Effective approaches for predicting environmental concentrations of pesticides: the apecop project

�� !"�� #�$�� !�%&��

Effective approaches for Assessing the PredictedEnvironmental Concentrations of Pesticides: A

proposal supporting the harmonised registration ofpesticides in Europe.

APECOP

QLK4-CT-1999-01238

Final ReportAugust 2003

Editor: M. Vanclooster

Section editors: V. Linnemann, N., Jarvis, A. Tiktak, M. Trevisan, A. Jones

Language editor: A. Armstrong

Authors:

M. Vanclooster, J.D. Piñeros Garcet, J.J.T.I. Boesten , F. Van den Berg, M. Leistra, J. Smelt,

N. Jarvis, S. Roulier, P. Burauel, H. Vereecken, A. Wolters, V. Linnemann, E. Fernandez, M.

Trevisan M., E. Capri, L. Padovani, M. Klein, A. Tiktak, A. Van der Linden, D. De Nie, G.

Bidoglio, F. Baouroui, A. Jones A. and A. Armstrong

Effective approaches for assessing the predicted environmental concentrations of pesticides (APECOP), Final Report 2

Acknowledgment

The authors would like to thank the many people belonging to the project consortium who

contributed to the execution of the APECOP project but which do not figure in the author list,

the different actors, stakeholders and end-users of the APECOP deliverables which influenced

the implementation strategy of the project, and the European Commission for providing

financial means allowing to execute this programme under FP5-Quality of Life Programme.

Citation

This document should be cited as follows:

M. Vanclooster, J.D. Piñeros Garcet, J.J.T.I. Boesten , F. Van den Berg, M. Leistra, J. Smelt,

N. Jarvis, S. Roulier, P. Burauel, H. Vereecken, A. Wolters, V. Linnemann, E. Fernandez, M.

Trevisan M., E. Capri, L. Padovani, M. Klein, A. Tiktak, A. Van der Linden, D. De Nie, G.

Bidoglio, F. Baouroui, A. Jones A. and A. Armstrong, 2003. APECOP: Effective approaches

for Assessing the Predicted Environmental Concentrations of Pesticides - Final Report.

Department of environmental sciences and land use planning, Université catholique de

Louvain, Belgium.


Table of contents

Acknowledgment........................................................................................................................ 2

Citation ....................................................................................................................................... 2

Table of contents ........................................................................................................................ 3

Executive summary .................................................................................................................... 7

What was the APECOP project? ............................................................................................ 7

What strategy was adopted to improve the actual exposure models? .................................... 8

How is preferential flow now considered in PEARL, MACRO and PELMO? ..................... 9

What was learned from the process studies on volatilisation from bare soil? ..................... 10

How is volatilisation modelled with PEARL, MACRO and PELMO? ............................... 10

Which validation strategy was considered to evaluate the modelling of the ground-water

exposure?.............................................................................................................................. 11

What is the validation status of MACRO, PEARL and PELMO for ground-water exposure

modelling? ............................................................................................................................ 12

Which validation strategy was considered to evaluate the scenarios for calculating

exposure to ground-water? ................................................................................................... 14

What is the validation status of the scenarios for ground-water exposure? ......................... 16

How can APECOP results be used in the EU registration process of PPP’s ....................... 17

1. Introduction .......................................................................................................................... 19

References ............................................................................................................................ 23

2. Improvement of preferential flow in PECgw models .......................................................... 24

2.1. Concepts for modelling preferential flow .................................................................... 25

2.2 Preferential flow in MACRO ......................................................................................... 27

2.3 Preferential flow in PELMO .......................................................................................... 29

2.3 Preferential flow in PEARL ........................................................................................... 30

References ............................................................................................................................ 33

3. Process studies of volatilisation from soil and canopies ...................................................... 35

3.1 Process studies of volatilisation from soil...................................................................... 35

3.2 Wind tunnel experiments of volatilisation from soil and canopy .................................. 38

3.2.1 The Wind tunnel...................................................................................................... 38

3.2.2 Experiments from bare soil ..................................................................................... 39

3.2.3 Experiments from plants ......................................................................................... 40

3.2 Field experiments of volatilisation from soil and canopy .............................................. 42


3.2.1 Introduction ............................................................................................................. 42

3.2.2 Volatilisation experiments from fallow soil............................................................ 42

3.2.3 Volatilisation experiments from canopies............................................................... 44

3.3.4 Prospects.................................................................................................................. 46

References ............................................................................................................................ 47

4. Modelling volatilisation processes ....................................................................................... 48

4.1 Concepts for modelling volatilisation from soil............................................................. 48

4.1.1 Introduction ............................................................................................................. 48

4.1.2 Description of model concepts ................................................................................ 48

4.1.3 Prospects for further developments......................................................................... 51

4.2 Concepts for modelling volatilisation from canopies..................................................... 52

4.2.1 Introduction ............................................................................................................. 52

4.2.2 Description of processes and factors ....................................................................... 52

4.2.3 Description of model concepts ................................................................................ 53

4.2.4 Prospects.................................................................................................................. 56

4.3 Volatilisation modelling with MACRO ......................................................................... 56

4.3.1. Application to the Jülich no-1 experiment ............................................................. 58

4.4 Volatilisation modelling with PELMO .......................................................................... 60

4.4.1 Volatilisation from soil............................................................................................ 60

4.4.2 Volatilisation from plants........................................................................................ 61

4.5 Volatilisation modelling with PEARL ........................................................................... 64

4.5.1 Introduction ............................................................................................................. 64

4.5.2 Improved model concept for volatilisation from soil in PEARL ............................ 64

4.5.3 Testing the new PEARL version for soil volatilisation........................................... 65

4.5.4 Testing the new PEARL plant volatilisation module.............................................. 66

4.5.5 Conclusions ............................................................................................................. 68

References ............................................................................................................................ 69

5. Overview of data sets used in the PEC model validation..................................................... 70

6. Validation of point scale models .......................................................................................... 71

6.1 Validation protocol......................................................................................................... 72

6.1.1 Introduction ............................................................................................................. 72

6.1.2 Validation steps ...................................................................................................... 73

6.1.3 Target values ........................................................................................................... 75

6.1.4 Statistical criteria..................................................................................................... 75


6.1.5 Models ..................................................................................................................... 76

6.2 Validation of PEARL ..................................................................................................... 77

6.2.1 Introduction ............................................................................................................. 77

6.2.2 Andelst..................................................................................................................... 77

6.2.3 Bologna ................................................................................................................... 81

6.2.4 Brimstone ................................................................................................................ 82

6.2.5 Lanna ....................................................................................................................... 82

6.2.6 Vredepeel................................................................................................................. 83

6.2.7 Conclusions ............................................................................................................. 84

6.3 Validation of MACRO ................................................................................................... 84

6.3.1 Andelst..................................................................................................................... 84

6.3.2 Lanna ....................................................................................................................... 84

6.3.3 Lebrija ..................................................................................................................... 86

6.3.5 Overall conclusions ................................................................................................. 95

6.4 Validation of PELMO .................................................................................................... 96

6.4.1 Introduction ............................................................................................................. 96

6.4.2 Andelst..................................................................................................................... 96

6.4.3 Lanna ....................................................................................................................... 98

6.4.4 Bologna ................................................................................................................... 99

6.4.5 Vredepeel............................................................................................................... 100

6.4.6 Conclusion............................................................................................................. 101

6.5 Intercomparison and assessment of the progress of the validation of the models ....... 101

6.5.1 Introduction ........................................................................................................... 101

6.5.2 Results ................................................................................................................... 101

6.5.3 Conclusion............................................................................................................. 104

References .......................................................................................................................... 106

7. PEC ground-water scenario evaluation .............................................................................. 108

7.1 Introduction .................................................................................................................. 108

7.2 Methodology for evaluating the FOCUS ground-water scenarios.............................. 110

7.2.1 Multiple aspects of scenario evaluation ............................................................... 110

7.2.2 Scenario representativeness definition .................................................................. 111

7.2.3 Determination of the FOCUS scenarios intension ................................................ 112

7.2.4 Determination of the extension universe............................................................... 113

7.2.5 Choice of the extension functions ......................................................................... 113


7.2.6 Reference extension determination ....................................................................... 113

7.2.7 Representativeness estimation.............................................................................. 114

7.3 Assessment of Pesticide Leaching at the Pan-European Level using a Spatially

Distributed Model (EuroPEARL) ...................................................................................... 114

7.3.1Materials and methods............................................................................................ 115

7.3.3 Results and discussion........................................................................................... 123

7.4 Scenario representativeness estimation using EuroPEARL......................................... 129

7.4.1 Introduction ........................................................................................................... 129

7.4.2 Discussion ............................................................................................................. 131

7.4.3 Conclusions ........................................................................................................... 134

7.5 Scenario representativeness estimation using a metamodel of EuroPEARL............... 134

7.5.1 Methodology ......................................................................................................... 135

7.5.2 Results and discussion........................................................................................... 138

7.5.3 Conclusions ........................................................................................................... 141

7.6 Synthesis....................................................................................................................... 142

References .......................................................................................................................... 144

8. Compliance of new approaches with current and future EU regulations........................... 146

8.1 Objectives..................................................................................................................... 146

8.2 Methodology ................................................................................................................ 146

8.3 Current EU legislation on Pesticides............................................................................ 146

8.4 Role of the new modelling approaches with present and future EU regulation........... 148

8.4.1 Potentials ............................................................................................................... 149

8.4.2 Recommendations ................................................................................................. 149

8.4.3 Weak points........................................................................................................... 150

8.4.4 Related Documentation ......................................................................................... 150

8.5 Implementation of APECOP models into EUSES....................................................... 150

Annex 1 : List of presentations and publications realised within the framework of APECOP

................................................................................................................................................ 154

Communication of project results at international meetings and workshops, communication

with stake-holders and end-users ....................................................................................... 154

List of publications............................................................................................................. 156


Executive summary

What was the APECOP project?

Environmental fate and exposure modelling of plant protection products (PPPs) is an integral

part of the risk assessment supporting the registration of PPPs in Europe and hence the

implementation of Council Directive EU/91/414. Within this context, ‘ground-water’ and

‘air’ are two environmental components which are subjected to the pressures induced by

contamination of PPPs and for which exposure modelling needs to be performed. To be

efficient in a harmonised registration process, the quality of the ground-water and air

exposure modelling needs to be assured. Quality assurance of exposure modelling implies the

implementation of a good modelling practice, and the validation, documentation and

maintenance of the modelling codes and scenarios. In recognition of the fact that there was

no agreed methodology for exposure modelling, the European Commission (DG-SANCO) set

up the FOrum for the Co-ordination of pesticide fate models and their USe (FOCUS). FOCUS

has published general guidance document and reports on the use of mathematical models for

predicting PECs in ground-water, surface-water and soil. Limited number of standardised

worst-case scenarios, to be used in the final calculations for ground-water exposure and

guidance to model selection, parameter selection and scenario selection became available in

2000. However, the FOCUS working groups also identified a range of uncertainties,

especially in relation to ground-water exposure modelling. These uncertainties were related

to the validity of the leaching models and scenarios.

To mitigate some of these uncertainties and to make a major step forward in air exposure

modelling, the research project entitled ‘Effective Approaches for Predicting Environmental

COncentrations of Pesticides (APECOP)’ was executed within the framework of the EU-FP5

Quality of Life Program. The major objectives of the project were i) to evaluate the validation

status of actual models and scenarios for predicting, within the framework of Council

Directive EU/91/414, environmental concentrations of PPPs in ground-water; and ii) to

reduce the uncertainties in the predictions, by improving the description of preferential

transport of PPPs in soils and volatilisation of these substances to air. For realising these

project objectives, experimental studies were compiled or executed to improve the process

understanding of PPPs fate and behaviour in the soil-plant-atmosphere system and to allow to


evaluate the performance of local scale PPPs exposure models. Modelling studies at the local

scale envisaged the evaluation of the performance of local scale models and modelling tools.

Modelling studies at the large EU scale allowed to evaluate the appropriateness of the

scenarios. The project was executed by a consortium of leading scientists in the domain of

exposure modelling of PPPs, both from high quality research centres and universities,

belonging to 6 different European countries. The research was executed in close collaboration

with well identified end-users, in particular the FOCUS steering Committee of DG-SANCO

and the European Crop Protection Agency. The major deliverables of the project are: i) a set

of improved models for describing ground-water and air exposure of PPPs, considering in

particular a better description of the preferential flow process in soil and the volatilisation

process from soil and plants; ii) methodological approaches for validating exposure models

and scenarios; iii) an assessment of the validation status of the models and the scenarios for

ground-water exposure modelling.

What strategy was adopted to improve the actual exposure models?

Within the FOCUS procedure, four PPP ground-water exposure models are currently

considered to calculate in a first tier environmental concentration of PPPs in ground-water:

PEARL, MACRO, PELMO and PRZM. However, given the fact that PRZM is conceptually

very similar to PELMO and since the model developer of PRZM does not belong to an

European research institute, only PEARL, MACRO and PELMO were considered for model

improvement within APECOP.

One of the major problems in the current ground-water exposure modelling is the modelling

of the preferential flow process in soils. Preferential flow in soils is the kinetic process by

which PPPs are leached fast from the root zones of crops by mechanisms that cannot be

described by classical soil matrix flow theory. Preferential flow can be driven by the presence

of structural pores in a macro-porous soil, by extreme variability of porosity in the soil matrix,

and by the hydrophobicity creating unstable wetting fronts and fingers. The lack of thorough

understanding of preferential flow, put a major constraint on the prediction of PPPs with the

current approaches. However, ignoring preferential flow is likely to induce considerable bias

in the present adopted risk assessment procedure. Preferential flow was already present in the

MACRO model at the onset of the APECOP, but suffered from a series of problems related to

its robustness. The PEARL and PELMO model did not have preferential flow functionalities.

Hence, a preferential flow module was added to PELMO and PEARL while the preferential

flow module of MACRO was upgraded.


Emission by volatilisation is another major process that is poorly represented in the present

exposure procedures. Volatilisation is a major source of PPPs residues in air, fog and rain, and

thus may lead to a long-range transport of residues remote from their application.

Volatilisation is therefore likely to have a major impact on environmental balance of PPPs.

Unfortunately, the volatilisation process is poorly understood. It is a result of a series of

dynamic processes occurring in the soil-crop canopy- atmosphere continuum. To elucidate

these process as a basis for improved volatilisation modelling, process studies of volatilisation

from soil were carried out in a photovolatility chamber. Wind tunnel experiments and field

volatilisation experiments using micro-meteorological techniques were further used to

evaluate the new volatilisation modules which were implemented in the PEARL, MACRO

and PELMO model.

How is preferential flow now considered in PEARL, MACRO and

PELMO?

Models accounting for flow in soil macropores can be broadly classified as either pore-scale

models or lumped dual- or multi-region models, and as either stochastic or deterministic.

Pore-scale models and stochastic approaches are conceptually the most accurate for

describing preferential flow but at present these approaches are difficult to apply in regulatory

practice due to lack of robust coding and parameterisation approaches. Therefore, it was decided

to concentrate in APECOP on approaches for modelling macroporous flow using deterministic

dual-and multi-porosity modelling approaches. Such an approach was considered as the only

feasible way to include heterogeneous flow in the operational regulatory exposure models. The

models take a macroscopic, continuum, approach by lumping individual preferential flow

pathways in the soil into two or more pore regions, each characterized by a porosity, a water

pressure (or water content), water flow rate and solute concentration.

The MACRO model already included different transport descriptions for the macropores and the

micropores flow domain . The existing macropore module was upgraded in the APECOP project

and the following new features were added to the model: i) improved numerical algorithms for

the solution of the flow and transport equations in the micropore and macropore domain; ii)

continuous van Genuchten hydraulic functions for the parameterisation of the soil hydraulic

processes; iii) a physico-empirical approach for modelling tillage effects on soil flow and

transport; iv) a kinetic two-site sorption model; and v) an inversion module for estimating model

parameters using a global search Bayesian parameter estimation approach.


A preferential flow module for PELMO was developed in the APECOP project. The upgraded

version of the PELMO model considers two flow domains, but the flow in the macroporous

flow domain is only activated when a critical threshold rainfall value is obtained. The

macroporous flow process is considered to be an internal catchment process with a re-

equilibration of mass between the macropores and the micropores at the bottom of the

macropores. The depth of the macropores are considered to be fixed in depth.

In the new beta version of the PEARL model, a diversity of pore types in the macroporous flow

domain are considered. A difference is made between continuous bypass pores and internal

catchment pores. In addition, a difference is made between permanent and time variant

macropores. The latter type of macropores are induced by soil moisture dependent swelling and

shrinkage processes and builds upon the availability of the shrinkage-swelling characteristic.

Flow and transport in both pore systems are modelled using classical approaches. The mixing

cell concept such as also used in MACRO and PELMO was introduced to define the top

boundary condition.

What was learned from the process studies on volatilisation from

bare soil?

The volatilisation experiments performed under controlled conditions of radiation, soil

moisture, air humidity, soil temperature and wind velocity, clearly elucidated the effect of

soil moisture on volatilisation. This could be explained by an increased sorption of the active

gradient when soil dries out. The inclusion of this process in the exposure models, together

with a better parameterisation of the aerodynamic resistances in the boundary layer and an

improved spatio-temporal discretisation of the process description at the soil surface, resulted

in a better description of the volatilisation from bare soil under both controlled and field

conditions.

How is volatilisation modelled with PEARL, MACRO and PELMO?

The volatilisation from bare soil in PELMO was upgraded by improving the spatio-temporal

discretisation at the soil surface (smaller time steps, smaller space steps at the soil surface),

improving the empirical description of temperature dependency of Henry’ s constant and

increased sorption of volatile PPPs in dry soils. Volatilisation from the crop canopy is

calculated together with other crop fate processes such as crop penetration, wash-off and

photodegradation. Volatilisation through the canopy boundary layer is modelled by means of

a Fickian diffusion approach.


A more advanced approach for describing volatilisation has been developed in PEARL.

Volatilisation from bare soil through the laminar boundary layer at the soil atmosphere

interface is modelled in PEARL by means of a diffusion-like process. Improved descriptions

of the aerodynamic resistances in the laminar boundary layers were introduced, together with

an improved spatio-temporal discretisation scheme of the flow processes at the soil surface,

and an improved description of the sorption process when soil dries out. A beta version of a

full canopy fate model considering wash-off, product penetration and volatilisation from

canopies is also available, but will be further tested before being released.

In MACRO, only the volatilisation of PPPs from the soil is considered. Use is made of a

simple empirical correction of the applied dose, based on an experimentally derived

relationship between the cumulative volatilisation in the 21 days following application as a

percentage of the dose, and the fraction of the compound calculated to be present in the air

phase. This latter is calculated as a moisture and temperature dependent equilibrium process

in the soil boundary layer.

The volatilisation of fenpropimorph from the bare soil in the field experiments were

successfully predicted by the PEARL and PELMO models. The prediction of the

volatilisation of fenpropimorph and parathion-methyl was somewhat poorer for MACRO but

judged to be sufficiently well described from a leaching point of view. In contrast to this, the

simple empirical method in MACRO was successful in simulating the volatilisation of

terbuthylazine from bare soil experiments.

Which validation strategy was considered to evaluate the modelling

of the ground-water exposure?

Field experiments were made available to evaluate the performance of the PEARL, MACRO

and PELMO in predicting ground-water exposure. The major criterion for considering a field

site was the availability of data appropriate for model validation. This implies the availability

of high quality data on water, solute, and heat transport in the soil, and data on fate and

transport of the PPP in soil and ground-water (if applicable). The availability of support from

the data set provider was considered to be essential. A multistage validation approach was

considered in APECOP. The different components of the emission models are validated

separately in a sequential process. The originality of the procedure resides in the adoption of a

blind validation strategy for the evaluation of the models on some of the field sites. For this

blind validation exercise, no field data were made available to the model users. Hence, it was

evaluated how good exposure models describe field behaviour if only laboratory data or


generic data are made available. Although it is well known that calibration may improve

considerably the model performance, field data are generally not available in an operational

context of registration and the blind validation level is therefore the most appropriate

validation level to be considered for a real life registration exercise.

What is the validation status of MACRO, PEARL and PELMO for

ground-water exposure modelling?

The overall results show that the validation status of the three exposure models is quite

variable depending on the dataset simulated. No large differences are found between models,

even if Richards equation based models are better to simulate hydraulic behaviour in the soil

system. Prior knowledge of real data improves significantly the performance of models. In

general, the blind predictive ability of the models for the system state variables and in

particular the absolute concentrations of the active gradients at the ppm level is still poor, and

more research is needed to identify the reasons for this: is it the models themselves, or a lack

of data to enable appropriate parameterisation? The improvements in model performance that

can be demonstrated following ‘physically justifiable calibration’ , suggest that it is mostly the

latter. This poor predictive ability in absolute terms does not invalidate the models for

screening appropriately active gradients for ground-water exposure or for ranking correctly

the active gradients in terms of leaching risk. However, the predictive ability for relative

ranking or screening was not tested in the project.

The results for the Andelst field site illustrates the need for high-quality datasets for correctly

assessing the validation status. If in this case only the mobile bentazone would have been

tested and only soil profiles would have been measured, then the deficiencies in the transport

part would not have been detected. In the same way, a good correspondence between

predicted and measured lack of movement in the Bologna loam soil could not be considered

as a sensitive test of the models. The same applies to the test for the Vredepeel dataset

because this test focussed on movement of the bulk of the PPPs rather than to the exceptional

events leading to leaching in drain and ground-water. Another methodological issue was

highlighted with the use of the statistical modelling evaluation indicators. Sometimes there

was a large discrepancy between the model efficiency and the graphical comparison of

measurements and calculations. Systematic research on the causes of such discrepancies is

therefore recommended.


The PEARL model was tested against five leaching datasets (Andelst, Bologna, Brimstone,

Lanna and Vredepeel). Three of the datasets were on clay soils (Andelst, Brimstone and

Lanna) and the other two were on a sandy loam (Bologna) and a sandy soil (Vredepeel). The

new PEARL version including preferential flow could not be finalised within the time frame

of the project, thus the results pertain only to the chromatographic version of the model.

In all model tests, calibration of the water flow submodel appeared necessary for an adequate

description of water flow. After this calibration, a more or less acceptable description could

be achieved in nearly all cases.

The three tests on the clayey soils indicated that the chromatographic version of the PEARL

model cannot adequately describe solute transport in well structured soils. It is expected that

this problem will partly be solved with the future version of PEARL which will include a

preferential flow module.

The MACRO model was fully tested on the Lanna and Lebrija data set. Simulation results for

the Andelst field site are only partially available. For the Lebrija data set, both column and

field data were available. The Lanna and Lebrija data set sites are characterized by clayey soil

types exhibiting preferential flow. Both the original and the upgraded model was used in the

analysis.

Both versions of the model produce acceptable fits to the data for the Lanna field site. The

new version of the model however, does not perform better than the original one. This

surprising result is partly due to the fact that the more accurate numerical solution resulted in

different parameter estimates. The differences in parameter estimates between the model

versions seem to overshadow the effects of different modelling concepts.

Both versions of MACRO gave largely acceptable simulations of bromide leaching, field soil

water contents, and the fate of the two pesticides (chloridazon and lenacil) in the recently

tilled clay soil of Lebrija. Underprediction of persistence early in the experiment, and

overprediction later, was noted. The reason for this is not clear, but it may be due to

departures from first-order degradation kinetics. Also for this field site, significantly different

parameterisations gave very similar predictions.

The similar modelling performance for the former and upgraded model suggests that, despite

the fact that the datasets must be considered unusually comprehensive, the model

parameterisation was still insufficiently constrained by measured validation data, resulting in

‘ill-posed’ problems. More research is needed to identify the data requirements for reliable

inverse estimation of model parameters related to macropore flow. However, the new version

of MACRO still represents a considerable improvement compared to the currently available


version of the model, partly in terms of the availability of new process descriptions (e.g.

tillage, kinetic sorption) and improved numerical accuracy, but also with respect to user

friendliness.

The PELMO model was tested on the Bologna, Vredepeel, Andelst and Lanna field sites,

comprising both clayey and non clayey soil types. The results showed that soil moisture was

often weakly estimated, and this is in contrast to other state variables, such as the soil

temperature. Calibration of the hydraulic variables may have improved the situation slightly

but was not performed as the effect on the pesticide parameters was relatively small. The

original PELMO could well describe the experiments performed at Vredepeel and Bologna

after calibration of the application rate to correct for the volatilisation losses at the beginning

of the experiments. However, preferential flow was not the dominant process at these

locations. In addition, a similar remark should be made here as for the PEARL model: these

test sites allow to evaluate the bulk behaviour of the PPP in soil rather than the leaching to

ground-water. The situation was completely different for the other field data sets at Lanna

and Andelst. Only the new version of PELMO with a preferential flow module was able to

simulate the situation in these soils correctly. It is presently not possible to conclude about

the quality of the new development. Initial simulations suggest more realistic description for

instance of pesticide volatilisation from soil surfaces. However, as not much experience has

been made with the new models it is much too early to define default (standard) values for

these new input parameters.

Which validation strategy was considered to evaluate the scenarios

for calculating exposure to ground-water?

A scenario in the context of the actual recommended FOCUS ground-water exposure

procedure is defined as a representative combination of crop, soil, climate and agronomic

parameters to be used in the first tier exposure modelling to ground-water with the exposure

models such as MACRO, PEARL and PELMO. Representative means that the selected

scenarios should represent physical sites known to exist, i.e. the combination of crop, soil,

climate and agronomic conditions should be realistic. FOCUS intended to construct scenarios

that represents an overall vulnerability approximating the 90th percentile leaching of all

possible situations (this percentile is often referred to as a realistic worst case). It was further

assumed that the vulnerability was equally attributed to soil and climate. To achieve this, they

first defined nine climatic areas inside the arable lands of Europe, so-called FOCUS areas.

Within each of these areas, the FOCUS team selected an approximate 80% vulnerable soil,


which implies that the concentration of a PPP should be less than the EU drinking water limit

in at least 80% of the corresponding area. The combination of the 80th percentile vulnerable

soil with an 80th percentile vulnerable climate was supposed to include at least a 90th

percentile vulnerable case. Hence, a FOCUS ground-water scenario is supposed to be a

combination of parameter values selected in such a way that the leaching concentration

calculated with a FOCUS exposure model equals the 90th percentile of leaching inside the

corresponding FOCUS area. It was an objective of the APECOP project to validate this

scenario definition statement.

The key ,, , question to answer was: ‘Are FOCUS scenario combinations of parameter values

selected in such a way that when used in combination with a FOCUS PEC ground-water

model, the calculated leaching values correspond to the real 90th percentile of leaching?’ To

answer this question, it was essential to define the ‘real’ 90th percentile of leaching. In an ideal

situation, this 90th percentile could be obtained from a detailed monitoring of the presence of

active substance in ground-water. However, our observational skills are limited in time and

space, and therefore the ‘real’ 90th percentile is unknown. If the 90th percentile of leaching

cannot be obtained from direct measurements, an alternative is offered by approximating

these leaching values using a ‘detailed’ assessment technique. This technique comes down to

approximating the ‘real’ 90th percentile of the leaching concentration by means of a spatially

distributed leaching model in combination with Pan-European soil, climate and agricultural

databases. Obviously, this type of validation has a lower power than a validation in which the

90th percentile of leaching is estimated from direct measurements, but it is the only pragmatic

way to proceed with scenario validation at this time. Given the computational burden

associated with the use of spatially distributed modelling technique, the validation could only

be applied to a single PPP, to two major agricultural crops (maize and winter wheat), using

only one single exposure model (PEARL). This choice implies that the vulnerability is

considered to be driven mainly by soil and climate, and not by crop properties.

Pan-European spatially distributed simulations were performed by means of the PEARL

model (i.e. the EuroPEARL model) for 1062 unique combinations of soil type and climate,

sampled throughout Europe. Soil properties, including soil horizon designations, were

obtained from the Soil Profile Analytical Database of Europe in combination with the

1.000.000 European soil map. Daily weather data for each unique combination were obtained

by scaling the time series of the 9 available FOCUS climate scenarios inferred from the

MARS data base using the mean yearly averaged temperature and precipitation map. Other

data such as irrigation data, crop data and pesticide properties were compiled from various


sources, such as inventories, field-studies and the literature. The 1062 unique combinations

together represent at maximum 75% of the total agricultural area of the European Union.

Austria, Sweden and Finland could not be included in the simulations, because there was

insufficient soil profile information for these countries. However, to consider also the soil-

climate and crop combinations for which no profile data where available, a meta PEC ground-

water model was derived by interpolating EuroPEARL modelling results in the input

parameter space using radial basis functions neural networks. The combination of the process

based deterministic EuroPEARL model with the meta-model allowed the project to: i) use (for

leaching simulations) the European 1:1 000 000 soil map that covers 97% of the European

agricultural area instead of the 75% covered by the profile database; ii) take into account

spatial variability of leaching inside the mapping units; iii) test the validity of the scenarios in

a statistical way by comparing the probability density function of the 90th percentile leaching

risks generated with the EuroPEARL metamodel with the results obtained using the FOCUS

procedure; and iv) to consider the model uncertainty in a explicit way.

What is the validation status of the scenarios for ground-water

exposure?

Results of the EuroPEARL model and its meta-model are presented in maps with a resolution

of 10x10 km2, which is the highest justifiable resolution based on the vectorial EU soil map

1:1 000 000.

The results indicate that the leaching concentration generally increases with precipitation and

irrigation and decreases with increasing organic matter content. Because of the strong

sensitivity of the leaching concentration to soil properties, there is a strong variability of the

calculated leaching concentration at relatively short distances. Results further indicate that

due to large irrigation amounts combined with large temporal variation of rainfall in the

Southern European countries, the trend in the calculated leaching risks from North to South is

less extreme than expected. This implies that areas of high leaching risk (‘hotspots’ ) as

assessed by means of the EuroPEARL model occur in all countries of the European Union,

including the Southern European countries.

Comparing results of the EuroPEARL model with the FOCUS calculations shows that, for the

limited cases which were analysed, the FOCUS ground-water scenarios for Southern

European countries may be biased and should be considered as needing revision. It should

also be noted that the variability in the soil properties in EuroPEARL model may be

underestimated since it is built on the use of dominant soil types in the European soil map.


The probability of obtaining exceptional soil properties, and hence extreme vulnerable

situations, decrease when dominant soil types are used. The EuroPEARL based evaluation of

the representativeness of the scenario may therefore be too optimistic. Comparing the meta-

modelling results with the FOCUS scenarios shows that the inclusion of the additional soil

variability could increase substantially the leaching percentiles, and hence the probability of

rejecting the actual FOCUS scenarios. However, these results should be interpreted with a lot

of caution, considering the uncertainties in the meta-model predictions for these areas

characterized by soil and climate properties situated outside the EuroPEARL calibration

dataset range. In addition, the poor quality of the Pan-European databases reduces

substantially the reliability of the meta-model predictions. In particular the poor relationship

between the European soil map and the analytical SPADE database introduces substantial

uncertainties in the validation procedure. In addition, the full validation test with the meta-

model was performed for only one single substance and one single crop. It is expected that

representativeness estimation can change if more accurate databases become available and if

more substances and crops are simulated.

How can APECOP results be used in the EU registration process of

PPP’s

The new approaches for describing preferential flow in soil and volatilisation from soil and

crop canopies result in new versions of the PEC ground-water models. The new releases of

the codes will be considered by the FOCUS version control group for inclusion in the current

first tier harmonised operational ground-water exposure assessment procedures. The

validation study of the local scale PEC ground-water models increases the validation status of

the models and hence increases the confidence of model users, i.e. regulators and applicants,

in the current recommended procedures developed by FOCUS. Some critical results related to

this model validation, however, suggest that PEC modelling is, and always will be, associated

with uncertainty.

The scenario evaluation result in a first transparent and systematic assessment of the validity

of the FOCUS ground-water scenarios. The validation test should however be extended for

more active substances and more complex land uses scenarios. If the results are confirmed,

then some of the current scenarios, in particular those for Southern Europe, should be

considered for revision. The positive results for other scenarios will increase the confidence of


regulators and applicants in the appropriateness of the FOCUS ground-water scenarios for

lower tier ground-water exposure assessments.

The spatially distributed modelling approach which was introduced to assess the validity of

the ground-water scenarios allowed the project to introduce a detailed description of the

variability of soil, crop and climate in ground-water exposure assessment. The spatially

distributed modelling approach could therefore be at the core of a higher tier refined

assessment procedure which considers variability of land properties in an explicit way. In

such a procedure, data coming from ground-water monitoring studies could be assimilated

with data generated by means of the spatially distributed model. This should be the basis for

further study and the development of scientifically based, harmonised, detailed, refined higher

tier exposure assessment procedure supporting the current EU registration process.


1. Introduction

Appropriate agricultural management at the European scale is needed to reduce ground-water

contamination risk by residues of toxic plant protection products,. The Council Directive

91/414/EEC, related to the registration of active substances of plant protection products, is

one of the instruments which has been developed to meet sustainability criteria in the

agricultural sector. The directive envisages, amongst others, the implementation of uniform

principles for assessing the risk associated with the use of plant protection products, and this

to support a harmonised registration at the EU level. Predicting the environmental

concentrations of pesticides in the different environmental components by means of

mathematical models is an essential part of such a risk assessment.

In recognition of the fact that there was no agreed methodology for PEC calculation, the

European Commission (DG-SANCO) set up the FOrum for the Co-ordination of pesticide

fate models and their USe (FOCUS). FOCUS has published a general guidance document and

reports on the use of mathematical models for predicting PECs in ground-water, surface-water

and soil (e.g. FOCUS, 1995). A limited number of standardised worst-case scenarios, were to

be used in the penultimate calculations for PEC ground-water and guidance to model

selection, parameter selection and scenario selection became available in 2000 (FOCUS,

2000). Standardised scenarios are needed because they increase the uniformity of the

regulatory evaluation process by minimising the influence of the person that performs the

PEC ground-water calculation and because they make PEC calculations and their

interpretation much easier for administrators, regulators and industry (Boesten et al., 2000).

The PEC ground-water calculation model and the scenarios should apply to the whole EU,

and this implies the consideration of the variability of soil, crop, climate, hydrology and

hydrogeology, agricultural practice, land use and pesticide use at the community level. An

appropriate PEC ground-water model should therefore ideally be sufficiently validated for the

different conditions in the EU; should use compiled databases on soil, climate, hydrology and

hydrogeology at EU level; should permit the generation of realistic worst case input for the

PEC models; and should be available and compatible with the identified PEC ground-water

models. A series of shortfalls from this ideal were identified by the FOCUS ground-water

working group and addressed in some cautionary notes related to the uncertainty of the


proposed procedures (FOCUS, 2000). With respect to ground-water, the major uncertainties

are related to the validation status of present PEC models and scenarios.

Indeed, validating a PEC ground-water model implies the quantification of the error that is

made when predicting e.g. leaching with different leaching modelling codes. Quantifying this

modelling error for all possible scenarios in which the model may potentially be used is

simply not possible since experimental data for all possible scenarios are not available. Hence,

when evaluating the model performance for a new condition (e.g. the fate of a new substance

at a new site), there is always a possibility that the model error will become unacceptably

large, and that the modelling code should be considered as invalidated for this site. Therefore,

the FOCUS framework considers the overall validation status of a model will be increased if

it has successfully passed several validation tests. As a model is increasingly tested, the

probability of success in a new scenario with similar properties will also increase. The PEC

ground-water models proposed by FOCUS have been through several validation exercises

before (e.g. Styczen, 1995; Thorsen et al. 1998; Vanclooster et al., 2000). However, the

exercises documented so far represent a very limited number of cases as compared to the

number of potential scenarios for which the models will be used in the registration context,

and the results of such exercises are rather variable. Therefore scope exists to increase the

number of validation studies of the FOCUS PEC ground-water models.

