General enquiries on this form should be made to:
Defra, Science Directorate, Management Support and Finance Team,
Telephone No. 020 7238 1612
E-mail:
[email protected]
SID 5



Research Project Final Report
Note
In line with the Freedom of Information
Act 2000, Defra aims to place the results
of its completed research projects in the
public domain wherever possible. The
SID 5 (Research Project Final Report) is
designed to capture the information on
the results and outputs of Defra-funded
research in a format that is easily
publishable through the Defra website. A
SID 5 must be completed for all projects.
1.
Defra Project code
2.
Project title
This form is in Word format and the
boxes may be expanded or reduced, as
appropriate.
3.
ACCESS TO INFORMATION
The information collected on this form will
be stored electronically and may be sent
to any part of Defra, or to individual
researchers or organisations outside
Defra for the purposes of reviewing the
project. Defra may also disclose the
information to any outside organisation
acting as an agent authorised by Defra to
process final research reports on its
behalf. Defra intends to publish this form
on its website, unless there are strong
reasons not to, which fully comply with
exemptions under the Environmental
Information Regulations or the Freedom
of Information Act 2000.
Defra may be required to release
information, including personal data and
commercial information, on request under
the Environmental Information
Regulations or the Freedom of
Information Act 2000. However, Defra will
not permit any unwarranted breach of
confidentiality or act in contravention of
its obligations under the Data Protection
Act 1998. Defra or its appointed agents
may use the name, address or other
details on your form to contact you in
connection with occasional customer
research aimed at improving the
processes through which Defra works
with its contractors.
SID 5 (Rev. 3/06)
Project identification
PS2228
Characterisation and modelling of time-dependent
sorption of pesticides
Contractor
organisation(s)
Central Science Laboratory
Sand Hutton
York
YO41 1LZ
54. Total Defra project costs
(agreed fixed price)
5. Project:
Page 1 of 18
£
58,380
start date ................
01 April 2007
end date .................
31 January 2009
6. It is Defra’s intention to publish this form.
Please confirm your agreement to do so. ................................................................................... YES
NO
(a) When preparing SID 5s contractors should bear in mind that Defra intends that they be made public. They
should be written in a clear and concise manner and represent a full account of the research project
which someone not closely associated with the project can follow.
Defra recognises that in a small minority of cases there may be information, such as intellectual property
or commercially confidential data, used in or generated by the research project, which should not be
disclosed. In these cases, such information should be detailed in a separate annex (not to be published)
so that the SID 5 can be placed in the public domain. Where it is impossible to complete the Final Report
without including references to any sensitive or confidential data, the information should be included and
section (b) completed. NB: only in exceptional circumstances will Defra expect contractors to give a "No"
answer.
In all cases, reasons for withholding information must be fully in line with exemptions under the
Environmental Information Regulations or the Freedom of Information Act 2000.
(b) If you have answered NO, please explain why the Final report should not be released into public domain
Executive Summary
7.
The executive summary must not exceed 2 sides in total of A4 and should be understandable to the
intelligent non-scientist. It should cover the main objectives, methods and findings of the research, together
with any other significant events and options for new work.
The overall aim of this research was to deliver knowledge on the factors controlling time-dependent
sorption of pesticides in soil and to propose a robust approach to incorporate descriptions of timedependent sorption into pesticide fate models. Data sets on time-dependent sorption were collected. Total
residues of the pesticide remaining in soil at different times from application and concentrations in soil
solution were recorded. Pesticide concentrations in the soil solution were either measured directly using a
centrifugation method or indirectly using an aqueous extraction method. Apparent sorption coefficients for
each time point were calculated as the ratio of sorbed to dissolved concentrations. After applying quality
criteria, 195 laboratory datasets were considered suitable for this project, covering 28 pesticides and 50
soils.
Three models were systematically fitted to the datasets:
(1) The Walker model gives an empirical description of the increase in apparent Kd values with time. It
has two parameters, the initial sorption coefficient and a rate constant describing the increase in sorption
over time. The increase in sorption relative to the initial apparent Kd (relative rate) was also calculated.
(2) The PEARLNEQ model assumes instantaneous sorption on one fraction of the sorption sites and
slow sorption on the remaining fraction. Four parameters were optimised against total pesticide residues
and concentrations in soil solution: the initial concentration of the pesticide, the degradation rate
constant, the sorption coefficient for instantaneous sorption and the ratio of non-equilibrium to
equilibrium sorption. The desorption rate constant was fixed to avoid interdependency of model
parameters. PEARLNEQ is incorporated into the pesticide leaching model PEARL.
(3) A diffusion model which describes time-dependent sorption by radial diffusion and sorption inside
particles. This model would be a possible alternative for incorporation into pesticide leaching models.
The optimised parameters were the same as for PEARLNEQ. The diffusion rate coefficient was fixed.
Most datasets were described reasonably well by the models except where the measurements were very
scattered. PEARLNEQ and the diffusion model gave a similar fit to the data. There was no clear
advantage of one model over the other.
The model parameters differed between pesticides and soils and varied considerably within the same
soil-pesticide combination. The experimental method used in the studies influenced the parameters to
some extent. In one subset of data, the increase in sorption with time was overall slightly larger when the
concentration of the pesticide in soil solution was determined after shaking the soil with aqueous solution
than after extraction by centrifugation. This was not confirmed in additional sorption studies carried out
within this project. The soil incubation temperature had a positive effect on the increase in sorption with
SID 5 (Rev. 3/06)
Page 2 of 18
time for some datasets. A consistent effect of moisture could not be demonstrated based on the
available data. It is expected that a standardisation of experimental protocols for time-dependent
sorption studies will improve the reproducibility of time-dependent sorption studies with the same
pesticide-soil combination somewhat.
Single and multiple regression techniques were used to determine which pesticide properties, soil
properties and experimental factors affect time-dependent sorption. The sorption rate coefficient of the
Walker model was positively related to the initial apparent Kd (i.e. sorption increased to a greater extent
for more strongly sorbed compounds). Other important factors were soil pH and organic carbon content
and the method used to extract the soil solution. In total, 62% of the variance in the log(rates) could be
explained by a multiple regression model. Only 37.6% of the total variance in the log (relative rates)
could be explained by a multiple regression model.
Soil pH was the most influential variable for the log(ratio) between equilibrium and non-equilibrium sorption
derived with PEARLNEQ and the diffusion model. Sorption increased more strongly over time in soil with
lower pH. The multiple regression models could only explain 36.5% of the variance in PEARLNEQ
log(ratios) and 23.0 % of the variance in the log(ratios) for the diffusion model.
Single regressions showed a negative relationship between the ratio of non-equilibrium to equilibrium
sorption within the PEARLNEQ model and pesticide availability in the soil solution. This supports the
theory that time-dependent sorption may be more prominent for more strongly adsorbed compounds.
There is also a negative relationship of the ratio with soil pH and a weak positive relationship with soil
organic carbon content. The compounds that were analysed are not ionic and the direct effect of pH on
sorption is likely to be limited. Slow diffusion of pesticide to sorption sites inside organic soil particles may
explain the positive influence of the organic carbon content on the ratio. However, soil pH, organic carbon
content and pesticide availability are correlated and it is not clear which of the variables have a direct
influence on time-dependent sorption.
Time-dependent model parameters (desorption rate constant, ratio of non-equilibrium to equilibrium
sorption) are required for higher tier regulatory modelling of pesticide leaching through soil. The aim of this
project was to develop generic guidance on how to generate model parameters for new pesticide-soil
combinations. The statistical relationships between model parameters and basic soil and pesticide
properties were tested against results from additional sorption studies carried out within this project. The
analysis showed that the relationships are too weak to allow a robust estimation of parameters for new
pesticide-soil combinations.
