Miljø- og Fødevareudvalget 2017-18
MOF Alm.del
Offentligt
1838197_0001.png
Det Jordbrugsvidenskabelige Fakultet
Baggrundsnotat til Vandmiljøplan III - midtvejsevaluering
Reestimation and further development in the
model N-LES, N-LES
3
to N-LES
4
Kristian Kristensen
Department of Genetics and Biotechnology, Faculty of Agricultural Sciences, Aarhus
University
Jesper Waagepetersen
Christen Duus Børgesen
Finn Pilgaard Vinther
Department of Agroecology and Environmnet, Faculty of Agricultural Sciences,
Aarhus University
Ruth Grant
Gitte Blicher-Mathiesen
National Environmental Research Institute, Aarhus University
December 2008
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
Index
1. Introduction ..................................................................................................... 3
2. Data................................................................................................................... 3
3. Choice of model type and explanatory parameters ..................................... 4
4. Results: N-LES
4
model.................................................................................... 6
4.1 Parameter estimates.......................................................................................................... 6
4.2 Random effects and coefficient of determination .......................................................... 10
5. Model validation ............................................................................................ 11
6. References ...................................................................................................... 22
Appendix 1 – data used and shown in figure 2............................................... 23
Appendix 2 – data shown in figure 4 ............................................................... 24
Appendix 3 – data shown in figure 5 ............................................................... 24
Appendix 4 – data used and shown in figure 6............................................... 25
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1. Introduction
This model N-LES
4
is the fourth version of an empirical model for prediction of nitrogen
leaching from arable lands. The first version was published by Simmelsgaard et al. (2000) and
the previous version (N-LES
3
) was described by Kristensen et al. (2003). The model predicts
the leaching based on nitrogen applications and crops in the year of leaching, the crops in the
previous year, the average nitrogen applications through the last five years and information on
soil type and drainage during the last two years. The model is developed in cooperation
between The Faculty of Agricultural Sciences (DJF) and National Environmental Research
Institute (NERI), both part of the University of Aarhus – and based on data collected by both
these institutions. This report describes the results of this latest version of N-LES together
with information on how it deviates from previous versions. In addition, the report describes
some of the effects and gives a preliminary evaluation of the model.
2. Data
This model version uses the same data sources as previous versions, i.e. data from the
agricultural catchment monitoring programme – the LOOP programme - (collected by NERI),
data from series of drainage water measurements collected by DJF and data from field
experiments carried out by DJF. However, previous sources have been updated with more
recent data when possible. Data from monitoring catchment 5 (LOOP5) together with one
observation from catchment 1 (LOOP1) and one from catchment 3 (LOOP3) have been
excluded because these sites are atypical for Danish agriculture and are no longer part of the
monitoring programme.
The more recent data show leaching levels and ranges that are generally somewhat lower than
those previously used. Average leaching and standard deviation of leaching used in N-LES
2
were thus, respectively, 74 and 62 kg N ha
-1
, those in N-LES
3
were 64 and 57 kg N ha
-1
,
respectively, while they in the present version were 52 and 45 kg N ha
-1
, respectively. The
range of leaching in the data used in the present version has become smaller because of some
very extreme observations in the excluded data, so the smallest and largest leaching levels
recorded are now 0 and 341 kg N ha
-1
compared to 0 and 446 kg N ha
-1
, respectively, in the
data used for N-LES
3
. In total, 1467 observations were used for estimation of the parameters,
while in the N-LES
1
, N-LES
2
and N-LES
3
models the numbers of observations were 598, 596
and 1299, respectively. The number of observations together with the year span and the level
of leaching for each source of data can be found in Table 5.
Calculation of drainage and thus leaching were done using rainfall corrections and
evaporation with the Makkink formula and crop coefficients (K
c
) that may exceed the value 1
(Plauborg et al., 2002). The monthly drainage was estimated using the model DAISY
3
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
(Abrahamsen and Hansen, 2000) and the yearly leaching, defined as from 1April in the year
of harvesting the summer crop to 31 March the following year, was then calculated.
The amount of nitrogen (N) fixed by a main crop, and any catch crop, has for the NERI data
been calculated using the Danish farm planning program “Bedriftsløsningen” from the Danish
Agricultural Advisory Service (Hvid, 2004), while nitrogen fixation for data from DJF has
been calculated according to Høgh-Jensen et al. (2004). The latter requires information on the
dry matter content of harvested legumes, which is usually included in the experiments at DJF,
but not in the NERI data from the monitoring programme. A comparison of the two methods
(Vinther, unpublished) shows that at typical clover levels (about 20 %), the same fixation
level was obtained with the two methods, while at high or low levels of clover content the
best estimate was achieved using the method of Høgh-Jensen et al. (2004).
3. Choice of model type and explanatory parameters
The basic structure in the model is the same as in previous versions (Simmelsgaard et al.,
2000 and Kristensen et al., 2003). This means that the model comprises both additive and
multiplicative parameters. The model has, however, been modified in a number of areas on
the basis of new knowledge and discussions with users of the model. The most important
modifications are:
•
The effect of crop is now included as an additive effect. The previous crop is included as
an additive effect in order to describe their effect on variation in residual N levels in the
soil. Both the crop and previous crop are now subdivided into a summer crop (main crop)
and a winter crop (sub crop). The winter crop can be the same as the summer crop and/or
the crop of the following year – or it can be an under sown grass for fodder or a catch
crop used to reduce leaching. The grouping of crops has been changed.
In the modelling process the difference between commercial farms and experimental
stations is now included as an additive effect.