In contrast to the studies related to the validation of the PEC ground-water modelling codes,

little attention in the relevant literature has been devoted to the validation of modelling

scenarios. Within the framework of the Tier 1 FOCUS PEC ground-water procedure,

vulnerability of ground-water to contamination resulting from the use of an active substance is

represented by means of nine realistic worst-case scenarios. Collectively, these nine scenarios

represent agriculture across Europe, for the purposes of a Tier 1 EU-level assessment of

leaching potential. The realistic worst case was identified with the concept that scenarios

should correspond to 90th percentile vulnerability situations (FOCUS, 2000). This is, in

reality, a function of all system properties (weather, soil, ground-water, crop, substance

application and chemical properties). A correct theoretical identification of this realistic worst

case would therefore imply the development of a few hundred scenarios at the EU-level,

which should all be run for the specified substance. A 90th percentile vulnerable scenario

could then be identified from the resulting frequency distribution. At the time of the start of

the FOCUS activity, Pan-European climate and soil databases, which are required to develop


this large number of scenarios, were not available. Therefore, a statistical approach to infer

optimal scenarios could not be adopted, thereby adding uncertainty to the Tier 1 assessment.

Again, scope exists to reduce this source of uncertainty, hereby improving the quality of PEC

ground-water modelling in the Tier 1 level risk assessment.

Further, in contrast to the PEC ground-water, surface-water and soil which was addressed by

the previous FOCUS working groups, little attention was paid so far to the calculation of PEC

to air in a harmonised way. This is a little bit surprising given the possible human

toxicological and ecotoxicological impacts of the volatile components of PPPs in air.

In order to mitigate the above mentioned uncertainty problems and to make a major step

forward in PEC air calculations, the project entitled ‘Effective approaches for predicting

environmental concentrations of pesticides, APECOP’ was included in the framework of the

EU-FP5 Quality of Life Program. The primary objectives of APECOP were:

• to evaluate the validation status of the PEC ground-water models and scenarios, as

proposed by FOCUS; and

• to propose effective strategies to reduce the uncertainty in the present PEC models and

scenarios, based on a detailed validation analysis. This reduction of uncertainty will be

obtained by proposing alternative approaches for considering pesticide exposure by

volatilisation and preferential flow.

The secondary objectives of the project were:

• to establish a network of high quality field data sets, as a basis for the validation of the

present PEC ground-water models and scenarios, and for the realisation of process studies

on pesticide volatilisation;

• to compile data of the high quality field sites in a centralised data base, accessible for the

project partners;

• to identify the validation status of the physically based PEC ground-water models as

proposed by FOCUS;

• to identify the validation status of the present ground-water scenarios as proposed by

FOCUS;

• to improve physically based PEC ground-water models, by improving the description of

the preferential flow and volatilisation from soil and plant surfaces; this latter being a first

attempt to model PEC’ s to air; and


• to analyse the sensitivity of easily accessible data, available in a pan-European data base

on PEC’ s at field scale, as a basis for the development of new improved robust and cost

effective PEC calculation schemes, applicable at the larger pan-European scale.

For reaching the project objectives a set of work packages were executed by 9 project

partners, belonging to 6 different European countries. The data collection phase provided the

ground truth data of pesticide environmental concentrations at 7 selected sites in Europe.

These data are the basis of an evaluation of the validity of present and improved local scale

conceptual PEC models. Process studies in the laboratory and controlled field conditions

were at the basis of the development of new modules for describing PPP volatilisation. The

analysis of data, presently available at the pan-European scale, and the sensitivity of the

conceptual models and alternative models to regional available data formed the basis for the

development of robust PEC models that can directly be integrated with present actual

regional data bases.

The major deliverables of the project are: i) a set of improved models for describing exposure

of PPPs to ground-water and air, considering in particular a better description of the

preferential flow process in soil and the volatilisation process from soil and plants; ii)

methodological approaches for validating exposure models and scenarios; iii) an assessment

of the validation status of the models and the scenarios for exposure modelling to ground-

water. End users of the project products are regulatory bodies at the European level (e.g.

Standing Committee on Plant Health, FOCUS steering committee) and industry (ECPA,

EPPO). They have been indirectly involved in the project through the project partnership.

The present report , summarises the major findings of the project. More details can be found

in the series of papers and documents realised within the framework of this project and

referred to in Annex 1.


References

Boesten J., R.L. Jones, M. Businelli, A.B. Delmas, B. Gottesbüren, K. Hanze, T. Jarvis, M. Klein, A.M.A. vander Linden, W.M. Maier, S. Rekolainen, H. Resseler, M. Styczen; K. Travis and M. Vanclooster, 2000. Thedevelopment of FOCUS scenarios for assessing pesticide leaching to ground-water in EU registration. In:‘The 1999 Brighton conference – Weeds’ 527- 536.

FOCUS, 1995. Leaching models and EU registration - EC Document DOC.4952/VI/95, 123ppFOCUS, 2000. FOCUS ground-water scenarios in the EU plant protection product review process - Report of the

FOCUS Ground-water Scenarios Workgroup, EC Document Reference Sanco/321/2000, 197ppStyczen, M., 1995. Validation of pesticide leaching models. In: Leaching models and EU registration. Final

report of the FOCUS work group. DOC. 4952/VI/95.Thorsen, M., Jørgensen, P.R., Felding, G., Jacobsen, O.H., Spliid, N.H., and Refsgaard, J.C, 1998. Evaluation of

a stepwise procedure for comparative validation of pesticide leaching models. J. of Environmental Quality 27(5) : 1183-1193.

Vanclooster M., J. Boesten, M. Trevisan, C. Brown, E. Capri, O.M. Eklo, B. Gottesbüren, V. Gouy and A.M.A.van der Linden, 2000. A European test of pesticide-leaching models: methodology and majorrecommendations. Agricultural Water Management 44: 1-21.


2. Improvement of preferential flow in PECgw models

Until recently, the prevailing conceptual model of water infiltration into soils was based on the

idea that ‘new’ incoming water displaced existing ‘old’ water uniformly, with water moving

downwards in the soil profile as a broad and well-defined ‘wetting front’ . Similarly, the

prevailing view of agrochemical transport was that leaching took place as a chromatographic

process, with the chemical as fully exposed to adsorption sites in undisturbed field soils as it

would be both in laboratory batch experiments on water-slurry mixtures, and column leaching

experiments on repacked soils. The idea that water flow and pesticide transport normally takes

place as a uniform displacement process in soils has now been abandoned, and has been replaced

by an understanding that the heterogeneity of undisturbed soils in the field often leads to

markedly non-uniform patterns of water flow and agrochemical displacement.

Preferential flow is the generic term used to describe this irregular wetting. It is an umbrella

term, covering several processes with different physical causes, but with the common feature

that non-uniform wetting leads to an increase of the effective velocity of the water flow

through a small portion of the soil unsaturated zone. For example, in structured soils,

macropores (shrinkage cracks, worm channels, root holes) may dominate the soil hydrology,

particularly in fine-textured soils, where they operate as high-conductivity flow pathways by-

passing the denser impermeable soil matrix (Beven and Germann, 1982). Preferential flow

also occurs in unstructured sandy soils in the form of unstable flow or fingering (Hillel, 1987)

caused by profile heterogeneities such as horizon interfaces or water repellency (Hendrickx et

al., 1993), or simply as heterogeneous flow in soils characterised by mixtures of materials of

differing texture and pore size distribution (Roth, 1995).

Preferential flow greatly increases the risk of leaching of surface-applied agrochemicals to

ground-water and surface-waters bodies (Jarvis, 2002), since infiltrating water is channeled

through only a small fraction of the total pore space, at rates which are too fast to allow

sufficient time for equilibration with slowly moving ‘old’ water stored in the bulk of the soil

matrix. Thus, much of the adsorption and degradation capacity of the chemically and

biologically reactive topsoil is ‘by-passed’ and a significant fraction of the applied

agrochemical quickly reaches subsoil layers where these attenuation processes are generally

less effective.


This chapter presents different concepts underlying existing preferential flow models and

presents descriptions of preferential flow routines that have been incorporated into two

existing models used in EU pesticide regulation, namely PELMO and PEARL. It also briefly

identifies and discusses improvements to the current description of macropore flow in the

MACRO model that were also made during the APECOP project.

2.1. Concepts for modelling preferential flow

Although some aspects of preferential flow are still imperfectly understood, great advances

have been made in the last two decades in the understanding of the mechanisms responsible

for generating and maintaining preferential flow and transport in soils. This improved process

understanding has also led to the development of a wide range of models. Comprehensive

reviews of existing preferential flow and transport models have recently been published

(Jarvis, 1998; Feyen et al., 1998), and so this is not repeated here. Instead, emphasis is placed

on approaches which are suitable to estimate pesticide leaching in the field under transient

conditions and at time scales appropriate for management purposes such as pesticide regulation

(i.e. of the order of seasons and years).

Little is known about the relative significance of finger flow, heterogeneous flow and

macropore flow for leaching of agrochemicals. However, intuitively, macropore flow ought to

be the most important process, for two main reasons: macropores are ubiquitous, and the

transport volume (often fractions of one percent of the soil volume) is appreciably smaller,

which should give shorter transit times and minimal adsorption interaction with the matrix.

Thus, in a numerical simulation study based on field experiments, the transit time for a non-

adsorbed chemical through a sandy vadose zone was predicted to be four times faster in the

presence of heterogeneous flow (Ju and Kung, 1997). From observed flow velocities (Beven

and Germann, 1982), we may expect macropore flow to decrease the transit time through the

unsaturated zone by up to two orders of magnitude (i.e. hours instead of years). On the other

hand, macropore flows are clearly highly intermittent, while heterogeneous flows in the

matrix are more or less continuous. It was eventually decided in APECOP to concentrate on

concepts for modelling macropore flow, since this was considered to be the most important of

the different preferential flow mechanisms: however, a few comments on suitable approaches


to modelling finger flow and heterogeneous flow can be made. Heterogeneous flow can be

modelled by solving Richards’ equation in a two-dimensional numerical framework,

explicitly representing the heterogeneity of soil hydraulic properties as spatially correlated

‘fields’ (Kung, 1990; Roth, 1995). Two-dimensional simulations using Richards’ equation can

also reproduce finger flow in homogeneous soils exhibiting strong hysteresis (Nieber, 1996).

Finger and heterogeneous flows can also be accounted for in empirical fashion in one-

dimensional models based on the convection-dispersion equation, by the inclusion of ‘mobile-

immobile’ regions in the soil matrix, although difficulties may arise in defining the immobile

water content in transient simulations where the total water content varies (Zurmühl and

Durner, 1996).

Models accounting for flow in soil macropores can be broadly classified as either pore-scale

models or lumped dual- or multi-region models, and as either stochastic or deterministic.

Pore-scale models and stochastic approaches are becoming increasingly popular in the

research arena, and are certain to be the subject of active research and development in the

coming years. However, at present and for the forseeable future, these approaches are difficult to

apply in regulatory practice for one or more reasons, including assumptions of steady-state flow

and/or the difficulty in obtaining the required parameter input. The concepts underlying these

models are also not compatible with existing models used in the regulatory process. Therefore, in

practice, deterministic dual-and multi-porosity models are the only feasible alternative. These

models take a macroscopic, continuum, approach by lumping individual preferential flow

pathways in the soil into two or more pore regions, each characterized by a porosity, a water

pressure (or water content), water flow rate and solute concentration. Mass exchange between

domains is treated as a source/sink term in the flow and transport equations. A large number of

models of this type have been developed in recent years (see reviews by Jarvis, 1998, Feyen et

al., 1998 and Jarvis, 2002), varying with regard to the degree of simplification and empiricism

involved in descriptions of water flow and solute transport (e.g. functional vs. mechanistic

models) and the way in which mass exchange between flow domains is represented (i.e.

approximate first-order mass transfer coefficients, explicit Fickian diffusion). Flow in

macropores has been modelled either as a function of macropore geometry and dimensions

(Chen and Wagenet, 1992) or in a lumped fashion, using either Richards’ equation (Gerke and

van Genuchten, 1993) or kinematic wave theory, assuming that flow is dominated by gravity

(Jarvis, 1994). Simpler approaches neglect flow in the macropores per se and assume that all


water and solute entering the macropores at the soil surface is immediately transferred to the soil

matrix at a depth of ‘internal catchment’ within the soil profile (e.g. Hendriks et al. 1999).

In the following sections, each of the FOCUS regulatory models included in APECOP is briefly

described and the macropore flow routines incorporated in the models are presented.

2.2 Preferential flow in MACRO

MACRO is the most widely-used example of the type of preferential flow and transport

model that is usually termed ‘dual-permeability’ . The soil pore system is divided into two

parts, one part with a high flow capacity and low storage capacity (macropores) and the

remainder with a low flow capacity and a high storage capacity (micropores). The boundary

between the pore regions is defined by a fixed water tension, and corresponding water content

and hydraulic conductivity. Classical continuum equations are used to model flow and

transport in the micropores (Richards equation and the convection-dispersion equation) while

flow in the macropores is calculated using the kinematic wave equation (Germann, 1985),

assuming gravity-dominated flow (i.e. neglecting capillarity). Transport in macropores is

calculated neglecting dispersion-diffusion, but accounting for adsorption by one parameter

that partitions the sorption constant between the two flow regions. The concentration of

pesticide entering the macropores at the soil surface is calculated using the mixing depth

concept outlined above. Mass exchange between the two pore regions is calculated using

approximate first-order equations based on an effective diffusion pathlength, accounting for

both convection and diffusion. Internal catchment can be modelled by setting the

macroporosity and macropore conductivity to very small values at a given depth in the soil,

but this cannot be combined with macropores that are continuous throughout the profile, since

only one macropore domain is considered in MACRO.

A new version of MACRO (5.0) has been developed in the APECOP project that includes

several improvements:

1.) The numerical routines have been upgraded, by converting to implicit solution methods for

water and heat flow and solute transport, which in turn allows for a larger number (up to 200)

of much thinner layers in the profile.


2.) Van Genuchten (1980) hydraulic functions are used instead of Brooks and Corey (1964).

This was prompted by the change in numerical methods described above, but it also allows

more flexibility in matching measured soil hydraulic properties. However, it requires one

additional parameter, since the boundary between macropores and micropores is no longer

given automatically by the Brooks-Corey air entry pressure.

3.) The effects of tillage practices, and subsequent soil sealing, on soil structure, hydraulic

functions, and macropore flow, can be simulated with a new physico-empirical approach.

The effects of tillage on the hydraulic properties are expressed through van Genuchten’ s α

parameter, while empirical relationships are used to track changes in parameters related to

soil structure (diffusion pathlength or effective aggregate size, kinematic exponent describing

macropore tortuosity), as a function of cumulative rainfall following a tillage event.

4.) Kinetic sorption has been incorporated into the model, following the ‘two-site’ model

approach described by Altfelder et al. (2000). This is applied only to the micropore region.

The updated version of MACRO (v5.0) requires six parameters in addition to those needed for

a chromatographic model based on Richards equation and the convection-dispersion equation:

the mixing depth, two additional hydraulic properties defining the boundary between

macropores and micropores (water tension and hydraulic conductivity), the kinematic

exponent, the macroporosity and the fraction of sorption sites equilibrating with macropore

water. The new version of the model is described in full detail in Larsbo and Jarvis (2003).

One important reason why macropore flow models have not been widely adopted in exposure

and risk assessments for agrochemical leaching is the difficulty of parameterisation.

Therefore, an automatic parameter calibration routine has been incorporated into the shell

program for version 5.0 of MACRO. This inverse capability is supplied by the SUFI

methodology described by Abbaspour et al. (1997). SUFI is a Bayesian global search

algorithm that is considered well-suited to complex simulation models such as MACRO, since

it minimises the risk of falling into local optima in the n-dimensional parameter space. The

user first defines a prior uncertainty range or domain for each parameter to be estimated. This

range is divided into a number of equal-size ‘strata’ and the first moment of each stratum

defines the parameter estimate. All combinations of parameters are then run, and the value of

a user-defined objective function (i.e. root mean square error or model efficiency, Loague and

Green, 1991) is calculated for each simulation by comparing to measured data. Based on a

critical tolerance for the objective function, the parameter ranges are refined (i.e. narrowed)


for subsequent iterations by rejecting parameter values that perform badly. This kind of

optimisation technique would have been too time-consuming for routine applications with

earlier versions of the model, but the enhanced speed of execution of version 5.0 makes such

methods more practicable.

2.3 Preferential flow in PELMO

PELMO is a deterministic ‘functional’ model based on capacity-approach to describe water

flow and the convection-dispersion equation for solute transport. The official FOCUSPELMO

does not consider any preferential flow. As a functional model like PELMO is based on a

simplified description for the movement of water in the soil, there is little sense in

incorporating a mechanistic description of preferential flow. Consequently a simple functional

approach has been adopted as follows:

Preferential flow in PELMO has been estimated based on a simplified ‘two-parameter’ linear

response model with a threshold :

( ) ( )( ) cccmicma

cmima

IR;IIRf1I,IRfI

IR;RI,0I

>+−−=−=≤==

(2.1)

where Ima and Imi are the amounts of water routed into macropores and matrix respectively, Ic is

the threshold daily rainfall amount which generates infiltration into macropores, R is the daily

rainfall and f is the fraction of the excess rainfall which is routed into macropores.

The concentration of pesticide entering the macropores at the soil surface cma are calculated

using the mixing depth concept, whereby incoming rainfall is assumed to mix perfectly with

the resident water in a shallow surface layer of soil according to following equation::

( )( )1nmafmidma

d1 ckzRc

zz

Q −γ+θ+=

∆(2.2)

where ∆z is the thickness of the top numerical layer, Q1 is the amount of pesticide stored in

the top numerical layer at the previous time step (including dissolved and adsorbed phases), R

is the rainfall amount during the time step, θmi is the matrix water content, γ is the bulk


density (reduced by the fraction of solids assumed to equlibrate with macropore liquid), n is

the Freundlich exponent and kf is the Freundlich sorption coefficient.

The flux of pesticide into the macropores is given by cma multplied by the infiltration rate into

macropores Ima, while the amount of pesticide added to, or extracted from, the matrix in the

given time step is given by :

mamar IcCR − (2.3)

A fixed number is defined for the depth of the macropores. At that soil depth percolate is

distributed in the soil matrix system again independent of the actual soil moisture conditions.

2.3 Preferential flow in PEARL

The PEARL model (Tiktak et al., 2000; Leistra et al., 2001) describes the fate and transport of

pesticides in soil and uses the SWAP model (Van Dam, 2000) as the submodel for soil water

flow. Preferential water flow is the driving force of preferential flow of solutes. The concepts

for preferential water flow are described first, and thereafter the concepts for preferential

solute flow.

SWAP is based on Richard’ s equation and has been modified to account for macropore flow

by introducing an adapted version of the FLOCR model (Hendriks et al., 1999). Two classes

of macropore are distinguished with respect to pore continuity. One domain is continuous

throughout the profile (i.e. the main bypass domain), and one domain represents macropores

ending at different depths in the profile, resulting in ‘internal catchments’ (i.e. the internal

catchment domain). Figure 2.1 shows a conceptual visualisation of these two classes of

macropores. As shown in this figure, the volume of macropores in the main bypass domain

consists of a network of interconnected macropores (e.g. structural and shrinkage cracks). It is

constant with depth up to the depth where the internal catchment domain stops; thereafter the

volume of pores in the main bypass domain decreases linearly with depth. The volume of the

internal catchment consists of macropores that are not interconnected and that end at different

depths. The decline of the number of internal catchment macropores is described by a power

law function.

Additionally, two types of macropore are included in the model to describe the dynamics of

the macropore volume resulting from swelling and shrinking: a permanent static macropore

volume independent of the soil moisture status and dynamic shrinkage cracks whose volume


depends on the shrinkage characteristic and the current soil moisture content. SWAP

simulates the swelling and shrinking dynamics via a simplified procedure: the soil level

remains fixed and swelling and shrinking influences only the pore volumes. Figure 2.2

visualises the permanent/static and the dynamic macropore volumes.

Water enters the macropores at the soil surface either as rain falling directly into the

macropores or as ‘runoff’ if the rainfall rate exceeds the infiltration capacity of the matrix.

Water flowing into the macropores accumulates at the bottom, while uptake into the matrix

takes place only in the saturated part of the macropore.

In FOCUSPEARL version 1.1.1 solute flow is based on the convection-dispersion equation.

This is also the case in the new model, but transport in the macropores is added. Surface

applied pesticides are introduced into the macropores using the mixing-cell concept developed

by Steenhuis and Parlange (1980). This concept has also been used in MACRO (Jarvis, 1994).

It is assumed that the two classes of macropore (main bypass and internal catchment) each

have a uniform pesticide concentration (assuming perfect mixing and ignoring adsorption and

transformation). Exchange of pesticide between macropores and the matrix is calculated as

the product of the water uptake rate and the pesticide concentration in the corresponding

domain (i.e. convective transport only). Thus the solute behaviour in the macropores is

described in a strongly simplified way (the only parameter being the thickness of the mixing-

cell layer).

Only a beta version of the new PEARL version including the above concepts could be

finalized within the time frame of the APECOP project. Thus only very limited time was

available for testing (as will be described later).


Figure 2.1. Conceptual model of the soil pore system used in the SWAP module describing preferential waterflow in cracking clay soils. The left part is the main bypass domain and the right part is the internal catchmentdomain. The dark colour represents water stored in the two domains.

Figure 2.2. Conceptual model of the soil pore system used in the SWAP module describing preferential waterflow in cracking clay soils. The left part is the main bypass domain and the right part is the internal catchmentdomain. The dark area shows the volume of the soil matrix in unswollen status; the grey area shows the increaseof the volume of the soil matrix due to swelling (and thus the volume of dynamic macropores); the white area isthe volume of permanent/static macropores.


References

Abbaspour, K.C., van Genuchten, M.T., Schulin, R., Schläppi, E., 1997. A sequential uncertainty domain inverseprocedure for estimating subsurface flow and transport parameters. Water Resources Research. 33: 1879-1892.

Altfelder, S., Streck, T., Richter, J. 2000. Nonsingular sorption of organic compounds in soil: the role of slowkinetics. Journal of Environmental Quality, 29: 917-925.

Beven, K., Germann, P., 1982. Macropores and water flow in soils. Water Resources Research, 18: 1311-1325.Brooks, R.H., Corey, A.T. 1964. Hydraulic properties of porous media. Hydrology Paper no.3, Colorado State

University, Fort Collins, Colorado, 27pp.Chen, C. and Wagenet, R.J. 1992. Simulation of water and chemicals in macropore soils. 1. Representation of

the equivalent macropore influence and its effect on soilwater flow. Journal of Hydrology, 130: 105-126.Feyen, J., Jacques, D., Timmerman, A. and Vanderborght, J. 1998. Modelling water flow and solute transport in

heterogeneous soils: a review of recent approaches. J. Agric. Eng. Res., 70: 231-256.Gerke, H.H. and M.T. van Genuchten, M.T. 1993. A dual-porosity model for simulating the preferential

movement of water and solutes in structured porous media. Water Resour. Res., 29, 305-319.Germann, P. 1985. Kinematic wave approach to infiltration and drainage into and from soil macropores.

Transactions of the ASAE, 28: 745-749.Hendrickx, J.M.H., Dekker, L.W., Boersma, O.H. 1993. Unstable wetting fronts in water repellent field soils

Journal of Environmental Quality, 22: 109-118.Hendriks, R.F.A., Oostindie, K., Hamminga, P. 1999. Simulation of bromide tracer and nitrogen transport in a

cracked clay soil with the FLOCR/ANIMO model combination. Journal of Hydrology, 215: 94-115.Hillel, D. 1987. Unstable flow in layered soils. Hydrol. Proc., 1, 143-147.Jarvis, N.J. 1994. The MACRO model (Version 3.1). Technical description and sample simulations. Reports and

Dissertations, 19, Dept. Soil Sci., Swedish Univ. Agric. Sci., Uppsala, Sweden, 51 pp.Jarvis, N.J. 1998. Modelling the impact of preferential flow on non-point source pollution. In H.M. Selim and L.

Ma (eds.), Physical nonequilibrium in soils: modeling and application, Ann Arbor Press, 1998, pp.195-221.Jarvis, N.J. 2002. Macropore and preferential flow. In: The Encyclopedia of Agrochemicals (ed. J. Plimmer),

vol.3, 1005-1013, J.Wiley & Sons, Inc., New York.Jarvis, N.J., 1994. The MACRO model (Version 3.1) - Technical description and sample simulations. Reports

and Dissertations, 19, Dept. Soil Sciences, Swedish Univ. Agric. Sciences, Uppsala, Sweden, 51 pp.Ju, S.-H. and Kung, K.-J.S. 1997. Impact of funnel flow on contaminant transport in sandy soils: numerical

simulation. Soil Science Society of America Journal, 61: 409-415.Kung, K-J.S. 1990. Preferential flow in a sandy vadose zone. 2. Mechanism and implications. Geoderma, 46: 59-

71.Larsbo, M., and Jarvis, N.J. 2003. MACRO5.0. Process descriptions and numerical solution procedures. Emergo

2003:5. Studies in the Biophysical Environment. Department of Soil Sciences, SLU, Uppala, Sweden.Leistra, M, A.M.A van der Linden, J.JT.I. Boesten, A .Tiktak & F. van den Berg (2001). PEARL model for

pesticide behaviour and emissions in soil-plant systems: description of the processes in FOCUSPEARLversion 1.1.1. Alterra-rapport 013, Alterra Wageningen. RIVM Report 711401009; RIVM Bilthoven.

Loague, K.M., Green, R.E. 1991. Statistical and graphical methods for evaluating solute transport models:overview and application. Journal of Contaminant Hydrology, 7: 51-73.

Nieber, J.L. 1996. Modeling finger development and persistence in initially dry porous media. Geoderma, 70,207-229.

Roth, K. 1995. Steady-state flow in an unsaturated, two-dimensional, macroscopically homogeneous, Miller-similar medium. Water Resources Research, 31: 2127-2140.

Steenhuis, T.S., Parlange, J.Y. 1980. Closed form solution for pesticide runoff in runoff water. Trans. ASAE 23:615-620.

Tiktak, A., F. van den Berg, J.J.T.I. Boesten, D. van Kraalingen,M. Leistra & A.M.A. van der Linden (2000).Manual of FOCUSPEARL version 1.1.1. RIVM Report 711401008, Alterra report 28, RIVM, Bilthoven, 144pp.

Van Dam, J. 2000. Field-scale water flow and solute transport : SWAP model concepts, parameter estimationand case studies. PhD thesis, Wageningen University, The Netherlands.


Van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturatedsoils. Soil Science Society of America Journal, 44: 892-898.

Zurmühl, T. and Durner, W. 1996. Modeling transient water and solute transport in a biporous soil. WaterResources Research, 32: 819-829.


3. Process studies of volatilisation from soil and canopies

Most of the PEC models in Europe are not able to predict the concentration of pesticides in

air. To remedy this problem a work package was executed to include or improve the

description of the volatilisation process in the PEARL, PELMO and MACRO models.

Experiments on the volatilisation behaviour of pesticides were performed at different spatial

scales to assess the effects of soil moisture and other environmental factors on volatilisation.

The experiments ranged from laboratory studies (3.1), via semi-field experiments (3.2) to

field studies (3.3). The experimental volatilisation studies were particularly dependent on

comprehensive, high-resolution measuring systems and data acquisition systems.

3.1 Process studies of volatilisation from soil

The laboratory photovolatility chamber (Figure 3.1) allows for simultaneous measurement of

volatilisation and photodegradation of 14C-labeled pesticides under controlled constant, but

variable climatic conditions (Kromer, 2001).

ML

MLU

PFM

P

GM

CIW

D2 D1

V

AO DF AF

AO

FM

P

PUF

S/PCPWR

KR

DQ

AIWB DF AF

RH/TPUF

S/PCPWR

GD

P

RH/T

IR

AF

RH/T

V

RH/T

APPARATUS 1

APPARATUS 2

data acquisition

RH

TIR

Ozone

AIR ANALYSIS SUN TEST AIR CONDITIONING

GM

Figure 3.1. Schematic of the photovolatility chamber (air conditioning unit and sun test apparatus are installed ina climate chamber). AF = activated charcoal filter, AI = air inlet, AO = air outlet, CIW = cooled intensive washbottle, CP = ceramic plate, D1 = first drying stage (silica gel), D2 = second drying stage (phosphorus pentoxide),DF = dust filter, FM = flow meter, DQ = double-walled quartz vessel with water jacket, GD = glass dome, GM =gas meter, IR = infrared sensor, KR = cryostat, ML = mercury lamp for ozone generation, MLU = ozoneanalyser, P = metal bellows pump, ∆P = moisture tension, PUF = polyurethane foam plugs, RH/T = control ofrelative humidity and air temperature, S/P = soil/plant container, V = control valve, WB = wash bottle, WR =water reservoir.

The apparatus had been previously established for detailed studies of direct and indirect

photodegradation processes of pesticides on surfaces (Apparatus 1) but had to be modified to

suit the particular needs of process studies on volatilisation (Apparatus 2). The set-up of the


chamber was improved to characterize the influence of soil moisture, soil temperature, and

evaporation on volatilisation behaviour from soil.

The photovolatility chamber and the air conditioning unit are installed in an environmental

chamber to obtain constant preconditioned climatic parameters. The air passing through the

chamber is purified via a filter system consisting of activated charcoal, dust filters and two

water filled wash-bottles to ensure clean and water-saturated air to prevent drying of the soil

surface and reduce photochemical reactions. Significant climatic parameters, including ozone

concentration, air humidity, air temperature, soil moisture and soil surface temperature, are

monitored continuously using various sensors (Figure 3.1).

In these apparati the glass domes are mounted on a base plate to allow the use of containers

with different surfaces, e.g. glass, teflon, soil dust, soil layers and soil/plant systems.

Containers of adjustable height can be used for soil bodies of different thickness and provide a

surface area of approx. 0.01 m2. The chamber volume is approx. 0.34 L. An air analysis

system was connected to record gaseous losses. 14C-labeled organic compounds in the sample

air are collected in the total-volume sampler consisting of a glass cartridge filled with three

polyurethane foam plugs. In addition, 14CO2 arising from the complete mineralisation of the

test compound is measured by a medium-volume sampler to gain complete radioactivity and

mass balances (Stork et al., 1997).

To characterize the influence of soil moisture and temperature on volatilisation, three process

studies with EC-formulated 14C-parathion-methyl applied to gleyic cambisol were performed

under various moisture conditions at the soil top layer, ranging from 1.18 to 8.67%weight (3-

6 days each) and at different temperatures (20 & 30 °C).


0 20 40 60 80 100 120 1400

5

10

15

20

25

30

35

Soil moisture (0 - 7 mm depth): 8.67 %weight

3.41 %weight

1.18 %weight

Cum

ulat

. vol

atili

zatio

n [%

AR

]

Time after application [h]

Figure 3. 2. Cumulative volatilised radioactivity after soil surface application of 14C-parathion-methyl to gleyiccambisol deter-mined in PUF plugs. Net applied radioactivity (AR) = 100%.

An unambiguous dependence of volatilisation on the water content in the top layer of the soil

was established within the three soil moisture studies (Figure 3.2), exemplified by a

cumulative volatilisation of 12.8% under medium conditions and 32.9% under moist

conditions after 6 days. Only a slight volatilisation of 2.4% AR in the experiment performed

under dry conditions was detected after 4 days. The pronounced enhancement of

volatilisation, e.g. changing from dry to moist conditions led to an increase by a factor of

approx. 14, allowed for an unambiguous quantification of the influence of soil moisture,

without being affected by varying environmental conditions. Thus, the constant

environmental scenario during the laboratory studies, especially soil moisture and soil

temperature, enabled the definite correlation between volatilisation and water content.

Considering the clear correlation between soil moisture and cumulative volatilisation

measured in the process studies evaporation of water does not have a significant influence on

the volatilisation behaviour.

The strong decrease of volatilisation rates after the first hours of the experiment corresponds

to the kinetics as observed in an interlaboratory comparison of volatilisation assessment

methods (Walter et al., 1996) and reveals a “phasing out” of volatilisation between the second

and third day after application.

In comparison with the studies on the influence of soil moisture, which were performed at a

soil temperature of 20 °C, enhanced amounts of metabolites were detected after raising the

temperature to 30 °C. The increase of mineralisation to 14CO2 compared to the soil moisture


studies illustrates the tendency towards enhanced degradation rates in soil after increasing the

temperature (Domsch, 1992).

3.2 Wind tunnel experiments of volatilisation from soil and canopy

Wind-tunnel experiments measure pesticide volatilisation and bio-mineralisation directly

under field-like conditions, combining the advantages of laboratory facilities, e.g. use of

radioisotopes, and field studies.

3.2.1 The Wind tunnel

As an extension of the lysimeter concept (Führ et al., 1998), a glass wind tunnel was set up

above a lysimeter with a soil surface area of 0.5 m2 to measure the gaseous emissions of the

applied pesticide mixture (Figure 3.3). A detailed description of the system was given by

Stork (1995) and Linnemann (2002).

The glass wind tunnel is 1.1 m high, 0.7 m wide and 2.7 m long. A single blower presses air

into the wind tunnel after intensive cleaning in various filter stages. Air filters and subsequent

sieves ensure a uniform air stream through the glass tunnel. The top of the wind tunnel can be

adjusted in height and thus be adapted to the level of growing plants. Constant wind velocities

from 0.3 to 3.5 m s-1 can be achieved with a minimum air volume. Realistic environmental

conditions are simulated and monitored inside this wind tunnel by a continuous, automatic

adjustment of e.g. the air temperature and wind speed to the outdoor situation. Due to the UV-

transparent glass design, sufficient light intensity and light quality is ensured so that

experiments with plants can be performed. The application of radioactive spray mixtures of

pesticides, following the good agricultural practice, was performed using a semiautomatic

sprayer.

The organic 14C-labeled air constituents were sampled using a high-volume sampler (HVS)

equipped with an adhesive-free glass fibre filter to trap particulate matter followed by three

polyurethane foam plugs. A medium-volume sampler (MVS) was used for measuring 14CO2

arising as the end product of biomineralization of 14C-labeled compounds in wind tunnel and

photovolatility chamber experiments (3.1).


Glass Exhaust / Air Analytics

solar radiation air temperature

CV2

air inletgas meter

CV1B

transitionadjustable topdeflector

air outlet

probestainlesssteel net

gas streammixer

gas meter withimpulse output

HVS

CV3

R

R

C

P

PFFFAFFF P

steel-plate

T8 P1H3

T9

A1

P3 Py2

T3

T5

I1

A2

T1H1

TDR

P2

T7

Py1

XAD1

XAD2

MVS

IS

T2H2

T6

I2

T4

P

bellows

I3

TDR

leachate

TDT

atmospheric pressure

123

PUF plugs

gas meter withimpulse output

control plot

lysimeter (0.5 m2)

air humidity

GLASS EXHAUST / AIR ANALYTICS GLASS WIND TUNNEL BLOWER / AIR CONDITIONING UNIT

Figure 3.3. Schematic of the wind tunnel for measuring volatilisation of pesticides from the soil surface underfield-like conditions. A1-2 = anemometer, AF = activated charcoal filter, B = brine tank, C = cooler, CV1-3 =converter, FF = fine filter, H1-3 = hygrometer, I1-3 = gas volume (sensors), IS = isokinetic sensor, P =pump/blower, P1-3 = pressure (sensors), PF = prefilter, PY1-2 = pyranometer, R = refrigeration, PUF =polyurethane foam, T1-9 = thermo sensors, TDR = time domain reflectometry, TDT = thermal desorption tubes,XAD = adsorbing resin (Amberlite XAD-7).

3.2.2 Experiments from bare soil

Volatilisation rates of four pesticides (14C-labeled parathion-methyl, fenpropimorph,

terbuthylazine and non-labeled chlorpyrifos) were determined in a 13-day wind-tunnel

experiment after simultaneous soil surface application to gleyic cambisol. (Table 3.1)

Table 3.1 : Wind-tunnel study on volatilisation from bare soil: Application details. (Applied water 720L ha)

Pesticide Parathion-methyl Fenpropimorph Terbuthylazine Chlorpyrifos

Specificradioactivity

1.94 MBq mg-1 a.i.,radiochemical purity:

> 99.0%

6.89 MBq mg-1 a.i.,radiochemical purity: >

99.0%

3.17 MBq mg-1 a.i.,radiochemical purity:

98.9%-

Net appliedradioactivity

12.06 MBq 7.91 MBq 102.55 MBq -

Net applied a.i.124.30 g ha-1

EC 40% a.i.†

651.80 g ha-1, (14C-labeled: 3.66%)

EC 80% a.i.

