two times higher in the field. Jene (1998) concluded that these differences were not system related but rather caused by different meteorological conditions for the field and lysimeter systems, which were separated by 20 km. Kasteel et al. (2007) performed a 3-yr experiment and also found considerably more (46%) cumulative water percolation in six lysimeters than in suction plates installed in the field but almost equal recoveries of bromide. From two-dimensional simulations for an isotropic heterogeneous porous medium, they concluded that the amount of water collected by the suction plates was sensitive to the hydraulic conductivity of the plate and not representative for the field. So while Kasteel et al. (2007) intended to test whether lysimeters behave similar to the field (which is considered to be the "true" system), they concluded that the suction plate measurement method needs to be improved to avoid the problem that the resistance of the plates dictates the amount of water collected in the field. After correcting for the artifacts resulting from the suction plates via simulations, Kasteel et al. (2007) found a slightly faster pore water velocity in the field than in the lysimeters, which is consistent with the results found by Flury et al. (1999) and Abdou and Flury (2004). Flury et al. (1999), Abdou and Flury (2004), and Kasteel et al. (2007) considered systems with deep groundwater tables, and they all attributed the faster pore water velocity in the field to the higher volume fractions of water at the bottom of the lysimeter caused by its seepage face.

Flury et al. (1999), Abdou and Flury (2004), and Kasteel et al. (2007) only considered solutes that are not transformed. In the assessment of pesticide leaching to groundwater in the EU, the EU drinking water limit of 0.1 µg L–1 (Smeets and Amavis, 1981) plays an important role. Leaching at a concentration level of 0.1 µg L–1 in water layers of 100 to 1000 mm corresponds to leaching losses of only 0.01 to 0.1% of a normal pesticide dose of 1 kg ha–1, which implies that more than 99% of the dose disappears via loss processes such as transformation and plant uptake. Thus, a longer residence time of the pesticide in the lysimeter may have a considerable effect on these loss processes. Effects calculated for pesticides may be larger than those calculated for solutes that are not transformed. It seems worthwhile, therefore, to assess the effect of the lower boundary condition of the lysimeter for pesticides in a modeling study.

A second point of debate is whether the number of application years and the duration of the lysimeter studies as prescribed by the OECD (2000) are adequate to ensure that the maximum concentration that would leach under normal agricultural practice is measured. FOCUS (2000) established groundwater scenarios for the EU assessment of pesticide leaching using applications over periods of decades to simulate possible cumulative effects of multiple applications. These scenarios include a 6-yr "warming-up period" in which the pesticide is applied but is ignored in the evaluation of the model output. However, the OECD (2000) prescribed a different design of lysimeter studies: pesticides that are applied in agricultural practice every two or more years are applied to the lysimeter in a single year; for such pesticides, the study stops 2 yr after this application. The OECD (2000) did not provide any justification for its application regimes and study duration. In addition, the two monographs on lysimeter studies (Führ and Hance, 1992; Führ et al., 1998b) do not provide experimental support for the application regime and the duration of the study as proposed by the OECD (2000). The OECD approach is probably based on pragmatic considerations: a 5-yr lysimeter study, for example, does not fit into the time scale of a pesticide registration procedure. In the context of pesticide registration, however, pragmatism is only defensible if it leads to conservative results. Otherwise, the pragmatism itself may lead to underestimating the leaching risk, which cannot be defended.

A third point of debate is the effect of the uncertainty in those pesticide–soil properties of the lysimeter system that are unknown. The OECD (2000) prescribes measurement of only the following soil properties (as a function of depth): texture, pH, cation exchange capacity, organic carbon, and water holding capacity. However, Boesten (1991, 2004) and Dubus et al. (2003) showed that pesticide leaching concentrations calculated with chromatographic simulation models are sensitive to half-life, sorption properties, and dispersion length in soil (considering only a limited number of soil profiles). Diels et al. (1996) found similar results using an analytical stochastic-convective transport model with steady-state water flow and including some 100 Belgian sandy soil profiles. However, the OECD (2000) does not prescribe measurements of half-life, sorption properties, and dispersion length in the framework of the lysimeter studies. As a consequence, these parameters may not be available; in such cases, the lysimeter study is a "black box" with respect to them. The FOCUS groundwater scenarios (i.e., the first tier of the leaching assessment at the EU level) use fixed values of these parameters, which is an important aspect of the vulnerability concept behind these scenarios (FOCUS, 2000). A higher-tier study such as a lysimeter should not be based on less-conservative parameter values than the lower-tier (that triggered the lysimeter study) without appropriate justification. (Pesticide risk assessment is usually based on tiered procedures, i.e., stepped procedures going from simple lower tiers to more sophisticated higher tiers as necessary.) However, if these parameters are unknown for the lysimeter–soil system considered, the leachate concentrations measured in the lysimeter study may remain below 0.1 µg L–1 by a coincidental favorable combination of the above-mentioned parameters. The probability that a risk assessment generates favorable results by coincidence should be kept to a minimum (especially if a lower tier has indicated a potential risk). Yon (1992), Bergström et al. (1994), and Jene (1998) found coefficients of variation of pesticide leaching losses of 50 to 90% between replicates in lysimeter studies with sandy soils. This indicates a large spatial variability between lysimeters taken from the same field. Jene (1998), for example, found flux concentrations of benazolin averaged over 2 yr of 0.4, 1.3 and 3.8 µg L–1 for three lysimeters taken from the same sandy field. One may expect an even larger variability between lysimeters taken from different sandy fields. Kördel et al. (1991) reported large differences between results of lysimeter studies for two pesticides with similar transformation and sorption properties. All this information indicates that the likelihood of a favorable result of the lysimeter study caused by a coincidental favorable combination of soil–pesticide parameters cannot be ignored a priori.