The effect of organic matter in the soil is now included both as an additive effect and a
multiplicative effect. The additive effect describes the nitrogen leaching from the organic
matter taking into account the relation between N and C and how this influences the
degradation and thus release of N (Thomsen et al., 2008). The multiplicative effect
describes the extent to which organic matter retains mineral nitrogen and water in the soil
and thus reduces leaching.
•
•
4
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
•
•
The effect of nitrogen removed in harvested crops is no longer included in the model as
the effect was very low. This effect was only included in some of the previous models.
A technology effect, which describes any significant development in leaching over time,
is included. This effect describes the changes over time that are not included in the model,
such as changes to crop rotations, soil tillage, use of pesticides and varieties, etc.
All additive effects that directly or indirectly describe added or removed N are now
summed before they are raised to a power. This makes the effect of a specific source of N
dependent on how much is added/removed by the other sources. For example, the effect
of 1 kg spring-applied plant-available N will be larger if the N level is 200 kg N than if it
is 150 kg N. Similarly, the effect will depend on crop, previous crop, and other nitrogen
sources. If the sum of the additive effects describing added or removed N becomes
negative, it is set at a very small value and some of it is additionally transferred to the part
of the additive effects that are not raised to a power. This is done in order to ensure a
slope on the N response curve also at low N levels.
In this version of N-LES the drainage has been made up on a monthly basis instead of on
a yearly basis as used in previous versions. Effect of yearly leaching is now calculated for
the period from 1 April to 31 March and the drainage for all fields is now calculated using
the model DAISY. In previous versions the model EVACROP (Olsen and Heidmann,
1990) was used for some of the data. The Daisy model yielded lower drainage values,
hence the calculated N leaching values were lower than previously.
An effect of previous year’s drainage is included in order to incorporate the leachable N
that may remain in the soil from previous years.
The effect of both the drainage in the year of leaching and in previous year has now been
subdivided into three periods: A summer period (April-August), an autumn period
(September-December) and a winter period (January-March) in order to take into account
the differences in importance of drainage in the different periods.
The N level is calculated as the average amount of N added in the five years previous to
the actual year of leaching. When information about the five previous years is not
available, average N values from the first five years with recordings have been used.
The newest methods for calculating water drainage and N fixation have been adopted
(Plauborg et al., 2002; Høgh-Jensen et al, 2004).
•
•
•
•
•
•
In connection with the setup of the model, several explanatory variables were investigated but
not included. These were the effects of spring and autumn-applied organic N in animal
manure as separate parameters (they are included in the N level), but the effects were non-
5
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
interpretable and thus not retained in the model. The effect of farm type (crop, cattle and pigs
in combination with organic farming) was investigated as additive effects, but the effects were
relatively small and thus not retained. The effect of temperatures, in three-four separate
periods, was investigated as additive effects, but the effects were not so clear and seemed to
be correlated with each other and other effects already included in the model.
4. Results: N-LES
4
model
For given values of the explanatory variables the prediction model can be written as:
ˆ
ˆ
Y
=
{
U
+
V
κ
ˆ
}
Mc
Where
ˆ
Y
is the predicted leaching
T
is the sum of direct and indirect N effects
U
is an additive effect of a temporal development (technology effect) and a mathematical
expression that ensures a slope to the N response curve also at low N levels
V
the positive values of T that are raised to an exponent
ˆ
c
is a constant to ensure same mean of leaching for predicted and observed values,
ˆ ˆ
ˆ
hvis
T
≥
0
⎧
θ
0
+
θ
1
/(
år
−
θ
2
)
⎪
ˆ ˆ
ˆ
U=
⎨
θ
0
+
θ
1
/(
år
−
θ
2
)
+
ϕ
T
hvis
T
<
0
⎪
⎩
min 0
if
T
>
0
⎧
T
V
=⎨
⎩
0 (0.001) if
T
≤
0
ˆ
ˆ
ˆ
ˆ
ˆ
ˆ
T
=
β
0
+
β
1
N
level
+
β
2
(
N
sping
+
N
fix
)
+
β
3
N
excretion
+
β
4
j
N
autumn
+
β
5
f
c
/
n
C
total
ˆ
M
=
[1
−
exp(
−
δ
ˆ
1
a
A
0
aa
−
δ
ˆ
1
a
A
0
sd
−
δ
ˆ
1
b
A
0
jm
)]exp(
−
δ
ˆ
2
a
A
1
aa
−
δ
ˆ
2
b
A
1
sd
−
δ
2
b
A
1
jm
) exp(
−
δ
ˆ
3
H
) exp(
−
δ
ˆ
4
L
)
A
0
aa
,
A
0
sd
and
A
0
jm
are the drainage in the months April-August, September-December, January-March
in the actual year.
A
1
aa
,
A
1
sd
and
A
1
jm
are the drainage in same months of the previous year
other parameters are defined in tables 1-3.
ˆ
ˆ
ˆ
+
γ
ˆ
summer crop
+
γ
ˆ
winter crop
+
λ
previous summer crop
+
λ
previous winter crop
+
η
experimantal station
4.1 Parameter estimates
Tables 1, 2 and 3 show the parameter estimates for the model together with an approximate
standard error. With the chosen parameter estimates, the level of leaching,
Y,
can be predicted
for given values of the explanatory variables using the equation above.
Table 1 shows the additive effects that comprise the effect of added N,
N level, N-spring, N-
autumn, N-excretion
and
N-C in soil.
6
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0007.png
Table 1. Parameter estimates and standard errors for additive non-classification
variables.
