49 S. Asseng et al. European Journal of Agronomy 12 2000 37–54
8.4 t ha−1, and grain protein ranged from 11.3 to 13.5
with a median of 12.8 . Soil residual N for
treatment B ranged from 9 to 49 kg N ha−1, with only a small increase in the median, compared to
treatment A, from 10 to 11 kg N ha−1. Model simulations showed [Fig. 14a and b]
that additional N fertiliser applied at the flag leaf stage DC39 treatments C, D, E and F had
minimal effects on grain yields, but increased grain protein to medians of 13.5, 14.0, 14.9 and 15.0
when applying 20, 40, 60 and 80 kg N ha−1, respec- tively. Such an increase in applied N at DC39
hardly affected the median for residual N
13 kg N ha−1 with up to 40 kg N ha−1 fertiliser, but increased the median to 18 and 31 kg N ha−1
with 60 and 80 kg N ha−1 fertiliser applied, respec- tively [Fig. 14c]. However, the upper range in
soil residual N was affected with each additional N application at DC39, reaching 68, 87, 106 and
125 kg N ha−1 with fertiliser applications of 20, 40, 60 and 80 kg N ha−1, respectively.
In a second experiment, grain yield, grain pro- tein and soil residual N were compared for zero
0 and maximum max N applications for the Eest, based on simulations with 14 years of histori-
cal weather data from Swifterbant and for the Bouwing, based on simulations with 31 years of
Fig. 15. Simulated effects of zero 0 and maximum max N-fertiliser application at the Eest and at the Bouwing.
historical weather
data from
Wageningen
Cumulative probability distributions of a grain yield, b grain
Fig. 15. The additional N-fertiliser applications
protein and c soil residual N for the Eest 0 ——— and max
in the maximum N treatment had minor effects on
- - - - - and the Bouwing 0 – – – – and max · · · · · · .
grain yield and protein, but increased the soil residual N substantially. The highest simulated
4. Discussion
yield was 9.9 t ha−1. Without any N-fertiliser application, the yields ranged between 3.8 and
4.1. Performance of the APSIM Nwheat model in 5.7 t ha−1. The Bouwing, with weather data from
the Netherlands Wageningen showed slightly lower yields in the 0
and max N treatments due to higher solar radiation The comprehensive test of model performance
about 0.5 MJ m−2 in average higher in spring and under the temperate maritime conditions of the
summer in Swifterbant compared to Wageningen, Netherlands has confirmed that the APSIM
as a result of less rainfall average year rainfall: Nwheat model can be applied in this climate. An
Swifterbant: 646 mm, Wageningen: 763 mm and acceptable model performance of the APSIM
an extended growing period due to lower mean Nwheat model has also been shown for the sub-
temperatures about 0.5°C in average lower in tropical and arid climates of southern Queensland,
spring and summer in Swifterbant compared to Australia Keating et al., 1995; Probert et al.,
Wageningen. Due
to the
lower yields
in 1995, 1998; Meinke et al., 1998, temperate conti-
Wageningen, grain protein was slightly higher as nental Michigan, USA and temperate maritime
New Zealand climate Meinke et al., 1998, well as residual soil N.
50 S. Asseng et al. European Journal of Agronomy 12 2000 37–54
winter dominant and uniform rainfall regions in in a year with less rainfall, as for instance in 1996,
with only 68 of the average rainfall, an above
eastern Australia Meinke et al., 1998 and the Mediterranean
climate of
Western Australia
average yield was observed and also simulated. In such a dry year, roots grew deeper into the soil
Asseng et al., 1998a,b. However, Asseng et al. 1998b and Meinke
profile with no physical soil constraints, utilising deeper soil layers and partially groundwater for
et al. 1998 have pointed out the poor perfor- mance of grain protein simulations of the APSIM
water uptake. The analysis of other growth condi- tions revealed that the above average yield in 1996
Nwheat model under Australian conditions, in particular with terminal droughts during grain
was due to higher solar radiation during the dry early summer 10
higher radiation during June – filling. In contrast to the Western Australian simu-
lations, predictions of protein content for the July in 1996 than the average for this period at
Wageningen as a result of fewer rain days and Netherlands were good with a RMSD as low as
1.6 , compared to 3.2
in Western Australia below-average temperatures during grainfilling
1.2°C below the Wageningen mean average tem- Asseng et al., 1998b. These better protein simula-
tions in the Netherlands were mainly due to the perature of 17°C for July.