The additional sorption data showed a large variability between the extent of time-dependent sorption for
the same compound in different soils. Modelling the behaviour of a substance with parameters averaged
over a small number of experimental studies will thus carry a high degree of uncertainty. The use of
conservative default values could be an alternative.
Project Report to Defra
8.
1
As a guide this report should be no longer than 20 sides of A4. This report is to provide Defra with
details of the outputs of the research project for internal purposes; to meet the terms of the contract; and
to allow Defra to publish details of the outputs to meet Environmental Information Regulation or
Freedom of Information obligations. This short report to Defra does not preclude contractors from also
seeking to publish a full, formal scientific report/paper in an appropriate scientific or other
journal/publication. Indeed, Defra actively encourages such publications as part of the contract terms.
The report to Defra should include:
 the scientific objectives as set out in the contract;
 the extent to which the objectives set out in the contract have been met;
 details of methods used and the results obtained, including statistical analysis (if appropriate);
 a discussion of the results and their reliability;
 the main implications of the findings;
 possible future work; and
 any action resulting from the research (e.g. IP, Knowledge Transfer).
INTRODUCTION
The sorption of a pesticide to soil constituents determines its availability to non-target organisms and its potential to
move to groundwater or surface waters. Pesticide fate modelling at the first tier of regulatory environmental risk
assessments assumes that pesticide sorption is instantaneous and fully reversible. This implies that sorption
coefficients are constant with time. However, sorption has been observed to increase with increasing time of
interaction with the soil (e.g. Walker and Jurado-Exposito, 1998; Cox and Walker, 1999). Research for DEFRA
SID 5 (Rev. 3/06)
Page 3 of 18
project PS2206 has confirmed that amounts of pesticide in the soil solution are constantly changing. Inclusion of
the change in sorption with time in pesticide fate models would significantly improve the accuracy of model
predictions. Indeed, many applicants are including such components already on an ad-hoc basis. Some models
(e.g. PEARL) have the functionality to describe time-dependent sorption (Leistra et al., 2000), whilst other
approaches have been proposed but not yet incorporated into the models (e.g. the diffusion-based approach
developed at Cranfield University within DEFRA project PL0541 and BBSRC project 63/D14743 (Beulke et al.,
2004; Renaud et al., 2004; van Beinum et al., 2005; 2006)). The mechanisms controlling kinetic sorption are not
fully understood and it is not clear which mathematical description gives the best reflection of the underlying
processes. There is a need for (i) comprehensive evaluation of alternative descriptions to allow recommendation
of the most robust, and (ii) development of guidance on how to select parameters for input to pesticide fate
models. The current project aimed to achieve this by exploiting existing sources of data including the dataset
generated within PS2206. The specific objectives were:
1.
To use existing data to identify the main controls on time-dependent sorption;
2.
To identify the most appropriate mathematical description of time-dependent sorption;
3.
To develop generic guidance on how to generate parameters to give a robust description of timedependent sorption;
4.
To generate new experimental data to independently test the guidance.
2
MATERIALS AND METHODS
2.1
Collation of datasets
Data sets on time-dependent sorption were collected from the open literature and PhD theses. Representatives
from industry and research organisations were encouraged to provide additional data. Raw data from recent
research for DEFRA project PS2206 (‘Time-dependent sorption processes in soil’) were provided by Warwick
HRI. Project PS2206 investigated time-dependent sorption of five pesticides (azoxystrobin, chlorotoluron,
cyanazine, imidacloprid, mecoprop-P) in two soils. The properties of the two test soils are given in Table 1. All
pesticide-soil combinations were incubated at five initial concentrations. Subsamples were taken at several time
intervals and analysed for total residues and concentration in soil solution. Parallel soil samples were extracted
with aqueous solution or centrifuged to determine the pesticide concentration in the soil solution. This resulted in
10 data sets per soil-pesticide combination for each pesticide. Most datasets for mecoprop-P were not included
after applying quality criteria.
Table 1. Properties of the test soils used in DEFRA project PS2206.
Soil name
Sand (%)
Silt (%)
Clay (%)
pH
Organic matter (%)
Soakwaters
46.6
19.2
34.2
7.08
2.74
Warwick
70.0
6.1
23.9
5.30
4.34
Data sets were included if time-dependent sorption was measured in laboratory incubation studies with sieved
soil where the total residue and the concentration in soil solution was analysed at several time intervals. Pesticide
concentrations in the soil solution were either measured directly using a centrifugation method or indirectly by
shaking with aqueous extraction solution. Where available, the equilibrium sorption coefficient and Freundlich
exponent from a standard shaken batch experiment with the same soil were recorded. Additional information
required to fit mathematical models to the data was compiled (e.g. soil moisture content during incubation,
amount of extraction solution added). Where necessary, the authors were contacted and asked for additional
details. The mass of pesticide sorbed was derived as the difference between the total mass and the amount in
solution. Apparent sorption coefficients (Kd app) for each time point were calculated as the ratio of sorbed to
dissolved concentrations.
Datasets were only included if there were at least 5 sampling times with measurements above the detection limit.
Datasets where the first sample was taken more than 7 days from application were not considered suitable for
this project. Sorption often increases strongly within the first days from application. The time of the first
measurement is thus critical for a robust estimate of the increase in sorption. Studies with negative calculated
apparent Kd values were also rejected.
After applying the criteria above, a total of 195 laboratory datasets were considered suitable for this project,
covering 28 pesticides and 50 soils. These covered a range of properties (Figures 1 and 2). Approximately 60%
of the studies used an aqueous extraction method, 40% used centrifugation. A large proportion of data (43%)
were measured at Warwick HRI within DEFRA project PS2206.
SID 5 (Rev. 3/06)
Page 4 of 18
Figure 2. Clay contents and organic carbon contents of the
50 soils included in the literature database.
80
1000
60
Clay content (%)
10000
-1
Koc (L kg )
Figure 1. Average Koc and DT50 values of the 28 pesticides
included in the literature database, on a logarithmic scale.
100
10
40
20
0
1
1
10
100
0
1000
4
6
8
10
Organic carbon content (%)
DT50 (days)
2.2
2
Fitting of mathematical models
Model predictions of pesticide behaviour in soil using pesticide fate models could be improved by including robust
descriptions of time-dependent sorption. It is thus necessary to evaluate modelling approaches and identify the
description that is most appropriate in the regulatory context. Ideally, the model should have a small number of
parameters that can be derived from simple experiments and describe the observed behaviour well.
A two-site sorption model (PEARLNEQ) and a diffusion model were evaluated in this project. Both models were
systematically fitted to the data. PEARLNEQ is incorporated in the pesticide leaching model PEARL (Leistra et
al., 2000). The diffusion model would be a possible alternative. An empirical model (‘Walker model’) was also
fitted to the data to quantify the increase in sorption with time. Due to its empirical nature, the Walker model is not
suitable for implementation into pesticide fate models that are currently used within pesticide registration.