Approximate
Parameter Description
Estimate
StdErr
Power
1.50
0.10
κ
Intercept
175
67
θ
0
Technology effect
2878
1
θ
1
Technology effect
1968
2
θ
2
Intercept in T
31
10
β
0
N-level
0.115
0.026
β
1
N-spring and N-fixation
0.094
0.023
β
2
N-excretion
0.103
0.052
β
3
N-autumn
Sandy soil
0.374
0.176
β
4s
N-autumn
Clay soil
0.167
0.071
β
4l
N-C in soil
0.728
0.160
β
5
Effect of negative T
0.5
-
ϕ
Intercept:
can be interpreted as the leaching of N from a hypothetical cereal crop field
followed by bare soil with a cereal followed by bare soil as previous crop and with a clay and
organic matter content of 0%, where no N has been added and where drainage has been,
respectively, infinitely large and zero in the current and previous leaching years.
Technology effect:
The variable is the specific year when the leaching year starts. Year 2001
is, for example, the leaching year that starts in April 2001 and finishes March 2002. The
technology effect explains the changes (which cannot be explained by the input variables) in
leaching that took place over the years. This could be the effect of new crop varieties, changes
in soil management or crop protection. Changes in temperature and increasing CO
2
levels can,
however, also have an effect. The technology effect is largest in the 1970s and reduces over
time. As the technology effect cannot in any sensible way be extrapolated to coming years, it
is recommended that in the future application of the model the year 2004 is used – the last
year of the data material.
N-level:
Has been calculated as the average addition of N (measured in kg ha
-1
year
-1
) in the
five years of leaching prior to the actual year of leaching (where no information on the five
previous years is available, average N levels from the first five years of data are used). Added
N comprises the sum of total-N in artificial fertilizer, animal manure, deposition from animals
on grass and biological N fixation.
N-spring:
Is the amount of added mineral N in artificial fertilizers and animal manure
(measured in kg ha
-1
year
-1
) in the period between 15
th
February and 1
st
September.
N-fixation:
Is the amount of fixed N (measured in kg ha
-1
year
-1
) by crops grown in the year of
leaching. For fields without legumes a fixation of 2 kg ha
-1
year
-1
is assumed.
7
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0008.png
N-excretion:
Is the total amount of N (measured in kg ha
-1
year
-1
) deposited on the field
during grazing.
N-autumn:
Is the autumn-applied artificial fertilizer and winter-applied (1.9 – 15.2)
ammonium-N (measured in kg ha
-1
year
-1
) in animal manure on either sandy soil (Jb 1-4) or
clay soil (Jb 5-8).
N-C in soil:
Is the effect of the amount of N in soil organic matter (based on tonne C ha
-1
and
the C/N relation in the top 0-25 cm). The parameter
β
5
describes the effect after the total
amount of C has been corrected by a factor that depends on the relation between C and N
(Thomsen et al., 2008). The factor varies between 0.35 and 1.00 for the soils used here and is
given by the expression
f
C
/
N
=
min(56.2
×
CN
−
1.69
,1.0) , where CN is the relation between C
and N in the top 0-25 cm.
Table 2 shows the additive effects that comprise the effect of crop, previous crop and
cultivation on an experimental station when compared to a commercial farm.
Table 2. Parameter estimates and standard errors for additive classification variables.
Parameter Description
Estimate Approximate
StdErr
Summer crop
γ
s1
γ
s2
γ
s3
γ
s4
γ
s5
Grass
a
+ Peas + Cereal/clover
Beets + Potatoes
Cereal + Grass for seed production +
Legume/spring cereal
Rape
Maize
18.6
-29.3
0
23.2
28.4
0
-100.6
-43.6
-11.5
-17.7
5.0
0
0
-51.6
-9.1
-15.9
-24.9
6.2
6.7
-
14.4
15.0
-
16.5
7.8
4.6
4.6
3.2
-
-
18.5
3.2
9.3
6.7
γ
v1
γ
v2
γ
v3
γ
v4
λ
s1
λ
s2
λ
s3
Winter crop
No crop (bare soil)
Grass for seed productions + Grass
a
Undersown grass + Winter rape + Autumn-sown
catch crop
Autumn-sown cereal
Previous summer crop
Grass for seed productions + Beets + Potatoes +
Peas + Maize +Legume/spring cereal
Grass
a
+ Rape + Fallow
Cereal + Cereal/clover
λ
v1
λ
v2
λ
v3
λ
v4
a)
Previous winter crop
No crop (bare soil)
Grass for seed productions
Grass
a
+ Under sown grass + Autumn-sown cereal
Winter rape + Other autumn-sown crop
η
Location of observation
Experimental station
Includes pure grass as well as grass-clover mixtures
8
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0009.png
Summer crop:
The effect of a crop group is estimated as being either larger or smaller than
group 3 (
Cereal + Grass for seed production + Legume/spring cereal
). Group 5 had the largest
indirect N effect of 28.4. Crops in group 5 therefore – all other things being equal – produced
the largest level of leaching. The lowest leaching levels were obtained with crops in group 2.
Winter crop:
The effect of a winter crop group is estimated as being either larger or smaller
than group 1 (
No crop, i.e. bare soil
). All winter crops had a negative value and they thus
produced less leaching than bare soil – all other things being equal. The lowest leaching levels
are achieved when the soil was covered with a grass crop (group 2).
Previous summer crop:
The effect of a previous crop group is estimated as being either larger
or smaller than group 3 (
Cereal + Cereal/clover
).
Previous winter crop:
The effect of a previous winter crop group is estimated as being either
larger or smaller than group 0 (
No crop, i.e. bare soil
). All winter crops reduced leaching in the
following leaching year when compared to bare soil and with all other things being equal.