Based on the above considerations, the present study addresses the following questions: (i) To what extent will the bottom boundary condition of the soil column in the lysimeter lead to pesticide leaching concentrations that are lower than leaching concentrations obtained for the same soil and weather conditions in a field soil? (ii) Are the number of application years and the duration of the lysimeter study prescribed by the OECD (2000) adequate for obtaining the maximum in time of the leaching concentration occurring in the long-term in agricultural practice? (iii) What is the likelihood that the leachate concentration measured in the lysimeter study remains below 0.1 µg L–1 by a coincidental favorable combination of soil-pesticide parameters not measured in the lysimeter experiment? These questions were explored by simulations with the PEARL model (based on chromatographic flow) as described by Leistra et al. (2001) using a cropped soil and multiyear weather data. The main argument for using a chromatographic leaching model is that lysimeter studies are usually performed with sandy soils (BBA, 1990). In such soils, flow through macropores is less important because the soil matrix is very conductive. Heterogeneous flow and transport occur also in sandy soils, but this can often be described satisfactorily by the dispersion process that is part of chromatographic flow models (Boesten and Gottesbüren, 2000). Furthermore, chromatographic leaching models are used for tier-1 risk leaching assessments at EU level (FOCUS, 2000), so the need for lysimeter studies is also triggered on the basis of such models.


   The PEARL Model
TOP
 ABSTRACT
 INTRODUCTION
 The PEARL Model
Simulation Procedures
 Results and Discussion
 Discussion and Recommendations
 REFERENCES
 
The PEARL model has been described in detail elsewhere (Leistra et al., 2001), so the model concepts are only briefly described here. PEARL is a one-dimensional model for pesticide leaching through soil. It uses the SWAP model as a submodel for simulating soil water flow and soil temperatures. SWAP is based on Darcy's law for water flow in soil and can simulate a wide range of hydrological boundary conditions, including fluctuating groundwater levels. SWAP has been described in detail by Van Dam et al. (1997). Briefly, it uses Richards' equation, and it uses the van Genuchten functions to describe the volume fraction of liquid and the hydraulic conductivity as a function of the soil water pressure head. The potential evapotranspiration rate, ETP, is based on daily estimates of Penman evaporation, which is multiplied by a crop factor for the growing season and a bare soil factor for the remainder of the year. The ETP is distributed over potential transpiration rate TP and potential soil evaporation rate EP on the basis of the leaf area index (see Van Dam et al., 1997, for details). The actual rate of water uptake from soil, S, is calculated as

Formula 1[1]
in which Fred is a factor (–) accounting for reduction of water uptake if the soil becomes too dry, and SP is the potential rate of water uptake (T–1), which is calculated as the quotient of TP divided by root depth. The actual soil evaporation rate, E (L T–1), is derived from the relationship

Formula 2[2]
in which ß is a parameter (L1/2), {Sigma}EP is the cumulative EP during a drying cycle (L), and {Sigma}E is the cumulative E during a drying cycle (L). Equation [2] applies only to the stage at which the actual soil evaporation rate is less than the potential rate. The bottom boundary condition for the field simulations is a volume flux of water at the bottom of the soil system, qBOT (L T–1), that is always downward and that decreases exponentially with increasing depth of the groundwater table:

Formula 3[3]
where a (L T–1) and b (L–1) are parameters, and where G is the depth of the groundwater level (L) below the soil surface. Equation [3] implies that qBOT equals its maximum value a if the groundwater level is at the soil surface (G = 0). This equation results in a fluctuating groundwater table within the soil profile (provided that the soil profile considered is deep enough). The option for simulating a lysimeter is free drainage if the pressure head at the bottom of the lysimeter is zero and a zero flux if this pressure head is negative. SWAP also simulates soil temperature using the combination of Fourier's law and the conservation equation for heat in soil using daily average air temperature as the upper boundary condition.

PEARL describes the partitioning of the pesticide over the phases in soil by

Formula 4[4]
where c* is the total concentration in soil system (M L–3), {varepsilon} is the volume fraction of gas (–), {theta} is the volume fraction of liquid phase (M3 M–3), cG and cL are concentrations (M L–3) in gas and liquid phase, {rho} is the dry soil bulk density (M L–3), and XEQ and XNE are contents sorbed (M M–1) at equilibrium and nonequilibrium sorption sites. The concentration in the gas phase is directly proportional to the concentration in the liquid phase using the concept of the Henry coefficient. Sorption at the equilibrium sites is described via a Freundlich isotherm with KF,EQ as the Freundlich coefficient for the equilibrium sorption sites and N as the Freundlich exponent. The coefficient KF,EQ is calculated as the product of the mass fraction of organic matter and the organic-matter/water distribution coefficient, KOM. The operational definition for equilibrium sorption sites is that they reach equilibrium with 24 h shaking. Sorption at the nonequilibrium sorption sites is described by a pseudo first-order rate equation:

Formula 5[5]
where t is time (T), KF,NE is the Freundlich coefficient for the nonequilibrium sorption sites (L3 m–1), kD is the desorption rate coefficient of the nonequilibrium sorption sites (T–1), and cL,R is the reference concentration in liquid phase (set at 1 mg L–1). The Freundlich coefficient for the nonequilibrium sites is assumed to be directly proportional to the coefficient of the equilibrium sites:

Formula 6[6]
where fNE is the proportionality factor (–). The rate of pesticide transformation in soil, Rt (M L–3 T–1), is described by

Formula 7[7]
where fT, f{theta} and fZ are factors describing the effect on the transformation rate of temperature, moisture content, and depth in soil, respectively, and where k is the first-order transformation rate coefficient (T–1) at reference conditions (i.e., top soil at 20°C and pF = 2). This first-order coefficient equals ln(2)/DegT50, where DegT50 is the half-life at these reference conditions (using the DegT50 acronym as recommended by FOCUS, 2006). Equation [7] implies that the pesticide at the nonequilibrium sorption sites is not subject to transformation. The temperature factor fT is described by the Arrhenius equation. The Walker relationship is used for describing the effect of soil moisture content:

Formula 8[8]
where {theta}FC is the volume fraction of water at field capacity (i.e., pF = 2) and B is an exponent (–). The depth factor fZ is described numerically. The rate of plant uptake of pesticide in soil, RU (M L–3 T–1), is described using the "transpiration stream concentration factor" approach:

Formula 9[9]
where fU is this factor (–). The mass flux of pesticide in soil is calculated using the convection–dispersion equation for the liquid phase plus a diffusion term for the gas phase.