Similar to the performance of the APSIM absence of water limitations. The performance test
for grain protein in the temperate maritime climate Nwheat model in Western Australia, larger devia-
tions occurred in the LAI simulations. These devia- also showed no tendency towards over- or
underprediction as reported for the Mediterranean tions had little effect on the accuracy of biomass
and grainfilling simulations Asseng et al., 1998b, growing conditions Asseng et al., 1998b, suggest-
ing that the basic model routines for grain protein even though LAI is critical for light interception
and photosynthesis in the model at least during after Ritchie et al. 1985 are reasonable, but
requires improvements for dealing with terminal lower LAI Ritchie et al., 1985. Most overestima-
tions occurred in a higher LAI range, where an drought conditions.
A comparison between simulations and long- increase in LAI only marginally affects light inter-
ception when the ground is fully covered with a term experiments or historical yield data often
allows an additional assessment of model–season LAI of about 3 Ritchie et al., 1985. However,
some of the overestimations also occurred at a interactions. However, such comparison is not
always valid, in particular when initial conditions lower LAI, early and at the end of the growing
season during senescence, when light interception had to be estimated. Initial conditions could have
been very different in some years in reality, and is more critical, but showed no major effects on
the performance of other model components. other conditions could have been present which
are not considered by the model, and were not A groundwater table was simulated with the
APSIM Nwheat model, but simulated water table recorded with the data e.g. disease, pest or weeds.
This usually results in higher simulated than depths did not follow measured depths not
shown, because the model is not capable of hand- observed values. Consequently, comparisons of
simulations with such data should be qualitative ling bypass flow through soil cracks or lateral
ground water inflow. Preferential pathways for soil and treated with caution for interpretation of
model performance. water movement have been suggested for similar
soils Lafolie, 1991 and an improved water table In analysing the general seasonal–yield inter-
action and its reproduction with a model, particu- simulation has been shown for some of these soils
by using a soil water model with a bypass routine larly of a rare dry season in the Netherlands, the
historical long-term yield data proved to be useful. Eckersten and Jansson, 1991. However, a simple
water balance test indicated water table fluctua- In all the seasons, the absence of severe water
deficits in the Netherlands simulations was due to tions, which were not accounted for by rainfall
inputs, suggesting additional lateral ground water a combination of usually evenly distributed rain-
fall, soils with high water holding capacities and inflow, which is impossible to simulate with any
of the above models. relative low temperatures during grainfilling. Even
51 S. Asseng et al. European Journal of Agronomy 12 2000 37–54
Some of the smaller discrepancies between simu- three of the 546 measured soil mineral N data
Groot and Verberne, 1991 could not be ade- lated and observed soil mineral N could be due to
bypass flow Rijtema and Kroes, 1991, as sug- quately reproduced by the model. These data were
also impossible to explain through a simple budget gested by small amounts of mineral N that had
built up in deeper layers, even though hardly any test De Willigen, 1991. One of these measured
data occurred on 2 March 1983 at the Eest, where water as the basic medium for N transport moved
through the soil according to the model. However, 52 kg N ha−1 was measured in the 0–30 cm soil
layer, after measurements of 7 kg N ha−1 in this this bypass movement of mineral N had only
marginal effects on the overall accuracy of soil N layer 3 weeks earlier. Since no fertiliser was applied
between 9 February and 2 March 1983, no reason- simulations. This confirms results with other crack-
ing clays without a shallow groundwater table, able explanation exists for this steep increase in
mineral N over a period of 21 days, when the where bypass flow through soil cracks had little
effect on the accuracy of soil N simulations with average mean temperature was −0.2°C maximum
mean temperature of 5.2°C . The model simulated the APSIM Nwheat model Probert et al., 1998.
