NGAUGE: A decision support system to optimise N fertilisation
of British grassland for economic and environmental goals
L. Brown*, D. Scholefield, E.C. Jewkes, D.R. Lockyer, A. del Prado
Institute of Grassland and Environmental Research, Soil Science and Environmental Quality,
North Wyke, Okehampton, Devon EX20 2SB, UK
Received 23 June 2004; received in revised form 9 February 2005; accepted 15 February 2005
The poor efficiency with which nitrogen (N) is often used on grassland farms is well documented, as are the potential
consequences of undesirable emissions of nitrogen. As fertiliser represents a major input of nitrogen to such systems, its
improved management has good potential for increasing the efficiency of nitrogen use and enhancing environmental and
economic performance. This paper describes the development, structure and potential application of a new decision support
system for fertiliser management for British grassland. The underlying empirically-based model simulates monthly nitrogen
flows within and between the main components of the livestock production system according to user inputs describing site
conditions and farm management characteristics. The user-friendly decision support system (‘NGAUGE’) has a user interface
that was produced in collaboration with livestock farmers to ensure availability of all required inputs. NGAUGE is an
improvement on existing nitrogen fertiliser recommendation systems in that it relates production to environmental impact and is
therefore potentially valuable to policy makers and researchers for identifying pollution mitigation strategies and blueprints for
novel, more sustainable systems of livestock production. One possible application is the simulation of the phenomenon of
pollution swapping, whereby, for example, the adoption of strategies for the reduction of nitrate leaching may exacerbate
emissions of ammonia and nitrous oxide. Outputs of the decision support system include a field- and target-specific N fertiliser
recommendation together with farm- and field-based N budgets, comprising amounts of N in both production and loss
components of the system. Recommendations may be updated on a monthly basis to take account of deviations of weather
conditions from the 30-year mean. The optimisation procedure within NGAUGE enables user-specified targets of herbage
production, N loss or fertiliser use to be achieved while maximising efficiency of N use. Examples of model output for a typical
grassland management scenario demonstrate the effect on model predictions of site and management properties such as soil
with NGAUGE suggest that it is possible to reduce nitrate leaching by up to 46% (compared with a fertiliser distribution from
existing fertiliser recommendations), and fertiliser by 33%, without sacrificing herbage yield. The greatest improvements in
efficiency are possible on sandy-textured soils, with moderate N inputs.
# 2005 Elsevier B.V. All rights reserved.
Keywords: Nitrogen; Decision support system; Fertiliser recommendations; Grassland; Model
Agriculture, Ecosystems and Environment 109 (2005) 20–39
* Corresponding author. Tel.: +44 1837 883509; fax: +44 1837 82139.
E-mail address: firstname.lastname@example.org (L. Brown).
0167-8809/$ – see front matter # 2005 Elsevier B.V. All rights reserved.
decades, of nitrogen (N) emissions from grassland
(Ryden, 1981, 1984; Scholefield et al., 1993) has
demonstrated the inefficiency with which N is
frequently used. The loss of N has both economic
and environmental consequences. The N loss path-
and emission of the gases nitrous oxide (N2O) and
ammonia (NH3). The increase in nitrate concentration
in water bodies in recent decades has been a cause of
concern because of the perceived potential threat to
human health and because of the ecological and
aesthetic consequences of eutrophication. In the UK,
agriculture is the main source of nitrate in most UK
rivers and groundwaters (Powlson, 2000) and is
estimated to account for 69% of the emission of N2O
(Salway et al., 2001), which contributes both to global
warming and to the depletion of the stratospheric
ozone layer. Ammonia emission and subsequent
deposition may contribute to water and soil acidifica-
tion (Van Breemen et al., 1982) and is one of the main
sources of the increased N supply to natural areas that
may cause eutrophication of terrestrial and aquatic
ecosystems (Isermann, 1990).
It has been shown (Scholefield et al., 1991) that
there is a strong linear relationship between total
annual inorganic N input to a grassland system and
percentage recovery of that N by plants, such that in
systems of low N flux, a larger proportion of the total
N isrecovered by the plantthan in systems of higher N
flux. Agricultural systems can be manipulated to
changed efficiency simply by increasing or decreasing
N input. Additionally, efficiency of plant uptake of N
changes seasonally with weather and soil conditions
and with physiological traits of the plant. Nitrogen
fertilisers are the major N input to a typical dairy farm
in the UK, comprising as much as 74% of the total N
input (Jarvis, 1993), and are the input to the grassland
N cycle that is most easily managed. It appears that
there is much potential, therefore, to manipulate the
efficiency of the system by appropriate management
of fertilisers. However, simply reducing the fertiliser
N input moves the system along the established
efficiency relationship, and although losses can be
reduced, production is also compromised. The
challenge lies in the development and implementation
of a system which lies above this line, i.e. is genuinely
of greater N efficiency for the same total flux of N.
This will involve both temporal and quantitative
adjustment to fertiliser patterns.
Fertiliser recommendations for N have been
produced in a similar format for England and Wales
since 1973. With the exception of the most recent
edition, recommendations have given little or no
consideration to the potential environmental impacts
of N application and have been rather generalised in
relation to site variables. In the current version
(RB209, MAFF, 2000), there is more site-specificity,
in terms of soil types (three classes), rainfall (three
classes) and previous management and N use.
Although the publication points out the importance
of achieving the right balance between profitable
agricultural production and environmental protection,
it also states that ‘the primary aim of the recommen-
dations is to maximise the economic return from the
use of fertilisers’. Improvement of the current UK
recommendation system to effect improvements in
efficiency would necessitate a change in emphasis
from production/economic targets to a system driven,
to a greater degree, by limitation of the undesirable
exports: nitrate lost to surfacewater and N2O and NH3
emitted to the atmosphere. The application of such an
approach would be especially beneficial in areas of
particular sensitivity such as Nitrate Vulnerable Zones
(NVZs, implemented under the Nitrates Directive, 91/
676/EC), where the nitrate concentration of water
in the selection of agricultural management. The
improved recommendation would seek to strike a
compromise between production and environmental
impact since the farmer still needs to achieve an
acceptable level of income.
The objective of the research presented in this
paper was to produce a decision support system (DSS)
which would enable the efficiency of N use in
grassland fields to be improved, by calculating the
field. In order to achieve this aim, the NGAUGE DSS
was developed, to provide field-specific monthly N
fertiliser recommendations, which improve the effi-
ciencywith which N is used, for user-specified targets.
This necessitated simulating flows of N on a site-
specific basis, with sensitivity to climate, soil
properties, sward management and on-going weather,
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3921
and the development of the means of determining the
best distribution of fertiliser N through the year to
improve the efficiency of N use.
2. Model development
An existing empirically-based model of N cycling
in grassland soils, NCYCLE (Scholefield et al., 1991),
was taken as the basis for the new model and DSS.
NCYCLE is an annual, empirical model, based on
published multi-site grassland data sets and has, since
its creation, been validated for many of its key
components (Rodda et al., 1995). NCYCLE simulates
N flows through the major processes of N transforma-
tion in the soil and therefore links the input,
production and loss components of the system.
Sensitivity to soil properties, sward management
and weather already exist within NCYCLE, although
the latter is not sufficiently detailed for the purposes of
the DSS development. NCYCLE is an annual model
and therefore does not have the appropriate temporal
resolution for prescribing fertiliser recommendations.
The sub-models within NGAUGE therefore calculate
N cycling through N components and processes on a
monthly basis. In addition, there are five main areas in
which NGAUGE extends the capabilities of the
original NCYCLE model:
(1) Inclusion of an optimisation procedure to identify
a fertiliser amount and distribution according to
criteria of herbage production and N losses to the
(2) Increased detail of average weather, and sensi-
tivity to within-year on-going weather.
