Agriculture, Ecosystems and Environment 294 (2020) 106865

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment

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

Benchmarking impact of nitrogen inputs on grain yield and environmental T

performance of producer fields in the western US Corn Belt
Fatima A.M. Tenorioa, Eileen L. McLellanb, Alison J. Eaglec, Kenneth G. Cassmana,
Daryl Andersend,1, Marie Krausnicke, Russell Oaklundf, John Thorburng, Patricio Grassinia,*
a
Department of Agronomy and Horticulture, University of Nebraska–Lincoln, Lincoln, NE, 68583-0915, USA
b
Environmental Defense Fund, Washington, DC, 20009, USA
c
Environmental Defense Fund, Raleigh, NC, 27607, USA
d
Little Blue Natural Resources District, PO Box 100, Davenport, NE, 68335, USA
e
Upper Big Blue Natural Resources District, York, NE, 68467, USA
f
Lower Platte North Natural Resources District, PO Box 126, Wahoo, NE, 68066, USA
g
Tri Basin Natural Resources District, Holdrege, NE, 68949, USA

ARTICLE INFO ABSTRACT

Keywords: Benchmarking crop yields against nitrogen (N) input levels can help provide opportunities to improve N ferti-
Maize lizer efficiency and reduce N losses on maize in the US Corn Belt by identifying fields most likely to benefit from
Yield improved N management practices. Here, we evaluated a large producer database that includes field-level data
Nitrogen on yield and applied N inputs from 9280 irrigated and rainfed fields over a 7-year period (2009–2015) in
Nitrogen balance
Nebraska (USA). A spatial framework, based on technology extrapolation domains, was used to cluster each field
Fertilizer
into spatial units with similar climate and soil type that represent 1.3 million ha of US farm land sown annually
with maize. Three metrics were employed to evaluate agronomic and environmental performance: partial factor
productivity for N inputs (PFPN, ratio between yield and N inputs), N balance (difference between N inputs and
grain N removal), and yield-scaled N balance (ratio between N balance and yield). Nitrogen inputs included N
from fertilizer and N contained in applied irrigation water. Average yield and N inputs were 40 and 44% higher
in irrigated versus rainfed fields. The N balance was ca. 2-fold greater in irrigated versus rainfed fields (81 versus
41 kg N ha−1). Of the total number of field-years, 58% (irrigated) and 15% (rainfed) had N balance ≥ 75 kg N
ha−1, which was considered a threshold to identify fields with potentially large N losses. Very large (> 150 kg N
ha−1) and negative N balance estimates were not apparent when analysis was based on field averages using a
minimum of three years' data instead of individual field-years. Nitrogen balance was smaller for maize crops
following soybean compared to continuous maize. Despite the larger N balance (on an area basis), irrigated fields
exhibited smaller yield-scaled N balance relative to rainfed fields. The approach proposed here can readily be
adopted to benchmark current use of N fertilizer for other cereal-based crop systems, inform policy, and identify
opportunities for improvement in N management.

1. Introduction contained in applied irrigation water (Skaggs et al., 1995; Connor et al.,
2011). Synthetic N fertilizer accounts for ca. half of total N input to
Nitrogen (N) is an essential nutrient to support crop growth and a global cropland, and increasing N fertilizer use since the middle of the
key pillar for global food security (Cassman et al., 2002; Tilman et al., 20th century has been a major contributor to rapid increases in cereal
2002; Mueller et al., 2012). Sources of N that contribute to crop N crop yields (Cassman et al., 2002; Tilman et al., 2002; Foley et al.,
supply include synthetic fertilizer, organic fertilizer such as animal and 2011). Nitrogen inputs exceeding crop N requirements (i.e., N surplus)
green manure, biological N fixation, mineralization of soil organic can lead to large N losses via denitrification, leaching, volatilization,
matter, dry and wet atmospheric deposition, nitrate-N (NO3−-N) in and run-off, straining the capacity of the earth to meet humanity’s need
shallow water tables, and, in the case of irrigated agriculture, N for clean water, clean air, and abundant and healthy food (Matson

⁎
Corresponding author.
E-mail address: [email protected] (P. Grassini).
1
Current address: Lower Platte North Natural Resources District, PO Box 126, Wahoo, NE 68066, USA.

https://doi.org/10.1016/j.agee.2020.106865
Received 14 September 2019; Received in revised form 8 February 2020; Accepted 11 February 2020
Available online 27 February 2020
0167-8809/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/BY/4.0/).
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

et al., 1998; Erisman et al., 2013; Steffen et al., 2015). In contrast, N N balance approach is the framework for assessing N use or manage-
fertilizer inputs consistently below crop N requirements (i.e., N deficit) ment developed by the European Union Nitrogen Expert Panel that
can lead to soil N mining and reduced soil quality (Sanchez, 2002; considers (i) minimum amount of N input required for production; (ii)
Sanchez and Swaminathan, 2005). The challenge is to find an effective maximum N surplus that is environmentally acceptable; and (iii)
balance between N inputs and crop N requirements, to achieve high minimum and maximum N use efficiency, defining a “safe operating
crop productivity while preserving soil quality and reducing environ- space”, which shows the most desirable range for N output and N input
mental footprint (Zhang et al., 2015; Lassaletta et al., 2014). (EU-NEP, 2015). Other examples of application of the N balance ap-
Benchmarking N input use in individual fields for a large number of proach include whole-farm level assessments, including dairy farms
cohort fields may help identify fields with greatest opportunities to (Schröder et al., 2003; Spears et al., 2003; Cela et al., 2014). Finally, N
improve productivity and reduce overall environmental impact. balance can also be expressed per unit of yield (hereafter referred to as
However, we are not aware of previous studies that used actual field- ‘yield-scaled N balance’) to recognize the different land requirements
level data to benchmark the efficacy of N inputs to produce grain while associated with low- and high-yield cropping systems to meet a given
minimizing N losses to the environment. Instead, studies addressing production goal (Schröder et al., 2003; Grassini and Cassman, 2012).
both productivity and environmental performance of agro-ecosystems Estimating N balance and other N-metrics in producer fields can help
in relation to N inputs can roughly be grouped in two categories. The understand potential N losses in current agro-ecosystems, on a per-area
first category includes the large number of studies conducted in ex- and per-output basis and, in the case of fields with large N balance and
perimental plots or field trials in which researchers selectively applied low PFPN, identify opportunities for improvement via better crop and
different N input levels or management practices and carefully mea- soil management practices (Cassman, 2017; McLellan et al., 2018).
sured yield and N losses (e.g., Harmel et al., 2008; Venterea et al., To establish a baseline and determine the variability among maize
2012). The second category includes in-silico modeling studies at re- fields in both production and environmental outcomes related to N
gional and global levels (e.g., Van Drecht et al., 2003; Howarth et al., input use, we developed an approach using producer-reported data, a
2006). In between these two extremes, we found few studies that ex- combination of N-metrics (PFPN, N balance, and yield-scaled N bal-
plicitly aimed to benchmark on-farm yield and N input use (e.g., Khanal ance), and a spatial framework to cluster fields into near-similar cli-
et al., 2014; Lassaletta et al., 2014; Basso et al., 2019). However, most mate-soil domains. We used Nebraska (NE), USA as a study case—a
of these studies have relied on N fertilizer use data reported at a high state that produces 43 MMT of maize annually in ca. 4 million ha
level of spatial aggregation (e.g., country, state). The major reason for (USDA-NASS, 2014–2018). The assessment was based on a large data-
scarcity of field-level studies is lack of producer data on yield and N base including field-level data on yield and N fertilizer rates collected
inputs. For example, for the Corn Belt, a large region in the north- from irrigated and rainfed maize over multiple years (total of 9280
central USA that produces ca. one third of global maize production, field-year observations). Specific objectives were to (i) determine cur-
data on N fertilizer rates applied to maize are available only at the state rent PFPN, N balance, and yield-scaled N balance for irrigated and
level, and at 5-year intervals (USDA-ERS, https://data.ers.usda.gov/ rainfed maize; (ii) evaluate the sensitivity of these N-metrics as a result
reports.aspx?ID=17883). Due to the lack of more detailed data, some of different levels of spatial and temporal aggregation (field averages,
studies have attempted to generate predictions of N fertilizer for small year averages, and individual field-year observations); and (iii) assess
regions or even individual fields following tortuous methods (e.g., fer- the influence of water regime and crop sequence on yield, N inputs, and
tilizer sales records, university-based N recommendations), but such N-metrics as a first step towards understanding how management
predictions have not been validated for their ability to reproduce actual practices affect these N performance metrics.
N fertilizer rates in producer fields (Khanal et al., 2014; Basso et al.,
2019). 2. Materials and methods
Accurate assessments of both the current situation and opportunities
for improvement require cost-effective approaches for evaluating on- 2.1. Study region, on-farm database, and field grouping based on climate
farm yield and environmental footprint in relation to N inputs, thus and soil
enabling identification of fields with poor N use efficiency. For such an
approach to be feasible, it would need to rely on a small number of The United States accounts for 28% of global maize production
parameters that are readily available from producers. To that end, we (FAOSTAT, 2013-2017). About 90% of maize in the USA is produced in
evaluated three metrics related to agronomic and environmental per- the north-central region, commonly referred to as the “Corn Belt”,
formance (hereby called ‘N-metrics’): partial factor productivity for N where maize is grown as monoculture or in a 2-y rotation with soybean
inputs from fertilizer and irrigation water (PFPN), N balance, and yield- (Grassini et al., 2014). Nebraska ranks third among USA maize pro-
scaled N balance. The PFPN – the ratio between grain yield and the ducing states, with irrigated area accounting for ca. 58% and 65% of
amount of applied N inputs (Cassman et al., 1996) – represents an N total NE maize cropland and production, respectively (USDA-NASS,
fertilizer efficiency metric and only requires data on yield and N inputs. 2014–2018). Nebraska is divided into 23 Natural Resources Districts
However, while PFPN provides an indication of N fertilizer efficiency (NRDs; www.nrdnet.org), with each NRD serving as a government en-
for grain production, it tells little about potential environmental impact tity authorized to establish regulations to conserve water and soil re-
and long-term sustainability of the resource base. It may also give a source quality and quantity (Exner et al., 2010; Ferguson, 2015). Some
biased assessment of agronomic performance of the cropping system. of the NRDs require producers with fields located within their bound-
For example, high PFPN values can result from a combination of low aries to report field-level data on yield and applied inputs every year. In
yields and nil N inputs; if this situation continues over time, it would the present study, we used data reported from maize fields located in
invariably lead to soil N mining, loss of soil quality, and, at scale, a four NRDs: Little Blue, Lower Platte North, Tri-Basin, and Upper Big
deficient cereal supply. Another metric is the partial N balance (here- Blue (Fig. 1). Producer-reported data included field location (township,
after simply referred to as ‘N balance’), which is defined as the differ- range, and section), maize yield (at standard moisture content of 155 g
ence between N inputs and grain N removal (Treacy et al., 2008; H2O kg−1 grain), N fertilizer rate, irrigation amount, some manage-
Oenema et al., 2012; McLellan et al., 2018). As in the previous example, ment practices (previous crop, irrigation system type, and manure ap-
a persistent negative N balance over time would invariably lead to soil plication), and NO3−-N concentration contained in applied irrigation
N mining. In contrast, a large N balance is a strong indicator of po- water. The database included irrigated and rainfed fields sown with
tentially large N losses. For example, in the case of maize, N losses maize during seven crop seasons (2009–2015) with contrasting weather
increase exponentially when N balance exceeds 75 kg N ha−1 (Zhao conditions. For example, 2012 exhibited warmer and dry conditions,
et al., 2016; McLellan et al., 2018). An example of the application of the with seasonal temperature and total rainfall averaging 22°C and 202