Experimental station
: The effect of the field being located on an experimental station is
estimated as being either larger or smaller than a field on a commercial farm. The estimate
shows that – all other things being equal – the leaching from a field on an experimental
station was smaller than for a similar field on a commercial farm.
Table 3 shows the parameter estimates of multiplicative variables that comprise the effect of
drainage and soil type and the correction factor.
Table 3. Parameter estimates and standard errors of multiplicative variables.
Approximate
Parameter Description
Estimate
StdErr
Drainage in year of leaching April-
0.000382
0.000112
δ
1a
December
Drainage in year of leaching January-
0.000659
0.000201
δ
1b
March
Drainage in previous year April-
0.000549
0.000390
δ
2a
August
Drainage in previous year
0.000424
0.000118
δ
2b
September-March
Amount of Humus, %
0.1866
0.0237
δ
3
Amount of clay, %
0.0494
0.0064
δ
4
c
Correction factor
1.256
-
Drainage in the leaching year:
The drainage is calculated using the exponential function.
Drainage has a very strong effect on leaching. At very large drainage events (where leaching
is not limited), a multiplication factor of 1 is used. If the drainage is 0 the multiplication factor
9
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0010.png
is 0 and for all other drainages the factor varies between 0 and 1. A given drainage event had
a larger effect in the winter period (January-March) than in the rest of the year.
Drainage in previous leaching year:
The drainage is calculated using the exponential
function. It has a less important effect on leaching than drainage in the leaching year. At low
drainage events (where leaching the previous year was virtually non-existent) the factor is
close to 1 and becomes smaller with increasing levels of drainage in the previous year.
Soil type:
Is characterised by percentage of organic matter and clay in the topsoil. Leaching
reduces with increasing soil organic matter content and clay content. An increase in soil
organic matter content from 2% to 4% thus reduces leaching by approx. 31%. An increase in
soil clay content from 6% to 10% correspondingly reduces leaching by approx. 18%.
Correction for skewness:
In order to obtain the same average for the predicted leaching as for
the measured values, all values are multiplied by the factor 1.256. The correction factor was
calculated after all other effects were estimated.
4.2 Random effects and coefficient of determination
Table 4 shows the contribution to the variability of ln(Y) for each of the random effects, the
coefficient of determination (R
2
) and the standard deviation.
Table 4. Number of observations, variance components, coefficient of determination and
standard deviation on differences between observed and predicted values.
Parameter Description
Estimate
Number of observation
1467
n
a)
2
Location
0.0419
σ
L
2
Year
0.0221
σ
Y
Residual, DJF
0.2593
σ
12
Residual, DMU without grassing animals
0.3965
σ
22
Residual, DMU with grassing animals
0.6350
σ
32
Coefficient of determination
b)
0.526
R
2
c)
Standard deviation
33.3
Std
a)
Field on farmlands or on experimental station
b)
Based on sum of squares for predicted and observed leaching
c)
Based on differences between observed and predicted leaching,
kg N ha
-1
year
-1
Variance:
The residual variance on ln(Y) corresponds to a coefficient of variation on leaching
of approx. 50-70%. If random effects are included, the coefficient of variation is 60-90%.
Residual variance includes several components. The most important are: 1) uncertainty
relating to concentrations in water in suction cups due to, e.g., soil variation, variation in the
application of fertilizer, etc., 2) uncertainty in the explanatory variables such as applied
fertilizer, nitrogen excretion, etc, and 3) the inability of the model to explain all the details.
10
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
The coefficient of variation therefore depends on the observation type, as many relations must
necessarily be more uncertain when the explanatory variables are based on interviews rather
than on experimental data, likewise the uncertainty relating to application of fertilizer will be
larger when some of this originates from grazing animals. The random variation of location
and year describes the additional variation from location to location (field to field or treatment
to treatment) caused by e.g. soil variation not explained by percentage organic matter and clay
and the farmers’ choice of management. In this version, this component of variation was
clearly reduced when compared to the previous versions, which is most probably mainly
caused by the exclusion of some fields (more extreme ones). The random variation of year
describes the additional variation from year to year caused by e.g. climate differences from
year to year.
Coefficient of determination
(R
2
): About 50% of the variation in the observed leaching was
explained by the model. The value is about the same as in the previous model, but somewhat
lower than in N-LES
1
and N-LES
2
, which is due, among other things, to the number of
observations increasing steeply from about 600 to about 1300-1500 and the new observations
being less extreme than earlier data.
Standard deviation
(Std): describes the average deviation between recorded and predicted
leaching. The value of 33.3 corresponds to approx. 60% of the average leaching in the data.
This may seem to be a rather high value. However, even with the best models, the standard
deviation cannot be less than the standard deviation of recorded leaching. It is associated with
a large level of uncertainty and in some experiments with measurements in replicated plots in
the same field it has been found to vary between approx. 1-200% with an average value of
around 40%. On this background the model is judged to give a good description of the
utilized data.
5. Model validation
The model has not been validated using independent data or any kind of cross-validation. The
previous version was validated using cross-validation (Larsen and Kristensen, 2007) and
similar methods may be used to validate this model. The following shows a number of tables
and figures that can be used for a preliminary validation of the model.
Table 5 shows the measured and predicted leaching for each locality and/or experiment. A
good agreement between recorded and modelled leaching can be seen. However, in a few
series predicted leaching deviated considerably from measured leaching. This could be due to
the effect of variables that the model does not take account of. Some experiments may, for
example, have taken place in a period with cold autumns and winters, which is expected to
reduce mineralization in the leaching period.