647.00 g ha-1 SC 50%a.i.†

696.53 g ha-1

EC480 g a.i. L-1

† applied compound was quantitatively 14C-labeled

The lid of the wind tunnel was adjusted to a height of 30 cm and the wind velocity was kept

constant at 1 m s-1 in a height of 20 cm. Irrigation (8 mm) was given on Day 8 after

application. At the end of the experiment, soil layers up to 10 cm were completely removed

and soil cores of deeper layers were taken, and leachate was also pumped off. Subsequently,

sample preparation and analysis of water samples and soil samples were performed. Soil


moisture was determined by gravimetric analysis and continuous TDR measurements in

different depth while the experiment.

The wind-tunnel study after soil surface application to gleyic cambisol revealed a

comprehensive picture of the fate of the applied pesticides, illustrated by 14C-recoveries

ranging from 94.4 to 103.9%. The observed order of cumulative volatilisation of 14C-labeled

compounds (parathionmethyl > terbuthylazine > fenpropimorph) deviated markedly from

previous studies on orthic luvisol (Figure 3.4). These findings demonstrated that comparisons

of volatilisation studies are to be handled with care due to the strong influence of micro-

climatic conditions and soil conditions.

0 2 4 6 8 10 120

5

10

15

20

25

Parathion-methyl Terbuthylazine Fenpropimorph Fenpropimorph acid Desethyl-terbuthylazine

Cum

ulat

. vol

atili

zatio

n [%

AR

]

Time after application [d]

Figure 3.4. Cumulative volatilisation of 14C-labeled pesticides (parathion-methyl, terbuthylazine,fenpropimorph) and metabolites (fenpropimorph acid, desethyl-terbuthylazine) after soil surface application togleyic cambisol determined in polyurethane foam (PUF) plugs. Net applied radio-activity (AR) = 100%.

Cumulative volatilisation of 4.3% AR (fenpropimorph) and 2.1% AR (fenpropimorph acid)

within 13 days was observed during the wind-tunnel study. Volatilisation rates reached a

maximum at 24 hours after application under moist conditions. During this time the loss

kinetics were primarily dictated by volatilisation of the pesticide from the liquid phase (3.1).

The results of the process studies and the wind tunnel suggest that Henry’ s law constant is the

driving factor in volatilisation from moist bare soil of most pesticides e.g. chlorpyrifos.

3.2.3 Experiments from plants

Detailed descriptions of the wind-tunnel studies and the underlying experimental data used for

the evaluation of the PEARL plant volatilisation module and the PELMO approach were

given by Stork et al. (1998) and Ophoff et al. (1999). One wind tunnel study from plants was

conducted in the framework of the APECOP project. In this study volatilisation rates of three


pesticides (14C-labeled parathion-methyl, non-labelled fenpropimorph and quinoxyfen) were

determined in an 10 days experiment after simultaneous spray application to winter wheat

sowed on orthic luvisol (Wolters 2003)

0 2 4 6 8 100

5

10

15

20

25

30

Parathion-methyl Quinoxyfen Fenpropimorph

Cum

ul. v

olat

il. [%

app

l. am

ount

]


Figure 3.5. Volatilisation of 14C-parathion-methyl, quinoxyfen, and fenpropimorph after application to winterwheat. Cumulative volatilised amount. Net applied active ingredients = 100%. 14C-labelled parathion-methyl wasquantitated by LSC, non-labelled quinoxyfen and fenpropimorph were determined by GC-MSD .

In total, 32.5% AR were volatilised by the end of the wind-tunnel study, corresponding to

29.2% parathion-methyl and 3.3% of metabolite. About 25% AR already volatilised within

the first 24 h, followed by a sharp decrease (Fig. 3-5), indicating other processes like foliar

uptake counteracting the volatilisation process. The major part of volatilisation took place

immediately after application due to a weak adsorption/bonding of the pesticide on the leaf

surface. Decreasing volatilisation rates were observed following the first 24 h, illustrating the

long-term release of protected or adsorbed pesticide residues. Increasing air temperatures and

irrigation caused only temporary increases of volatilisation rates. Irrigation generally

increases volatilisation from soil and plant surfaces and this effect was enhanced by an

additional increase of temperature on Day 7. In total, 6.0% of the applied fenpropimorph

volatilised in the wind tunnel. Cumulative volatilisation of quinoxyfen over the course of the

experiment was about 15.0% of the net applied amount apparently indicating a higher

volatilisation tendency of quinoxyfen in comparison with fenpropimorph despite to much

lower vapour pressure (factor 100). From this it follows that there are more factors

influencing the volatilisation from plant surfaces than just the vapour pressure (e.g. crop

types, meteorological conditions, formulation, etc.).


3.2 Field experiments of volatilisation from soil and canopy

3.2.1 Introduction

The flux of a pesticides into air is influenced by many factors such as the physico-chemical

properties of the pesticide, characteristics of the treated surfaces (soil or plant),

meteorological conditions, application methods etc. (Linnemann & Wolters 2003 a, Leistra,

2002). The influence of some factors on volatilisation can be measured in chambers in the

laboratory and in micro-ecosystems. Examples are the penetration into the leaves and

(photochemical)- transformations (Linnemann & Wolters 2003 b). However, field

measurements provide the most realistic volatilisation rates for pesticides because the net

effect of all factors is represented, including plant/pesticide interactions. In addition, the

dynamics of the volatilisation rate are best reflected by field measurements, because of the

continuous and simultaneous change of the numerous parameters. A disadvantage of field

experiments may be the lack of continuous volatilisation data, especially during the night and

in periods with very low wind speed, during which micrometeorological methods do not

function well then. The fact that volatilisation under these conditions is also generally low,

weakens this disadvantage. Mass balances can be best obtained with 14C labelled pesticides in

laboratory studies or in micro-ecosystems like wind tunnels.

Simulations with computation models would be very helpful to interpret and explain the

results of the field experiments. Such models should describe both volatilisation from treated

surfaces (soil or canopy) and the competing processes like (photochemical) transformation,

leaching in soil and penetration into leaves. Such computation models are being developed

(van den Berg, 2003; Leistra, 2003) and comprehensive field data sets are necessary to

calibrate and validate these models.

There is a lack of well-documented field experiments, especially experiments on volatilisation

from plants. To remedy this lack of data, three field experiments were performed in the

framework of the APECOP project. In addition, the raw data from two recent field

experiments, were also made suitable for model validation studies.

3.2.2 Volatilisation experiments from fallow soil

Field experiment at Jülich, Germany (APECOP code: SJue 1)

In this field experiment the pesticides fenpropimorph, parathion-methyl and terbuthylazine

were sprayed as a mixture onto a bare soil (orthic luvisol, 15% clay, 1.1% C org, pH-KCl 7.2)


on a 1 ha field near Jülich-Merzenhausen, in May 1995 (Stork et al., 1996). Volatilisation was

measured with the Aerodynamic (AD) and Bowen-ratio (BR) methods, both

micrometeorological methods. Pesticide concentrations in air and meteorological parameters

were measured as required by each method at 13 times during the 15-day experimental period.

Residues in the soil profile were measured at four times. Rainfall and temperatures (air and

soil) were measured continuously. Temperature at the soil surface was measured with an

infrared meter. Change in moisture status in the top mm of the soil was recorded by

measuring the albedo of the soil surface, which gave better information concerning drying out

than soil moisture profiles measured at several times. The adsorption of the pesticides by soil

of the experimental field was recently measured by Forschungszentum Jülich to complete the

physico-chemical data of the pesticides for modelling purposes.

The experiment started in a warm and dry period, resulting in very dry soil surface (3%).

Because of very strong sorption of pesticides on air-dry soil, the initial volatilisation rates

declined very rapidly to very low rates (<0.01-0.05% of applied h-1). The rates increased

considerably (0.3-0.18% of applied h-1) with remoistening of soil by rainfall events. The

volatilisation rates measured with the two methods generally differed only slightly.

Field-measured volatilisation rates were generally about an order of magnitude higher than

those in a wind tunnel experiment that was performed simultaneously under climatic

conditions as measured in the field (Stork et al., 1996). The differences were mainly caused

by the different reactions of the upper millimeters of the soil surface in the wind tunnel to

radiation, drying conditions and by problems with the simulated rainfall intensities .

Field experiment in Bologna, Italy (APECOP code: SBol 2)

This experiment was carried out on a bare silty soil (19% clay, 1.5% C org, pH 7.8) at a field

near Bologna (Ferrari et al., 2003). In September 1999, a quasi circle with 25 m diameter was

sprayed with the pesticides malathion, ethoprophos and procymidone. The volatilisation was

measured with the theoretical profile shape (TPS) method. Pesticide concentrations in air and

meteorological parameters were measured as required by the method, almost continuously

during the 16-day experimental period. Residues in soil were measured at the end of the

experiment. Average daily air temperature ranged between 7.5 and 22 °C. Almost no rain fell

in the first 6 days. A diurnal rhythm was measured, with an increase in volatilisation rates in

the daytime. Total volatilisation was estimated to be 41, 23, and 19% of applied for

procymidone, malathion and ethoprophos, respectively over the 16-day experimental period.

Highest volatilisation were measured during the seventh to the tenth day after application


when rainfall increased soil moisture content. The great effect of soil moisture condition on

volatilisation rate corresponds with that observed in the experiment in Jülich.

3.2.3 Volatilisation experiments from canopies

Experiment in sugar beets at Jülich, Germany (APECOP code: PJue 2)

The experiment was carried out in a cooperative study by Forschungszentrum Jülich and

Alterra on a field near Merzenhausen (Smelt et al., 1997). The fungicide fenpropimorph (Vp

2.3 mPa) and the herbicide clopyralid (Vp 1.3 mPa) were sprayed on a 2 ha sugar beet crop at

the end of June. The volatilisation rates were measured with the AD and BR methods at ten

times during a six day period. Highest volatilisation rates for fenpropimorph (3% of dosage

h-1) were measured immediately after spraying. Volatilisation rates gradually declined to <

0.1% h-1 at the end of the 2nd day and remained very low (about 0.01% h-1) thereafter. The

residue on the sugar beet leaves gradually declined from 68% of the dosage (one hour after

appl.) to 13% (six days later). The pattern of decline of the leaf residue demonstrated that

besides volatilisation other processes considerably contributed to the dissipation. An almost

perfect match of volatilisation rates was measured for fenpropimorph in terms of rates and

kinetics between the field and a simultaneous wind tunnel experiment (Stork et al., 1997).

Volatilisation rates for clopyralid were very low (always < 0.1% of dosage h-1) and could not

be measured very accurately because of interfering metabolites and the complex analytical

method. This implies that the data for clopyralid are less suitable for model calibration. Much

of the systemic herbicide clopyralid and its metabolites penetrated into the sugar beet leaves

(> 80% of dosage) as measured in the simultaneous wind tunnel experiment. However, the

great difference in volatilisation rate between the two pesticides (with only a factor of two

difference in vapour pressure, but 1800 times higher water solubility for clopyralid than for

fenpropimorph) makes clopyralid an interesting compound for modelling processes at the

plant surface.

Experiment in a potato crop in The Netherlands (APECOP code: PAlt 3)

In June 2002 the volatilisation rates of the fungicide fenpropimorph and the insecticide

chlorpyrifos were measured after spraying on a well-developed potato crop on a 4 ha field in

the Wieringermeer Polder. Volatilisation rates were measured with the AD and BR methods

at nine times during a six day period. Volatilisation was also measured with the relaxed eddy

accumulation method (REA) and the plume dispersion method (PD) during the first 24 hours

after application. Pesticide concentrations in air and meteorological parameters were


measured as required by each method. Additional parameters for computation models, like

actual leaf temperatures (by IR method) were measured also. Leaf temperatures were

considerably higher than air temperature on sunny days. Residues on the leaves at two levels

in the canopy were measured at five times. Rinsability of the residues on leaves, using

solvents with decreasing polarity, was measured at four times.

On the day of application and the next day it was bright sunny weather with average air

temperatures around 20 °C. Volatilisation rates as a function of time are presented in figure 3.

6A. Volatilisation rate of fenpropimorph was only measurable on the day of spraying, next

rate was below the detection limit (< 0.001% of dosage h-1). The low volatilisation rates of

fenpropimorph and especially the very fast decline with time in this experiment, differ from

that measured in the experiment in sugar beets in Jülich. Other processes than volatilisation

formed the major routes of the very fast dissipation from the leaves, which was comparable

with the dissipation of chlorpyrifos (Figure 3.6B). The pesticide fluxes determined with the

four methods agreed fairly well under the conditions where comparison was possible. Very

low fluxes could be best measured with the equipment used for the AD and BR methods in

this experiment.

Time after application (h)

0 20 40 60 80 100 120 140 160

Res

idue

on

leav

es (%

of a

pplie

d)

0

20

40

60

80

100

120

chlorpyrifos fenpropimorph

B

Time after application (h)

0 20 40 60 80 100 120 140 160

Flux

(mg

m-2

h-1

)

0.01

0.1

1

10

chlorpyrifosfenpropimorph

A

Figure 3.6. Volatilisation rates (A) and decline of pesticide residues on potato leaves (B) for the experiment inWieringermeer.

Experiment in a wheat crop at Jülich, Germany (APECOP code: PJue 3)

In May 2002 a mixture of parathion-methyl, fenpropimorph and quinoxyfen was sprayed on a

4 ha winter wheat crop near Merzenhausen-Jülich. This experiment suffered from a number of

problems. Undissolved particles of the parathion formulation blocked nozzles several times

during application, which lead to an interrupted application period and a broad range of

deposit values in the crop. High concentrations in the upwind samples interfered with the


measurements of quinoxyfen. Furthermore, analysis of air samples indicated degradation of

parathion-methyl on the adsorbent (XAD-4). Due to the results on parathion-methyl and

quinoxyfen being afflicted with a high degree of uncertainty, these data have little value for

interpretation and model validation purposes. Heavy rainfall after application retarded the

measurements during the (very important) hours immediately after application. The first

measurements (18 hours after application) delivered a flux of 2.7-4.4% of applied h-1 for

fenpropimorph. Due to lower radiation (and temperatures) and reduced wind speed in the

evening, the flux of fenpropimorph declined rapidly to < 1% at 20.00 h. A detailed discussion

on plant residues, rinsability of the pesticides and meteorological conditions is presented by

Wolters (2003).

3.3.4 Prospects

The field experiments provide comprehensive data sets on volatilisation of pesticides, with

different physical and chemical properties, from soil and canopies for testing computation

modules being developed or improved in the near future (van den Berg, 2003; Leistra and

Wolters, 2003). All data of the four qualified field experiments have been made available to

the APECOP members (data in ASCII files on APECOP website).


References

Berg, F. van den. 2003. Concepts for modelling volatilisation from soil. Chapter 4.1, this issue.Domsch, K.H. 1992. Pestizide im Boden: Mikrobieller Abbau und Nebenwirkungen auf Mikroorganismen, p.

65-66, 1st edition, VCH, Weinheim, Germany.F. Ferrari, M. Trevisan, and E. Capri. Predicting and measuring environmental concentration of pesticides in air

after soil application. (submitted for publication)Führ, F., P. Burauel, M. Dust, W. Mittelstaedt, T. Pütz, G. Reinken, and A. Stork. 1998. Comprehensive tracer

on environmental behaviour of pesticides: The lysimeter concept. p. 1-20. In F. Führ, R.J. Hance, J.R.Plimmer, and J.O. Nelson (eds.) The lysimeter concept. ACS Symp. Ser. 699. Am. Chem. Soc., Washington,D.C., USA.

Kromer, T. 2001. Photovolatility chamber for simultaneous measurement of photodegradation and volatilizationof environmental chemicals on surfaces using the insecticide parathion-methyl. PhD thesis, University ofBonn, Germany.

Leistra, M. 2002. Volatilization of pesticides from plant surfaces. Notes in the framework of a literature study.Progress report July 2002. Alterra, Wageningen, The Netherlands.

Leistra M. & A. Wolters. 2003. Concepts for modelling volatilisation from plant canopies. Chapter 4.2, thisissue.

Linneman, V and Wolters, W., 2003 A. Process studies of volatilisation from soil. Chapter 4. this issueLinneman, V and Wolters, W. , 2003 B. Wind tunnel experiments of volatilisation from soil and canopy.

Chapter 4. this issueLinnemann, V. 2002. Transport of volatile hydrocarbons through an undisturbed soil core into the atmosphere

after contamination of the ground-water with the fuel additive methyl-tert-butyl ether (MTBE). PhD thesis,University of Bonn, Germany.

Smelt, J.H., R.A. Smidt, F. van den Berg, A.M. Matser, A. Stork and H. Ophoff. 1997.Volatilization of fenpropimorph and clopyralid after spraying onto a sugar beet crop. Report 136, DLO-Winand

Staring Centre Wageningen & Forschungszentrum Jülich. pp 45.Stork, A., K. Matzerath, H. Ophoff, F. Führ, (IRA/KFA Jülich), J. Smelt, F. van den Berg, R. Smidt (SC-DLO)

1996. Comparative behaviour of pesticides under field-like and field conditions- volatilization and soilresidues of fenpropimorph, paration-methyl and terbuthylazine In: The Environmental Fate of Xenobiotics,Proceedings of the X Symposium Pesticide Chemistry p.271-278. September 30 - October 2, 1996, Piacenza,Italy. Del Re, A.A.M., Capri, E., Evans, S.P. and Trevisan, M. (Eds.)

Stork, A., H. Ophof, J.H. Smelt, and F. Führ. 1997. Volatilisation of pesticides: measurements under simulatedfield conditions. In: ACS Symposium series 699, The Lysimeter Concept , Environmental Behaviour ofPesticides, Chapter IX, p 21, F. Führ, R.J. Hance, J.R. Plimmer & J.O. Nelson (Eds). American ChemicalSociety, Washington, DC.

Stork, A., R. Witte, and F. Führ. 1997. 14CO2 measurement in air: Literature review and a new sensitive method.Environ. Sci. Technol. 31:949-955.

Stork, A.. 1995. Wind tunnel for the measurement of gaseous losses of environmental chemicals from the soil-/plant system under practice-like conditions with direct air analytics using 14C-labelled chemicals. PhDthesis, University of Bonn, Germany.

Walter, U., M. Frost, G. Krasel, and W. Pestemer. 1996. Assessing volatilization of pesticides: A comparison of18 laboratory methods and a field method. Berichte aus der BBA 16, Federal Research Centre for Agricultureand Forestry (BBA), Braunschweig, Germany.

Wolters, A., 2003. Pesticide Volatilization from Soil and Plant Surfaces: Measurements at Different Scalesversus Model Predictions. PhD thesis, University of Technology Aachen, Germany, in press.


4. Modelling volatilisation processes

In the current EU risk assessment procedure under Council Directive 91/414, the PEARL,

PELMO and MACRO models are being used to calculate environmental concentrations of

pesticides for the registration procedure. To reduce the uncertainty in these concentrations,

one goal of the APECOP was to improve or include the description of the volatilisation

processes in these models. Based on process studies (Chapter 3 this issue) and literature

screening new volatilisation concepts were developed and existing modules have been

improved.

4.1 Concepts for modelling volatilisation from soil

4.1.1 Introduction

Volatilisation of pesticides from the soil after spraying of the soil surface represents an

important source of input of pesticides into the atmosphere. The most important factors that

affect volatilisation are the physico-chemical properties of a pesticide (vapour pressure,

sorption coefficient and water solubility), the atmospheric conditions (e.g. air temperature,

atmospheric stability) and the soil conditions (e.g. moisture and organic matter content).

Current models that include a description of volatilisation may estimate the volatilisation of

pesticides from soil that have been incorporated or injected into the soil reasonably well, but

they are not adequate to describe the volatilisation of soil surface applied pesticides.

4.1.2 Description of model concepts

The initial volatilisation rate after spraying on the soil surface can be estimated fairly well

using the effective vapour pressure of the pesticide, defined as the vapour pressure that is in

equilibrium with the concentration in the liquid phase and the mass of pesticide sorbed on the

soil surface (Woodrow et al., 1997). A similar empirical approach has also been used by Smit

et al. (1997) to estimate the cumulative volatilisation in the 21 days for surface applied

pesticides (eq. (4.1)). CV is application as a percentage of the dose and Fa is the fraction of the

compound calculated to be present in the air phase.


9a

9aa

e33.6F;0.0CV

e33.6F;)F100log(*6.119.71CV−

−

≤=

>+=(4.1)

Based on equilibrium partitioning, Fa is given by :

aw

aa

KHKHk

Fθ+

θ+

γ

θ= (4.2)

where θa and θw are the volumetric air and water contents, γ is the bulk density, k is the

effective sorption partitioning coefficient and KH is the dimensionless Henry’ s law constant.

KH is calculated from the molecular weight, vapour pressure and solubility following the

approach described by Smit et al (1997), accounting for the effects of air temperature on both

vapour pressure and solubility.

Another simple description of the volatilisation process is based on the assumption that a

laminar air layer exists through which the pesticide has to diffuse before it escapes into the

atmosphere. Using this concept, the volatilisation flux density is given by:

d)cc()T(D = J air0,g

aa,v

−⋅− (4.3)

with Jv,a , volatilisation flux density through the boundary air layer (kg m-2 s-1) ; Da (T), the

coefficient for diffusion in air (m2 s-1) at temperature T ; cg,0, concentration in the gas phase at

the soil surface (kg m-3) ; cair, the concentration in the turbulent air (kg m-3) ; d, the thickness

of laminar boundary air layer (m).

The shortcomings of this concept have been discussed by van den Berg et al. (1999).

Moreover, the thickness of this layer is assumed to be constant, so it does not depend on

meteorological conditions. The effect of these conditions can be taken into account using the

concept of the resistances to transport of substance from the surface to the atmosphere (Baker

et al., 1996; Wang et al., 1997; Asman, 1998). The volatilisation flux density is given by:


ba

air0,ga,v rr

)cc( = J +

−− (4.4)

in which ra, the aerodynamic resistance (s m-1); and rb, the boundary layer resistance (s m-1).

The aerodynamic resistance (ra) is the resistance to transport between the roughness length for

momentum z0m and the height of the internal boundary layer (zbl), into which the pesticide has

mixed. This height depends on the length of the sprayed field, the roughness length and the

stability conditions of the atmosphere (see van der Molen et al., 1990). Hence, the

aerodynamic resistance is given by:

*

m0h

blh

m0

bl

a u

Lz

Lz

zz

ln

rκ

ψ+

ψ−

= (4.5)

in which zbl, the height of internal boundary layer (m) ; z0m, the roughness length for

momentum (m) ; Ψh, the stability correction for heat and substance (dimensionless); L, the

Obukhov length (m) ; κ, the Karman constant (dimensionless) ; and u*, friction velocity (m

s-1).

The boundary resistance rb, the resistance to the transport between the source height (i.e. the

soil surface) and z = z0m, can be described by (Hicks et al., 1987):

3/2

b PrSc

u2

r

κ=

∗

(4.6)

in which Sc, Schmidt number (-) ; and Pr, the Prandtl number (-) .

The Prandtl number is mostly assumed to be constant, whereas the Schmidt number depends

on the diffusion coefficient of pesticide in air.

The concentration of the pesticide in the gas phase at the soil surface depends on the amount

in the soil system and the partitioning of pesticide between the soil phases. The sorption

coefficient is mostly assumed to be constant for all water contents in the soil. At low moisture


contents, however, the coefficient for sorption to soil increases and this results in lower

concentrations of the pesticide in the gas phase of the soil system. Thus, volatilisation of

pesticides from the soil surface decreases strongly when the soil surface dries out. A simple

approach to take this effect into account is to specify a maximum sorption coefficient for air-

dry soil and a moisture content below which the sorption coefficient increases. For water

contents in soil below this level, an exponential relationship may be used to describe the

effect of the moisture content on the sorption coefficient:

wmax,deff,d eKK ⋅α−⋅= w < wlow (4.7)

in which Kd,eff is the effective sorption coefficient (L kg-1) ; Kd,max , the maximum sorption

coefficient (L kg-1) ; Kd, the sorption coefficient under moist soil conditions (L kg-1) ; �� a

coefficient (-) ; w, the moisture content (kg kg-1) ; and wlow, the moisture content below

which sorption coefficient increases (kg kg-1).

The moisture content below which the sorption coefficient increases is set equal to the

moisture content at wilting point. Below wilting point, air humidity in the soil pores is no

longer saturated.

4.1.3 Prospects for further developments

At present, daily meteorological data are used to calculate the transport resistances, so the

diurnal pattern of volatilisation cannot be described. This limitation can be remedied by a

further model improvement to process hourly meteorological data.

As standard meteorological measurements at weather stations are taken above a grass surface,

a procedure has to be developed to translate weather data for a location with grass to a

location with bare soil. These data are needed to calculate the resistances to transport of

pesticide from the soil surface into the air.


4.2 Concepts for modelling volatilisation from canopies

4.2.1 Introduction

Volatilisation of pesticides from plant surfaces is one of the main pathways of their emission

to the environment. After volatilisation and transport of the pesticides in air, people can be

exposed to the vapour and there can be deposition (both wet and dry) in residential areas,

drinking water catchments, nature conservation areas, etc.

In earlier versions of models for the soil-plant system, the processes in the plant canopy were

described in a rather empirical way. It was desirable to describe the processes in a more

mechanistic and detailed way, accounting for the influence of the environmental factors. Only

then, a model can be used as a tool for explaining and predicting (on longer term) the rate and

extent of pesticide volatilisation from plant canopies in the field.

4.2.2 Description of processes and factors

In an extensive literature study, an inventory was made of the current knowledge on the

processes for pesticides in plant canopies (Leistra, 2002). This served as a basis for the

development of concepts and equations for modelling. Substantial effort was needed to obtain

the most reliable input parameters and relationships for modelling.

The distribution of pesticide deposit on the plant surface is dependent on application factors

such as dosage, formulation, spray volume and droplet size spectrum. The initial rate of

volatilisation of a pesticide tends to be proportional to its dosage. Common formulations like

emulsifiable concentrates and wettable powders show little effect on the rate of pesticide

volatilisation; special formulations have to be developed to reduce vapour losses. Some

experiments show that besides the main fraction of pesticide deposit which is well exposed to

the air flow etc., a smaller fraction is only poorly exposed (enclosed by plant parts; deeper in

the canopy, etc.).

Pesticide molecules in the deposit tend to escape into the air by molecular movement, which

is highly dependent on the temperature. The escape is counteracted by the molecular

interaction forces in the deposit. This results in a certain saturated vapour pressure of the


pesticide at the surface. The divergent values reported for the vapour pressure of a pesticide

should be investigated carefully to be able to select the most reliable value.

In the boundary layer approach, the resistance to the volatilisation of pesticides is described in

terms of an equivalent thickness of the air boundary layer. In this layer with laminar air flow,

the pesticide has to diffuse from the deposit surface to the turbulent air with fast removal of

substances. This vapour diffusion is considered to be the rate-limiting process. The thickness

of the air boundary layer may be expected to vary in space and time. Its value is influenced by

factors like wind-speed, atmospheric turbulence and surface roughness. It should be realised

that the geometry of a plant canopy is complex, so there may not be a single thickness of the

air boundary layer for all heights in the canopy.

Penetration of a pesticide into the plant leaves is a process competing with its volatilisation.

The rate of penetration can be rather high (e.g. of systemic pesticides), dependent on physico-

chemical properties, product formulation and moisture condition. Pesticide deposited on the

plants may be washed-off by rainfall, especially in the first few hours and days after

application. The pesticides are exposed to sunlight, so direct and indirect photochemical

transformation may occur. More details on the processes competing with volatilisation at the

plant surfaces have been given by Leistra (2002).

4.2.3 Description of model concepts

The following empirical relationship was derived by Smit et al. (1998) from literature data to

describe the cumulative volatilisation of pesticides during the first seven days after

application to plants (fully covering the soil) in the field and in climate chambers:

)VPlog(466.0528.1)CVlog( += for VP ≤ 10.3 mPa (4.8)

where CV is the cumulative volatilisation during seven days after application (% of dosage),

and VP, the vapour pressure (mPa).

The computation model (PEARL plant volatilisation module) simulates volatilisation of a

pesticide from plant surfaces on an hourly basis. The module is based on the boundary layer

concept (4.2.2). The saturated vapour concentration of the pesticide at the leaf surface is

calculated as a function of the temperature, from its vapour pressure at reference temperature

(20 oC). The potential rate of pesticide volatilisation from the leaf surfaces is calculated by:


( )lam

t,as,aapot,vol d

CCDJ

−= (4.9)

with Jvol,pot , the potential rate of pesticide volatilisation, (kg m-2 d-1) ; Da , the diffusion

coefficient of the pesticide in air (m2 d-1) ; Ca,s , the vapour concentration at the leaf surface

(kg m-3) ; Ca,t , the vapour concentration in the turbulent air (kg m-3) ; and dlam , the

equivalent thickness of the laminar air boundary layer (m).

The actual rate of volatilisation is obtained by accounting for the areic mass of pesticide

deposit on the plants. The rate of pesticide penetration into the leaves is described by the first-

order rate equation. Rainfall intensity and wash-off coefficient determine the rate of pesticide

wash-off from the leaves. The rate of phototransformation is dependent on solar irradiation

intensity.

The equation for the conservation of pesticide mass on the plant surface reads:

phwpenact,volp RRRJ

dt

dA−−−−= (4.10)

with Ap , the areic mass of pesticide on the plants (kg m-2) ; Jvol,act , the actual rate of

pesticide volatilisation (kg m-2 d-1) ; Rpen , the rate of penetration into the plants (kg m-2 d-1) ;

Rw , the rate of wash-off from the plant by rainfall (kg m-2 d-1) ; and Rph the rate of

photochemical transformation (kg m-2 d-1).

Both a well exposed and a poorly exposed fraction of the deposit can be distinguished; each

of them requires its own conservation equation.

The actual rate of pesticide volatilisation is described by taking the mass on the plants into

account:

Jvol,act = (Ap/Ap,ref)*Jvol,pot (4.11)

with Jvol,act , the actual rate of pesticide volatilisation (kg m-2 d-1) ; fmas a factor for the effect

of pesticide mass on the plants ; Ap the areic mass of pesticide on the plants, (kg m-2) ; Ap,ref ,

the reference areic mass of pesticide on the plants (1.0 10–4 kg m-2 (= 1 kg ha-1)).


The vapour pressure at the deposit/leaf surface is assumed to be saturated, dependent on the

temperature. The saturated vapour pressure of the pesticide at the prevailing temperature is

calculated by the Clausius-Clapeyron equation. The coefficient for diffusion of the pesticide

in air at the reference temperature is estimated according to the following equation:

75.1

refref,aa T

TDD

= (4.12)

with Da, the diffusion coefficient of pesticide in air (m2 d-1) ; Da,ref , the diffusion coefficient

in air at a reference temperature (m2 d-1).

The rate of pesticide penetration into the leaves is calculated by:

ppenpen AkR = (4.13)

with Rpen , the rate of pesticide penetration into the leaves (kg m-2 d-1) and kpen , the rate

coefficient of penetration (d-1). The rate of pesticide wash-off from the leaves by rainfall is

then dependent on rainfall intensity and a wash-off coefficient:

prww AWkR = (4.14)

with Rw , the rate of pesticide wash-off from the leaves (kg m-2 d-1) ; kw , the coefficient for

pesticide wash-off (mm-1) and Wr , the rainfall intensity (mm d-1).

The rate of pesticide transformation by solar irradiation is described by first-order kinetics:

pphph AkR = (4.15)

with Rph , the rate of phototransformation on the leaves, (kg m-2 d-1) and kph , the rate

coefficient of phototransformation, (d-1).

The rate coefficient kph is dependent on sunlight irradiation intensity:

ref,phref

actph k

II

k

= (4.16)


with Iact , the actual solar irradiation intensity (W m-2) ; Iref , the reference solar irradiation

intensity 500 W m-2 ; kph,ref , the rate coefficient of phototransformation at reference

irradiation intensity (d-1)

The coefficient kph,ref is one of the quantities to be calibrated in the computation on the basis

of the measurements or it has to be derived from other studies on the pesticide.Further details

on the model are given by Leistra & Wolters (2003).

4.2.4 Prospects

The ultimate aim of modelling is to explain and predict (as far as possible) pesticide

volatilisation from plant canopies under divergent field conditions. Detailed interpretation

(using a model) of wind tunnel experiments with radio-labelled pesticides should provide

essential input data for field modelling work. It is difficult to obtain the rate coefficients for

the various processes in the field, because use of radio-labelled compounds is not allowed

there. Detailed input data from the laboratory or estimation methods for such data are often

not available.

In a more advanced approach, two types of atmospheric resistance to pesticide volatilisation

are defined in the air above the emitting surface (van den Berg et al., 2003). The boundary

layer resistance rb applies to the range between source height and roughness length of

momentum z0m. The aerodynamic resistance ra controls the range between z0m and height of

the internal boundary layer above the field in which the pesticide has mixed by small-scale

turbulence. This approach may be usable for plant canopies, provided that specific values for

the coefficients can be obtained.

4.3 Volatilisation modelling with MACRO

It was decided within APECOP to include a simple volatilisation routine in the MACRO

model which even though it might not be considered sufficiently accurate for assessments of

volatilisation losses per se, would aim to improve the accuracy of model leaching predictions

for volatile substances. This section describes such a simple empirical routine that has been


introduced into the FOCUS version of MACRO. The empirical approach described by Smit et

al. (1997), (4.1.2, eq. (4.1) and (4.2))�was adopted for the improvement of MACRO.

The approach includes the Henry coefficient, which is a function of the temperature. The air

temperature at the time of application Tair is approximated from a sinusoidal variation of air

temperature at the soil surface given by following equation:

( )

−π+=

365121IRRDAY

2sinANNAMPANNTAVTair (4.17)

with Tair, the air temperature at the time of application (°C) ; ANNTAV, the annual average

temperature at the site (°C) ; ANNAMP, the amplitude in the annual temperature wave ;

IRRDAY, the Julian date of application (i.e. half the difference between July and January

temperatures).

The bulk density (GAMMA) of the soil is already an input parameter in the MACRO model.

The water content θw of the surface soil will of course vary with the time of application, and is

predicted by MACRO for the surface numerical compartment in each FOCUS simulation.

However, for several reasons, it was decided not to utilise these predicted water contents in

order to estimate volatilisation: firstly, it is the water content of the surface few millimetres of

the soil which will be critical, and this is likely to differ significantly from the average water

content of the surface compartment which is of the order of several centimetres thick in the

FOCUS parameterisations. Secondly, the hydraulic properties of the surface few millimetres

are likely to differ from those of the underlying soil, and this is again not reflected in the

parameterisation. Thirdly, the actual water content is likely to vary considerably during the 21

day period due to intermittent rainfall and evaporation. Instead, an estimated water content is

utilized based on the season of application and the soil water retention curve for the top

numerical layer. For summer applications, defined as the period from mid-June (i.e. Julian

day 165) to the end of August, a water content equivalent to the wilting point (i.e. a tension of

15,000 cm) is assumed, while for all other application days, a water content at a tension of

300 cm is assumed. The volumetric air content is simply given by the total porosity (TPORV)

minus the water content.

The use of a Freundlich isotherm in MACRO implies that the partitioning between solid,

liquid and air phases depends on the total concentration in soil. Thus, the effective sorption

coefficient k, depends on the solution concentration Ceff in the surface soil immediately after

spraying :


1nefff Ckk −= (4.18)

where k , is the effective sorption coefficient ; kf, the Freundlich coefficient (given by the

product of the organic carbon content and the koc value of the compound) ; n, the Freundlich

exponent.

The effective concentration Ceff is calculated assuming that the dose reaching the soil surface

(i.e. after interception) is uniformly mixed in the surface 1 cm of soil, such that :

( )1nefff

eff Ck01.0A

C −γ+θ= (4.19)

with A is the dose (mass m-2).

This equation is solved iteratively to yield the value of Ceff that satisfies the mass balance.

4.3.1. Application to the Jülich no-1 experiment

Tables 4.1 and 4.2 below show the input parameters used to calculate volatilisation for

Fenpropimorph, Terbuthylazine and Parathion-methyl in the Jülich no-1 experiment.