(3) Simulation of losses of NH3from, and miner-
alisation of applied organic manures, and con-
sideration of the magnitude and timing of this
source of N in calculating fertiliser recommenda-
(4) Provision of farm-gate N budgets (excluding
import and export of animals).
(5) More detailed simulation of nitrification and
denitrification to enable prediction of N2O
emissions separately from dinitrogen (N2) and
nitric oxide (NO).
2.1. Model components
2.1.1. Plant uptake
At different times in the growing season, soil
inorganic N is recovered by harvested herbage with
contrasting efficiency. This was demonstrated in
experiments such as those of Morrison et al. (1980)
and Hopkins et al. (1990), in which equal amounts of
fertiliser were applied in each time period (i.e.
‘month’), giving a range of annual amounts of N
of,e.g.0–750 kg N ha?1year?1.Plotswerecutona4-
weekly basis and N in herbage determined. These
multi-site trials provide a source of information on N
recovery at different rates of fertiliser N and at sites
with different soil types and land-use histories. Data
from these experiments were used to derive a set of
curves (Fig. 1), which describe the relationship
between inorganic N flux (the sum of all the inputs
to the soil) and plant N flux (including N in roots) for
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–39 22
Fig. 1. Monthly N uptake (‘h’) factors used in the optimisation process.
each month. Inorganic N includes N in fertiliser, N
mineralised from soil organic matter and manures, N
in urine (if grazed), N input from the atmosphere and
‘carried-over’ leachable N that was not leached in the
In order to calculate total N in plant from published
data on N in herbage, assumptions were made about
the u factor (defined in Scholefield et al. (1991), as the
proportion of N in the whole plant that is harvested by
the animal or by cutting). There are few data to act as a
guide to what the value of this factor may be. Parsons
et al. (1983) report a value of 0.63 from measurements
of carbon on continuously grazed pastures in south-
west England, which is similar to the value of 0.62
assumed for the NCYCLE model (Scholefield et al.,
1991). Hansson and Pettersson (1989), on perennial
grassland leys, report values of 0.71 and 0.77, and
Ourry et al. (1988) report values between 0.45 and
0.49. In order for the criterion of annual mass balance
to be satisfied, internal consistency between miner-
alisation, plant N uptake and losses must be observed.
By assuming a monthly distribution for mineralisation
(see Section 2.1.3), the value of u and plant N in each
month can be fixed from the herbage data. Existing
datasets and systems simulations of NCYCLE were
used to quantify u at a range of N input values. These
data were then used to provide relationships between
herbage N and u for each month. The values of u lie,
for example, in a 300 kg N ha?1system between 0.2
(December) and 0.67 (June), with smaller values in
winter months, when growth of the lamina region of
the grass plant is limited, and large values in May and
June, a time in which partitioning of N to the lamina
and reproductive regions would be expected. From an
N balance perspective, the lag between peak in
herbage production (usually observed in May,
particularly for cut systems) and mineralisation
(frequently peaking in July or later) requires that a
largeproportionof the N takenupby the plant must be
recovered in harvested material in the early summer
The N uptake curves define the efficiency of the
plant in recovering N in each month and are analogous
to the annual ‘h’ factor relationship used in NCYCLE.
The comparison of these relationships between
months is fundamental to the prediction of losses at
each level of N input, and therefore to the operation of
The concentration of N in cut herbage was
calculated using relationships between fertiliser N
and %N in herbage, derived from Morrison et al.
(1980). In the absence of better data for grazed
modified to produce a monthly relationship.
2.1.2. N cycling through the grazing animal
The amount of herbage N ingested by the animal is
determined by the u factor for each month, as
discussed earlier. The relationships used to calculate
and product were taken directly from Scholefield et al.
(1991), and were based on empirical relationships
describing the influence of herbage %N on N
partitioning. As with NCYCLE, it was considered
that most of the urine N is mineralised within a few
days, and therefore enters the inorganic N pool within
the month of excretion. Twenty-two percent of the N
in dung mineralises in each month (i.e. passes into the
soil inorganic N pool), as discussed in Scholefield
et al. (1991).
Mineralised N was considered to be derived from
four components: (i) the previous land use; (ii) the
herbage production in the current year; (iii) dung; and
(iv) applied manures (Section 2.1.9). As in NCYCLE,
the previous land use was categorised as long-term
grassland, mixed ley and arable or long-term arable.
Annual starting values for each of these were
determined from the zero N fertiliser plots of cut-
plot experiments of Morrison et al. (1980) and
Hopkins et al. (1990), assuming no N loss and that,
on an annual basis, 0.62 of the N in the whole plant is
harvested by cutting (as NCYCLE). The starting
values were 134, 76 and 27 kg N ha?1for long-term
grassland, mixed ley–arable and long-term arable,
respectively. These were then moderated by factors
describing the effect of sward age, soil texture and
drainage status. This annual total mineralised N was
allocated to different months according to relation-
ships describing the effect of soil moisture and
temperature on mineralisation. For the former, it was
between the soil’s permanent wilting point and field
capacity. This is supported by the work of Stanford
and Epstein (1974), who found the highest N
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3923
mineralisation between matric suctions of 0.3 and
0.1 bar (equivalent to 80–90% water-filled pore
space), and that between this optimum range and
15 bar, there was a near linear relationship between
mineralisation and soil water content. Reichman et al.
(1966) report that ammonification and nitrification
were almost directly proportional to soil moisture
content at suctions of 0.2–15 bar. For the effect of
temperature on mineralisation, a factor was calculated
based on a linear increase of mineralisation rate with
temperature, between a minimum (0) at 2 8C and a
maximum (1)at20 8C. Macduff and White (1985) and
Blantern (1991) support a linear relationship between
mineralisation and temperature between 2 and 20, and
4 and 13 8C, respectively. For each month, soil
moisture content was calculated from soil moisture
deficit (30-year average data for each zone and each
month; see Section 2.1.10), using algorithms supplied
by the National Soil Resources Institute, which
assume for each of the five textural classes an
effective depth of operation of the deficit, moisture
content at field capacity and permanent wilting point,
porosity and bulk density (C. Brown, personal
communication). Mineralisation from the current
year’s residues was calculated using empirically-
derived functions, which relate monthly plant N to
observed or estimated mineralisation. These values
were then modified by factors which account for the
effect of soil texture, drainage status, sward age and
weather zone (using relationships with temperature
and moisture as described earlier).
Denitrification was modelled as a function of soil
inorganic N, water-filled pore space (WFPS) and
temperature. Water-filled pore space was related to
monthly denitrification using a relationship derived
from the controlled laboratory experiments of Schole-
field et al. (1997). Denitrification rate was assumed to
increase linearly with temperature from 2 to 20 8C.
The relationship between temperature and denitrifica-
tion rate has been found to be linear by Cho et al.
between 7 and 16 8C, although at higher temperatures
(e.g. 15–35 8C), a Q10of 2 (i.e. a doubling in reaction
rate for an increased temperature of 10 8C) has been
reported (Stanford et al., 1975). A rapid decrease in
denitrification below 5 8C has been observed (Bailey
and Beauchamp, 1973), but minimum temperatures
for denitrification may vary widely (Aulakh et al.,
1992). Initially, the annual denitrification totals of
NCYCLE were used together with weighting factors
for soil texture, drainage status, temperature zone and
rainfall zone to predict denitrification in each month.
In order that denitrification could be calculated
dynamically during the optimisation process (i.e.
without recourse to annual totals), relationships were
derived from these meta-data to predict denitrification
from inorganic N in each month, retaining sensitivity
to climate using the temperature and WFPS weighting
factors described earlier.