   Simulation Procedures
TOP
 ABSTRACT
 INTRODUCTION
 The PEARL Model
 Simulation Procedures
Results and Discussion
 Discussion and Recommendations
 REFERENCES
 
Risk Assessment Case
In the simulations, the following risk assessment case is considered. Leaching of a pesticide has to be assessed for EU pesticide registration. The pesticide considered is mainly used in the climatic zone covered by the "Hamburg" groundwater scenario as defined by FOCUS (2000). Let us assume that simulations with the PEARL model for this scenario have indicated that groundwater concentrations at 1 m depth exceed the EU limit of 0.1 µg L–1. Let us further assume that the company that wants to register the pesticide has conducted a lysimeter study that is considered to represent realistic worst-case conditions (hoping to demonstrate that these simulations overestimated the risk). Thus all simulations were done for soil and weather conditions taken from the FOCUS Hamburg scenario. This case was selected because (i) a large fraction of the regulatory lysimeter studies has been conducted under German weather conditions and (ii) the soil organic carbon profile of this scenario (see Table 1) fulfills the criterion of a maximum organic carbon content of 1.5% recommended by the German guideline (BBA, 1990).


View this table:
[in this window]
[in a new window]

  		TABLE 1. Properties of the soil profile used in the simulations (taken from the Hamburg scenario of FOCUS, 2000). The symbols {theta}s, {theta}r, {alpha}, n, Ks, and {lambda} are parameters of the van Genuchten relationships; see Leistra et al. (2001) for the definition of these symbols.

 
General Characteristics of the Simulations
All simulations were made with PEARL v3.3.3 (available at http://viso.ei.jrc.it/focus). Weather, crop, and soil data of the Hamburg scenario were used as defined by FOCUS (2000). The weather data consist of a series of 66 yr based on 20-yr daily weather data (1975–1994) taken from the MARS database (Terres, 1998). This series of 20-yr weather data was repeated three times, and the last 6 of the 20 yr were added at the start of the series to give the 66 yr (FOCUS, 2000). These initial 6 yr act as a so-called warming-up period for application of the pesticides. Average annual rainfall in the 20 yr was 786 mm, and annual rainfall ranged between 540 and 999 mm. The crop was spring cereals (as parameterized by FOCUS, 2000) because this is a common crop in lysimeter studies. The crop emerged on 1 April, and crop harvest was on 20 August. The crop reached its maximum leaf area index (LAI) of 3.9 and its maximum rooting depth of 90 cm on 5 June. Both LAI and rooting depth increased linearly from zero at emergence to the maximum value at 5 June and remained constant at this maximum level until harvest. The factor for reduction of water uptake by plants Fred (Eq. [1]) decreased from 1 at pressure heads ranging from –500 to –900 cm to 0 for heads below –16 000 cm, that is, wilting point (the value of –500 cm applies to high transpiration rates and the value of –900 cm applies to low transpiration rates; Van Dam et al., 1997). The multiplication factor for potential evaporation from bare soil was 1.0, and the crop factor for potential evapotranspiration during the growing season decreased linearly from 1.0 at emergence of the crop to 0.8 at 5 June, after which it remained constant at 0.8 until harvest. The top 60 cm of the soil was sandy loam, which was underlaid with sand (Table 1). The parameter ß of Eq. [2] was 0.79 cm0.5. The dispersion length was 5 cm as used by FOCUS (2000) for the PEARL model; this value is close to the median length of 6.2 cm found by Vanderborght and Vereecken (2007) in their review for the pooled distribution of both column and field measurements of dispersion length for depths between 81 and 200 cm (see their Fig. 11).

Simulations were done both for a field and a lysimeter system. The soil profile was 450 cm deep in the field simulations and 100 cm deep in the lysimeter simulations. For the field simulations, qBOT as defined by Eq. [3] was used as the bottom boundary at 450 cm depth with a = 10 mm d–1 and b = 1.4 m–1 as specified by FOCUS (2000). This implies a fluctuating groundwater level. The bottom boundary condition of the lysimeter does not require any further parameters. The thickness of the numerical compartments was 1.25 cm in the top 100 cm for both the lysimeter and field simulations and 2.5 cm below 100 cm depth for the field simulations.

Simulations were performed for a number of hypothetical pesticides with varying DegT50 and KOM values. All other pesticide properties (Table 2) were equal to FOCUS model substance D (FOCUS, 2000) except the long-term sorption parameters, which were set equal to the default values for the PEARL model as recommended by Leistra et al. (2001), that is, a nonequilibrium Freundlich sorption coefficient that is 0.5 times the equilibrium coefficient (fNE = 0.5) and using a desorption rate coefficient (kD) of 0.01 d–1. The long-term sorption process was included because ample evidence shows that this process occurs for pesticides in soil (Wauchope et al., 2002). Thus, one can expect that the simulations are more realistic when this process is included.


View this table:
[in this window]
[in a new window]

  		TABLE 2. Substance properties common for all pesticides as used in the simulations.

 
Effect of the Bottom Boundary Condition
The effect of the bottom boundary condition of the lysimeter on pesticide leaching concentrations was tested by performing simulations both for the field and the lysimeter systems. Pesticide was applied at a rate of 1 kg ha–1 on 31 March (i.e., 1 d before emergence of the spring cereals). Simulations were made both for application every 2 yr (using a 46-yr simulation period) and every 3 yr (using a 66-yr simulation period). Results of the first 6 yr (the warming-up period) were not evaluated, resulting in 40- and 60-yr evaluation periods. Leaching was characterized by flux concentrations averaged (i) over the whole 40 or 60 yr and (ii) over 2 or 3 yr for application every 2 or 3 yr, respectively. The 2- or 3-yr averages were characterized by the 80th percentile concentration from a series of 20 periods (using 20 biennial average values for the 40-yr period and 20 triennial averages for the 60-yr period; see FOCUS, 2000). This 80th percentile was calculated by taking the 17th value of the 20 ranked values. The evaluation depth for calculating these concentrations was 100 cm for both the field and the lysimeter systems.