Table 4.1. MACRO input parameters for Jülich no-1 experimentParameter Value Source

Bulk density (g cm-3) 1.57 Jülich

Water content @150 m (%) 9.8 Jülich

Water content @300 cm (%) 26.8 Jülich

Saturated water content (%) 40.4 Jülich

Organic carbon content (%) 1.22 Jülich

Annual mean air temperature (oC) 11 Jarvis

Annual amplitude in air temperature (oC) 7 Jarvis


Table 4.2. MACRO input parameters for compound properties, and application information. Bold figuresindicate selected value, if more than one value was available.Parameter Fenpropimorph Terbuthylazine Parathion-methyl Source

Koc

(cm3g-1)

3569

798

310

277

243

230

Jülich

PETE

Solubility

(mg/L)

70

4.3

8.5 55 PETE

Perkow

Molecular

Weight

304 230 263 PETE

Vapour pressure

(Pa)

0.0023 0.00015 0.0002

0.002

PETE

Perkow

Application date 11/5 11/5 11/5 Jülich

Dose (kg ha-1) 1.44 1.18 0.604 Jülich

Table 4.3 shows a comparison of measured and calculated volatilisation losses. It shows that

the accumulated loss of Terbuthylazine after 21 days was accurately estimated by the simple

empirical method. Volatilisation of Fenpropimorph was somewhat overestimated, while that

of Parathion-methyl was underestimated by c. 40%. Given that the objective was to find a

simple method to correct the dose for leaching purposes, rather than to accurately predict

volatilisation itself, this level of accuracy is probably acceptable and sufficient, given that

leaching is roughly linearly dependent on dose.

Table 4.3. Comparison of measured and predicted volatilization lossesSite Compound Predicted % of

dose volatilised

after 21 days

volatilisation after 21 days

(% of dose)

estimated from flux

measurements

Jülich Fenpropimorph 15.7 6.0

Terbuthylazine 9.4 9.0

Parathion-methyl 14.9 25.0


4.4 Volatilisation modelling with PELMO

4.4.1 Volatilisation from soil

In the original FOCUSPELMO volatilisation is estimated in rather a simple way. Henry’s

constant is assumed to be independent on temperature. The volatilisation flux is estimated

based on the concentration in air in the top soil, the laminar layer and the diffusion coefficient

in air. The minimum time step is one day.

A new routine was implemented in PELMO that considers a temperature dependent Henry’s-

constant, H, which must be known for two different temperatures. PELMO will use both

values to extrapolate the constant for all other temperatures according to an exponential

model. It is a similar equation as the Q10-approach used to estimate biodegradation for

different temperatures. The model assumes constant increasing factors for Henry’s constant

for a given increase of air temperature. If for example Henry's constant is doubled from 20 °C

to 30 °C, PELMO will consider a increase by a factor of 4 for any temperature difference of

40 °C.

The current volatilisation routine of PELMO is based on a time step of one day. That can only

be a rough estimation as volatilisation is a fast process that is very much dependent on the

daily amplitude of air temperature and humidity. In order to describe the volatilisation on a

more reasonable way the minimum time step of 1 day in PELMO was reduced to 1 hour.

The soil compartments currently defined in FOCUSPELMO are of homogeneous thickness.

In order to simulate realistic dispersion, depths of 2.5 to 5 cm are recommended as

compartment size. However, shortly after spraying the pesticide is usually distributed within a

very small part of the top soil only and a direct mixing depth of 5 centimetres is an unrealistic

assumption that may have a drastic effect on the simulation results. In order to improve the

volatilisation routines a small surface layer of 1 mm was implemented where all pesticide is

stored directly after spraying. The volatilisation itself is estimated based on following

equation:

a

airairVol d

cDif = J

×− (4.20)


with Jvol , the volatilisation flux (kg m-2 d-1) ; Difair , the diffusion coefficient in air (m2 d-1) ;

Cair, the concentration of the pesticide in the soil air of the thin surface layer (kg m-3) ; and da,

the thickness of boundary layer (m).

The implementation of a small top layer where all volatilisation from soil is assumed to occur

generally leads to a significant increase of the volatilisation flux compared to the original

FOCUSPELMO (about a factor of 50 directly after spraying). Comparisons with experimental

data performed in the past based on the original FOCUSPELMO showed that the new

approach may overestimate volatilisation from soil surfaces.

An equation based on a power function for the increased sorption at low water contents which

is similar to eq. 4.5 in chapter 4.1.2 has been implemented in the new PELMO development

(4.21). However, it is limited to the thin surface-water layer of 1 mm.

ssdeff,d FKK ⋅= for θ = θad (4.21)

mdeff,d 10KK ⋅= for θad < θ < θwp (4.22)

Kd,eff = Kd for θwp < θ (4.23)

in which Fss is the increase of soil sorption when soil is air dry (comp. to ref. conditions) ;

θ, the volume fraction liquid phase (m3 m-3) ; θwp the volume fraction liquid phase at wilting

point (m3 m-3) ; θad the volume fraction liquid phase when soil is air dry (m3 m-3) ; Kd,eff,

the effective sorption coefficient (L kg-1) ; and Kd = sorption coefficient under moist soil

conditions (L kg-1).

The exponent m is given by:

])loglog(

)loglog(Flog[m

wp10

ad10

wp1010

ss10

Θ−ΘΘ−Θ

= (4.24)

In the present status PELMO automatically estimates the soil moisture when soil is air dry as

10 % of the wilting point. This function cannot be modified by the user.

4.4.2 Volatilisation from plants


Currently FOCUSPELMO estimates the fate of pesticides on plant surfaces by considering

two processes, the wash-off from plants and a lumped degradation rate on plants for all other

processes.

In order to improve the fate of pesticides on plants a new model was implemented into

PELMO that simulates the environmental fate of pesticides after application on an hourly

basis, including volatilisation from leaves, penetration into leaves, wash-off and photo-

transformation. The model algorithms are based on the PEARL plant volatilisation module

(c.f. 4.2.3). The model has the option of distinguishing between two deposit classes: a well-

exposed and a poorly-exposed class. The deposit in the latter class may be enclosed by plant

parts (e.g. in leaf axils), it might be located on the lee side of the air flow, or it is assumed to

be located deeper in the canopy.

Volatilisation of pesticide from the deposit/leaf surface is determined by vapour diffusion

through the laminar air boundary layer. The potential rate of volatilisation (Jvol,pot) of pesticide

from the deposit/leaf surface is calculated by equation 4.9.

In this concept penetration, wash-off and photo-degradation processes. With all these

processes, a complete mass conservation equation can be drawn up (4.10).

Reliable prediction of the processes affecting the environmental fate of pesticides under field

conditions is important, especially with regard to the implementation of the boundary-layer

concept in PELMO, enabling an estimation of the volatilisation of pesticides from plant and

soil surfaces. Application of the improved PELMO version to an environmental scenario of

the wind-tunnel study after application of fenpropimorph to radish plants allows for the

estimation of the relevant plant and soil processes, as summarised in figure 4.1.


0 1 2 3 4

0.01

0.1

1

10

100

Perc

enta

ge o

f net

app

lied

dose

[%]


plant volatilisation plant deposit degradation in soil soil deposit degradation on plants soil volatilisation penetration into the plants

Figure 4.1. PELMO calculation for application of 14C-labelled fenpropimorph to radish plants (semi-logarithmic plot).A poorly-exposed scenario was used for the plant processes. For the computation ofthe soil processes, a soil layer thickness of 1 cm was assumed. The fraction of the applied doseintercepted by the crop was estimated to be 87.1%.

Plant processes, including volatilisation from the crop, penetration into the leaves and

photodegradation, were computed by PELMO assuming a poorly-exposed scenario using the

boundary layer thickness. For the calculation of the soil processes, including volatilisation,

degradation and root-uptake, an advanced volatilisation module developed within the

framework of the APECOP project was applied to the soil deposit (Van den Berg et al., 2003).

Within the calculations, PELMO was not used for the prediction of pesticide leaching. The

scenario for the PELMO simulation included a default value for soil layer thickness of 1 cm

and the fraction of the applied dose intercepted by the crop was taken to be 87.1%.

Computations on the plant processes in PELMO were in agreement with the predictions,

illustrated by a cumulative volatilisation from the plants of approximately 50% of the applied

dose.


4.5 Volatilisation modelling with PEARL

4.5.1 Introduction

The description of the volatilisation process in FOCUS-PEARL version 1.1.1 (Leistra et al.,

2001) is based on the assumption that a laminar boundary air layer exists through which the

pesticide has to diffuse before it escapes into the atmosphere (see Section 4.1). This concept

does not take the effect of changing meteorological conditions into account, so the concept of

resistances to transport of pesticide from the soil surface into the air as presented in section

4.1 was adopted to improve the volatilisation description in FOCUS-PEARL 1.1.1. To

describe the concentration of the pesticide in the gas phase at the soil surface during dry soil

surface conditions more accurately, the description of the sorption process was modified to

include the effect of soil moisture on the sorption coefficient.

4.5.2 Improved model concept for volatilisation from soil in PEARL

The resistance to pesticide transport from the soil surface to the turbulent air has been

described in section 4.1. Because PEARL requires meteorological data on a daily basis, the

effect of changes in the atmospheric stability during the day could not be taken into account.

Therefore, atmospheric conditions were assumed to be neutral, because volatilisation occurs

mostly during the day and during daytime atmospheric conditions are commonly neutral.

Under neutral conditions, the aerodynamic resistance as given in eq. (4.5) can be simplified

to:

*

0

bl

a u

zz

ln

rκ

= (4.25)

in which zbl, is the height of internal boundary layer (m) ; z0, the roughness length for

momentum (m) ; κ, the Karman constant (dimensionless) ; and u*, the friction velocity (m s-

1).


The height of the internal boundary layer zbl can be calculated iteratively using the equation

given by Van der Molen et al. (1990) and this procedure was implemented in the new PEARL

version.

The boundary layer resistance to the transport between the source height (i.e. the soil surface)

and z = z0m is described using the equation presented by Hicks et al. (1987, see equation 4.6).

The Prandtl number was taken to be 0.72, whereas the Schmidt number depends on the

diffusion coefficient of pesticide in air. In PEARL, the coefficient for diffusion in the gas

phase is an input parameter. The actual value at a specific time is adjusted to the prevailing air

temperature near the soil surface. Hence, the flux density of volatilisation in the new PEARL

version is given by eq. (4.4). It is assumed that the concentration of the pesticide in the

turbulent air is negligible compared to the vapour concentration at the soil surface.

The concept as described by eqs. (4.4)-(4.6) was implemented in FOCUS-PEARL 1.1.1 as an

option for the user to select. If this option is selected, then z0 has to be specified. The effect of

soil moisture on the sorption coefficient was accounted for by implementing the concept

given in section 4.1. In the new version, the user must specify the sorption coefficient for air-

dry soil. The actual sorption coefficient at a specific time is adjusted to the prevailing

moisture content.

4.5.3 Testing the new PEARL version for soil volatilisation

The new version of PEARL was tested against the volatilisation rate of fenpropimorph as

measured in the Jülich-1 field experiment. This field experiment is described in section 3.3 of

this report. For the PEARL parameterisation, the vapour pressure, the water solubility and the

Koc of fenpropimorph were taken to be 2.3 mPa, 4.3 mg L-1 and 3569 L kg-1, respectively. A

boundary air layer was assumed and the thickness of this layer was set at 1 mm. This

thickness is about the same as the value for the roughness length during daytime as calculated

from wind speed data and assuming neutral atmospheric conditions. The amount of

fenpropimorph applied was assigned to the top 1 mm of the soil profile. The van Genuchten

parameters for each soil layer were obtained using the Staring Reeks shell and database. The

input data on the clay and organic matter contents were taken from the measurements on the

composition of the layers in the Jülich soil.


The preliminary results of the comparison between the rates of volatilisation of

fenpropimorph calculated with FOCUS-PEARL and the new PEARL model, and the rates

measured in the field are shown in figure 4.2 during the day of application. The volatilisation

rate computed with FOCUS-PEARL was up to a factor 4 higher than most of the rates

measured. Using the new PEARL model, however, the volatilisation rate computed was not

much different from that measured. This improvement is due to the inclusion of the effect of

an increased sorption coefficient at low soil water contents in the new PEARL. Both FOCUS-

PEARL and the new PEARL model underestimated the volatilisation rate on the 6th day after

the day of application. As the hydrological model only produces output on a daily basis, the

wetting and drying out of the soil surface during the day cannot be taken into account. On this

day 10 mm of rain fell and in the simulation this was distributed equally over 24 h.

Modification of the hydrological model to handle input and output on a hourly basis can be

expected to give a more accurate description of the soil moisture condition at the surface.

Figure 4.2. Comparison of the volatilisation rate of fenpropimorph calculated with FOCUSPEARL and themodified PEARL model, with those measured during the field experiment at Jülich in May 1995. Start ofsimulation, 11 May 0.00

4.5.4 Testing the new PEARL plant volatilisation module

The PEARL plant volatilisation module was calibrated and tested on the basis of results from

wind tunnel experiments in which pesticides were sprayed on plants. The set-up of the wind

0

100

200

300

0 5 10 15 20

Time (days since start of simulation)

Vol

atili

satio

n ra

te (µ

g/(m

²h)) PEARL New

FOCUS-PEARLMeasurements


tunnel experiments has been described by Wolters et al. (2003). The use of radio-labelled

compounds was essential for estimating the rate and extent of the processes competing with

volatilisation. An illustration of calibration of the model on the basis of an experiment with14C-fenpropimorph sprayed on bean plants is given in figure 4.3. In this case, volatilisation

was by far the main process, with lower extents of penetration and phototransformation.

Volatilisation was favoured by the rather high temperature and by filtering the inlet air (low

extent of indirect phototransformation). Only by assuming a fraction of the deposit to be

poorly exposed, the gradually continuing volatilisation after the first day could be simulated

(Figure 4.2). Details on these computations have been given by Leistra & Wolters (2003). The

present sub-model was found to be suitable for incorporation into current models for pesticide

behaviour in soil-plant systems in the field (Wolters et al., 2003).

Both in the computations and in the measurements, the environmental conditions in the wind

tunnel studies determined the relative importance of the competing processes. Volatilisation

dominated at higher temperatures and air flow rates, whereas penetration into the plants

dominated at lower temperatures and air flow rates.

0 1 2 3 4 5Time (days)

0

10

20

30

40

50

60

70

80

90

100

Per

cent

age

of th

e do

se

Deposit comp. Volat. comp. Penetr. comp.Transf. comp. Volat. meas. Deposit meas.Plant meas. Transf. meas.

Figure 4.3. Results of measurements and single computations (PEARL plant volatilisation module) for thefungicide fenpropimorph sprayed on bean plants in a wind tunnel. Fraction of 0.2 of the deposit poorly exposedto the processes. Rate coefficients for the poorly-exposed deposit were set at 20% of the coefficients for the


corresponding processes of the well-exposed deposit. kpen = 3.10 d-1; dlam = 1.0 mm; kph,ref = 0.18 d-1. For thecalculations, a diffusion coefficient in air of 0.36 m2 d-1 and a vapour pressure of 3.50 mPa at 20 °C were used.

Two further data sets on the volatilisation of pesticides from crops in the field from previous

experiments were also prepared for testing the module. The first data set dealt with the

volatilisation of fenpropimorph and quinoxyfen from a sugar beet crop in Merzenhausen,

Germany. The second set, data measured the volatilisation of fenpropimorph and chlorpyrifos

from a potato crop in Wieringermeer, The Netherlands.

4.5.5 Conclusions

The stronger sorption of pesticide at low soil water contents can be described with a simple

exponential function that gives the relation between sorption coefficient and water content.

The water content below which the sorption coefficient increases can be taken to be the

moisture content of the soil layer at wilting point.

At present, daily meteorological data are used to calculate the transport resistances, so the

diurnal pattern of volatilisation cannot be described. This limitation can be remedied by a

further model improvement to process hourly meteorological data.

The introduction of the aerodynamic transport resistance for neutral atmospheric conditions

in the concept to describe volatilisation is a first step to improve this description. Further

improvement is needed to take atmospheric stability for the description of plant and soil

volatilisation processes on the field scale into account.

The development of the PEARL Plant Volatilisation module allows for the simulation of the

environmental fate of pesticides after application to plant surfaces, including volatilisation,

wash-off, penetration and photodegradation. Further improvements of this approach will

require a deeper understanding of the underlying processes and with this the use of more

reliable values for the rate coefficients (penetration, phototransformation, etc.).


References

Asman, W.A.H., 1998. Factors influencing local dry deposition of gases with special reference to ammonia,Atmos. Environ., 32, 415-421.

Baker, J.M., Koskinen, W.C., and Dowdy, R.H., 1996. Volatilization of EPTC: Simulation and measurement.J.of Environ. Qual., 25, 169-177.

Hicks, B.B., Baldocchi, D.D., Meyers, T.P., Hosker R.P., and Matt, D.R., 1987. A preliminary multipleresistance routine for deriving dry deposition velocities from measured quantities. Water, Air and SoilPollution, 36, 311-330.

Leistra, M. & A. Wolters (2003). Computations on the volatilisation of the fungicide fenpropimorph from plantsin a wind tunnel (submitted for publication).

Leistra, M. (2002). Volatilisation of pesticides from plant surfaces. Notes in the framework of a literature study.Progress Report July 2002. Alterra, Wageningen, The Netherlands.

Leistra, M., van der Linden, A.M.A., Boesten, J.J.T.I., van den Berg, F., 2001. PEARL model for pesticidebehaviour and emissions in soil-plant systems; Descriptions of the processes in FOCUSPEARL v 1.1.1.Alterra-rapport 013, RIVM report 711401009, ISSN 1566-7197.

Smit, A.A.M.F.R., van den Berg, F., Leistra, M., 1997. Estimation method for the volatilization of pesticidesfrom fallow soils. Environmental Planning Bureau series 2, DLO Winand Staring Centre, Wageningen, TheNetherlands.

van den Berg, F., A. Wolters, N. Jarvis, M. Klein, J.J.T.I. Boesten, M. Leistra, V. Linnemann, J.H. Smelt & H.Vereecken (2003). Improvement of concepts for pesticide volatilisation from bare soil in PEARL PELMOand MACRO models within the framework of the EU–APECOP project. Proc. 12th Symposium PesticideChemistry (in press).

Van den Berg, F., Kubiak, R., Benjey, W.G., Majewski, M.S., Yates, S.R., Reeves, G.L., Smelt, J.H., van derLinden, A.M.A., 1999. Emission of pesticides into the air. Water, Air and Soil Pollution, 115, 195-218.

Van der Molen, J., Beljaars, A.C.M., Chardon, W.J., Jury, W.A., van Faassen, H.G., 1990. Ammoniavolatilization from arable land after application of cattle slurry. 2. Derivation of a transfer model. NetherlandsJournal Agricultural Science, 38, 239-254.

Wolters, A., M. Leistra, V. Linnemann, J.H. Smelt, F. van den Berg, M. Klein, N. Jarvis, J.J.T.I. Boesten & H.Vereecken (2003). Pesticide volatilisation from plants: Improvement of the PEARL, PELMO and MACROmodels. Proc. 12th Symposium Pesticide Chemistry (in press).

Wang, D., Yates, S.R., Gan, J., 1997. Temperature effect on methyl bromide volatilization in soil fumigation, J.Environ. Qual., 26, 1072-1079.

Woodrow, J.E., J.N. Seiber, Baker, L.W., 1997. Correlation techniques for estimating pesticide volatilisationflux and downwind concentrations. Environ. Sci. Technol. 31, 523-529.


5. Overview of data sets used in the PEC model validation

The performance of the local scale PEC models was evaluated using experimental data from

seven field sites. The major characteristics of these sites are summarised in Table 1. The

major criterion for considering the site was the availability of data appropriate for model

validation. This implies the availability of high quality data on water, solute, and heat

transport in the soil, and data on pesticide fate and transport in soil and ground-water (if

applicable). The availability of support from the data set provider was considered to be

essential. The data sets for the model validation are now fully documented. Details of the data

set can be found back on the APECOP data base which is listed in Annex 3.

Table 5.1. Major characteristics of the APECOP experimental field sitesName Country Climate Soil type Available

referenceLanna Sweden Cold humid Silty clay over

clayLarsson andJarvis, 1999

Brimstone United Kingdom Moderate seaclimate

Cracking heavyclay

Bromilow et al.,1998; Harris and

Catt, 1999;Armstrong et al.,

2000.Andelst the Netherlands Moderate sea

climateSilty clay loam In press.

Vredepeel the Netherlands Moderate seaclimate

Humic sand Boesten and vander Pas, 1999,

2000Lebrija Spain Mediterranean Clay (drained

marshland)Andreu et al.,

1996, Rieu et al,1998;

Coria Spain Mediterranean Silty clay loam In press.Bologna Italy Mediterranean Loam Araldi, 1997,

Scarabello, 1999


6. Validation of point scale models

The Directive 91/414/EEC, developed by Member States to harmonise registration procedures

for pesticides within the EU, places great importance on the use of validated models to

calculate Predicted Environmental Concentrations (PECs), as a basis for assessing the

environmental risks and health effects. An appropriate PEC model should therefore be

sufficiently validated for the different conditions in the EU. Unfortunately, the present

pesticide emission models are characterised by an intrinsic low validation status. Model

predictions are subjected to the following errors: invalid model structure, incorrect

mathematical formulation of the process descriptions, incorrect parameterisation, and

incorrect input definition. Validation results in the characterisation of the modelling error and

comprises the testing of the model against experimental data. Excellent simulation fits have

rarely been reported in the literature, and this is not surprising given the complexity of the

system to be described. One of the APECOP objectives was therefore to evaluate the

validation status of the PEC ground-water models, and to propose effective strategies to

reduce the uncertainty in the present PEC models, based on a detailed validation analysis.

The validation study of a model should answer key questions such as “How near are my

estimates to real world conditions?” If a ‘valid’ model is taken to mean a ‘true’ model, then

validation of a computer model is, in principle, an impossible task (Oreskes et al, 1994;

Refsgaard and Storm, 1996), since no model can be ‘true’ in this absolute sense. However, a

more pragmatic definition of validation (e.g. as envisaged in FOCUS) means that it has to be

substantiated that a model can be parameterised for a number of different sites with

acceptable results, for a given context of application, implying that the errors made with

different leaching models must be quantified. As a new site is always different from earlier

test sites it can never be proven that the computer model will perform adequately in the new

situation. However, with many positive tests, the probability of success at a new site with

similar properties increases, at least in the minds of the users, thereby building confidence in

the model for this given purpose (Rykiel, 1996). Similarly, as substances have different

properties and are subject to different reactions, it is not proven that a model that has been

validated for one substance will also simulate a different substance correctly. The uncertainty

related to the description of substance transformation and transport processes will be

substance-dependent. It is not possible to remove this uncertainty, even if the model

simulation of flow and conservative solutes is perfect. Again, only through a number of

simulations of substances with similar properties can this uncertainty be reduced. Thus, model


validation, in this pragmatic sense, envisages quantifying the error that is made when

predicting environmental concentrations with different PEC models for a given range of

scenarios, in order to build confidence in the use of the model for a given purpose (i.e.

regulation), and to define the limits of applicability of the model (Rykiel, 1996).

Three models (PELMO, PEARL and MACRO) are tested in APECOP. The selected PEC

ground-water models have been subjected to several previous validation exercises before (e.g.

Styczen, 1995; Thorsen et al. 1998; Vanclooster et al., 2000). The results have shown that

large differences between these models can be obtained. The variability in model predictions

therefore introduces a significant source of uncertainty in pesticide fate evaluation. However,

the exercises documented so far represent a very limited number of cases compared to the

number of potential scenarios for which the models will be used in the registration context. In

addition, most validation studies reported in literature comprise the calibration of the model

against the data to be simulated. Model calibration is a crucial action in pesticide fate

modelling. It is sometimes considered that model calibration can help reduce the uncertainty

in the modelling through the derivation of values for input parameters that help to improve the

simulation of experimental data. The uncertainty in model predictions resulting from the

influence of the modeller should also be reduced in the validation study by providing detailed

guidance on model parameterisation and model use. Very few validation studies reported so

far, have assessed the prediction capacity of the PEC models in a pure predictive mode , i.e.

without calibration or so-called blind validation. In APECOP, a systematic validation

procedure has been implemented which allows to address all these aspects in systematic

validation study.

6.1 Validation protocol

6.1.1 Introduction

Stepwise validation protocols such as those presented by Anderson and Woessner (1992),

Styczen (1995), Thorsen (1998) and Vanclooster et al. (2000) can be adopted to reduce the

uncertainty associated with the parameters, model input algorithms and code, thus adding

credibility to the simulation results. This may also provide some guidance for future model

users regarding which model performs best for which scenario. Based on previous experience

(Vanclooster et al., 2000), a multistage validation protocol, in which the different components


of the pesticide emission models are validated separately in a sequential process, is adopted in

APECOP.

6.1.2 Validation steps

The validation protocol was drawn up based on the project objectives and the availability of

data for the selected sites. As such, the hydrological component of the emission models was

validated using the measured water balance, as reported in the data bases of the seven sites.

The solute transport component of the emission models was validated using available tracer

data, while the pesticide fate and transport component was validated using measured terms of

the pesticide balance. A clear distinction between calibration and extrapolation/prediction was

drawn. The validation exercise was executed by 6 modelling partners with proven experience

with the use of the present PEC models. Only two modelling partners per code were

considered in order to minimise user subjectivity. For carrying out the validation, data from

seven field sites (either historical data or new data) were made available (cf. section 5).

One original aspect of the procedure was the adoption of a blind validation step for some of

the field sites. For this blind validation exercise, no field data were made available to the

model users. Hence, the performance of the PEC models was evaluated for the case when

only laboratory data or generic data are available for pesticide fate parameters. Although it is

well known that calibration may considerably improve model performance, field data are

generally not available in the regulatory context and blind validation is therefore an

appropriate test for this purpose. The evaluation protocol can be summarised in the following

steps:

Blind validation

In this step, the model performance is analysed without calibration of data. It should be noted

that, in a regulatory context, calibration of the pesticide components of the models will in

general not be possible, because it requires several field experiments for the same

pesticide/soil combination that are usually not available. This justifies the inclusion of a blind

validation step in the procedure. In pesticide registration procedures, a first assessment is

made on the basis of pesticide properties measured under well-defined laboratory conditions.

So for decision-makers, it is crucial to know the potential of a model to predict pesticide

behaviour without calibration.


The procedure was divided in two levels as a function of the information available. If the

modeller already had information about water, heat, solute, and pesticide in the soil system

then level 1 was skipped.

Level 1: Blind validation based on generic input parameters which gives an indication of

model performance if little information is available. No laboratory data are available. This

means that the modeller is only informed about land use and soil characteristics (soil type,

map information, horizons, texture, pH, crop type, land management practice, pesticide

application time and dose, etc.), and the boundary conditions (climate, hydrogeology,

fertilizer management). In this procedure, no additional laboratory results are made available

related to soil hydrology, adsorption kinetics, degradation rates, etc.) and the user will need to

resort to generic data sources such as pedotransfer functions for hydraulic properties, EPA or

other chemical databases for sorption and degradation properties, koc values etc. It must be

emphasized that ‘blind validation’ at level 1 represents a test not only of the model itself (i.e.

the process descriptions) but also of the routines used to parameterise the model and the

appropriateness of the data sources. In other words, this test cannot distinguish between model

error and parameter error (Loague and Green, 1991).

Level 2: Blind calibration at level 2 represents a test in which all results of laboratory studies

(both on hydrology and pesticides) can be used but also all available information on boundary

conditions (e.g. time series of ground-water levels, rainfall and air temperatures); “blind”

implies that the modeller has no information about water, heat, solute, and pesticide in the soil

system itself (so no moisture profiles, no pesticide concentrations in soil profile and ground-

water etc.) or if these are known, they are not used in the parameterisation process.

For this reason, level 1 was used only at new sites (Andelst, Coria, Lebrija, Bologna). Some

parts of the historical data sets (Vredepeel, Brimstone, Lanna) have already been published in

the scientific literature, and can therefore not be considered as blind data sets. For these

historical data sets, only level 2 is relevant.

Model calibration

In a second phase, calibration was considered as a parameter estimation technique. During

calibration, an object function is optimised such that the differences between model calculated

and observed pesticide concentrations are minimised, using laboratory data (if available) at

first and field data subsequently. It is crucial that calibration of pesticide behaviour must be

carried out using justifiable procedures. It was not considered admissible to calibrate pesticide


parameters such as sorption coefficients and half-lives, where these have been measured at the

site, without scientific justification.

Model prediction

In the third phase, the ability of the model to extrapolate or to predict was assessed. During

this phase, no re-adjustment of the model parameters was done and the whole model with

calibrated parameter values is tested if data are available.

6.1.3 Target values

In model testing, particular attention needs to be paid to the target quantity to be considered.

The sub-models for simulating water flow, solute flow, heat flow, and pesticide behaviour

could be tested using system-state variables such as the soil moisture content, the soil solute

concentration, the soil temperature, and the soil pesticide (bentazone, ethoprophos,

imidacloprid, aclonifen, isoproturon, chloridazon) concentration. Alternatively, balance terms

such as the cumulative soil water drainage or the cumulative amount of pesticide leached can

be used. The balance terms may be more appropriate in a risk analysis context. Yet the

balance terms are hard to measure on a field scale, which makes the test of pesticide-leaching

models at this scale cumbersome. Correct cumulative amounts leached can be inferred from

lysimeter datasets. However, flow behaviour in lysimeters does not reflect the complete

variability of pesticide flow that might occur in the field, and bias may result in the risk

assessment. To cope with this dilemma, a pragmatic solution was followed. For the field

datasets, the simulated state variables are compared with the field-measured state variables in

space and time. Adopting such an approach, one tacitly assumes that if the model matches

measured state variables in space and time, the model will have a large probability of

simulating the balance terms appropriately. However, it is recognized that this assumption

may not be correct when only resident soil concentrations are available for model comparison

and preferential flow and transport significantly affect leaching (Jarvis, 1999).

6.1.4 Statistical criteria

Either modelling statistics or simulation graphics can be used to assess the performance of the

model. When both system output and model output are considered to be deterministic, the

arithmetic difference between the measured and calculated output is the most straightforward

measure of the model deviation. The large amount of data, however, suggests the use of more


comprehensive model deviation indicators as reported in Vanclooster et al. (2000). In this

study, only the model efficiency (EF) was used (Loague et al., 1989):

( ) ( )

( )∑

∑∑

−−−

=N

NN

1

11

2

i

2ii

2

i

O-O

OPOO

EF (6.1)

where Pi is the predicted value ; , Oi the observed; O , the mean observed value, and N the

number of observations. This indicator is equal to 1 when predicted (P) and observed (O)

values are equal and becomes negative when the ratio between simulation error (P-O) and

H[SHULPHQWDO�GLVSHUVLRQ��2� ��LV�OHVV�WKDQ�RQH��8QGHU�WKHVH�FRQGLWLRQV��WKH�PRGHO�SUHGLFWLRQVare not significantly better than the mean of the observed data. When EF becomes less than

minus 1 the fit could be considered unacceptably poor (Walker et al., 1995). For each

combination of model and data set only one EF value was calculated via summing over all

sampling depths and sampling times.

6.1.5 Models

The following versions of the models were used: PEARL 1.1.1; MACRO 4.3; PELMO 3.2

All derived input data was tagged with tables indicating how the value was obtained (Table

6.1). These tables were then filled out by the different modellers at the different steps (blind

simulation, calibration with laboratory data, calibration with field data, prediction).

Table 6.1. Example of the table containing the information related to model input parameters

that must be coupled with input files.

Input name Value Source and comments

Segment thickness (mm) 50 Depth of soil sampling in data-set

Water solubility (mg/l) 500 From PETE data base

Lower boundary condition (option) 2 Free-draining profile as indicated in data-set

Clay layer 0-5 cm % 2.7 From data-set

Clay layer 50-55 cm % 3 Lacking data. We use the same data of 75-80 layer

Campbell exponent 3 Model’ s value default


6.2 Validation of PEARL

6.2.1 Introduction

The PEARL v 1.1.1 model was tested against five leaching datasets from four EU countries

(Andelst, Bologna, Brimstone, Lanna and Vredepeel). Three of the datasets were clay soils

(Andelst, Brimstone and Lanna) and the other two were a sandy loam (Bologna) and a sand

(Vredepeel). Each dataset was only tested by one partner (Andelst and Bologna by Alterra and

the other three by JRC).

Only very limited time was available for testing the new PEARL version including

preferential flow. Thus this was calibrated only to the bentazone leaching of the Andelst data

set.

6.2.2 Andelst

Detailed results for this site are given by Scorza Junior, 2002, Mechteld ter Horst, 2003).

Testing procedure for v 1.1.1

The stepwise calibration procedure was followed as described in Section 6.1 (blind validation

level 1, blind validation level 2, calibration). All calibrations were carried out using the PEST

software package (Doherty, 2000).

Results and discussion for v 1.1.1

The model efficiency (EF) for the moisture profiles was better for level-1 than for level-2 (see

Table 6.2). However, calibration of Van-Genuchten parameters produced the best EF. In

level 2 moisture characteristics were used measured in the laboratory using soil samples from

one pit. Obviously this was an inadequate procedure due to spatial variability in the field. The

calibrated runs resulted in a good description of measured daily drain flow. After this

calibration of water flow, measured soil temperatures were simulated well.

Level-1 simulations produces no drainflow because free drainage was assumed. Thus for all

substances simulated drain water concentrations were zero. Table 6.2 shows that this resulted

in a EF of -15 for bromide drain water concentrations (and strange enough in -0.2 for

bentazone and -0.8 for imidacloprid; so EF values differed strongly although all simulated

values were zero). The very low EF values for bromide and bentazone concentrations in

ground-water at level-1 were caused by simulated concentrations that were far too high (see

Figure 6.1). This is probably caused by the exclusion of drainage in level-1 leading to

increased substance fluxes to ground-water.


Table 6.2. EF’ s for different steps in the validation procedure for PEARL v 1.1.1 and the Andelst dataset.blind

validation

level 1

blind

validation

level 2

calibrated

moisture profiles -0.7 -0.9 +0.5

bromide concentrations in drain water -15.1 -3.8 -1.9

bromide concentrations in ground-water -75.8 -0.9 -0.3

bentazone concentrations in drain water -0.2 -0.2 -0.2

bentazone concentrations in ground-water -123.3 -2.6 -0.3

imidacloprid concentrations in drain water -0.8 -0.8 -0.8

imidacloprid concentrations in ground-water -1.3 -1.3 -3.5

Measured concentration profiles of bromide in December and at the end of the winter were

used to calibrate the dispersion length. The calibration thus focussed on bulk leaching in

winter. The uncalibrated dispersion length of 5 cm resulted in simulated profiles that

exhibited less spreading than the measured profiles. Calibration resulted in a good description

of measured profiles and a dispersion length of 61 cm. This is far longer than observed in soil

column experiments and is possibly caused by spatial variability of the solute flux at the field

scale. The calibration improved also the EF values for bromide concentrations in drainwater

and ground-water (see Table 6.2).

The total amount of bentazone in soil declined considerably faster than calculated on the basis

of laboratory-measured half-lives. Because it was the aim to test PECgw (so mainly the

transport part of the model), the transformation rate of bentazone in soil was calibrated. After

this calibration, PEARL v 1.1.1 simulated the bentazone concentration profiles well.

However, PEARL v 1.1.1 underestimated the drainwater concentrations of bentazone: by far

the highest concentrations were measured in the first drain


Figure 6.1. Simulated and measured concentrations in ground-water between 1 and 1.2 m depth at theexperimental field Andelst. Calculations were carried out with PEARL v 1.1.1.

flow event a few weeks after application when PEARL v 1.1.1 predicted zero concentrations.

This happened also in the blind validation calculations and explains why the EF of bentazone

EF for drain water did not change between the calibration steps (see Table 6.2). PEARL v

1.1.1 simulated the persistence of imidacloprid in soil reasonably well without any further

calibration. However, the measured concentration profiles were not described well: PEARL v

1.1.1 predicted deeper leaching than was observed. In contrast to this, PEARL v 1.1.1

predicted too low concentrations of imidacloprid in drain-water. The simulation of the

ground-water concentration was more or less reasonable (see Figure 6.1). However, the EF of

the calibrated run was worse than the EF’ s of the blind validation runs (see Table 6.2). The

blind validation runs produces zero concentrations in ground-water (so lines are not visible in

300 350 400 450 500

Con

cent

ratio

n in

gro

undw

ater

(m

g dm

-3)

0.0

5.0

10.0

15.0

20.0

0 100 200 300 400

Con

cent

ratio

n in

gro

undw

ater

(µ g

dm

-3)

0.0

20.0

40.0

60.0

80.0

100.0

MeasuredBlind validation level 1Blind validation level 2Calibrated

Time (days)

0 100 200 300 400

Con

cent

ratio

n in

gro

undw

ater

(µ g

dm

-3)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Bromide

Bentazone

Imidacloprid


bottom graph of Figure 6.1). This seems more or less an artefact of the EF approach: in the

context of pesticide registration it seems difficult to defend that zero is a better predictor than

the simulated line in the bottom part of Figure 6.1.

