Computers and Electronics in Agriculture 185 (2021) 106126

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture

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

Original papers

Ten years of corn yield dynamics at field scale under digital agriculture
solutions: A case study from North Italy
Ahmed Kayad a, b, Marco Sozzi a, Simone Gatto c, Brett Whelan d, Luigi Sartori a,
Francesco Marinello a, *
a
University of Padova, Department TESAF, viale dell’Università, 16, I-35020 Legnaro (PD), Italy
b
Agricultural Engineering Research Institute (AEnRI), Agricultural Research Centre, Giza 12619, Egypt
c
Porto Felloni srl, 44023 Lagosanto (FE), Italy
d
Precision Agriculture Lab., Sydney Institute of Agriculture, The University of Sydney, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: Farmer’s management decisions and environmental factors are the main drivers for field spatial and temporal
Yield maps yield variability. In this study, a 22 ha field cultivated with corn for more than ten years using different pre­
Site-specific nitrogen scription maps of nitrogen application rates was investigated. Prescription maps were developed based on
Soil moisture sensors
archived yield maps, soil analysis and recently integrated with Sentinel 2 satellite images. In addition, farmer
Soil electrical conductivity
GNDVI
experience and availability of variable rate application (VRA) requirements had an influence on the development
Digital agriculture of the homogeneous management zones. The initial approach with VRA was quite simple, based on a simple
partitioning of the field into three rectangular zones (defined mainly based on previous yield maps and farmer
experience). The partitioning changed with time and knowledge, evolving to the final five irregularly shaped
zones (defined based on Farm works decision support software). Furthermore, since 2010 the farmer began using
soil moisture sensor for irrigation decisions. Results of the present study highlight an improvement in corn yield
and a reduction in total applied nitrogen. Corn yield improved on average by 31% on a ten years basis to reach
more than 14 ton/ha dm. in 2018. At the beginning of VRA, yield maps showed a high spatial variation between
field zones compared to reduced variation in the following seasons. In addition, the nitrogen applied reduced by
around 23% while the total yield was improving. These results showed an increase in the partial factor pro­
ductivity from less than 54 to around 87 kg of corn grain per kg of nitrogen applied. This promising result shows
that farmer management decisions can improve every season by continuous monitoring of crop performance,
understanding field variability and taking advantage of recently developed decision support software tools.

1. Introduction million tons) between 2007 and 2018 respectively (EUROSTAT, 2019).
This huge reduction in corn cultivated area is due to many reasons such
Currently, there are a lot of challenges facing agriculture and food as diminishing profitability, reduction of dairy farms that utilize corn
security such as crop land shrinking, limited resources, growing popu­ feed and the cross-compliance on crop rotations of Common Agriculture
lation to exceed 9 billion by 2050 and environmental protection issues Policy (CAP) of European Union. Recently, the global consumption from
(West et al., 2014). In Italy, the total agricultural area is about 12.9 nitrogen fertilizers has increased dramatically and is expected to exceed
million ha, and more than 78% of this area consists of farms that are 186 million tons of nitrogen (N) per year by 2050 (Muschietti-Piana
more than 10 ha area in 2013 (EUROSTAT, 2019). More than 5% of the et al., 2018). On the other hand, nitrate leaching causes an average of
Italian agricultural land is sown to corn (FAOSTAT, 2020). Corn is one of 10–30% of nitrogen losses (Schepers and Raun, 2008) and in corn, these
the most important crops in North Italy with an average yield ranging losses are larger than other cereals (Luce et al., 2011).
between 8 and 10.6 t/ha over the last ten years (EUROSTAT, 2019). The Precision agriculture (PA) is a potential solution to these challenges
whole corn cultivated area reduced from over one million hectares and considered as one of the digital agriculture elements (Bolfe et al.,
(producing 9.7 million tons) to 622 thousand hectares (producing 6.2 2020; Rubio and Más, 2020). PA aims to improve crop yield quantity

* Corresponding author.
E-mail address: [email protected] (F. Marinello).

https://doi.org/10.1016/j.compag.2021.106126
Received 17 August 2020; Received in revised form 19 March 2021; Accepted 21 March 2021
Available online 3 April 2021
0168-1699/© 2021 Elsevier B.V. All rights reserved.
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