2
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

Fig. 1. Location of irrigated (blue dots; n = 8413 field-years)

and rainfed maize fields (red dots; n = 867 field-years). Fields
were grouped into two ‘technology extrapolation domains’
(TED 1 & 2) based on climate and soil similarity and fields in
TED2 were further grouped based upon water regime (irri-
gated [2I] and rainfed [2R]), resulting in three TED-water
regime combinations (TED 1I, 2I and 2R). Note that all fields
in TED1 were irrigated. Stars indicate location of the me-
teorological stations. Inset shows maize harvested area (in
green; USDA-NASS, 2017) and location of area of interest
within Nebraska (NE). Lines show borders of NE Natural
Resources Districts (NRDs). Producer data from four NRDs
were used for the present study: Little Blue (LB), Lower Platte
North (LPN), Tri Basin (TB), and Upper Big Blue (UBB). (For
interpretation of the references to colour in this figure legend,
the reader is referred to the webversion of this article.)

mm, respectively, across the study area. In contrast, 2014 was cooler description of the TED spatial framework is available at http://www.
and wet, with seasonal temperature and total rainfall averaging 20°C yieldgap.org/web/guest/cz-ted. For our analysis, we grouped fields
and 544 mm, respectively. Water table depth was consistently below into two TEDs (TED 1 and 2) which, together, account for ca. 1.3
the rooting depth across the region where the reporting fields were million ha land in the US sown with maize every year. Both TEDs have
located. high temperature seasonality and same GDD range (i.e., 3792–4829
Field boundaries were mapped using Google Earth® based on the °Cd). In contrast, TED 1 had higher PAWHC (300–350 versus 250−300
field location as provided by the NRDs. Associated data were screened mm) and higher water limitation (i.e., lower aridity index) compared to
for erroneous and incomplete entries, using quality control measures TED 2. The TED 1 only included irrigated fields, while TED 2 included
that set acceptable ranges for yield, N inputs, and applied irrigation. For both irrigated (I) and rainfed fields (R), which were disaggregated for
example, fields that reported maize yields > 20 Mg ha−1, N fertilizer the analysis. Hence, fields were grouped into three TED-water regime
amounts > 350 kg N ha−1, and/or applied irrigation > 1200 mm were (TED-WR) combinations: TED 1I, TED 2I, TED 2R. After applying
excluded from the database (ca. 0.1% of total observations). Fields re- quality control measures and grouping the fields into the three TED-
ceiving manure application were excluded because (i) on average, only WRs, the database contained a total of 9280 field-year observations; of
5% of maize fields in NE receive manure (USDA-ERS, 2005), and (ii) it these, 91 and 9% corresponded to irrigated and rainfed fields, respec-
is difficult to estimate the release and amount of N from applied manure tively. On average, there were 511, 691 and 124 fields per year in TEDs
(van Kessel and Reeves, 2002). Only pivot-irrigated fields were con- 1I, 2I, and 2R, respectively.
sidered for our study as surface (flood) irrigation accounts for a small
fraction of irrigated maize area in NE (ca. 14%) and its area has steadily 2.2. On-farm data quality assessment
declined over time (USDA-ERS, 2010). Because the majority (> 85%)
of maize across the US Corn Belt region is grown continuously or in a Previous studies have found that NRD producer-reported data
maize-soybean rotation (Farmaha et al., 2016), fields sown with maize aligned well with data collected by other independent sources (Grassini
after wheat, alfalfa, or other crops besides maize and soybean were et al., 2014). In this study, we further evaluated the quality of the NRD
excluded from the analysis. Our study only included fields sown with data by comparing average annual N fertilizer rates and yield derived
maize for grain; other maize fields sown for seed production or silage from the NRD database against independent estimates from both pro-
were excluded. The group of reporting fields remained the same during ducer survey data (Grassini et al., 2015; Gibson et al., 2019) and official
the 2009–2015 time period in the four NRDs. statistics (USDA-NASS, http://quickstats.nass.usda.gov/; USDA-ERS,
A robust comparison of producer fields in terms of yield, N inputs, https://data.ers.usda.gov/) for each TED-WR. Survey data included two
and N-metrics requires grouping fields based on those factors with crop seasons (2010 and 2011) and a total of 48 field-years located
greatest influence on yield potential, yield stability and, indirectly, on within the same TED-WRs (1I, 2I, and 2R). For consistency, we used the
nutrient cycling and other variables influencing crop responses to N 2010–2011 time period for all yield and fertilizer paired comparisons.
inputs. In the present study, maize fields were grouped into technology Unfortunately, USDA ERS data on N fertilizer amount for irrigated and
extrapolation domains (TEDs; Rattalino Edreira et al., 2018). Briefly, a rainfed maize were aggregated at state level and only available for
TED corresponds to a unique combination of annual growing-degree 2010; hence, the comparison between average N rate comparison
days (GDD), aridity index (ratio between precipitation and reference against the NRD database was performed at different levels of spatial
ET), temperature seasonality (as quantified with standard deviation for aggregation.
monthly temperature), and plant available water holding capacity
(PAWHC). Within a defined region, such as the US Corn Belt, the TED
2.3. Retrieval of weather and soil data and simulation of yield potential for
framework categorizes soils into cohort groups, within which climate
each TED-water regime in each year
and soils are of sufficient similarity that crop responses to management
practices (including N fertilizer) are expected to be similar. Detailed
Weather and soil data were retrieved to assess differences among