In a monthly time-step model, it is not possible to
account for the effects of individual rainfall events,
although it is widely recognised (Jarvis et al., 1991; Li
et al., 1992) that the occurrence of rain events, and
time since a rainfall event, may be major determining
factors of denitrification rate, and that good relation-
ships between denitrification rates and controlling
variables may be obscured by the considerable
temporal variation that occurs with denitrification.
2.1.5. N-oxides sub-models
126.96.36.199. N2and N2O from denitrification. This sub-
model was conceptually based on the ‘hole-in-the-
pipe’ model described by Firestone and Davidson
(1989). This scheme postulates two levels of regula-
tionfortrace N-gas production: factorsthatcontrolthe
rate of the overall process dictate the movement of N
through the ‘process pipe’ (denitrification and
nitrification processes); and factors that control the
partitioning of the reacting N species to NO, N2O or
N2(i.e. control the size of the holes in the pipe through
which the different N-gases ‘leak’).
In NGAUGE, N2O and N2were assumed to be the
only gaseous products of the denitrification process.
Although NO has been proved to be produced during
the microbial process of nitrification and denitrifica-
tion (Firestone and Davidson, 1989), many studies
have indicated that the NO gas does not constitute a
major denitrification product (e.g. Anderson and
Levine, 1986; Skiba et al., 1992; Neff et al., 1995;
Parsons and Keller, 1995).
In order to predict N2and N2O, the monthly values
for denitrification were divided according to three
factors: soil moisture content (WFPS), mineral N flux
and mineralised N in the soil, using three functions to
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–39 24
represent the effect of these factors on the N2:N2O
ratio as proposed by Parton et al. (1996). The level of
nitrate was expressed as mineral N in order to be
compatible with the main model.
Thus, the N2:N2O ratio was calculated as follows:
N2: N2O ¼minðFrðMin NÞ;FrðMineralisÞÞ
where Fr(WFPS) is the effect of soil WFPS on the
ratio, Fr(Mineralis) is the effect of mineralisation rate
on the ratio and Fr(Min N) is the effect of mineral N
level in the soil on the ratio.
188.8.131.52. N2O and NO from nitrification. The monthly
nitrification rate within NGAUGE was developed on
the basis that the main substrates to be nitrified would
be originated from the pools of ammonium (NH4+)
mineralised from the organic matter (including
excreta) and NH4+from the mineral fertiliser.
The zero-order kinetics approach described by
Gilmour (1984) was implemented into the model with
the nitrification rate constant being a function of
temperature and soil moisture. The functions were as
NIT rate ¼ K½NHþ
K : month?1
NIT rate ¼
(kg N ha?1month?1), [NH4+]iis the level of NH4+
in the soil at the beginning of the month and K/Kmax
and KmaxWare the soil temperature and moisture
content factors, respectively, which affected the nitri-
fication rate. The effect of temperature was modelled
according to the Arrhenius equation.
KmaxW was derived from Macduff and White
(1985), who used three different functions for soils
under permanent wilting point, between permanent
wiltingpointandfield capacity andoverfieldcapacity.
From the predicted net nitrification pool, NO
emissions were simulated on a monthly basis,
following the approach of Davidson et al. (1993), in
which NO fluxes are governed by the total amount of
NH4+-N nitrified (nitrification), a factor describing the
potential maximum percentage nitrified as NO
(Max%NIT) and a modifier accounting for the soil
NIT rateis nitrificationrate
moisture (WFPSf). The functions were as follows:
WFPSf¼ 0:0181 ? WFPSþ0:0165
WFPSf¼ ?0:0667 ? WFPSþ4:6667
NOðg N-NO ha?1month?1Þ
¼ Max%NIT ? WFPSf? NIT rate
Nitrous oxide emissions from nitrification were cal-
culated in the model following the approach of Mosier
to predict daily N2O loss from soils from nitrification
and denitrification. According to this study, the total
amount of N2O emitted from the nitrification process
(N2Onit) is governed by the maximum potential rate of
N2O from nitrification, assumed in NGAUGE to be
110 g N ha?1day?1, based on maximum recorded
field values (Yamulki, personal communication), a
normalised (0–1) factor accounting for the amount
of NH4+nitrified (En) and a soil moisture normalised
(0–1) modifier (Ec) as follows:
N2Onitðg N ha?1day?1Þ ¼ 110 ? Ec ? En
Ec ¼ 0:1
ElseEc ¼ ½3
1 þ 1:335 ? e?1:24? NIT rate
where RWC is the soil relativewater content, which is
equal to the difference between measured soil water
content and soil water content at wilting point divided
by the difference between soil water content at field
capacity and the soil water content at wilting point.
if RWCðwater contentÞ ¼ ½0?4?
2RWC ? 5? ? 0:1
2.1.6. Nitrate leaching
Leachable nitrate, peak and average nitrate-N
concentrations are presented by the model on an
annual basis. For each month, soil inorganic N flux
was calculated as the sum of atmospheric input,
dung and manures, fertiliser, urine and ‘leachable N’
carried over from the previous month. From this total,
uptake of N by the plant, NH3volatilisation and N lost
by denitrification were subtracted. The fate of the
remaining ‘leachable N’ depends on the month in
question; for January, February and December, it was
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3925
assumed that ‘leachable N’ contributes to the total
annual leaching, and for other months it was passed to
the succeeding month as a component of the inorganic
that which would be measured in the field as soil
mineral N at the end of each month.
Peak and average concentrations of nitrate-N in
leachate were calculated on an annual basis using the
relationships derived by Rodda et al. (1995), which
predict peak concentration from leached N, according
to soil drainage and textural class (Fig. 2), and average
concentration from leached N and drainage class.
2.1.7. Ammonia volatilisation
An NH3emission factor of 1.6% was suggested for
ammonium nitrate (Van der Weerden and Jarvis,
1997), which is the most widely-used fertiliser N form
in Great Britain. This factor is currently used in the
UK ammonia emission inventory (Misselbrook et al.,
2000). It is generally the case that uniform emission
factors are assumed across seasons (Pinder et al.,
2004) and there are insufficient data available to
determine different empirical relationships describing
NGAUGE, prediction of NH3 volatilisation from
fertiliser and its sensitivity to weather was achieved
using the model of Misselbrook et al. (2004). In this
model, it is assumed that emission from ammonium
nitrate fertiliser is moderated from a maximum value
by temperature only. In NGAUGE, this gives emission
factors ranging from 1.2 to 2.2% of applied N.
Emission of NH3from urine and dung deposited
while grazing was calculated as 15% of urine and 2%
of dung, as NCYCLE (Scholefield et al., 1991).
Insufficient data were available with which to vary
this factor by month or weather zone. For applied
manures, emission factors for NH3volatilisation were
determined according to the properties of the
farmyard manure (FYM) or slurry, its application
date and method of application (using data of
Misselbrook et al., 2000). These emission factors
range, for example, from 60% of applied N for dairy
slurry with 10% dry matter, surface-applied in
summer to 3% for dairy slurry with 2% dry matter,
injected in winter.
The optimisation procedure is the means by which
the best fertiliser distribution is calculated. There are
two main concepts behind the operation of the
(i) Goal-seeking to a specified target.
(ii) Satisfaction of optimisation criteria.
The procedure was based on the set of monthly plant
uptake (h factor) relationships, described earlier.
Initially, the average herbage N production of the
farm is used as the target for the optimisation, but a
field-specific target can be set by the user, and may be
herbage N, N loss or fertiliser N. For one of these, the
user selects the value desired (e.g. 300 kg herbage
N ha?1, 50 kg N ha?1loss, or 300 kg N ha?1fertiliser
applied). The end point of the optimisation is achieved
when the model reaches the targetvalue, satisfyingthe
optimisation criteria (Fig. 3).