We conclude that PEARL v 1.1.1 could describe well (1) water flow, (2) the bulk leaching of

the tracer and of the mobile pesticide bentazone, (3) persistence of the persistent imidacloprid.

However PEARL v 1.1.1 could not describe preferential drainflow of both pesticides and

penetration of the sorbing pesticide into the soil. This illustrates the need for high-quality

datasets: if only the mobile bentazone had been tested and only soil profiles had been

measured, the deficiencies in the transport part would not have been detected.

Testing procedure for new PEARL version including preferential flow

Only bentazone leaching was considered. All parameters (except those describing preferential

flow) were taken from the calibrated runs of v 1.1.1. The thickness of the mixing-cell layer

and the saturated hydraulic conductivity were calibrated using the measured concentrations in

drain-water.

Results and discussion for PEARL version including preferential flow

Calculations with initial estimates of the calibration parameters showed a zero concentration

of bentazone for the first and most important drain-flow event after about 20 days (see Figure

6.2). To obtain the calculated results shown in figure 6.2 a mixing-cell thickness of 1 mm was

used and the saturated hydraulic conductivity of the soil matrix had to be lowered to about 1

cm/d (thus forcing more water to flow into the macropores). These are preliminary results

because only very limited time was available for performing this calibration due to initial

problems with the numerical solution procedure.


Figure 6.2. Measured and simulated bentazone concentrations in drainwater at the Andelst experimental field.Only calibrated calculations are shown for PEARL v 1.1.1 and the new PEARL version including preferentialflow.

6.2.3 Bologna

Testing procedure

Van-Genuchten parameters were calibrated using measured moisture profiles. It was

attempted to use measured tracer concentrations in ground-water to calibrate solute flow. The

dosage and the half-life in soil were calibrated to obtain a good description of persistence. It

was tested whether PEARL v 1.1.1 could explain the observed movement of pesticides in the

soil profile and their concentration in ground-water.

Results and discussion

Calibration of van-Genuchten parameters resulted in a considerable improvement of

simulated moisture profiles. For bromide, only concentrations in ground-water were available.

These appeared not useful for calibrating the dispersion length. Simulations with uncalibrated

dosages and half-lives resulted in an strong overestimation of measured total amounts in the

field (measured typically five times lower than simulated). The calibration of the dosage was

necessary to account for initial loss processes (e.g. photodegradation at the soil surface) that

are not included in PEARL v 1.1.1. For aclonifen, measurable concentrations were only found

in the top 5 cm layer over the sampling period of about 1 year.. This lack of movement of

aclonifen was simulated well by PEARL v 1.1.1. Ethoprophos concentration profiles were

available only for the first two months after application. Highest concentrations were always

found in the top 5 cm. This was also simulated reasonably well by PEARL v 1.1.1. Available

PEARL with macropores

Time (days)

0.0 100.0 200.0 300.0 400.00

20

40

60

80

100Drain set 1Drain set 2simulated

PEARL 1.1.1

Time (days)

0.0 100.0 200.0 300.0 400.0Con

c. B

enta

zone

in d

rain

flow

(µg

dm-3

)

0.0

20.0

40.0

60.0

80.0

100.0


measurements of ground-water concentrations of aclonifen (in July-November 1995 after

application in May 1995) were all below the detection limit. PEARL v 1.1.1 simulated in all

cases zero concentrations in ground-water.

We conclude that PEARL v 1.1.1 could simulate the lack of movement of aclonifen and

ethoprophos reasonably well. However, this cannot be considered as a sensitive test of

PEARL v 1.1.1 for PECgw assessment.

6.2.4 Brimstone

Testing procedure

Van-Genuchten parameters were calibrated using measured moisture profiles (which were

available for 38 sampling times) and using measured drain-flow over a six-year period. No

further calibration was carried out. Only drain water concentrations of isoproturon were

available for testing of pesticide behaviour.


The average predicted moisture content was quite close to the measured one. However, on

occasions differences between measured and simulated volume fraction of water were as high

as 0.10. The EF for daily drain-flow was 0.44. Measured fluctuations of the ground-water

table were considerably larger than simulated fluctuations. Measured concentration peaks of

isoproturon in drain-water occurred several weeks earlier than predicted by PEARL v 1.1.1

(probably due to preferential flow).

The results indicate that PEARL v 1.1.1 is unable to describe pesticide leaching adequately in

this clay soil. However, the relevance of this soil for assessment of leaching to ground-water

in the context of EU registration seems questionable in view of the extremely high clay

content of 0.54-0.62.

6.2.5 Lanna

Detailed results for this field site are given by Bouraoui et al., 2003.

Testing procedure

Van-Genuchten parameters were calibrated using measured moisture profiles and measured

drain-flow. The dispersion length was calibrated using measured bromide concentration

profiles. No further calibration was carried out.



Calibration resulted in a good simulation of measured moisture profiles (EF of 0.8). Drain

flow was simulated well during the first nine months but thereafter PEARL v 1.1.1

underpredicted the drain flow (13% too low averaged over whole experimental period). This

underprediction is possibly caused by preferential water flow that is not simulated by PEARL

v 1.1.1. Calibrated dispersion lengths were 15-30 cm and these resulted in a reasonable

description of measured bromide concentration profiles. Bentazone concentration profiles

were also described well. However, PEARL v 1.1.1 underestimated bentazone drain-water

concentrations in the first three months in which drain flow occurred. In the following five

months PEARL v 1.1.1 simulated the bentazone drain-water concentrations reasonably well.

The results of this study imply that PEARL v 1.1.1 is inadequate for simulating leaching of

small fractions of the dose in this structured clay soil. PEARL v 1.1.1 predictions of drain-

water concentrations would probably have been worse if a pesticide had been tested of which

only a small fraction leached (in this study about 8% of the dose leached out of the drains).

6.2.6 Vredepeel

Detailed results for this field site are given by Bouraoui et al., 2003.

Testing procedure

Van-Genuchten parameters were calibrated using measured moisture profiles. The dispersion

length was calibrated using measured bromide concentration profiles. No further calibration

was carried out.


Calibration resulted in a good simulation of measured moisture profiles (EF of 0.82).

Calibration of the dispersion length showed that the best description of bromide concentration

profiles was obtained using the initial estimate of 3 cm (EF of 0.94). Concentration profiles of

bentazone and ethoprophos were described well (EF of 0.80 and 0.89, respectively). So after

calibration PEARL v 1.1.1 could describe movement of the pesticides quite well. However,

the model test cannot be considered as a sensitive test of PEARL v 1.1.1 for PECgw because

the test (and calculated EF-values) focussed on the leaching of the bulk of the pesticides and

not on leaching of a small fraction of the dose which is most relevant for pesticide

registration.


6.2.7 Conclusions

In all model tests, calibration of the water flow sub-model appeared necessary for an adequate

description of water flow. After this calibration, a more or less acceptable description could

be achieved in nearly all cases. This indicates that the problem is caused by the inaccuracy in

the estimation of the model input.

Three tests with clay soils indicated that PEARL v 1.1.1 cannot describe solute transport in

such soils good enough because it does not describe preferential flow. First calibrations with

the new PEARL version (that includes preferential flow) showed promising results for one of

these clays soils (no time was available for calibrating the other two clay soils). The

correspondence between predicted and measured lack of movement in the Bologna loam soil

cannot be considered as a sensitive test of PEARL v 1.1.1 in the context of the assessment of

pesticide leaching to ground-water for EU registration. The same applies to the test for the

Vredepeel dataset because this test focussed on movement of the bulk of the pesticides.

6.3 Validation of MACRO

The upgraded MACRO model (v5.0) has been compared with the existing FOCUS version of

MACRO (v.4.3) in validation tests at three of the field sites included in APECOP, namely

Andelst (Netherlands), Lanna (Sweden), and Lebrija (Spain). All three sites are clay soils, that

were expected to show strong evidence of macropore flow. References to the descriptions of

the three experimental sites can be found in Chapter 5. Since the Lanna dataset was already

published (Larsson and Jarvis, 1999), no blind validation test was possible for this site.

6.3.1 Andelst

The application of MACRO4.3 to Andelst has been published in Scorza Junior (2002). Our

simulations with MACRO5.0 are still not quite finished, and so we are not yet in a position to

be able to compare the performance of the two model versions at this site.

6.3.2 Lanna

Table 6.3 compares calculated model efficiencies for the various kinds of data available at

Lanna for the two versions of MACRO, and in the case of MACRO5.0, for two different goal

functions used in the optimisation procedure. Table 6.4 shows the parameters that were

included in the calibration procedure, using the inverse modelling tool, SUFI (Abbaspour et

al., 1997).


Table 6.3. EF’ s calculated for all measured data points (Lanna).MACRO 4.3 MACRO 5.0

Goal function in SUFI Root meansquare error

Root meansquare error

Modelefficiency

Soil moisture profiles 0.88 0.58 0.47Bromide profiles 0.66 0.47 0.45Bentazone profiles 0.90 0.58 0.79Bromide concentration in drainage 0.37 0.33 0.69Bentazone concentration in drainage 0.29 0.43 0.57Drainflow 0.29 0.45 0.47

Table 6.4. Estimated parameter values (Lanna). Ranges indicate the posterior uncertainty domains (ASCALE =diffusion pathlength, mm; KSM = saturated matrix hydraulic conductivity, mm h-1; RGWFLOW = Residencetime, ground-water, days; DEG = degradation rate coeff., days-1, 20oC).

MACRO 4.3 MACRO 5.0ParameterRoot mean squareerror

Root mean squareerror Model efficiency

ASCALE (0-30cm) 253 (223-400) 169 (141-313) 289 (235-300)ASCALE (30-60cm) 61 (46-76) 357 (140-400) 203 (107-300)ASCALE (60-100cm) 65 (44-86) 357 (226-400) 32 (10-203)ASCALE (100-175cm) 31.5 (2-179) 350 (10-400) 117 (10-139)KSM (0-1cm) 0.112 (0.1-0.172) 0.0765 (0.073-0.087) 0.067 (0.01-0.12)RGWFLOW 4.5 (1-5.2) 33.8 (1-40.3) 40 (20-60)KOC cm3 g-1 4.84 (4.58-5.38) 3.11 (1-5.22) 2.17 (1.66-4.68)DEG (0-30cm) 0.04 (0.03-0.046) 0.106 (0.07-0.113) 0.112 (0.11-0.14)DEG (30-175cm) 0.01 (0.004-0.028) 0.0315 (0.003-0.06) 0.031 (0.025-0.048)

Table 6.3 shows that both versions of the model produce acceptable fits to the data, and that

the new version does not perform better than the existing FOCUS version, at least for this

dataset. The choice of goal function appears to be important when several data sources are

combined into one overall measure of goodness of fit: the model performance can be

significantly influenced (Table 6.3), and just as importantly, the best-fit parameter estimates

may also change due to the optimisation methodology selected (Table 6.4). In principle, the

model efficiency should be preferred, since it has the advantage of the numerical value being

independent of the units of the data source. Nevertheless, the differences in parameter

estimates between different model versions seem to overshadow the effects of different

optimisation methods: version 5 predicts much longer residence times of local shallow lateral

ground-water flow (values which seem much more reasonable given the size of the plots, 0.4

ha), slightly weaker sorption and considerably larger degradation rate coefficients in both

topsoil and subsoil. Table 6.4 also shows that one important parameter regulating the strength

of macropore flow (the diffusion path-length) could be identified only within very wide


posterior uncertainty bounds, especially in the topsoil, and deeper subsoil, and for the new

version of the model. The reason for this is not clear, but it may be related to the simplified

first-order description of mass exchange between the flow domains, which fails to capture the

short-term fluctuations observed in drain-flow concentrations (Figure 6.3).

Figure 6.3. Bentazone concentrations in drainflow at Lanna simulated with MACRO

6.3.3 Lebrija

Field and laboratory experiments

Lebrija is representative of the irrigated agriculture in the area of Sevilla, Spain, with sugar

beet, maize and cotton as the main crops. At Lebrija, field experiments were carried out in a

sugar beet crop treated with chloridazon combined with lenacil. Details of these experiments

can be found in Cuevas et al., (2001, 2002a,b, 2003a,b). In addition, laboratory experiments

with undisturbed soil columns were carried out to evaluate tracer (bromide) transport and the

fate and mobility of the two herbicides. Experiments with hand packed columns were also

made to obtain more information on the mobility of the pesticides. Additional laboratory

experiments were carried out to establish the degradation and sorption properties of the

herbicides.


Calibration procedure

First, we calibrated the saturated hydraulic conductivity (KSATMIN, mm h-1), the effective

diffusion pathlength (ASCALE, mm), and the dispersivity (DV, cm) against the bromide

column breakthrough experiment. For version 4.3, the field water contents were also included

in this calibration step, but for technical reasons, this was not possible using version 5.0. Two

values of ASCALE and KSATMIN were considered, one for the tilled layer (zero to 0.15 m

depth) and another for the non-tilled layer (0.15 to 0.2 m depth). A single value of DV was

considered for the two layers. Then we used the data from the columns with the herbicides to

calibrate the reference degradation rate coefficient (DEG, days-1) and the sorption distribution

coefficient (ZKD, cm3 g-1). The same values were considered for the two layers. For

parameter estimation, we used the sequential uncertainty domain parameter fitting (SUFI,

Abbaspour et al., 1997), a global search algorithm. The initial uncertainty domains of the

calibrated parameters were set to KSATMIN: 2-200, ASCALE: 2-100, DV: 1-5; DEG: 0.004-

0.2, and ZKD : 0.6-1.8. Parameters of the soil water retention function were derived from

measured pF curves obtained at a nearby site with the same soil type. Remaining parameters

were set either to default values, known values, or values estimated by expert judgement.

In the final step, we ran the model against the data from the field experiments using the

calibrated parameters. For the pesticide data, values below the detection limit were set to half

the detection limit, both for observed and simulated values. Both untransformed and log-

transformed data were used when calculating overall EF values for all depths in the profile,

since the latter gives more equal weights to all data points.

Results

Figures 6.4 to 6.6 show a graphical comparison between the measurements and the

simulations with the two versions of the model. The bromide leaching experiment showed

little apparent indication of macropore flow, with the leaching pattern suggesting a

convective-dispersive process (Figure 6.4). This is attributed to the primary and secondary

tillage operations that were carried out immediately prior to sampling the columns, producing

a fine tilth in the seedbed. This fine tilth is reflected in the small values for the effective

diffusion path-length in the tilled layer obtained by calibration with both model versions

(Table 6.5). Although both versions of the model predicted well the total amount of bromide

leaching from the columns, the agreement between measured and simulated values was better

for MACRO 4.3 than for MACRO 5.0 (Figure 6.4). The reason for this is not clear, but we


believe that this is unlikely to be due to the inherent differences in process descriptions or

numerical solution procedures, but must rather be due to different outcomes from the inverse

procedure used for parameter estimation. This is suggested by the fact that widely different

parameter estimates were obtained, especially for the saturated conductivity (both layers) and

the subsoil effective diffusion path-length. The subsoil (non-tilled) value for ASCALE

obtained with v4.3 (95mm) seems more in line with previous experience when applying the

model to heavy clay soils (e.g. Larsson and Jarvis, 1999). SUFI is not a completely automatic

inverse procedure, but depends on subjective decisions of the user following each series of

simulations, or iteration, concerning how the prior uncertainty domains should be reduced. In

our study, the simulations with versions 4.3 and 5.0 were carried out by different researchers,

and their strategies were not identical. For the 5.0 simulations, a smaller tolerance was

chosen, which allows for a greater reduction of the uncertainty (Table 6.5). However, this

represents a more risky procedure which may lead to incorrect parameter estimates if the

inverse problem is not ‘well-posed’ . For the contents of chloridazon in the soil columns,

version 5.0 gave slightly better results than 4.3 (Figure 6.5), while the contrary was found for

lenacil (Figure 6.6). The degradation rate coefficients derived by calibration of v5.0 were

significantly smaller for both compounds (Table 6.5).

Table 6.5. Parameter values and posterior uncertainty ranges obtained by calibration with SUFIParameter MACRO 4.3 MACRO 5.0Effective diffusion pathlength(tilled, 0-15 cm), ASCALE, mm

2.8 (2.0 - 3.6) 0.8 (0.7 - 0.9)

Effective diffusion pathlength(non-tilled, 15-20 cm), ASCALE, mm

95 (67 - 103) 7.9 (7.0 - 9.0)

Saturated hydraulic conductivity(tilled, 0-15 cm), KSATMIN mm h-1

34 (30 - 62) 75 (70 - 80)

Saturated hydraulic conductivity(non-tilled, 15-20 cm), KSATMIN mm h-1

96 (68 – 100) 8.1 (7.75 - 8.5)

Dispersivity, DV, cm 2.3 (2.0 - 3.8) 2.1 (2.0 - 2.3)Lenacil :Sorption coefficient, ZKD, cm3 g-1 0.77 (0.72 - 1.8) 0.64 (0.60 - 0.68)Degradation rate coefficient, DEG, days-1 0.031 (0.014 - 0.039) 0.022 (0.016 - 0.028)Chloridazon :Sorption coefficient, ZKD, cm3 g-1 0.66 (0.6 - 1.8)* 0.75 (0.6 - 0.9)Degradation rate coefficient, DEG, days-1 0.199 (0.128 - 0.200) 0.126 (0.102 - 0.151)* prior uncertainty domain not reduced


Figure 6.4. Bromide leaching in the columns

Figure 6.5. Measured and simulated contents of chloridazon in the columns in mg/m-3, at different depths anddays after the herbicide application. Each point represents the average of five measurements. The horizontal barsrepresent the standard deviation.

0 100 200

Chloridazon content (mg m-2)

0 5000 10000 150000

5

10

15

20

0 1000 2000 3000

Day 15 Day 22

0 200 400 600 800 1000

Dep

th (c

m)

0

5

10

15

20

MeasuredMacro 5.0Macro 4.3Day 29 Day 36

Applied water (ml)1000 2000 3000 4000 5000

Cum

ulat

ive

brom

ide

(mg

m-2

)

0

500

1000

1500

2000

2500

MeasuredMacro 5.0Macro 4.3


Figure 6.6. Measured and simulated contents of lenacil in the columns in mg m-3, at different depths and daysafter the herbicide application. Each point represents the average of five measurements. The horizontal barsrepresent the standard deviation.

Lenacil content (mg m-2)

0 2000 4000 6000 80000

5

10

15

20

0 1000 2000 3000 4000 5000

Day 15 Day 22

0 2000 4000 6000 8000

Dep

th (c

m)

0

5

10

15

20

0 1000 2000 3000

Day 29 Day 36

0 300 600 900 1200 1500 1800


Day 43


In general, MACRO gave good predictions of the field soil water contents, but the soil water

profiles simulated by version 4.3 agreed slightly better with the measured data than the

predictions from version 5.0 (Figure 6.7). This is primarily because this data was included in

the calibration procedure for v4.3, but not for v5.0. On April 11, however, when the soil was

relatively dry, v4.3 predicted too low water contents in the topsoil layers. This is shown in

figure 6.7 and also by the EF values at 5 and 15 cm depth (Table 6.6).

Figure 6.7. Measured and simulated soil water profiles in the field, at different days of the crop season. Eachpoint represents the average of six measurements. The horizontal bars represent the standard deviation.

0.2 0.4 0.6

Volumetric soil water content (cm3 cm-3)

0.0 0.2 0.4 0.60

20

40

60

Measured

Macro 5.0

Macro 4.3

0.2 0.4 0.6

0.2 0.4 0.6

Dep

th (c

m)

0

20

40

60

0.2 0.4 0.6

0.2 0.40

20

40

60

14 November 22 December

12 February12 January

15 March 11 April


Table 6.6: EF’ s for volumetric water content at different depths.EF for volumetric soil water content

Depth (cm) Blind Simulation Macro 4.3 Macro 5.0 5 -0.04 -0.06 0.5615 0.45 0.36 0.7025 0.94 0.96 0.6135 0.85 0.89 0.5245 0.60 0.70 0.1655 0.38 0.51 -0.07All depths 0.48 0.50 0.47

Figures 6.8 and 6.9 show the herbicide profiles measured in the field for chloridazon and

lenacil, together with the values simulated by both versions of the model. MACRO 5.0 gave

slightly better predictions than MACRO 4.3, for both herbicides, as can also be deduced from

the EF values (Table 6.7), but the overall differences are small even though the values of

several water flow and pesticide fate parameters differed significantly between the two model

versions. This suggests that small differences in soil hydraulic functions, together with

improved numerical routines in version 5.0 (i.e a finer discretization), sufficiently affected the

simulation results to lead to significantly different parameterisations of the model, especially

since the optimisation procedure contained an element of subjectivity. The dataset may also

have been insufficient or too ‘noisy’ to properly constrain the model. It is possible that

additional data (e.g. water flows or flux concentrations) may have helped to stabilize the

inverse procedure.


Figure 6.8. Measured and simulated contents of chloridazon in the field, at different depths and days after theherbicide application. Each point represents the average of eight measurements. The horizontal bars represent thestandard deviation.

0 100 200 300 400 500

Chloridazon content (mg m-3)

0 1000 20000

10

20

30

40

50


0 1000 2000

0 200 400 600 800

Dep

th (c

m)

0

10

20

30

40

50

Day 6 Day 12

Day 20 Day 32


Figure 6.9. Measured and simulated contents of lenacil in the field, at different depths and days after theherbicide application. Each point represents the average of eight measurements. The horizontal bars represent thestandard deviation.

Table 6.7. Modelling efficiency (EF) for herbicide content at different depths (BS = Blind simulation withMACRO 4.3).

EF for chloridazon content EF for lenacil contentDepth (cm) BS MACRO 4.3 MACRO 5.0 BS MACRO 4.3 MACRO 5.0 2.5 0.84 0.85 0.94 -0.1 -0.01 0.8110 -1.31 -1.31 -0.96 -0.13 -0.18 0.3120 -0.83 -0.83 -0.83 - - -45 - - 0.67 - - -All depths 0.69 0.70 0.78 0.19 0.25 0.76All depths1 -0.19 -0.02 0.37 0.39 0.41 0.481 calculated on log-transformed data

Lenacil content (mg m-3)

0 400 800 12000

10

20

30

40

50


0 300 600 900

0 200 400 600

Dep

th (c

m)

0

10

20

30

40

50

0 150 300 450

Day 6 Day 12

Day 20 Day 32

0 150 300 4500

10

20

30

40

50Day 54


Tables 6.5 and 6.6 also show that calibration had no effect on water balance predictions, but

did give small but significant improvements in model performance with respect to herbicide

leaching. However, even the blind simulations gave positive model efficiencies for Lenacil

(Table 6.7), which shows that model performance is not always poor in the absence of

calibration.

Following calibration, both versions of MACRO gave largely acceptable simulations of

bromide leaching, field soil water contents, and the fate of the two pesticides (chloridazon and

lenacil) in a recently tilled clay soil. Underprediction of persistence early in the experiment,

and overprediction later, was noted. The reason for this is not clear, but it may be due to

departures from first-order degradation kinetics. Macropore flow was not apparent in these

experiments, and only limited leaching occurred, since the fine tilth produced by tillage led to

an efficient equilibration of potentials and concentrations between macropores and soil

aggregates. This was reflected in the very small calibrated values (< 3mm) of the effective

diffusion path-length in the model. It is concluded that the effects of tillage should be

considered in pesticide exposure assessments using macropore flow models, especially when

the compounds are applied in connection to seedbed preparation. Significantly different

parameterisations of version 4.3 and 5.0 of MACRO gave very similar predictions. This was

attributed to small differences between the models in process descriptions (e.g. soil hydraulic

functions) and numerical solution methods, combined with the effects of subjective choices

made during an inverse parameter optimisation procedure that was insufficiently constrained

by data. More research is needed to identify robust parameter estimation methods and data

requirements for macropore flow models.

6.3.5 Overall conclusions

In the validation tests conducted, no marked improvement of model performance was

observed compared to the existing FOCUS version of MACRO, although significant

differences in parameterisations were found. This suggests that despite the fact that these

three datasets must be considered unusually comprehensive, the model parameterisation was

still insufficiently constrained by measured validation data, resulting in ‘ill-posed’ problems.

More research is needed to identify the data requirements for reliable inverse estimation of

model parameters related to macropore flow.


However, the new version of MACRO (5.0) still represents a considerable improvement

compared to the old version of the model, in terms of i.) the availability of new process

descriptions such as tillage and kinetic sorption. For example, the simulations for Lebrija

showed that tillage is an important mechanism in reducing the impact of preferential flow on

leaching in heavy clay soils for herbicides applied at emergence, ii.) improved numerical

accuracy and useability. The model runs much faster, which in turn enables advanced,

automated, methods for parameter estimation.

6.4 Validation of PELMO

6.4.1 Introduction

The FOCUSPELMO model was tested against four leaching data sets from four EU countries

(Andelst, Bologna, Lanna and Vredepeel). Two of the datasets were clay soils (Andelst and

Lanna) and the other two were a sandy loam (Bologna) and a sand (Vredepeel). Two partners

ICAA and Fraunhofer were involved in the testing.

FOCUSPELMO was also tested against 2 volatilisation datasets from two EU countries

(Bologna and Jülich). The tests were performed by the same partners (ICAA and Fraunhofer).

Due to the capacity approach in FOCUSPELMO there are limited possibility calibrating

moisture profiles. Therefore, the stepwise procedure as described in Section 6.1 (blind

validation level 1, blind validation level 2, calibration) was not followed completely for the

leaching data sets but stopped before the final calibration step. However, some examples are

given to demonstrate the immense effect even a limited calibration may have on the parameter

model efficiency. Instead of the final calibration step, simulations were performed using the

new APECOP-development where preferential flow and improved volatilisation routines have

been implemented.

6.4.2 Andelst

The soil temperatures at different soil depths were excellently simulated (model efficiency EF

between 0.90 and 0.95). Also the moisture profiles were modelled satisfactory (EF = 0.20,

table 6.8).


Table 6.8. EF’ s for different steps in the validation procedure for FOCUSPELMO and the Andelst dataset.blindvalidationlevel 1

blindvalidationlevel 2

new PELMOdevelopment

soil temperatures at 5 cm, 50 cm, and 100 cm 0.90 to 0.95Moisture profiles 0.20 - 0.20Bromide concentrations in drain water -15.98 - -12.10Bentazone concentrations in drain water -0.18 -3.05 -1.72 (-0.05*)Imidacloprid concentrations in drain water - -0.46 0.16(* Simulation using the blind validation level 1 data set for bentazone)

Due to the soil type, preferential flow is a dominating process at Andelst. However,

FOCUSPELMO is not able to consider this fast transport process. Consequently it was not

able to simulate the initial substance peaks in the drain pipes in April 98. After implementing

a preferential flow option into PELMO the situation significantly improved as the

comparisons of the model efficiencies between original and new model version demonstrate.

The advantage of considering preferential flow in PELMO was significantly bigger for the

pesticide simulations than for the simulation with the tracer.

Figure 6.10. PELMO simulations on ANDELST

Unfortunately bentazone declined considerably faster in soil than calculated on the basis of

laboratory-measured half-lives. Consequently the model efficiencies are extremely low for the

2 PELMO simulations where this laboratory data was considered (-3.05 and -1.72 for the

original model and the new development). However, based on the input data used for level 1

the new development is able to do a much better job. The improvement of the simulation

quality is really obvious when looking at the imidacloprid concentrations in drain water. The

original model simulated no imidacloprid in the drains for the whole simulation period

Andelst: Imidacloprid in Drainage

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

200 250 300 350 400 450 500Days after Jan 1, 1998

c (µ

g/L)

PELMO Drain 1 Drain 2 PELMO pref.flow


whereas the new development was able to mimic the maximum concentrations around day

250 as well as the total shape of the concentrations in excellent quality. This simulation result

is confirmed by the change in the model efficiency (FOCUSPELMO: -0.46, new

development: 0.16) and are supported by an excellent description of the imidacloprid

concentrations in the soil profile.

It can be concluded that PELMO could simulate the fate of imidacloprid better than the fate of

bentazone mainly because the laboratory measurements could not explain what happened in

the field. However, PELMO had difficulties simulating the fate of bromide in the experiment.

This remained a problem also the implementation of the new preferential flow routines.

6.4.3 Lanna

No blind simulations were performed for this data set as the experimental has been published

earlier. As already found in other validation studies the moisture profiles are simulated badly

by PELMO. However, obviously the soil moisture profile had no big influence on the

performance of the other parameters. Especially, the concentration profile of bentazone was

simulated really well (see table 6.9).

Table 6.9. EF’ s for FOCUSPELMO and the PELMO improvement for the Lanna dataset.FOCUSPELMO

new PELMOdevelopment

moisture profiles -1.36bromide concentrations in soil 0.12bentazone in soil 0.76bromide concentrations in drain water -3.67 -1.02bentazone concentrations in drain water -1.82 -1.64

Figure 6.10. PELMO bromide simulations on Lanna

Lanna: Bromide concentration in drainage

0

2

4

6

8

10

12

14

16

0 50 100 150 200 250 300 350 400 450

Time (d)

c (m

g/L)

Exp. FOCUSPELMO PEMO new


The model efficiencies summarised for the drain water concentration seem to mislead the

quality of the simulations as the respective graphical representation of the bromide

concentrations demonstrate that the PELMO simulations could model the situation at Lanna

quite well. Also the respective bentazone simulation was not as bad as the model efficiency of

-1.82 for FOCUSPELMO and -1.64 for the new development may indicate.

Figure 6.11. PELMO bentazone simulations in the drainwater for the Lanna site

It can be concluded that PELMO could not describe the bromide profile in soil adequately,

but showed a reasonable performance when estimating bromide concentrations in the drains.

Bentazone profiles were excellently modelled already with the original FOCUSPELMO.

However, only the new preferential flow development was able to simulate bentazone in the

drains shortly after application even though the concentration shape was not perfect.

6.4.4 Bologna

Similar as for the previous data sets no calibration of soil moisture was performed. The model

efficiency as shown in table 6.10 indicates that the soil moisture was only poorly simulated by

PELMO.

Table 6.10. EF’ s for different steps in the validation procedure for FOCUSPELMO and the Bologna dataset.Blind

ValidationLevel 1

blindvalidation

level 2

"calibration"

moisture profiles -3.27Aclonifen concentrations in soil -148.51 -2.94 0.66Ethoprophos in soil -7.39 0.04 0.90

Lanna: Bentazone in drainage

0

50

100

150

200

250

0 50 100 150 200 250 300 350 400 450

Days after Oct. 12, 1994

c (µ

g/L)

Exp. PELMO Pref.flow PELMO


However, one should be careful in evaluating the quality of simulations only by the parameter

model efficiency as explained by the following example: The model efficiency indicates an

immense improvement of the simulation quality of the pesticide profiles when comparing the

different simulation levels. However, the only thing that was modified when moving from

level 2 to the "final calibrated version" was an adaptation of the application rate using the

actual concentration in soil to correct for the volatilisation losses at the beginning of the

experiment. No other pesticide or scenario parameter has been changed. The shape of the

level 2 and the calibrated simulation was identical.

However, in either case, it can be concluded that PELMO could simulate both the aclonifen

and the ethoprophos profiles in soil well. During the experiments no pesticide could be

detected in the ground-water. That was also simulated by PELMO.

It can be concluded that FOCUSPELMO could simulate the fate of movement of aclonifen

and ethoprophos reasonably well. However, it can be questioned whether that can be

considered as a hard test for assessing ground-water concentrations. As the original

FOCUSPELMO could simulate the situation at Bologna satisfactory new simulations with the

new development was performed for this data set.

6.4.5 Vredepeel

No blind simulations could be performed for this data set as the experimental has been

published earlier. The results demonstrate excellent simulation results for all parameters.

However, the simulation of the concentration profiles were all based on different type of

calibration work such as adapting the plant uptake factors (bromide) or application rates

(ethoprophos to calibrate volatilisation at the beginning of the experiment).

Table 6.11. EF’ s for different steps in the validation procedure for FOCUSPELMO and the Vredepeel dataset.FOCUSPELMO

soil temperatures at 2.5 cm 0.93moisture profiles 0.69bromide concentrations in soil 0.95bentazone concentrations in soil 0.96ethoprophos in soil 0.95

It can be concluded that, although PELMO was able to describe the situation at Vredepeel

excellently, it cannot be considered a hard model test as pesticide concentration in drains

were not monitored.


6.4.6 Conclusion

The results showed that the soil temperatures routine used by PELMO is working very well

without any type of calibration. However, soil moisture were often weakly estimated.

Calibration of the driving variables may have improved the situation slightly but was not

performed as the effect on the pesticide parameters was relatively small. The original

FOCUSPELMO could describe the experiments performed at Vredepeel and Bologna

excellently after calibration of the application rate to correct for the volatilisation losses at the

beginning of the experiments. However, preferential flow was not the dominant process at

these locations. The situation was completely different for the other field data sets at Lanna

and Andelst. Only the new development of PELMO in which a preferential flow module was

implemented was able to simulate the situation in these soils correctly after calibration.

6.5 Intercomparison and assessment of the progress of the

validation of the models

6.5.1 Introduction

The validation status of the three models was on 22 cases, using 3 models, 7 datasets and 6

model users. An overview is given in table 6.12.

Table 6.12. Number of teams involved in the exercises of evaluation divided by dataset and models.Data set PEARL PELMO MACROVredepeel (NL) 2 2Lanna (SE) 1 2 1Brimstone (UK) 1Bologna (IT) 2 2Coria (SP) 1Lebrija (SP) 1 2Andelst (NL) 1 2 2

For the new datasets (Bologna, Coria, Lebrija and Andelst) the protocol defined in section 6.1

was followed. For existing datasets (Lanna, Vredepeel, Brimstone) the blind simulation step

level 1 was omitted. In this section, the modelling results will be compared.

6.5.2 Results

In table 6.13, the EF values are reported for PEARL, MACRO and PELMO models calculated

for the Andelst data set by 3 different teams. These results show a sufficient level of fit for


soil water content after calibration for all models. The agreement was also acceptable for the

blind simulation. For the MACRO model, the blind level 1 run was not performed. For the

chemical transport, the simulations are quite different depending on the type of solute and

model. Generally, imidacloprid was poorly simulated by all models, while bentazone, which

is a mobile, non-persistent pesticide, showed an acceptable fit after calibration. Simulation of

solute transport with the PELMO model did not improve after calibrating the hydrological

component of the model, and so only the blind simulations are reported. This is probably

because preferential flow dominates water flow at Andelst and the model is not able to

simulate this process. Similarly, calibration of another chromatographic model, PEARL, did

not improve predictions of ground-water concentrations for bromide (from 0.2 to -0.34) and

imidacloprid (from -1.30 to -3.5). The new version of the PELMO model improved the

simulation of imidacloprid (EF from -0.46 to 0.16) in drain water but was not able to improve

the simulations of the behaviour of bromide and bentazone.

Table 6.13. Model efficiencies for PEARL and MACRO models using Andelst data setSoil moisture PEARL MACRO PELMOBlind validation level 1 -0.70 0.20Blind validation level 2 -0.90 0.06 0.23Calibration 0.49 0.17 0.23

conc. in drain flow conc. in ground-waterPEARL MACRO PELMO PEARL MACRO PELMO

Blind validation level 1Bromide -14.09 -1.59 -20.14 poorBentazone -0.19 -0.18 poor -0.73Imidacloprid -0.81 poor -1.30 -1.88Blind validation level 2Bromide -3.82 -5.58 -0.90 -1.72Bentazone -0.21 -0.18 -3.05 -2.60 -0.53Imidacloprid -0.80 -4.73 -0.46 -1.30 -0.87CalibrationBromide -1.86 -0.48 -0.34 -3.19Bentazone -0.17 0.63 -0.32 -0.39Imidacloprid -0.76 -9.01 -3.5 -0.64

Tables 6.14 and 6.15 report results obtained using PEARL, MACRO and PELMO using

Lanna and Bologna datasets. PELMO is unable to adequately simulate soil moisture

behaviour but simulates the solutes well, confirming that soil moisture does not dominate

chemical fate. Instead, PEARL shows a good fit at Lanna, where preferential flow occurs, but


a bad fit at Bologna, which has a Mediterranean climate. The PEARL simulations of water

and solute profiles at Lanna are almost as good as the macropore flow model MACRO, which

reinforces the fact that preferential flow can only be reliably detected from flux

measurements, and may not significantly alter either water content or resident concentration

profiles (Jarvis, 1999), at least within the bounds of spatial variability usually found in the

field. However, although not shown here, PEARL was, as expected, poorer at simulating flux

concentrations in drain-flow compared to MACRO, missing the initial rapid breakthrough to

the drains.