3
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

selected TED-WR combinations. Averages of weather variables re- expensive and laborious measurements. Hence, we focused on those N
trieved for each TED-year during the crop season (emergence to phy- inputs that account for the largest fraction of total N inputs and that are
siological maturity) were calculated for the 2009–2015 time period per readily available from producer fields. In our study, we excluded fields
TED. Average dates of emergence and physiological maturity in each receiving manure application as this is not a common practice in NE. In
year were simulated using Hybrid-Maize model (Yang et al., 2004, the case of atmospheric N deposition, NE is situated far from industrial
2017) based on average sowing date and hybrid maturity data available areas and overall annual N deposition has been estimated to be very
for each TED-WR (Morell et al., 2016; Gibson et al., 2019) and mea- small (< 10 kg N ha−1; NADP, USDA-REEIS, https://reeis.usda.gov/
sured daily weather data from three or four meteorological stations web/crisprojectpages/1007486-the-national-atmospheric-deposition-
located within each TED (Fig. 1). Weather variables included incident program-nadp.html). A key question is the relationship between N
solar radiation, minimum and maximum temperature (Tmin and Tmax, mineralization and immobilization, which may differ for a given field-
respectively), precipitation, and Penman-Monteith grass-referenced year. However, given the stable stoichiometry between C and N, so long
evapotranspiration (ETo; Allen et al., 1998). Soil variables including as SOM content does not change over time, the magnitude of these two
percentage of soil organic matter, PAWHC, and topographic wetness processes would converge over the long-term. Because SOM is near
index (TWI) for each field were retrieved from Soil Survey Geographic steady state in the US Corn Belt (Baker and Griffith, 2005; Verma et al.,
database (SSURGO, https://websoilsurvey.nrcs.usda.gov). PAWHC re- 2005; Dolan et al., 2006; Blanco-Canqui and Lal, 2008), we assumed N
presents the amount of water (mm) that the soil can hold between field released from SOM mineralization (which includes the inorganic soil N
capacity and wilting point within the rootable depth. TWI indicates the at sowing) to be similar to soil N immobilization over the full field-year.
likelihood of surface runoff (run-on) from (to) an area based on slope In contrast, the amount of N contained in applied irrigation water
and surrounding area, with bottom and upland areas having highest (hereafter referred to as “N irrigation”) cannot be neglected for irri-
and lowest values, respectively (Sørensen et al., 2006). gated fields (Grassini et al., 2014; Ferguson, 2015). Hence, we con-
Yield potential (Yp) is defined as the yield attained by an adapted sidered N from both fertilizer and applied irrigation water for our cal-
crop cultivar when grown with non-limiting nutrient and water supplies culation of PFPN, N balance, and yield-scaled N balance.
and with pests and diseases effectively controlled (Evans, 1993; van Nitrogen added via irrigation was calculated from reported irriga-
Ittersum et al., 2013). Water-limited yield potential (Yw) is influenced tion amount and NO3−-N concentration in groundwater. For field-years
by the same factors that define Yp but also determined by precipitation with no data to estimate N irrigation (because irrigation amount and/or
amount and distribution and soil properties that influence water NO3−-N concentration were not available), we used the average N ir-
availability such as PAWHC and field slope. In our study, we estimated rigation calculated for other fields located within the same TED-WR-
Yp and Yw for three purposes. First, the ratio between Yw and Yp year. Because field-level irrigation amounts were not reported for TED
provides an objective estimate of the degree of water limitation in 2I, we estimated a constant irrigation amount for all fields within a
rainfed versus irrigated fields in TED 2. Second, comparison of average TED-year, using the relationship between seasonal water deficit and on-
producer yield against simulated Yp (irrigated fields) or Yw (rainfed farm irrigation amount for silt loam soils reported by Gibson et al.
fields) provides an estimate of the yield gap (difference between pro- (2018) for the same region. In the case of NO3−-N concentration, we
ducer yield and Yp or Yw), which is useful to understand yield per- used the average value estimated across fields in TED2. We note that N
formance in relation to the N balance for a given field-year. For ex- irrigation accounts for a relatively small portion of the N inputs (ca.
ample, a large yield gap and a large N balance suggests an opportunity 11%), so the estimation of N irrigation for TED 2I is unlikely to bias
to produce more yield with the same or even smaller N balance. Third, results.
expressing producer yield as a percentage of the Yp (or Yw) for a given Partial factor productivity for N inputs (PFPN) was calculated as the
TED-WR-year (hereafter referred to as ‘relative yield’) allows a fair ratio between yield and N inputs. The N balance was calculated as the
comparison of producer yields and N balance across years with con- difference between N inputs and grain N removal. Maize grain N re-
trasting weather conditions, which is critical in the case of rainfed fields moval was estimated based on producer yield and grain nitrogen con-
that depends on the erratic fluctuation in precipitation amount and centration (GNC, at standard 15.5% grain moisture content). The latter
distribution across years. was estimated for each field-year using the predictive model developed
We used Hybrid-Maize model (Yang et al., 2006, 2017) to estimate by Tenorio et al. (2019) for maize in the US Corn Belt. We note that the
Yp (irrigated) and Yw (rainfed) for each TED-WR-year combination. goal is not to achieve zero N balance because that would lead to mining
Hybrid-Maize model has been widely evaluated for its ability to esti- of soil organic matter. Instead, here we used a threshold of 75 kg N
mate yield potential in well-managed crops that grew without nutrient ha−1 to identify fields with large N balance and, hence, potentially
limitations and kept free of biotic stresses (Yang et al., 2004; Grassini large N losses (Zhao et al., 2016; McLellan et al., 2018). Using data from
et al., 2009a). Because the goal was to estimate the maximum possible individual field-years may give a biased assessment of producer per-
yield that results from the best possible management in each TED-WR, formace in relation with applied N inputs. For example, a severe
we selected the combination of sowing date, hybrid maturity, and plant drought (e.g., year 2012) would reduce yield and lead to a relatively
density that give the highest yield in each TED-WR based on previous large N balance in rainfed fields. Likewise, a severe soil mining can be
survey data (Farmaha et al., 2016; Gibson et al., 2019). Data inputs and (wrongly) inferred from a field that (purposely) received little N ferti-
model parameters used to simulate Yp or Yw are shown in Supple- lizer in an specific year because of large residual soil N from previous
mentary Tables S1-S2. Producer yield exceeded simulated Yp (or Yw) in crop as measured using soil nitrate tests. To evaluate the degree to
4% of the total field-year observations, likely due to inaccuracies in which our estimates of N balance may be biased due to the aformen-
weather, soil, or producer yield data. For the purposes of this analysis, tioned factors, we calculated the N balance at three different levels of
relative yield was set at one when producer yield exceeded Yp (or Yw). aggregation: (i) individual field-years, (ii) individual fields with N
balance averaged across years, and (iii) individual years with N balance
2.4. Calculation of partial factor productivity for nitrogen (N) inputs, N averaged across fields. In the case of (ii), we included only those fields
balance, and yield-scaled N balance with at least three years of data. Finally, the yield-scaled N balance was
calculated as the ratio between N balance and producer yield.
The N inputs included N from synthetic fertilizer, applied irrigation Frequency distributions were used to assess variation in yield, N
water (in the case of pivot-irrigated fields), manure, atmospheric dry inputs, and N-metrics. Deviation from normality was tested using
and wet deposition, inorganic soil N at sowing, and soil organic matter D’Agostino-Pearson normality test. In addition, a three-way analysis of
(SOM) mineralization during the crop season. Quantification of all N variance (ANOVA) was used to quantify the influence of TED-WR, year,
input sources for a large population of producer fields would require previous crop, and their interactions at explaining observed variation

