200 A. Korsaeth, R. Eltun Agriculture, Ecosystems and Environment 79 2000 199–214
limiting nutrient for producing food and maintaining the level of soil N is essential for sustaining agricul-
tural productivity. Agronomic practices affect the balance between
production and mineralisation of soil organic N. The characteristically slow turnover of the N pools implies
that low input farming systems may sustain fairly high productivity at the cost of a gradual decline of
the soil organic N. The degree to which an agronomic system maintains the soil organic N level is thus one
of several measures of its sustainability. This crite- rion is conditional, however, since net changes also
depend on the history of the soil. It is difficult to avoid a net reduction of soil organic N in soils with a
high content of mineralisable organic N, especially in arable cropping systems Christensen, 1990; Uhlen,
1991; Heenan et al., 1995; Thomsen and Christensen, 1998. At the other extreme, most cultivation regimes
will result in a net accumulation of soil organic N on soils with very low initial soil organic N levels Fet-
tell and Gill, 1995; Poulton, 1995; Raun et al., 1998. Thus the sustainability of cultivation regimes with
respect to soil organic N levels can only be compared on equal terms, i.e., when the regimes are run side
by side on the same type of soil and under identical climatic conditions.
Changes in the soil N pool can only be observed over an extended time Poulton, 1995, since the an-
nual changes are small as compared to the total pool size e.g., Uhlen, 1991. Long term experiments are
therefore needed Dick, 1992. It is, however, desirable to obtain information on trends at an early stage, in or-
der to correct the development in the present cropping systems. Good correlations have been found between
calculated N balances and changes in soil N Uhlen, 1989; Nyborg et al., 1995, and thus nitrogen mass
balance N balance studies are a suitable approach to analyse agroecosystems Wood et al., 1991.
Systems with a similar difference between inputs and outputs of N may, however, differ greatly in their
losses of N to the environment, as high losses may be counteracted by high inputs. One major path for
N losses from agriculture is the transport via surface and drainage runoff, and the resulting pollution of wa-
ter resources has become an increasingly important problem Kristensen et al., 1995. In an ideal crop-
ping system, should not only the reduction of soil N be minimised but also the N runoff. Reliable measure-
ments of N runoff are difficult to obtain and require rather expensive lysimeter trials e.g., Bergström and
Brink, 1986; Uhlen, 1994; Thomsen and Christensen, 1998. Finding a simple but robust way to estimate
N runoff would thus be a great advantage when eval- uating cropping systems. Different factors have been
found to correlate with N runoff, such as fertiliser level Bergström and Brink, 1986; Bergström, 1987, pre-
cipitation Jenkinson, 1990, soil mineral-N content in early autumn Bergström and Brink, 1986; Vagstad
et al., 1997 and crop yield Vagstad et al., 1997. Hal- berg et al. 1995 stated in more general terms, that
the N balance fertiliser N plus N applied with ma- nure minus harvested N is an expression of the to-
tal potential for N losses from an agroecosystem, and thus an indirect indicator of N runoff. The direct rela-
tionship between N balances and N runoff at the field level has, however, seldom been investigated.
The Apelsvoll Cropping System Experiment was established in 1988–1989 with the aim of developing
sustainable and environmentally sound cropping sys- tems Eltun, 1994. The experiment consists of six dif-
ferent cropping systems with rotations of either mainly arable or mainly forage crops, ranging from low to
high inputs of N and other production factors. On the basis of results from the first 8 years of this experi-
ment one entire crop rotation, the objective of the study presented here was, i to quantify the major N
flows in the cropping systems, using a mass-balance approach, in order to analyse possible changes in the
soil N content, and ii to test the usefulness of mass balance calculations for predicting N runoff.