Table 6.14. EF’ s for PEARL, MACRO and PELMO models using Lanna data set. Only the best results arereported.

PELMO New PELMO PEARL MACROSoil moisture profiles -1.36 0.81 0.88Bromide soil profiles 0.12 0.69 0.66Bentazone soil profiles 0.76 0.52 0.90Bromide in drain water -3.67 -1.02 0.69Bentazone in drain water -1.82 -1.64 0.57

Table 6.15. Model efficiency values for PEARL and PELMO models using Bologna data set. Only the bestresults (modeller and protocol level) are reported.

PELMO PEARLSoil moisture -3.27 -2.80Aclonifen 0.66 0.53Ethoprophos 0.90 0.90

In table 6.16, we report data obtained using an existing data set from Vredepeel (Vanclooster

et al., 2000).

Table 6.16. EF’ s values for PEARL and PELMO models using Vredepeel data set. Only the best results arereported.

PELMO team 1 PELMO team 2 PEARLSoil moisture 0.69 0.20 0.82Soil temperature 0.93 0.93 0.87Bromide 0.95 0.92 0.94Bentazone 0.95 0.96 0.89Ethoprophos 0.95 0.83 0.78

At this site, water, heat and solute behaviour are well simulated, as shown by the EF values.


In table 6.17, we report data obtained in Lebrija.

Table 6.17: Modelling efficiency (EF) for volumetric water content and pesticide content in Lebrija usingMACRO model.

EF for volumetric soil water contentDepth (cm) Macro 4.3 Macro 5.0Moisture 0.50 0.47Chloridazon content 0.70 0.78Lenacil content 0.25 0.76

Both versions of MACRO gave largely acceptable simulations of bromide leaching, field soil

water contents, and the fate of the two pesticides (chloridazon and lenacil) in Lebrija clay soil

(table 6.17). Macropore flow was not apparent in these experiments, and only limited leaching

occurred. Significantly different parameterisations of version 4.3 and 5.0 of MACRO gave

very similar predictions. This was attributed to small differences between the models in

process descriptions (e.g. soil hydraulic functions) and numerical solution methods, combined

with the effects of subjective choices made during an inverse parameter optimisation

procedure that was insufficiently constrained by data.

Brimstone farm dataset was simulate only with PEARL model. The simulation was done after

van Genuchten parameters calibration using measured moisture profiles and drain flow. The

EF for daily drain flow was 0.44. Measured fluctuations of the ground-water table were

considerably larger than simulated fluctuations. Measured concentration peaks of isoproturon

in drain water occurred several weeks earlier than predicted by PEARL (probably due to

preferential flow). The results indicate that PEARL is unable to describe pesticide leaching

adequately in this clay soil.

6.5.3 Conclusion

These results show that the validation status of the three FOCUS model is quite variable

depending on the scenario simulated. No large differences are found between models,

although Richards equation based models are better able to simulate water behaviour. Prior

knowledge of real data improves the performance of models. The models predict the soil

resident concentration accurately, but not the flux to ground-water. If the models are tested for

their predictive ability (predictions of 0.1 mg/l without specific information on scenario) their

performances are still poor. But if the models are used as first tier of exposure modelling their

ability could be considered acceptable. More research is needed to identify the reasons for


this: are process descriptions in models not good enough, or a lack of data to enable

appropriate parameterisation ? The improvements in model performance that can be

demonstrated following physically justifiable calibration, suggest that it is mostly the latter.

Thus, comprehensive datasets, such as those reported here are required for proper validation

of FOCUS PEC models.


References

Abbaspour, K.C., van Genuchten, M.T., Schulin, R., Schläppi, E., 1997. A sequential uncertainty domaininverse procedure for estimating subsurface flow and transport parameters. Water Resources Research, 33:1879-1892.

Anderson, M.P. and Woessner, W.W.,1992: The role of postaudit in model validation. Advances in WaterResources, 15, 167-173.

Bouraoui, F., J. Boesten, N. Jarvis and G. Bidoglio. 2003. Testing the PEARL model in the Netherlands and inSweden. Proceedings XII symposium Pesticide Chemistry, Piacenza, Italy.

Cuevas, M.V., Calderón, M.J., Fernández, J.E., Hermosín, M.C., 2002a. Degradación de cloridazona y lenaciloen un suelo arcilloso del suroeste de España. In: Programa y resúmenes del XIII Congreso Científico delINCA. 12-15 de noviembre de 2002, San José de las Lajas, Cuba, p. 153. 4 pp, in CD-ROM.

Cuevas, M.V., Calderón, M.J., Fernández, J.E., Hermosín, M.C., Moreno, F., Cornejo J., 2001. Assessing herbicideleaching from field measurements and laboratory experiments. Acta Agrophysica. 57, 15-25.

Cuevas, M.V., Fernández, J.E., Calderón, M.J., Hermosín, M.C., Moreno, F., Cornejo, J., 2002b. Adsorption anddesorption of chloridazon and lenacil in a soil with a sugar beet crop. In: Proc. VII Congress European Soc.of Agronomy, Córdoba, Spain, 15-18 July, pp.351-352.

Cuevas, M.V., Fernández, J.E., Roulier, S., Calderón M.J., Hermosín, M.C., Stenemo, F., Larsbo, M., Jarvis N.2003b. Comparing macro 4.3 with 5.0 for simulating the fate of chloridazon and lenacil in a clayey soil ofsouthwest Spain. In: Proc. Of the XII Symposium Pesticide Chemistry. Piacenza, Italy, 4-6 June.

Cuevas, M.V., Hermosín, M.C., Calderón, M.J., Fernández, J.E., Velázquez, I., Lozano M.T. 2003a. Degradationand adsorption of chloridazon and lenacil as affected by temperature. In: Proc. Of the XII SymposiumPesticide Chemistry. Piacenza, Italy, 4-6 June.

Doherty, J. 2000. Visual PEST (user’ s manual). Watermark Numerical Computing. Corinda, Australia.FOCUS, 2000. FOCUS ground-water scenarios in the EU plant protection product review process Report of the

FOCUS Ground-water Scenarios Workgroup, EC Document Reference Sanco/321/2000, 197ppGarratt J.A., Capri E., Trevisan M., Errera G., Wilkins R.M. (2002). Parameterisation evaluation and comparison

of pesticide leaching models to data from a Bologna field site, Italy. Pest Management Science, 58, 3-20.Jarvis, N.J. 1999. Using preferential flow models for management purposes. In: Proceedings of the International

Workshop ‘Modelling of transport processes in soils at various scales in time and space’ , Leuven, Nov.1999, 521-535.

Larsson, M.H., Jarvis, N.J., 1999. Evaluation of a dual-porosity model to predict field-scale solute transport inmacroporous soil. Journal of Hydrology, 215: 153-171.

Loague, K.M. and R.E. Green. 1991. Statistical and graphical methods for evaluating solute transport models:overview and application. Journal of Contaminant Hydrology 7, 51-73.

Loague, K.M., R.S. Yost, R.E. Green, T.C. Liang. 1989. Uncertainty in a pesticide leaching assessment forHawaii. Journal of Contaminant Hydrology 4, 139-161.

Oreskes, N., Shrader-Frechette, K. Belitz, K. ,1994: Verification, validation and confirmation of numericalmodels in earth sciences. Science, 164, 641-646.

Refsgaard, J.C. and Storm, B. ,1996: Construction, calibration and validation of hydrological models. In: Abbott,M.B. Refsgaard, J.C. (Eds): Distributed Hydrological Modelling, Kluwer Academic Publishers, 41-54.

Rykiel, E.J. 1996. Testing ecological models: the meaning of validation. Ecological Modelling 90, 229-244.Scorza Junior, R.P. 2002. Pesticide leaching in macroporous clay soils: field experiment and modeling. Ph D

Thesis, Wageningen University, 230 pp.Styczen, M., 1995: Validation of pesticide leaching models. In: Leaching models and EU registration. Final

report of the FOCUS work group. DOC. 4952/VI/95.Thorsen, M., Jørgensen, P.R., Felding, G., Jacobsen, O.H., Spliid, N.H., Refsgaard, J.C, Evaluation of a

stepwise procedure for comparative validation of pesticide leaching models. Journal of EnvironmentalQuality 27 (5) : 1183-1193.

Vanclooster M., J. Boesten, M. Trevisan, C. Brown, E. Capri, O.M. Eklo, B. Gottesbüren, V. Gouy A.M.A.van der Linden, 2000. A European test of pesticide-leaching models: methodology and majorrecommendations. Agricultural Water Management 44: 1-21.


Walker A., Calvet R., Del Re A.A.M., Pestemer W., Hollis J.M. ,1995. Evaluation and improvement ofmathematical models of pesticide mobility in soils and assessment of their potential to predict contaminationof water systems. Biologischen Budensanstalt fur Land- und Forstwirtschaft, Berlin-Dahlem, pp. 115.


7. PEC ground-water scenario evaluation

7.1 Introduction

In contrast with the large number of publications available on model validation, only few

studies dealt with the validation or evaluation of scenarios. However, the efficiency of a PEC

ground-water assessment in a Tier 1 exercise will depend on both the validity of the PEC

ground-water model and the correctness of the scenarios. As already reported in the

introduction, only a limited number of ground-water scenarios were, for pragmatic reasons,

identified in the FOCUS PEC ground-water procedure. Given the limited availability of Pan-

European environmental data at the start of the FOCUS activity, a statistical approach to infer

optimal scenarios could not be adopted. The present PEC calculations may therefore be biased

in representing worst case scenarios, given the under-sampling of the population of European

soil types, climatic conditions and agricultural practices. Therefore, it is of paramount

importance to make progress in the validation of the ground-water scenarios and to quantify

and reduce this bias.

A scenario in the context of the FOCUS procedure is defined as a representative combination

of crop, soil, climate and agronomic parameters to be used in modelling. In this context

representative means that the selected scenarios should represent physical sites known to

exist, i.e. the combination of crop, soil, climate and agronomic conditions should be realistic

(FOCUS, 1995). For PEC ground-water calculations, FOCUS intends to construct scenarios

that represent an overall vulnerability approximating the 90th percentile leaching of all

possible situations (this percentile is often referred to as a realistic worst case) (FOCUS,

2000). They further assumed that the vulnerability was equally attributed to soil and climate.

To achieve this, they first defined nine so-called FOCUS areas. Within each of these areas,

they selected an approximate 80% vulnerable soil, which implies that the concentration of a

PPP should be less than the EU drinking water limit in at least 80% of the corresponding area.

Hence, a FOCUS ground-water scenario is a combination of parameter values selected in such

a way that the leaching concentration calculated with a FOCUS PEC ground-water model

equals the 90th percentile of leaching inside the corresponding FOCUS area. It was the

objective to validate this scenario definition statement in the APECOP project.

The key question to answer is: ‘Are FOCUS scenario combinations of parameter values

selected in such a way that when used in combination with a FOCUS PEC ground-water

model, the calculated leaching values correspond to the real 90th percentile of leaching?’ . To


answer this question, it is essential to define the ‘real’ 90th percentile of leaching. In an ideal

situation, this 90th percentile could be obtained from a detailed monitoring of the presence of

active substance in ground-water. However, our observational skills are limited in time and

space, and therefore the ‘real’ 90th percentile is unknown. If the 90th percentile of leaching

cannot be obtained from direct measurements, an alternative is offered by approximating

these leaching values using a ‘detailed’ assessment technique. This technique comes down to

approximating the ‘real’ 90th percentile of the leaching concentration by means of a spatially

distributed leaching model in combination with Pan-European soil, climate and agricultural

databases. Obviously, this type of validation has a lower power than a validation in which the

90th percentile of leaching is estimated from direct measurements, but it is the only pragmatic

way to proceed with scenario validation at this time. Given the computational burden

associated with the use of spatially distributed modelling technique, the validation could only

be applied to a limited number of pesticides and to two major agricultural crops (maize and

winter wheat).

Simulations were performed by means of a spatially distributed model (i.e. the EuroPEARL

model) for 1062 unique combinations of soil type, climate and country, sampled throughout

Europe. Soil properties, including soil horizon designations, were obtained from the Soil

Profile Analytical Database of Europe. Daily weather data were obtained from the MARS

database. Other data, such as irrigation data, crop data and pesticide properties, have been

compiled from various sources, such as inventories, field-studies and the literature. The 1062

unique combinations together represent at maximum 75% of the total agricultural area of the

European Union. Austria, Sweden and Finland could not be included in the simulations,

because there was insufficient soil profile information for these countries. However, to

consider additional soil-climate and crop combinations for which no profile data where

available, a metamodel was derived by interpolating EuroPEARL modelling results in the

input parameter space using radial basis functions neural networks. The combination of the

process based deterministic EuroPEARL model with the metamodel allows: (i) the use (for

leaching simulations) the European 1:1,000,000 soil map, which covers 97% of the European

agricultural area instead of the 75% covered by the profile database; (ii) the consideration of

spatial variability of leaching inside the mapping units; and (iii) a statistical test of the validity

of the scenarios.


The aim of this chapter is (i) to present the methodology that was adopted for validation of the

FOCUS ground-water scenarios, and (ii) to present the main results.

7.2 Methodology for evaluating the FOCUS ground-water scenarios

For assessing predicted environmental concentrations (PECs) of plant protection products

(PPPs) in ground-water, the FOCUS ground-water scenarios working group published nine

standard scenarios which, in combination with the FOCUS leaching models, implement a

harmonised risk assessment procedure at the Pan-European scale (FOCUS, 2000). It is

supposed, in the registration procedure, that realistic worst cases at the European scale are

represented by the standard scenarios. This realistic worst case was identified with the

concept that scenarios should correspond to 90th percentile vulnerability situations (FOCUS,

2000).

Unfortunately, due to limited data availability, expert judgement was considered to

characterise some of the properties of the standard scenarios. However, to assure the quality

of the PEC assessment, it is of paramount importance to reduce the potential bias in such a

procedure. In addition, if significant bias were to be present in the proposed scenarios,

corrections to them would need to be proposed.

In contrast to the number of validation studies for PEC exposure models (see e.g. Vanclooster

et al., 2000 and chapter 6), few, if any, validation studies for PEC exposure scenarios are

presented in the literature. Therefore, no methodological approach is presently available to

validate exposure scenarios in a transparent way. The objective of this section is to present

such an approach. Due to the novelty of the method, a theoretical background is presented

here, while the results of the validation are presented later on.

7.2.1 Multiple aspects of scenario evaluation

Standard scenarios are complex combinations of different soil, climate and land use attributes

and may therefore conform to the envisaged use in some aspects, but not in others. The

objective of the evaluation of scenarios is therefore to demonstrate the quality of a scenario in

its multiple aspects in a transparent way. Scenario evaluation may result in suggestions to

improve the scenario and to reduce the uncertainty in the final assessments.

Given the complexity and the multiple characteristics of a scenario, an ‘accept or reject’

approach is inappropriate. Rather, a multi-criterion approach was pursued to assess the

following aspects of the scenario:


1. Relevance. Do the scenarios correspond to the use for which they have been constructed?

2. Consistency. Is the definition of the scenario coherent and logical?

3. Representativeness. Is the scenario indeed representative for realistic worst case

conditions?

4. Realism. Do places exist where the soil and climate are well described by the scenarios

parameters, as compared with measurements? Do the scenarios describe existing

practices?

For the case of the FOCUS ground-water scenarios, it was considered that the ‘realism’ of the

scenarios was not an issue. Indeed, scenarios were defined by experts and it is very likely that

the conditions of the standard scenario may occur somewhere in Europe. For testing the

‘relevancy’, scenario conditions must be coherent with the envisaged use. To illustrate the

‘relevancy’ issue consider the following example. Scenarios have been developed as part of

the implementation of the 91/414 directive at the EU level. The conditions of the standard

scenarios inferred from western European data bases and experiences, will not be relevant for

evaluating PPP emission in the tropical areas of the ‘Département d’ Outre-Mer’ of France

which also belongs to the EU. This illustrates that the scenarios are not relevant to all

situations. For testing the ‘consistency’ , it should be analysed if the scenario textual

description respect the rules of a reference logic. Semantic techniques such as propositional

calculus (Gries and Schneider, 1993) could be applied to analyse the consistency. However,

we can give again one example: the scenarios are supposed to represent arable land, but

grasslands are parameterised, most of them located in non-arable land.

Unfortunately, due to time constraints, the ‘consistency’ and ‘relevancy’ aspects of the

scenarios could not be evaluated in APECOP. Therefore most attention was paid to the

evaluation of the ‘representativeness’ of scenarios.

7.2.2 Scenario representativeness definition

Representativeness is defined in a terms used in set theory, and particularly the extension of

the Boolean logic to symbolic objects (Bock and Diday, 2000). In this context, scenarios are

defined by their ‘intension’ (i.e. an ensemble of properties, attributes and operators fully

defining the objects represented by the scenario). The scenario ‘extension’ is the set of objects

corresponding to the properties and attributes defined by the scenario intension. The extension

is a subset of a extension universe. It is useful to give an example: if a scenario represents

places in Europe with an altitude above 1000 m, then a object is a ‘place’ on the Earth

described by its coordinates and altitude. The extension universe is constituted by the set of


all places on Earth as depicted in a chosen database, while the intension is ‘places in Europe

with altitudes over 1000 m’ . The extension are the objects in the chosen database for which

the altitude is greater or equal to 1000 m. The extension functions are the functions,

operations, models, etc, needed to decide if an object from the extension universe is part of

the extension. In the given example, the extension function can be the GIS software used.

Using those definitions for evaluating the representativeness of a scenario, we compare a

‘tested extension’ (TE) with a ‘reference extension’ (RE). This conceptual framework is

presented in figure 7.1, for the case of the FOCUS scenarios.

FOCUSscenariointension

All possibleleachingconcentrations

testedextension(TE)

referenceextension(RE)

Extension function1:FOCUS exposuremodel

Extension function2:MM exposuremodel

Extension function 2:Reference exposure model

Extension universe:all leaching concentrationsin the EU

Figure 7.1. Conceptual framework of the scenarios representativeness estimation. The estimation is based on thecomparison of a tested extension (TE) with a reference extension (RE). The tested extension is the set of worst-cases concentrations calculated using the FOCUS scenarios and models, while the reference extension is the setof worst cases concentrations obtained using a reference exposure model.

7.2.3 Determination of the FOCUS scenarios intension

The FOCUS scenario intension is determined by an interpretation of the FOCUS report

(FOCUS, 2002). Former and current FOCUS members are part of the APECOP project,

ensuring that the interpretation of the FOCUS scenarios is close to the actual use of the

scenarios. The ‘intension’ of the FOCUS ground-water scenarios, as interpreted by APECOP,

can be formalised as follows:

‘Places (i) inside the major European agricultural arable areas (ii) where

realistic worst cases of pesticide leaching occur and (iii) where the soil and

climate have characteristics as defined by a set of chosen parameters. A worst

case of pesticide leaching is equal or greater than the 90th percentile of

leaching’.


7.2.4 Determination of the extension universe

The extension universe is determined taking into account that the intension has a geographical

component (European arable areas), so the database must be georeferenced; the objects of the

database are “places”. Additionally, the intension defines objects by the leaching occurring at

a given place. In consequence, the extension universe database must be a European map of

leaching, i.e. a collection of objects with two characteristics: location and leaching. The

leaching map that has been chosen is a raster map with pixel size of 10x10 km2. Leaching is

defined at 2 m depth .

7.2.5 Choice of the extension functions

The representativeness estimation is dependent on the extension functions 1 and 2 used

(figure 7.1). In the present case the extension function 1 is the FOCUS procedures and

models. Extension function 2 must be chosen to be a more accurate modelling procedure than

the FOCUS procedure. APECOP has implemented two complementary modelling techniques

to be used as extension function 2: the EuroPEARL numerical model (NM) approach and the

Metamodel (MM) approach.

The extension functions are needed to obtain the extension universe database, i.e. the map of

leaching in the European Union, and the reference extension (the set of worst-case leaching

concentrations). PEC ground-water models are the chosen extension functions. Obviously,

this type of evaluation has a lower level of security than a validation in which the reference

leaching concentration is estimated from direct measurements. It is expected that through the

implementation of the European water framework directive, reference monitoring data will

become available and that direct comparison with measurements using techniques such as

presented by Worrall (2002) may become possible in the future. In consequence, leaching

models are used within the APECOP framework.

7.2.6 Reference extension determination

Using either the metamodel or EuroPEARL, two distinct reference extensions are obtained:

RE_MM and RE_NM. This means that two different leaching maps are obtained, and that two

different representativeness estimations will result from the validation exercise.Since the

intension defines that the scenarios represent places with leaching equal or greater than the

90th percentile, RE_NM and RE_MM are sets of leaching values, extracted from the leaching

map, and equal or greater than the 90th percentile.


7.2.7 Representativeness estimation

As depicted in figure 7.1, the representativeness (REPR) is estimated by the comparison

between TE (the leaching as predicted by means of the PPP FOCUS exposure model and

scenarios) with RE_MM or RE_NM, the ensemble of worst-cases concentrations occurring

inside major agricultural arable areas. In consequence two different estimations of the

representativeness are obtained: REPR_MM and REPR_NM, as calculated using either MM

or NM.

The EuroPEARL approach leads to a deterministic estimation of the leaching, with one value

of leaching for each pixel of the map. In consequence, REPR_NM is a binary value: 1 if the

leaching calculated by FOCUS is included into RE_NM and zero otherwise:

NMRETEif=NMREPR

NMRETEif=NMREPR{

_0__1_

⊆/

⊆ (7.1)

The metamodel approach leads to a probabilistic estimation of leaching, with a pdf for each

value of leaching; in consequence, the representativeness is a probability value:

)_(_ MMRETEPMMREPR ⊆= (7.2)

In other words, to evaluate the representativeness, we compare the 90th percentile of leaching

concentration of the ‘reference extension’ obtained by a “reference” assessment technique

(i.e. p0.9MM or p0.9

NM ) with the 90th percentile of leaching concentration obtained with a

FOCUS scenario and FOCUS model, i.e. p0.9FOCUS. In consequence, eq. (7.1) and (7.2) can be

translated as follows:

)(_ 9.09.0MMFOCUS ppPMMREPR ≥= (7.3)

i.e. the probability that the FOCUS scenario is strict enough, if the metamodel is used as

reference, and

)(_ 9.09.0NMFOCUS pptruthvalueNMREPR ≥= (7.4)

i.e. a Yes or No answer to the question ‘Are FOCUS scenarios strict enough?’ , when the

numerical model is used as reference.

7.3 Assessment of Pesticide Leaching at the Pan-European Level

using a Spatially Distributed Model (EuroPEARL)

This section describes the implementation of a Pan-European spatially distributed PPP

leaching model, referred to as EuroPEARL. Spatially distributed leaching models consider the

variability of environmental and land use properties in an explicit way to assess large scale

exposure. Such models provides the user with maps of the predicted leaching concentration.


These maps contain a wealth of additional information, particularly high and low risk areas.

Also frequency distributions and percentiles of the predicted leaching concentration can be

directly inferred from these maps. In addition, bias due to undersampling, as is the case when

using a limited range of standard scenarios, will be reduced. Earlier efforts to model spatial

patterns of pesticide leaching were reported for individual member states (Tiktak et al.,

1996ab, 2002a) or regions (e.g. Capri et al., 2000).

The implementation of this model has become possible after the release of a series of Pan-

European environmental and land-use data bases, such as the European Soil Map, the

European Soil Database (Madsen-Breuning and Jones, 1995), and the Pan-European Climate

Database (Vossen and Meyer-Roux, 1995). The objective of this section is to present and

evaluate the methodology which was used to implement the spatially distributed leaching

model and to illustrate the possibilities of the model to assess the PPP leaching risk at the

Pan-European level. Results will be discussed on the basis of four theoretical plant protection

products with different physico-chemical properties. The representativeness of the FOCUS

ground-water scenarios will be discussed on the basis of frequency distributions of the

predicted leaching concentration.

7.3.1Materials and methods

The local scale exposure model

An appropriate PPP leaching model meets the following two criteria: (i) the model must have

sufficient conceptual resolution to be applicable to the entire area for which it is developed

and for the full range of PPPs, and (ii) the number of model inputs that cannot be directly or

indirectly (through transfer functions) be obtained from measurements should be minimised.

Tiktak et al., 2002b investigated the balance between these two conflicting demands. They

suggested that due to the large number of PPPs with different properties, a moderately

detailed leaching model should be recommended. This consideration has led to the

development of the PEARL model, which is a one-dimensional, dynamic, multi-layer model

of the fate of a pesticide and relevant transformation products in the soil-plant system (figure

7.2). The SWAP model (van Dam, 2001), which describes the flow of water in the Soil Water

Atmosphere Plant system, is an integral component of PEARL. A comprehensive overview of

the PEARL model is given by Tiktak et al. (2001) and Leistra et al. (2001).

The SWAP model (van Dam, 2001) uses a finite-difference method to solve Richard’ s

equation. The hydraulic properties of soils are described by closed form equations (van

Genuchten, 1980). The upper boundary of the model is situated at the top of the crop canopy.


Daily rainfall fluxes are input to the model; the reference evapotranspiration is calculated

according to the FAO modified Penman-Monteith approach (Allen et al., 1998). In this study

irrigation was automatically performed under the following conditions: (i) an irrigation

system should be present, and (ii) the pressure head in the rooting zone should be below a

critical value. SWAP has different options for the lower boundary condition; in this study the

ground-water level was prescribed. Interception by the crop canopy is calculated with an

empirical equation. In this equation, the amount of water intercepted asymptotically reaches

aLAI, where a (-) is an empirical parameter and LAI is the Leaf Area Index. The potential

evapotranspiration is partitioned over transpiration and soil evaporation on the basis of the

LAI. The actual soil evapotranspiration is calculated according to Black et al. (1969).

Potential transpiration is distributed over the root zone using the root density distribution as a

weighing factor. Water uptake is reduced for layers with low pressure heads or anaerobic

conditions.

precipitationirrigation

transpirationevaporation of

intercepted water

soilevaporation throughfall

fluctuatinggroundwaterlevel

soil waterfluxes

ponding

water uptakeby plant roots

seepage

lateraldischarge toditches andfield-drains

unsa

tura

ted

zone

satu

rate

d zo

ne

depositionapplication

dissipation at thecrop canopy

volatilisationwash-off

convectiondispersiondiffusion

pesticideuptake

leaching

transformationsolid-liquid

gaspartitioning

SWAPhydrology

PEARL(pesticides)

cropcalendar

heatflow

injection

surfacerun-off

Figure 7.2 Overview of processes included in the PEARL and SWAP models.

PEARL considers a soil system where PPPs and relevant metabolites reside in an equilibrium

domain and in a non-equilibrium domain. The equilibrium domain is partitioned into three

phases, i.e. and adsorbed phase, a dissolved phase and a gaseous phase. Sorption in the

equilibrium domain is described by a Freundlich isotherm. The Freundlich coefficient is


calculated from the coefficient for distributing the substance over organic matter and water.

PEARL has an option for pH dependent sorption behaviour (Van der Linden and Boesten,

2001). Sorption to the non-equilibrium sites is calculated with a first-order rate equation. The

partitioning of the PPP between the gas phase and the liquid phase is described by Henry’ s

law. The flux in the liquid phase of the soil is described by an equation including convection,

dispersion and diffusion. The flux in the gas phase is described by Fick’ s law. The

transformation of PPPs is described with a first-order rate equation and a number of reduction

factors, which account for the influence of temperature, soil moisture and depth in soil. The

uptake of PPPs is taken proportional to the root water uptake and an empirical transpiration

stream concentration factor. The initial condition for the model is defined by profiles of the

concentration of pesticide in the equilibrium and non-equilibrium domains of the soil system.

In this study, an initially pesticide free soil is assumed. PEARL has several options for

application of PPPs, i.e. spraying to the soil surface, injection and incorporation by tillage.

Derivation of the spatial schematisation

The aim of the current study is to establish the PPP leaching risk in Europe using a spatially

distributed modelling approach, i.e. EuroPEARL. A resolution of 10x10 km2 was chosen,

which is the highest justifiable resolution of the 1:1 000 000 EU soil map. Running a

comprehensive model for all relevant grid cells would require too much computation time.

Therefore, simulations are carried out for unique combinations of spatially distributed model

inputs (cf. Tiktak et al. 1996ab, 2002a). These unique combinations (here referred to as

‘plots’ ) are assumed representative for one or more grid cells within the area to be mapped.

The plots were constructed by overlaying the following three maps:

1. The 1:1,000,000 soil map of the European Union (Jamagne et al., 1995). This map

features a total number of 735 Soil Mapping Units (SMU’ s). Each map unit is an

association of Soil Typological Units (STU’ s) occurring within the limits of a discrete

physiographic entity. It is composed of a dominant soil type and of subdominant

associated soils.

2. A map showing 8 major climate zones of the European Union (figure 7.3a). The climate

zone map is based on maps of long-term averages of annual precipitation and temperature,

which were constructed using data from approximately 1500 weather stations (Vossen and

Meyer-Roux, 1995). The definition of the zones is shown in table 7.1.

3. The map showing the countries of the European Union. The country is not required in the

EuroPEARL model itself, rather it is used to guarantee that the correct soil profile is

linked to the Soil Mapping Units (see section 0).


All original maps were available digitally and were converted to raster maps with a resolution

of 10x10 km2. As discussed in the introduction, the PPP exposure assessment should apply to

agricultural areas only. Therefore, the overlay was masked with a map showing agricultural

land-use (Mücher et al., 1998). The final result was a map with 1442 relevant unique

combinations of soil type, climate zone and country. The size of the units was between 100

km2 and 19,600 km2; the average plot size was 1,037 km2 (figure 7.3b).

Table 7.1.Major climate zones of the Europe Union, based on mean annual rainfall and meanannual temperature. Classification according to FOCUS (2000).Zone number Mean annual rainfallMean annual

temperature

Surface area†

(mm) (oC) (km2)

Cold < 600 < 5 83,000

Temperate 1 < 600 5 – 12.5 36,300

Temperate 2 600 – 800 5 – 12.5 429,400

Temperate 3 800 – 1000 5 – 12.5 269,000

Temperate 4 > 1000 5 – 12.5 167,800

Warm 1 < 800 > 12.5 329,800

Warm 2 800 – 1000 > 12.5 148,400

Warm 3 > 1000 > 12.5 32,600

† Agricultural land only

Linking the soil profiles with the map of unique combinations

The Soil Profile Analytical Database of Europe (Madsen-Breuning and Jones, 1995) has been

compiled through the collaboration of national experts of the EU countries (12 at that time)

and has been extended to include data from Eastern European and Scandinavian countries.

The profile database (SPADE) contains information at the soil profile level. The soil profiles

are not georeferenced and are estimated profiles, meaning that the national experts have given

a best possible description of typical soil profiles of their countries. Their estimate is not

necessarily based on soil profile descriptions and measurements. Parameters include the full

FAO soil name (FAO-Unesco, 1990), country, dominant soil textural class and horizon

designations. At the horizon level, the database contains information on bulk density, organic

matter, pH in water, and textural distribution, which is the information required by the

PEARL model. The total number of profiles in the database amounts to 621.


Figure 7.3. Basic maps for EuroPEARL. Areas without agricultural land-use and areas where insufficient soilprofile information was available, are not shown. This implies that the Northern European countries are notrepresented. a) Climate zones. The letter T refers to ‘temperate’ , the letter W to ‘warm’ ; b) Size of the plotsexpressed in number of gridcells. Each gridcel has an area of 100 km2. c) Percentage of the area equipped forirrigation around 1995. Figure is based on data from Siebert and Döll (1995). d) Organic matter of the top meter;e) Clay content of the top meter; f) pH of the top meter. All soil properties have been obtained by combining theSoil Map of Europe with the Soil Profile Analytical Database of Europe (see figure 7.4 and text).

Due to the structure of the European soil map, it was not possible to establish a direct link

with the profile database. As mentioned before, the Mapping Units of the soil map are

associations of dominant and subdominant Soil Typological Units. The total coverage of all

individual STU’ s within one SMU is 100%. The STU’ s and not the SMU’ s are the carriers of


basic soil information such as the FAO soil name and textural class. The link between the

profile database and the soil map was therefore made in a two step approach (figure 7.4).

Soil Map of Europe

Soil Profile Analytical Database

SMUSMU number;Country;STU 1 number;STU 1 coverage;

…...STU n number;STU n coverage;

Dominant STU number1

STU number1,2

Country2;FAO Soil Name2;Texture class2;Otherparameters.

Soil profile number(STU number)2;Country2;FAO Soil Name2;Texture class2;Horizon numbers3;Other parameters

Soil Horizon number3

Horizon depth;Profile number;Organic matter;Texture;pH.

1

2

3

Figure 7.4. Link between the European Soil Map and the Soil Profile Analytical Database of Europe. Parameterswith suffix 1 have been used for the link between the SMU and the STU; parameters with suffix 2 have beenused for the link between the STU and the Soil Profile number, and parameters with suffix 3 have been used forthe link with the soil horizon database.

In a first step, the dominant STU within each SMU was determined. Then, a soil profile was

assigned to the dominant STU. This second link was made at different confidence levels. The

most reliable link could be obtained if the author of a profile has explicitly stated the

corresponding STU. If the STU was not specified, a profile was assigned on the basis of the

full FAO soil name, the textural class and the country code. By including the country code in

the query, it was assured that soil profiles from the soil profile database of a given country

could only be matched to Soil Mapping Units within that country. In those situations where it

was still not possible to assign a soil profile, the query was repeated with the full FAO soil

name and the country code only. Finally, a query was carried out using the major soil type

only (e.g. Cambisol instead of Eutric Cambisol). Using this procedure, 1062 Unique

Combinations could be assigned a soil profile, representing approximately 75% of the total

agricultural area of the European Union. Unfortunately, Austria, Sweden and Finland had to

be omitted, because there was insufficient soil profile information available for these

countries.


Model parameterisation

Parameter values were assigned to the 1062 plots described above. To avoid data redundancy,

all model parameters were stored in a relational database. The database contains a hierarchy,

as shown in figure 7.5. At the highest level, a distinction can be made between spatially

constant and spatially distributed parameters. Simulations are carried out for a single crop and

a single substance in a time, which implies that the substance name and the crop type are

stored at the highest hierarchical level. Notice that some of the crop properties are spatially

constant, while other crop properties are assumed to be climate dependent. The ‘plot’ is the

central entry for all spatially distributed model inputs (see also section 0). Long-term average

rainfall and temperature are given at the plot level. All other spatially distributed parameters

are related to climate zone or soil profile number.

RunControl;Substance;Crop;Plot.

SubstanceSubstance properties, such asthe half-live, thepartitioning coefficient andapplication type.

PlotClimate zone;Soil profile;Long-term averageweather conditions.

ControlStart and end date;Other control parameters.

Climate zoneDaily weather series;Emergence andharvest date of crops;Phenologicaldevelopment stages;Application date ofpesticides.

Soil profileSoil horizon;Groundwater level;Irrigation switch.

Soil horizonSoil physical unit;Layer thickness;Texture;Organic matter;pH.

Crop propertiesCritical pressure heads for drought stress and irrigation;Extinction coefficient for solar radiation.

Development stageLAI;Crop factor;Rooting depth.

Soil physical unitParameters of theMualem-van Genuchtenfunctions;Dispersion length

Spatially distributed variables

Daily weatherTemperature;Rainfall;Wind speed;Humidity;Radiation.

Figure 7.5. Structure of the EuroPEARL database.Weather

Within each climate zone, a single weather station was selected from the MARS database

(Vossen and Meyer-Roux, 1995). This weather station was assumed to correctly describe the

seasonal dynamics of weather conditions within the entire zone. Daily precipitation and

temperature for each plot were obtained by combining the maps of the long-term average

annual precipitation and temperature with daily records from the 9 weather stations:


stationaplotastationdplotdstationa

plotastationdplotd TTTT

P

PPP ,,,,

,

,,, and −+== (7.5)

where P is precipitation, T is temperature, and the suffixes a and d refer to daily and annual,

respectively.