4
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

on yield, N inputs, N balance, PFPN, and yield-scaled N balance. Table 2

Proportion of sum of squares (%SS) attributable to each term was Comparison for yield and N fertilizer among the Natural Resources District
computed after excluding the error. Mean contrasts were used to assess (NRD) database (this study), independent survey producer data (Grassini et al.,
the overall effect of water regime and crop sequence on the different 2015; Gibson et al., 2019), and official statistics (National Agricultural Statistics
Service [USDA-NASS]; Economic Research Service [USDA-ERS]) for each
parameters. Tukey’s test was used to determine statistically significant
technology extrapolation domain x water regime (TED-WR) combination. Va-
differences among averages (α = 0.05). To evaluate biases in the
lues are 20102011 averages, except for average N fertilizer reported by ERS,
analysis due to data imbalance among TED-WRs, the ANOVA was re-
which corresponds to an average statewide value reported for year 2010. Par-
peated 5x using resampling of equally 300 observations in each TED- enthetic values indicate the range. Note that a single aggregated value was
WR-previous crop combinations to obtain a balanced experimental available from USDA-NASS; hence, the range is not shown.
design. The test indicated that using either a balanced versus un-
TED-WR Yield (Mg ha−1)
balanced number of observations or different subsets of randomly se-
lected fields had little impact on the results. Hence, we reported only NRD Survey NASS/ERS
the results derived from the ANOVA using the entire database. Yield
versus N balance plots were assessed to determine the frequency of 1I 12.7 (6.3–17.6) 12.8 (10.7–15.0) 12.4
2I 12.5 (6.3–17.3) 13.0 (9.9–15.1) 12.0
fields with small or large N balance and low or high yield. The analysis
2R 8.9 (1.9–13.8) 8.8 (6.4–9.7) 8.8
was also performed using relative yield (as % of Yp or Yw) to account N fertilizer rate (kg N ha−1)
for weather variation across years, TEDs, and WR. Fields were subse- 1I 210 (84–325) 218 (168–246) 189*
quently grouped in four categories: (A) high relative yield, N bal- 2I 204 (78–325) 215 (157–263)
ance < 75 kg N ha−1; (B) low relative yield, N balance < 75 kg N ha−1; 2R 138 (76–246) 144 (140–151) 131*

(C) high relative yield, N balance ≥ 75 kg N ha−1; (D) low relative

* Significantly different from mean NRD estimate (t-test, α = 0.05).
yield, N balance ≥ 75 kg N ha−1. Following Lobell et al. (2009) and
van Ittersum et al. (2013), we used 80% and 70% of Yp and Yw as
contrast, average statewide N fertilizer data reported through NASS was
thresholds to distinguish high versus low yields in irrigated and rainfed
7–10% (irrigated) and 5% (rainfed) lower than average N fertilizer rate
fields, respectively. These values represent reasonable yield goals, with
as reported to the NRDs (t-test, p < 0.05). Inclusion of other regions of
the smaller yield goal in the case of rainfed crops aiming to account for
NE with lower maize yields and, probably, lower fertilizer N amounts in
higher production risk associated with erratic rainfall across years.
the calculation of the statewide NASS average may explain these dif-
ferences. Indeed, our study area has slightly higher average irrigated
3. Results and rainfed maize yields (13.4 and 9.3 Mg ha−1, respectively) com-
pared with the state averages (12.7 and 9.0 Mg ha−1; USDA-NASS,
3.1. On-farm yield, N inputs, and N-metrics across climate x soil x water 2013–2017).
regime domains Average producer yield represented ca. 81% of simulated Yp for
irrigated fields and ca. 70% of Yw for rainfed crops (Table 3). Yield
Averages of meteorological variables were similar between the three potential was 5% higher in TED 1I versus TED 2I due to longer crop
TED-WRs, except for ETo, which tended to be higher in TED 1 versus cycle length (i.e., days from emergence until physiological maturity) in
TED 2 (Table 1). Maize fields exhibited higher PAWHC and TWI, and the former as a result of lower seasonal temperature (Table 1). The Yw
lower soil organic matter, in TED 1 compared with TED 2. Within TED for rainfed maize in TED 2 was ca. 30% lower and three times more
2, the slightly higher soil organic matter and TWI in irrigated fields variable compared with the simulated Yp for irrigated maize in the
compared with rainfed fields is likely related both to historical selection same TED. Frequency distributions for producer yield in irrigated and
of the best soils for pivot irrigation installation and to greater crop rainfed fields were negatively skewed, with the majority of the fields
residue return with higher-yield irrigated production. Weather and soil closer to highest yields (Fig. 2a, c). TED-WR had the greatest influence
parameters exhibited relatively small year-to-year and field-to-field on yield and N inputs, accounting for 70–90% of SS excluding the error,
variation, respectively, as indicated by their respective coefficients of with the rest of the modelled variation mostly explained by year, TED-
variation (CVs ≤ 16%); total precipitation was an important exception, WR x year interaction, and, in the case of N inputs, also by previous
exhibiting large variation across years in both TEDs (CVs = 30–35%) crop (Table 4). This result was expected as the TED-WR stratification
(Table 1). aimed to account for differences in climate, soil, and water supply be-
Averages for NRD yield and N fertilizer were in reasonable agree- tween regions and water regimes. Average producer yield was ca. 40%
ment with estimates derived from independent survey data, collected lower (and 5x more variable) in rainfed versus irrigated fields (Table 3).
from fields located in same NRD, with differences among databases Consistent with the yield difference, average N input was 44% higher in
within ± 4% of NRD averages and statistically not different from zero irrigated versus rainfed fields (Fig. 2b, d). In contrast to crop yield, the
(t-test, p > 0.10) (Table 2). Similarly, there was good agreement be- degree of inter-annual variation for N inputs was identical for both
tween NRD and NASS maize yields as indicated by the lack of statisti- water regimes (CV = 6%). Distribution of field-level N inputs was
cally significant difference between databases (t-test, p > 0.15). In

Table 1
Averages for weather and soil variables for each technology extrapolation domain x water regime (TED-WR) combination. Averages for weather variables during the
2009-2015 period were computed based on seasonal (emergence-to-physiological maturity) values, while averages for soil variables were computed based on the
values retrieved for each individual field. Parenthetic values indicate inter-annual and field-to-field coefficient of variation (in %) for weather and soil variables,
respectively. Different letters indicate statistically significant (p < 0.05, Tukey’s test) differences among TED-WRs. Means of weather variables for irrigated and
rainfed fields located in TED 2 were identical; hence, a single mean for each weather variable is shown for fields in TED 2.
TED Water regime Solar radiation (MJ m-2d-1) Tmin (ºC) Tmax (ºC) Total ETo (mm) Total rainfall (mm) Soil organic matter (%) PAWHC (mm) TWI

a a a a a c a
1 Irrigated 20.9 (7) 13.0 (6) 27.0 (6) 720 (5) 371 (35) 1.9 (10) 313 (3) 10.3 (4)a
2 Irrigated 20.9 (6)a 14.1 (7)a 27.6 (5)a 673 (5)b 393 (31)a 2.5 (13)a 285 (3)b 10.1 (6)b
Rainfed 2.4 (16)b 287 (3)b 9.9 (6)c

Tmin: minimum temperature; Tmax: maximum temperature; ETo: grass-based reference evapotranspiration; PAWHC: plant available water holding capacity; TWI:
topographic wetness index.

5
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

Table 3 normally distributed in irrigated fields but positively skewed in rainfed

Average 7-y (2009- 2015) producer yield, yield potential (Yp; irrigated crops) fields, indicating that a relatively smaller number of fields received
or water-limited yield potential (Yw; rainfed crops), and relative yield (ratio much larger N inputs than the rest of the fields. In irrigated fields, N
between producer yield and Yp or Yw) for each technology extrapolation do- fertilizer exhibited a negatively skewed distribution (skewness =
main x water regime (TED-WR) combination. Parenthetic values indicate the
-0.19). Average N irrigation represented 11% of the N input in irrigated
inter-annual coefficient of variation (in %).
fields, exhibiting larger inter-annual variation compared with N ferti-
TED-WR Producer yield (Mg ha−1) Yp or Yw (Mg ha−1) Relative yield lizer (CV = 41 versus 4%) as a result of variation in irrigation amounts
across years in response to water demand as affected by weather
1I 13.9 (8)a 17.1 (7)a 0.81 (11)a
2I 13.3 (4)b 16.6 (10)b 0.80 (9)b
(Fig. 2b, inset).
2R 8.2 (28)c 11.8 (31)c 0.69 (13)c Frequency distribution for N balance and PFPN showed contrasting
patterns between water regimes: the N balance was negatively and
Different letters indicate statistically significant differences among TED-WRs positively skewed in irrigated and rainfed fields, respectively (Fig. 3a,
(p < 0.05, Tukey’s test). d) while PFPN exhibited the inverse trend (Fig. 3b, e). However, yield-

Fig. 2. Frequency distributions for producer yield

(left) and N inputs (right) in irrigated (a, b); and
rainfed fields (c, d). Number of field-years were 8413
(irrigated) and 867 (rainfed). Average (solid line),
maximum and minimum (dashed lines) annual simu-
lated yield potential (irrigated) or water-limited yield
potential (rainfed) are shown. Inset in panel b shows
averages for N fertilizer and N irrigation. Average
(X ± standard error) and skewness (S) are shown.
Irrigated data from technology extrapolation domains
(TEDs) 1 and 2 were pooled as frequency distributions
for yield and N inputs were almost identical.