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3926
Fig. 2. Relationships between nitrate leached and peak nitrate-N concentration (after Rodda et al., 1995).
As the optimisation progresses towards its target,
the optimisation criteria must be met in each iteration.
The objective of the development of NGAUGE was to
improve the efficiency with which N is used on
grassland farms and the optimisation criteria were
selected to reflect that. Three criteria are used, the one
inoperation inanygivenrunisdependent onthetarget
set by the user. All are a combination of maximising
herbage and the efficiency ratio (ER), defined as kg N
in herbage per kg N loss. These criteria may be
combined in a number of ways that favour either one
role in the running of the model may be considered as
outlined in Fig. 3. While the optimisation criteria in
run-time, andforthe purposes of theultimateenduser,
havebeen set, the procedure canbereadilyre-coded to
enable other logical optimisation criteria to be met.
To begin the optimisation, an initial amount of
fertiliser is allocated to all months and all pools are
calculated (i.e. plant, herbage, product, mineralisa-
tion, denitrification, volatilisation, leaching, urine,
dung). Fertiliser is provisionally transferred between
and the values of variables required for the optimisa-
tion criteria (currently herbage and ER) are compared.
The pair of months (i.e. one donating and one
receiving fertiliser) with the best combination of
herbage and ER is identified and the transfer of N is
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–39 27
Fig. 3. Outline of the operation of the optimisation process.
effected. This new pattern of fertiliser is the ‘optimal’
distribution of fertiliser at this N level, but this
fertiliser amount may be insufficient to achieve the
specified target. If the target variable (e.g. herbage)
has not yet met its target value, the procedure
effectively returns to the top of Fig. 3 and another
unit of fertiliser is applied to all months.
2.1.9. Manure management
In order for N pools to be calculated, account must
be taken of N supply from application of organic
manures. Two slurry types and two FYM types are
available for selection in NGAUGE, each associated
with default values of ammoniacal N, organic N and
user. Following application of manure (with amount
and month of application specified by the user),
volatilisation of NH3 is simulated. The remaining
inorganic N from slurry or FYM (ammoniacal
N ? volatilised N) is assumed to enter the soil
inorganic N pool in the month of application. The
mineralisation of organic N in manure (transfer of N
between the manure organic N and soil inorganic N
pool) is simulated each month according to applica-
tion date, C:N ratio (which is specified within the
model according to manure type) and cumulative
degree days above 5 8C, according to the factors
derived by Chadwick et al. (2000).
NGAUGE has not been designed to optimise
manure application dates and amounts because it is
considered that this will be determined by funda-
mental constraints of the system, such as volume of
slurry storage available and by legislation. However,
the N from manure is taken into account when
calculating an optimal fertiliser recommendation.
Location (i.e. weather) has a substantial effect on
both the initial calculations of N pools and the
outcome of optimisation for a particular target. In the
initial run of NGAUGE, the weather is assumed to be
‘average’ for that location (rain and temperature
zone). Thirty-year average (1961–1990) weather data
were obtained from seven weather stations represent-
ing the range of agricultural weather conditions in
Britain (Penecuik, High Mowthorpe, Waddington,
Wattisham, Shawbury and Yeovilton). Data used were
monthly total rainfall, monthly average daily tem-
perature and soil moisture deficit, calculated for grass
on a medium soil by the MORECS system (Thompson
rainfall, temperature and atmospheric input may be
considered independently, thus increasing the com-
with NCYCLE, there are three zones for atmospheric
N input, setting values at 15, 25 and 35 kg N ha?1but
in the new model there are six zones for rainfall (based
on average summer rainfall) and six zones for
temperature, giving a total of 36 possible tempera-
Weather impacts on plant growth both directly and
(ii) denitrification (influencing the amount of inor-
ganic N in soil);
(iii) plant growth directly.
The effects of soil water and temperature on mi-
neralisation and denitrification were described in S-
ection 2.1.3. For the effect of weather on plant growth,
both temperature and soil moisture factors were co-
nsidered. Plant N uptakewas assumed to be limited by
temperature according toa linear relationship between
factors of 0 at 5 8C and 1 at 20 8C (as Dowle and
Armstrong, 1990). The effect of water availability was
included by assuming that the growth factor was 1 at
the moisture content at 2 bar of suction for each soil
texture and 0 at the corresponding moisture content at
15 bar suction (permanent wilting point). Although
water is considered available between suctions of 0-
.05–15 bar (Hall et al., 1977), that held at suctions of
less than 2 bar is generally considered easily available
(Brady, 1984). Similar limitation factor approaches to
the calculation of the effect of moisture on plant gr-
owth have previously been adopted. Dowle and Ar-
mstrong (1990), for example, assumed that maximum
growth was possible between field capacity and wil-
ting point, declining linearly outside this range to 0 at
100% soil moisture in the root zone and, at the op-
posite end of the range, at permanent wilting point.
184.108.40.206. Updating weather. Within an actual year of
operation of the model, the observed N pools may be
significantly affected by weather, and users may need
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3928
to change their fertiliser management in the light of
incident weather in order to achieve their specified
targets. NGAUGE has sensitivity to on-going monthly
weather in order to achieve these objectives, which
impacts on three major sub-models: denitrification,
mineralisation and plant uptake.
Using the 30-year weather data described earlier,
data were analysed for each rainfall and temperature
zone, to produce five classes of data, representing the
10th, 30th, 50th, 70th and 90th centile. For the user,
these centiles are accessed by the selection of weather
categories for the preceding months (very wet, wet,
average, dry and very dry for rainfall and very warm,
warm, average, cold and very cold for temperature).
These weather data, and denitrification and miner-
alisation factors derived from them are held in arrays
and are used to re-calculate pools from the beginning
of the year to the end of the last full month before
today’s date. For example, for denitrification, arrays
exist for temperature (of dimensions month, zone,
centile) and water-filled pore space (of dimensions
month, zone, centile, soil texture). To take an example
ofthis, the 50thcentile fora clayloam inzone 3would
give a water-filled pore space denitrification factor of
0.177, for the 10th centile this would be 0.02 and for
the 90th centile 2.82.
Having re-calculated all pools according to the
weather for the preceding months, the original
recommendation (which was based on average
weather) is updated to take account of the new
weather information. This involves re-running the
optimisation procedure. In this updating mode,
although all months are included in the procedure
and its calculations, movement of fertiliser can only
take place between months that are forward of today’s
date (including the current month). Clearly, there are
often cases in which it may no longer be possible to
reach the user-specified target, particularly where this
is related to loss. Achievement of the target may now
be associated with different losses, different herbage
total or different fertiliser totals, and it is necessary
that the user is aware of this.
3. Model validation
The performance of NGAUGE was evaluated in
two ways: (a) assessment of the closeness of
predictions and observations of N loss and transfor-
mation; and (b) investigation of the effect of
NGAUGE fertiliser recommendations on N losses
on paddocks of commercial dairy farms. Resultsof the
latter will be presented in another paper. Data from a
purpose-built cut-plot experiment in mid Devon, UK
were used to evaluate the predictions of NGAUGE
against field measurements. The site was on an old
sward (more than 20 years old), which had received
moderate fertiliser additions for the past 13 years. The
average annual rainfall was 1025 mm, 550 mm of
which was in excess of evapotranspiration. The plots
(each 10 m ? 3 m) were laid out in a randomised
design with their long axes aligned with the direction
of slope (approximately 58) and were hydrologically
isolated to a depth of 30 cm with vinyl sheet. Drainage
via runoff and lateral flow was channelled to tipping
bucket flow monitors with flow proportional samplers
(Scholefield and Stone, 1995). Half of the plots were
re-seeded in the year prior to measurements, to
provide a contrast in sward age. Three N treatments
were applied, corresponding to approximately 230,
300 and 420 kg N ha?1year?1. Applications, as
ammonium nitrate, were made monthly according
to a pattern prescribed by the model. Mineralisation
was determined on a monthly basis using the method
of Hatch et al. (1990). Herbage was cut to 25 mm and
were then dried in a forced draught oven at 85 8C for
18 h and weighed. Goodness of fit of observed and
predicted fluxes was assessed using the method of
Measured mean values of net mineralisation were
generally greater than those predicted by the model in
both years, although due to the large variation in
observations on each treatment, observed and pre-
dicted rates were not significantly different in 9 out of
12 treatment years. Dry matter yields were generally
under-predicted in year 2, and over-predicted in year
1, suggesting that there is no systematic error in the
model’s predictions. Unusually low yields were
observed in year 1 (e.g. less than 5 t ha?1year?1of
drymatter froma fertiliser
350 kg N ha?1year?1) on the old sward, and the
reason for this was not clear. Nitrate leaching was
generally well predicted by NGAUGE: in 8 out of 12
cases, there was no significant difference between
modelled and measured values. The good agreement
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–39 29
between modelled and measured values is shown in
under-predict large values of leaching, but there are
insufficient points at the high end of the range to
determine whether this is genuinely the case. Peak
nitrate concentration was also well predicted, with no
significant difference betweenmodelled andmeasured
values in 8 out of 12 treatment years.
4. User interface description
NGAUGE was programmed in Borland Delphi 5.
This is an object-oriented language, which associates
portions of code with ‘events’ that happen to objects
(e.g. a click on the ‘run’ button). It was written in a
modular structure,using procedures and functions that
can be called from any part of the program.
The user interface was designed and constructed in
consultation with farmers, advisors, computer pro-
grammers and others with experience in DSS software
development. User preferences suggested that a
modular design with as few screens as possible
shouldbe aimed at.Thus,NGAUGEhastwo inputand
three output screens, each with logical positioning of
check boxes, menus, edit boxes, tables and graphs.
The first screen is used to enter generalised data about
N use on the whole farm. From this, the model
calculates the average N flows on the farm and
provides a target herbage yield as an initial basis for
fertiliser optimisation on individual fields (output
screen 1). It also calculates an N balance for the whole
farm. This was included in order to give farmers a
general appreciation of the magnitude of N inputs and
losses on their farms, and the potential for improve-
ment with optimisation. In the current stage of the
DSS, it is based on the generalised data entered about
the farm as a whole on input screen 1, and does not
make calculations based on individual field inputs.
The second input screen is used to enter data about
individualfields for which optimised fertiliser patterns
are required. Output screen 2 displays optimised and
non-optimised N budgets according to target, with a
facility to graph these, a histogram of the optimised
fertiliser distribution and a means to alter the
optimisation target (e.g. Fig. 5). Output screen 3 is
dedicated to updating the inputs and targets according
5. Use of NGAUGE for prediction of existing and
optimised N flows on livestock farms
The degree to which optimisation is able to
improve upon the predicted performance of a
conventional system is dependent on the character-
istics of the system (weather, soil type, fertiliser use,
etc.) and the optimisation performed. Some examples
of NGAUGE runs are given below to exemplify its
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–39 30
Fig. 4. Comparison of observed and predicted annual nitrate leaching from a clay loam soil with six treatments of contrasting sward age and N
input over 2 years.
capability. For each scenario, the results from an
optimised and non-optimised run are given. The
conventional or non-optimised fertiliser distribution is
based on MAFF (2000), for a fertiliser input of
300 kg N ha?1.
5.1. Effect of soil texture
Soil texture exerts an important effect on N2O, N2
and NO losses by operating on both levels of
regulation of N-gas products; it affects the process
rate at which N is moving through the ‘‘pipe’’
(nitrification and denitrification net rates) and it
controls the sizes of the holes through which the N-
oxides ‘‘leak’’ (are transported to the atmosphere). In
Table 1, non-optimised and optimised outputs are
compared for a well-drained sandy loam and a poorly-
climatic characteristics (11.5–12 8C and 400–450 mm
average growing season temperature and rainfall,
respectively).The effect of soil texture may be seen by
comparing runs A and C, non-optimised runs for a
sandy loam and clay loam, respectively. N2O and N2
fluxes in the clay loam soil were much higher than in
the sandy loam soil, reflecting the better diffusion of
thegaseous N compounds with lower water-filled pore
space. The simulated N2O:N2from denitrification and
NO:N2O ratios in the sandy loam were greater (by
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3931
Fig. 5. Example of NGAUGE output screen 2, showing output of a non-optimised field-based run at a fertiliser input of 248 kg N ha?1and the
corresponding optimised run using the same herbage N (327 kg N ha?1) as a target.
factors of 2.6 and 64, respectively) than in the clay
loam soil because of the enhanced water-filled pore
space in the latter. Soil moisture governs whether
nitrification or denitrification is the dominant process
and strongly influences the corresponding turnover as
well as the ratio of NO production over consumption
rates. The slightly larger N2O emission in optimised
than non-optimised systems arises because the
optimisation criteria are based on total N loss, rather
than individual N loss processes. The effect is
particularly apparent on well-drained soils, in which
the largest component of loss is leached N.
For both soil textures, the efficiency with which the
herbage yield was achieved (described by ER) was
improved by optimisation. For grazed systems
(Table 1), greater reductions in losses were possible
following optimisation on the sandy loam (runs A and
B)than clay loam(runsCandD) systems(35and17%
reduction in loss for the sandy loam and clay loam,
respectively). On heavier-textured soils, denitrifica-
tion is often the major route of N loss, and the period
with greatest potential for denitrification coincides
with the period for greatest potential for grass growth.
The fertiliser distribution formaximum plantuptakeis
thus always compromised by the criterion that ER
must increase when fertiliser is moved between
months in the optimisation procedure.
In these grazed systems with higher N inputs and N
returns from the grazing animals continuing into
the sandy loam soil at both locations becomes more
polarised towards the beginning of the year (see
Fig. 6). The residual effects of these early applications
will be carried through as soil mineral N into the later
months. In the non-optimised system (run A), leached
N from the sandy loam site accounted for 33% of the
fertiliser applied. This percentage was reduced to 19%
for optimised systems on the sandy loam soil (run B).
The fertiliser distribution from the optimisation
in the case of the clay loam soil, avoiding large
applications in the period of maximum denitrification
risk (March–May, Fig. 6). For the sandy loam soil, the
denitrification risk is smaller because of the relatively
smaller retention of water within the soil. Fertiliser
distribution can also be a major factor affecting the
proportion of N2over N2O that it is actually emitted to
the air.Comparing two sandy loam soils with the same
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3932
Predicted N outputs from non-optimised and optimised grazed systems on soils of contrasting texture at location 1
Outputs (kg N ha?1year?1)
SL and WD
CL and PD
SL = sandy loam; CL = clay loam; WD = well-drained; PD = poorly-drained; N = non-optimised; O = optimised; ER = efficiency ratio (kg N in herbage per kg N lost);
DM = herbage dry matter yield.
management characteristics in the same agroclimatic
area but with a different fertiliser distribution (runs A
and B), NGAUGE predicted that a more temporally
even fertiliser distribution (run A) resulted in a lower
N2O:N2ratio (22% smaller). This result agrees well
with other studies (Firestone and Davidson, 1989).
The optimisation process reduced the amount of
fertiliser required to reach the herbage target by 9 and
18% for the grazed clay loam and sandy loam,
5.2. Effect of cutting/grazing management
The reduced efficiency of fields grazed by animals
relative to cut-only fields can be seen for two
contrasting soil textures in location 1 by comparison
of runs A and C (grazed management, Table 1) with
runs E and G (cut-only, Table 2). The ‘grazed’
scenarios simulate the effect of grazing with dairy
cows from April to September, inclusive. The reduced
ER of grazed areas, relative to cut systems (e.g. 2.8 for
run A, compared to 5.9 for run E) is due to the greater
total N in the system (as a result of animal excretion)
and the addition of volatilised N to the total loss.
Under the grazed system, all N losses (NO, N2, N2O,
NH3 and leaching) were greater than under cut
systems, because of the greater total throughput of soil
inorganic N. (To aid examination of the effects of
grazing alone, this comparison does not address the
potential applications of manure to cut fields, which
may take place in reality. The effect of manure
application is examined in Section 5.4.)
The ER for both cut and grazed systems was
improved by optimisation, compared with the non-
optimised runs, with the improvement under cut
systems being greater than that under grazed. The
larger proportion of the N input, i.e. that from returns
of dung and urine from grazing animals, cannot
directly be optimised, although it is affected by the
5.3. Effect of weather zone
Selection of different temperature and rainfall
zones has a significant effect on both the simulated
from optimisation. To demonstrate this, two locations
were compared: location 1 has an average growing
season temperature of 11.5–12 8C (temperature zone
2) and an average growing season rainfall of 400–
450 mm (rainfall zone 4), while location 2 has an
average temperature of 9–10 8C (temperature zone 5)
and a rainfall of 300–350 mm (rainfall zone 2). The
sites were identical in all other respects. For an non-
optimised cut system with 300 kg N ha?1fertiliser
applied, the herbage dry matter yields from location 2
were 14 and 13% smaller for sandy loam (Table 3, run
M) and clay loam (run O), respectively, than from
location 1 (Table 2, runs E and G). Annual
mineralisation calculated in a non-optimised run
was 44 and 36% smaller at location 2 than location
1, for the sandy loam and clay loam soils, respectively.
The cooler, drier location 2 also had smaller N losses
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–39 33
Fig. 6. Fertiliser distributions from optimised runs for a grazed sandy loam (run B, black bars) and grazed clay loam (run D, white bars) at
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–39
Predicted N outputs from non-optimised and optimised cut-only systems on soils of contrasting texture at location 1
RunSoil typeLocation ManagementRun Outputs (kg N ha?1year?1) Peak NO3?-N
Leached N NH3
SL and WD1
CL and PD
SL = sandy loam; CL = clay loam; WD = well-drained; PD = poorly-drained; N = non-optimised; O = optimised; ER = efficiency ratio (kg N in herbage per kg N lost);
DM = herbage dry matter yield.
Predicted N outputs from non-optimised and optimised systems with contrasting grazing management and soil texture at location 2
Run Soil typeLocation ManagementRun Outputs (kg N ha?1year?1) Peak NO3?-N
SL and WD2
CL and PD
SL and WD
CL and PD
SL = sandy loam; CL = clay loam; WD = well-drained; PD = poorly-drained; N = non-optimised; O = optimised; ER = efficiency ratio (kg N in herbage per kg N lost);
DM = herbage dry matter yield.
than location 1; denitrification was 65% smaller on
clayloamsoilatlocation 2(runO)thanlocation 1(run
G), for the non-optimised run.
Nitrous oxide and N2fluxesin thewarmer and drier
area were generally higher than in the colder and
wetter area, due in part to the increased mineralisation
and greater inorganic N throughflow in the system. In
drier areas, the N2O:N2 and NO:N2O ratios were
generally higher than in wetter areas.
5.4. Accounting for manures
NGAUGE was used to investigate the effect of N
from applied manures on N cycling in grassland
systems and the degree to which fertiliser use may be
reduced by taking account of this source. From a
starting distribution of 300 kg N ha?1fertiliser (based
on RB209 as earlier) applied to a cut sward on a well-
drained sandy loam soil (as run E), the application of
30 t ha?1dairy slurry was simulated in February, May
and November. This resulted in a significant increase
in leached N (Table 4, run Q), which was particularly
due to the November application. Fertiliser use was
then optimised to achieve the same herbage yield with
more efficient N use, resulting in a 15% reduction in
fertiliser use (run R). The effect of optimisation on
nitrate leaching is compromised in this run by the
application of slurry, the timing and amount of which
is not determined by the optimisation process but is
considered as a fixed input. Injecting rather than
surface spreading the slurry allowed a further
7 kg N ha?1fertiliser to be saved (run T), and NH3
volatilisation to be substantially reduced (75%)
demonstrating its effect as an NH3abatement strategy.
However, nitrate leaching was predicted to increase
(37%) with this application method. NGAUGE
predicted an increase of 17% in N2O emissions when
injecting slurry (run Q compared to run S, Table 4).
This effect of larger N2O loss from injected than
(e.g. Dosch and Gutser, 1996).
The simulation of existing fertiliser, manure and
grazing practices in the non-optimised mode of
NGAUGE enables the user to investigate the likely
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3935
Table 4The effect of dairy slurry application on N outputs and optimisation from a cut system on a sandy loam soil at location 1
Outputs (kg N ha?1year?1)
SL and WD
SL and WD
SL = sandy loam; CL = clay loam; WD = well-drained; PD = poorly-drained; N = non-optimised; O = optimised; ER = efficiency ratio (kg N in herbage per kg N lost);
DM = herbage dry matter yield.
effects of changed management in any of these areas
on both production and losses of N through the main
processes of volatilisation, denitrification and leach-
ing. The simulation of all of these processes also
allows the potential effects of ‘pollution swapping’ to
be monitored, as strategies for the abatement of
individual loss processes are implemented. NGAUGE
could, for example, be used to investigate the effect of
the manure management changes associated with the
recent NVZ guidelines on losses of N via both nitrate
leaching, at which the legislation is aimed, and
gaseous losses. This legislation affects both amount
and timing of manure application to grassland of
particular characteristics. The effect on production of
the constraints imposed by this legislation could also
be assessed, for individual fields and farms.
optimised mode to investigate the effect of manage-
ment is the simulation of extending the grazing season
into periods in which the animals would traditionally
be housed. This is an increasingly popular practice in
as the southwest of England and south Wales. The
potential economic benefits of such grazing manage-
ment have been demonstrated (Frame and Laidlaw,
2001), but debate continues about the potential
environmental impacts of applying N, as fertiliser or
grazing returns, outside the conventional grazing
season. To assess the system fully, simulation would
need to include the effects of the timing and amount of
fertiliser application, the presence of grazing animals
and the application of reduced amounts of animal
of the system and has usefully been applied to the
assessment of extended grazing (Webb et al., 2005).
NGAUGE has a number of key advantages and
applications. First, it allows N to be used more
efficiently while still retaining the focus of the system
on production targets. Secondly, and perhaps more
importantly, it enables the focus to be shifted, and
the changing, and multiple, objectives of modern
agricultural systems. An example of this is the facility
to use N loss, rather than production, as a target for
optimisation. To make maximum use of this capability
and to enable NGAUGE to contribute to an existing
practical problem faced by the farming community,
some changes to the operation of the DSS may be
required, viz. makingpeak nitrate concentration rather
than ‘N loss’ the target of optimisation.
The scope for improvement in efficiency through
optimisation is limited by site factors, but more
importantly by the level of N input to the system. The
latter is obvious from the shape of the N response
curve in plants: there is greatest scope for improve-
ment with steepest gradient of the curve. At low N
inputs, N response is dominated by mineralisation
(largely unmanageable) while at high N inputs,
response to incremental N input is very low.
While the model is capable of optimising the
efficiency of N use for a particular grassland system,
the optimised pattern of herbage production (high
yields in early summer) may not be compatible with
the farmer’s preferred stock management. UK live-
stock management encompasses a range of degrees of
reliance on grazed grass, with some farms operating
zero grazing systems with indoor feeding of cut and
conserved forage, and others utilising grazed grass
throughout the year. The distribution of herbage
production predicted by optimisation would benefit
more the former, which reflects the basis for the
popularity of silage-based grassland production.
The NGAUGE DSS provides site-specific fertiliser
to the existing UK fertiliser recommendation system
(MAFF, 2000), the potential losses of N are taken into
account in the production of this recommendation,
both by ensuring that the target is achieved with the
greatest ratio of herbage N to N lost, and by providing
the facility for N losses to be entered as a target.
While there have been several other approaches to
decision support for N fertiliser management, origi-
nating in The Netherlands (Dairy Farming Model, Van
De Ven, 1996), France (AzoPa ˆt, Decau etal., 1997 and
Delaby et al., 1997) and New Zealand (NLE, Di and
Cameron, 2000; and OVERSEER, Wheeler et al.,
2003), NGAUGE is unique in its combination of
farmer-friendly user interface, sophisticated descrip-
tion of process and optimisation capability, that
enables both production and environmental losses to
be quantified. In addition, NGAUGE is capable of
interfacing with budget- and indicator-based systems
of fertiliser management (e.g. Jarvis et al., 1996;
Schroder et al., 2003). Because of this combination of
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3936
ease of use and complexity of simulation, the DSS
should be of benefit to a variety of users. In addition to
its use by farmers and their advisors, it could be used
by policy makers to explore mitigation options for
enabling compliance with N loss legislation (e.g. for
NVZ regulation compliance, as mentioned earlier);
and by researchers to explore impacts of novel farm
managements on pollution swapping and fundamental
controls on the efficiency of the system. To aid
progress towards these objectives the model could be
further developed to enable a wider range of forage
crops to be considered and to explore the advantages
of within- and between-farm optimisation.
NGAUGE provides a basis for improved decision-
making about fertiliser management on grassland
of the magnitude of N losses and provides a means of
improving the efficiency with which N used on
grassland fields. The potential for improvement in
efficiency was found to be dependent on site
characteristics and existing management, with the
greatest improvement possible on sandy-textured soils
with moderate N inputs. It was possible to reduce
nitrate leaching by up to 46% and annual fertiliser use
by up to 33%, without compromising herbage yield.
Field-specific fertiliser recommendations are pro-
vided, according to the user-specified target, soil
texture and drainage status, weather, land-use history
and manure use. The optimisation procedure was
developed with dual criteria of increased herbage
production and reduced losses for a given N input,
of undesirable losses compared to existing recom-
of these criteria in future applications, to further shift
recommendation, for example, where limitation of a
specific loss pathway is of particular importance.
The development of NGAUGE was funded by
DEFRA, London (NT1601, NT1603). We thank
Eunice Lord (ADAS) for analysis of the long-term
weather data and helpful comments during the
development of NGAUGE, and Colin Brown (NSRI)
for the soil moisture algorithms. The input of farmers
to the user interface evaluation is gratefully acknowl-
edged. IGER is sponsored by the Biotechnology and
Biological Sciences Research Council.
Anderson, I.C., Levine, J.S., 1986. Relative rates of nitric-oxide and
nitrous-oxide production by nitrifiers, denitrifiers, and nitrate
respirers. Appl. Environ. Microb. 51 (5), 938–945.
Aulakh, M.S., Doran, J.W., Dossier, A.R., 1992. Soil denitrifica-
tion—significance, measurement, and effects of management.
Adv. Soil Sci. 18, 1–57.
Bailey, L.D., Beauchamp, E.G., 1973. Effects of moisture, added
NO3?and macerated roots on NO3?transformation and redox
potential in surface and subsurface soils. Can. J. Soil Sci. 53,
Blantern, P., 1991. Factors affecting nitrogen transformations in
grazed grassland soils with specific reference to the effects of
artificial land drainage and N fertiliser. Ph.D. thesis, University
of Exeter, UK.
Brady, N.C., 1984. The Nature and Properties of Soils, ninth ed.
Macmillan, New York.
Chadwick, D.R., John, F., Pain, B.F., Chambers, B.J., Williams, J.,
Cho, C.M., Sakdinan, L., Chang, C., 1979. Denitrification intensity
Davidson,E.A., Matson,P.A., Vitousek,P.M., Riley, R., Dunkin,K.,
emissions of NO and N2O in a seasonally dry tropical forest.
Ecology 74 (1), 130–139.
Decau, M.L., Delaby, L., Roche, B., 1997. AzoPa ˆt: une description
quantifie ´e des flux annuels d’azote en prairie pa ˆture ´e par les
vaches laitie `res: 2. Les flux du syste `me sol-plante. Fourrages
Delaby, L., Decau, M.L., Peyraud, J.L., Accarie, P., 1997. AzoPa ˆt:
une description quantifie ´e des flux annuels d’azote en prairie
pa ˆture ´e par les vaches laitie `res: 1. Les flux asocies a ` l’animal.
Fourrages 151, 297–311.
Di, H.J., Cameron, K.C., 2000. Calculating N leaching losses and
critical nitrogen application rates in dairy pasture systems using
a semi-empirical model. J. Agric. Res. 43 (1), 139–147.
Dosch, P., Gutser, R., 1996. Reducing N losses (NH3, N2O, N2) and
immobilization from slurry through optimized application tech-
niques. Fertil. Res. 43, 165–171.
Dowle, K., Armstrong, A.C., 1990. A model for investment apprai-
sal of grassland drainage schemes on farms in the UK. Agric.
Water Manage. 18, 101–120.
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3937
Firestone, M.K., Davidson, E.A., 1989. Microbial basis of NO
and N2O production and consumption in the soil. In: Andraea,
M.O., Schimel, D.S. (Eds.), Exchange of Trace Gases Between
Terrestrial Ecosystems and the Atmosphere. Wiley, pp. 7–21.
Frame, J., Laidlaw, A.S., 2001. Extending the grazing season.
Forage Matters 5, 1–3.
nitrification inhibition. Soil Sci. Soc. Am. J. 48 (6), 1262–1266.
Hall, D.G.M., Reeve, M.J., Thomasson, A.J., Wright, V.F., 1977.
Water retention, porosity and density of field soils. Soil Survey
Technical Monograph 9, Harpenden.
Hansson, A.C., Pettersson, R., 1989. Uptake and above- and below-
ground allocation of soil mineral-N and fertilizer-15N in a
perennial ryegrass ley (Festuca pratense). J. Appl. Ecol. 26,
Hatch, D.J., Jarvis, S.C., Philips, L., 1990. Field measurement of
nitrogen mineralisation using soil core incubation and acetylene
inhibition of nitrification. Plant Soil 124, 97–107.
Hopkins, A., Gilbey, J., Dibb, C., Bowling, P.J., Murray, P.J., 1990.
Response of permanent and reseeded grassland to fertilizer
nitrogen: 1. Herbage production and herbage quality. Grass
Forage Sci. 45, 43–55.
Isermann, K., 1990. Share of agriculture in nitrogen and phosphorus
emissions into the surface waters of Western Europe against the
background of their eutrophication. Fertil. Res. 253–269.
Jarvis, S.C., 1993. Nitrogen cycling and losses from dairy farms.
Soil Use Manage. 9 (3), 99–105.
Jarvis, S.C., Barraclough, D., Williams, J., Rook, A.J., 1991.
Patterns of denitrification loss from grazed grassland—effects
of N fertiliser inputs at different sites. Plant Soil 131, 77–88.
Jarvis, S.C., Wilkins, R.J., Pain, B.F., 1996. Opportunities for
reducing the environmental impact of dairy farming manage-
ments: a systems approach. Grass Forage Sci. 51 (1), 21–31.
Li, C., Frolking, S., Frolking, T.A., 1992. A model of nitrous oxide
evolution from soil driven by rainfall events: 1. Model structure
and sensitivity. J. Geophys. Res. 97, 9759–9776.
Macduff, J.H., White, R.E., 1985. Net mineralization and nitrifica-
tion rates in a clay soil measured and predicted in permanent
grassland from soil temperature and moisture content. Plant Soil
MAFF, 2000. Fertiliser Recommendations for Agricultural and
Misselbrook, T.H., Sutton, M.A., Scholefield, D., 2004. A simple
process-based model for estimating ammonia emissions from
agricultural land after fertilizer applications. Soil Use Manage.
Misselbrook, T.H., Van der Weerden, T.J., Pain, B.F., Jarvis, S.C.,
Chambers, B.J., Smith, K.A., Phillips, V.R., Demmers, T.G.M.,
2000. Ammonia emission factors for UK agriculture. Atmos.
Morrison, J., Jackson, M.V., Sparrow, P.E., 1980. The response of
perennial ryegrass to fertiliser nitrogen in relation to climate and
soil. Report of the joint ADAS/GRI Grassland Manuring Trial—
GM20. Grassland Research Institute, Agricultural Development
and Advisory Service, Rothamsted Experimental Station, Hur-
Mosier, A.R., Parton, W.J., Hutchinson, G.L., 1983. Modeling
nitrous-oxide evolution from cropped and native soils. Ecol.
Bull. 35, 229–241.
Neff, J.C., Keller, M., Holland, E.A., Weitz, A.W., Veldkamp, E.,
burning of a secondary tropical rain forest. J. Geophys. Res.
Atmos. 100 (D12), 25913–25922.
Ourry, A., Boucaud,J., Salette, J., 1988. Nitrogenmobilization from
stubble and roots during re-growth of defoliated perennial
ryegrass. J. Exp. Bot. 39, 803–809.
from tropical pasture and rain-forest soils. Biol. Fertil. Soils 20,
The physiology of grass production under grazing: II. Photo-
swards. J. Appl. Ecol. 20, 127–139.
Parton, W.J., Holland, E.A., Del Grosso, S.J., Hartman, M.D.,
Martin, R.E., Mosier, A.R., Ojima, D.S., Schimel, D.S., 1996.
Generalized model for N2and N2O production from nitrification
Pinder, R.W., Pekney, N.J., Davidson, C.I., Adams, P.J., 2004. A
process-based model of ammonia emissions from dairy cows:
improved temporal and spatial resolution. Atmos. Environ. 38,
Powlson, D.S., 2000. Tackling nitrate from agriculture: foreword.
Soil Use Manage. 16, 141.
Reichman, G.A., Grunes, D.L., Viets Jr., F.G., 1966. Effects of soil
moisture on ammonification and nitrification in two northern
plain soils. Soil Sci. Soc. Am. Proc. 30, 363–366.
Rodda, H.J.E., Scholefield, D., Webb, B.W., Walling, D.E., 1995.
Management model for predicting nitrate leaching from grass-
land catchments in the United Kingdom: 1. Model development.
Hydrol. Sci. J. 40, 433–451.
Ryden, J.C., 1981. N2O exchange between a grassland soils and the
atmosphere. Nature 292, 235–237.
Ryden, J.C., 1984. The flow of nitrogen in grassland. Proc. Fertil.
Soc. 229, 3–44.
Salway, A.G., Murrells, T.P., Milne, R., Ellis, S., 2001. UK Green-
house Gas Inventory, 1990 to 1999. Annual Report for submis-
sion under the Framework Convention on Climate Change.
National Environmental Technology Centre.
Scholefield, D., Hawkins, J.M.B., Jackson, S.M., 1997. Use of a
flowing helium atmosphere incubation technique to measure the
effects of denitrification controls applied to intact cores of a clay
soil. Soil Biol. Biochem. 29, 1337–1344.
Scholefield, D., Stone, A.C., 1995. Nutrient losses in runoff water
following application of different fertilisers to grassland cut for
silage. Agric. Ecosyst. Environ. 55, 181–191.
Scholefield, D., Tyson, K.C., Garwood, E.A., Armstrong, A.C.,
Hawkins, J., Stone, A.C., 1993. Nitrate leaching from grazed
grassland lysimeters—effects of fertiliser input, field drainage,
age of sward and pattern of weather. J. Soil Sci. 44, 601–
Scholefield, D., Lockyer, D.R., Whitehead, D.C., Tyson, K.T., 1991.
A model to predict transformations and losses of nitrogen in UK
pastures grazed by beef cattle. Plant Soil 132, 165–177.
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3938
Schroder, J.J., Aarts, H.F.M., ten Berge, H.F.M., van Keulen, H., Download full-text
Neeteson, J.J., 2003. An evaluation of whole-farm nitrogen
balances and related indices for efficient nitrogen use. Eur. J.
Agron. 20 (1–2), 33–44.
Skiba, U., Hargreaves, K.J., Fowler, D., Smith, K.A., 1992. Fluxes of
nitric and nitrous oxides from agricultural soils in a cool tempe-
rate climate. Atmos. Environ. A: Gen. 26 (14), 2477–2488.
Stanford, G., Dzienia, S., Vander Pol, R.A., 1975. Effect of tem-
perature on denitrification rate in soils. Soil Sci. Soc. Am. Proc.
Stanford, G., Epstein, E., 1974. Nitrogen mineralization water
relations in soils. Soil Sci. Soc. Am. Proc. 38, 103–107.
Thompson, N., Barrie, I.A., Ayles, M., 1981. The Meteorological
Office Rainfall and Evapotranspiration Calculation System
(MORECS). The Meteorological Office, London.
Van Breemen,N., Burrough,P.A., Velthorst,E.J., Van Dobben,H.F.,
De Wit, T., Ridder, T.B., Reijnders, H.F.R., 1982. Soil acidifica-
tion from atmospheric sulphate in forest canopy throughfall.
Nature 299, 548–550.
Van De Ven, G.W.J., 1996. A mathematical approach to comparing
environmental and economic goals in dairy farming on sandy
soils in The Netherlands. Ph.D. thesis, Wageningen Agricultural
University, Wageningen, 239 pp.
Van der Weerden, T.J., Jarvis, S.C., 1997. Ammonia emission
factors for N fertilizers applied to two contrasting grassland
soils. Environ. Pollut. 95, 205–211.
Webb, J., Anthony, S.G., Brown, L., Lyons-Visser, H., Ross, C.,
Cottrill, B., Johnson, P., Scholefield, D., 2005. The impact of
ammonia and nitrous oxide and on nitrate leaching in England
and Wales. Agric. Ecosyst. Environ. 105, 307–321.
Wheeler, D.M., Ledgard, S.F., deKlein, C.A.M., Monaghan, R.M.,
Carey, P.L., McDowell, R.W., Johns, K.L., 2003. OVERSEER1
nutrient budgets—moving towards on-farm resource account-
ing. Proc. N. Z. Grass. Assoc. 65, 191–194.
Whitmore, A.P., 1991. A method for assessing the goodness of
computer simulation of soil processes. J. Soil Sci. 42 (2), 289–
L. Brown et al./Agriculture, Ecosystems and Environment 109 (2005) 20–3939