2. Materials and methods
2.1. Experimental site and treatments A detailed description of the experimental design
Eltun, 1994 and soil characteristics Riley and El- tun, 1994 have been presented elsewhere. Briefly, a
3.3 ha field lysimeter was established in 1988–1989 at Apelsvoll Research Centre in central southeast Nor-
way 60
◦
42
′
N, 10
◦
51
′
E, altitude 250 m. The climate of the region is humid continental with a mean an-
nual precipitation of 600 mm and a mean annual tem- perature of 3.6 and 12.0
◦
C in the growing season May–September. The experimental area, which slopes
A. Korsaeth, R. Eltun Agriculture, Ecosystems and Environment 79 2000 199–214 201
2–8 towards northeast, was used as pasture from 1935 to 1975. During the following 10 years the field
was cropped with a rotation including 10 root crops, 40 small grains and 50 ley, using an average of 10
tonnes slurry ha
− 1
per year containing approximately 50 g kg
− 1
total-N plus regular amounts of inorganic fertiliser. The field was lying fallow in 1986 and 1988
and under winter wheat in 1987. The major soil group at the experimental site is im-
perfectly drained brown earth Oxiaquic Cryoboroll, USDA; Gleyed melanic brunisoils, Canada Soil
Survey with dominantly loam and silty sand textures, and containing about 60 g kg
− 1
organic matter in the top soil. Six cropping systems, each with two repli-
cates, were distributed on 0.18 ha trial units within a randomised complete block design in the field
lysimeter. Each cropping system has an 8-year crop rotation with eight plots, so that all crops in the rota-
tion are present every year. They include conventional arable cropping CON-A, integrated arable cropping
INT-A, ecological arable cropping ECO-A, con- ventional forage cropping CON-F, integrated forage
cropping INT-F, and ecological forage cropping
Table 1 Characteristics of the cropping systems at Apelsvoll Research Centre, southeast Norway
Management Cropping system
Conventional Integrated
Ecological Conventional
Integrated Ecological
arable CON-A arable INT-A
arable ECO-A forage CON-F
forage INT-F forage ECO-F
Crop rotation Barley
a
Barley
a
Barley
b
Barley
b
Barley
b
Barley
b
Winter wheat Winter wheat
Clover grass 1st year ley
1st year ley 1st year ley
Oats Oats
Spring wheat
c
2nd year ley 2nd year ley
2nd year ley Barley
Barley Potatoes
3rd year ley 3rd year ley
3rd year ley Potatoes
Potatoes Barley
b
Swedes
d
Swedes
d
Swedes
d
Spring wheat Spring wheat
Clover grass Spring wheat
Spring wheat Green fodder
Oats Oats
Winter wheat
c
Oats Oats
Spring wheat
c
Barley Barley
Oats
c
Green fodder Green fodder
Oats and peas Fertiliser
e
+ +
+ +
Slurry
f
0.4 2.0
1.5 1.2
Soil tillage Spring ploughing
g
Spring harrowing Spring ploughing
Spring ploughing
g
Spring ploughing Spring ploughing
Plant protection Chemical
Integrated
h
Mechanical Chemical
Integrated
h
Mechanical
a
Early potatoes in the period 1990–1994.
b
With undersown clover grassley.
c
With undersown crop.
d
Fodder beet in the period 1990–1994.
e
Inorganic fertiliser.
f
Cattle slurry as big animal units per ha.
g
Autumn ploughing in the period 1990–1994.
h
Chemical protection, but with reduced amounts of pesticides and application times compared to the conventional systems.
ECO-F. Characteristics of the cropping systems are shown in Table 1. Each trial unit is separately drained
with PVC pipes at a depth of 1 m with spacing of 7.5 m. Surface runoff is collected at the lower end
of each trial unit and led to a sedimentation tank. Grass covered border zones separate the trial units.
Drainage water runoff and surface water runoff from the sedimentation tank is transported in sealed plastic
pipes to measuring stations equipped for discharge measurements by tipping buckets and for volume
proportional water sampling.
The experimental site was irrigated in cases of moderate or extreme moisture deficit, using a mobile
rain-gun, which delivered the same amount of water to all crops.
2.2. Measurements Cattle slurry slurry was sampled 1–2 weeks be-
fore application and analysed for total-N N
slurry
us- ing the Kjeldahl method. Ammonium-N and nitrate-N
in the slurry N
slurry, inorg
were extracted with 2M KCl and determined colorimetrically with an auto-
202 A. Korsaeth, R. Eltun Agriculture, Ecosystems and Environment 79 2000 199–214
analyser Traacs, Bran and Luebbe, Germany. Pre- cipitation was sampled on a monthly basis from a
rain gauge placed at the experimental area, and anal- ysed for total-N N
wet dep.
using the Kjeldahl method Allen et al., 1974.