Table 4
Analysis of variance for the effect of technology extrapolation domain x water regime combination (TED-WR), year, previous crop, and their interactions on yield,
nitrogen (N) inputs, N balance, partial factor productivity for N inputs (PFPN), and yield-scaled N balance. Mean difference of each parameter for soybean-maize
versus continuous maize and irrigated versus rainfed fields in TED2 are shown.
Percentage of total sum of squares (%SS)†

Source of variation d.f. Yield ‡ (Mg N inputs ‡

(kg N N balance ‡
(kg N PFPN ‡
(kg grain kg−1 N balance per yield
ha−1) ha−1) ha−1) N) unit‡ (kg N Mg−1 grain)

TED-WR 2 72*** 87*** 49*** 5*** 7***

Year 6 14*** 3*** 16*** 48*** 41***
Previous crop 1 < 1*** 5*** 23*** 17*** 5***
TED-WR x year 12 13*** 4*** 9*** 26*** 39***
TED-WR x previous crop 2 < 1*** < 1*** 1*** <1 < 1**
Year x previous crop 6 < 1*** <1 < 1* 1*** 3***
TED-WR x year x previous crop 12 < 1*** < 1* 1*** 2*** 4***
Mean estimate difference
previous crop (soybean vs maize) 0.2*** −20*** −22*** 7*** −2***
water regime (irrigated vs rainfed in TED 2) 5*** 96*** 35*** −4*** −3***
Model sum of squares 41 27900 8825481 2662321 366282 85024
Error sum of squares 9238 21821 8731687 9219978 1078117 155610

†
Proportion (in %) of total sum of squares (SS) after excluding error.
‡
F-test significant at P < 0.05*, < 0.01**, and < 0.001***.

6
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

Fig. 3. Frequency distributions for nitrogen (N) balance (left), partial factor productivity for N inputs (PFPN, center), and yield-scaled N balance (right) in irrigated
(a, b, c) and rainfed fields (d, e, f). Number of field-years were 8413 (irrigated) and 867 (rainfed). Average (X ± standard error) and skewness (S) are shown. Irrigated
data from technology extrapolation domains (TEDs) 1 and 2 were pooled as frequency distributions for yield and N inputs were almost identical.

scaled N balance was positively skewed in both water regimes (Fig. 3c, agronomic practices. About 41 and 52% of the irrigated and rainfed
f), indicating that a relatively small number of fields in irrigated (1%) fields fell into the low relative yield categories (i.e., below 80% and
and rainfed (9%) exhibited very large yield-scaled N balance (> 15 kg 70% of Yp and Yw, respectively, categories B and D in Fig. 5), in-
N Mg−1 grain). dicating room to further increase yields within the observed range of N
The TED-WR term of our ANOVA explained ca. half of the modelled balance (Fig. 5). Of particular concern are those fields exhibiting large
variation in N balance; the rest of the variation was accounted for by N balance and low relative yield (category D), representing 29 and 3%
year, TED-WR x year, and previous crop (Table 4). In contrast, TED-WR of total irrigated and rainfed fields, respectively. Attaining high yields
explained a small portion of modelled variation in PFPN and yield- with a smaller N balance (category A) is a realistic goal: 26% and 47%
scaled N balance (< 10%), with most variation accounted for by year, of irrigated and rainfed fields, respectively, exhibited N balance < 75
TED-WR x year, and, in the case of PFPN, by previous crop as well. The kg N ha−1 and attained or even exceeded their respective yield goals.
large portion of unaccounted variation in N balance, PFPN, and yield-
scaled N balance (78, 75, and 65% of total SS, respectively) suggests 3.3. Yield, N inputs and N-metrics as influenced by TED, water regime, and
that magnitude of field-to-field variation was as important as variation previous crop
due to TED-WR, year, previous crops, and their interactions.
Average N input rates were 44% larger in irrigated versus rainfed
3.2. Benchmarking yield and N balance in producer fields fields, but higher yields in irrigated fields meant that PFPN was re-
markably similar between water regimes (Table 4, Fig. 3). And while N
Similar to the observed pattern in average yield, the average N balance was 49% larger in irrigated versus rainfed fields, yield-scaled N
balance, calculated using all field-year observations, decreased in the balance was smaller in irrigated fields. For a given TED-WR, yield and N
following order: TED 1I (86 kg N ha−1), 2I (77 kg N ha−1) and 2R (38 inputs were 2% lower and 10% larger, respectively, in maize after
kg N ha−1) (Fig. 4a–c). About 61, 54, and 15% of the field-years in TED maize versus maize after soybean (Table 4, Fig. 6). As a result, PFPN and
1I, 2I, and 2R, respectively, exhibited N balance ≥ 75 kg N ha−1. N balance was higher and lower, respectively, in maize after soybean
Results were similar when field averages (i.e., averages for each field compared to maize after maize. Consistent with these results, frequency
based on at least 3 years of data) were used for the analysis instead of of fields with N balance ≥ 75 kg N ha−1 was lower in soybean-maize
individual field-year observations (Fig. 4d–f), except that the range of N than in maize-maize: 40% versus 72% (irrigated fields) and 13% versus
balance narrowed considerably. For example, cases with very large N 20% (rainfed fields).
balance (> 150 kg N ha−1) or negative N balance were not apparent
when the analysis was based on field averages instead of field-years. 4. Discussion
Average annual N balance did not vary substantially among years in
the case of irrigated maize (CV = 10%) (Fig. 4g, h). In contrast, rainfed Benchmarking crop yields against external input use provides in-
maize exhibited a large year-to-year variation (CV = 46%), with larger sight about opportunities to increase producer profit while using the
(smaller) N balance corresponded to years with lower (higher) yield same or less amount of input. There are many examples using this ap-
(Fig. 4i). For instance, highest N balance in TED 2R (rainfed) occurred proach in the literature. For example, in a classic study, French and
in 2012, which corresponded to a drought year with very low yield. The Schultz (1984) developed a boundary function for the relationship be-
year-to-year variation in N balance in irrigated fields was mostly due to tween yield and seasonal water supply for wheat in Australia; these
variation in N irrigation (CV = 34–56%), but not in N fertilizer (CV = authors documented large variation in yield across a wide range of
3–5%). water supply, which was attributable to management. This framework
Analysis of yield variation across field-years, for a given N balance has been subsequently used in a multitude of studies to assess crop
level, is confounded by year-to-year variation in weather. Expressing water productivity and identify opportunities for improvement (Sadras
producer yields as percentage of Yp (irrigated fields) or Yw (rainfed and Angus, 2006; Passioura, 2006; Grassini et al., 2009b, 2011). As far
fields) using field averages allows an objective assessment of available as we know, Hochman et al. (2014) is the only study that used a similar
room for improving yield at a given N balance level through better approach to benchmark crop yields in relation with N inputs. These

7
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

Fig. 4. Nitrogen (N) balance and producer maize yield in irrigated fields in TED 1 (1I; left) and TED 2 (2I; center) and rainfed fields in TED 2 (2R; right). Each
datapoint represents a field-year observation (a, b, c), field averages based on 3 years of data or more (d, e, f), and annual averages based on all fields in a given year
(g, h, i). Horizontal arrows indicate N balance =75 kg N ha−1, which was used as a threshold to identify fields with large N balance. Average yield (Ya) and N
balance are shown (and indicated with blue crosses). Percentage of field-years (a, b, c), field averages (d, e, f), and years (g, h, i) with N balance ≥ 75 kg N ha−1 is
also shown. n = number of observations.