Simulations were performed for 200 hypothetical pesticides with KOM and DegT50 values generated by Monte Carlo simulations. The KOM was uniformly distributed between 0 and 100 L kg–1. Preliminary calculations showed that for KOM values above 100 L kg–1, the initial 6-yr warming-up period was not long enough. The DegT50 was uniformly distributed between 1 and 100 d. This sampling procedure was followed to achieve a range of different KOM and DegT50 values for the test. The distributions were generated using subroutines described by Press et al. (1986). From the simulation results, ratios of lysimeter leaching concentrations divided by field leaching concentrations were calculated for the different types of concentration described above. Probability densities of these ratios were calculated with the procedure described by Harrell and Davis (1982).

Adequateness of Prescribed Number of Application Years and Duration of Lysimeter Studies
The principle of the test was to compare pesticide leaching after a single pesticide application for a typical 2-yr lysimeter experiment as prescribed by OECD (2000) with pesticide leaching for the corresponding field system with application every 2 or 3 yr (as used by FOCUS, 2000). The procedures of the simulations for the field system were the same as described above. The simulations for the 2-yr lysimeter experiment were for a 3-yr period starting on 31 Mar. 1912 and finishing 30 Mar. 1915 with application of 1 kg ha–1 on 31 Mar. 1913 (i.e., 1 d before emergence of the spring cereals). So the simulations lasted for 2 yr after this application. The first simulation year was included to ensure that the initial hydrological conditions had no effect on the results. This 3-yr period was selected out of the 20 yr of Hamburg weather because rainfall for the first 2 yr after application was 800 mm and 826 mm, respectively, which is consistent with the requirements in the BBA guideline for lysimeter studies (800 mm per year; BBA, 1990). Simulations were made for the same 200 hypothetical pesticides as described in the previous section.

The output of the simulation of the 2-yr lysimeter experiment was characterized by the maximum of the two annual average leachate concentrations, which is used directly for decision making in the German and EU risk assessments (Nolting and Schinkel, 1998; EFSA, 2005). For the field leaching simulations, the output was characterized by the 80th percentile of 20 biennial or triennial average leaching concentrations at 1 m depth.

Likelihood of Lysimeter Leachate Concentrations below 0.1 µg L–1 from Coincidental Favorable Combinations of Unknown Pesticide–Soil Parameters
In the test we assume that the "true" realistic worst case that needs to be protected in the leaching risk assessment consists of a FOCUS Hamburg scenario with (i) pesticide application every 2 or 3 yr and (ii) median or average values for the parameters DegT50, KOM, Freundlich exponent, and dispersion length. The principle of the test is to treat these parameters as stochastic variables with realistic probability density functions and to assess, via Monte Carlo simulations, the likelihood that a specific lysimeter experiment generates leaching concentrations below 0.1 µg L–1 for a number of model pesticides that showed concentrations exceeding 0.1 µg L–1 in simulations for this FOCUS Hamburg groundwater scenario.

The Monte Carlo simulations were performed for the lysimeter system using the same 3-yr weather period as before (31 Mar. 1912 to 30 Mar. 1915), again with application of 1 kg ha–1 on 31 Mar. 1913 (i.e., 1 d before emergence of the spring cereals crop). The simulations were performed with the following stochastic variables: (i) a DegT50 assuming a normal distribution with a coefficient of variation of 25%, (ii) a KOM assuming a normal distribution with a coefficient of variation of 25%, (iii) a Freundlich exponent N assuming a uniform distribution with a range of 0.8 to 1.0, and (iv) a dispersion length assuming a lognormal distribution with a median of 5 cm and a standard deviation of the logarithm of this length of 0.86 (–). The coefficient of variation of 25% for the DegT50 is supported by values ranging from 18 to 50% measured for 18 U.K. soils with metamitron, metazachlor, simazine, linuron, and propyzamide by Walker and Thompson (1977) and Allen and Walker (1987). Including long-term sorption kinetics implies that the DegT50 used applies only to the pesticide in the liquid phase and sorbed to the equilibrium sorption sites (as described by Eq. [7]). This definition differs from that used by Walker and Thompson (1977) and Allen and Walker (1987). An analysis of the differences between these types of DegT50 by Boesten and van der Linden (2001) showed that differences of 10 to 50% can be expected. It is unlikely that the DegT50 as used by PEARL is much less variable across different soils than the conventional DegT50 as calculated by Walker and Thompson (1977) and Allen and Walker (1987), so a coefficient of variation of 25% seems defensible for the DegT50 as used here. The coefficient of variation of 25% for the KOM is supported by measurements by Walker and Thompson (1977), Elabd and Jury (1986), and Allen and Walker (1987), who found a range from 20 to 48% for the pesticides metamitron, metazachlor, metribuzin, simazine, linuron, propyzamide, and napropamide. The width of the distribution of the Freundlich exponent N is supported by distributions of N as measured for 18 U.K. soils by Allen and Walker (1987) for metazachlor, metribuzin, and metamitron (showing that N varied from 0.62 to 1.0 for metamitron, from 0.68 to 1.2 for metazachlor. and from 0.71 to 1.0 for metribuzin). The median dispersion length of 5 cm was based on FOCUS (2000); the standard deviation of the logarithm of this length of 0.86 was based on the pooled distribution of both column and field measurements of dispersion length for depths between 81 and 200 cm from the review by Vanderborght and Vereecken (2007).

Simulations were made for two series of hypothetical pesticides with average values of DegT50 and KOM along two lines in the DegT50–KOM plane (DegT50 = 20 d and KOM = 35 L kg–1). Each simulation consisted of 500 runs for 500 draws of combinations of DegT50, KOM, N, and dispersion length. All parameter variations were assumed to be independent of each other. The distributions were generated using subroutines described by Press et al. (1986) using a different seed for each simulation. The dispersion length was set to a minimum value of 0.625 cm if lengths were drawn below 0.625 cm. This was done because the numerical solution procedure in PEARL prescribes that the compartment thickness should be smaller than or equal to two times the dispersion length (Leistra et al., 2001); thus, dispersion lengths close to zero would require excessive computing. The restriction implies an overestimation of leaching concentrations (Boesten, 2004) and so an underestimation of the percentage of runs with maximum concentrations below 0.1 µg L–1. However, the effect of this restriction is expected to be small because less than 1% of the runs were affected by this restriction (illustrated by the probability density function of the dispersion length shown in Fig. 1 ).