Soil properties

Horizon distribution, textural distribution, pH and organic matter were taken directly from the

soil profile database (section 0; figure 7.3d-f). A continuous pedotransfer approach was used

to relate the bulk density to the organic matter content (Tiktak et al., 1996a). Parameter values

for the Mualem-van Genuchten functions to describe the soil physical properties (Van

Genuchten, 1980) were taken from the HYPRES database (Wösten et al., 1999). Because

there was insufficient data to quantify the mean ground-water depth, it was set constant at two

meters below soil surface. This was justified, because the most important model outputs

showed a very low sensitivity to the depth of the ground-water table as long as the ground-

water table was below 1 m, which is usually the case in agricultural soils (Tiktak et al.,

2002a).

Crop properties

The growth of a crop is described as a function of development stage, which ranges from zero

at crop emergence to 2 at crop harvest. The dependence of crop development on actual

weather conditions was described by making the crop development stage dependent on the

temperature sum since emergence. Emergence date, harvest date, LAI and rooting depth were

obtained from nine representative field sites, i.e. one for each climate zone (FOCUS, 2000).

Crop factors in relation to the modified Penman-Monteith approach were taken from Allen et

al. (1998). Critical pressure heads for drought stress and irrigation were taken from Van Dam

(2000). Braden (1985) analysed interception data for a number of crops and found that

parameter a in the interception equation could be set to 0.25 for most common crops.

Substance properties

Simulations were carried out for four dummy pesticides with different properties. Individual

parameter values chosen were in the range of values that can be found for registered plant

protection products in Europe but are not intended to be attributable to real compounds. A

summary of the most important pesticide properties is given in table 7.2. Properties of

substance A, B and D were taken from FOCUS (2000). Pesticide E was added, because it

shows pH dependent sorption behaviour (Van der Linden et al., 2001).


Table 7.2 Overviewof the most important properties of the pesticides considered in this study.Property1 A B D E

M (g mol-1) 300 300 300 240

Pv,s (Pa) 0 0.0001 0 0.01

Sw (mg L-1) 90 90 90 50

Kom,ac,eq (L kg-1) 60 10 35 500

Kom,ba,eq (L kg-1) 60 10 35 23

Pka - - - 4.6

DT50,ref (d) 60 (20 oC) 20 (20 oC) 20 (20 oC) 50 (20 oC)

M is the molar mass, Pv,s is the saturated vapour pressure, Sw is the solubility in water, Kom,ac,eq isthe coefficient of equilibrium sorption on organic matter under acidic conditions, Kom,ba,eq is thecoefficient of equilibrium sorption on organic matter under basic conditions, pKa is the negativelogarithm of the dissociation constant, and DT50,ref is the half-live under reference conditionsSimulation length, crops, irrigation and pesticide applications

The model was run for a 26-year period. The first six years of the simulation were considered

as “warm-up” years, which implies that the final model results refer to a 20 year period.

Simulations were carried out for two crops, winter wheat and maize. These crops were

chosen, because they are grown in almost the entire European Union. Following common

agricultural practice (FOCUS, 2000), winter wheat is not irrigated and maize is irrigated if an

irrigation system is present. The presence of an irrigation system was derived from

inventories by Siebert and Döll (1995), who present maps of the fraction of land equipped for

irrigation (figure 7.3c). In this study, it was assumed that an irrigation system is likely to be

present if more than 2.5% of the total area is equipped for irrigation.

All pesticides were surface applied at a rate of 1 kg ha-1 one day before crop emergence (pre-

emergence pesticides). This implied that the pesticides were applied in autumn in the case of

winter wheat and spring in the case of maize.

7.3.3 Results and discussion

EuroPEARL was used to obtain maps of the predicted pesticide mass flux at 2 m below soil

surface, which is the target depth considered in European legislation procedures (FOCUS,

2000). Water and substance balances were calculated for Europe as a whole, and for the

climate zones described in table 7.1. The balances were included to show the large-scale

variability of pesticide mass fluxes across Europe. The substance and water balances were

calculated for the entire agricultural area of Europe, so no consideration was given to the

actual crop area. As shown by Tiktak et al. (2002), this may affect the average water balances.

Water balances

Figure 7.6 shows the spatial patterns of the predicted 20 years average water fluxes in winter

wheat and maize. The differences between the spatial patterns of water inputs in winter wheat

and maize are remarkable. In the case of winter wheat the highest water inputs occur in north-


western Europe (particularly Ireland), which is also the wettest area in terms of precipitation.

In the case of maize, high water inputs also occur in Southern Europe, where irrigation is a

common practice (figure 7.6). Irrigation also has a large effect on the spatial patterns of

predicted actual evapotranspiration rates. In the case of maize, the highest evapotranspiration

rates occur in warm and sunny climates (Southern Europe). In winter wheat, which is usually

not irrigated, low predicted evapotranspiration rates occur in dry regions (particularly Spain,

Sicily and Greece).

Table 7.320 Years predicted average water fluxes (mm yr-1) for winter wheat and maize. Water balances arepresented for Europe as a whole and for seven climate zones as described in table 7.1.

a) Winter wheat

P I Ei Et Es Q Et,pot Es,pot

Europe 821 0 33 247 200 337 282 554

Temperate 1 650 0 39 211 208 191 248 542

Temperate 2 720 0 25 219 211 264 229 398

Temperate 3 900 0 31 258 239 371 264 426

Temperate 4 1119 0 47 238 228 605 253 410

Warm 1 644 0 36 274 121 210 418 1123

Warm 2 898 0 28 275 196 385 292 528

Warm 3 1069 0 61 330 179 494 354 571

b) Maize

P I Ei Et Es Q Et,pot Es,pot

Europe 821 117 30 335 224 346 407 492

Temperate 1 650 59 28 275 227 178 409 451

Temperate 2 720 34 32 269 203 249 300 366

Temperate 3 900 10 33 301 233 342 325 407

Temperate 4 1119 13 52 278 219 583 320 365

Warm 1 644 434 10 525 243 298 725 944

Warm 2 898 180 25 384 223 430 457 460

Warm 3 1069 185 16 330 283 618 491 580

P is precipitation, I is irrigation, Ei is interception loss, Et is transpiration, Es is soilevaporation, and Q is seepage flux. The suffix pot refers to potential.

As shown in table 7.3, the largest part of the predicted evapotranspiration deficit for winter

wheat in Southern countries can be attributed to predicted reduction of evaporation from the

soil surface. As soil evaporation rates are primarily affected by daily rainfall patterns, it can

be concluded that the unpredictable rainfall patterns in Mediterranean regions and not the

mean annual precipitation depths explain the large evapotranspiration deficits in winter wheat.

Despite all this, there is a strong resemblance between the spatial patterns of predicted

precipitation surplus for winter wheat and maize. In both cases, the largest predicted


precipitation surplus occurs in north-western Europe, Italy, and the mountainous regions of

central Europe. Predicted precipitation surplus is lowest in East England, East Germany and

Spain. Notice that there is also a strong resemblance with the precipitation map (which is the

same map as the water input map for winter wheat). Apparently, the larger water inputs by

irrigation in maize are compensated by larger predicted evapotranspiration rates, eventually

leading to the same predicted precipitation surplus.


Figure 7.6. Average water fluxes (mm a-1) for winter wheat (left) and maize (right). Input is the sum ofprecipitation and irrigation, ETact is the sum of interception, soil evaporation and transpiration, and output is theground-water recharge. Fluxes were calculated for a 20 years period (1971-2000).

The mean average precipitation surplus amounts to approximately 350 mm yr-1 (table 7.3).

The table shows that the differences between the regions are larger than the differences

between the crops. It is obvious that climate zones with high precipitation fluxes generally

also have a high precipitation surplus, which is a confirmation of the conclusions drawn from

the maps.

Predicted pesticide leaching fluxes and concentration in leaching water

Table 7.4 shows the predicted 20 years average mass fraction of pesticides leached below 2 m

depth for the major climate zones. Mass fractions are expressed as a percentage of the applied

dosage. It is obvious that there are large differences between the three pesticides considered.

The average leaching fraction generally decreases in the order E > B > A > D. The large

sensitivity to the physico-chemical properties of the pesticide is entirely in line with

investigations by Boesten and van der Linden (1991), who found that changing Kom or DT50

by roughly a factor of two changes the amount leached by roughly a factor of 10. The figure

further shows that the predicted leaching fraction is larger for autumn applied pesticides

(winter wheat) than for spring applied pesticides (maize). This confirms earlier findings

(Boesten and van der Linden, 1991; Tiktak et al., 1996b) that the leaching fraction is

extremely sensitive to the amount of rainfall in the period directly after product application.

Notice that the differences between winter wheat and maize are usually larger for the warm

(Mediterranean) climate zones, where there is a distinct wet season. The predicted average

leaching fraction for the major climate zones shows two distinct series, i.e. one for the

temperate climates (T4 > T3 §�7��!�7��DQG�RQH�IRU�WKH�ZDUP�FOLPDWHV��:��!�:��§�:��This was expected on the basis of differences in precipitation surplus. Notice that the

differences between the warm and temperature climates are large in the case of maize and

small in the case of winter wheat. Apparently, differences in seasonal rainfall patterns have a

much stronger effect on the predicted leaching fraction than temperature differences (cf.

Tiktak et al., 1994).


Table 7.4. 20-year average predicted mass flux of pesticide leached below 2 m depth. Mass fluxes are presented forthe climate zones as described in table 7.1. The letter ‘T’ refers to temperate, the letter W to ‘Warm’ .

a) Winter wheat

Percentage of dosage applied leached below 2 m depth

T1 T2 T3 T4 W1 W2 W3

Substance A 0.22 1.14 1.12 3.25 0.73 1.02 2.96

Substance B 0.60 3.58 2.83 9.07 2.88 1.95 8.02

Substance D 0.02 0.24 0.17 1.04 0.33 0.19 1.15

Substance E 0.76 2.96 2.39 5.40 2.40 2.91 6.67

b) Maize

Percentage of dosage applied leached below 2 m depth

T1 T2 T3 T4 W1 W2 W3

Substance A 0.14 0.72 0.71 1.96 0.30 0.65 1.22

Substance B 0.08 0.55 0.75 1.62 0.25 0.20 0.39

Substance D 0.01 0.05 0.05 0.20 0.01 0.02 0.04

Substance E 0.38 1.42 1.36 2.75 0.73 1.28 1.99

Figure 7.7 shows maps of the leaching risk of pesticide in Europe as calculated with the

EuroPEARL approach. According to harmonised European procedures (FOCUS, 2000), the

leaching risk of a pesticide is approximated by the 80th percentile of the leaching

concentration due to weather conditions. This percentile was calculated in a two step

approach. First, for each year, the mass flux of pesticide leached was divided by the annual

precipitation surplus, which is approximated by the predicted water flux at 1 m depth. Then,

from a series of 20 years, the 80th percentile was chosen. The maps show that high and low

risk areas occur at relatively short distances. This suggests that soil properties, which show a

strong variability at short distances (figure 7.3), have a large effect on the leaching risk.

Additional analyses by


Figure 7.7. Leaching risk of pesticides, approximated by the 80th percentile of the predicted leachingconcentration. Concentrations were calculated according to FOCUS (2000). See further text.

Piñeros-Garcet et al. (2003) indeed showed a large sensitivity of the leaching concentration to

organic matter and (to a lesser extent) soil texture. The maps further show that high and low

risk areas can occur everywhere in Europe. This could be expected on the basis of the maps of

precipitation surplus (figure 7.6), which show that a high precipitation surplus can occur in


both Western and Southern Europe (particularly Italy). For the driest climates, however, this

finding was not expected, as the predicted mean precipitation surplus in these climates is low

(Spain, Southern Italy and Greece). In these cases, the explanation should be sought in the

large temporal variability of rainfall. This effect is shown in figure 7.8, which shows the

predicted leaching concentration in a dry year (represented by the 20th percentile) and the

predicted leaching concentration in a wet year (represented by the 80th percentile). The figure

clearly shows that the North-South trend of the 20th percentile is much stronger than the

gradient of the 80th percentile. Apparently, the leaching problem in dry climates is confined to

worst-case conditions (i.e. wet years).

Figure 7.8. Maps showing the 20th and 80th percentiles of the leaching concentration of substance B.The maps show the effect of weather conditions on the leaching concentration.

7.4 Scenario representativeness estimation using EuroPEARL

7.4.1 Introduction

Despite the fact that EuroPEARL results pertain to 74% of the total area only, a first estimate

of the 80th percentile of the leaching concentration in space can be obtained from the

frequency distributions of the simulated leaching concentration within each FOCUS area.


These percentiles can be compared with results from the FOCUS standard scenarios.

Scenarios are assumed to be representative of realistic worst case conditions if the simulated

leaching concentration is higher than or equal to the 80th percentile of the leaching

concentration in space. Results of this exercise are presented in figures 7.9 and 7.10 for winter

wheat and maize, respectively.

Both figures show that the 80th percentile of the leaching concentration simulated by

EuroPEARL is generally higher than or equal to the FOCUS results for five out of eight

scenarios. The Sevilla, Porto and Thiva scenarios underestimate the 80th percentile of the

leaching concentration, which would imply that these scenarios are not strict enough. This

finding should, however, be treated with extreme caution. First, there are still questions about

the quality of the Pan-European soil databases (see discussion and section 7.5). Second, the

EuroPEARL results pertain to only 74% of the total area.

Figure 7.9. 80th percentile of the predicted leaching concentration for each FOCUS area (‘PEARL’ ) comparedwith results of eight FOCUS standard scenarios (‘FOCUS’ ). Simulations were carried out for winter wheat. CH= Chateaudun, HA = Hamburg, KR = Kremsmünster, OK = Okehampton, SE = Sevilla, PI = Piacenza, PO =Porto and TH = Thiva.


Figure 7.10. 80th percentile of the predicted leaching concentration for each FOCUS area (‘PEARL’ ) comparedwith results of eight FOCUS standard scenarios (‘FOCUS’ ). Simulations were carried out for maize. CH =Chateaudun, HA = Hamburg, KR = Kremsmünster, OK = Okehampton, SE = Sevilla, PI = Piacenza, PO = Portoand TH = Thiva.

7.4.2 Discussion

The current version of EuroPEARL can be seen as an attempt to fully implement a Pan-

European, mechanistic and spatially distributed leaching model for plant protection products.

Based on common knowledge of the leaching process, the behaviour of the model can be

judged ‘plausible’ . Nevertheless, the model predictions are subject to a high degree of

uncertainty. Errors result first from the way how the system is conceived in the selected

model (the conceptual level); and second from the way how the model inputs and parameters

have been generated (Loague and Corwin, 1996; Vanclooster et al., 2002).

Model errors at the conceptual level arise when processes are inappropriately described by the

model, or when process descriptions are forced to be used in applications for which they were

not initially intended. A conceptual limitation of EuroPEARL is for instance related to the

spatial schematisation of the environmental system, i.e. the plant-soil-subsoil system. The

properties of the environmental system vary extremely in space and time and this variability is

now encoded by spatially distributing the environmental properties in a discrete way. Thereby

it is considered that the transport of PPPs from the land surface to the ground-water body

passes through a set of 10 x 10 km2 parallel soil columns, each characterised by effective


properties (e.g. effective soil properties). Variability of fate and transport processes at the

surface or within these large soil columns, and of PPP between these columns, is completely

ignored in EUROPEARL. Techniques for assessing the small scale variability are still poorly

developed and cannot be implemented at the Pan-European scale. An extreme example of this

small scale variability is the is the ignoring of preferential flow, a process for which

consensus exists that it is extremely important for correctly describing PPP transport in soils

(Flühler et al, 2001; chapter 2). The most important reasons for ignoring preferential flow is

the lack of parameters to quantify the preferential flow process for regional-scale model

applications. As progress has recently been made in this research area (Scorza Júnior, 2002),

an attempt is made to include preferential flow in the PEARL model (Section 2). However,

basic soil information for this preferential flow model such as quantitative soil structure

information (Rawls et al, 1996) is not yet available at the Pan-European scale, so it remains

questionable whether a regional-scale version of this preferential flow model will become

available shortly. Another conceptual limitation is related to the simplification of the fate and

transport processes of PPPs at the soil surface. Surface hydrological components are not

implemented in detail, and transport of PPP with surface runoff or eroded soil particles has

not been considered. In addition, the land use component has been simplified to a serious

extent that may affect the predicted PPP balance at the soil surface.

Input and parameter generation errors depend on the quality of the underlying databases and

the quality of the parameter generation techniques such as the quality of the used pedotransfer

functions (Tiktak et al., 1999). The most important databases for EuroPEARL are the soil and

climate databases.

For characterising the spatial distribution of soil properties throughout Europe, the European

Soil map at the scale of 1:1.000.000 was used in combination with the Soil Profile Analytical

Database, release I (Jamagne et al., 1995). This version of the database has some serious

limitations. The most serious limitation is that 25% of the total agricultural area of the

European Union could not be assigned a soil profile. Moreover, only for a limited number of

Soil Mapping Units could a direct link be established between the Soil Mapping Unit and the

analytical data in the profile database, so a less certain link based on decision rules had to be

established in many cases. Although each Soil Mapping Unit is an association of a number of

Soil Typological Units, the soil profile database did not contain sufficient information to link

more than one typological unit. This implies that information on the underlying spatial

variability was lost. Tiktak et al. (1996b) showed that ignoring spatial variability in a leaching

study may affect the final results. A further concern pertains to the way that the analytical data


has been obtained. In many cases, available data were based on estimated profiles, potentially

introducing a bias between the individual countries. These concerns have led to the

development of a new version of the database (SPADE II), which will become available soon.

Another concern relates to generation of EUROPEARL input data from basic soil data

available in SPADE. For some of the EuroPEARL input parameters (e.g. soil hydraulic

properties) pedotransfer rules have been adopted. The validity of these pedotransfer rules may

however be small, in particular for some of the sensitive hydraulic parameters, such as the

saturated hydraulic conductivity. For other sensitive EuroPEARL model parameters such as

the hydrodynamic dispersion coefficient no pedotransfer rules exist. The low validation level

of some of the pedotransfer rules, or the total absence of pedotransfer rules for some of the

sensitive parameters will add another source of uncertainty to the predicted PECs.

To obtain the spatial distribution of daily weather data, a simple scaling procedure has been

adopted. A central assumption in this approach is that data from one weather stations could be

used to correctly describe the seasonal dynamics of weather conditions within each climate

zone. Figure 7.8 shows that the annual variability due to weather conditions is larger in warm

climate zones than in temperate climate zones. Moreover, maps of winter and summer

precipitation (not shown) show a distinct dry season in the warm climate zones. These

conclusions suggest that the adopted procedure is generally applicable. However, the maps

also show some anomalies. It is therefore important that this procedure needs to be validated

against more detailed weather data as available in the MARS database (Vossen and Meyer-

Roux, 1995).

Given all these uncertainties, we believe that the maps generated by means of the

EuroPEARL model should be treated with a lot of care. The PEC maps of PPPs generated by

means of the EuroPEARL should rather be considered as the potential concentration of a PPP

in ground-water systems conditioned to the emitted hypothesis invoked in the modelling

exercise. These potential concentrations should be considered as proxy variables of the actual

concentrations which might be found in ground-water systems and should be compared to

and/or combined with the results generated through more detailed higher tier modelling and

through detailed monitoring of the ground-water system. Notwithstanding this intrinsic high

uncertainty associated with the PECs generated by means of large scale spatially distributed

leaching models, we believe that the presented methodology makes a major step forward in

modelling potential ground-water contamination by the use of PPPs, in particular in view of

Pan-European harmonised registration and risk assessment procedures. In contrast to the

current procedure (FOCUS, 2000), the methodology presented in this paper considers the


variability of the environmental system in an explicit and statistical verifiable way.

Considering variability in such a verifiable way will increase the quality of the exposure

assessment, and should result in a more balanced and scientifically based process of

registration.

7.4.3 Conclusions

The PPP leaching model PEARL in combination with European soil and climate databases

could be used to describe the leaching risk of PPPs at the scale of the European Union. Using

the approach described in this study, 75% of the total agricultural area could be

parameterised.

Simulations were carried out for four pesticides with different properties. Results showed that

the leaching risk generally increased with precipitation and irrigation and decreased with

increasing organic matter content. Because of the strong sensitivity of the leaching

concentration to soil properties, there was a strong variability of the calculated leaching

concentration at relatively short distances. Results further indicated that due to large irrigation

amounts combined with large temporal variation of rainfall in the Southern European

countries, the trend in the calculated leaching risks from North to South was less extreme than

expected. This implies that areas of high leaching risk (‘hotspots’ ) as assessed by means of

the EuroPEARL model occur in all countries of the European Union, including the Southern

European countries.

7.5 Scenario representativeness estimation using a metamodel of

EuroPEARL

To assess the representativeness of the scenarios, it is essential to define the ‘real’ 90th

percentile of the leaching concentration. In an ideal situation, this 90th percentile is obtained

from a detailed monitoring of the presence of active substance in ground-waters. However,

our observational skills are limited in time and space, and therefore the ‘real’ 90th percentile is

unknown. If the 90th percentile of leaching cannot be obtained from direct measurements, an

alternative is offered by simulating these leaching values by means of a spatially distributed

leaching model. This type of validation has a lower power than a validation in which the 90th

percentile of leaching is estimated from direct measurements, but it is the only pragmatic way

to proceed with scenario validation at this time. However, as shown in section 7.4, it is

currently not possible to perform simulations with a deterministic model (EuroPEARL) for

the entire European agricultural area. To overcome this problem, the EuroPEARL results


were interpolated and extrapolated to the rest of the area, using a metamodel of EuroPEARL.

The combination of EuroPEARL with the metamodel allows: (i) the use (for leaching

simulations) of the 1:1,000,000 European soil map, which covers 97% of the European

agricultural area instead of the 75% covered by the profile database; (ii) the consideration of

spatial variability of leaching inside the soil mapping units; and (iii) a statistical test of the

validity of the scenarios.

7.5.1 Methodology

The conceptually based, spatially distributed numerical model as described in section 7.3 (i.e.

the EuroPEARL model) suffers from some limitations. First the conceptually based model

approach is a computer-intensive exercise. Second, the modelling procedure can not presently

be implemented for approximately 25 to 48% of the European agricultural area, since the Soil

Profile Analytical Database (SPADE) does not (yet) cover all European soil types. Finally, the

model ignores the variability of soil properties within the Soil Mapping Units (SMUs) by

using only the dominant soil typological unit (STU) within each SMU. To avoid these

limitations, a multi-scale hierarchical spatial model approach has been implemented. In this

approach, three spatial levels were considered (figure 7.11): (i) the point scale for which a

point scale leaching concentration (PS_PEC) is calculated; (ii) the grid scale corresponding to

a regular grid of 10x10 km2, for which a grid scale leaching concentration (GS_PEC) has

been calculated and (iii) the regional scale (the FOCUS area or Europe), for which the

regional scale leaching concentration (RS_PEC) is calculated.

A multi-scale hierarchical model approach to infer the leaching concentration

At each spatial level, the leaching concentration is fully characterised by a probability density

function (pdf), to infer the required percentile. The local scale pdf of the leaching

concentration reflects the uncertainty related to (i) the use of the point scale exposure model,

(ii) the use of effective point scale model parameters thereby ignoring pore scale variability;

and (iii) the temporal variability of the annual leaching concentration. The grid scale pdf

further considers the variability of the soil parameters within a 10x10 km2 grid. The regional

scale pdf finally encodes the variability of the grid scale parameters within the region. The

upscaling from the point scale to the grid scale is done by considering that the grid scale is a

mixed distribution (Bishop, 1995) from each individual point scale pdf. In the same way, the

regional scale pdf is a mixed distribution of the grid scale pdfs. Finally, from the regional-

scale pdf, a 90th percentile of leaching can be calculated. If this calculation is repeated through


a Monte Carlo procedure, the metamodel error (with respect to the numerical model) can be

taken into account, and a pdf can be calculated for the p90REF .

dens

ity

dens

ity

point scale leaching

regional scale leaching

grid scale leaching dens

ity

leaching

leaching

p90REFde

nsity

leaching

p90 pdf ofregional scale leaching

montecarlo

Figure 7.11. The multi-scale spatially distributed exposure modelling approach. Three scales are considered: thepoint scale, the grid scale and regional scale. For each scale, a leaching pdf is calculated. From the regional scalepdf, a Monte-Carlo procedure determines the p0.9

REF pdf, taking into account the uncertainty of the metamodel.

Once the p0.9REF pdf is determined, it is used to estimate the representativeness of the FOCUS

scenarios, and to perform the representativeness test (Figure 7.12[Juan1])

Reference leaching:

FOCUS scenarios

Tested leaching:

Scenarios representativeness test: P(p0.9

MM<p0.9FOCUS)

p0.9FOCUS

pdf(p0.9MM)

Metamodel

pdf(p0.9MM)

leaching

Hierarchical multiscalespatial modelling PEARL

Figure 7.12 The scenarios representativeness test. Thereference leaching pdf is estimated using a metamodeland a hierarchical multiscale approach. The testleaching is estimated from the FOCUS scenarios usingthe PEARL numerical exposure model. Reference andtested leaching are compared through a statisticalprocedure which estimates .


A metamodel of PEARL

The multi-scale hierarchical spatial modelling approach is based on the possibility of

estimating the point scale leaching concentrations pdf from a reduced number of parameters

present in the soil and climate maps. For such a purpose, a metamodel was derived from the

EuroPEARL model. The metamodel relates the annual leaching concentration pdf at 2 m

depth as derived from EuroPEARL simulations to a set of basic soil and climatic input

parameters. The fitting of a metamodel to the EuroPEARL results was realised using radial

basis artificial neural networks. Artificial neural networks outperform other regression

methods in terms of flexibility and versatility. Radial basis neural networks are described in

detail in Bishop (1995). Input parameters for the metamodel were long term average mean

seasonal and annual climatic data (rain and temperature), the top- and subsoil available water,

organic matter, sand and clay content, and packing density. Outputs of the metamodel are

mean and standard deviation of the log-transformed annual leaching concentration at 1 m

depth.

Data handling

Figure 7.13 gives a brief overview of the steps followed from the original data up to the

leaching maps and representativeness test.

X2

PEARL Y=PEARL(X2)

SPADE profile database

X1=

X3

MM (metamodel)calibration with

X3,Y

Scaling

MMPS_PEC=MM(X1)

Leaching maps

Reference leachingat the regional scale(RS_PEC)

leaching

• Europeansoil map

• Focus areasmap

• Climate maps

+

+

00- 0.0010.001- 0.010.01-0.10.1- 0.50.5- 11- 55- 1010-100

O ut o f FO CUS area

80t h perce ntile leaching (µg/l)

pdf(p0.9MM)

Hierarchicalmultiscalespatialmodelling

Figure 7.13: Estimation of the reference leaching. From a database X1 constituted by the 1:1.000.000 Europeansoil map, the European climatic maps and the climatic FOCUS areas maps, a second database X2 is created, bylinking the European soil map to the SPADE database. Then, the numerical exposure model PEARL is run overX2 , to obtain a leaching estimation Y. Additionally, X2 (which contains, amongst others, continuous soil profileparameters) is transformed to X3 (which contains soil parameters classes) by a scaling and classificationprocedure. X3 and Y are used to calibrate a metamodel MM. MM is used to estimate PS_PEC, the leaching pdfat the point scale. Finally, a hierarchical multiscale spatial modeling is implemented to estimate both theleaching maps and the pdf of the 90th percentile (pdf(p0.9

MM)) of chosen regions. p0.9MM is considered to be the

reference leaching against which the tested leaching p0.9FOCUS will be compared, so p0.9

REF = p0.9MM.


7.5.2 Results and discussion

To infer the 90th percentile of the leaching concentration, the following steps were followed:

(i) metamodel calibration, (ii) application of the metamodel at the Pan-European scale and

generation of the leaching maps, and (iii) estimation of the 90th percentile at the Pan-European

scale.

Metamodel calibration

The inputs and outputs of EuroPEARL were used to train the artificial neural network.

Figure 7.14 presents the results of the metamodel validation for the mean of the annual point

scale leaching. The efficiency of the metamodel (R2=0.89) is sufficient, so it has been

concluded that it can be used in the multi-scale spatial procedure proposed before.

Pearl numerical model (leaching µg/l)

AN

N m

etam

odel

(lea

chin

g µg

/l)

R2=0.89

Figure 7.14. Metamodel validation scatterplot for the prediction of the mean of log-transformed annual leaching.

Mapping the leaching concentration at the Pan-European scale

The metamodel gives both the mean and the standard deviation of the leaching concentration

at the point scale. In consequence, the metamodel can be used to estimate the pdf at the grid

scale. From the grid scale pdf, the 80th percentile of the leaching concentration can be

estimated. Results are presented for FOCUS substance D and compared with the 80th

percentile in time as simulated with EuroPEARL (figure 7.15).


Figure 7.15. 80th percentile of the predicted leaching concentration of substance D, calculated with EuroPEARL(left) and the metamodel (right). Simulations have been carried for winter wheat.

From figure 7.15 it can be seen that the spatial coverage of the metamodel is higher than the

spatial coverage of EuroPEARL (25-48% for EuroPEARL; 74% for the metamodel). It can

also be seen that there are considerable differences between the metamodel results and the

EuroPEARL results. From figure 7.16[Juan2], it can be seen that the differences between the

two maps are mainly due to differences in the extreme high or low values of leaching, while

the median of the two maps are in the same order of magnitude. There are three main sources

of error that can explain the differences between the metamodel and EuroPEARL, i.e. (i)

errors due to extrapolation, (ii) errors due to the use of different databases, and (iii) errors due

to the use of dominant STUs by EuroPEARL. These three aspects are differences are

discussed further.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 110

-6

10-5

10-4

10-3

10-2

10-1

100

101

102

103

p 80

from

map

s (µ

g/l)

Cumulated density

Numerical modelMetamodel

Figure 7.16: Cumulated probability densityfunctions (cdf) of the 80th percentileleaching maps, as calculated by theeuropearl numerical model and by themetamodel.

log10(p0.8)Substance D

[-10 , -3]

[-3 , -1][-1 , 0][0 , 1]

[1 , 4]no data


[-10 , -3]

[-3 , -1][-1 , 0][0 , 1]

[1 , 4]no data


[-10 , -3]

[-3 , -1][-1 , 0][0 , 1]

[1 , 4]no data


The first reason for the discrepancies is that the metamodel results are extrapolated beyond

the range of the numerical model. This is particularly true in the Northern European countries,

where EuroPEARL simulations are lacking. As a result, extreme high values were predicted

with the metamodel for Northern Europe. We therefore concluded that the metamodel cannot

be used for validation of the Jokioinen scenario.

Differences also result from the use of different database in the generation of the leaching

maps. EuroPEARL uses the SPADE database to estimate soil properties, while the metamodel

directly uses the European Soil Map 1,000,000. One would expect a good correlation between

soil properties at the European Soil Map and SPADE, but unfortunately this is not true. This

finding is illustrated in figure 7.17, which shows the extremely poor relationship between

organic matter derived from the soil map and organic matter derived from SPADE. We

concluded that improvement of the Pan-European databases should be a main research task

for the upcoming years.

OM (g/g) in SPADE database

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

M -

L -

VL-

OM

cla

sses

in E

urop

ean

soil

map

VL L M H

Figure 7.17. Relation between the topsoil organic matter content in the SPADE database and in the European soilmap. Organic matter at the European Soil Map is given in classes only (VL = very low, L = Low, M = Mediumand H = high).

Differences may finally be caused by the fact that EuroPEARL uses dominant STUs only (see

section 7.3), while the metamodel uses all STUs. The incorporation of variability at lower

spatial scales usually increases the mean leaching concentration (Van der Zee and Boesten,

1991), but the extend of this effect has not been quantified within the APECOP project.


Estimation of the 90th percentile of the leaching concentration

The metamodel can be used to estimate the regional scale leaching pdf and the 90th percentile

of the leaching concentration. This exercise has been carried out for Europe as a whole.

Results are presented in figure 7.18.

100

101

102

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 20

met

amod

el p

90

dens

ity

p90 (µg/l)

FOC

US

p 90 O

keha

mpt

on, s

ubst

ance

D

p 0.9

MM

Figure 7.18. The reference 90th percentile leaching concentration for Europe as a whole (p0.9MM) as estimated by

the metamodel. Simulation were carried out for substance D and winter wheat.Figure 7.18 shows that the reference leaching concentration is approximately 43 µg L-1, which

is 16 times higher than the 80th percentile of the leaching concentration inferred directly from

the EuroPEARL leaching maps (2.5 µg L-1). The estimated reference leaching concentration

is also an order of magnitude higher than the leaching concentration obtained from the

FOCUS scenarios (results not shown here). This would invalidate the current FOCUS ground-

water scenarios. However, as described above, there are considerable errors in the prediction

of the leaching concentration by the metamodel. As long as the effects of the different error

sources on the leaching concentration have not been quantified, results of Pan-European

leaching studies, including the metamodel results, should be treated with extreme care.

7.5.3 Conclusions

This section shows that a methodology to validate the representativeness of the FOCUS

ground-water scenarios is currently operational. In this methodology, a numerical model

(EuroPEARL) is combined with a metamodel. However, the poor quality of the Pan-European

soil databases introduces errors in the procedure. Using the available databases, with all

defects, the representativeness of the FOCUS scenarios has been found to be low, and the

FOCUS scenarios seem to underestimate the 90th percentile of leaching. However, as the

effects of the different error sources as described above have not yet been quantified, this

finding must be treated with extreme care.


7.6 Synthesis

In this section, a conceptual fraimwork for scenarios validation and a methodology to validate

the representativeness of the FOCUS groundwater scenarios was presented and applied. Four

validation criterions were defined: representativeness, realism, consistency and relevancy.

Validating the representativeness of the FOCUS scenarios came down to answering the

question ‘Are FOCUS scenario combinations of parameter values selected in such a way that

when used in combination with a FOCUS PEC ground-water model, the calculated leaching

values correspond to the real 90th percentile of leaching?’ .

We adopted a two-step approach to answer this question. First, simulations were carried out

by means of a spatially distributed model (i.e. the EuroPEARL model) for 1062 unique

combinations of soil type, climate and country, sampled throughout Europe. Soil properties,

including soil horizon designations, were obtained from the Soil Profile Analytical Database

of Europe. Daily weather data were obtained from 9 points of the MARS database and scaled

from those points to cover the rest of Europe. Other data like irrigation data, crop data and

pesticide properties have been compiled from various sources, such as inventories, field-

studies and the literature. The 1062 unique combinations together represented at maximum

75% of the total agricultural area of the European Union (52% if the area of the STU inside

the SMU is taken into account). To consider also the soil-climate and crop combinations for

which no profile data where available, in a second step a metamodel of EuroPEARL was

applied. The combination of the process based deterministic EuroPEARL model with the

metamodel allowed: (i) the use (for leaching simulations) the European 1:1,000,000 soil map,

which covers 74% of the European agricultural area instead of the 52-75% covered by the

profile database; (ii) the consideration of spatial variability of leaching inside the mapping

units; and (iii) a statistical test of the validity of the scenarios.Results of the EuroPEARL

model were presented in maps with a resolution of 10x10 km2. Results indicated that the

leaching concentration generally increased with precipitation and irrigation and decreased

with increasing organic matter content. Because of the strong sensitivity of the leaching

concentration to soil properties, there was a strong variability of the calculated leaching

concentration over relatively short distances. Results further indicated that due to large

irrigation amounts combined with large temporal variation of rainfall in the Southern

European countries, the trend in the calculated leaching risks from North to South was less

extreme than expected. This implies that areas of high leaching risk (‘hotspots’ ) as assessed

by means of the EuroPEARL model occurred in all countries of the European Union,


including the Southern European countries. Although the EuroPEARL calculations apply to

only 52-75% of the agricultural area, we made a direct comparison between the EuroPEARL

results and the FOCUS calculations. This comparison showed that the 80th percentile of the

leaching concentration simulated by EuroPEARL was generally lower than or equal to the

FOCUS results for five out of eight scenarios. The Sevilla, Porto and Thiva scenarios

underestimated the 80th percentile of the leaching concentration, which would imply that these

scenarios are not strict enough.