authors presented an input-yield production frontier that benchmarked TED-WR, crop sequence, and their interactions (ca. 75 versus 25%, re-
the efficiency of applied N fertilizer in terms of crop production; spectively; Table 4). Similarly, although fields were grouped into TED-
however, the approach had a (data-intensive) modeling component to WRs, and N balance was averaged across years, there was still large
estimate crop N requirement and did not explicitly focus on assessing variation in N balance at any given yield level and vice versa. For in-
potential N losses or estimating the N balance. In contrast, our study stance, at a yield level of ca. 13 Mg ha−1, the N balance in irrigated
provides a cost-effective approach to benchmark yields in relation to N fields varied from 20 to 150 kg N ha−1 (Fig. 4). While some of this
balance of individual producer fields using several readily-available variation can be attributed to remaining spatial and temporal variation
parameters. in climate and soil within each TED-WR combination, these findings
At issue is the degree to which the observed variation in N balance suggest that management practices likely have a large influence on on-
across producer fields is attributable to variation in agronomic man- farm N balance. It is still uncertain, however, how much of that var-
agement. Our study showed that field-to-field variation in N balance iation is manageable through cost-effective agronomic technologies. In
was much larger than the portion of variance accounted for by year, this regard, a key challenge to improved N fertilizer efficiency is that

Fig. 5. Relative yield and nitrogen (N) balance in irrigated fields in TED 1 (1I; left) and TED 2 (2I; center) and rainfed fields in TED 2 (2R; right). Relative yield was
calculated based on producer yield expressed as percentage of yield potential (Yp; irrigated) or water-limited yield potential (Yw; rainfed). Each datapoint represents
a field average based on at least 3 years of data. Vertical line indicates N balance = 75 kg N ha−1, which was used as a threshold to identify fields with small and
large N balance. Horizontal lines indicate 80% and 70% of Yp and Yw, which are reasonable yield goals for irrigated and rainfed fields, respectively. Frequency of
fields in each of four (yield x N balance categories) combinations are shown.

8
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

Fig. 6. Average producer yield, nitrogen (N)

fertilizer rate, N inputs, N balance, partial
factor productivity for N inputs (PFPN), and
yield-scaled N balance in the three technology
extrapolation domain-water regime (TED-WR)
combinations: irrigated TED 1(1I), irrigated
TED 2 (2I), and rainfed TED 2 (2R). Separate
averages are shown for fields sown with maize
after maize (empty bars) or after soybean
(solid bars). Averages were calculated based on
annual averages, with vertical lines indicating
the standard errors. Different letters indicate
statistically significant differences among TED-
WR x previous crop combinations (Tukey’s
test; p < 0.05). Percentage of fields sown with
maize after maize or soybean in each TED-WR
combination is shown in (A).

producers apply fertilizer without knowing the magnitude of total crop balance and yield gap is useful to inform meaningful agronomic inter-
N demand, which is largely determined by Yp (or Yw in the case of ventions and orient policy (Fig. 5). Firstly, our findings demonstrated
rainfed fields) of the crop season ahead. If the season is unfavorable, the that the goal of achieving high yields without a large N balance is not
amount of N fertilizer they apply may be too large compared with crop an oxymoron as 25% of the fields in our study cases achieved these two
N requirements that year. In contrast, if the year has Yp (or Yw) well goals simultaneously (category A in Fig. 5). Secondly, the framework
above average, the applied N fertilizer may be insufficient to meet crop can help avoid the “one-size-fits-all” solutions promoted by some en-
N requirements. Uncertainty in yield and N demand is most important vironmental advocacy groups that propose restricting the amount of N
in rainfed fields because Yw fluctuates dramatically from year to year fertilizer that can be applied across all fields regardless of crop yields
(inter-annual CV = 31%) as a result of contrasting in-season pre- and N demand. This approach would punish producers who are already
cipitation amounts and temporal distribution, while N fertilizer remains producing high yields while achieving small positive N balance. In-
fairly constant (inter-annual CV = 6%) as rainfed producers did not try stead, agronomic and extension efforts should focus on those fields with
to adjust N fertilizer rates in response to the large annual yield varia- large positive N balance and large yield gaps (category D in Fig. 5),
tion. Not surprisingly, our study showed that N balance was smaller and which roughly represent 30% of the irrigated fields in our study and
more variable in rainfed versus irrigated fields as a result of its higher likely contributed disproportionately more to the overall N footprint
climatic risk (Fig. 4). We note that NE is a harsh environment for compared with the other fields. Similar findings have been reported for
rainfed maize production; in contrast, irrigated maize yield (and its irrigated wheat in Mexico (Ahrens et al., 2010). Finally, the framework
stability) in NE are comparable to those of rainfed maize production in is useful for individual producer and crop consultants to diagnose their
the most favorable environments in the eastern and central portions of current N fertilizer management, serving as a starting point to identify
the US Corn Belt (Grassini et al., 2014). Hence, results from this study inefficiencies and possible solutions. For example, if the current yield
for irrigated maize in NE are likely to be comparable to those for maize gap is small, it may be wise for producers to look for opportunities to
grown in favorable rainfed environments in the US Corn Belt. reduce N input use without reducing crop yields, which would lead to
While it may be difficult for producers to optimize N balance based greater input-use efficiency and extra producer revenue as it has been
on in-season weather, there may be other options that can help reduce documented in the case of irrigation water management in NE (Irmak
the N balance regardless of the year-specific weather, and with little (if et al., 2012; Gibson et al., 2019).
nil) yield penalty. Irrigated maize in rotation with soybean received Our assessment makes two key contributions relative to estimation
smaller N fertilizer amounts and achieved higher yields, which is con- of N balance. First, our study showed that calculation of N balance for
sistent with previous findings (Grassini et al., 2011; Farmaha et al., individual fields should rely on more than one year to avoid the con-
2016), leading to substantially smaller N balance (Fig. 6). In connection founding effect of weather and episodic adjustments in N fertilizer rates
to this finding, we note that future studies addressing the N balance in to account for large residual soil N from previous crop or other factors.
agro-ecosystems should aim to include the entire crop sequence into the For example, the analysis based on all field-year observations would
analysis rather than individual crops. This is critical in the case of have pointed out to an important number of fields with apparent soil
maize-soybean rotation considering the typical negative N balance mining (i.e., negative N balance) or very large N balance; this pattern
during the soybean cycle, as a result of large seed N removal without was not apparent when the analysis was based on average N balance
addition of N fertilizer, as documented by a number of studies (Connor using three or more years (Fig. 4). Second, our assessment clearly in-
et al., 2011; Santachiara et al., 2017; Ciampitti and Salvagiotti, 2018). dicated that using a suite of N-metrics is more robust compared with the
While the goal of having N balance < 75 kg N ha−1 seems realistic for use of single indicators. For instance, results in this study showed that
continuous maize systems, this threshold may need to be re-examined (low-input) rainfed systems exhibited lower N balance with almost
in the case of maize-soybean sequences where an apparent large N same PFPN compared to (high-input) irrigated systems. However, in a
balance during the maize cycle may actually be needed if the goal is to broader scale, to reach the same total grain production target, the low-
keep the N balance for the entire crop sequence above a level at which input system would need ca. 40% more cropland, which would lead to
there is sufficient N to maintain soil organic matter at steady state. an overall N balance (on a regional basis) that is similar or even higher
Our proposed framework to categorize fields into low/small N compared with the high-input irrigated systems. In other words, as

9
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

reported by previous studies (e.g., Grassini and Cassman, 2012), when Cela, S., Ketterings, Q.M., Czymmek, K., Soberon, M., Rasmussen, C., 2014.
the N balance was scaled by yield (i.e., yield-scaled N balance), the Characterization of nitrogen, phosphorus, and potassium mass balances of dairy
farms in New York State. J. Dairy Sci. 97, 7614–7632.
apparent advantage of low-input versus high-input systems vanished. Ciampitti, I.A., Salvagiotti, F., 2018. New insights into soybean biological nitrogen fixa-
So, while N balance at a field-level would be the proper indicator to tion. Agron. J. 110, 1185–1196.
evaluate environmental footprint in relation to crop-system perfor- Connor, D.J., Loomis, R.S., Cassman, K.G., 2011. Crop Ecology. Productivity and
Management in Agricultural Systems. Cambridge Univ. Press, Cambridge, UK.
mance, yield-scaled N balance is a more relevant metric for regional Dolan, M.S., Clapp, C.E., Allmarras, R.R., Baker, J.M., Molina, J.A.E., 2006. Soil organic
and global assessments that account for possible changes in land use for carbon and nitrogen in Minnesota soil as related to tillage, residue and nitrogen
agricultural production to meet future food demand. management. Soil Till. Res. 89, 221–231.
Erisman, J.W., Galloway, J.N., Seitzinger, S., Bleeker, A., Dise, N.B., Petrescu, A.M.R.,
Leach, A.M., de Vries, W., 2013. Consequences of human modification of the global
5. Conclusions nitrogen cycle. Philos. Trans. Biol. Sci. 368 20130116.
Evans, L.T., 1993. Crop Evolution, Adaptation, and Yield. Cambridge University Press,
Cambridge, UK.
We followed an approach that consists of producer field-level data, a
Exner, M.E., Perea-Estrada, H., Spalding, R.F., 2010. Long-term response of groundwater
suite of N-metrics and a spatial framework to diagnose N input use in nitrate concentrations to management regulations in Nebraska’s Central Platte
relation to crop yield and environmental outcomes in rainfed and irri- Valley. Sci. World J. 10, 286–297.
gated maize fields in the western US Corn Belt. Our study indicated that Farmaha, B.S., Eskridge, K.M., Cassman, K.G., Specht, J.E., Yang, H., Grassini, P., 2016.
Rotation impact on on-farm yield and input-use efficiency in high-yield irrigated
there is substantial room to improve yield and/or reduce N balance maize–soybean systems. Agron. J. 108, 2313–2321.
through agronomic management (e.g., crop rotation instead of con- Ferguson, R.B., 2015. Groundwater quality and nitrogen use efficiency in Nebraska’s
tinuous maize). Achieving high yields with relatively small positive N Central Platte River Valley. J. Environ. Qual. 44, 449–459.
Foley, J.A., Ramankutty, N., Braunman, K.A., Cassidy, E.S., Gerber, J.S., Johnston, M.,
balance are not conflicting goals and ca. 25% of the producers are al- Mueller, N.D., O’Connell, C., Ray, D.K., West, P.C., Balzer, C., Bennett, E.M.,
ready reaching these goals simultaneously. Although NE was used as a Carpenter, S.R., Hill, J., Monfreda, C., Polasky, S., Rockström, J., Sheehan, J., Siebert,
case study for proof of concept, the approach can be extended to other S., Tilman, G.D., Zaks, D.P.M., 2011. Solutions for a cultivated planet. Nature 478,
337–342.
cereal-based systems around the world. This study demonstrated the French, R.J., Schultz, J.E., 1984. Water use efficiency of wheat in a Mediterranean-type
value of a comprehensive and confidential field-level producer data in environment. I. The relation between yield water use and climate. Aust. J. Agric. Res.
providing useful information to producers, supply-chain companies, 35, 743–764.
Gibson, K.E.B., Yang, H., Franz, T., Eisenhauer, D., Gates, J.B., Nasta, P., Farmaha, B.S.,
and policy makers in improving yield and reducing N losses from Grassini, P., 2018. Assessing explanatory factors for variation in on-farm irrigation in
agricultural production. US maize-soybean systems. Agric. Water Manage. 197, 34–40.
Gibson, K.E.B., Gibson, J., Grassini, P., 2019. Benchmarking irrigation water use in
producer fields in the US central Great Plains. Environ. Res. Lett. 14, 054009.
Declaration of Competing Interest
Grassini, P., Cassman, K.G., 2012. High-yield maize with large net energy yield and small
global warming intensity. Proc. Natl. Acad. Sci. U.S.A. (PNAS) 109, 1074–1079.
The authors declare that there are no conflicts of interest. Grassini, P., Yang, H., Cassman, K.G., 2009a. Limits to maize productivity in the Western
Corn-Belt: a simulation analysis for fully irrigated and rainfed conditions. Agric. For.
Meteorol. 149, 1254–1265.
Acknowledgements Grassini, P., Hall, A.J., Mercau, J.L., 2009b. Benchmarking sunflower water productivity
in semiarid environments. Field Crops Res. 110, 251–262.
Financial support was received from the Nebraska Corn Board Grassini, P., Yang, H., Irmak, S., Thorburn, J., Burr, C., Cassman, K.G., 2011. High-yield
irrigated maize in the Western U.S. Corn Belt. II. Irrigation management and crop
(NCB), the Environmental Defense Fund (EDF), and the Agriculture and water productivity. Field Crops Res. 120, 133–141.
Food Research Initiative (AFRI) of the United States Department of Grassini, P., Torrion, J.A., Cassman, K.G., Yang, H., Specht, J.E., 2014. Drivers of spatial
Agriculture (USDA), Grant # 12431808. We are grateful to staff, and temporal variation in soybean yield and irrigation requirements in the western
US Corn Belt. Field Crops Res. 163, 32–46.
boards, and producers of the Little Blue, Lower Platte North, Tri-Basin, Grassini, P., Torrion, J.A., Yang, H., Rees, J., Andersen, D., Cassman, K.G., Specht, J.E.,
and Upper Big Blue Natural Resources Districts (NRDs) for providing 2015. Soybean yield gaps and water productivity in the western U.S. Corn Belt. Field
access to the producer data. We also thank Drs. Trenton Franz, Juan Crops Res. 179, 150–163.
Harmel, D., Qian, S., Reckhow, K., Casebolt, P., 2008. The MANAGE database: nutrient
Ignacio Rattalino Edreira, and Haishun Yang for providing useful
load and site characteristic updates and runoff concentration data. J. Environ. Qual.
comments. 37, 2403–2406.
Hochman, Z., Prestwidge, D., Carberry, P.S., 2014. Crop sequences in Australia’s northern
grain zone are less agronomically efficient than implied by the sum of their parts.
Appendix A. Supplementary data
Agric. Syst. 129, 124–132.
Howarth, R.W., Swaney, D.P., Boyer, E.W., Marino, R., Jaworski, N., Goodale, C., 2006.
Supplementary material related to this article can be found, in the The influence of climate on average nitrogen export from large watersheds in the
online version, at doi:https://doi.org/10.1016/j.agee.2020.106865. Northeastern United States. Biogeochem. 79, 163–186.
Irmak, S., Burgert, M.J., Yang, H., Cassman, K.G., Walters, D.T., Rathje, W.R., Payero,
J.O., Grassini, P., Kuzila, M.S., Brunkhorst, K.J., Van DeWalle, B., Rees, J.M., Kranz,
References W.L., Eisenhauer, D.E., Shapir, oC A., Zoubek, G.L., Teichmeier, G.J., 2012. Large
scale on-farm implementation of soil moisture-based irrigation management strate-
gies for increasing maize water productivity. Trans. ASABE. 55, 881–894.
Ahrens, T.D., Lobell, D.B., Ortiz-Monasterio, J.I., Li, Y., Matson, P.A., 2010. Narrowing Khanal, S., Anex, R.P., Gelder, B.K., Wolter, C., 2014. Nitrogen balance in Iowa and the
the agronomic yield gap with improved nitrogen use efficiency: a modeling approach. implications of corn–stover harvesting. Agric. Ecosyst. Environ. 183, 21–30.
Ecol. Appl. 20, 91–100. Lassaletta, L., Billen, G., Grizzetti, B., Anglade, J., Garnier, J., 2014. 50 year trends in
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Evapotranspiration: Guidelines nitrogen use efficiency of world cropping systems: the relationship between yield and
for Computing Crop Water Requirements. FAO Irrigation and Drainage. FAO, Rome, nitrogen input to cropland. Environ. Res. Lett. 9, 105011.
Italy pp. 56. Lobell, D.B., Cassman, K.G., Field, C.B., 2009. Crop yield gaps: their importance, mag-
Baker, J.M., Griffith, T.J., 2005. Examining strategies to improve the carbon balance of nitudes, and causes. Annu. Rev. Environ. Resour. 34, 179–204.
corn/soybean agriculture using eddy covariance and mass balance techniques. Agric. Matson, P.A., Naylor, R., Ortiz-Monasterio, I., 1998. Integration of environmental,
Forest Meteorol. 128, 168–177. agronomic, and economic aspects of fertilizer management. Science 280, 112–115.
Basso, B., Shuai, G., Zhang, J., Robertson, G.P., 2019. Yield stability analysis reveals McLellan, E.L., Cassman, K.G., Eagle, A.J., Woodbury, P.B., Sela, S., Tonitto, C.,
sources of large-scale nitrogen loss from the US Midwest. Sci. Rep. 9, 5774. Marjerison, R.D., van Es, H.M., 2018. The nitrogen balancing act: tracking the en-
Blanco-Canqui, H., Lal, R., 2008. No-tillage and soil-profile carbon sequestration: an on- vironmental performance of food production. BioSci. 68, 194–203.
farm assessment. Soil Sci. Soc. Am. J. 72, 693–701. Morell, F.J., Yang, H., Cassman, K.G., Van Wart, J., Elmore, R.W., Licht, M., Coulter, J.A.,
Cassman, K.G., 2017. Driving blind: the need for robust N fertilizer efficiency metrics. Ciampitti, I.A., Pittelkow, C.M., Btouder, S.M., Thomison, P., Lauer, J., Graham, C.,
Fert. Focus (July/August), 44–46. Massey, R., Grassini, P., 2016. Can crop simulation models be used to predict local to
Cassman, K.G., Gines, G.C., Dizon, M.A., Samson, M.I., Alcantara, J.M., 1996. Nitrogen- regional maize yields and total production in the US Corn Belt? Field Crops Res. 192,
use efficiency in tropical lowland rice systems: contributions from indigenous and 1–12.
applied nitrogen. Field Crops Res. 47, 1–12. Mueller, N.D., Gerber, J.S., Johnston, M., Ray, D.K., Ramankutty, N., Foley, J.A., 2012.
Cassman, K.G., Dobermann, A., Walters, D.T., 2002. Agroecosystems, nitrogen-use effi- Closing yield gaps through nutrient and water management. Nature 490, 254–257.
ciency, and nitrogen management. Ambio 31, 132–140. Oenema, J., van Ittersum, M., van Keulen, H., 2012. Improving nitrogen management on

10
F.A.M. Tenorio, et al. Agriculture, Ecosystems and Environment 294 (2020) 106865

grassland on commercial pilot dairy farms in the Netherlands. Agric. Ecosyst. Crops Res. 240, 185–193.
Environ. 162, 116–126. Tilman, D., Cassman, K.G., Matson, P.A., Naylor, R., Polasky, S., 2002. Agricultural sus-
Passioura, J.B., 2006. Increasing crop productivity when water in scare–from breeding to tainability and intensive production practices. Nature 418, 671–677.
field management. Agric. Water Manage. 80, 176–196. Treacy, M., Humphreys, J., Mc Namara, K., Browne, R., Watson, C.J., 2008. Farm-gate
Rattalino Edreira, J.I., Guilpart, N., Sadras, V., Cassman, K.G., van Ittersum, M.K., Schils, nitrogen balances on intensive dairy farms in the south west of Ireland. Irish J. Agric.
R.L.M., Grassini, P., 2018. Water productivity of rainfed maize and wheat: a local to Food Res. 47, 105–117.
global perspective. Agric. Forest Meteorol. 259, 364–373. van Drecht, G., Bouwman, A.F., Knoop, J.M., Beusen, A.H.W., Meinardi, C.R., 2003.
Sadras, V.O., Angus, J.F., 2006. Benchmarking water-use efficiency of rainfed wheat in Global modeling of the fate of nitrogen from point and nonpoint sources in soils,
dry environments. Aust. J. Agric. Res. 57, 847–856. groundwater, and surface water. Global Biogeochem. Cycles 17, 1115.
Sanchez, P.A., 2002. Soil fertility and hunger in Africa. Science 295, 2019–2020. van Ittersum, M.K., Cassman, K.G., Grassini, P., Wolf, J., Tittonell, P., Hochman, Z., 2013.
Sanchez, P.A., Swaminathan, M., 2005. Hunger in Africa: the link between unhealthy Field 349 crops research yield gap analysis with local to global relevance—a review.
people and unhealthy soils. Lancet 365, 442–444. Field Crops Res. 350 (143), 4–17.
Santachiara, G., Borrás, L., Salvagiotti, F., Gerde, J.A., Rotundo, J.L., 2017. Relative van Kessel, J.S., Reeves, J.B., 2002. Nitrogen mineralization potential of dairy manures
importance of biological nitrogen fixation and mineral uptake in high yielding soy- and its relationship to composition. Biol. Fert. Soils 36, 118–123.
bean cultivars. Plant Soil 418, 191–203. Venterea, R.T., Halvorson, A.D., Kitchen, N., Liebig, M.A., Cavigelli, M.A., Grosso, S.J.D.,
Schröder, J.J., Aarts, H.F.M., ten Berge, H.F.M., van Keulen, H., Neeteso, J.J., 2003. An Motavalli, P.P., Nelson, K.A., Spokas, K.A., Singh, B.P., Stewart, C.E., Ranaivoson, A.,
evaluation of whole-farm nitrogen balances and related indices for efficient nitrogen Strock, J., Collins, H., 2012. Challenges and opportunities for mitigating nitrous
use. Eur. J. Agron. 20, 33–44. oxide emissions from fertilized cropping systems. Front. Ecol. Environ. 10, 562–570.
Skaggs, R.W., Breve, M.A., Gilliam, J.W., 1995. Predicting effects of water table man- Verma, S.B., Dobermann, A., Cassman, K.G., Walters, D.T., Knops, J.M., Arkebauer, T.J.,
agement on loss of nitrogen from poorly drained soils Eur. J. Agron. 4, 441–451. Suyker, A.E., Burba, G.G., Amos, B., Yang, H., Ginting, D., Hubbard, K.G., Gitelson,
Sørensen, R., Zinko, U., Seibert, J., 2006. On the calculation of the topographic wetness A.A., Walter-Shea, E.A., 2005. Annual carbon dioxide exchange in irrigated and
index: evaluation of different methods based on field observations. Hydrol. Earth rainfed maize-based agroecosystems. Agric. Forest Meteorol. 131, 77–96.
Syst. Sci. 10, 101–112. Yang, H., Dobermann, A., Lindquist, J.L., Walters, D.T., Arkebauer, T.J., Cassman, K.G.,
Spears, R.A., Kohn, R.A., Young, A.J., 2003. Whole-farm nitrogen balance on Western 2004. Hybrid-maize: a maize simulation model that combines two crop modeling
dairy farms. J. Dairy Sci. 86, 4178–4186. approaches. Field Crops Res. 87, 131–154.
Steffen, W., Richardson, K., Rockström, J., Cornell, S.E., Fetzer, I., Bennett, E.M., Biggs, Yang, H., Dobermann, A., Cassman, K.G., Walters, D.T., 2006. Features, applications, and
R., Carpenter, S.R., de Vries, W., de Wit, C.A., Folke, C., Gerten, D., Heinke, J., Mace, limitations of the hybrid-maize simulation model. Agron. J. 98, 737–748.
G.M., Persson, L.M., Ramanathan, V., Reyers, B., Sörlin, S., 2015. Planetary bound- Yang, H., Grassini, P., Cassman, K.G., Aiken, R.M., Coyne, P.I., 2017. Improvements to the
aries: guiding human development on a changing planet. Science 347, 1259855. hybrid-maize model for simulating maize yields in harsh rainfed environments. Field
Tenorio, F.A.M., Eagle, A.J., McLellan, E.L., Cassman, K.G., Howard, R., Below, F.E., Clay, Crops Res. 204, 180–190.
D.E., J Coulter, J.A., Geyerg, A.B., Joos, D.K., Lauer, J.G., Licht, M.A., Lindsey, A.J., Zhang, X., Davidson, E., Mauzerall, D., Searchinger, T., Dumas, P., 2015. Managing ni-
Maharjan, B., Pittelkow, C.M., Thomison, P.R., Wortmann, C.S., Sadras, V.O., trogen for sustainable development. Nature 528, 51–59.
Grassini, P., 2019. Assessing variation in maize grain nitrogen concentration and its Zhao, X., Christianson, L.E., Harmel, D., Pittelkow, C.M., 2016. Assessment of drainage
implications for estimating nitrogen balance in the US North Central region. Field nitrogen losses on a yield-scaled basis. Field Crops Res. 199, 156–166.

11