11
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0012.png
Table 5. Observed and predicted leaching for each locality or experiment. Only fixed
effects have been used to predict the leaching (i.e. without including the random effect of
the individual field or year).
Locality/experiment
Years
No.
Observed
Predicted
obs.
kg N/ha
kg N/ha
Avg.
Min
Max
Avg.
Min
Max
Loop nr 1
Loop nr 2
Loop nr 3
Loop nr 4
Loop nr 6
Drainage water
Supplementary square grid
Increasing N
Ploughing of grass-clover
Catch crop, tillage and N
Long-term catch crop
Organic matter input
Low input rotations
Fodder crop rotation
Organic cereal rotations
Organic fodder crop
rotation
Residual effect of grassland
Drainage water, continued
Slurry and catch crop
Low input fodder crops
All
1991-2004
1991-2004
1991-2004
1991-2004
1991-2004
1973-1996
1989-1993
1974-1990
1990-1994
1988-1991
1994-1996
1990-1992
1990-1992
1990-1992
1998-2004
1995-2001
1998-1999
1998-2004
1988-1989
1998-2000
1972-2004
81
82
56
84
106
94
73
103
30
64
24
44
31
24
294
168
36
33
12
28
1467
31
80
57
39
101
67
65
44
77
52
51
50
46
38
44
29
23
45
80
49
52
0
0
6
0
1
7
1
3
52
4
19
12
7
12
3
0
4
5
15
16
0
109
284
341
127
334
240
192
169
108
160
117
129
121
73
248
144
80
140
232
155
341
39
78
65
53
87
65
52
50
90
53
45
50
56
51
41
27
23
61
70
45
52
0
0
8
0
6
6
17
12
72
20
12
21
30
22
8
4
16
18
29
16
0
109
183
204
123
231
165
145
113
112
95
108
85
131
75
121
71
32
123
111
120
231
Table 6 likewise shows measured and estimated leaching for 20 groups of crop. Large
deviations were found for fields with first year and older grass, with winter rape/winter cereal,
with maize and especially with other crops, but here there were only two observations and
both were atypical agricultural crops (winter rape with under sown grass and summer rape
followed by bare soil). For the remaining crops the averages of observed and predicted
leaching seemed to be reasonably close. The maximum predicted values were in most cases
smaller than the observed maximum and the minimum predicted values were in many cases
larger than the observed minimum. This is expected because the model aims at finding the
best predicted value for a given field.
12
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0013.png
Table 6. Observed and predicted leaching for each combination of summer/winter crop.
Only systematic effects have been used to predict the leaching (i.e. excluding the random
effect of the individual field or year).
Crop
No.
Observed kg N/ha
Predicted kg N/ha
summer/winter
obs.
Avg.
Min
Max
Avg.
Min
Max
Grass for seed production
21
23
1
59
24
10
51
First year grass/grass
96
43
1
248
31
4
113
Older grass/grass
87
52
0
319
40
4
121
Grass/winter cereal
61
75
6
341
75
17
204
Winter cereal/under sown grass
29
33
7
88
31
9
83
Spring cereal/under sown grass
341
37
0
221
36
4
184
Fodder beets/bare soil
68
55
6
184
56
8
208
Sugar beets/bare soil
43
39
2
106
49
1
109
Potatoes/bare soil
3
52
50
55
45
42
51
Cereal/winter cereal
141
52
0
174
55
0
137
Cereal/other crop
36
43
8
127
41
17
79
Cereal/bare soil
378
63
0
240
66
0
172
Spring rape/winter cereal
5
92
14
171
91
37
143
Peas/winter cereal, other crop
43
55
0
201
63
0
182
Winter rape/winter cereal
17
66
3
162
86
3
165
Maize/winter cereal
2
66
25
107
43
37
48
Maize/bare soil
24
114
28
334
120
17
231
Cereal-mixture/under sown grass
34
29
6
185
30
10
53
Cereal-mixture/bare soil
36
60
14
235
57
21
111
a)
Other crops
2
143
109
178
77
74
81
All
1467
52
0
341
52
0
231
a)
Winter rape with under sown grass and spring rape with bare soil.
13
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0014.png
Figure 1 shows the observed leaching plotted against predicted leaching. The plot shows that
the variation between predicted and observed leaching was much smaller when the leaching
was small than when the leaching was large.
400
300
Observed leaching
200
100
0
0
100
200
300
400
Predicted leaching
Figure 1. Plot of observed leaching against predicted leaching.
Figure 2 shows the effect of N for spring cereal followed by bare soil and cereal followed by
winter cereal. The calculation assumed that N level was identical to
N spring + N fixed,
that
there were no other N allocations to the crop and that the previous crop was cereal flowed by
bare soil. The leaching depends on soil type and precipitation and the two figures show two
extremes – a coarse sandy soil in a region with high precipitation and a sandy loam soil in a
region with low precipitation. The figure shows the increasing marginal effect of applied
nitrogen as the level of nitrogen increased and that the leaching – for the same crop –
depended very much on where it was grown. The marginal effect seemed to be lower than
expected from experimental data (Simmelsgaard and Djurhuus, 1998). The average slope was
about 0.30 and 0.15 for spring cereal followed by bare soil in the figure to the left and to the
right, respectively.
14
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0015.png
150
Predicted leaching, kg N/ha
100
50
0
0
50
100
150
200
Applied Nitrogen in spring, kg N/ha
150
Predicted leaching, kg N/ha
100
50
0
0
50
100
150
200
Applied Nitrogen in spring, kg N/ha
Figure 2. Predicted nitrogen leaching for different levels of spring-applied fertilizer for
spring cereal. In each plot with (full line) and without (dashed line) autumn-sown winter
crop. Left: on sandy soil (Jb 1) and high precipitation (Jyndevad). Right: Clay soil (Jb 6)
and low precipitation (Roskilde). Data are shown in appendix 1.