2.2.1
Walker model
Walker (1987) proposed an empirical description of the relationship between the strength of sorption and the time
from application:
Kd( t )  Kdini  k ads
relative rate 
t
k ads
 100
Kd ini
[1, 2]
where Kd(t) is the apparent sorption coefficient (L kg-1) at time t (days), Kdini is the initial sorption coefficient
(L kg-1), and kads is a rate constant describing the increase in sorption over time (increase in apparent Kd per unit
square root of time). The model was fitted to the apparent Kd values calculated from the measurements. The
fitted adsorption rate constant k ads provides a quantitative measure of the increase in sorption with time. An
alternative measure is the relative rate (% per unit square root of time), which characterises the increase in
sorption relative to the initial apparent Kd (at time zero). The model was fitted to the data by minimising the sum
of squared residuals using Microsoft Excel Solver. First-order degradation kinetics were fitted to the decline in
total residues.
2.2.2
PEARLNEQ
PEARLNEQ (Boesten et al., 2007) assumes that sorption is instantaneous on one fraction of the sorption sites
and slow on the remaining fraction (Figure 3). Degradation of molecules present in liquid phase and sorbed to the
equilibrium site follows first-order kinetics. Molecules sorbed on the slow non-equilibrium sorption site are
considered not to degrade. Equilibrium and non-equilibrium sorption are described by:
n
 C  F
X EQL  K F ,EQL CL,R  L 
C 
 L,R 

 C
dX NEQ

 k des  K F ,NEQ CL ,R  L
C
dt

 L ,R

[3]




nF


 X NEQ 


[4]
where:
CL = concentration in the liquid phase (mg L-1)
CL,R = reference concentration in the liquid phase (mg L-1)
XEQL = mass sorbed at equilibrium sites (mg kg-1)
XNEQ = mass sorbed at non-equilibrium sites (mg kg-1)
KF,EQL = equilibrium Freundlich sorption coefficient (L kg-1)
KF,NEQ = non-equilibrium Freundlich sorption coefficient (L kg-1)
nF = Freundlich exponent (-)
kdes = desorption rate coefficient (d-1)
t = time (d)
SID 5 (Rev. 3/06)
Page 5 of 18
Figure 3. Schematic representation of the PEARLNEQ model showing the soil solution on the right and the equilibrium and nonequilibrium sorption sites on the left.
Freundlich:
KF,EQL nF
equilibrium
Freundlich
KF,NEQ nF
non-equilibrium
sorption
sorption
Desorption Rate Constant:
kdes
Ratio KF NEQ:KF EQL
The model has five parameters: the initial concentration of the pesticide, degradation rate constant, equilibrium
sorption coefficient, ratio of non-equilibrium sorption to equilibrium sorption and the desorption rate constant.
According to the model authors, the equilibrium sorption coefficient can be derived from a standard shaken batch
experiment. A computer programme coupled with the optimisation package PEST is available to estimate the
remaining four parameters from experimental data (PEARLNEQ). It was, however, found in this project that the
sorption coefficient from a standard shaken batch study is sometimes a poor estimate of sorption at the beginning
of the incubation study. Version 4 of PEARLNEQ was thus modified to allow the estimation of all five model
parameters (personal communication M. ter Horst), i.e. the equilibrium Freundlich sorption coefficient that gave
the best fit to the measured data was estimated alongside the other four parameters. The Freundlich exponent
(nf) was fixed at the value that was derived from a batch sorption experiment, or fixed at 1 if no measurements
were available.
The model was fitted simultaneously to the total pesticide residues in soil plotted against time and the
concentrations of the pesticide in soil solution. The smallest possible value for the optimised ratio between nonequilibrium and equilibrium sorption was set to 0.01. An example fit is shown in Figure 4 (grey line). Apparent Kd
values calculated from the simulated data are shown in the right-hand graph. The solution concentration is
controlled by three processes: degradation, equilibrium sorption and the transfer of pesticide between the solution
and the non-equilibrium phase. The decline in soil solution would be parallel to that of the total soil residues if
equilibrium sorption was linear (Freundlich exponent = 1) and instantaneously at equilibrium over the whole
incubation period. In this case, degradation would be the only driving factor. Concentrations generally decline
faster than expected from degradation alone if equilibrium sorption is non-linear with Freundlich exponents <1.
This is because the curvature of the isotherm leads to stronger sorption at smaller concentrations. An increase in
apparent Kd values with increasing time from application can thus be observed even if sorption is instantaneously
at equilibrium (black line in Figure 4). Non-equilibrium sorption results in an additional decline in solution
concentrations and in a stronger increase in apparent Kd values (grey line).
Figure 4. Total mass, concentration in soil solution and apparent Kd simulated with PEARLNEQ with or without nonequilibrium sorption.
Mass (microg)
30
20
10
Non-equilibrium sorption
Equilibrium sorption
Measurements
0
0
20
40
Time (days)
2.2.3
60
40
Non-equilibrium sorption
Equilibrium sorption
Measurements
0.3
Kd app (kg/L)
Concentration (microg/L)
0.4
0.2
0.1
0.0
80
Non-equilibrium sorption
Equilibrium sorption
Measurements
30
20
10
0
0
20
40
Time (days)
60
80
0
20
40
60
80
Time (days)
Diffusion model
An alternative model was developed by van Beinum et al. (2005, 2006). The model assumes that sorption is
limited by slow diffusion inside soil organic particles. Sorption is instantaneous on part of the organic matter
(equilibrium fraction). This fraction is assumed to be present in the form of very fine organic material. The other
fraction consists of particles for which sorption is limited by diffusion. Pesticide molecules move into the particle
by diffusion within the solution inside the particle pores before they are sorbed on the internal surfaces. Diffusion
is the rate-limiting step. Once the pesticide has reached an internal sorptive surface, it is instantaneously bound.
SID 5 (Rev. 3/06)
Page 6 of 18
The diffusion model was implemented using the ORCHESTRA modelling toolkit (Meeussen, 2000). The organic
matter fraction of the soil was divided into an equilibrium and non-equilibrium fraction (Figure 5). Sorption in the
non-equilibrium fraction is described by diffusion into small porous spheres (d = 10 μm, porosity = 0.5). Spherical
diffusion is solved numerically by dividing the sphere into ten concentric layers (Figure 6) and calculating diffusion
across the boundaries between the layers. Freundlich sorption is calculated within each layer. The model
parameters are similar to those for PEARLNEQ, but the desorption rate coefficient is replaced by a diffusion rate
coefficient (D in m 2 day-1). The ratio of equilibrium to non-equilibrium sorption and the diffusion coefficient are
derived by fitting the model to experimental data.
The rate at which pesticides diffuse into the particles is dependent on the diffusion coefficient divided by the
squared particle diameter (D/a2). The model does not account for variable sizes of organic particles and the
diameter of the particles (10 μm) is arbitrary. Note that, as the true particle diameter is unknown, the diffusion
coefficient used in the model has no intrinsic meaning but is only a relative measure for the rate.
subject to
transformation
equilibrium sorption
intraparticle diffusion
Figure 5. Schematic representation of the diffusion model showing the soil
solution outside the porous spheres on the left and the equilibrium and nonequilibrium sorption sites on the right (for simplification only five of the ten
concentric layers are represented in the diagram)
2.2.4
Figure 6. Model representation of organic
particle: migration into the particle is
represented by diffusion between ten
concentric layers
Goodness of fit criteria
The goodness of fit was evaluated for each dataset based on Chi 2 statistics (FOCUS, 2006) and visual
assessment. A model fit was discarded if the model could not adequately describe the data (Chi 2 error value
above 8% and (i) systematic deviation of the modelled line from the measured data or (ii) very large variability in
measurements). Most datasets were described well by the models except where data were very scattered. The
number of datasets that were considered acceptable was 166 for the Walker model and 170 for PEARLNEQ and
the diffusion model.