Results from the metamodel pointed in another direction: the leaching concentrations

predicted by the metamodel were an order of magnitude higher than the leaching

concentrations predicted by EuroPEARL and by the FOCUS procedure. As the quality of the

metamodel itself was good (R2 = 0.89), the differences occurred during the application of the

metamodel and were caused by the extreme poor quality of the Pan-European soil databases.

We reported three main sources of error that could explain the differences between the

metamodel and EuroPEARL, i.e. (i) errors due to extrapolation, (ii) errors due to the use of

different databases, and (iii) errors due to ignoring spatial variability within soil mapping

units by EuroPEARL.

In conclusion, the validation status of the current FOCUS ground-water scenarios cannot yet

be determined with sufficient accuracy. To answer this question, the quality of the Pan-

European soil databases needs to be improved. It should be emphasised, however, that the

proposed methodology is based on modelling. To come to a more powerful validation

statement, results should be assimilated with detailed monitoring of active substances in

European ground-water systems.


References

Anonymous. 1991. The Authorisation Directive. Official Journal of the European Communities No. L 230,19.8.1991. Brussels, Belgium.

Allen, R.G., L.S. Pereira, D. Raes and M. Smith, 1998. Crop evapotranspiration: guidelines for computing cropwater requirements. FAO Irrigation and Drainage paper (56), 300 pp. FAO, Rome, Italy. Available athttp://www.kimberly.uidaho.edu/water/fao56.

Bishop, C.M. 1995. Neural networks for pattern recognition. Oxford University Press.Black, T.A., W.R. Gardner, and G.W. Thurtell. 1969. The prediction of evaporation, drainage and soil water

storage for bare soil. Soil Sci. Soc. Am. J. (33):655-660.Bock, H.-H., E. Diday (Eds.). 2000. Analysis of symbolic data. Springer verlag, München.Braden, H. 1985. An energy balance and Evapotranspiration model for common crops. Mitteilungen Deutsche

Bodenkundliche Geselschaft (42):294-299.Brouwer, W.W.M., J.J.T.I. Boesten, J.B.H.J. Linders and A.M.A. van der Linden. 1994. The behaviour of

pesticides in soil: Dutch guidelines for laboratory studies and their evaluation. Pesticide Outlook (5):23-28.Capri, E., L. Padovani and M. Trevisan. La previsione della contaminazione delle acque sotterranee da prodotti

fitosanitari. Quaderni di tecniche di protezione ambientale (69). Pitagora Editors, Bologna, Italy (In Italian).FAO-Unesco. 1995. Soil Map of the World. Revised Legend. FAO, Rome.FOCUS, 2000. FOCUS ground-water scenarios in the EU review of active substances. Report of the FOCUS

ground-water scenarios workgroup. EC Document. Sanco/321/2000 rev.2., 202pp. Available athttp://arno.ei.jrc.it/focus.htm.

Gries D., F.B. Schneider,1993. A Logical Approach to Discrete Math, Springer-Verlag.Jamagne, M., C. Le Bas, M. Berland, and W. Eckelman. 1995. Extension of the EU database for the soils of

central and eastern Europe. In: D. King, R.J.A. Jones, and A.J. Thomassen (eds.). EUR 16232 EN. Office forOfficial Publications of the European Communities, Luxembourg, pp. 101-114.

Leistra, M., A.M.A. van der Linden, J.J.T.I. Boesten, A. Tiktak and F. van den Berg. 2001. PEARL model forpesticide behaviour and emissions in soil-plant systems. Description of processes. Alterra report 013, Alterra,Wageningen, the Netherlands. Available at www.pearl.alterra.nl.

Madsen-Breuning, H. and R.J.A. Jones. 1995. The establishment of a soil profile analytical database for theEuropean Union. In: D. King, R.J.A. Jones, and A.J. Thomassen (eds.). EUR 16232 EN, Office for OfficialPublications of the European Communities, Luxembourg, pp. 55-63.

Mücher, C.A., K. Steinnocher, J.L. Champeaux, S. Griguolo, K. Wester and P.Loudjani. 1998. Land covercharacterisation for environmental monitoring of pan-Europe. In: Proc. 18th EARSEL Symp. on OperationalRemote Sensing for Sustainable development, ITC, Enschede, 11-13th May 1998, pp 107-113.

Rawls, W.J., D.L. Brakensiek, and S.D. Logsdon, 1996. Estimation of macropore properties for non-till soils.Trans. Am. Soc. of Ag. Eng. (39):91-95.

Resseler, H., H. Schäfer, G. Görlitz, M. Hermann, J. Hosang, R. Kloskowski, R. Marx, R. Sarafin, B. Stein, andR. Winkler. 1997. Recommendations for conducting simulation calculations for the registration procedure.Nachrichtenblad Deutsche Pflantzenschutzdienst (49):305-309.

Scorza Júnior, R.P., 2002. Pesticide leaching in macroporous clay soils: field experiment and modelling. Ph.D.Thesis, Wageningen University and Research Centre, Wageningen, the Netherlands, 234pp.

Siebert, S. and P. Döll. 2001. A Digital Global Map of Irrigated Areas – An update for Latin America andEurope. Report A0102, Centre for Environmental Systems Research, University of Kassel, Germany.

Tiktak, A., A. Leijnse, and H.A. Vissenberg, 1999. Uncertainty in a regional-scale assessment of cadmiumaccumulation in the Netherlands. J. Environ. Qual. ((28):461-470.

Tiktak, A., A.M.A. van der Linden, and I. Leine. 1996a. Application of GIS to the Modeling of PesticideLeaching on a Regional Scale in the Netherlands. In: D.L. Corwin and K. Loague (Ed.). Application of GIS tothe Modeling of Non-Point Source Pollutants in the Vadose Zone. SSSA-Special Publication number 48, pp.259-281.

Tiktak, A., A.M.A. van der Linden and R.C.M. Merkelbach. 1996b. Modeling pesticide leaching at a regionalscale in the Netherlands. RIVM report no. 715801008, Bilthoven, the Netherlands.

Tiktak, A., F.A. Swartjes, R. Sanders and P.H.M. Janssen, 1994. Sensitivity analysis of a model for pesticideleaching and accumulation. In: J. Grasman and G. van Straten (eds.). Predictability and non-linear modellingin natural sciences and economics. Kluwer Academic, Dordrecht, the Netherlands, pp. 471-484.


Tiktak, A., F. van den Berg, J.J.T.I. Boesten, M. Leistra, A.M.A. van der Linden and D. van Kraalingen. 2001.Pesticide Emission Assessment at Regional and Local Scales: User Manual of FOCUSPEARL version 1.1.1.RIVM Report 711401008, Alterra Report 28, RIVM, Bilthoven, 142 pp. Available at www.pearl.alterra.nl .

Tiktak, A., D.S. De Nie, A.M.A. van der Linden and R. Kruijne. 2002a. Modelling the leaching and drainage ofpesticides in the Netherlands: The GeoPEARL model. Agronomie (22):373-387.

Tiktak, A., J.J.T.I. Boesten and A.M.A. van der Linden. 2002b. Nationwide assessments of non-point sourcepollution with field-scale developed models: The pesticide case. In G.J. Hunter and K. Lowell (eds.).Proceedings of the 5th international symposium on spatial accuracy assessment in natural resources andenvironmental sciences (Accuracy 2002), Melbourne, 10-12 July 2002, pp. 17-31.

Vanclooster M., J. Boesten, M. Trevisan, C. Brown, E. Capri, O.M. Eklo, B. Gottesbüren, V. Gouy A.M.A.van der Linden. 2000. A European test of pesticide-leaching models: methodology and majorrecommendations. Agricultural Water Management (44): 1-21.

Vanclooster, M., M. Javaux, F. Hupet, S. Lambot, A. Rochdi, J.D. Piñeros-Garcet and C.L. Bielders. 2002.Effective approaches for modelling chemical transport in soils supporting soil management at the largerscale. In: M. Pagliai and R. Jones (eds.). Sustainable Land Management – Environmental Protection.Advances in GeoEcology 35, Catena Verlag, Reiskirchen, Germany, pp. 171-184.

Vanclooster, M., J.D. Piñeros-Garcet, J.J.T.I. Boesten, N. Jarvis, H. Vereecken, E. Fernandez, A. Tiktak, M.Kleine and G. Bidoglio, 2001. Effective approaches for assessing the predicted environmental concentrationsof pesticides. EGS General Assembly 2001, Geophysical Research Abstracts 3.

Van der Zee, S.E.A.T.M., and J.J.T.I. Boesten. 1991. Effects of soil heterogeneity on pesticide leaching toground-water. Water Resour. Res., 27: 3051-3063.

Van Dam, J.C. 2000. Field-scale water flow and solute transport. SWAP model concepts, parameter estimationand case studies. Ph.D. thesis, Wageningen University, Wageningen, the Netherlands, 167 pp

Van der Linden, A.M.A., A. Tiktak and M. Leistra. 2001. Incorporation of soil-pH dependent behaviour inpesticide leaching assessment. In: A. Walker. Pesticide Behaviour in Soils and Water. BCPC SymposiumProceedings No. 78, MCPC, Farnham, UK, p. 45-50.

Van Genuchten, M.Th. 1980. A closed form for predicting the hydraulic conductivity of unsaturated soils. SoilSci. Soc. Am. J. (44):892-898.

Vossen, P. and J. Meyer-Roux. 1995. Crop monitoring and yield forecasting activities of the MARS project. In:D. King, R.J.A. Jones and A.J. Thomassen (eds.). European Land Information Systems for agro-environmental Monitoring. EUR EN 16232, Office for the official publications of the EuropeanCommunities, Luxembourg, pp. 11-30.

Worrall, F.. 2002. Direct assessment of ground-water vulnerability from borehole observations, In: K.M.Hiscock, M.O. Rivett, and R.M. Davison (Eds.). Sustainable Ground-water Development. GeologicalSociety, London, Special Publications (193): 245-254.

Wösten, J.H.M., A. Nemes, A. Lilly, and C. Le Bas, 1999. Development and use of a database of hydraulicproperties of European soils. Geoderma (90): 169-185.


8. Compliance of new approaches with current and future

EU regulations

8.1 Objectives

The current FOCUS models and scenarios have been identified within the context of the

implementation of EU Directive 91/414 and are based on the state-of-the-art of pesticide

exposure modelling. The various APECOP work packages have attempted to develop

improved procedures for estimating PEC of pesticides to ground-water, based on thorough

validation of models and scenarios. The objective of this work package was to analyse the

feasibility of introducing the novel and effective approaches developed by the APECOP

project into current EU and future regulation.

8.2 Methodology

This study address two key issues: Firstly, an assessment of the impact of the improved

modelling procedures to calculate PEC into the environment in the context of current

legislation; and secondly, an assessment of how APECOP models could be implemented into

current exposure assessment software.

8.3 Current EU legislation on Pesticides

Pesticides are chemical preparations containing one or more active substances that are

intended to control harmful organisms (such as pest control). Pesticides can pose risks to

humans, animals and the environment in a variety of ways due to their intrinsic properties. In

contrast with other chemical substances (existing and new substances), pesticides are not

allowed to be placed on the market without being authorized.

For access to the market to be given, a series of experimental procedures must be carried out

with a view to provide regulators with adequate information on the impact of the product on

the

• health of human beings, creatures and plants;

• means for safeguarding the environment and


• being a secure safe, efficient and humane methods of controlling pests.

Assessing the predicted environmental concentration of pesticides and related products in the

environment is essential for further evaluation of any risk and decision making for the

inclusion of active substances in Annex I of the Directive on Plant Protection Products (PPP -

Council Directive 91/414/EEC). Annex I is a positive list of active substances that have been

shown to be without unacceptable risk to people or the environment and consequentially

allows the authorization of the product.

Member States will only be able to authorise the marketing and use of plant protection

products whose active substances are listed in Annex I, except where transitional

arrangements apply.

An inclusion of an active substance in Annex I and decisions taken with respect to the

granting of authorization take account of the agricultural, plant health and environmental

(including climatic) conditions in question. Member States must have regard to all normal

conditions under which the plant protection product may be used. Environmental conditions

are one of the most important parameters that are considered when taking decisions about

granting the authorisation for use of plant protection product in certain areas of EU.

The aim of Directive 414 is to ensure that the Member States will use harmonized and

uniform principles for taking equivalent decisions on authorization for PPP before placing

them on the market within the EU. The authorized products (active substances, metabolites,

breakdown products or residues) should have no harmful effect on human or animal health,

directly or indirectly (e.g. through drinking water, food or feed) or on ground-water and they

should have no unacceptable influence on the environment, having particular regard to the

fate and distribution in the environment, particularly contamination of water including

drinking water and ground-water.

The Members States should ensure that compliance with requirements is established by

official or officially recognised tests and analyses carried out under agricultural, plant health

and environmental conditions relevant to use the PPP in question and representative of these

prevailing where the product is intended to be used, within the territory of the Member State

concerned.

Mutual recognition of authorizations between Member States is needed to prevent repetition

of tests and analyses already carried out. Agricultural, plant health and environmental

(including climatic) conditions relevant to the use of the product should be comparable in the

regions concerned.


The requirements for the dossiers to be submitted are included in subsequent annexes to the

Directive:

• Annex II includes the list of requirements for the dossier for the inclusion of an active

substance in Annex I. Among the requirements it considers fate and behaviour of the

substances in soil the rate and route of degradation, adsorption and desorption,

mobility of active substance in at least three soil types and, where relevant,

metabolites and breakdown products should also be addressed. For the water and air

compartment, biodegradationand hydrolysis, as well as degradation, adsorption and

desorption should be described. An additional important property for the air

compartment is the volatility of the substance.

• Annex III includes the list of requirements for the dossier related to the inclusion of a

plant protection product. One of the requirements is testing for distribution and

dissipation in soil, water and air and should be included in dossier where relevant for

better description of fate and behaviour of the substance in the environment.

• Annex VI includes the uniform principles that Member States should adopt for

evaluation and authorization of plant protection products.

In summary then, the models used for the calculation of certain parameters required by the

legislation should:

• make a best possible estimation of all relevant processes involved taking into account

realistic parameters and assumptions,

• be submitted to an analysis to identify critical decision points or items of data for

which uncertainties could lead to a false classification of risk,

• be reliably validated with measurements carried out under circumstances relevant for

the use of the model,

• be relevant to the conditions in the area of use.

8.4 Role of the new modelling approaches with present and future

EU regulation


8.4.1 Potentials

The enhanced and validated local scale models presented in the APECOP project (section 2

to 6) consider all requirements of the PPP legislation mentioned in the previous section. They

focus primarily on fate and distribution of the PPP in the environment.

Two models involved in this project (PELMO and PEARL) are also discussed with respect to

their applicability in the scenarios for treated wood-in-service and storage prior to shipment,

for calculation of the emission from treated wood that might reach ground-water (OECD

Series on emission scenario documents recommended in EU for risk assessment of biocide

active substances).

We can expect that the new approach described in this project will extend its applicability

from PPP to biocides and probably also to substances that are included in the new chemical

policy (REACH).

The greatest advantage of the EuroPEARL – metamodel approach (section 7) compared to the

local scale models is the inclusion of pan-European environmental databases in an attempt to

provide a more realistic view on impact of PPP. The information on European environmental

conditions (soil properties and climate characteristics) is crucial when taking decisions on

restrictions of PPP use in certain areas of EU. The facility for estimating

PEC values (soil, ground-water, surface-water, air) at the European level in the form of maps

showing the spatial variation of vulnerability will make decisions considering the mutual

recognition of authorisation in different Member States easier, faster and more consistent.

It is important that the model gives the possibility of working at different levels or scales: for

example, decision makers may wish to obtain a general overview of the situation for a

particular substance which can be calculated very quickly but due to the need to scale up

environmental parameters, the result would be less precise that when the model is run at a

more detailed scale. As the EuroPEARL approach uses realistic environmental data, the

model includes a significantly higher number of different soil types. This means that the

vulnerability of many more soils can be assessed.

8.4.2 Recommendations

As the output of the EuroPEARL meta-model approach are now spatial and compatible with

Geographical Information Systems (GIS), it would be very interesting to include in the


database information about ground-water reservoirs and protected regions such as national

parks etc.

8.4.3 Weak points

The information on environment is given in 10x10 km scale. The EUSES program considers

local scenarios at a lower scale down to 100 m. The EuroPEARL approach does not give

possibility for the estimation of exposure at this local level. The meta-model could remediate

to this but suffers from the lack of quality data basis to be operational at these lower scales.

8.4.4 Related Documentation

Council Directive 91/414/EEC concerning the placing of plant protection products on the

market. OJ L 230, 19. 8. 1991.

Council Directive 97/57/EC establishing Annex VI to Directive 91/414/EEC concerning the

placing of plant protection products on the market. OJ L 165, 27.9.1997.

Council Directive 75/440/EEC concerning the quality required of surface-water intended for

the abstraction of drinking water in the Member States. OJ L 194 , 25.07. 1975

Council Directive 80/778/EEC relating to the quality of water intended for human

consumption. OJ L 229, 30.8.80.

8.5 Implementation of APECOP models into EUSES

The concern regarding the potential risks of chemicals and in particular existing chemicals

was already a policy priority in the late 1980s. The Council of the European Communities, in

approving the Fourth Community Action Programme on the Environment (1987-1992), stated

that one of the priority areas was the evaluation of the risks to the environment and human

health posed by chemical substances. This Action Programme underlined the need for a

legislative instrument, which would provide a comprehensive structure for the evaluation of

the risks posed by "existing" chemicals. In particular, the Action Programme stated that such


a legislative instrument "will establish a procedure for treating priority lists of chemicals for

immediate attention, as well as setting out the means for gathering information, requiring

testing and evaluating the risks to people and the environment". Consequently, the European

Commission proposed a series of legal instruments, which were aimed at meeting the

objectives outlined in the Action Programme. One of these instruments was the Existing

Substances Regulation.

In 1993 the Council adopted Council Regulation (EEC) 793/93 or the Existing Substances

Regulation (ESR), thereby introducing a comprehensive framework for the evaluation and

control of "existing" chemical substances. The Regulation was intended to complement the

already existing rules governed by Council Directive 67/548/EEC for "new" chemical

substances. An “Existing" chemical substance is in the EU defined as any chemical substance

listed in the European INventory of Existing Commercial Substances (EINECS), an inventory

containing 100,195 substances. The Regulation 793/93 foresees that the evaluation and

control of the risks posed by existing chemicals will be carried out in four steps:

• STEP I Data collection;

• STEP II Priority setting;

• STEP III Risk assessment;

• STEP IV Risk reduction.

EINECS was drawn up by the European Commission in application of Article 13 of Directive

67/548, as amended by Directive 79/831, and in accordance with the detailed provisions of

Commission Decision 81/437. It lists and defines those chemical substances, which were

deemed to be on the European Community market between 1 January 1971 and 18 September

1981. In terms of Article 1(4) of the amended Directive 67/548, these are substances to which

the pre-marketing notification provisions of the Directive do not apply. Any chemical

substance that has been marketed after 18 September 1981 is called a new chemical.

Substances on priority lists must undergo an in-depth risk assessment covering the risks posed

by the priority chemical to man (covering workers, consumers and man exposed via the

environment) and the environment (covering the terrestrial, aquatic and atmospheric eco-

systems and accumulation through the food chain). This risk assessment follows the


framework set out in Commission Regulation (EC) 1488/94 and implemented in the detailed

Technical Guidance Documents (TGD) on Risk Assessment for New and Existing

Substances. After adoption of the risk assessment, three publications are produced:

• the comprehensive risk assessment report (as a book, on the ECB Web Site and in the

International Uniform ChemicaL Database (IUCLID)),

• a summary thereof (as an EUR report and on the ECB Web Site)

• a listing of the conclusions in the Official Journal of the European Communities.

Within this context, the EC commissioned the development of a decision-support instrument

software system that allows the implementation of the Technical Guidance Document (TGD)

on risk assessment for New and Existing Substances. This programme is known as EUSES

(European Union System for the Evaluation of Substances). EUSES is designed to enable

government authorities, research institutes and chemical companies to carry out rapid and

efficient assessments of the general risks posed by substances to man and the environment.

EUSES is intended mainly for initial and refined risk assessments rather than comprehensive

assessments. After consideration of certain limits to environmental and human-health effects,

EUSES ranks a substance by degree of concern, and this outcome constitutes the first draft of

a Commission proposal for a priority list.

Within the original EUSES model, PECs are calculated considering a “hypothetical” region

with standard environmental characteristics. This situation does not allow an evaluation of

the substances in other market environments where differing soil or climatic conditions could

lead to soil and ground-water degradation.

The JRC’ s Soil and Waste Unit have been developing a spatial adaptation to the EUSES

model (referred to internally as GeoEUSES) that is similar to the approach undertaken in the

APECOP project for the EuroPEARL model (i.e. adapting a non-spatial model to deliver map

based output). Within this project, EUSES was linked to a number of Geo-referenced datasets

(e.g. Corine Land Cover, European Soil Database, EUROSTAT Population Density, etc).

These datasets were used in order to calculate the regional variability of the Predicted

Environmental Concentration (PEC) due to differing patterns of actual environmental

characteristics (i.e. size of the region, organic carbon content of soil, fraction of agricultural,

urban and natural soil, precipitation, etc).


As in the investigation of the EuroPEARL model, the assessment of the Geo-EUSES indicates

that among the regional parameters of the model, the organic carbon content of soil plays an

important role in defining potentially vulnerable areas. Reliable data for this parameter on a

pan-European basis is lacking.

A new version of the EUSES program is in development, which covers, in addition to new

and existing chemicals, also the assessment of pesticides and biocides. The pesticide model

that is the engine of EUSES 2 is PEARL. This implies that it should be possible to

successfully integrate EuroPEARL and EUSES 2 but a significant level of software re-

engineering will be required.


Annex 1 : List of presentations and publications realised

within the framework of APECOP

Communication of project results at international meetings and

workshops, communication with stake-holders and end-users

Venue Date Title Objectives

Orleans March 2000 PEGASE – Kick off

meeting

Establish links with the PEGASE project

Louvain-

la-Neuve

April 2000 Down scaling

scenarios

Establish links with ECPA

Julich June 2000 Inverse modelling Discuss advanced procedures for

parameter estimation

Julich October

2000

PEGASE – First

project meeting

Establish links with the PEGASE project

Nice March 2001 EGS- General

Assembly

Present APECOP projects to the

scientific community

Firenze July 2001 Int. conf. Sustainable

land management and

environmental

protection

Present effective approaches for

modelling solute transport in soil

Bonn Sept 2001 Fresenius workshop:

Pesticides fate and

transport in the soil

and the environment

Presentation of intermediate project

results

Nice April 2002 EGS- General

Assembly

Present scenario validation protocol

Lisbon June 2002 2nd European

modelling workshop

Present intermediate project results to

members of industry and registration

Bilthoven July 2002 FOCUS-APECOP Present intermediate project results to


meeting delegates of ECPA and the FOCUS

steering committee

Canberra August 2002 Accuracy 2002 Presentation of regional approaches to

predict pesticide leaching

Corfu Sept 2002 2nd European

congresson pesticides

and micropollutants in

the environments

Present intermediate results on scenario

validation at the scientific community

Bruxelles April 2003 Focus Steering

Committee meeting

Present advances and preliminary

conclusions of the project on the FOCUS

steering committee

Piacenza June 2003 XII Int. symposium:

Pesticides in air,

plant, soil and water

system

Present project results to the large

scientific community, including members

of industry and registratio officers

Darmstadt July 2003 Fresenius workshop:Behaviour ofPesticides in Soils,Water and air

Presentation results of the volatilisation

studies

Hamburg April 2003 SETAC Europe13th Annual Meeting


studies

Basel August 2002 Intertopic Workshop:Environmental risk

assessment - Integratingthe exposure andeffects Information,10th IUPAC

International Congress onthe chemistry of cropprotection

Presentation scenario validation

protocole


studies


List of publications

Anonymous, 2003. Effective approaches for predicting environmental concentrations of pesticides: the APECOPproject. International conference. ‘Environment for Better Health’ Arhus, Denmark on 8-11 May 2003.

Berg, F. van den, Wolters A., Jarvis N., Klein M., Boesten J.J.T.I., Leistra M., Linneman V. Smelt J.H.,Vereecken H., 2003. Pesticide volatilization from soil: Improvement of concepts for pesticide volatilizationfrom bare soil in PEARL, PELMO, and Macro models . In: Pesticide in Air, Plant, Soil & Water.Proceedings XII Symposium Pesticide Chemistry June 4-6, 2003 Piacenza, Italy, p 973 - 983. A.A.M. DelRe, E. Capri, L. Padovani, and M. Trevisan, (Eds.)

Cuevas M.V., Fernandez J.E., Celderon M.J., Hermosin M.C., Moreno F., Cornejo J., 2002. Adsorption anddesorption of chloridazon and lenacil in a soil with a sugar beet crop. Pest Management Science, 58, 3-20.

Cuevas M.V., J.E. Fernandez, S. Roulier, M.J. Calderon, M.C. Hermosin, F. Stenemo, M. Larsbo and N. Jarvis,2003. Comparing MACRO 4.3 with 5.0 for simulating the fate of chloridazon and lenacil in a clayey soil ofSoutwest Spain. In: A. Del Re, M. Trevisan and E. Capri (eds), XII Pesticide chemistry symposium. Pesticidein air, plant soil and water system. Piacenza, Italy 4-6 June 2003.

Cuevas M.V., M.J. Calderon, J.E. Fernandez, M. C. Hermosin, 2002. Degradacion de cloridazona y lenacil enun suelo arcilloso del suroeste de Espana. XIII Inca congress, Cuba.

Cuevas M.V., M.J. Calderon, J.E. Fernandez, M.C. Hermosin, F. Moreno and J. Cornejo, 2001. Assessingherbicide leaching from field measurements and laboratory experiments. Acta Agrophysica 57 :15-25

Cuevas MV, Hermosin MC, Calderon M.J., Fernandez J.E., Velazquez I., Lozano MT, 2003. Degradation andadsorption of chloridazon and lenacil as affected by temperature. In: Proceedings of the VII Congress of theEuropean Society for Agronomy, Cordoba, Spain, 15-18 July 2002, pp.351-352.

Ferrari F., M. Trevisan and E. Capri, 2003. Predicting and measuring environmental concentrations of pesticidesin air after soil application. J. Env. Quality. In press.

Garratt J.A., Capri E., Trevisan M., Errera G., Wilkins R.M. (2002). Parameterisation evaluation andcomparison of pesticide leaching models to data from a Bologn. In: Proceedings of the XII SymposiumPesticide Chemistry (A.A.M. Del Re, E. Capri, L. Padovani, M. Trevisan, eds.). June 4-6, 2003, Piacenza (Italy),pp. 367-374.

Leistra, M. & A. Wolters (2002). Computations on the volatilisation of the fungicide fenpropimorph from plantsin a wind tunnel (submitted to Water Air Soil Pollut.).

Leistra, M. (2002). Volatilisation of herbicides from plant surfaces in micro-ecosystems. Proceedings 12th

Symposium European Weed Research Society, Arnhem, p. 112-113.Piñeros Garcet J.D, Vanclooster M.,, A. Tiktak, D. De Nie and A. Jones, 2003. Methodological approach for

evaluating higher tier PEC ground-water scenarios supporting the prediction of environmental concentrationsof pesticides at the pan European scale. In: A. Del Re, M. Trevisan and E. Capri (eds), XII Pesticidechemistry symposium. Pesticide in air, plant soil and water system. Piacenza, Italy 4-6 June 2003.

Piñeros Garcet J.D., A. Ordonnez, J. Roosen and M. Vanclooster, 2003. Meta – modelling analysis for assessingnitrogen leaching in crop rotations. Ecological modeling. Submitted

Piñeros Garcet J.D., D. Denie, M. Vanclooster, A. Tiktak, M. Klein and A. Jones,2002. Validation of thescenarios designed for the EU registration of pesticides. EGS-General Assembly, Nice, April 2002 (EGS02-A-06731)

Piñeros Garcet JD, A. Tiktak, D. de Nie, M. Vanclooster, M. Klein and A. Jones, 2002. Pesticides leachingmetamodels and validation of the scenarios designed for EU registration. 10th IUPAC International Congresson the chemistry of crop protection. p. 151.

Piñeros Garcet JD, M. Vanclooster, 2002. Assessing the quality of the environmental quality assessment tools.In: T. Albanis (ed.). Proceedings of the 2nd European Conference on pesticides and related organicmicropollutants in the environment, Corfu, Greece. p 245.

Scorza Junior, R.P. , Smelt J.H., Boesten J.J.T.I., Hendriks, R.F.A. and van der Zee, S.E.A.T.M., 2003.Preferential flow of bromide, bentazone and imidacloprid in a Dutch clay soil (submitted).

Scorza Junior, R.P., 2002. Pesticide leaching in macroporous clay soils: field experiment and modelling. Ph.D.dissertation Wageningen University, 234 p.

Smelt J.H., R.P. Scorza Junior, R.F.A. Hendriks and J.J.T.I. Boesten, 2003. Preferential transport of imidaclopridin a cracking clay soil. In: A. Del Re, M. Trevisan and E. Capri (eds), XII Pesticide chemistry symposium.Pesticide in air, plant soil and water system. Piacenza, Italy 4-6 June 2003.


Smelt J.H., Scorza Júnior, R.P, Hendriks R,F.A., Boesten J.J.T.I. 2003. Preferential transport of imidacloprid in acracking clay soil. In: Pesticide in Air, Plant, Soil & Water. Proceedings XII Symposium Pesticide ChemistryJune 4-6, 2003 Piacenza, Italy, p 319-326. A.A.M. Del Re, E. Capri, L. Padovani, and M. Trevisan, (Eds.)

Smelt, J.H., R.F.A. Hendriks, L.J.T. van der Pas, A.M. Matser, A. van den Toorn, K. Oostindie, O.M. van Dijk-Hooijer, J.J.T.I. Boesten, R.P. Scorza Júnior. 2003. Transport of water, bromide ion, nutrients and thepesticides bentazone and imidacloprid in a cracking, tile drained clay soil at Andelst, the Netherlands.Alterra, Green World Research, Rapport 289. (in press).

Tiktak A, J. Boesten and T. van der Linden, 2002. Nationwide assessments of non-point oure pollution wothfeild scale developed models : the pesticide case. Int. Conference. Accuracy 2002.

Tiktak A., D. de Nie, T. vander Linden and R. Kruijne, 2002. Modelling the leaching and drainage of pesticidesin the Netherlands : the GeoPEARL model.

Tiktak A., D.S. De Nie, J.D. Piñeros Garcet, A. Jones and M. Vanclooster, 2003. Assessment of the pesticideleaching at the pan European level using a spatially distributed model. In: A. Del Re, M. Trevisan and E.Capri (eds), XII Pesticide chemistry symposium. Pesticide in air, plant soil and water system. Piacenza, Italy4-6 June 2003.

Tiktak A., D.S. De Nie, J.D. Piñeros Garcet, A. Jones and M. Vanclooster, 2003. Assessing the pesticideleaching risk at the pan European level: the EuroPEARL approach. J. of Hydrology. Submitted.

Trevisan M., L. Padovani, C. Vischetti and E. Capri, 2001. Perspectives of the prediction of pesticideenvironmental concentrations and related questions. 2nd symposium of pesticides in food and the environmentin Mediterranean Countries. Valencia, Spain 9-12 May 2001.

Trevisan M., L. Padovani, N. Jarvis, S. Roulier, F. Bouraoui, M. Klein and J.J.T.I. Boesten, 2003. Validationstatus of the present PEC groundwater models. In: A. Del Re, M. Trevisan and E. Capri (eds), XII Pesticidechemistry symposium. Pesticide in air, plant soil and water system. Piacenza, Italy 4-6 June 2003.

Vanclooster M, J.D. Piñeros Garcet, J. Boesten, F. Van den Berg, M. Leistra, J. Smelt, N. Jarvis, P. Burauel,H. Vereecken, A. Wolters,V. Linnemann, J.E. Fernandez, M. Trevisan, E. Capri, M. Klein, A. Tiktak, A. Vander Linden, D. De Nie, G. Bidoglio, F. Baouroui, A. Jones, A. Armstrong and L. Bontoux, 2001. Validationof models and scenarios supporting the estimation of predicted environmental concentrations of activesubstances in ground-water. 3rd Int. Fresenius conference. Behaviour of pesticides in plants, soils andground-water. Bonn, September 2001. Invited papers.

Vanclooster M. and JD Piñeros Garcet, 2003. Advances in assessessing the quality of PEC ground-waterassessment tools. Submitted

Vanclooster M., Piñeros Garcet J.D. , Boesten , Van den Berg F. , Leistra M., Smelt J., Jarvis N., Burauel P.,Vereecken H., Wolters A., Linnemann V., Fernandez E., Trevisan M., Capri E., Klein M., Tiktak A., Van derLinden A., De Nie D., Bidoglio G., Baouroui F., Jones A., and A. Armstrong, 2003. Effective approaches forpredicting environmental concentrations of pesticides. In: A. Del Re, M. Trevisan and E. Capri (eds), XIIPesticide chemistry symposium. Pesticide in air, plant soil and water system. Piacenza, Italy 4-6 June 2003.

Vanclooster M., JD Piñeros Garcet, J. Boesten, N. Jarvis, H. Vereecken, E. Fernandez, A. Tiktak, M. Kleine, G.Bidoglio, 2001. Effective approaches for assessing the predicted environmental concentrations of pesticides.EGS General Assembly March 2001, Geophysical Research Abstracts 3/CD

Vanclooster, M., M. Javaux, F. Hupet, S. Lambot, A. Rochdi, J.D. Piñeros-Garcet and C. Bielders, 2002.Effective approaches for modelling chemical transport in soils supporting soil management at the largerscale. In: M. Pagliai and R. Jones (Eds.). Sustainable land management – environmental protection. A soilphysical approach. Advances in geo-ecology 35: 171-184.

Van den Berg, F. A. Wolters, N. Jarvis, M. Klein, J.J.T.I. Boesten, M., Leistra, V. Linnemann, J.H. Smelt, H.Vereecken (2003): Pesticide volatilization from soil: Improvement of concepts for pesticide volatilizationfrom bare soil in PEARL, PELMO, and Macro models. Proceedings. XII. Symp. Pestic. Chem., 973-984.

Wolters, A. V. Linnemann, M. Herbst, M. Klein, A. Schäffer, and H.Vereecken (2003), Pesticide Volatilizationfrom Soil: Field-like Measurements vs. Predictions of European Registration Models. J. Environ. Qual.,32:1183-1193.

Wolters, A. , M. Leistra, V. Linnemann, J.H. Smelt, F. Van den Berg, M., Klein, N. Jarvis, J.J.T.I. Boesten, H.Vereecken (2003): Pesticide volatilization from plants: Improvement of the PEARL, PELMO, and Macromodels. Proceedings. XII. Symp. Pestic. Chem., 985-994.

Wolters, A. , T. Kromer, V. Linnemann, A. Schäffer, H. Vereecken. (2003), A new tool for laboratory studies onvolatilization: extension of applicability of the photovolatility chamber. Environ. Toxicol. Chem., 22 (4),791-797.


Wolters, A. (2003): Pesticide volatilization from soil and plant surfaces: Measurements at different scales versusmodel predictions. Aachen 127 p., RWTH, Diss.v.27.4.2003

Wolters, A. T. Kromer, V. Linnemann, H. Ophoff, A. Stork, H. Vereecken (2002), Volatilization of[14C]fluoranthene and [14C]diflufenican after soil surface application under field-like conditions:measurement and comparison with different model approaches. Agronomie. 22 (4), 337-350.

Wolters, A. V. Linnemann, M. Herbst, H. Vereecken (2002): Volatilization of pesticides from soil:measurement under field-like conditions and comparison with model calculation. 10th IUPAC InternationalCongress on the chemistry of crop protection, 4.-9.8.02, Basel, CH.

Wolters, A.; Kromer, T.; Linnemann, V.; Schäffer, A.; Vereecken, H.: A new tool for laboratory studies onvolatilization: extension of applicability of the photovolatility chamber. SETAC Europe Conference: OrganicSoil Contaminants, 2.-5. September 2001, Kopenhagen, Denmark.

Effective approaches for predicting environmental concentrations of pesticides: the apecop project

Documents

Transcript of Effective approaches for predicting environmental concentrations of pesticides: the apecop project