15
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0016.png
Similarly rather low values for responses to nitrogen application were found for other crops.
In order to investigate whether this could be caused by defects in the model, four different
sets of data were selected. In each set the crop, previous crop and soil conditions were made
as homogeneous as possible. For each of the sets the leaching was plotted against the level of
nitrogen application to the soil, a simple regression analysis was performed and the regression
line superimposed on the plot (Figure 3). The figures seemed to show an accordance between
the low predicted leaching for increasing nitrogen and the data from which they were
estimated.
Spring sown cereal after spring sown cereal on Jb 1
Y=28+0.29x
300
Autumn sown cereal after autumn sown cereal on Jb 6, 7 or 8
Y=18+0.13x
150
250
Leaching, kg/year
150
100
Leaching, kg/year
50
100
150
200
250
300
200
100
50
50
0
0
0
50
100
150
200
250
300
350
Average level of nitrogen, kg/year
Average level of nitrogen, kg/year
Grass after grass on Jb 1
350
Y=-60+0.41x
Maize after maize or beets
250
Y=-36+0.19x
300
200
250
Leaching, kg/year
200
Leaching, kg/year
200
250
300
350
400
450
500
550
150
150
100
100
50
50
0
150
0
100
150
200
250
300
350
400
Average level of nitrogen, kg/year
Average level of nitrogen, kg/year
Figure 3. Plot of leaching against average nitrogen level (over five years) for four
different crops with the best linear regression line.
16
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0017.png
It was questioned whether the low response to nitrogen application could be due to a partial
confounding with the technology effect because of the decreasing nitrogen application over
time from about 1975 to 2005. This was examined by estimating the nitrogen response in a
model where the technology effect was excluded. This did not change the response to
increasing nitrogen application (Table 7), although the absolute amount of leaching was
predicted to be higher (in 2005) when the decline over years was not taken into account.
Table 7. Change in nitrogen leaching in models with and without the technology effect
incorporated for selected examples.
Leaching with N application in model
with tech.
without tech.
Summer
Winter
Precipitation
JB-
effect
effect
crop
crop
level
no 100* 200* 300* 100* 200* 300*
Grass
none
Jyndevad
1
123 157 194 133 165 201
mixture
Grass
Grass
Jyndevad
1
27
32
48
29
39
57
mixture
mixture
Winter
none
Jyndevad
1
88
117 151
97
125 157
barley
Winter
none
Roskilde
6
49
64
80
54
68
84
barley
Winter
Winter
Jyndevad
1
73
101 132
81
108 136
barley
barley
Winter
Winter
Roskilde
6
41
55
71
46
60
75
barley
barley
Maize
none
Jyndevad
1
117 151 188 128 161 196
Maize
* kg N/ha
none
Roskilde
6
61
77
94
67
83
99
The technology effect was almost the same in N-LES
4
as in N-LES
3
(Figure 4). The shifting
position of the lines is caused by the smaller response to nitrogen applications in N-LES
4
than
in N-LES
3
.
17
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0018.png
90
90
Predicted leached, kg N/ha
80
Predicted leached, kg N/ha
1980
1990
2000
2010
80
70
70
60
60
50
50
40
40
30
1970
30
1970
1980
1990
2000
2010
Year
90
Year
Predicted leached, kg N/ha
80
70
60
50
40
30
1970
1980
1990
2000
2010
Year
Figure 4. Nitrogen leaching for a cereal crops as a function of year predicted with N-
LES
3
(blue solid line) and N-LES
4
(green dashed line). Top left: Spring cereal with 50 kg
applied N. Top right: Spring cereal with 100 kg applied N. Bottom: Spring cereal with
150 kg applied N. The values are predicted means of all combinations of soil type and
precipitation. Data are shown in appendix 2.
18
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
Comparisons with N-LES
3
The results of the new N-LES
4
were compared with the previous version (N-LES
3
) by
predicting the leaching by both models:
The average yearly leaching was calculated for five LOOP-areas and compared
The leaching in three standard crop rotations were calculated and compared
The LOOP programme is part of the National Monitoring Programme for the Aquatic
Environment initiated in 1990. The LOOP areas consist of five small agricultural catchments
(5-15 km
2
) varying in soil type, rainfall, and livestock density. The monitoring consists of
yearly interviews with farmers regarding farming practices (crops, fertiliser, yields,
cultivation, and livestock) and intensive measurement of soil water and groundwater at 5-8
selected sites of each catchment as well as measurement of stream water.
The comparisons for the LOOP areas showed a clear trend of N-LES
4
predicting less leaching
than N-LES
3
in the years until about 1997, whereas in the following years the predicted
leaching values were almost identical (Figure 5). The discrepancies in the period 1990-97 are
due to high levels of N applications and a lower N response in NLES
4
than in NLES
3
.
Comparisons between two crop rotations, a maize rotation and a grass-clover rotation, are
shown in Figure 6, and for continuous spring barley in Table 8. No catch crops were included
and the spring whole-seed crop was under sown with grass-clover. Liquid cattle manure
equivalent to 1.5 animal units per ha was applied to all rotations and supplemented with
inorganic N according to the Danish N-norms. N removed by crops was calculated from norm
yields and standard N contents. The comparisons were made for a coarse sandy soil (JB1) and
a sandy loam soil (JB6) in combination with high and low precipitation, corresponding to a
runoff at 1 m depth of 549-575 and 251-279 mm, respectively, depending on crops.