Figure 1
View larger version (11K):
[in this window]
[in a new window]

  		FIG. 1. The lognormal probability density function of the dispersion length as used in the Monte Carlo simulations considering coincidental favorable combinations of unknown pesticide–soil parameters. The median is 5 cm, and the standard deviation of the logarithm of the dispersion length is 0.86 (–).

 
The output for each run was the maximum of the two annual average concentrations leaching from the lysimeter as before. From the 500 maximum concentrations generated by the 500 runs, percentiles of the probability density function were calculated; from this density function, the percentage of runs with a maximum concentration below 0.1 µg L–1 was calculated. This percentage was considered the most relevant summary of the output because it gives the probability that the lysimeter generates a concentration below 0.1 µg L–1. For each DegT50–KOM combination, at least three simulation sets of 500 runs each were made; for each of these simulation sets, this percentage was calculated. Additionally, for each DegT50–KOM combination, the maximum of the two annual average leachate concentrations was calculated for the mean values of DegT50, KOM, and Freundlich exponent and for the median value of the dispersion length.


   Results and Discussion
TOP
 ABSTRACT
 INTRODUCTION
 The PEARL Model
 Simulation Procedures
 Results and Discussion
Discussion and Recommendations
 REFERENCES
 
Effect of the Bottom Boundary Condition
The simulated course of the groundwater level for a typical 5-yr period (Fig. 2A ) shows an annual pattern: groundwater levels are shallow in winter and deep in summer. Figure 2B shows that the percentage saturation close to the bottom of the lysimeter (at 97 cm depth) is calculated to be considerably higher than that at the same depth in the field (as was also found by Kasteel et al., 2007). This is the result of the lysimeter bottom boundary condition: water can only flow out of the lysimeter if the bottom layer of the lysimeter is water saturated. Therefore, the percentage saturation at 97 cm depth in the lysimeter is in winter periods close to 100%. The percentage saturation at 97 cm depth in the field (Fig. 2B) shows the same pattern as the groundwater level in Fig. 2A.


Figure 2
View larger version (24K):
[in this window]
[in a new window]

  		FIG. 2. (A) The groundwater level in the field soil and (B) the fraction of the pore volume filled with water ("Percentage saturation") at 97 cm depth in the field and lysimeter soils as a function of time as simulated with the SWAP model. The year numbers indicate 1 January of the corresponding year.

 
Average annual amounts of water percolated at 1 m depth for the 20-yr simulation period between 1907 and 1926 were calculated to check differences in the water balance. The 20-yr annual average percolation for the field was 311.9 mm; that for the lysimeter was 312.0 mm. This difference is so small because individual annual amounts of simulated evaporation and transpiration in both systems were almost identical (difference always <1%). However, individual annual amounts of percolated water sometimes differed considerably between the field and the lysimeter: the ratio of annual lysimeter percolation and annual field percolation ranged from 0.74 to 1.55. The small difference between the 20-yr annual average percolations is probably specific for the soil–crop–climate conditions in this study, and thus this result cannot be extrapolated to other conditions.

The comparison of simulated pesticide leaching concentrations was limited to pesticides that generated 80th percentile leaching concentrations above 0.1 µg L–1 for field conditions. This was done to exclude pesticides with a low leaching potential, that is, pesticides for which lysimeter studies would not be needed. As a result of this limitation, only output of 138 and 133 out of 200 pesticides was considered for application every 2 and 3 yr, respectively. Figure 3 shows that the lysimeter bottom boundary condition resulted in leaching concentrations that were 50 to 95% of those simulated for the field bottom boundary condition. The figure shows also that the lysimeter/field ratio was very similar for the different types of leaching concentrations considered; the median of this ratio ranged between 77 and 79% for the different types of leaching concentrations.


Figure 3
View larger version (26K):
[in this window]
[in a new window]

  		FIG. 3. Cumulative probability density of the ratio of leaching concentrations for the lysimeter divided by leaching concentrations for the field. Ratios are for 80th percentiles of a series of 20 2- or 3-yr average flux concentrations for application every 2 or 3 yr, respectively, and for average flux concentration over 40 and 60 yr for application every 2 or 3 yr, respectively. The probability density was based on 200 hypothetical pesticides with randomly generated organic-matter/water distribution coefficient, KOM, values between 0 and 100 L kg–1 and DegT50 values between 1 and 100 d, but results were included only if the 80th percentile concentration of the pesticide was above 0.1 µg L–1.

 
A possible explanation of the lower concentrations for the lysimeter bottom boundary condition is that the residence time of the pesticide in the lysimeter was longer than the residence time in the top 1.0 m in the field caused by the higher volume fraction of water in the bottom part of the lysimeter (Fig. 2B). A longer residence time in the lysimeter gives less leaching because there is more time for other loss processes such as transformation. This explanation was explored in additional simulations with a tracer that was applied every 2 yr (at the same time and application rate as the pesticides). The tracer was defined as a hypothetical substance that was not sorbed, not transformed, and not taken up by plants. Flux concentrations were calculated at 1 m depth for both the field and the lysimeter as a function of time for six typical years (so covering three applications) out of the simulation period of 46 yr. Figure 4 shows that the maxima of the flux concentrations in the lysimeter occurred later than those in the field, indicating a longer residence time in the lysimeter (similar to the results of Kasteel et al., 2007).


Figure 4
View larger version (21K):
[in this window]
[in a new window]

  		FIG. 4. Flux concentrations of a tracer (i.e., a substance that is not sorbed, not transformed, nor taken up by plants) at 1 m depth in the soil as a function of time as calculated for the field and lysimeter systems. Year numbers indicate 1 January of the corresponding year. Tracer (1 kg ha–1) was applied on 31 March in 1911, 1913, and 1915.