The harvested crops and straw residues when re- moved were weighed four parallels on each rotation
plot and analysed for total-N N
harvest
using Kjeldahl digestion. The proportion of legumes was determined
visually before harvest. The water samples surface and drainage runoff
were analysed on a monthly basis for total-N with the Kjeldahl method. Ammonium-N and nitrate-N in
runoff was determined as the slurry samples, as de- scribed earlier.
Soil samples were taken at five depths 5–10, 20–25, 35–40, 50–55 and 65–70 cm with two replicates from
each model farm in 1988 Riley and Eltun, 1994 and from topsoil 0–25 cm at the same sites in 1995, and
analysed for total-N by the Kjeldahl method.
2.3. Estimates The input of N with seeds N
seed
was estimated us- ing measured N content of harvested grain and pota-
toes from the CON-A system. Literature values were selected for the legumes Lunnan, 1988, 1989 and the
N content in swede seedlings was set to 0.1 g per plant.
Volatilisation of NH
3
-N from cattle slurry N
volat. slurry
was estimated individually for each rotation plot and application time by the method of Horlacher and
Marschner 1990. The estimates for losses from cat- tle slurry applied to arable rotation plots were taken
directly from their framework as a percentage of the amount of NH
4
-N applied, by considering the factors expected infiltration rate of the slurry, mean daily air
temperature, precipitation and the time between appli- cation and precipitation or incorporation of the slurry.
The estimated N losses from slurry applied to ley were multiplied by 1.5, in order to compensate for a lower
expected rate of infiltration compared to arable fields Thompson et al., 1990.
Symbiotic N fixation, N
fixation
g N m
− 2
, was sim- ulated by a modified model of Hansen 1995. The
annual amount of fixed N was modelled as a product of the crop yield, Y g DM m
− 2
, the proportion of legumes, L g g
− 1
, the N content of the legumes, N
leg
g N g
− 1
DM, the fraction of N
leg
originating from fixation, F g N g
− 1
N, and a factor accounting for net accumulated fixed N below stubble height, R, as
shown in Eq. 1:
N
fixation
=
2
X
i= 1
4
X
j = 1
3
X
k= 1
Y
i
L
ij
N
leg ij
F
j
R
j k
1 Subscript i indicates the seasonal cut number, j the
legume type and k the year of ley. F
j
is given by: F
j
= F
max, j
− a
j
N
net inorg
F
j
≥ 2
where F
max, j
denotes the maximal fraction originat- ing from fixation, a
j
m
2
g
− 1
N
net inorg
is a constant and N
net inorg
g N m
− 2
is the net amount of inorganic N applied defined by:
N
net inorg
= N
fertiliser
+ N
slurry inorg
− N
volat. slurry
3 All parameter values are shown in Table 2.
Dry atmospheric deposition N
dry dep.
was set to 2 kg N ha
− 1
per year based on Tørseth and Manø 1996. Denitrification N
denitrification
was set to 7 of the net amount of inorganic N applied Eq. 3,
based on experiments executed under fairly compara- ble conditions Ryden, 1985; Svensson et al., 1991;
Maag, 1995. Net losses of NH
3
to the atmosphere by volatilisation from crops N
volat. crop
were set to 2 kg N ha
− 1
per year according to Holtan-Hartwig and Bøckman 1994. Non-symbiotic N fixation, reported
to be less than 0.8 kg ha
− 1
per year Haynes, 1986, was not considered.
2.4. Calculations and statistics All flows were calculated as total-N kg N ha
− 1
. Calculations of transported N via surface and
drainage runoff were based on measured N concen- trations and volumes of surface and drainage water
Eltun and Fugleberg, 1996. Nitrogen runoff occur- ring during one agrohydrological year, lasting from 1
May to 30 April Høyås et al., 1997, was attributed to the cropping season in the same hydrological year.
Yearly changes in soil N 1N were calculated as: 1
N = N
fertiliser
+ N
slurry
+ N
wet dep.
+ N
dry dep.
+ N
seed
+ N
fixation
− N
harvest
− N
volat. slurry
− N
volat. crop
− N
denitrification
− N
runoff
4
A. Korsaeth, R. Eltun Agriculture, Ecosystems and Environment 79 2000 199–214 203
Table 2 Parameters and variables used for estimating the biological fixation of nitrogen in Eqs. 1 and 2
Parameter Value
Reference
a
Description Y
kg ha
− 1
measured Yield of crop including legumes
L g g
− 1
measured Content of legumes in harvest
N gN gDM
− 1
Nitrogen content in legumes N
11
0.0394 [1]
White clover, 1st cut N
12
0.0349 [1]
White clover, 2nd cut N
21
0.0314 [1]
Red clover
b
, 1st cut N
22
0.0291 [1]
Red clover, 2nd cut N
3
0.028 [2]
Grey peas N
4
0.030 [2]
Common vetch F
max
gN
fixed
gN
− 1
Maximum fraction of fixed N in legumes F
max,1
0.923 [3]
White clover F
max,2
0.923 [3]
Red clover F
max,3
0.654 [4]
Grey peas F
max,4
0.654 set
Common vetch a
j
m
2
gN
net inorg −
1
Constant for fertiliser N response a
1
and a
2
0.026 [3]
White and red clover a
3
0.028 [3]
Grey peas a
4
0.028 set
Common vetch R
jk
– Correction for net accumulation of fixed N below stubble height
R
11
and R
21
1.27 [5]
White and red clover, 1st year of ley
c
R
12
and R
22
1.27 [5]
White and red clover, 2nd year of ley R
13
and R
23
1.00 [5]
White and red clover, 3rd year of ley R
31
1.045 [4]
Grey peas green fodder, oatspeas R
41
1.27 set
Common vetch green fodder
a
References are: [1] Lunnan 1989, [2] Lunnan 1988, [3] Hansen 1995, [4] Hansen 1993 and [5] Steen Kristensen et al. 1995.
b
The properties for alsike clover were assumed to be the same as for red clover.
c
Including the accumulation of fixed N in the year the ley was sown.
where N
runoff
is the sum of N lost via surface and drainage runoff.
Averaged data are presented with standard errors of the mean SE. Analysis of variance Gomez and
Gomez, 1984 was performed on the results of har- vested N and N runoff, using a split-plot model with
cropping system as major plot and year as subplot. When analysing data at the rotation plot level dry mat-
ter yield and legume content, a split-split-plot model was used, with cropping system as major plot, rota-
tion plot as subplot and year as sub-subplot. Paired comparisons LSD were performed using appropriate
standard errors of the mean difference and student t values Gomez and Gomez, 1984.
Various mass N balance calculations were con- ducted in order to find a simple predictor for N
runoff
Starting with the major N flows, N
fertiliser
and N
harvest
, the balance calculation was expanded stepwise to
include all the N flows in Eq. 4. Thus the most complex way to express the potential for N
runoff
Bal
complex
was calculated as: 1
Bal
complex
= N
fertiliser
+ N
slurry
+ N
wet dep.
+ N
dry dep.
+ N
seed
+ N
fixation
− N
harvest
− N
volat. slurry
− N
volat. crop
− N
denitrification
5 To assess the usefulness of the balances to predict
N
runoff
, we used the classical linear regression model Johnson and Wichern, 1992:
Y = β +
β
1
z
1
+ β
2
z
2
+ β
3
z
3
+ ε
6 where Y is N
runoff
for each system and year, β –
3
are parameters, z
1
–
3
are predictors and ε is the random error.
204 A. Korsaeth, R. Eltun Agriculture, Ecosystems and Environment 79 2000 199–214
In Step A the annual N balances Bal
x
were used as only predictor z
1
= Bal
x
, z
2
= z
3
= 0. Since precip-
itation has been found to be an important factor for N runoff Jenkinson, 1990, annual total precipitation
Prec
t
was alternatively tested z
1
= Prec
t
, z
2
= z
3
= 0.
In Step B both calculated N balances and precipita- tion were included as predictors z
1
= Bal
x
, z
2
= Prec
t
, z
3
= 0. In Step C we added the total precipitation
of the previous year Prec
t − 1
as a third predictor z
1
= Bal
x
, z
2
= Prec
t
, z
3
= Prec
t − 1
. The rationale for so doing was the assumption that the amount of
leachable N, which is not lost in 1 year e.g. due to shortage of water for N transport in a dry year,
increases the runoff potential the following year due to nitrate storage below the root zone. In order to
eliminate the effect of the climatic variations on the regressions, we finally regressed N balances averaged
over all years against average N runoff Y=average N
runoff
, z
1
= average Bal
x
, z
2
= z
3
= 0 Step D.
All statistical tests were performed at the 0.05 level of probability.
3. Results