Generally, calculations with N-LES
4
resulted in higher leaching than when calculated with N-
LES
3
, and the difference was most pronounced with high precipitation. The largest difference
between N-LES
4
and N_LES
3
, corresponding to 33 kg N/ha, was found for continuous barley
on a sandy loam soil and at high precipitation (Table 8).
19
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0020.png
LOOP1 Storstrøm
1 5
7
LOOP2 Nordjylland
17
5
1 0
5
15
0
1 5
2
12
5
1 0
0
10
0
7
5
7
5
5
0
5
0
2
5
2
5
0
1 9 1 9 1 9 1 9 1 9 2 0 2 0 2 0 2 06
9 0 9 2 9 4 9 6 9 8 0 0 0 2 0 4 0
H st å
ø
r
0
19
90 1 2 19
99
94 1 6 19
99
98 20 0 2 2 20
0
00
04 2
006
H st å
ø
r
LOOP3 Vejle
1
75
LOOP4 Fyn
175
1
50
150
1
25
125
1
00
100
75
75
50
50
25
25
0
19
90 1 2 19
99
94 1
996 19
98 2 0 20
00
02 2 4 20
00
06
H st år
ø
0
1990 1992 1994 1996 1998 2000 2002 2004 2006
H st år
ø
LOOP6 Sønderjylland
1
75
1
50
1
25
1
00
75
50
25
0
1 0 19
99
92 19
94 19
96 19
98 20
00 2
002 2
004 2 6
00
H st å
ø
r
Figure 5. Predicted yearly leaching for each of five loops. Blue squares and lines are for
N-LES
3
while red dots and lines are for N-LES
4
. Data are shown in appendix 3.
20
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0021.png
180
150
120
90
60
30
0
150
120
90
60
30
0
150
120
90
60
30
0
150
120
90
60
30
0
150
120
90
60
30
0
120
90
60
30
0
120
90
60
30
0
120
90
60
30
Spring barley
Spring barley
Spring barley
Grass-clover
Grass-clover
Maize
Maize
Average
Whole-seed
Average
0
Precipitation: Low
Soiltype: Sandy loam (JB6)
Precipitation: Low
Soil type: Coarse sand (JB1)
Precipitation: High
Soiltype: Sandy loam (JB6)
Precipitation: High
Soil type: Coarse sand (JB1)
Figure 6. Leaching (kg N/ha) from a maize and a grass-clover rotation at four
combinations of precipitation and soil type. Results of N-LES
4
calculations are shown
with grey bars and N-LES
3
with white bars. Data are shown in appendix 4.
Table 8. Leaching (kg N/ha) calculated with N-LES
4
and N-LES
3
from continuous
barley at four combinations of precipitation and soil type.
High precipitation
Low precipitation
Coarse sand
Sandy loam
Coarse sand
Sandy loam
N-LES
4
113
101
70
59
N-LES
3
86
68
61
45
21
Leaching (kg N/ha)
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
6. References
Abrahamsen, P. and Hansen S (2000). Daisy: an open soil-crop-atmosphere system model.
Environmental Modelling and Software. 15, 313-330
.
Høgh-Jensen, H., Loges, R., Jørgensen, F. V., Jensen, E. S. & Vinther, F. P. (2004) An
empirical model for quantification of symbiotic nitrogen fixation in grass-clover mixtures.
Agricultural Systems 82, 181-194.
Hvid, S. K. (2004) Beregning af kvælstoffiksering, Planteavlsorientering Nr. 07-497,
Landbrugets Rådgivningscenter.
Kristensen, K., Jørgensen,U.,Grant, R.. 2003. Genberegning af modellen N-LES
.
12 pp.
Baggrundsnotat til Grant, R. og Waagepetersen, J. 2003. Vandmiljøplan II - slutevaluering.
Danmarks Miljøundersøgelser, Miljøministeriet. ISBN:87-7772-776-2”. Available at
http://nywww.agrsci.dk/var/agrsci/storage/original/application/phpE1.tmp.pdf and at
http://www.dmu.dk/1_viden/2_publikationer/3_ovrige/rapporter/VMPII/Genberegning_af_m
odellen_NLES.pdf. An English translation with the title “Recalculation of the model N-LES
.
”
is available at: http://130.226.173.223/farmn/dokumentation/Nles3%20english.doc.
Larsen, S., Kristensen, K. 2007. Udvaskningsmodellen N-LES
3
– usikkerhed og validering.
DJF rapport, Markbrug. 132.
Plauborg, F., Refsgaard, J.C., Henriksen, H.J., Blicher-Mathiesen, G., Kern-Hansen, C. 2002.
Vandbalance på mark- og oplandsskala. DJF rapport, Markbrug. 70, 45 pp.
Simmelsgaard, S. E. and Djurhuus, J. 1998. An empirical model for estimating nitrate
leaching as affected by crop type and long-term N fertilizer rate. Soil Use and Management,
14
,
30-36
.
Simmelsgaard, S.E., Kristensen, K., Andersen, H.E., Grant, R., Jørgensen, J.O., Østergaard,
H.S. 2000. Empirisk model til beregning af kvælstofudvaskning fra rodzonen. N-LES.
Nitrate Leaching EStimator. DJF rapport, Markbrug. 32, 67 pp.
Thomsen, I. K., Petersen, B. M., Bruun, S., Jensen, L. S. and Christensen, B. T. 2008.
Estimating soil C loss potentials from C to N ratio. Soil Biol. Biochem. 40
,
849-852.