2.3
Statistical analyses
The quantitative measures of time-dependent sorption derived for the Walker model and the optimised
parameters of the PEARLNEQ and diffusion model were subjected to multi-variate statistical analysis to:
1. Determine the primary controls on time-dependent sorption (e.g. properties of the pesticide or soil);
2. Determine whether any relationships exist between the optimised model parameters and basic soil and/or
pesticide properties. This would make it possible to estimate the parameters for new pesticide-soil
combinations for which measurements are not available.
All available information on pesticide and soil properties and experimental method that could have influenced
time-dependent sorption were collected for each dataset. Single and multiple regression techniques were used to
determine the combination of factors that best explains the values for Rate and RelativeRate (Walker model) and
for the ratio between non-equilibrium and equilibrium sorption (PEARLNEQ and diffusion model).
Table 2 shows all covariates that were included in the statistical analysis. The degradation half-life (DT50),
equilibrium sorption coefficient (Kf) and the total initial mass of pesticide (TC) were taken from the model fit. As a
consequence, these covariates are not independent of the derived rate constant and ratio. However,
independently derived values were in many cases not available or not suitable, and therefore the fitted values
offered the best representation of the pesticide properties. Pesticide availability (AVA) is defined as the fraction of
pesticide that was in soil solution at time zero, calculated from Kf and the soil:solution ratio. The extraction
method (EXTMETHOD) was either ‘centrifugation’ or ‘extraction’ and was included as a binominal variable.
Examination of the data showed that all parameters were highly skewed and that the data needed to be analysed
on a log-transformed scale. Frequency distribution plots of raw data showed that some of the explanatory
variables (DT50, KF, N, AVA, TC) are also skewed and log transformation of these was also appropriate.
SID 5 (Rev. 3/06)
Page 7 of 18
Interaction between variables may be expected if the effect of one factor depends on another factor, for example
if the soil pH may have a stronger effect in soils with a high organic carbon content. Interactions that were
included in the analysis were: KF.DT50, AVA.DT50, SOILSAND.SOILPH, SOILOC.SOILPH and
SOILSAND.SOILOC.
The data were analysed using stepwise linear regression (Montgomery and Peck, 1982). All main effects,
together with the selected interactions were tested and entered or dropped in turn, based on their partial Fstatistics (tested at the 5% significance level). In other words, the model was started using the best single term.
Then, with this term in place, the term improving the model most was tested and only kept if the improvement was
significant at the 5% significance level. Then, as well as testing for the addition of terms, terms already in the
model were also tested for dropping out of the model (if, for instance, they have become redundant as a result of
other terms added to the model after they were originally added themselves). This process was carried on until no
further significant improvement could be made. Simulations (and corresponding 95% confidence intervals) were
then produced and these were compared to the observed data.
Table 2. Covariates that were included in the statistical analysis
Covariate
Description
SOILCLAY
clay content of the soil (%)
SOILSAND
sand content of the soil (%)
SOILOC
organic carbon content of the soil (%)
SOILPH
soil pH
INCTEMP
incubation temperature (C)
INCTIME
incubation time (days) for which measurements were taken (not including non-detect measurements)
EXTMETHOD
extraction method (centrifugation method or extraction with CaCl2 solution)
DT50
degradation half-life derived in model optimisation (days)
KF
equilibrium Freundlich coefficient derived in model optimisation (L kg -1)
NF
Freundlich exponent from batch sorption experiment
AVA
fraction of pesticide in soil solution at time zero
TC
total concentration of pesticide in the soil at the start of the incubation (mg kg-1)
2.4
Sorption experiments
Sorption experiments were undertaken by Warwick HRI to provide independent datasets to test the guidance
developed in this project. Five soils and five pesticides were selected (Table 3). These were chosen after the first
150 datasets collated from the literature had been analysed. Soils with low pH, high organic carbon content and
clay content and strongly sorbed pesticides were included to fill gaps in the available data set. Three replicate
batches for each combination were incubated at 15oC and either 40 or 60% maximum water holding capacity
(mwhc). Parallel samples were taken from each replicate at intervals. One soil sample (10 g) was extracted with
organic solvents (20 mL acetonitrile) for 1 hour on a wrist-action shaker to determine total pesticide residues
remaining. The second sample was shaken with 20 mL 0.01 M CaCl2 solution for 1 hour on a reciprocal shaker at
240 rpm. A third sample was transferred into a polythene syringe body (10 mL) stoppered with cotton wool and
centrifuged at 10,000 rpm over an HPLC vial to extract soil pore water.
Table 3. Pesticides and soils used in the sorption experiments
Koc1
(L kg-1)
DT502
(d)
Soil
Series
Texture
Sand
(%)
Silt
(%)
Propyzamide
123.3-351.0
14.1-98.2
106
Salop
Clay
36
23
41
6.8
4.2
Diuron
202.8-304.3 57.9-188.2
245
Bromsgrove
Sandy Loam
67
18
15
5.1
4.8
Paclobutrazol
135.6-437.6 55.3-123.5
261
Arrow
Sandy Loam
68
17
15
1.9
6.2
265
Fladbury
Clay
47
14
39
3.4
6.2
Pesticide
Metamitron
66.6-98.6
15.4-20.4
Clay
Organic
pH
(%) carbon (%) (H2O)
Metazachlor
75.6-132.5 10.3-26.4
282AA
Elmton
Sandy Clay Loam 61
19
20
5.4
7.3
1 Range of 24-hour batch Koc values in the five soils.
2 Range of DT50 values in the five soils. Derived by fitting a first-order curve to the measurements at 15 oC and 40% mwhc.
Equilibrium sorption was measured in shaken batch experiments for 1 hour and for 24 hours. The experiments
were performed in triplicate for each soil-pesticide combination and at five concentration levels. Pesticide solution
(in 0.01M CaCl2) was added to centrifuge tubes with 10 g moist soil and left on a reciprocal shaker (240 rpm) for
either 1 hour or 24 hours before centrifugation (2 min at 6000 rpm). Soil:solution ratios were 20:10 or 20:5 (g/mL)
depending on the soil.
SID 5 (Rev. 3/06)
Page 8 of 18
All study compounds were analysed by high performance liquid chromatography (HPLC) with an ultra violet (UV)
absorption detector at 220 nm. The HPLC system used was a Kontron 320 pump, Kontron 332 UV detector,
Kontron 360 Autosampler with Kromasystem 2000 PC based data system. A Lichrosorb C18 column (250 x 46
mm) was fitted and a mobile phase of 60:40 acetonitrile:water (metamitron) or 75:25 acetonitrile:water
(propyzamide, diuron, paclobutrazol and metazachlor) was used.
3
RESULTS
3.1
Distributions of model parameters
3.1.1
Walker model
The Walker model was fitted to a total of 195 data sets. For 29 of the data sets the goodness of fit was not
considered adequate. Figure 7 shows the distributions of the adsorption rate constants and the relative rates
derived by fitting the Walker model to the remaining 166 data sets. The distribution of the rate constants is
relatively wide with a median of 0.28 L kg-1 day-1/2 and a 90th percentile of 1.58 L kg-1 day-1/2.
The relative rate characterises the increase in sorption relative to the initial level. For example, if the initial
apparent Kd value is 5 L kg-1 and the relative rate is 10% day-1/2, then the apparent Kd increases by 0.5 L kg-1 for
each unit square root of day. This gives a total increase of 5 L kg -1 over 100 days (100 days =10 units square root
of day), i.e. the apparent Kd increases to twice its original level. The median of the relative rate for the 166
datasets is 14.6 % day-1/2 (corresponding to an increase by a factor of 2.5 over 100 days) and the 90 th percentile
is 62.8 % day-1/2 (factor 7.3 over 100 days).