 
A consequence of the above interpretation is that the groundwater level in the field may have a strong influence on the effect shown in Fig. 3. For example, if another scenario would have been considered with a shallower groundwater level (e.g., on average, about 0.8 m), the effect could be in the opposite direction. This was confirmed by additional simulations for pesticide application every 2 yr where the parameter a of Eq. [3] was changed from 10 to 3 mm d–1 (leading to a lower water flux at the bottom of the soil system and thus to shallower simulated groundwater levels). Results were analyzed by calculating the cumulative probability density of the ratio of the leaching concentrations for the lysimeter divided by leaching concentrations for the field (as in Fig. 3). The median value of this ratio was found to be 112% for the 80th percentile leaching concentrations, confirming that the effect can be in the opposite direction when field systems with shallower groundwater levels are considered.

Adequateness of Prescribed Number of Application Years and Duration of Lysimeter Studies
Figure 5 shows that the ratio between the maximum leaching concentration for the lysimeter divided by the 80th percentile leaching concentration for the field varied from about 0.1 to about 100% for application every 2 yr and from 0.3 to about 200% for application every 3 yr. The lysimeter concentrations, therefore, were sometimes much lower than the field concentrations. The difference between application every 2 and 3 yr is the result of lower leaching concentrations simulated for application every 3 yr. These lower concentrations could be expected because for application every 3 yr, the leached mass is divided by the cumulative water percolation over 3 yr, which is about 50% higher than the cumulative water percolation over 2 yr.


Figure 5
View larger version (20K):
[in this window]
[in a new window]

  		FIG. 5. Cumulative probability density of the ratio of the maximum annual leaching concentration for the lysimeter divided by the 80th percentile leaching concentration for the field. The maximum is the maximum of two annual average leaching concentrations following a single application. The 80th percentile is derived from a series of 20 2- or 3-yr average flux concentrations for application every 2 or 3 yr, respectively. The probability density was based on 200 hypothetical pesticides with randomly generated organic-matter/water distribution coefficient, KOM, values between 0 and 100 L kg–1 and DegT50 values between 1 and 100 d, but results were included only if the 80th percentile concentration of the pesticide was above 0.1 µg L–1.

 
Because residence time of a pesticide in a lysimeter increases with increasing sorption, low ratios from Fig. 5 may be pesticides for which the 2-yr duration of the lysimeter study was not long enough to reach the maximum leaching concentration. To test this, the ratio was plotted against the KOM of the pesticide, which resulted in a close relationship (Fig. 6 ). As a further check, the ratio was also plotted against the DegT50, but this resulted in graphs with huge scatter. This suggests that low ratios are pesticides for which the 2-yr duration of the lysimeter study was not long enough to reach the maximum leaching concentration. This was checked in detail by a calculation for a pesticide with DegT50 = 50 d and KOM = 90 L kg–1, which had 80th percentile leaching concentrations of 0.17 µg L–1 for application every 2 yr and 0.09 µg L–1 for application every 3 yr. The lysimeter/field ratio for this pesticide was 0.5% for application every 2 yr and 0.9% for application every 3 yr, which fits well within the clusters of points shown in Fig. 6. In a run for the lysimeter made that extended the 2-yr period to 7 yr, the results showed that the maximum was reached in the fifth year. Figure 6 implies, therefore, that a single application season combined with a duration of a lysimeter study of 2 yr (as recommended by OECD, 2000) is not defensible for pesticides that are applied every two or more years in agricultural practice if their KOM is above ~40 L kg–1. The OECD (2000) indicated that it may be appropriate to extend the duration of the lysimeter study to 3 yr if the pesticide has a DegT50 longer than 180 d. However, Fig. 6 suggests that an extension of the duration is more strongly linked to the KOM than to the DegT50.


Figure 6
View larger version (19K):
[in this window]
[in a new window]

  		FIG. 6. The ratio of the maximum annual leaching concentration for the lysimeter divided by the 80th percentile leaching concentration for the field as a function of the organic-matter/water distribution coefficient, KOM. The maximum is the maximum of two annual average leaching concentrations following a single application. The 80th percentile was derived from a series of 20 2- or 3-yr average flux concentrations for application every 2 or 3 yr, respectively. The points are simulations for 200 hypothetical pesticides with DegT50 values ranging from 1 to 100 d, but results were included only if the 80th percentile concentration of the pesticide was above 0.1 µg L–1.

 
Further inspection of the model output of the simulation for this pesticide with DegT50 = 50 d and KOM = 90 L kg–1 showed that at the start of the fifth year, only 0.7% of the pesticide dose was left in the soil profile. Nevertheless, the maximum concentration leached in this fifth year. An argument in favor of a study duration of 2 yr could be that the soil can be sampled for pesticide at the end of the lysimeter study to check for any pesticide. However, this will not always be a reliable procedure because it is often impossible to detect only 1% of the dose. Concentrations in soil of this pesticide were simulated to be no higher than 0.01 mg kg–1 2 yr after application; this is close to the determination limit in soil for many pesticides.

