Biosystems Engineering (2003) 84 (4), 467–480 Available online at www.sciencedirect.

com
doi:10.1016/S1537-5110(02)00284-2
PA}Precision Agriculture

Soil, Water and Yield Relationships in developing Strategies for the Precision
Application of Nitrogen Fertiliser to Winter Barley
I.T. James; R.J. Godwin
National Soil Resources Institute, Cranfield University, Silsoe, Bedfordshire MK45 4DT, UK; e-mail of corresponding author:
i.t.james@cranfield.ac.uk

(Received 28 September 2002; accepted in revised form 12 December 2002)

To investigate on-farm strategies for the site-specific application of nitrogen fertiliser, two hypotheses were
tested: that maximum economic output is either obtained by applying more nitrogen to the historically highest
yielding soil and less to the lowest yielding soil, or by the reverse strategy.
Field experiments were conducted in 1997, 1998 and 1999 on a commercially managed field of winter
barley (Hordeum vulgare), comprising a clay loam and a sandy loam, in Cambridgeshire, UK. The experiments
were carried out in a strip-based design using on-farm equipment. In all 3 yr, at typical application rates of
100–200 kg[N] ha
1, there were significant differences between the yield of the clay loam and the sandy loam
soil. When rainfall in the February to July period of the growing season was higher than average, the sandy loam
soil yielded significantly higher than the clay loam soil and the reverse occurred when the rainfall was lower.
Despite the above factors, analysis showed that the maximum of the yield response curve for each soil,
occurred at the same application rate in each growing season. Given this relationship, there was no economic
benefit from variable rate application of nitrogen to different soil units based upon historic yield, or any other
form of information.
# 2003 Silsoe Research Institute. All rights reserved
Published by Elsevier Science Ltd

1. Introduction from the harvests of 1997–1999, on a commercially

managed field of winter barley (Hordeum vulgare) in
The requirement for the development of strategies for Cambridgeshire, UK. The trials were carried out using a
the spatially variable application of nitrogen fertiliser strip design (Thompson & Robert, 1995; Ferguson et al.,
resulted from the commercial introduction of agricul- 1999), which was considered to be the most suitable
tural technologies that allowed the farmer to measure method of evaluating the nitrogen trials over a range of
yield patterns within fields and have spatially variable soil variability, within a commercial system.
control of input applications. Whilst the origin of It was research conducted in the USA that pioneered
precision farming in the UK is often associated with investigations into the variable rate application of
this commercial introduction of yield mapping combines fertiliser. The work of Carr et al. (1991) and Mulla
in 1993 (Dampney & Moore, 1999), a principal cause of et al. (1992) is widely regarded as the pioneering work in
within-field yield variability in UK production stems increasing profitability through spatially variable ferti-
from the amalgamation of small fields into larger ones, liser management of wheat and barley crops, respec-
which has taken place since the Second World War tively. Prior to the commencement of this study in 1996,
(Stafford, 2000). published material investigating the variable application
Input management based upon soil, crop and climate of nitrogen was dominated by North American research.
is key to the research detailed in this paper, which began Research into the application of nitrogen fertiliser in the
in 1996 with the objective of evaluating variable rate UK was focused on the adaptation of uniform rate
nitrogen application strategies in a farming situation in guidelines to increase nitrogen use efficiency. The
the UK (Peters et al., 1999). Field trials were conducted requirement for more targeted application of nitrogen

1537-5110/03/$30.00 467 # 2003 Silsoe Research Institute. All rights reserved

Published by Elsevier Science Ltd
468 I.T. JAMES; R.J. GODWIN

Notation

Ai area of management unit i, ha PG price of grain, £ t
1

a, b, c regression coefficients PN price of nitrogen (as element N), £ kg
1
MG gross margin per unit area, £ ha
1 R2 Coefficient of determination
N app
nitrogen application rate, kg[N] ha
1 RP price ratio of nitrogen to grain
Niapp
nitrogen application rate of management unit Y yield, t ha
1
i, kg[N] ha
1 YI yield of management unit i, t ha
1
NMAX nitrogen application rate at the calculated Dyv change in volumetric water content over the
maximum yield response, kg[N] ha
1 growing season, %
NMERN nitrogen application rate at the calculated Subscripts
most economic rate of nitrogen, kg[N] ha
1 I Management unit identifier, i 5 1–n

fertiliser was highlighted by Sylvester-Bradley (1993) could be used to generate nitrogen application rate
and in the context of a UK precision agriculture system maps. This was a similar approach to Schnug et al.
by Earl et al. (1996). This research provided a precursor (1993) but cost would be limiting, decision support
to the Home Grown Cereals Authority research project systems would be required and weather prediction
which started after this research, providing an extensive would still be necessary to achieve optimum nitrogen
investigation into similar principles of Precision Agri- fertiliser rates.
culture (Godwin et al., 2003; Welsh et al., 2003a,b; The principal aim of this research was to investigate
Wood et al., 2003). variable rate nitrogen application management strate-
As outlined by Mulla et al. (1992), it was the gies, which could be developed and adopted by farmers
subsequent technological developments that facilitated in a commercial situation. These would use historic yield
the practical implementation of the site-specific manage- information that took into account the quantities,
ment concepts in a precision agriculture system on location and timeliness of fertiliser applications and
commercial farms, ‘with the objective of managing adhered to current best management practices in terms
inputs and cultivations on a sub-field scale, as required of economic and environmental pressures.
by the crop to produce the maximum economic benefit The experiments were designed to achieve this aim
to the farmer, with complementary environmental through the following objectives:
benefit through improved input use efficiency’. The
(1) to establish a pilot experiment to investigate
commercial introduction of real-time yield mapping
experimental methodologies for conducting practi-
combine harvesters resulted in the rapid production of
cal, on-farm experiments with commercially avail-
‘yield maps’, which allowed farmers to make an
able yield-mapping equipment;
assessment of yield variation within the field.
(2) to determine the interactions between soil, water
This technological advance in precision agriculture
and nitrogen in crop growth, using a selection of
was ahead of the agronomic management systems to
detailed field-survey techniques; and
implement the new technology and was more complex
(3) to determine whether historic yield information can
than the single uniform rate application ethos of
be used successfully to manage variable rate
traditional intensive cereal production in the UK. As
application of nitrogen fertiliser to maximise yield
the number of yield mapping and variable rate applica-
and economic output.
tion units sold increased, producers would ask ‘what do
I do with it?’ and adoption rates for precision
agriculture technologies were limited (Daberkow, 1997;
Fountas, 2001). Nutrient removal strategies were devel- 2. Materials and methods
oped for the variable rate application of phosphorous
(P) and potassium (K) fertilisers, and were commercially 2.1. Site details and characterisation
available in the UK from the mid-1990s. It was the
variable rate application of nitrogen, however, that was The field experiments were conducted in the harvest
seen as one of the most important developments in years of 1997, 1998 and 1999. In each year, the crop
precision agriculture, with the potential to offer both grown was winter barley (Hordeum vulgare, cv. Angora).
economic and environmental benefits. The trials were conducted within an 18 ha area of Short
Ferguson et al. (1996) suggested that ‘combinations of Lane field, Gamlingay, Cambridgeshire, UK
yield maps, remotely sensed images and soil sampling’ (TL249530). This field was selected because of the
 SOIL, WATER AND YIELD RELATIONSHIPS IN DEVELOPING STRATEGIES 469

contrasting soil properties, i.e. clay loam and sandy minimal occurrence of the deep sandy loam within the
loam texture, found within the experimental area. The actual trial plots. The range of soil types, allowed direct
trial area had a shallow relief, with a south facing comparisons to be made between the performances of
maximum slope of 1:75. crops on different soils under the same management
The site is located in one of the major cereal growing regime.
regions of the UK. The annual precipitation of this The soil volumetric moisture content of each soil type
region is on average 574 mm with one of the highest was determined using calibrated neutron probe appara-
number of degree days above 108C and one of the tus (Bell, 1976), over a 1 m depth, from March until
longest growing seasons (249 days) in the UK (Smith & harvest in each year. Laboratory analysis and calcula-
Trafford, 1976); the location is ideally suited to the tions showed that the two soils had similar available
mass-production of cereal crops. The field was cropped water capacities, with a difference of only 1% v/v, the
with continuous winter barley from 1993 to 1999. All sandy loam, however, had a greater saturated hydraulic
field operations prior to the first variable rate trials in conductivity (14 mm h
1 c.f. 5 mm h
1 for the clay
1997 were completed on a conventional uniform loam). Daily measurements of rainfall and ambient
application basis, with reliable machinery and skilled temperature were recorded on the farm and records were
operators. available from 1956.
Soil textural surveys of the field were completed in
1997 and 1999. Prompted by an electromagnetic
induction survey (James, 2000; Godwin & Miller, 2.2. Site management
2003), a 24–50 m grid spaced, 12 m depth, core survey
revealed that there were principally two soil types within The principal agricultural practice controlled in the
the trial area, a clay loam and a sandy loam. Three soils experiments was fertiliser application. All other proce-
are shown in Fig. 1, the two sandy loam soils are dures, including cultivations, crop establishment and
combined in this experiment due to their similar textural protection were completed by the farmer according to
classification in the root zone of the crop and the good agricultural practice on a whole field basis. Crop
harvesting of the experimental area was undertaken by
the authors according to the processes described below.
The application of nitrogen fertiliser, as ammonium
nitrate (345% [N] m/m), was completed using a Nodet
DP12, 12 m boom pneumatic spreader. The coefficient
of variation of the distribution along the boom was
Trial plot location 12%, compared to coefficients of variation in excess of
20% for spinning-disc type spreaders (Culpin, 1992).
The trial plots were intensively sampled for soil mineral
nitrogen, available phosphorus, available potassium,
available magnesium and pH in each year. Phosphate,
potash, magnesium and manganese fertilisers were
applied according to recognised best practice, on a
uniform field basis.
Plot harvesting was completed with a combine
harvester equipped with a global positioning system
(GPS) linked to a radiometric yield monitor, which was
calibrated using a weighing trailer with errors in the
N measurement of static mass of less than 002% m/m
(Godwin et al., 1999). During the harvesting process,
careful consideration was given to reducing the yield
mapping errors identified by Blackmore and Marshall
100 0 100 200 300 m
(1996), principally through operator practice and data
correction. The absolute grain mass for each strip was
measured with the weighing trailer and compared to the
Fig. 1. Map of the trial plot location, including the soil grain mass data recorded by the yield monitor. This
classification for the 18 ha research site; & clay loam over
loamy sand; & deep sandy loam and sandy loam over sand; comparison revealed a measurement error of 11% in
the deep sandy loam and sandy loam over sand are combined as the yield monitor measurement of mass, over the
sandy loam in this experiment complete trial area. The magnitude of error was similar
470 I.T. JAMES; R.J. GODWIN

to the results determined from analysis of the radio- identified as the management decision units for the
metric system by Blackmore et al. (2003). Measurement experiment, which was used as a pilot to the main body
of yield data on the combine was subject to autocorrela- of the work reported by Welsh et al. (2003a, 2003b).
tion from the movement of the combine. Lark et al. Yield measurements were calculated for each soil type
(1997) measured the spatial resolution of a capacitance and the variable rate strategies were based upon the soil
yield monitor to be 20–25 m. Analysis by James (2000) textural variation. Each strip was 6 m wide and 500 m
calculated that with the radiometric yield monitoring long, in a direction parallel to the field tramlines and
system and the forward speed used in this experiment, across each of the soil types, see Fig. 2.
spatial resolution was 6–8 m. The corrected yield data The aim of using the strip-based approach was to
were plotted, strip-wise, in a geographical information investigate and measure the effect of soil variation on the
system and then sampled along the strip using a 10 m yield response to applied nitrogen, not to remove the
spaced grid to ensure isolated point measurements of effect of within-field soil variation by traditional
yield within each strip. randomised ANOVA design (Fisher & Yates, 1963).
Spatial replication was not possible as it would be very
2.3. Trial design difficult to locate zones for replication with exactly the
same soil properties due to changes in soil texture and
A strip design was used in this experiment to apply depth. The large spatial soil variation in this field is
different rates of nitrogen over the different soil types illustrated in Godwin and Miller (2003). The errors

1997

Application rate, kg[N] ha-1

0
50
100
120 1998
125
140 N
150
160
170
175
200
250
100 0 100 200 m

HY1 1999

HY2
MERN 98
MERN 97

Fig. 2. Nitrogen application rates in kg[N] ha
1 for the 1997, 1998 and 1999 experiments; outer plots are experimental controls in
each year; scale bar refers to plot diagrams
 SOIL, WATER AND YIELD RELATIONSHIPS IN DEVELOPING STRATEGIES 471

associated with this experiment are therefore limited to (1) that by applying the historically most economic rate
the absolute accuracy of the instrumentation. The of nitrogen (MERN) to each soil the most
variation in crop performance within any given experi- economically successful yield response would be
mental unit could, however, be determined by consider- produced for both soil types}the MERN 97 and
ing the 95% confidence interval of discreet yield MERN 98 strategies; and
measurements from each treatment plot as described (2) that if nitrogen application rates were to average
above. 150 kg[N] ha
1 over both management units, a yield
The design included two experiments, the first to benefit could be created by applying more nitrogen
develop yield response curves to 50 kg[N] ha
1 incre- to the highest yielding soil and less to the lower
ments of applied nitrogen fertiliser and the second to yielding soil}the historic yield (HY1 and HY2)
test variable rate application strategies derived from the strategies.
yield response curves.
These strategies were designed to provide an answer
2.3.1. Standard application rate experiment to the common question, ‘once within-field variability in
Nitrogen fertiliser was applied with the calibrated yield potential is identified, should more or less nitrogen
pneumatic spreader in combinations of 50 kg[N] ha
1 be applied to the good areas of the field and the opposite
doses to apply the total amounts indicated in Table 1. In to the poor?’ (Welsh et al., 2003a, 2003b). Two
1997, the application rates ranged from 50 to hypotheses, HY1 and HY2, were tested: that maximum
200 kg[N] ha
1; these were increased in 1998 from 100 economic output is obtained by applying more nitrogen
to 250 kg[N] ha
1. In 1999, nitrogen fertiliser was to the historically highest yielding soil and less to the
omitted from one strip, the ‘zero strip’ in order to lowest yielding soil, by concentrating inputs on the areas
extend the nitrogen response curves and to evaluate the of highest yield potential (HY1); and secondly that
effect of residual nitrogen on crop yield. This meant that maximum economic output is achieved by applying
in 1999 a complete range from 0 to 250 kg[N] ha
1 was more nitrogen to the historically lowest yielding soil and
applied. less to the highest yielding soil such that the yield
potential of the lowest yielding soil is increased by
2.3.2. Variable application rate experiment fertiliser input and the potential of the highest yielding
Two variable rate experiments were introduced in soil is relied upon to provide yield with less input (HY2).
1999 with the objective of examining the yield response The MERN value for each soil type was calculated by
to different application rates for the two soil types. equating the first derivative of the quadratic regression
Following the 1998 experiment the relationship between estimate of the yield response function to the price ratio
yield response and soil moisture was identified. Based of nitrogen to grain (Kachanoski et al., 1996). The
upon this observation there were two hypotheses: quadratic yield response model assumes that yield Y in

Table 1
Nitrogen application rates (kg[N] ha
1) for 1997, 1998 and 1999
Strategy Nitrogen application rate, kg ha
1

Application dates
1997 1998 1999
14 March, 8 April 1 March, 30 March 23 March, 17 April

Standard rate trials

0
50 50
100 100 100
150 150 150
200 200 200
250 250
Variable rate trials (clay loam, sandy loam)
HY1 175,125
HY2 125,175
MERN 97 170,120
MERN 98 120,160
Note: HY, historic yield; MERN, most economic rate of nitrogen.
472 I.T. JAMES; R.J. GODWIN

t ha
1 is related to applied nitrogen Napp in kg[N] ha
1 3. Results and discussion
by the following formula:
Measured yield data were compared with rainfall and
Y 5 a þ bN app
cðN app Þ2 ð1Þ soil moisture data to determine the relationship among
where a, b and c are regression coefficients. crop yield responses to applied nitrogen fertiliser, for a
The first-order differential can be described as crop grown on two different soil types, over three
different seasons.
dY
5 b
2cN app ð2Þ
dN app
The nitrogen rate at the maximum of the yield response 3.1. Crop yield data}standard application rate
curve NMAX in kg[N] ha
1 is derived by equating the experiment
differential to zero:
dY The mean yield value (Table 2) for each soil type and
5 b
2cN app 5 0 ð3Þ nitrogen rate combination was calculated and plotted as
dN app
a yield response curve to applied nitrogen, Fig. 3; the
b 95% confidence interval was calculated for a qualitative
NMAX 5 ð4Þ comparison of mean yield values.
2c
The 1997 results were characterised by an increasing
The most economic rate of nitrogen (MERN) is yield response to nitrogen application from 50 to
identified when the differential is equated to the price 200 kg[N] ha
1. It is possible that even higher applica-
ratio RP, which is the price of nitrogen PN in £ kg
1 tion rates would have provided even greater yield return.
divided by the price of grain PG in £ t
1: In 1997, the clay loam soil yielded between 06 and
dY PN 09 t ha
1 higher than the sandy loam soil; this difference
app
5 b
2cN app 5 RP 5 ð5Þ was significant at all application rates except
dN PG
150 kg[N] ha
1.
b
RP In 1998, the mean yield on the sandy loam soil was
NMERN 5 ð6Þ
2c significantly higher than on the clay loam soil. The clay
loam mean yield values were between 07 and 23 t ha
1
where NMERN is the most economic rate of nitrogen
less in 1998 than 1997 and between 10 t ha
1 less and
application in kg[N] ha
1.
0.3 t ha
1 greater on the sandy loam soil. In 1998 there
The historic yield strategies HY1 and HY2 were based
was a relatively flat response to increased nitrogen
on experience of yield response patterns from the 1997
application; yield depression at 200 and 250 kg[N] ha
1
and 1998 trials. The application was split over two dates,
was caused by lodging of over extended plants,
at the time of the second application a decision was
accelerated in their growth and development by a mild
made as to whether soil moisture would be in excess or
winter, that were severely weather damaged in the later
shortage, dependent upon rainfall up to that date. The
rainfall of 1998.
two hypotheses were complicated by the observation
The yield response of 1999 was similar to 1997 in form
that in a high rainfall year the sandy loam soil yielded
but was between 04 and 17 t ha
1 lower on the clay
higher than the clay loam soil and vice versa. To
loam and 09–16 t ha
1 lower on the sandy loam. The
overcome this in the 1999 experiment and to test the two
result of not applying nitrogen in the ‘zero strip’ was to
hypotheses, four strategies were included:
(1) ‘MERN 97’}the most economic rate of nitrogen
Table 2
from the 1997 uniform rate trial; Yield results in t ha
1 for the standard rate trials of 1997 to 1999
(2) ‘MERN 98’}the most economic rate of nitrogen
from the 1998 uniform rate trial; Application rate, kg[N] ha
1 Mean yield, t ha
1
(3) ‘HY1’}applying 175 kg[N] ha
1 to the clay loam Clay loam Sandy loam
soil and 125 kg[N] ha
1 to the sandy loam soil; and
(4) ‘HY2’}applying 125 kg[N] ha
1 to the clay loam 1997 1998 1999 1997 1998 1999
soil and 175 kg[N] ha
1 to the sandy loam soil. 0 27 27
50 61 48 55 44
In HY1 and HY2, the uniform application rate of 100 64 57 55 55 58 51
150 kg[N] ha
1 was selected to be the standard because 150 72 56 59 66 62 54
there was no significant yield benefit above this 200 78 55 62 70 60 58
250 51 65 55 59
application rate in the 1997 or 1998 experiments.
 SOIL, WATER AND YIELD RELATIONSHIPS IN DEVELOPING STRATEGIES 473

9 9 9

8 8 8

7 7 7
Mean yield Y, t ha-1

6 6 6

5 5 5

4 4 4

3 3 3

2 2 2
0 50 100 150 200 250 300 0 50 100 150 200 250 300 0 50 100 150 200 250 300
(a) N app, kg[N] ha-1 (b) N app, kg[N] ha-1 (c) N app, kg[N] ha-1

Fig. 3. Mean yield (Y), t ha
1 response to applied nitrogen fertiliser (Napp), kg[N] ha
1 for the clay loam soil (^) and the sandy
loam soil (&) in (a) 1997; (b) 1998 and (c) 1999; error bars represent the 95% confidence interval; curves are the negative
quadratic regression models for the clay loam ( ) and sandy loam soils ( )

Table 3
The calculated most economic rate of nitrogen (MERN) and the theoretical application rate for maximum yield response (NMAX)
from the 1997, 1998 and 1999 standard application experiments, for the clay loam (Cl) and sandy loam (Sl).
Year Soil Quadratic yield response function [see Eqn (1)] R2 value MERN, kg[N] ha
1 NMAX, kg[N] ha
1

1997 Cl Y 5 5:20 þ 0:014N app
ð7:0  10
6 ÞðN app Þ2 096 102 75
Sl Y 5 4:44 þ 0:015N app
ð1:0  10
5 ÞðN app Þ2 086 765 573
1998 Cl Y 5 5:32 þ 0:007N app
ð3:0  10
5 ÞðN app Þ2 098 48 112
Sl Y 5 3:79 þ 0:029N app
ð9:0  10
5 ÞðN app Þ2 098 141 163
1999 Cl Y 5 2:95 þ 0:032N app
ð8:0  10
5 ÞðN app Þ2 096 179 203
Sl Y 5 2:87 þ 0:028N app
ð7:0  10
5 ÞðN app Þ2 098 173 201
Note: MERN values are calculated at a price ratio RP of 384  10
3, values in brackets are outside the range of applied
nitrogen; R2, coefficient of determination. Y, yield t ha
1; Napp, nitrogen application rate in kg[N] ha
1.

reduce yield to 27 t ha
1 from both soils, 32 to than those applied in the experiments, however in each
38 t ha
1 less than the maximum yield of 1999. As in year (except for 1999 at 250 kg[N] ha
1 on the clay
1997, the 1999 yield response curves were not limited by loam), yields from applications above 150 kg[N] ha
1
lodging of the crop. Results showed that from the three were not significantly higher than at 150 kg[N] ha
1
experiment years, there were three different yield when the 95% confidence interval is considered.
response patterns, a complicating factor in trying to Regression analysis of the mean yield data showed
determine spatially variable application strategies for that the negative quadratic model, Eqn (1), was more
nitrogen, based upon historic yield response. The fact suited to the 1998 and 1999 yield response data than in
that similar results have been found in the barley trials 1997, Table 3. This was because of the nature of the
of the HGCA project at Cranfield University (Welsh 1997 data which did not show a maximum ‘peak’
et al., 2003a, 2003b), in the silty clay loam over chalk within the range of applied nitrogen, as a consequence
soils of the Andover and Panholes series in Hampshire, the 1997 NMERN and NMAX values were outside the
supports these findings. 50–200 kg[N] ha
1 application range.
In each year, the maximum yield for each soil type In 1998, the NMAX values are 112 kg[N] ha
1 for
occurred at the same nitrogen application rate, within the clay loam and 163 kg[N] ha
1 for the sandy
the range applied: 200 kg[N] ha
1 in 1997; loam, indicating a theoretical opportunity for targeting
150 kg[N] ha in 1998 and 250 kg[N] ha
1 in 1999. In

1
100–112 kg[N] ha
1 to the clay loam and 141–
1997 and 1999 it is possible that the true maximum yield 163 kg[N] ha
1 to the sandy loam. This would be
could have been obtained at an application rate higher counter-intuitive following the 1997 results, however,
474 I.T. JAMES; R.J. GODWIN

and would require experience and data from previous standard. The lower yield of the HY1, HY2 and MERN
wet seasons. In 1999, there is a difference of only 97 strategies are also illustrated, with all mean yields
2 kg[N] ha
1 in NMAX and 6 kg[N] ha
1 in NMERN. As a falling on or marginally below the 1999 yield response
consequence, spatially variable application of nitrogen is curve for both soils.
not required, it is more important to ensure the
economic optimum amount of nitrogen is applied.
3.3. Rainfall and soil moisture data

3.2. Crop yield data}variable application rate The rainfall totals for 1993–1999 are shown in Fig. 5.
experiment The rainfall data for each year are split into two, 6-
month periods, August to January and February to
The mean yield for each variable rate strategy is July. The relationship between yield and rainfall was
detailed in Table 4, in this table the yield performance of demonstrated in the trials of 1997–1999. When rainfall
each strategy is compared with the standard uniform in February to July was low (less than 250 mm), the clay
application rate of 150 kg[N] ha
1. The mean yield loam soil yielded higher than the sandy loam over loamy
values from each of the variable rate strategies are sand soil, when rainfall in February to July was high the
compared with the uniform application response func- opposite was true.
tions in Fig. 4. All of the variable rate strategies are The relationship between split rainfall totals and
within the 95% confidence interval of the standard mean yield for each soil type from 1993 to 1999 is shown
rate means and the yield response functions, except for in Fig. 6. The mean yield values calculated are those of
the MERN 98 yield on the sandy loam soil. As a result, the whole field area and not just the trial plots, the yield
the MERN 98 strategy yielded higher than the uniform values shown were all corrected for systematic yield

Table 4
Variable application rate trial results compared with the 150 kg[N] ha
1 standard trial mean yield
Trial Application rate, kg[N] ha
1 Yield, t ha
1 Standard trial mean yield, t ha
1

Clay loam Sandy loam Mean Clay loam Sandy loam Mean

HY 1 175 125 150 53 50 52 56

HY 2 125 175 150 56 53 55 56
MERN 97 170 120 145 55 49 52 56
MERN 98 120 160 140 61 60 61 56
Note: HY, historic yield; MERN, most economic rate of nitrogen.

9 9

8 8
MERN 98 MERN 98
MERN 97
7 7
Mean yield Y, t ha-1

6 6

5 5
HY 2 HY 1 HY 2
4 4
HY 1
MERN 97
3 3

2 2
0 100 200 300 0 100 200 300
(a) N app, kg[N] ha-1 (b) N app, kg[N] ha-1

Fig. 4. Yield (Y), t ha
1 response to applied nitrogen (Napp), kg[N] ha
1 for (a) clay loam; (b) sandy loam. Including the variable
rate application means (n), the 1997 standard means (^), the 1998 standard means (&) and the 1999 standard means (  ); error
bars represent the 95% confidence interval. Curves are the negative quadratic regression models for 1997 (}), 1998 (. . . . .) and
1999 (- -); MERN, most economic rate of nitrogen; HY, historic yield
 SOIL, WATER AND YIELD RELATIONSHIPS IN DEVELOPING STRATEGIES 475

800

700

600

Rainfall, mm
500

400

300

200

100

0
1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999
Year

Fig. 5. Total rainfall for August to July, 1985–1999; each total is split into August to January & and February to July ; the 15
year mean for August to January is 300 mm, February to July, 232 mm and total rainfall, 532 mm

mapping errors and all differences shown are significant The rainfall return periods of each split, calculated
at the 95% confidence interval. The graph shows that using the method of Chow (1951), are presented in
the relationship between February to July rainfall and Table 5. The probability calculations show the context
the mean yields of the two soils determined from the of the split 6-month rainfall totals in terms of the 15-yr
three trial years also extends to the mean yield values of period of 1985–1999. The results also highlight the
1993–1996. In 1994, 1995, 1997 and 1999 the difference difficulty of predicting rainfall because the period
in yield between the two soils was less than or equal to rainfall totals of 1997–1999 range from 488 to 581 mm
01 t ha
1. In 1996, the driest of the 7 years investigated, but the rainfall return periods of these totals only range
the clay loam yielded 11 t ha
1 more than the sandy from 18 to 28 yr.These differences in rainfall cause
loam soil. In 1993 and 1998, with the two wettest differences in yield response between the two soils, yet it
February to July periods, the sandy loam soil yielded 07 is possible that within a 3-yr period, either the minimum
and 04 t ha
1 more than the clay loam, respectively. It is or the maximum rainfall can occur. This is characteristic
the February to July rainfall, not the total rainfall that of UK rainfall patterns within this 15-yr period resulting
was critical in determining which soil yielded the in great difficulty in rainfall prediction, which based
highest. When rainfall in this period was low the crop upon the yield response information above makes it very
growing in the sandy loam soil was drought stressed and difficult to predict which of the two soils will provide the
yields were lower than the clay loam soil. When rainfall greatest yield.
in this period was greater than approximately 225 mm, An explanation for the relationship between the yield
the clay loam became water-logged and anaerobic and response of each soil and rainfall is illustrated by the
the sandy loam soil provided greater crop yield. This different volumetric moisture content patterns in Fig. 7.
field trend is observed at the plot level too. Each graph shows a different pattern: the 1997 curve
The total rainfall in 1994 was also high but the clay [Fig. 7(a)] was quite different from the 1998 and 1999
loam yield was marginally greater than the sandy loam curves [Fig. 7(b) and (c)] which were similar to each
yield, this was because the rainfall in the second period other. The clay loam curve in 1997 showed extraction at
was low, as was the case in 1999 when yields were all depths, which was not seen in 1998 or 1999, where
similarly close to each other. This relationship was the extraction at 02 and 04 m predominated. This was a
principal driver in the temporal instability of yield, function of greater water availability from rainfall in
however, the graph shows that neither soil is more 1998 and 1999, as discussed above.
temporally stable than the other, which is confirmed by In 1998 and 1999, the volumes of water extracted were
analysis of the temporal instability of yield in relation to similar from 02 to 06 m, only differing greatly at 08
soil type (James, 2000). The rainfall in the second period and 10 m. The sandy loam curve in 1997 was different
was shown to be critical to the yield response, however, from the clay loam curve and extraction was greatest at
most of this rainfall followed nitrogen application, 06 m; in fact, extraction was greater at all depths than at
complicating nitrogen fertiliser recommendations. 02 m, a quite different behaviour from 1998 and 1999.
476 I.T. JAMES; R.J. GODWIN

450 8 observed; these findings are similar to those of Braum

400 7 et al. (1999).
350 6
300
Rainfall, mm

Yield, t/ha
5
250 3.4. Economic analysis
4
200
150
3 Economic analyses of the trial results were required
100 2 because the ultimate measure of the success of variable
50 1 rate application methods on UK farms is whether or not
0 0
they provide a greater economic return to the farmer
(a) 1993 1994 1995 1996 1997 1998 1999
than uniform strategies, not necessarily through im-
proved yields but also through reduced input costs.
350 8 There are two approaches taken in variable rate
7 application, the first is to achieve the economically
300
6
optimum yield from unlimited inputs and the second to
250 achieve the optimum yield with limited inputs, for
Rainfall, mm

Yield, t/ha
5
200 example when environmental restrictions on the use of
4
150
nitrogen apply. The principal question to answer was
3 whether the economic benefits of variable rate applica-
100 2 tion are greater than the uniform application strategies
50 1 used by the majority of farmers in the UK.
0
Economic benefits can be gained from even the most
0
(b) 1993 1994 1995 1996 1997 1998 1999 simple of precision farming systems. Even without variable
rate application strategies, yield maps can provide
700 8 invaluable information for producers to help plan tillage
600 7 operations and crop protection. Accurate yield maps
6
illustrate variation in yield which cannot be measured as
500 easily or as accurately in any other way. The yield map
Rainfall, mm

Yield, t/ha

5
400 information can be used to investigate and identify the
4 cause of differences in yield performance and variability.
300
3 There is an additional cost from implementing a
200 2 precision farming system which is dependent upon the
100 1 scale of technology used. This cost comprises the
additional machinery required to perform tasks such
0 0
as yield mapping and variable rate application and the
1993 1994 1995 1996 1997 1998 1999
(c) Year
additional management time requirement. Several esti-
mates of this cost, over a precision farmed area of
Fig. 6. The relationship between mean yield in 1993–1999 and 250 ha, have been calculated: Earl et al. (1996) estimated
(a) total rainfall , (b) August to January rainfall and (c) the cost to be £1086 ha
1; later Godwin et al. (2003)
February to July rainfall &; yield values are for the clay loam calculated values of £5–£18 ha
1 for four systems from a
soil (^) and the sandy loam soil (n); all differences in yield are
significant at the 95% confidence interval
basic non-differential GPS system to a fully integrated,
manufacturer installed system. They calculated that
these costs would reduce to £125 and £450 ha
1 for a
The extraction of larger volumes from depth in 1997 typical precision farming operation over 1000 ha.
was a function of water availability to the plant in a Commodity pricing is an important factor in the
dry summer following a dry winter, where deeper economic analysis of nitrogen fertiliser application. The
rooting was necessary in the sandy loam. The relation- gross margin per unit area MG in £ ha
1 relationship
ship between rainfall and soil moisture as a function of with yield response Y is a function of the grain revenue
soil texture has been shown. There is an association less the cost of the nitrogen input for each management
between this relationship and the yield variation unit, divided by the total area, as described in

A1 ððY1  PG Þ
ðN1app  PN ÞÞ þ    þ An ððYn  PG Þ
ðNnapp  PN ÞÞ

MG 5 Pn ð7Þ
i 5 1 Ai
 SOIL, WATER AND YIELD RELATIONSHIPS IN DEVELOPING STRATEGIES 477

Table 5
Calculated rainfall return periods for the growing seasons of 1997–1999, based upon 15 years of rainfall data for the site
Year Rainfall, mm Return period, yr

August to January February to July Total August to January February to July Total

1997 272 216 488 18 18 18

1998 258 323 581 16 40 28
1999 392 160 552 40 14 23

v, %
v, %
v, %

-20 -16 -12 -8 -4 0 -20 -16 -12 -8 -4 0 -20 -16 -12 -8 -4 0
0.0 0.0 0.0

-0.2 -0.2 -0.2

Depth, m

Depth, m

Depth, m
-0.4 -0.4 -0.4

-0.6 -0.6 -0.6

-0.8 -0.8 -0.8

-1.0 -1.0 -1.0

(a) (b) (c)

Fig. 7. Change in soil volumetric moisture content (Dyv), % v/v throughout the growing season change graphs for the clay loam soil
(^) and the sandy loam soil (n) in (a) 1997; (b) 1998 and (c) 1999

where Ai and Yi are the area in ha and the yield in t ha
1 each year, on both soils, the maximum economic benefit
for each management unit i. from each soil occurred at the same application rate as
Analysis by James (2000) showed that gross margin the maximum observed yield, except for the sandy loam
analysis is highly sensitive to grain price and signifi- soil in 1999, where the relatively small increase in yield
cantly less sensitive to changes in nitrogen price. For the from 200 to 250 kg[N] ha
1 did not justify the applica-
calculated MERN of each soil to be significantly tion of the additional nitrogen. Over the 3-yr period,
different, the price ratio of nitrogen to grain would and both soils, there is only a £3 ha
1 difference between
have to increase from the 1999 value of 38  10
3 to the mean economic return from 150 and 200 kg[N] ha
1.
27  10
3. In terms of the 1999 yield response patterns This suggests that there is an economic and environ-
and prices (£028 kg
1 of N and £73 t
1 of barley), a mental benefit from using 150 kg[N] ha
1 rather than
break even ratio of 27 is equivalent to either a rise in the 200 kg[N] ha
1; however, the producer is more likely to
cost of nitrogen to £200 kg
1 or a drop in the value of apply 200 kg[N] ha
1 to minimise risk to ensure that the
grain to £1050 t
1. The effect of the insensitivity of the maximum yield is obtained, given the sensitivity of the
gross margin to nitrogen price is compounded by the gross margin analysis to yield output.
large agronomic yield benefit from each extra unit of The mean gross margins of the 150 and
nitrogen applied. The sensitivity to the relative area of 200 kg[N] ha
1 uniform rate strategies were compared
each management unit was also important but depended with the four variable rate application strategies,
upon the yield response of each area and the number of Table 7. None of the four historic yield strategies
areas considered as management units. provided an economic benefit over either of the 150 and
The simple calculation above, Eqn (7), was used to 200 kg[N] ha
1 uniform applications, with losses ranging
calculate the gross margin for the yield from each from £5 to £76 ha
1. Once the specific costs of precision
standard application rate on each soil type, Table 6. The agricultural systems are included the losses are increased
prices used in the calculation were the average over the further. The results of this experiment show that historic
three 3-yr period, the price of nitrogen was yield crop management strategies are not suitable for the
£028 kg[N]
1 and the price of grain was £73 t
1. In variable rate application of nitrogen fertiliser.
478 I.T. JAMES; R.J. GODWIN

Table 6
Gross margin calculations for the standard rate trials
Application rate, kg[N] ha
1 Gross margin MG, £ ha
1

Clay loam Sandy loam

Overall mean
1997 1998 1999 Mean 1997 1998 1999 Mean

0 197 197 197 197 197

50 431 336 384 388 307 347 366
100 439 388 374 400 374 395 344 371 386
150 484 367 389 413 440 411 352 401 407
200 513 346 397 419 455 382 367 401 410
250 302 405 353 332 361 346 350
Note: Calculations are based upon a nitrogen price of £028 kg[N]
1 and a grain value of £73 t
1; gross margin calculations
assume that the area of each soil is equal

Table 7
Gross margin calculations for the variable rate trials in comparison with standard rate overall means
Gross margin MG, £ ha
1 Difference in MG, £ ha
1 of variable versus. standard rate
Strategy
Clay Sandy Standard rate, kg[N]/ha
1
Loam Loam 150 200

HY 1 338 330
73
76

HY 2 374 338
51
54
MERN 97 354 324
68
71
MERN 98 412 393
5
8
Note: calculations are based upon a nitrogen price of £028 kg[N]
1 and a grain value of £73 t
1; gross margin calculations
assume that the area of each soil is equal and the cost of variable rate application has not been considered. HY, historic yield;
MERN, most economic rate of nitrogen.

Research investigating real-time approaches to cano- more expensive inputs such as potash, phosphate and
py management in cereals (Wood et al., 2003) has some herbicides provide a greater potential for this
illustrated a more complex relationship between crop approach.
nitrogen requirements within a field and identified a
more effective approach to increasing nitrogen fertiliser
efficiency. This approach has produced a yield benefit 4. Conclusions
which translated into a benefit of £20–£60 ha
1, exclud-
ing the additional cost of the mapping system. There is a requirement for increased nitrogen use
There is a requirement for increased nitrogen fertiliser efficiency for both economic and environmental sustain-
use efficiency for both economic and environmental ability. This experiment was conducted to establish
reasons. Economically, it is important for UK producers methodologies for developing strategies for improved
to reduce costs when global grain values are relatively nitrogen efficiency in a precision farming system. The
low. Environmentally, increased efficiency should result results of this experiment related to one field, in three
in lower pollution risk without excessive restriction on seasons, with two principal soil types, and as such give
fertiliser use, which would limit yield output. The an indication of the effect of soil spatial variability on
potential for targeting nitrogen resources increases when winter barley response to applied nitrogen.
the cost of nitrogen is increased by tighter environ- Based upon historic yield information, there was no
mental legislation. simple variable rate application strategy that provided a
The impact of the findings of this research is yield or economic benefit compared with a 150 or
important to the development of precision agriculture 200 kg[N] ha
1 uniform application of nitrogen fertiliser,
in the UK. The results show that the dynamic nitrogen the exact rate to be dependent upon the residual soil
cycle and relatively ‘cheap’ nitrogen mean that justifica- nitrogen. Achieving the most economic application rate
tion for the variable rate application of nitrogen is critical to maximising economic output. It is very
based upon historic yield is difficult. More stable or difficult to predict the most economic rate of nitrogen
 SOIL, WATER AND YIELD RELATIONSHIPS IN DEVELOPING STRATEGIES 479

(MERN) for a future season based upon historic yield Braum S M; Hinds P; Malzer G L; Bell J; Mulla D; Robert P C
information alone. (1999). Terrain attributes and soil nitrogen: spatial effects on
In all 3-yr, there were significant differences between corn yield responses to nitrogen fertilization for a northern,
glaciated landscape. In: Proceedings of the Fourth Interna-
the yield of the clay loam and the sandy loam soil. When tional Conference on Precision Agriculture (Robert P C;
rainfall in the February to July period of the growing Rust R H; Larson W E, eds), St Paul, MN, 19–22 July, 1998,
season was high, the sandy loam soil yielded signifi- pp 195–202. ASA, CSSA, SSSA, Madison, WI, USA
cantly higher than the clay loam soil. When rainfall in Carr P M; Carlson G R; Jacobsen J S; Nielson G A; Skogley E
the February to July period was low, the clay loam soil O (1991). Farming soils, not fields: a strategy for increasing
fertiliser profitability. Journal of Production Agriculture,
yielded significantly higher than the sandy loam soil. 4(1), 57–61
Despite the differences in mean yield between the clay Chow V T (1951). General formula for hydrologic frequency
loam and the sandy loam soils, the maximum yield of analysis. American Geophysical Union Transactions, 32,
the response curve for each soil type occurred at the 231–237
same application rate in each growing season. Given Culpin C (1992). Farm Machinery (12th Edn). Blackwell
Scientific Publications, Oxford
that in each growing season the maximum yield
Daberkow S G (1997). Adoption rates for recommended crop
response occurred at the same application rate for each management practices: implications for precision farming.
soil, neither historic yield or ‘real-time’ strategies would In: Proceedings of the First European Conference on
provide a basis for applying different rates of nitrogen to Precision Agriculture (Stafford J V, ed), pp 941–948. BIOS
the different soils. Scientific Publishers, Oxford, UK
This pilot experiment has illustrated the potential and Dampney P M R; Moore M (1999). Precision agriculture in
England: current practice and research-based advice to
the difficulties in conducting on-farm experimentation farmers. In: Proceedings of the Fourth International
for investigating variable rate nitrogen application in the Conference on Precision Agriculture (Robert P C; Rust R
UK. Using a strip-based trial design and commercial H; Larson W E, eds), St Paul, MN, 19–22 July, 1998, pp
farm equipment it was possible to determine yield 661–674. ASA, CSSA, SSSA, Madison, WI, USA
response functions to applied nitrogen; however, future Earl R; Wheeler P N; Blackmore B S; Godwin R J (1996).
Precision farming}the management of variability. Land-
research must consider enhanced control within the wards, Journal of the Institution of Agricultural Engineers,
experiment design. The variation in yield response and 51(4), 18–23
weather patterns seen in just one field, over 3 yr Ferguson R B; Gotway C A; Hergert G W; Peterson T A (1996).
illustrates the difficulties in estimating the benefit of Soil sampling for site-specific management. In: Proceedings
site-specific application of nitrogen where other sources of the Third International Conference on Site-Specific
Management for Agricultural Systems, 1996 (Robert P C;
of variation could be different from this location.
Rust R H; Larson W E eds), pp 13–22, Bloomington/
Minneapolis, MN. June 23–26, 1996. ASA, CSSA, SSSA,
Madison, WI, USA
Acknowledgements Ferguson R B; Hergert G W; Schepers J S; Crawford C A
(1999). Site-specific nitrogen management of irrigated corn.
In: Proceedings of the Fourth International Conference on
The research was funded by DEFRA, AGCO (UK) Precision Agriculture (Robert P C; Rust R H; Larson W E,
Ltd. and in part by The Chadacre Trust. Eayrs Brothers eds), St Paul, MN, 19–22 July, 1998, pp 733–744. ASA,
of Cambridgeshire are thanked for the use of their fields, CSSA, SSSA, Madison, WI, USA
without which this research would not have been Fisher R A; Yates F (1963). Statistical Tables for Biological,
possible. Mark Moore and Robert Walker are gratefully Agricultural, and Medical Research (6th Edn). Oliver and
Boyd, Edinburgh
acknowledged for their technical help in the field. Fountas S (2001). Farmers’ attitudes to precision farming. In:
Proceedings of the third European Conference on Precision
Agriculture (Grenier G; Blackmore B S, eds), Montpellier,
France, June 18–20, 2001, pp 515–519. Agro Montpellier,
References France
Godwin R J; Miller P C H (2003). A review of the technologies
Bell J P (1976). Neutron probe practice. Report No.19, for mapping within-field variability. Biosystems Engineering
Institute of Hydrology, Wallingford, UK (Special Edition on Precision Agriculture), 84(4), doi:
Blackmore B S; Fountas S; Moore M; Godwin R J (2003). 10.1016/S1537-5110(02)00283-0, this issue
Spatial trends in yield maps. Biosystems Engineering Godwin R J; Richards T E; Wood G A; Welsh J P; Knight S
(Special Edition on Precision Agriculture), 84(4), this issue (2003). An economic analysis of the potential for precision
Blackmore B S; Marshall C J (1996). Yield mapping; errors farming in UK cereal production. Biosystems Engineering
and algorithms. In: Proceedings of the Third International (Special Edition on Precision Agriculture), 84(4), doi:
Conference on Site-Specific Management for Agricultural 10.1016/S1537-5110(02)00282-9
Systems (Robert P C; Rust R H; Larson W E, eds), Godwin R J; Wheeler P N; O’Dogherty M J; Watt C D;
Bloomington/Minneapolis, MN, 23–26 June, 1996, Richards, T E (1999). Cumulative mass determination for
pp 403–416. ASA, CSSA, SSSA, Madison, WI, USA yield maps of non-grain crops. Computers and Electronics
480 I.T. JAMES; R.J. GODWIN

in Agriculture, Special Issue: Spatial Yield Recording of Smith L P; Trafford B D (1976). Climate and Drainage.
Non-Grain Crops, 23(2), 85–101 Technical Bulletin No. 34, Ministry of Agriculture, Fisheries
James I T (2000). Investigation of management strategies for and Food, HMSO, London, ISBN 0112408958
the variable rate application of nitrogen fertiliser to winter Stafford J V (2000). Implementing precision agriculture in the
barley in the UK. PhD Thesis (unpublished), National Soil 21st century. Journal of Agricultural Engineering Research,
Resources Institute, Cranfield University, Silsoe, UK 76, 267–275
Kachanoski R G; Fairchild G L; Beauchamp E G (1996). Yield Sylvester-Bradley R (1993). Scope for more efficient use
indices for corn response to applied fertiliser: application in of fertiliser nitrogen. Soil Use and Management, 9(3),
site-specific crop management. In: Precision Agriculture: 112–117
Proceedings of the third International Conference (Robert P Thompson W H; Robert P C (1995). Evaluation of mapping
C; Rust; Larson W E, eds), pp 425–432, ASA, CSSA, SSSA, strategies for variable rate applications. In: Site-Specific
Madison, WI Management for Agricultural Systems (Robert P C; Rust R
Lark R M; Stafford J V; Bolam H C (1997). Limitations on the H; Larson W E, eds), pp 303–323. ASA, CSSA, SSSA
spatial resolution of yield mapping for combinable crops. Madison, WI
Journal of Agricultural Engineering Research, 66, 183–193 Welsh J P; Wood G A; Godwin R J; Taylor J C; Earl R;
Mulla D J; Bhatti A U; Hammond M W; Benson J A (1992). A Blackmore B S; Knight S (2003a). Developing strategies for
comparison of winter wheat yield and quality under uniform spatially variable nitrogen application in cereals, Part I:
versus spatially variable fertiliser management. Agriculture, winter barley. Biosystems Engineering (Special Edition on
Ecosystems and Environment, 38, 301–311 Precision Agriculture), 84(4), doi: 10.1016/S1537-
Peters M W; James I T; Earl R; Godwin R J (1999). Nitrogen 5110(03)00002-3, this issue
management strategies for precision farming. In: Precision Welsh J P; Wood G A; Godwin R J; Taylor J C; Earl R;
Agriculture ‘99, Papers presented at the second European Blackmore B S; Knight S (2003b). Developing strategies for
Conference on Precision Agriculture (Stafford J V, ed), spatially variable nitrogen application in cereals, Part II:
Odense, Denmark, 11–15 July, 1999, pp 719–728. SCI, wheat. Biosystems Engineering (Special Edition on Precision
Sheffield Academic Press, Sheffield, UK Agriculture), 84(4), doi: 10.1016/S1537-5110(03)00003-5,
Schnug E; Murphy D; Evans E; Haneklaus S; Lamp J (1993). this issue
Yield mapping and the application of yield maps to computer- Wood G A; Welsh J P; Godwin R J; Taylor J C; Earl R;
aided local resource management. In: Proceedings of the First Knight S (2003). Real-time measures of canopy size as a
Workshop, Soil Specific Crop Management, A workshop on basis for spatially varying nitrogen applications to winter
research and development issues, 1993 (Robert P C; Rust R wheat sown at different seed rates. Biosystems Engineering
H; Larson W E, eds), Minneapolis, MN, 14–16 April, 1992, (Special Edition on Precision Agriculture), 84(4), doi:
pp 87–93. ASA, CSSA, SSSA, Madison, WI, USA 10.1016/S1537-5110(03)00006-0, this issue