100
100
80
80
Frequency (%)
Frequency (%)
Figure 7. Distribution of adsorption rate constants (WM Rate) and relative sorption rates (WM RelRate) from fitting the
Walker model to the 166 datasets.
60
40
20
60
40
20
0
0
0
3.1.2
1
2
WM Rate (L kg-1 day-1/2)
3
0
100
200
300
400
WM RelRate (% day-1/2)
500
PEARLNEQ
The goodness of fit was not considered adequate for 25 out of the 195 data sets described by the PEARLNEQ
model. Figure 8 and 9 show the distributions of the desorption rate coefficients and of the ratios between nonequilibrium and equilibrium sorption (Ratio) derived by fitting the PEARLNEQ model to the remaining 170 data
sets. The median was 0.05 day-1 for the rate coefficient and 0.65 for the ratio (i.e. non-equilibrium sorption = 65%
of equilibrium sorption). The distributions are very wide. This is because the parameter that represents the extent
of non-equilibrium sorption (ratio) and the parameter that determines the rate of the increase in sorption are often
interrelated (i.e. changes in one parameter can be partly compensated by changes in the second parameter). As
a result, both parameters carry a high degree of uncertainty. To avoid the problem of interdependency, the
PEARLNEQ model was fitted to the datasets a second time with a fixed rate coefficient (0.05 day -1) so that only
the parameter representing the extent of non-equilibrium sorption (ratio) was fitted to the data. This resulted in a
much narrower distribution of ratios (Figure 9). The median of the new distribution of ratios was 0.39 (i.e. nonequilibrium sorption = 39% of equilibrium sorption). In 87% of the cases, the ratio was less than 1 (i.e. nonequilibrium sorption was smaller than equilibrium sorption). The 10 th-90th percentile interval was 0.09–1.11.
Sorption did not increase with time for 8 datasets. The optimised ratio reached the lower boundary allowed in the
fitting (0.01) for these studies.
SID 5 (Rev. 3/06)
Page 9 of 18
Figure 8. Distribution of sorption rate constants (PEARLNEQ Rate) and the ratios of non-equilibrium sorption to equilibrium
sorption (PEARLNEQ Ratio) from fitting the PEARLNEQ model to the 170 datasets.
Frequency (%)
100
80
60
40
20
0
0.0
0.1
0.2
0.3
0.4
PEARLNEQ Rate (day-1)
0.5
100
100
80
80
Frequency (%)
Frequency (%)
Figure 9. Distribution of the ratios of non-equilibrium sorption to equilibrium sorption from fitting the PEARLNEQ model
with a variable sorption rate constant (left) and a fixed value of 0.05 d -1 (right).
60
40
20
40
20
0
0
0
3.1.3
60
1
2
3 4 5 6 7 8 9 10 11
PEARLNEQ Ratio (-)
0
1
2
3 4 5 6 7 8 9 10 11
PEARLNEQ Ratio (-)
Diffusion model
Two of the parameters of the diffusion model are interrelated (the diffusion rate constant and the ratio of nonequilibrium sorption to equilibrium sorption). The model was initially fitted to all datasets with variable parameters
and the goodness of fit was assessed. The median of the diffusion rates for all acceptable fits was 2 x 10 -11 m2
day-1. The diffusion rate constant was fixed to this value and the fits were repeated. A total of 170 fits were
considered acceptable. The median ratio of non-equilibrium sorption to equilibrium sorption for these 170 fits is
0.66. The 10 th-90th percentile interval is 0.10-1.95. Sorption showed no increase in sorption with time for 8
datasets (the optimised ratio reached the lower boundary of 0.01).
3.2
Comparison between the PEARLNEQ model and the diffusion model
PEARLNEQ is often fitted to regulatory data on time-dependent sorption. The derived parameters are used in the
pesticide leaching model PEARL to calculate higher-tier predicted environmental concentrations in groundwater.
In this project, the diffusion model was tested as a potential alternative to PEARLNEQ.
Apparent Kd values (sorbed divided by dissolved concentration) were calculated from the fitted total simulated
mass and solution concentration and compared with apparent Kd values calculated from the measurements.
Some examples are shown in Figure 10. Note that the simulated results for PEARLNEQ presented in Figure 10
are based on model fits with a variable desorption rate constant. The diffusion model was fitted with a variable
diffusion rate constant. This was necessary for a valid comparison of the ability of the two models to describe the
observed behaviour.
Two patterns in apparent Kd values could be observed: (1) sorption increases rapidly at the beginning of the
experiment and reaches or approaches a plateau, or (2) sorption increases continuously at a lower rate and a
plateau is not reached within the experimental period. Both models were able to describe these patterns well.
Close examination of the curves shows that the diffusion model tends to simulate a slightly faster initial increase
in sorption than PEARLNEQ. Overall the difference is too small to assign a preference for either model.
The models are very similar in that they use the same number of fitted parameters and these parameters
describe similar properties. Both models describe non-equilibrium sorption by a capacity (ratio Neq/Eq) and a rate
constant. The diffusion model simulates a slightly sharper initial increase due to the multiple layers that are used
to represent the non-equilibrium sorption phase. The different concepts of the two models (pore diffusion in the
diffusion model and kinetic sorption in PEARLNEQ) mean that there is no direct correlation between the diffusion
coefficient in the diffusion model and the rate coefficient in PEARLNEQ.
SID 5 (Rev. 3/06)
Page 10 of 18
Figure 10. Apparent Kd values calculated from measurements and from data simulated with the diffusion model and PEARLNEQ.
Kd app (L/kg)
Kd app (L/kg)
10
8
6
4
0.7
7
0.6
6
5
4
3
2
Diffusion model
PEARLNEQ
Measurements
2
8
Kd app (L/kg)
12
Diffusion model
PEARLNEQ
Measurements
1
0
20
40
60
80
100
0.4
0.3
0.2
Diffusion model
PEARLNEQ
Measurements
0.1
0
0
0.5
0.0
0
20
Time (days)
40
60
80
0
20
Time (days)
40
60
80
Time (days)
The PEARLNEQ desorption rate constant and the diffusion rate constant were then fixed in order to derive robust
ratios of non-equilibrium to equilibrium sorption. The desorption rate constant was set to 0.05 day-1 and the
diffusion rate constant to 2 x 10-11 m2 day-1. Chi2 error values were calculated to assess the goodness of fit of both
models. Chi2 error values for the diffusion model are plotted against those for PEARLNEQ in Figure 11. Total
pesticide residues were described equally well by both models (all points are close to the 1:1 line in Figure 11).
The diffusion model provided a better fit to the measured concentrations in solution in 60% of the cases (points
below the 1:1 line in Figure 11). PEARLNEQ described the data better in 40% of the cases. The differences
between the two models is partly due to the fact that the rate constants were not included in the fitting. As a
result, some data sets were described better by one model than the other. The goodness of fit is thus biased by
the choice of the fixed rate constants.
Chi ConLiq diffusion model (%)
25
20
15
10
5
30
20
10
2
2
Chi mass diffusion model (%)
Figure 11. Chi2 error values for the total mass of pesticide in soil (left) and the concentration in the liquid phase (right)
simulated with the diffusion model and PEARLNEQ.