Likelihood of Lysimeter Leachate Concentrations below 0.1 µg L–1 from Coincidental Favorable Combinations of Unknown Pesticide–Soil Parameters
Table 3 shows that all model pesticides used in the Monte Carlo simulations on coincidental favorable combinations of unknown pesticide–soil parameters had an average KOM of 35 L kg–1 or lower. This was done to avoid a strong influence of the effect of the number of application years and of the duration of the lysimeter studies on the results of these Monte Carlo simulations. The leaching of these model pesticides is further characterized in Table 3 by their 80th percentile field leaching concentrations and their lysimeter leaching concentrations as derived from deterministic simulations with the average parameter values. Table 3 shows that the probability of lysimeter leachate concentrations below 0.1 µg L–1 is higher in general for lower field and lysimeter leachate concentrations as calculated with average parameters (as could be expected). The three highest probabilities (22, 27, and 58%) were found for the three lowest leaching concentrations as calculated with average parameters. This relationship is further explored in Fig. 7 , where the probabilities of Table 3 are plotted against the 80th percentile field leaching concentration with application every 3 yr. Figure 7 shows that the probability is strongly influenced by the pesticide sorption and degradation properties (i.e., the average KOM and DegT50 used in the Monte Carlo simulations): the line calculated for a KOM of 35 L kg–1 and DegT50 values of 20 to 100 d shows much higher probabilities than the line for a DegT50 of 20 d and KOM values ranging from 5 to 35 L kg–1. For example, the line for KOM = 35 L kg–1 in Fig. 7 shows a probability of around 10% for a 80th percentile leaching concentration of 5 µg L–1 whereas the line for a DegT50 of 20 d and KOM values ranging from 5 to 35 L kg–1 shows a probability below 1% at this leaching concentration. The probability point of the line for DegT50 = 20 d with a value below 10% in Fig. 7 is the run with an average KOM value of 10 L kg–1. So the probability seems to be lower for mobile pesticides than for moderately sorbed pesticides.


View this table:
[in this window]
[in a new window]

  		TABLE 3. Probability that a lysimeter study generates a maximum annual leaching concentration below 0.1 µg L–1 for all combinations of average DegT50 (half-life at reference conditions) and KOM (organic-matter/water distribution coefficient) used in the Monte Carlo simulations considering coincidental favorable combinations of unknown pesticide–soil parameters together with the 80th percentile leaching concentrations for application every 2 or 3 yr in the field calculated for the average DegT50 and KOM and the maximum annual average leaching concentration for the lysimeter study calculated for the average DegT50 and KOM. Each probability is the average of at least three Monte Carlo simulations (each consisting of 500 runs). The 80th percentile leaching concentration is the 80th percentile of a series of 20 2- or 3-yr average flux concentrations for application every 2 or 3 yr, respectively.

 

Figure 7
View larger version (15K):
[in this window]
[in a new window]

  		FIG. 7. Probability that the lysimeter study generates a maximum annual leaching concentration below 0.1 µg L–1 as a function of the 80th percentile leaching concentration in the field calculated for application every 3 yr. Points are averages of at least three Monte Carlo simulations (of 500 runs each), and bars indicate standard deviations. The 80th percentile leaching concentration is the 80th percentile of a series of 20 3-yr average flux concentrations. Closed symbols are for an average organic-matter/water distribution coefficient, KOM, of 35 L kg–1 and average DegT50 values ranging from 20 to 100 d; open symbols are for average KOM values ranging from 5 to 35 L kg–1 and an average DegT50 of 20 d.

 
Table 3 and Fig. 7 show that the combined uncertainty in DegT50, KOM, Freundlich exponent, and dispersion length may result in a large uncertainty in the leachate concentrations generated by the lysimeter system. Figure 7 shows that at an 80th percentile field leaching concentration of 1 µg L–1, the probability of a favorable lysimeter result (<0.1 µg L–1) may be as high as 25%. It is up to risk managers to decide on an acceptable upper limit of this probability. However, probabilities of tens of percents are unlikely to be accepted by risk managers because the lysimeter is a higher-tier study conducted to demonstrate that a potential risk does not exist in reality. In view of the strong dependence of the probability on pesticide sorption and degradation parameters, it is advisable to assess this probability carefully for each lysimeter study used in pesticide registration procedures.

To determine to what extent each of the stochastic parameters (DegT50, KOM, Freundlich exponent, and dispersion length) contributes to the probability shown in Fig. 7, additional Monte Carlo simulations were performed for a KOM of 35 L kg–1 and a range of DegT50 values (see Fig. 7) in which only one of these parameters was stochastic and the others where kept fixed to their average or median values. The results are plotted in Fig. 8 against the 80th percentile leaching concentration with application every 3 yr as in Fig. 7. Figure 8 shows that the contribution of each of the parameters to the probability was more or less equal. The figure also shows that the combined variability of all parameters resulted in a much larger probability than the variability of one single parameter (as might be expected).


Figure 8
View larger version (21K):
[in this window]
[in a new window]

  		FIG. 8. Probability that the lysimeter study generates a maximum annual leaching concentration below 0.1 µg L–1 as a function of the 80th percentile leaching concentration in the field calculated for application every 3 yr. Points are averages of at least three Monte Carlo simulations (of 500 runs each), and bars indicate the standard deviations. The 80th percentile leaching concentration is the 80th percentile of a series of 20 3-yr average flux concentrations. Each line was calculated using an average organic-matter/water distribution coefficient, KOM, of 35 L kg–1 and average DegT50 values ranging from 20 to 100 d. Upper line is for stochastic DegT50, KOM, Freundlich exponent, and dispersion length. Other lines are simulations in which all parameters were kept fixed except the parameter indicated in the legend.

 

   Discussion and Recommendations
TOP
 ABSTRACT
 INTRODUCTION
 The PEARL Model
 Simulation Procedures
 Results and Discussion
 Discussion and Recommendations
REFERENCES
 
The above analysis ignored any uncertainty in the hydrological characteristics of the lysimeter and any systematic hydrological differences between lysimeter and the field, except for the lower boundary condition whose effect was demonstrated in Fig. 3. The OECD (2000) prescribes that the amount of rainfall plus irrigation should be appropriate considering the agricultural area relevant to the risk assessment. However, lysimeters may show more evapotranspiration than a corresponding agricultural field (Hansen, 1982; Boesten, 1994) due to local advective transport of heat to the lysimeter referred to as the "clothesline" or "oasis" effect (Hillel, 1980). This transport may be caused, for example, by hard surfaces (e.g., car parks) in the neighborhood. Therefore, it is advisable to check also whether the amount of water percolated out of the lysimeter was appropriate. This can be done by determining whether the percolation can be described with water flow models using defensible values of the input parameters (Boesten, 1994; Herbst et al., 2005). Downward movement of water is the driving force of pesticide leaching, which justifies focusing on the percolated amount of water.