Others, who had simulated some of the data in this case a rather constant value of about
10 kg N ha−1 for February 1983. At another time, sets from the Eest, PAGV or the Bouwing with
different models, reported problems with simulat- no increase in soil mineral N was measured at 13
or 41 days or at any other following measuring ing the sharp decline in soil mineral N after an
increase from a fertiliser application Eckersten day after applying 80 kg N ha−1 fertiliser on 16
February 1983 at PAGV, while the model showed and Jansson, 1991; Grant, 1991; Groot and De
Willigen, 1991; Kersebaum and Richter, 1991; an increase in soil mineral N in the top soil layer
by the amount of the fertiliser N application, as Lafolie, 1991; Mirschel et al., 1991; Rijtema and
Kroes, 1991; Whitmore et al., 1991 and overesti- expected. Any other fertiliser N applied in
February in any of the experiments by Groot and mations of the relatively low observed mineral N
contents later in the season Bergstrom et al., 1991; Verberne 1991 has been accounted for in soil
measurements shortly after application Groot and Carbon et al., 1991; Eckersten and Jansson, 1991;
Lafolie, 1991; Mirschel et al., 1991. In contrast, Verberne, 1991 and was accordingly simulated
with the APSIM Nwheat model. the APSIM Nwheat model with the SOILN
module by Probert et al. 1998 adequately simu- lated the soil mineral N dynamics. It reproduced
4.2. Simulation experiments the sharp declines in soil N after fertiliser applica-
tions and the low observed soil mineral N contents The simulation experiment with different rates
of N fertiliser showed that N applied at tillering late in the season in each of the simulations for
the Eest, PAGV or the Bouwing. The soil N DC23 and at the beginning of stem elongation
DC31 increased grain yield and grain protein module in APSIM, SOILN, divides the soil organic
matter into two pools BIOM and HUM. BIOM with little effects on residual soil N, except for one
season with little or no rainfall after the fertiliser is a carbon and N pool notionally representing the
more labile, soil microbial biomass and microbial application. N applications at flag leaf stage
DC39 had a small effect on grain yield but products, whilst HUM comprises the rest of the
soil organic matter. The flows between the pools increased grain protein substantially, with a
N-fertiliser application of up to 40 kg N ha−1. A are calculated in terms of carbon as a function of
soil water, temperature and C:N ratio. The corre- similar relative effect has been reported by Ellen
and Spiertz 1980 with an increase of 7 yield
sponding nitrogen flow depends on the C:N ratio of the receiving pools. NO
3 moves according to its
from 5.6 to 6.0 t ha−1 and 10 grain protein
content from 8.2 to 8.9 by applying an addi-
soil water concentration with the flow of water and can be lost from the bottom of the profile in
tional 40 kg N ha−1 at the flag leaf stage and an increase of 11
in yield from 5.6 to 6.3 t ha−1 drainage water Probert et al., 1998.
Despite the general good soil N simulations, and 28
in grain protein content from 8.2 to
52 S. Asseng et al. European Journal of Agronomy 12 2000 37–54
10.4 by applying an additional 80 kg N ha−1.
Netherlands showed slightly better growing condi- tions for wheat near the coastal region at
Note that relative effects are compared here because of the very different N application
Swifterbant than more inland at Wageningen due to slightly more favourable climatic growing condi-
amounts previous to the flag leaf stage N applica- tion 40 kg N ha−1 in early April in the Ellen and
tions higher radiation and lower temperatures, but were otherwise very similar. An average yield
Spiertz 1980
experiment, but
up to
140 kg N ha−1 in February and 90 kg N ha−1 for both locations of 4.3 t ha−1 was simulated
without any fertiliser N application by utilising during tillering and the beginning of stem elonga-
tion in the simulation. Simulated yield and grain mineral N left after the previous crop initial soil
N in the model , N from mineralisation of crop protein contents for the same or similar locations
were in the range of observed data with yields of residues and soil organic material, and about
35–50 kg N ha−1 year−1 continuous wet and dry 6.4–8.2 t ha−1
Spiertz and
Ellen, 1978,
4.0–8.1 t ha−1 Spiertz and Van de Haar, 1978, deposition of N in the Netherlands Neeteson and
Hassink, 1997. 4.2–6.4 t ha−1
Ellen and
Spiertz, 1980,
4.5–8.3 t ha−1 Groot and Verberne, 1991, and The N treatment with up to 140 kg N ha−1
fertiliser in February, a total of 90 kg N ha−1 at 5.6–9.4 t ha−1 Darwinkel, 1998 and grain protein
contents of 9.7–12.5 Spiertz and Ellen, 1978,
tillering and beginning of stem elongation plus 40 kg N ha−1 at flag leaf stage was the most eco-
8.1–9.9 Spiertz and Van de Haar, 1978, 7.4–
10.9 Ellen and Spiertz, 1980, 7.6–15.6
Groot nomic treatment, with large effects on increasing
grain yield and grain protein and low amounts of and Verberne, 1991, 11.2–13.0
Darwinkel, 1998. Fischer et al. 1993 reported a grain pro-
residual N left in the soil after the harvest. The medians for this treatment were 8.5 t ha−1, 14.0
tein of up to 15 for irrigated wheat in Australia
for yields of up to 9 t ha−1. grain protein and 13 kg N ha−1 soil residual N.
The economic and environmental optima in winter The
simulation showed
that up
to 40 kg N ha−1 at DC39 had little effect on residual
wheat seem to coincide Whitmore and Van Noordwijk, 1995, which has also been demon-
N in most of the years, but did increase residual N substantially in a year with a dry period in June,
strated with data from England by Chaney 1990. The optimum N-fertiliser amount and splitting
just after the N application. However, such a season with a dry period around the flag leaf stage
regime agree with recent field experimental-based recommendations
for N-fertiliser
application occurred only once in 14 years. Chaney 1990
found small non-significant increases in soil nitrate for high-yielding winter wheat crops in the
Netherlands Darwinkel, 1998. up to an optimum fertiliser rate for yield, but once
the optimum was reached, further addition of fertiliser increased only soil nitrate after the wheat
harvest significantly. In the simulation experiment,
5. Conclusions