0
0
5
10
15
20
25
Chi2 mass PEARLNEQ (%)
0
0
10
20
30
Chi2 ConLiq PEARLNEQ (%)
3.3
Influence of experimental conditions on time-dependent sorption
3.3.1
Extraction method
The amount of pesticide available in soil solution is either characterised by shaking the soil with aqueous solution
or by centrifugation. Warwick HRI applied both methods to parallel samples from two soils treated with five
pesticides at five initial concentrations (DEFRA project PS2206). This provides an excellent dataset for
comparison between the two experimental methods.
The apparent Kd value increased to a larger extent when the shaking method was used than for the centrifugation
method in most of the studies by HRI. Figure 12 shows a comparison of the sorption rate constant of the Walker
model derived with the two methods. Most points lie above the 1:1 line.
The effect of the extraction method on the increase in apparent Kd values over time is mainly due to the fact that
the sorption isotherms for the tested pesticides are non-linear. For Freundlich exponents < 1, the apparent Kd
value increases with decreasing concentrations of the pesticide in the soil solution. Degradation leads to a decline
in solution concentrations with time. The increase in the apparent Kd is stronger for smaller initial concentrations.
Shaking of the soil with an aqueous solution leads to smaller pesticide concentrations in soil solution than
centrifugation due to dilution. The increase in sorption over time is thus more pronounced when the pesticide in
soil solution is extracted by shaking compared with centrifugation. The finding that sorption rate constants of the
Walker model are larger for the shaking than the centrifugation method is, therefore, an artefact than can be
explained by the empirical nature of the model.
SID 5 (Rev. 3/06)
Page 11 of 18
PEARLNEQ ratio shaking (-)
Figure 13. Comparison of PEARLNEQ ratios from datasets
using the shaking or centrifugation method to extract
pesticide in soil solution.
3
-1
WM-Rate Shaking (L kg d
-1/2
)
Figure 12. Comparison of adsorption rate constants (Walker
model) from datasets using the shaking or centrifugation
method to extract pesticide in soil solution.
2
1
0
0
1
2
2
1
0
3
-1
WM-Rate Centrifugation (L kg d
-1/2
0
)
1
2
PEARLNEQ ratio centrifugation (-)
Figure 13 shows the ratios between non-equilibrium and equilibrium sorption estimated with the PEARLNEQ
model for the datasets from DEFRA project PS2206. The extent of non-equilibrium sorption (ratio) was overall
slightly larger when the soil was shaken with aqueous solution than for the centrifugation method. The
discrepancy between the two extraction methods was much smaller for the ratios of the PEARLNEQ model than
for the rate constants of the Walker model (Figure 12). This is due to the fact that PEARLNEQ accounts for nonlinear sorption within the model.
The ratios between non-equilibrium and equilibrium sorption estimated with the diffusion model were also
somewhat larger for the shaking than the centrifugation method (not shown).
3.3.2
Soil incubation temperature and moisture
Some authors investigated time-dependent sorption for the same pesticide-soil combination incubated at different
temperatures and moisture contents. For several datasets, the increase in apparent Kd values was somewhat
greater at higher temperatures. An example is shown in Figure 14. This is partly due to the fact that degradation
generally increases with increasing temperature. The decline in concentrations leads to an increase in apparent
Kd values due to non-linear sorption (see above).
The PEARLNEQ ratio between non-equilibrium and equilibrium sorption is plotted against study temperature in
Figure 15 for some of the datasets. Each line represents a study where a pesticide was incubated in the same
soil at different temperatures. Temperature did not always influence the ratio, but a positive effect was found in
some studies. PEARLNEQ accounts for non-linear sorption and the estimated model parameters are thus
independent of sorption non-linearity. Degradation may have an additional effect on time-dependent sorption.
PEARLNEQ assumes transformation in the soil solution and the equilibrium pool only. These are likely to be
depleted faster at higher temperatures. This will shift the ratio of non-equilibrium to equilibrium sorption towards
the non-equilibrium state (i.e. increase the ratio). There may also be a direct effect of temperature on the diffusion
of pesticide into soil particles and on sorption reactions.
Figure 15. PEARLNEQ ratios for different incubation
temperatures.
2.5
1.6
1.2
5°C
20°C
0.8
0.4
PRL-Ratio (-)
-1
Apparent Kd (mL g )
Figure 14. Apparent Kd values of chlorotoluron at different
incubation temperatures (data provided by HRI).
2.0
1.5
1.0
0.5
0.0
0.0
0
10
20
0
30
10
20
Temperature (oC)
Time (d)
Wetter soil conditions are likely to favour degradation and enhance the diffusion of pesticide into and out of soil
particles. Data for comparison of time-dependent sorption at different moisture contents were, however, limited.
There was no consistent effect of soil moisture (in % of maximum water holding capacity) during the incubation on
the extent of time-dependent sorption (data not shown).
SID 5 (Rev. 3/06)
Page 12 of 18
3.4
Statistical relationships between the extent of time-dependent sorption and soil/pesticide
properties
Multiple regression analyses were undertaken to determine which factors affect time-dependent sorption. The
initial apparent Kd of the pesticide was identified as the most important factor in the regression for the empirical
sorption rate coefficient (Walker model). The absolute increase in sorption over time is greater for larger
optimised initial apparent Kd values. Other important factors were the soil pH, the method used to extract the soil
solution soil and the organic carbon content. Figures 16 and 17 show the individual regressions for log(Rate)
against log(Kd) and pH. In total, 62% of the variance in the log(rates) could be explained by the multiple
regression model.
Figure 17. Relationship between sorption rate constant
fitted with the Walker model and the soil pH.
1
1
0
0
log(SM-Rate)
log(SM-Rate)
Figure 16. Relationship between sorption rate constant
fitted with the Walker model and the sorption coefficient of
the pesticide at the start of the experiments.
-1
-2
-3
-1
-2
-3
R2 = 0.43
R2 = 0.21
-4
-4
-2
-1
0
Log(Kd)
1
4
2
5
6
7
8
9
pH
Soil pH was identified as the most important factor influencing log(relative rate) derived with the Walker model.
The relative rate characterises the increase in apparent Kd values relative to the initial level. The regression could
be improved slightly by including (i) the interaction between the degradation rate constant and pesticide
availability in soil solution at the start of the experiment, and (ii) the method used to extract the soil. The
relationship between the relative rate and all factors included in the regression is weak. Only 37.6% of the total
variance in the log (relative rates) could be explained by the regression model.
Soil pH was also the most influential variable for the log(ratio) between equilibrium and non-equilibrium sorption
derived with PEARLNEQ (r2 for the single regression = 0.22) and the diffusion model (r 2 = 0.18). All other
variables tested had a very weak influence on the regressions. The regression models could only explain 36.5%
of the variance in PEARLNEQ ratios and 23.0 % of the variance in the ratios for the diffusion model (Figure 18).
1
log(observed Diff-Ratio)
log(observed PRL-Ratio)
Figure 18. Ratio of the PEARLNEQ model (left) and the diffusion model (right) predicted with the regression model and
observed values in a 1:1 plot.
0
-1
-2
1
0
-1
-2
-2
SID 5 (Rev. 3/06)
-1
0
log(predicted PRL-Ratio)
1
-2
Page 13 of 18
-1
0
log(predicted Diff-Ratio)
1
3.5
Results of additional sorption studies and analysis of combined datasets
Sorption experiments with five pesticides and five soils were carried out within this project. The soils were
incubated at two different moisture contents. PEARLNEQ was fitted to the total residues and concentrations in the
aqueous phase measured after CaCl2 extraction or centrifugation (the volume of water extracted by centrifugation
was too small for most soils incubated at 40% mwhc). The desorption rate was fixed at 0.05 day -1. The fit was
acceptable for 62 of the 80 datasets.