22
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0023.png
Appendix 1 – data used and shown in figure 2
Applied values of variables that do not depend on amount of applied N
All data
Depending on soil and
Coarse
Sandy loam
precipitation
sandy soil,
soil, low
high
precipitation
Year of
precipitation
harvest
2005
N-level
See below
C in soil
65 t/ha
55 t/ha
N-spring
See below
C/N factor
0.56
0.98
N-fixation
2 kg/ha
Drainage Apr-Dec
315 mm
109 mm
N-excretion
0 kg/ha
Drainage Apr-Dec
245 mm
138 mm
N-autumn
0 kg/ha
Drainage prev. year Apr-
54 mm
34 mm
Aug
Crop
Spring cereal Drainage prev. year Sep-
517 mm
217 mm
Mar
Previous crop
Spring cereal Amount of humus
3.2 %
2.5 %
Location of
Commercial Amount of clay
4.7 %
12.7 %
obs.
Applied N-level, N spring application and predicted N leaching
Soil type and
Winter crop
N-level and N spring application
precipitation
0
50
100
150
200
Coarse sandy soil, high
-
74
87
101
116
133
precipitation
+
32
40
49
59
71
Sandy loam soil, low
-
40
46
53
61
68
precipitation
+
18
23
27
33
39
23
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0024.png
Appendix 2 – data shown in figure 4
Predicted leaching as function of year and applied N for spring cereal with N-LES
3
and
N-LES
4
Year
50 kg N/ha
100 kg N/ha
150 kg N/ha
N-
N-
N-
N-
N-
N-
LES
3
LES
4
LES
3
LES
4
LES
3
LES
4
1975
54
60
67
67
81
75
1980
45
48
57
55
72
63
1985
39
43
52
50
67
58
1990
36
40
49
47
63
56
1995
34
38
46
46
61
54
2000
32
37
45
45
59
53
2005
30
36
43
44
58
52
Appendix 3 – data shown in figure 5
Predicted by N-LES
3
Predicted by N-LES
4
Year of
harvest
LOOP1 LOOP4 LOOP3 LOOP2 LOOP6 LOOP1 LOOP4 LOOP3 LOOP2
1991
54
87
88
145
163
46
71
81
113
1992
54
77
86
141
148
45
64
82
103
1993
47
74
84
128
152
43
61
79
89
1994
46
71
75
111
149
43
62
76
88
1995
42
73
84
112
125
33
62
78
85
1996
39
63
78
104
113
36
57
74
80
1997
39
64
72
94
110
34
55
69
78
1998
39
58
82
90
112
38
60
78
88
1999
32
50
75
75
90
30
51
71
76
2000
40
49
67
80
85
34
48
70
80
2001
36
51
66
73
90
34
51
68
77
2002
37
49
63
71
84
35
47
64
75
2003
37
42
54
70
79
36
48
59
74
2004
35
46
55
74
76
33
47
58
79
2005
30
45
63
68
80
31
44
68
80
2006
35
47
61
72
82
30
42
59
71
LOOP6
142
134
134
138
113
115
113
112
92
88
92
89
85
84
89
95
24
 MOF, Alm.del - 2017-18 - Endeligt svar på spørgsmål 246: Spm. om oversendelse af et notat med en beskrivelse af processen bag udarbejdelsen af NLES 4 modellen, til miljø- og fødevareministeren
1838197_0025.png
Appendix 4 – data used and shown in figure 6
High precipitation
Coarse sandy soil (JB1)
Sandy loam soil (JB6)
Crop rotation
no. and crops
N-level
for crop
rotation
a
271
271
271
271
177
177
177
177
166
166
166
166
N leaching
N leaching
N-
N-
N-level
spring
b
predicted with:
spring
b
predicted with:
for crop
+N
+N
rotation
a
fixation N-LES
3
N-LES
4
fixation N-LES
3
N-LES
4
43
188
290
312
78
78
159
141
110
101
101
101
115
80
40
46
86
111
171
137
114
112
112
112
137
74
19
22
97
97
133
128
88
86
86
86
275
275
275
275
174
174
174
174
169
169
169
169
45
182
300
322
81
81
148
133
115
103
103
103
103
73
40
46
78
98
145
118
102
100
100
100
108
54
16
19
74
75
105
101
70
67
67
67
1. Barley
1. Whole-seed
1. Grass-clover
1. Grass-clover
2. Barley
2. Barley
2. Maize
2. Maize
3. Barley
3. Barley
3. Barley
3. Barley
a
N level corresponds to average total N applied to the crop rotation.
b
N-spring corresponds to the amount of inorganic N in manure plus added fertilizer N.
Low precipitation
Coarse sandy soil (JB1)
Sandy loam soil (JB6)
Crop rotation
no. and crops
N-level
for crop
rotation
a
271
271
271
271
177
177
177
177
166
166
166
166
N leaching
N leaching
N-
N
N-level
spring
b
predicted with:
spring
b
predicted with:
for crop
+N
+N
rotation
a
fixation N-LES
3
N-LES
4
fixation N-LES
3
N-LES
4
43
188
290
312
78
78
159
141
110
101
101
101
72
50
25
29
54
69
102
83
71
70
70
70
96
51
13
16
68
68
90
87
62
60
60
60
275
275
275
275
174
174
174
174
169
169
169
169
45
182
300
322
81
81
148
133
115
103
103
103
60
43
23
28
46
58
85
70
60
59
59
59
72
36
10
13
49
49
68
66
46
44
44
44
1. Barley
1. Whole-seed
1. Grass-clover
1. Grass-clover
2. Barley
2. Barley
2. Maize
2. Maize
3. Barley
3. Barley
3. Barley
3. Barley
a, b
N Notes a and b: see table above
25