Available data indicate that there may also be considerable random variation between replicate lysimeters. Jene (1998) found cumulative amounts of percolated water after 2 yr of 560, 570, and 660 mm for three cropped lysimeters taken from the same field. Kasteel et al. (2007) found cumulative amounts of percolated water after about 3 yr ranging from 800 to 1100 mm for six bare lysimeters taken from the same field. However, this variability in cumulative percolation can be handled in the risk assessment procedure by ignoring results from a replicate that has a cumulative percolation amount that is considered too low in view of percolation amounts in the agricultural area that has to be protected.

The presented simulations with a chromatographic model showed that lysimeter studies based on the existing OECD guideline do not necessarily generate realistic worst-case leaching concentrations, making it is difficult to defend the use of lysimeter leachate concentrations directly for decision making in the risk assessment. Thus, it is advisable that each lysimeter study (as part of any registration risk assessment) be assessed via modeling for the likelihood that the leachate concentration is below 0.1 µg L–1 considering all systematic and random factors identified in this study. (See Verschoor et al., 2001, for an example of procedures to be followed.) Only by including this information can risk managers make informed decisions on the usefulness of the lysimeter study for the leaching assessment.

The above recommendation is based on model simulations. If the model does not describe the reality either in the lysimeter or the field, then the model simulations may lead to a wrong interpretation of the lysimeter study. The alternative is to assess the above-mentioned factors experimentally for a range of pesticide–soil–weather conditions, including (i) comparison between field and lysimeters, (ii) lysimeter studies lasting considerably longer than 2 or 3 yr with pesticide applications in more than 2 yr, and (iii) parallel lysimeter studies with a range of soils. This is, of course, a more reliable procedure than a simulation study provided that the experiments are well conducted (see, e.g., the experimental and methodological problems of Jene, 1998, and Kasteel et al., 2007) and that the population of experiments is large enough to justify general conclusions (i.e., also extrapolating to pesticides that were not studied). Because such a database is not available yet and because it will take many years to generate it, using modeling to support interpretation of available lysimeter studies seems the best procedure currently available.

The BBA (1990) recommended using a sandy soil low in organic carbon as a realistic worst-case for German conditions. In addition, the OECD (2000, p. 14) claims that worst-case leaching conditions can be achieved by "e.g. a light sandy soil with low adsorptive and water holding capacity allowing rather fast movement of chemicals and water through the soil profile." However, the BBA (1990) and OECD (2000) provided no evidence that such a soil is indeed a worst-case in general (although it may be expected on the basis of chromatographic flow theory for nonionic pesticides). Smelt et al. (1983) conducted lysimeter studies with a sand soil with a low pH and two loam soils and found much more leaching of metabolites of aldicarb for the sand than for the loams (10–16% of dose for the sand vs. ~1% for the loams). These metabolites also caused widespread contamination of drinking water wells on sandy soils with low pH in Long Island (Zaki et al., 1982). Yon (1992) conducted lysimeter studies with a sand soil and a loam-over-clay soil and found more leaching of an unspecified pesticide in the loam-over-clay soil. Bergström et al. (1994) conducted lysimeter studies with a sand and a clay and found higher leaching losses for dichlorprop and bentazone in the clay. In lysimeter studies with another sand and another clay, they found higher leaching losses for bentazone in the sand. Brown et al. (2000) conducted lysimeter studies with isoproturon and five soil types (four mineral soils and a peat). One of these five was a sand. They found that the sand gave the smallest total leaching losses of isoproturon of the four mineral soils. Thus, in three out of five cases, sandy soils showed lower leaching losses of pesticide than did heavier textured soils. In general, one would not expect this on the basis of chromatographic flow theory. It indicates that preferential flow may have played a significant role in part of the studies (preferential flow was also observed by Brown et al. (2000) in part of their nonsandy soils). Admittedly, this is only a very limited set of data covering only a few soils and pesticides. Also, factors other than the flow mechanism (such as differences in sorption and degradation parameters) may have caused part of the differences. Thus, the effect of soil type on leaching may depend on the properties of the pesticide (e.g., sandy soils may be more vulnerable than other soils if the pesticide degrades much slower at low pH). Moreover, the risk assessment for pesticide leaching should not necessarily be based on studies with the soil types that generate the highest leaching concentrations because this risk assessment should be driven by its protection goals. In the Netherlands, for example, most of the drinking water pumping stations that collect groundwater from freatic aquifers are located in areas with sandy soils, whereas a considerable fraction of the heavier-textured soils are marine sediments located in the coastal areas with brackish groundwater unsuitable for drinking water purposes. Adequate assessment procedures for lysimeter studies with sandy soils will therefore remain a worthwhile risk assessment issue in the foreseeable future.


   ACKNOWLEDGMENTS
 
I thank Vincent Vulto for performing most of the Monte Carlo simulations and Jan Vanderborght (Forschungszentrum Jülich), Minze Leistra (Alterra) and two anonymous reviewers for useful critical comments and suggestions.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 The PEARL Model
 Simulation Procedures
 Results and Discussion
 Discussion and Recommendations
 REFERENCES
 




This article has been cited by other articles:


Home page
 Vadose Zone JHome page
H. Vereecken, P. Burauel, J. Groeneweg, E. Klumpp, W. Mittelstaedt, H.-D. Narres, T. Putz, J. van der Kruk, J. Vanderborght, and F. Wendland
Research at the Agrosphere Institute: From the Process Scale to the Catchment Scale
Vadose Zone J., August 11, 2009; 8(3): 664 - 669.
[Abstract] [Full Text] [PDF]

Home page
 Vadose Zone JHome page
J. Mertens, G. Kahl, B. Gottesburen, and J. Vanderborght
Inverse Modeling of Pesticide Leaching in Lysimeters: Local versus Global and Sequential Single-Objective versus Multiobjective Approaches
Vadose Zone J., August 11, 2009; 8(3): 793 - 804.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (6)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Boesten, J. J. T. I.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Boesten, J. J. T. I.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Boesten, J. J. T. I.
Related Collections
Right arrow Agricultural Pesticides
Right arrow Vadose Zone Risk Assessment
Right arrow Ground Water Quality

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome