You are on page 1of 11

Field Crops Research 118 (2010) 158–168

Contents lists available at ScienceDirect

Field Crops Research


journal homepage: www.elsevier.com/locate/fcr

Estimating maize nutrient uptake requirements


T.D. Setiyono a , D.T. Walters a , K.G. Cassman a , C. Witt b , A. Dobermann c,∗
a
Department of Agronomy and Horticulture, University of Nebraska-Lincoln, PO Box 830915, Lincoln, NE 68583-0915, USA
b
International Plant Nutrition Institute (IPNI) Southeast Asia Program, PO Box 500 GPO, Penang 10670, Malaysia
c
International Rice Research Institute (IRRI), DAPO Box 7777, Metro Manila, Philippines

a r t i c l e i n f o a b s t r a c t

Article history: Generic, robust models are needed for estimating crop nutrient uptake requirements. We quantified
Received 17 February 2010 and modeled grain yield–nutrient uptake relations in maize grown without significant biotic and abiotic
Received in revised form 15 May 2010 stresses. Grain yield and plant nutrient accumulation in above-ground plant dry matter (DM) of commer-
Accepted 16 May 2010
cial maize hybrids were measured at physiological maturity in on-station and on-farm experiments in
Nebraska (USA), Indonesia, and Vietnam during 1997–2006. These data were used to model the nutrient
Keywords:
requirements for yields up to 20 Mg ha−1 using the QUEFTS (QUantitative Evaluation of the Fertility of
Nutrient use efficiency
Tropical Soils) approach. The model required estimation of two boundary lines describing the minimum
Nutrient uptake
Grain nutrient removal
and maximum internal nutrient efficiencies of N, P and K (IE, kg grain per kg nutrient in plant DM), which
Maize were estimated at 40 and 83 kg grain kg−1 N, 225 and 726 kg grain kg−1 P and 29 and 125 kg grain kg−1
QUEFTS K, respectively. The model predicted a linear increase in grain yield if nutrients are taken up in balanced
Crop nutrient requirements amounts of 16.4 kg N, 2.3 kg P and 15.9 kg K per 1000 kg of grain until yield reached about 60–70% of the
Nitrogen yield potential. The corresponding IEs were 61 kg grain kg−1 N, 427 kg grain kg−1 P and 63 kg grain kg−1 K.
Phosphorus The model predicted a decrease in IEs when yield targets approached the yield potential limit. A spher-
Potassium ical model was derived from QUEFTS model outputs and found to be particularly suitable for practical
applications such as estimating fertilizer needs. The proposed spherical model offers generality across
environments and management practices, allowing users to estimate the optimal N, P and K uptake
requirements based on two inputs: estimated yield potential and yield target. Further improvements in
modeling the relationship between N uptake and grain yield can be made by taking into account differ-
ences in harvest index. Accuracy in the simulation of N uptake using the spherical model was improved
from an RMSE of 35 kg N ha−1 to 25 kg N ha−1 when harvest index was accounted for, suggesting that
the relationship between N uptake and actual yield is affected by both yield potential and efficiency in
biomass partitioning.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction Current nutrient management strategies for irrigated maize still


lack a practical but robust and globally applicable tool for esti-
Increasing demand for food, fiber and biofuel has significant mating balanced nutrient requirements for achieving a targeted
impact on fertilizer consumption at both local and global scales. maize yield goal. Most nutrient recommendation algorithms in the
At issue is whether maize fertilizer use can be kept in balance Midwest USA, for example, are state-specific and none of them
for achieving production goals and preserving soil quality and the explicitly deal with estimating N, P or K requirements as a function
environment (Dobermann and Cassman, 2002). Nutrient mining of yield potential, yield target, and nutrient supply from indige-
is a major cause of low crop yields and unsustainable agriculture nous sources (Dobermann and Cassman, 2002). Indigenous sources
in parts of the developing world, particularly Africa. At the other include nutrient supply from soil, available nutrients from wet/dry
extreme, nutrients such as N and P often move beyond the bounds deposition, heterotrophic N2 fixation, and in irrigated systems, the
of the agricultural field because the management practices fail to useful nutrient supply brought in with applied water. Nutrient bud-
achieve adequate congruence between nutrient supply and crop geting approaches or environmental indices for phosphorus (P)
nutrient demand (van Noordwijk and Cadisch, 2002). are increasingly used for guiding fertilizer applications, but they
also rely on robust estimates of crop nutrient removal (per unit
yield) and of nutrient recycling. The situation is often worse in
many developing countries, where blanket recommendations for
∗ Corresponding author. Tel.: +63 2 580 5600. fertilizer use still predominant. Quantitative fertilizer recommen-
E-mail address: a.dobermann@irri.org (A. Dobermann). dation algorithms that account for demand, supply, and efficiency

0378-4290/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.fcr.2010.05.006
T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168 159

terms are well suited to site-specific applications and are applicable


across wider range of environmental and management conditions.
Understanding nutrient efficiency and yield–uptake relations in
maize is thus a crucial step toward developing more robust fertilizer
recommendation algorithms.
Many studies have been conducted to quantify nutrient removal
with maize grain (Heckman et al., 2003), understand nutrient
uptake dynamics under different growth conditions (Karlen et
al., 1988; Walker and Peck, 1974), or measure and model the
decline in critical plant nutrient concentration with increasing
crop biomass accumulation (Greenwood et al., 1990). However,
estimates of nutrient requirements are often derived from field
experiments conducted at relatively few locations, mostly research
stations, whereas only few studies have attempted to model nutri-
ent uptake–yield relationships in maize across widely differing
farming environments. Fig. 1. Box plots of grain yield of maize for different sub-sets of data used in this
Dutch researchers (van Duivenbooden et al., 1996; van Keulen, study: NE, USA-HY (high-yield site at Lincoln, NE, USA), NE, USA-F (on-farm sites in
1977, 1986; van Keulen and Van Heemst, 1982) studied the Nebraska, USA), VN-F (on-farm sites in North Vietnam), and ID-F (on-farm sites in
Indonesia). Solid and dashed lines indicate median and mean, respectively. The box
relationship between yield and plant nutrient accumulation for
boundaries indicate the upper and lower quartiles, the whisker caps indicate 90th
major crops. They found a generic linear range followed by a and 10th percentiles, and the circles indicate the 95th and 5th percentiles.
parabolic–plateau and concluded that an upper boundary exists
at which a nutrient’s concentration in grain (and straw) becomes
diluted to the maximum possible extent when that nutrient is the 2. Material and methods
sole factor-limiting yield. Janssen et al. (1990) and Smaling and
Janssen (1993) further developed these concepts in the QUEFTS 2.1. Data
(QUantitative Evaluation of the Fertility of Tropical Soils) model
using two linear boundaries that described the range from maxi- The database used for this analysis included more than 2500
mum accumulation (“excess”) to maximum dilution (“deficiency”) measurements of maize yield, total dry matter and nutrient uptake
of N, P, and K in maize. Those envelopes were then mathematically collected from various studies conducted by the authors in collab-
combined into linear–parabolic–plateau curves for estimating opti- oration with other partners during 1997–2006. Given the diversity
mal (balanced) nutrient requirements of all three macronutrients. of experimental conditions, specific details of these field studies
The QUEFST approach has been applied in studies on rice in Asia are not reported here because this paper aims at evaluating the
(Witt et al., 1999) and West Africa (Haefele et al., 2003), wheat in relationship between grain yield and plant nutrient accumulation
China (Liu et al., 2006) and India (Pathak et al., 2003), and maize for a wide range of possible nutrient supplies and environmental
in Africa (Janssen et al., 1990; Saidou et al., 2003) and China (Liu et conditions. Methods for plant sampling and nutrient anlaysis, how-
al., 2006). It has also provided a scientifically sound and practically ever, are described below. The database included two contrasting
robust, generic basis for new site-specific nutrient management maize-growing regions, which also differed significantly in maize
concepts in these cereal crops (Dobermann et al., 2002; Khurana et yield potential and actual yields obtained in our studies (Fig. 1):
al., 2008; Witt et al., 2008). Although the assumptions and concepts
used in the QUEFTS model are reasonable, the original purpose of (a) High-yielding maize in Nebraska, USA: multiple locations
this model was to quantify yield for a given nutrient supply from the across Nebraska, representing different environments, soil
soil. Even though QUEFTS can be used for estimating plant uptake types and cropping practices in the Western Corn Belt. Specific
for a certain yield level (Smaling and Janssen, 1993), no straightfor- sub-sets included (i) an on-farm trial on potassium response
ward algorithm is available to solve for uptake at a given attainable conducted on a sandy soil at one location over 2 years (A. Dober-
yield using QUEFTS. In Witt et al. (1999), for example, a solver mod- mann and C. Shapiro, unpublished), (ii) a high-yielding maize
ule in Microsoft® Office Excel was used to calculate nutrient uptake experiment conducted at Lincoln, NE for 6 years (Adviento-
for a given rice yield goal. Borbe et al., 2007), (iii) two large irrigated maize fields and one
Previous studies on nutrient uptake requirements and inter- large favorable rainfed maize field at Mead, NE sampled for 2
nal nutrient efficiencies in maize were mainly conducted for grain years (Verma et al., 2005), and (iv) on-farm nutrient response
yields up to about 10 Mg ha−1 (Janssen et al., 1990; Liu et al., 2006; trials conducted at 11–12 locations across Nebraska for 3 years
van Duivenbooden et al., 1996). Our study involves a large database (Dobermann et al., 2006; Wortmann et al., 2009).
of nutrient uptake from irrigated and favorable rainfed maize envi- (b) Lowland and upland maize in Southeast Asia: two sub-regions,
ronments when crops were grown in well-managed on-farm and representing different environments, soil types and cropping
on-station trials in subtropical and tropical (North Vietnam and practices. Specific sub-sets included (i) on-farm research tri-
Indonesia) and temperate (multiple sites in Nebraska, USA) cli- als on site-specific nutrient management conducted for 2 years
mates. This database covers a wide yield range, from about 1 to in 24 farmers’ fields in two contrasting environments (alluvial
nearly 20 Mg ha−1 . The output of this analysis is therefore expected soils vs. degraded soils) in the Red River Delta, North Vietnam
to describe nutrient uptake and yield relations for maize grown (Son et al., 2004), and (ii) on-farm research trials on site-specific
with minimal biotic or abiotic stress other than limitation of N, nutrient management conducted for 2 years in 25 farmers’
P and K. The specific objectives of this study were to: (1) deter- fields across five different regions of Indonesia, namely North
mine the envelope functions describing relationships between Sumatra, South Sumatra, Central Java, East Java, and South
grain yield and nutrient (N, P, K) uptake by irrigated maize across a Sulawesi (Witt et al., 2008).
wide range of yields and environments; (2) quantify the balanced
N, P, and K uptake requirements of irrigated maize; and (3) derive Cropping systems varied widely. In Nebraska, maize was either
a novel spherical model as a simple, robust decision tool for esti- grown as continuous maize or following soybean. In Indone-
mating nutrient uptake requirements of irrigated maize. sia, double and triple cropping systems such as maize–maize,
160 T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168

rice–maize–maize and rice–rice–maize were predominant. In (b) for each nutrient, fit boundary lines to the data cloud of grain
North Vietnam, maize was grown during winter, following two yield vs. nutrient uptake; the two boundary lines describe the
crops of irrigated rice. Soil types included Mollisols for most maximum and minimum accumulation of the nutrient in the
Nebraska sites, whereas Inceptisols, Entisols, Alfisols, Vertisols, plant, respectively;
Ultisols and Oxisols were represented at the southeast Asian loca- (c) simulate curves of balanced uptake requirements of N, P and K
tions. All on-farm experiments were researcher-managed, with at different yield potential levels;
good control of weeds, diseases and insect pests. One site in (d) use the simulated nutrient requirement curves to derive and
Nebraska and three sites in Indonesia were under favorable rainfed parameterize a new spherical model for practical applications
conditions. All other sites had full sprinkler or furrow irriga- such as estimating nutrient uptake requirements for a certain
tion. At all sites, commercially available maize hybrids adapted yield goal.
to the specific environment were grown. Plant population densi-
ties were mostly in the range of 60,000–80,000 plants ha−1 , but
Only few samples fell outside normally expected ranges of the
also included some high-density treatments with up to about
harvest index (Fig. 2). Preliminary analysis indicated that they had
110,000 plants ha−1 (at Lincoln, NE). Tillage practices ranged from
minimal effect on the fitted boundary lines (envelopes). Therefore,
intensive plowing at some locations in Asia, to disking, ridge-till or
unlike in Witt et al. (1999), no data points were removed based on
no-till. The data set contained many different nutrient management
HI. Instead, as suggested by Witt et al. (1999), we used the 2.5th
treatments to establish a wide range of nutrient dilution and accu-
and 97.5th percentiles of nutrient IE for defining the final envelope
mulation situations, including single-nutrient omission plots such
function coefficients. Values outside this range were discarded as
as +PK (N omitted), +NK (P omitted) and +NP (K omitted) as well
potential measurement errors. Janssen et al. (1990) included the
as treatments with optimal fertilizer management and treatments
r-value (uptake required to produce any measurable grain yield)
with excessive NPK rates. Nitrogen applications ranged from 0 to
in their envelopes of the relationship between yield and nutrient
375 kg N ha−1 (average of 187 kg N ha−1 ). Phosphorus applications
uptake. However, because this value is small, difficult to define,
ranged from 0 to 130 kg P ha−1 (average of 22 kg P ha−1 ). Potas-
and has little impact on the modeled uptake requirements in yield
sium applications ranged from 0 to 160 kg K ha−1 (average of 44 kg
ranges of practical interest (Witt et al., 1999) it was not included in
K ha−1 ). At locations with known deficiencies, other nutrients (Mg,
our models.
Zn, S) were applied to ensure that nutrients other than N, P, or K
Using the envelope coefficients derived from the irrigated maize
were not limiting.
uptake data in our study, a spreadsheet version of the QUEFTS
Plant measurements followed similar protocols at all sites. Har-
model was used to generate optimum yield vs. nutrient uptake
vest areas for plant sampling varied from 6 m2 in Vietnam and
curves for yield potential levels ranging from 5 to 20 Mg ha−1 . An
Indonesia to 9–13.5 m2 in the Nebraska trials. A 6-plant sample at
automatic iteration module was added to facilitate the data gen-
R6 stage of maize (physiological maturity) was collected to deter-
erating steps using Microsoft® Visual Basic for Applications (VBA).
mine nutrient concentrations in grain and stover and obtain the
A spherical model (Dobermann et al., 2006) was used as a simpli-
harvest index. All plant parts were oven-dried to constant weight
fied algorithm describing the yield–uptake relations predicted by
at 70 ◦ C and ground for chemical analysis. Quality control mea-
the QUEFTS. In this spherical model, yield target is simulated as a
sures included the chemical analysis of standard plant samples with
function of nutrient uptake and yield potential:
known N, P and K concentrations. Plot grain yield and final plant
population density were measured from the whole harvest area.          3  
3 x 1 x
Final dry matter yield was calculated from the grain yield mea- y = if x ≤ c, b − ,b (1)
2 c 2 c
sured at harvest and the harvest index obtained from the 6-plant
sample collected at R6 stage. Plant nutrient accumulation in grain,
cobs, and vegetative parts was calculated from nutrient concentra- where y is yield target (Mg ha−1 ), x is nutrient uptake for a given
tions measured in the R6 sample and the estimated final dry matter yield target (kg ha−1 ), b is yield potential (Mg ha−1 ), and c is uptake
fractions. at yield approaching the maximum yield (kg ha−1 ).
Grain yields are reported at a standard moisture content of To solve nutrient uptake for a given yield goal y and yield poten-
0.155 g H2 O g−1 fresh weight. All nutrient concentrations in grain tial b the following simplified cubic solving algorithm (Nickalls,
and straw are reported on an oven-dry weight basis (about 3% resid- 1993) was used:
ual moisture after drying at 70 ◦ C). The harvest index, i.e., grain DM    4   −y 
as a proportion of above-ground plant DM, was obtained from the x = if y ≥ b, x ≥ c, c 2 cos + cos−1 (2)
oven-dry 6-plant sample. The nutrient harvest index is the nutrient 3 b
accumulation in grain as a proportion of total nutrient accumula-  
tion in above-ground plant DM. The internal efficiency (IE, kg kg−1 ) b × 103
x = min x, (3)
of a nutrient is defined as the amount of grain yield in kg ha−1 YA
(adjusted to 15.5% moisture content) produced per kg plant N, P or
K accumulation in above-ground plant DM (oven-dry weight). The The ‘if’ functions in Eqs. (1) and (2) have three arguments (sep-
reciprocal internal efficiency (RIE, kg kg−1 ) was calculated from the arated by commas): (i) logical test (true or false), (ii) the value if
average IE of all data and is the amount of a nutrient in the plant logical test equals true, and (iii) the value if logical test equals false.
DM needed to produce 1000 kg grain (on 15.5% moisture basis). YA is the envelope coefficient (slope) for the maximum nutrient
accumulation (kg grain kg nutrient uptake−1 ). The ‘min’ expres-
sion in Eq. (3) allows correction of uptake when y approaches b
2.2. Model development (otherwise uptake approaches infinity as y approaches b).
The spherical equation was fitted to the simulated QUEFTS
Data analysis largely followed Witt et al. (1999). Only a brief yield–uptake data using SigmaPlot 10 (Systat Software, Richmond,
description will be provided here, focusing on some differences in CA). Regression analysis was conducted to evaluate the relationship
our study. Key steps included: between the coefficients b and c in Eq. (1). In theory, the uptake at
which yield reaches a maximum is positively correlated with yield
(a) screen the data set and remove outliers/potential errors; potential.
T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168 161

Fig. 2. Distribution of grain yield and harvest index of irrigated maize in studies conducted in Nebraska and Southeast Asia.

3. Results and discussion whereas grain yields in our experiments averaged 13.8 Mg ha−1 .
The relatively higher maize grain yields in individual sub-sets as
3.1. Grain yield and above-ground biomass compared to national and state average are an indication of what
can be achieved with good management, and that the data used
Maize grain yield (15.5% m.c.) in the present study ranged from in our analysis represented conditions of mostly excellent man-
0.3 to 19.0 Mg ha−1 with an average of 12.0 Mg ha−1 (Table 1). The agement, with only small yield losses due to abiotic and biotic
grain yield mean in this study was much higher than the world aver- stresses. Maximum grain yield in this study is less than the theoret-
age (1997–2006) of 4.5 Mg ha−1 (FAO, 2008). Average grain yield ical potential yield of 25 Mg ha−1 (Tollenaar, 1983), but was within
for the various data sub-sets (Fig. 1) was 30–150% higher than the the vicinity of the reported maximum yields (19.4–20.9 Mg ha−1 )
corresponding national maize grain yield averages (1997–2006) of in USA (Tollenaar and Lee, 2002).
8.2, 3.0, and 3.0 Mg ha−1 in USA, Indonesia, and Vietnam, respec- Mean grain yield in NE (13.8 Mg ha−1 ) was more than twice
tively (FAO, 2008). During the 2000–2005 period, the average grain the average maize yield achieved in the Southeast Asia studies
yield of irrigated maize in Nebraska was 11.0 Mg ha−1 (NASS, 2008), (5.8 Mg ha−1 ). This was mainly due to differences in the cli-

Table 1
Descriptive statistics of grain yield (at 15.5% moisture), total above-ground dry matter (TDM), harvest index, concentrations of N [N], P [P], and K [K] in grain and stover (all
on oven-dry weight basis), plant N, P, and K accumulation in grain, and TDM, and nutrient harvest index (kg nutrient in grain per kg nutrient in TDM) at maturity of irrigated
maize in Nebraska and South East Asia (all data).

Parameter Unit n Mean SD Minimum 25% Q Median 75% Q Maximum


−1
Grain yield Mg ha 2579 12.01 4.14 0.32 9.98 13.28 15.03 19.01
Total dry matter Mg ha−1 2579 20.83 7.05 0.84 17.48 22.33 25.75 37.71
Harvest Index kg kg−1 2579 0.50 0.06 0.17 0.46 0.51 0.55 0.62
[N] in grain g kg−1 2341 13.30 2.15 4.90 11.93 13.36 14.76 19.63
[P] in grain g kg−1 2363 2.63 0.79 0.55 2.01 2.59 3.19 5.15
[K] in grain g kg−1 2361 3.63 0.80 0.99 3.12 3.60 4.06 9.73
[N] in stover g kg−1 2339 8.11 2.22 2.21 6.54 8.02 9.61 19.88
[P] in stover g kg−1 2363 0.52 0.41 0.05 0.22 0.42 0.70 4.15
[K] in stover g kg−1 2361 21.82 7.04 1.54 18.21 22.80 26.44 41.71
N uptake grain kg ha−1 2341 148.5 53.6 3.4 122.3 155.5 181.9 305.5
P uptake grain kg ha−1 2363 29.8 13.8 0.6 19.3 29.0 40.5 70.8
K uptake grain kg ha−1 2361 40.3 14.5 2.4 31.4 41.8 50.4 96.9
N uptake total kg ha−1 2341 232.2 81.8 6.8 189.5 239.1 286.9 471.2
P uptake total kg ha−1 2363 35.2 15.1 1.1 23.2 34.4 46.7 79.9
K uptake total kg ha−1 2361 269.1 122.8 4.3 186.9 270.4 360.2 597.9
N harvest index kg kg−1 2341 0.64 0.08 0.21 0.59 0.64 0.70 0.97
P harvest index kg kg−1 2363 0.84 0.10 0.21 0.79 0.86 0.91 0.98
K harvest index kg kg−1 2361 0.17 0.08 0.06 0.13 0.16 0.19 0.68

n = number of observation; SD = standard deviation; Q = quartile.


162 T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168

Table 2
Descriptive statistics of the internal efficiency of N, P and K (IE, kg grain kg nutrient−1 ) and its reciprocal term (RIE, kg nutrient Mg grain yield−1 ) for irrigated maize in Nebraska
and South East Asia.

Data set Parameter n Mean SD Min. 25% Q Med. 75% Q Max.

All IE-N 2341 56 11 22 49 55 62 125


IE-P 2363 400 134 172 302 367 468 1157
IE-K 2361 56 26 23 40 50 64 220
RIE-N 2341 18.4 3.5 8.0 16.1 18.2 20.5 46.3
RIE-P 2363 2.7 0.8 0.9 2.1 2.7 3.3 5.8
RIE-K 2361 20.4 6.9 4.6 15.5 20.0 25.3 43.3

Optimal NPK IE-N 191 50 6 30 46 49 54 70


rates IE-P 194 321 60 207 280 313 352 654
IE-K 191 51 21 29 37 43 60 117
RIE-N 191 20.5 2.5 14.4 18.7 20.2 21.9 33.2
RIE-P 194 3.2 0.5 1.5 2.8 3.2 3.6 4.8
RIE-K 191 21.8 6.3 8.5 16.8 23.2 26.8 34.1

Nebraska IE-N 2005 56 9 33 49 55 61 96


IE-P 2005 394 132 184 297 362 463 1110
IE-K 2005 50 16 24 39 48 59 133
RIE-N 2005 18.4 3.0 10.4 16.3 18.2 20.5 30.7
RIE-P 2005 2.8 0.8 0.9 2.2 2.8 3.4 5.4
RIE-K 2005 21.6 6.2 7.5 17.0 20.9 26.0 42.0

SE Asia IE-N 337 59 18 22 47 55 66 125


IE-P 359 434 140 172 341 397 491 1157
IE-K 357 89 41 23 61 82 109 220
RIE-N 337 18.5 5.6 8.0 15.1 18.3 21.1 46.3
RIE-P 359 2.5 0.7 0.9 2.0 2.5 2.9 5.8
RIE-K 357 14.0 7.3 4.6 9.2 12.1 16.5 43.3

n = number of observations; SD = standard deviation; Q = quartile; Med. = median; Min. = minimum; Max. = maximum.

matic yield potential. Simulations with the Hybrid-Maize model 3.2. Nutrient concentrations, uptake and internal efficiencies
(Yang et al., 2006) showed that maize yield potential at the
Nebraska sites was typically within a range of 14–20 Mg ha−1 for Average nutrient concentrations in grain were 13.3 g N kg−1 ,
the location-specific growing conditions and management prac- 2.6 g P kg−1 and 3.6 g K kg−1 (oven-dry weight basis), while con-
tices (hybrid maturity and plant density) used in our studies. At centrations in stover (stems, leaves, husks) were 8.1 g N kg−1 ,
recommended fertilizer rates, actual yields in Nebraska were typ- 0.5 g P kg−1 and 21.8 g K kg−1 (Table 1). However, due to the
ically within 80–90% of the simulated yield potential. In contrast, wide range of environmental and management conditions, nutri-
simulated maize yield potential was 8–10 Mg ha−1 in North Viet- ent concentrations varied tremendously in both grain (4.9–19.6 g
nam and 10–14 Mg ha−1 at the Indonesian sites (A. Dobermann, N kg−1 , 0.6–5.2 g P kg−1 and 1.0–9.7 g K kg−1 ) and stover (2.2–19.9 g
J.M. Pasuquin and J. Timsina, unpublished), and actual yields with N kg−1 , 0.1–4.2 g P kg−1 and 1.5–41.7 g K kg−1 ). Lowest nutrient
recommended nutrient supply were mostly within 70–80% of the concentrations were mainly observed in nutrient omission plots. In
yield potential. Maximum grain yield measured in North Vietnam contrast, maximum nutrient concentrations occurred in situations
was 7.2 Mg ha−1 , whereas in Indonesia it reached 12.8 Mg ha−1 in of excessive supply in situations where other nutrient deficiencies
East Java and 13.7 Mg ha−1 in Central Java. Hence, in addition to and environmental conditions limited plant growth. On average,
climate, it is likely that generally poorer soils and high disease and nutrient accumulation in the above-ground plant dry matter was
insect pressure caused slightly greater yield gaps in Southeast Asia 232 kg N ha−1 , 35 kg P ha−1 and 269 kg K ha−1 at an average yield
as compared to the Nebraska sites. of 12 Mg ha−1 (Table 1). The nutrient harvest index, i.e., nutrient
Total above-ground dry matter (TDM) ranged from 0.8 to accumulation in grain as a proportion of nutrient accumulation in
37.7 Mg ha−1 with an average of 20.8 Mg ha−1 , while harvest index above-ground plant DM, was highest for P (0.84), followed by N
(HI) ranged from 0.17 to 0.62 with an average of 0.50 (Table 1). (0.64) and K (0.17).
Similar to grain yield, TDM and HI were higher in Nebraska Concentrations of N and P in grain were higher in Nebraska
(23.6 Mg ha−1 and 0.51) than in Southeast Asia (10.9 Mg ha−1 , 0.46). (13.5 g N kg−1 and 2.7 g P kg−1 ) than in Southeast Asia (12.0 g
Half of all observations were within a narrow HI range of 0.46–0.55. N kg−1 and 2.1 g P kg−1 ). Most notable, average K concentration in
The average HI of 0.5 is in agreement with the range values for mod- stover was more than twice as much in Nebraska (23.7 g K kg−1 )
ern varieties of most intensively cultivated grain crops, which fall than in Southeast Asia (11.2 g K kg−1 ). Similarly, accumulation of
within a HI of 0.4–0.6 (Hay, 1995). The relatively small deviation in K in total dry matter was nearly three times higher in Nebraska
harvest index indicated that lower grain yield in Southeast Asia as (299 kg K ha−1 ) than in Southeast Asia (106 kg K ha−1 ), while the K
compared to grain yield in the Nebraska was related to the differ- harvest index was much smaller in Nebraska (0.15) as compared
ence in biomass production (Table 1). Differences in the distribution to Southeast Asia (0.28). This largely reflects the generally high K
of harvest index as compared to grain yield are shown in Fig. 2. The status of Nebraska soils used for growing maize (Wortmann et al.,
range from the 2.5th to 97.5th percentile is narrower for HI than 2009).
for grain yield. The congruence of median and mean of the normal Across all sites and seasons, the median measured IEs were
distribution is closer for the distribution plot of HI than for grain 55 kg grain per kg plant N, 367 kg grain per kg plant P and 50 kg
yield. For grain yield, there appears to be more than just one peak grain per kg plant K, equivalent to RIE of 18.2 kg N, 2.7 kg P and
in the data distribution, associated with Nebraska and Southeast 20.0 kg K per Mg grain yield produced (Table 2) or a N:P:K ratio
Asia data sets. Due to a greater proportion of data from Nebraska, of 6.8:1:7.6 in plant DM. As expected, IE varied widely, from 22 to
which tends to be in the high-yield region, the median is stretched 125 kg kg−1 for N, 172 to 1157 kg kg−1 for P, and 23 to 220 kg kg−1
toward the higher grain yield side of the mean (Fig. 2). for K. This wide variation in IE indicates how difficult it is to esti-
T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168 163

mate nutrient uptake requirements from empirical studies that are yield produced. In field experiments conducted from 1985 to 1995
limited in terms of diversity in environmental and management in five regions of China, Liu et al. (2006) found average IEs of 42 kg
situations. The median IE and RIE for N and P were quite similar grain per kg plant N, 255 kg grain per kg plant P and 51 kg grain per
between the two mega regions in our study, whereas the IE of K kg plant K, equivalent to average nutrient requirements of 23.6 kg
was significantly larger in Southeast Asia (82 kg kg−1 ) as compared N, 3.9 kg P and 19.8 kg K per Mg grain yield produced. Hence, dif-
to Nebraska (48 kg kg−1 ), for the reasons cited above. Median IEs ferences between these studies and our new data mainly occurred
across all sites and seasons when optimal rates of fertilizers were for N and P. They were mostly caused by (i) a newer generation of
applied were 49 kg grain per kg plant N, 313 kg grain per kg plant P commercial maize hybrids used at most of the sites in our study
and 43 kg grain per kg plant K. The lower IEs were due to the exclu- and (ii) generally higher yield levels and less abiotic stresses in our
sion of data from N, P, and K omission plots that tend to have higher experiments, most of which were irrigated. Both factors may also
IEs (data not shown). The overall ranges in IEs were also less in data be the cause of higher HI in our studies compared to others that
that include only optimal N, P, and K rates, with IE ranging from 30 were used to estimate internal nutrient efficiencies. With regard
to 70 kg kg−1 for N, 207 to 654 kg kg−1 for P, and 29 to 117 kg kg−1 to (i), it has been shown, for example, that modern maize hybrids
for K. maintain higher leaf photosynthesis during grain filling and tend to
Our measurements of IEs, largely conducted in farmers’ fields, have greater stress tolerance, higher HI and greater translocation of
were higher for N and P, but lower for K than those found in previous nutrients such as N (Ding et al., 2005; Echarte et al., 2008; Tollenaar
studies. For example, summarizing results of more than 50 older et al., 2004; Tollenaar and Lee, 2002), resulting in an increased
on-station experiments, van Duivenbooden et al. (1996) reported N harvest index and thus also increased IE of N as compared to
IEs of 43 kg grain per kg plant N, 285 kg grain per kg plant P and inbreds or older hybrids. A low HI in some older studies may also
60 kg grain per kg plant K. This was equivalent to average nutri- be caused by suboptimal N management, resulting in reduced grain
ent requirements of 23.4 kg N, 3.5 kg P and 16.6 kg K per Mg grain filling.

Fig. 3. Relationships between grain yield and N, P and K uptake (a–c) and grain yield and N, P ad K removal with grain (d–f). YU and YG curves are the QUEFTS-simulated
balanced N, P, or K uptake or grain removal requirements to achieve a certain grain yield target for the given maximum nutrient dilution (YD) and nutrient accumulation
(YA) envelopes fitted to the datasets. Balanced nutrient uptake curves are shown for two levels of yield potential: 20 Mg ha−1 (Nebraska) and 14 Mg ha−1 (Southeast Asia).
164 T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168

Table 3 d in our study (40 and 83 kg grain kg N−1 , 225 and 726 kg grain kg
Envelope coefficients relating grain yield of maize (GY) to the maximum accumula-
P−1 , 29 and 129 kg grain kg K−1 ) were comparable to the values
tion (a) and maximum dilution (d) of N, P, and K in the above-ground dry matter of
maize at maturity. For comparison, previously published value of a and d are shown for maize proposed by Janssen et al. (1990), especially for K, but
for studies conducted in Africa (Janssen et al., 1990) and in China (Liu et al., 2006). significantly larger than the values reported by Liu et al. (2006) for
maize in China (Table 3) and the literature data summarized by van
kg grain kg nutrient−1
Duivenbooden et al. (1996).
N P K With the new a and d coefficients derived from our data, the
Maximum nutrient accumulation (a) QUEFTS model was used to simulate balanced nutrient uptake
This study 40 225 29 requirements for N, P, and K across yield potentials ranging from 6
Janssen et al. (1990) 30 200 30
to 20 Mg ha−1 . In QUEFTS, interaction of the three dominant nutri-
Liu et al. (2006) 21 126 19
ents N, P, and K in relation to yield follows Liebscher’s law of the
Maximum nutrient dilution (d) optimum, i.e., a production factor that is in minimum supply con-
This study 83 726 125
tributes more to yield if other factors are closer to their optimum
Janssen et al. (1990) 70 600 120
Liu et al. (2006) 64 384 90 level (de Wit, 1992). Interaction of the three nutrients therefore
governs the slope of the yield–uptake curve up to a yield level

This illustrates the need to regularly update information on crop


nutrient update requirements and nutrient removal with grain. Fre-
quently, old information is still used in fertilizer recommendations
and nutrient budgeting for various purposes, leading to erroneous
results. Using the lower IEs reported earlier for N and P in maize,
for example, would lead to potentially excessive fertilizer recom-
mendations for these nutrients.

3.3. Estimating the optimum nutrient uptake for a specific yield


target

The boundary lines describing yield–uptake relations for N, P,


and K obtained from this study are shown in Fig. 3 and Table 3. For
each nutrient, the lower boundary line represents the maximum
accumulation (a) of that nutrient in the plant, which is equivalent
to the minimum IE (see Table 2) and reflects either a restriction
on the plant-internal utilization of that nutrient due to other con-
straints, or, in the case of K, an excessive (“luxury”) uptake beyond
amounts required to produce grain. In contrast, the upper bound-
ary line represents the maximum dilution (d) of that nutrient in the
plant, which is equivalent to maximum IE and generally reflects a
severe deficiency of that nutrient. The slope coefficients for a and

Table 4
Balanced uptake requirements, internal efficiencies (IE, kg grain kg nutrient−1 ), and
reciprocal internal efficiencies (RIE, kg nutrient Mg grain yield−1 ) of N, P, and K for
irrigated maize as simulated with the QUEFTS model to achieve certain grain yield
targets and assuming a maximum yield of 20 Mg ha−1 . All values are expressed on
elemental basis.

Yield Required nutrient Internal efficiency Reciprocal internal


(Mg ha−1 ) uptake (kg ha−1 ) (kg grain kg efficiency (kg
nutrient−1 ) nutrient Mg grain−1 )

N P K N P K N P K

1 16 2 16 61 427 63 16.4 2.3 15.9


2 33 5 32 61 427 63 16.4 2.3 15.9
3 49 7 48 61 427 63 16.4 2.3 15.9
4 66 9 64 61 427 63 16.4 2.3 15.9
5 82 12 80 61 427 63 16.4 2.3 15.9
6 99 14 95 61 427 63 16.4 2.3 15.9
7 115 16 111 61 427 63 16.4 2.3 15.9
8 131 19 127 61 427 63 16.4 2.3 15.9
9 148 21 143 61 427 63 16.4 2.3 15.9
10 164 23 159 61 427 63 16.4 2.3 15.9
11 181 26 175 61 427 63 16.4 2.3 15.9
12 197 28 191 61 426 63 16.5 2.3 15.9
13 216 31 209 60 423 62 16.6 2.4 16.1
14 234 33 227 60 419 62 16.7 2.4 16.2
15 257 37 249 58 409 60 17.2 2.4 16.6
16 285 41 276 56 394 58 17.8 2.5 17.3
Fig. 4. The balanced N, P, K uptake requirements for certain yield targets (YU) as
17 317 45 307 54 376 55 18.7 2.7 18.1
simulated with the QUEFTS model based on the boundary lines obtained from this
18 356 51 345 51 354 52 19.8 2.8 19.2
study, those proposed by Janssen et al. (1990), and those proposed by Liu et al.
19 409 58 397 46 326 48 21.5 3.1 20.9
(2006) at two maximum yield levels: maximum yield in Nebraska (Ymax-NE) and
20 563 80 546 36 249 37 28.2 4.0 27.3
maximum yield in Liu et al. (2006) (Ymax-Liu).
T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168 165

Fig. 5. The balanced requirements of total nutrient uptake (a–c) and nutrient removal with grain (d–f) as simulated with QUEFTS (symbols), and the fitted spherical models
(lines, r2 = 0.998) for grain yield at several maximum yield levels (6–20 Mg ha−1 ).

that plateaus at the environmental yield potential, which is deter- Asia, respectively. The balanced grain nutrient removal was 59, 77,
mined by solar radiation and temperature. Table 4 illustrates this and 19% of total N, P, and K uptake, respectively.
for a specific environment with a yield potential of 20 Mg ha−1 , A comparison of optimal nutrient uptake–yield curves using
such as occurs in parts of Nebraska. The model predicted a linear boundary line coefficients derived from this study and other stud-
increase in grain yield if nutrients are taken up in balanced amounts ies is shown in Fig. 4. The curves based on the proposed new a and
of 16.4 kg N, 2.3 kg P and 15.9 kg K per 1 Mg of grain until yield d coefficients have a larger slope (=less nutrients required per unit
reached about 60% of the yield potential. The corresponding opti- yield target), especially as compared to the coefficients reported in
mal IEs were 61 kg grain kg−1 N, 427 kg grain kg−1 P and 63 kg grain Liu et al. (2006). The proposed revised model, based on the more
kg−1 K for balanced nutrition. The model predicted a decrease in IEs recent and wider-ranging date used in our study, is likely to rep-
when yield targets approached the yield potential. In Fig. 3a–c, the resent a relationship that is globally applicable as long as factors
corresponding curve is superimposed on the entire data set, and other than N, P, and K are not limiting.
a similar QUEFTS-simulated curve is shown for the Southeast Asia Our final objective was to develop a simple, empirical model that
environment, with a yield potential of only 14 Mg ha−1 . One notable could be used as a tool to calculate the required nutrient uptake
difference in our study to those by others for maize, wheat and rice for a certain yield target, in an environment with a well-defined
(Liu et al., 2006; Pathak et al., 2003; Witt et al., 1999) is the presence yield potential – without having to use a more complex model
of large amounts of data points in the high-yield regions, near the such as QUEFTS. To that end, we found that the proposed spherical
yield potential in both environments. model (Eq. (1)) fitted the simulated QUEFTS model outputs very
Net removal of nutrients with the grain is used in many practical well (r2 = 0.998), for both yield–uptake and yield–grain nutrient
algorithms for guiding fertilizer recommendations, based on the removal relations (Fig. 5). In Fig. 5, the yield potential parameter
assumption that what is removed must be replenished to maintain b in Eq. (1) was based on the range found in our dataset. Variable
a sustainable production system (Dobermann and Cassman, 2002). c (uptake or grain removal when yield approaches maximum) was
However, grain nutrient removal coefficients (kg removed per unit linearly related (r2 = 1.00) to yield potential (Fig. 6). Therefore, the
yield) are not constant. Using the same QUEFTS-based simulation combination of Eq. (1) and the regression equations in Fig. 6 provide
approach, the balanced grain nutrient removal associated with the a simpler approximation of yield–uptake or yield–grain nutrient
NE and SE Asia yield potentials is shown in Fig. 3e and f. Fig. 3d removal relationships for each nutrient (N, P, or K). This approach
suggests, for example, that when yield reached the plateau, grain has the advantage of a straightforward calculation of uptake or
N removal was 271 and 195 kg N ha−1 for Nebraska and Southeast grain nutrient removal at a specified yield target. Eq. (2)/(3) in con-
166 T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168

Fig. 6. Regression lines (r2 = 1.0) for the relationships between the spherical model coefficients c and b for grain yield vs. nutrient uptake and grain yield vs. grain nutrient
removal, with symbols representing the c and b coefficients associated with the simulations in Fig. 5.

QUEFTS approach had a root mean squared error (RMSE) of 37 kg


N ha−1 and a normalized RMSE of 15% (Fig. 7a). However, the
observed negative relationship between the normalized N uptake
(N uptake/maximum N uptake at a particular yield level) and HI
(Fig. 8) suggest that HI can be used to improve estimation of N
uptake for a given yield target. Stepwise regressions based on the
observed grain yield showed an improvement in regression r2 (data
not shown), suggesting that regression coefficients for normalized
N uptake vs. HI are influenced by grain yield level in relation to
the yield potential ceiling for the environment in question. Based
on these observations, the following algorithm was developed for
adjustment of the predicted N uptake based on HI:

UN HI adjusted = UN f (HI) (4)

f (HI) = a(HI) + b (5)

a= − 0.2619 Yg+1.8629, a= min(a, −1.28), a= max(−2.40, a)

b = 0.1456 Yg − 0.3126, b = min(b, 2.07), b = max(1.43, b)

where UN HI adjusted is N uptake adjusted based on HI, UN is N


uptake predicted by the spherical-QUEFTS approach, Yg is yield
target (Mg ha−1 ), and HI is the harvest index (kg grain/kg above-
ground DM). Such an adjustment of N uptake predictions resulted
in a 10 kg N ha−1 smaller RMSE (Fig. 7b), and also eliminated the
trend of under-predicting N uptake at higher observed N uptake
(compare Fig. 7a and b). Improvement in prediction of N uptake
based on correction for HI as yields approach yield potential should
occur for all nutrients that have much higher concentration in grain
than in vegetative tissues at physiological maturity because the

Fig. 7. Comparisons of the observed and simulated N uptake without adjustment


based on harvest index (a) and with adjustment based on harvest index (b). Simu-
lation errors (simulated–observed) are shown as inserts. RMSE = root mean square
error; and NRMSE = normalized root mean square error (RMSE/mean observed).

junction with regression equations in Fig. 6 can be used for that


purpose.
Further improvements could be made to increase precision
of the nutrient uptake and removal estimates. For example,
Fig. 8. Relationship between observed harvest index and observed normalized
when compared to measured data at recommended nutrient N uptake (N uptake/maximum N uptake for a particular yield level) (r2 = 0.46,
management levels, N uptake simulated with the above spherical- p < 0.001).
T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168 167

magnitude of the concentration ratio between grain and vegetative for doing such location-specific adjustment is, however, that reli-
tissues decreases as yields approach the plateau (data not shown). able grain yield and nutrient uptake data are collected under best
However, it is generally important to follow good N management management practices and conditions that fully reflect farming
practice before making adjustments to nutrient uptake or removal environments, not just those of few research stations.
coefficients based on a lower HI.
Acknowledgements

4. Conclusions Funding for this research was provided by the State of Nebraska,
the International Plant Nutrition Institute (IPNI), the International
Due to the large variation in IE values found in this and other Fertilizer Industry Association (IFA), Canpotex International Pte.
studies on maize, quantifying internal nutrient efficiencies (and/or Ltd., the International Potash Institute (IPI), the Fluid Fertilizer
net removal) through a generic modeling approach is likely to be Foundation (FFF), Pioneer Hi-Bred Intl., and the Swiss Agency for
more reliable than attempting to specify an average value from Development and Cooperation (SDC). The authors wish to thank
limited empirical data. Our study demonstrates the relationship the following scientists and their institutions for conducting the
between grain yield and total nutrient uptake (or nutrient removal field experiments. Nebraska: D. Binder, G. Teichman, C. Wortmann,
with grain) can be modeled using the QUEFTS model or a simpler, R. Ferguson, C. Shapiro, D. Tarkalson, G. Hergert (all University
spherical model across a wide range of maize environments and of Nebraska); Indonesia: Sari S. Girsang (AIAT North Sumatera),
grain yield levels. This approach not only allows estimating the Andarias M. Murni (AIAT Lampung), Supadmo (AIAT Central Java),
nutrient requirements for achieving a specified yield target, but also Suwono (AIAT East Java), Peter Tandisau (AIAT South Sulawesi),
provides a powerful tool for identifying nutritionally and econom- Sania Saenong and Subandi (ICRI), Sunendar Kartaatmadja (IPNI)
ically optimal yield targets in relation to yield potential at a given and J.M. Pasuquin (IPNI); Vietnam: T.T. Son, N.V. Chien, and V.T.K.
location. The proposed spherical model can easily be incorporated Thoa (NISF).
in existing or new fertilizer prescription algorithms or models. It is
robust across environments and management practices, allowing
users to estimate the optimal N, P and K uptake requirements based References
on two inputs: estimated site yield potential and yield target.
Adviento-Borbe, M.A.A., Haddix, M.L., Binder, D.L., Walters, D.T., Dobermann, A.,
Using this approach yield potential is the key location-specific 2007. Soil greenhouse gas fluxes and global warming potential in four high-
parameter required for estimating nutrient uptake requirements yielding maize systems. Global Change Biol. 13, 1972–1988.
Bai, J., Chen, X., Dobermann, A., Yang, H., Cassman, K.G., Zhang, F., 2010. Evaluation
for a specific location, whereas functions describing IE can be
of NASA satellite- and model-derived weather data for simulation of maize yield
modeled with generic parameters. Site yield potential can be esti- potential in China. Agron. J. 102, 2–16.
mated using a simulation model such as Hybrid-Maize if long-term de Wit, C.T., 1992. Resource use efficiency in agriculture. Agric. Syst. 40, 125–151.
historical weather data are available (Yang et al., 2006). Use of Ding, L., Wang, K.J., Jiang, G.M., Liu, M.Z., Niu, S.L., Gao, L.M., 2005. Post-anthesis
changes in photosynthetic traits of maize hybrids released in different years.
satellite-derived data for solar radiation provides opportunities Field Crops Res. 93, 108–115.
to estimate yield potential in regions that only have long-term Dobermann, A., Cassman, K.G., 2002. Plant nutrient management for enhanced pro-
data on temperature and rainfall (Bai et al., 2010). Calibration of ductivity in intensive grain production systems of the United States and Asia.
Plant Soil 247, 153–175.
the QUEFTS model for maize grown across temperate and trop- Dobermann, A., Ferguson, R.B., Hergert, G.W., Shapiro, C.A., Tarkalson, D., Walters,
ics environments required estimating slopes of two boundary lines D.T., Wortmann, C.S., 2006. Nitrogen response in high-yielding corn systems of
describing the maximum accumulation (a) and dilution (d) of N, Nebraska. In: Schlegel, A.J. (Ed.), Proceedings of the Great Plains Soil Fertility
Conference. Denver, CO, March 7–8, 2006. Kansas State University, Manhattan,
P and K in the plant in relation to grain yield. We propose to use KS, pp. 50–59.
aN = 40, dN = 83, aP = 225, dP = 726, aK = 29 and dK = 125 as standard Dobermann, A., Witt, C., Dawe, D., Abdulrachman, S., Gines, G.C., Nagarajan, R.,
model parameters in QUEFTS for modern maize hybrids with a HI Satawathananont, S., Son, T.T., Tan, P.S., Wang, G.H., Chien, N.V., Thoa, V.T.K.,
Phung, C.V., Stalin, P., Muthukrishnan, P., Ravi, V., Babu, M., Chatuporn, S.,
of approximately 0.50 when grown under irrigated or favorable Kongchum, M., Sookthongsa, J., Sun, Q., Fu, R., Simbahan, G.C., Adviento, M.A.A.,
rainfed conditions with no severe abiotic constraints. 2002. Site-specific nutrient management for intensive rice cropping systems in
On condition that plant growth is primarily limited by nutrient Asia. Field Crops Res. 74, 37–66.
Echarte, L., Rothstein, S., Tollenaar, M., 2008. The response of leaf photosynthesis
supply, the proposed model predicted a linear increase in maize
and dry matter accumulation to nitrogen supply in an older and a newer maize
yield if nutrients are taken up in balanced amounts of 16.4 kg N, hybrid. Crop Sci. 48, 656–666.
2.3 kg P and 15.9 kg K per Mg of grain, until yield reached about FAO, 2008. FAOSTAT Database – Agricultural Production. Food and Agriculture Orga-
60–70% of the yield potential. The corresponding IEs were 61 kg nization of the United Nations, Rome. Available from: http://apps.fao.org.
Greenwood, D.J., Lemaire, G., Gosse, G., Cruz, P., Draycott, A., Neeteson, J.J., 1990.
grain kg−1 N, 427 kg grain kg−1 P and 63 kg grain kg−1 K for a bal- Decline in percentage N of C3 and C4 crops with increasing plant mass. Ann.
anced nutrition. With yield approaching the yield potential, IE of Bot. 66, 425–436.
nutrients decreased. Hence, managing maize crops at yield levels Haefele, S.M., Wopereis, M.C.S., Ndiaye, M.K., Barro, S.E., Ould Isselmo, M., 2003.
Internal nutrient efficiencies, fertilizer recovery rates and indigenous nutrient
of more than 60–70% of the site yield potential requires greater supply of irrigated lowland rice in Sahelian West Africa. Field Crops Res. 80,
amounts of nutrients per unit increase in grain yield, and greater 19–32.
care in applying nutrients to avoid losses to the environment, espe- Hay, R.K.M., 1995. Harvest index: a review of its use in plant breeding and crop
physiology. Ann. Appl. Biol. 126, 197–216.
cially for N. Heckman, J.R., Sims, J.T., Beegle, D.B., Coale, F.J., Herbert, S.J., Bruulsema, T.W., Bamka,
The generic methodology used in our study can also be applied W.J., 2003. Nutrient removal by corn grain harvest. Agron. J. 95, 587–591.
to developing more region- or location-specific envelopes describ- Janssen, B.H., Guiking, F.C.T., Van der Eijk, D., Smaling, E.M.A., Wolf, J., van Reuler,
H., 1990. A system for quantitative evaluation of the fertility of tropical soils
ing the relationship between grain yield and nutrient uptake. This
(QUEFTS). Geoderma 46, 299–318.
could allow taking into account genetic, environmental or man- Karlen, D.L., Flannery, R.L., Sadler, E.J., 1988. Aerial accumulation and partitioning of
agement differences that may affect the IE of nutrients, such as nutrients by corn. Agron. J. 80, 232–242.
Khurana, H., Phillips, S., Bijay, S., Bijay, S., Alley, M., Dobermann, A., Sidhu, A., Yad-
those between open-pollinated and hybrid varieties, if there are
vinder, S., Yadvinder, S., Peng, S., 2008. Agronomic and economic evaluation of
systematic differences in HI or nutrient uptake and translocation. site-specific nutrient management for irrigated wheat in northwest India. Nutr.
Such envelopes are likely to be narrower than the global models Cycling Agroecosyst. 82, 15–31.
developed in our study, and may thus be useful for further increas- Liu, M.Q., Yu, Z.R., Liu, Y.H., Konijn, N.T., 2006. Fertilizer requirements for wheat and
maize in China: the QUEFTS approach. Nutr. Cycling Agroecosyst. 74, 245–258.
ing the precision of estimating nutrient uptake requirements for NASS, 2008. National Agricultural Statistics Service. USDA, Washington. Available
purposes of precision nutrient management. A key requirement from: http://www.usda.gov/nass.
168 T.D. Setiyono et al. / Field Crops Research 118 (2010) 158–168

Nickalls, R.W.D., 1993. A new approach to solving the cubic: Cardan’s solution van Keulen, H., 1986. Crop yield and nutrient requirements. In: van Keulen, H., Wolf,
revealed. Math. Gaz. 77, 354–359. J. (Eds.), Modelling of Agricultural Production: Weather, Soils and Crops. Pudoc,
Pathak, H., Aggarwal, P.K., Roetter, R.P., Kalra, N., Bandyopadhaya, S.K., Prasad, S., Wageningen, pp. 155–181.
van Keulen, H., 2003. Modelling the quantitative evaluation of soil nutrient sup- van Keulen, H., 1977. Nitrogen Requirements of Rice with Special Reference to Java.
ply, nutrient use efficiency, and fertilizer requirements of wheat in India. Nutr. Ministry of Agriculture, Agricultural Technical Assistance Programme, Bogor,
Cycling Agroecosyst. 65, 105–113. Indonesia.
Saidou, A., Janssen, B.H., Temminghoff, E.J.M., 2003. Effects of soil properties, mulch van Keulen, H., Van Heemst, H.D.J., 1982. Crop response to the supply of macronu-
and NPK fertilizer on maize yields and nutrient budgets on ferralitic soils in trients. Agric. Res. Rep. 916. Pudoc, Wageningen.
southern Benin. Agric. Ecosyst. Environ. 100, 265–273. van Noordwijk, M., Cadisch, G., 2002. Access and excess problems in plan nutrition.
Smaling, E.M.A., Janssen, B.H., 1993. Calibration of QUEFTS: a model predicting Plant Soil 247, 25–40.
nutrient uptake and yields from chemical soil fertility indices. Geoderma 59, Verma, S.B., Dobermann, A., Cassman, K.G., Walters, D.T., Knops, J.M.H., Arkebauer,
21–44. T.J., Suyker, A.E., Burba, G.G., Amos, B., Yang, H.S., Ginting, D., Hubbard, K.G., Gitel-
Son, T.T., Chien, N.V., Thoa, V.T.K., Dobermann, A., 2004. Site-specific nutrient son, A.A., Walter-Shea, E.A., 2005. Annual carbon dioxide exchange in irrigated
management in irrigated rice systems of the Red River Delta, Vietnam. In: Dober- and rainfed maize-based agroecosystems. Agric. Forest Meteorol. 131, 77–96.
mann, A., Witt, C., Dawe, D. (Eds.), Increasing Productivity of Intensive Rice Walker, W.M., Peck, T.R., 1974. Relationship between corn yield and plant nutrient
Systems Through Site-Specific Nutrient Management. Science Publishers, Inc., content. Agron. J. 66, 253–256.
International Rice Research Institute, Enfield, NH, USA and Los Baños, Philip- Witt, C., Dobermann, A., Abdulrachman, S., Gines, H.C., Wang, G.H., Nagarajan, R.,
pines, pp. 217–242. Satawathananont, S., Son, T.T., Tan, P.S., Tiem, L.V., Simbahan, G.C., Olk, D.C., 1999.
Tollenaar, M., 1983. Potential vegetative productivity in Canada. Can. J. Plant Sci. 63, Internal nutrient efficiencies of irrigated lowland rice in tropical and subtropical
1–10. Asia. Field Crops Res. 63, 113–138.
Tollenaar, M., Ahmadzadeh, A., Lee, E.A., 2004. Physiological basis of heterosis for Witt, C., Pasuquin, J.M., Dobermann, A., 2008. Site-specific nutrient management
grain yield in maize. Crop Sci. 44, 2086–2094. for maize in favorable tropical environments of Asia. In: Proc. 5th International
Tollenaar, M., Lee, E.A., 2002. Yield potential, yield stability, and stress tolerance in Crop Sci. Congress, Jeju, Korea.
maize. Field Crops Res. 75, 161–169. Wortmann, C.S., Dobermann, A., Ferguson, R.B., Hergert, G.W., Shapiro, C.A.,
van Duivenbooden, N., de Wit, C.T., van Keulen, H., 1996. Nitrogen, phos- Tarkalson, D.D., Walters, D.T., 2009. High-yielding corn response to applied
phorus and potassium relations in five major cereals reviewed in respect phosphorus, potassium, and sulfur in Nebraska. Agron. J. 101, 546–555.
to fertilizer recommendations using simulation modeling. Fert. Res. 44, Yang, H.S., Dobermann, A., Cassman, K.G., Walters, D.T., 2006. Features, applications,
37–49. and limitations of the Hybrid-Maize simulation model. Agron. J. 98, 737–748.

You might also like