The PEARLNEQ ratio between equilibrium and non-equilibrium sorption for the new datasets ranged from the
minimum that was allowed for in the model fitting (0.01) to 4.21 with a median of 0.66. The median is considerably
larger than the median of the distribution of the literature data (0.39). There were large discrepancies between the
ratios for the same pesticide (Figure 19, left; Table 4). The ratios for each of the five soils also showed a large
variability (Figure 19, right). There was no consistent influence of soil moisture during incubation and no
consistent effect of the method used to extract the soil solution on the ratio between non-equilibrium and
equilibrium sorption.
PEARLNEQ Ratio (-)
PEARLNEQ Ratio (-)
Figure 19. PEARLNEQ ratios for the new sorption studies
4
3
2
1
4
3
2
1
0
0
Propyzamide
Paclobutrazol
Metazachlor
Soil 106
Soil 261
Soil 282AA
Diuron
Metamitron
Soil 245
Soil 265
Table 4: PEARLNEQ ratios derived from sorption experiments with five pesticides and five soils
(CaCl2 extraction, incubation at 40% mwhc)
Soil 106
Soil 245
Soil 261
Soil 265
Soil 282AA
Propyzamide
2.70
1.80
0.38
0.17
0.36
Diuron
0.66
0.73
0.44
0.71
n.a.
Paclobutrazol
n.a.
n.a.
n.a.
1.39
0.73
Metamitron
0.55
0.73
0.14
n.a.
0.01
Metazachlor
0.74
1.51
0.01
0.13
0.02
n.a. = fit not acceptable
The results from the additional studies were combined with those from the literature. Relationships between log
transformed ratios and selected soil and pesticide properties were evaluated for the complete set of data. There
was a significant relationship between log(ratio) and log(availability) of the pesticide in solution at the start of the
experiment (p<0.001; Figure 20 a). The smaller the availability the greater the increase in sorption over time. This
supports the theory that time-dependent sorption may be more prominent for compounds that are more strongly
adsorbed in soil. There is a negative relationship of the ratio with soil pH (p<0.001; Figure 20 b) and a weak
positive relationship with soil organic carbon content (p<0.01; Figure 20 c). The compounds are not ionic and the
direct effect of pH on sorption is likely to be limited. Slow diffusion of pesticide to sorption sites inside organic soil
particles is a possible cause of time-dependent sorption. This could explain the positive influence of the organic
carbon content on the ratio. However, soil pH and organic carbon content are correlated with each other and with
pesticide availability (Figure 20 d-f). It is, therefore, not clear which of the variables have a direct influence on
time-dependent sorption.
SID 5 (Rev. 3/06)
Page 14 of 18
Figure 20: Relationships between the PEARLNEQ ratio and pesticide availability (a), ratio and soil pH (b), ratio and organic
carbon content (c), soil pH and organic carbon content (d), availability and pH (e), availability and organic carbon content
(f) for the combined dataset.
8
a
d
R2 = 0.23
0
7
pH
log(PRL-Ratio)
1
-1
-2
6
5
Literature data
New experiments
Literature data
New experiments
2
R = 0.09
-3
4
-2.5
-2.0
-1.5
-1.0
log(Availability)
-0.5
0.0
1
0
2
4
6
Organic carbon content (%)
8
0.0
b
e
log(Availability)
log(PRL-Ratio)
-0.5
0
-1
-2
Literature data
New experiments
-1.0
-1.5
-2.0
-2.5
R2 = 0.25
-3
Literature data
New experiments
R2 = 0.13
-3.0
4
5
6
pH
7
8
1
4
5
6
pH
7
8
0.0
c
f
log(Availability)
log(PRL-Ratio)
-0.5
0
-1
-2
Literature data
New experiments
-1.0
-1.5
-2.0
-2.5
R2 = 0.04
-3
Literature data
New experiments
R2 = 0.14
-3.0
0
SID 5 (Rev. 3/06)
2
4
6
Organic carbon content (%)
8
Page 15 of 18
0
2
4
6
Organic carbon content (%)
8
3.6
Development and testing of draft guidance
One of the objectives of this project was to develop generic guidance on how to generate parameters that give a
robust description of time-dependent sorption in regulatory modelling. Experimental studies on time-dependent
sorption involve considerable effort and costs. It would thus be advantageous if parameters for new pesticide-soil
combinations could be derived from basic soil and/or pesticide properties. The regression model that was derived
from statistical analyses of literature data was used to predict PEARLNEQ ratios between non-equilibrium and
equilibrium sorption for the additional sorption studies undertaken within this project. The observed ratios are
plotted against the predicted values in Figure 21. The results suggest that this approach cannot be
recommended. There is a large scatter of the data around the 1:1 line. Any prediction of parameters for new
pesticide-soil combinations using the regression model would thus carry a high degree of uncertainty.
log(observed PRL-Ratio)
Figure 21. Ratios of the PEARLNEQ model predicted with the regression model and those observed in the new sorption
studies
1
0
-1
-2
-2
-1
0
log(predicted PRL-Ratio)
1
Another option would be to measure time-dependent sorption in a small number of studies and use the mean of
the parameters in the leaching modelling. Sorption experiments with five soils were carried out for five pesticides
in this project. There was a large variability between the parameters in the five soils (Figure 19, Table 4). Using
the parameters from a small number of experiments to characterise the time-dependent sorption behaviour of a
compound can potentially introduce a large error in the modelling.
There was considerable variability in time-dependent sorption data collated for the same soil-pesticide
combination. This is partly caused by differences in the experimental methods used (e.g. extraction method and
duration, incubation temperature). It is expected that a standardisation of experimental protocols for timedependent sorption studies will improve the reproducibility of these studies somewhat. This will be investigated
further within DEFRA project PS2235.
The use of default model parameters that give a conservative description of time-dependent sorption could be an
alternative. This approach was tested for two example compounds. The batch Koc value for pesticide 1 is
17 L kg-1 and the first-order DT50 is 77 days. The batch Koc value for pesticide 2 is 82 L kg-1 and the first-order
DT50 is 19 days. Two combinations of possible default values were evaluated. For each substance, four
simulations were carried out with FOCUS PEARL 3_3_3 and 80th percentile concentrations in leachate at 1-m
depth were calculated:
1. Time-dependent sorption was switched off in the first simulation (‘no TDS’). The 24-hour batch Kom and nf
values and the first-order DT50 value of the compounds were used in the modelling.
2. The current FOCUS groundwater scenarios work group recommends a default value of 0.01 day-1 for the
PEARLNEQ desorption rate constant and a value of 0.3 for the ratio between non-equilibrium and equilibrium
sorption. The 24-hour batch values were used to characterise sorption at equilibrium. The DT50 for modelling
had to be adjusted because PEARLNEQ assumes degradation only in the liquid and equilibrium phase, but
not in the non-equilibrium phase. The DT50 must be shorter than the overall half-life to give the same decline
in total residues.
3. The desorption rate constant was fixed at the median of 0.05 day-1 from the analysis of literature data. The
ratio between non-equilibrium and equilibrium sorption was set to the 10th percentile of the literature data
(0.0873). The rate constant and ratio were again combined with the 24-h batch sorption data and an adjusted
DT50.
4. The two default options were compared with the use of sorption parameters that characterise the actual
behaviour of the compounds in the studied soils. These were derived by fitting the ratio and DT50 value to the
measured time-dependent sorption data for the two compounds using PEARLNEQ. The desorption rate was
set to 0.05 day-1 in the fitting and leaching modelling. Sorption at equilibrium was fixed at the 24-hour batch
values.
SID 5 (Rev. 3/06)
Page 16 of 18
Figure 22 shows the results for four groundwater scenarios. As expected, time-dependent sorption reduces
leaching compared with instantaneous equilibrium sorption (No TDS). The effect is greater for pesticide 2 than for
pesticide 1. For both substances, the default parameters recommended by FOCUS result in slightly smaller
concentrations in leachate than the median rate constant and 10 th percentile ratio from the literature review (Study
default). The 80th percentile concentrations for substance 1 is smaller when the fitted parameters are used than
for the study default option. This is because the fitted ratio for this compound (0.229) is larger than the tested
default of 0.0873. The concentrations based on fitted parameters slightly exceeds those for the FOCUS default
option. The fitted ratio for substance 2 (0.594) is large and PEARL simulated much smaller concentrations in
leachate using the fitted parameters than for the tested default options.
0.6
Chateaudun
Kremsmuenster
0.5
Hamburg
Okehampton
0.4
0.3
0.2
0.1
0.0
No TDS
FOCUS
default
Study
default
Pesticide 2 80th percentile (ug/L)
Pesticide 1 80th percentile (ug/L)
Figure 22. 80th percentile concentrations at 1-m depth for pesticide 1 (left) and pesticide 2 (right) simulated with FOCUS
PEARL for various combinations of time-dependent sorption (TDS) parameters.
0.5
Chateaudun
Hamburg
Kremsmuenster
Okehampton
0.4
0.3
0.2
0.1
0.0
Fitted
No TDS
FOCUS
default
Study
default
Fitted
The study default approach uses the 10th percentile ratio from the literature data. The actual ratio is smaller than
this value for 10% of the literature datasets. Concentrations in leachate would thus be larger than simulated with
default option 2 for these soil-pesticide combinations. The 10th percentile was used as an example for a possible
default value, any recommendations on default values for regulatory modelling should be based on the required
level of conservatism. The most conservative approach would be not to include time-dependent sorption in the
modelling. The FOCUS default values are also expected to give a worst-case estimate of time-dependent
sorption in a large number of cases. However, simulations were only carried out for two substances and more
tests are required before general conclusions can be drawn. The use of default values will be investigated further
within DEFRA project PS2235.
3.7
Conclusions and general recommendations
The overall aim of this research was to deliver knowledge on the factors controlling time-dependent sorption of
pesticides in soil and to propose a robust approach to incorporate descriptions of time-dependent sorption into
pesticide fate models. A total of 220 data sets on time-dependent sorption were collected. Residues of the
pesticide remaining in soil at different times from application and concentrations in soil solution were recorded.
Three models were systematically fitted to the data: an empirical model (Walker model), a mechanistic two-site
model (PEARLNEQ) and a diffusion model. Statistical analyses were undertaken to identify any relationships
between the model parameters and pesticide properties, soil properties and experimental factors. Additional
sorption studies were carried out to test the relationships against an independent dataset. Various options to
derive model parameters for new pesticide soil combinations were considered.
PEARLNEQ and the diffusion model gave a similar fit to the data. There was no clear advantage of one model
over the other. PEARLNEQ described the data reasonably well although it did not always match the initial fast
decline in pesticide concentrations in soil solution. PEARLNEQ is used within the pesticide leaching model
PEARL for higher tier regulatory modelling of pesticide leaching through soil. The results from this study suggest
that PEARLNEQ is an adequate model concept.
The parameters of all three models showed a large variability. The experimental method used (e.g. extraction
method and duration, incubation temperature) influenced the results although consistent effects were not always
found. It is expected that a standardisation of experimental protocols for time-dependent sorption studies will
improve the reproducibility of these studies. But some variability between repeated studies with the same soilpesticide combination will remain. This will be investigated further within DEFRA project PS2235.
Time-dependent sorption parameters for a pesticide can differ markedly between different soils even if the same
experimental method is used. This makes it very difficult to derive representative sorption parameters for
regulatory modelling with reasonable experimental effort. Modelling the behaviour of a substance with parameters
averaged over a small number of experimental studies will potentially introduce a large error into the modelling.
Regression analyses identified only weak relationships between the key model parameters and basic soil or
pesticide properties. There was a large scatter of the data around the regression lines for individual and multiple
SID 5 (Rev. 3/06)
Page 17 of 18
variables. The relationships were tested against new data from experiments undertaken within this project. The
analysis showed that any prediction of parameters for new pesticide-soil combinations using the regression
models would carry a high degree of uncertainty.
The use of default values in regulatory modelling of time-dependent sorption appears to be the only alternative
that can be recommended based on current knowledge. The default input parameters for PEARLNEQ
recommended by the FOCUS groundwater group (desorption rate constant = 0.01 day -1, ratio of non-equilibrium
to equilibrium sorption = 0.3) were tested in this project. These were compared with default values derived from
the results of this study (rate constant = 0.05 day -1, ratio = 0.0873 day-1). Both default options are expected to
give a conservative estimate of time-dependent sorption in the majority of cases. There are, however, some
pesticide-soil combinations that show very little or no increase in sorption over time. The default options will
under-estimate leaching for these cases. More work is required before general conclusions on the use of default
values can be drawn. This issue will be investigated further within DEFRA project PS2235.
References to published material
9.
This section should be used to record links (hypertext links where possible) or references to other
published material generated by, or relating to this project.
Beulke S, Brown CD, Fryer CJ, van Beinum W (2004). Influence of kinetic sorption and diffusion on
pesticide movement through aggregated soils. Chemosphere 57:481-490.
Boesten JJTI, Tiktak A, van Leerdam RC (2007) Manual of PEARLNEQ v4. ALTERRA, Wageningen:
WOT Natuur & Milieu (Workdocuments 71) - p. 34.
Cox L, Walker A (1999). Studies of time-dependent sorption of linuron and isoproturon in soils.
Chemosphere, 38, 2707-2718.
FOCUS (2006). Guidance document on estimating persistence and degradation kinetics from
environmental fate studies on pesticides in EU registration, report of the FOCUS work group on
degradation kinetics, EC document reference Sanco/10058/2005 Version 2.0, 434 pp.
Leistra M, van der Linden AMA, Boesten JJTI, Tiktak A, van den Berg F (2000). PEARL model for
pesticide behaviour and emissions in soil-plant systems. Description of processes. Alterra report 13,
RIVM report 711401009.
Meeussen JCL (2003) ORCHESTRA: An object-oriented framework for implementing chemical equilibrium
models. Environmental Science and Technology 37:1175-1182.
Montgomery DC, Peck E A (1982). Introduction to linear regression analysis. New York, J. Wiley & Sons.
Renaud FG, Leeds-Harrison PB, Brown CD, van Beinum W (2004). Determination of time-dependent
partition coefficients for several pesticides using diffusion theory. Chemosphere 57:1525-35.
van Beinum W, Beulke S, Brown CD (2005). Pesticide sorption and diffusion in natural clay loam
aggregates. Journal of Agricultural and Food Chemistry 53:9146-9154.
van Beinum W, Beulke S, Brown CD (2006). Pesticide sorption and desorption by lignin described by an
intra-particle diffusion model. Environmental Science and Technology 40:494-500.
Walker A (1987): Evaluation of a simulation model for prediction of herbicide movement and persistence in
soil. Weed Research, 27, 143-152.
Walker A, Jurado-Exposito M (1998). Adsorption of isoproturon, diuron and metsulfuron-methyl in two
soils at high soil:solution ratios. Weed Research, 38, 229-238.
SID 5 (Rev. 3/06)
Page 18 of 18