and quality while reducing environmental footprint and operating costs required water (Domínguez-Niño et al., 2020). The capacitance type of
(Chlingaryan et al., 2018). Plant health and subsequently crop yield soil moisture sensors is considered as the most widely used type which
depends on many different attributes such as soil properties, irrigation, relies on the determination of the dielectric permittivity of the soil
weather, topography and agricultural management practices (Tola et al., around the sensor (Campbell, 1990; Kizito et al., 2008). These types are
2017). Yield mapping is an important element of PA. It can be used by characterized by the low cost and requires little maintenance (Visconti
farmers to identify underperforming areas where there may be soil et al., 2014).
compaction or fertility issues (Higgins et al., 2019). Previous assump­ Site-specific application of nitrogen is a PA approach to optimize
tions in PA expected that soil characteristics would dominate the effect nitrogen fertilizer use efficiency and to reduce nitrogen leaching (Del­
on crop performance and subsequently crop yield patterns would be gado et al., 2005; Khosla et al., 2002). This approach aims to adjust
similar each year, leading to relatively similar yield maps each year. This nitrogen fertilizer application to match crop requirements at different
assumption works only when the yield-limiting parameters are perma­ spatial and temporal extents. Assessing optimum nitrogen application
nent (Blackmore et al., 2003; Vrindts et al., 2003). for different MZ in the same field can be approached using a range of
Defining management zones (MZ) is one of the most applied PA techniques. One of the common methods is by using crop response
approaches to achieve more efficient practices by dividing the field into curves (Bramley et al., 2013). Alternatively, delineation by using crop
different homogeneous zones according to yield-limiting parameters models such as CERES-Maize in DSSAT (Jones et al., 2003) and APSIM
and actual needs to optimize profit (Moral et al., 2011; Peralta et al., (Keating et al., 2003) which take into consideration different soil con­
2015b; 2015a; Vrindts et al., 2005). The MZ maps can be used for var­ ditions, agricultural practices and weather factors (Pampolino et al.,
iable tillage depth, planting density, irrigation and fertilizer rate (Nawar 2012). Some studies investigated the possibility of combining crop
et al., 2017). Previous research work on the delineation of MZ used growth models and remotely sensed data for N fertilizer management in
different approaches such as soil electrical conductivity (Cillis et al., corn (Jin et al., 2018; Thompson et al., 2015; Wang et al., 2021). In
2018; Johnson et al., 2003), soil colour (Fleming et al., 2000), archived addition, other strategies includes yield goal methodologies (Stanford,
yield maps (Blackmore, 2003; Blackmore et al., 2003) or a combination 1973), soil nitrate tests (Shapiro et al., 2008), maximum return to ni­
between these factors (Taylor et al., 2007; Vrindts et al., 2005). trogen (Sawyer et al., 2006) and N-rich strip treatments (Solie et al.,
Identification of subfield regions with homogeneous yield-limiting 2012). (McBratney et al., 2005) reported that much efforts were focused
factors (homogeneous management zones) is a complex task (Peralta on zones management with less attention to decision support systems for
and Costa, 2013). Achieving the ideal prescription map for each appli­ practical world which hinders the full adoption of PA. Currently, there
cation requires a deep understanding of field spatial variability and the are different commercial software that provides a whole farm manage­
interaction between the whole system of soil, crop, weather and man­ ment packages which could combine different layers of information such
agement practices (fertilizer, irrigation, etc.). In order to define a MZ as soil properties, weather data, archived yield maps, remotely sensed
map, different information, mostly regarding soil properties and data and pricing information for corn and fertilizers to support farmers
archived yield maps are required. The effectiveness of the defined MZ decision for optimum N rates.
depends on the accuracy of this information and its processing method. An “ideal” prescription map may need different seasons of moni­
In addition, the final assessment of the quality of the defined MZ map is toring to prove its profitability. In this study, archived yield maps and as
the prospective yield map of the coming season, inputs cost and envi­ applied variable rate nitrogen application for ten seasons were analyzed
ronmental pollution reduction. in connection with utilized technologies to assess the impact of this
MZ delineation approaches were summarized by (Miao et al., 2018) technology on the long term.
to three main methods. Firstly, using crop yield maps by applying The commercial availability of precision and digital agriculture tools
empirical procedures to delineate the field into three zones. Secondly, in Italy, sensors, equipment, software and knowledge is relatively new at
using the advantage of soil sensors to describe soil electrical conduc­ the farmer’s level. The awareness of within-field variability for farmers
tivity, organic matter, etc. Thirdly, by integrating both yield maps and was well established through their work in fields year after year while
soil properties in one MZ map which expected to be the most efficient reliable commercial solutions have only been available since 2010. This
approach. means that most farmers started with just simple approaches with slow
Yield monitoring sensors mounted on harvesters have been steps towards investing capital for the whole set of digital agriculture
commercially available for over twenty years and subsequently archived available tools. Meanwhile, the current availability of new sensors with
yield maps are available from a quite large number of fields. Further­ reduced dimensions, cost and better performance allowed an increase of
more, archived remote sensing data from different satellites and aerial data and information at field level for wider range of farmers (Kayad
sensors plays an important role discovering the history of previous et al., 2020; Muangprathub et al., 2019; Zhai et al., 2020). This infor­
seasons crop variability. Different studies investigated the ability of mation is expected to support the digital transformation in agriculture
remotely sensed data to evaluate crop within field variability using and improve crop yield. Subsequently, evaluating the long-term impact
vegetation indices (Kayad et al., 2019; 2016; Madugundu et al., 2017b; of precision and digital agriculture applications in commercial fields has
Schwalbert et al., 2018). (Kayad et al., 2019) reported that the Green seen few available research studies (Yost et al., 2017). Therefore, this
Normalized Difference Vegetation Index (GNDVI) could describe corn study focuses on the long-term impact of PA in corn yield spatiotemporal
yield spatial variability within field scale at crop growing stages be­ variability with the real situation and history of farmer investment. The
tween R4 to R6. The availability and accessibility of these data resources specific objectives are as follows:
allows a comparison of the spatial yield variability from different sea­
sons and supports farmers decisions for nitrogen site-specific – To evaluate the improvement in corn yield in both within-field
application. variability and between different seasons under different nitrogen
Irrigation is a crucial practice for reducing crop production vulner­ prescription maps and irrigation decision methods.
ability caused by the lack of precipitation. The shortage in available – To compare the cumulative corn yield spatial variability with the soil
water for plants impairs crop production while excess may effect on root apparent electrical conductivity (ECa) maps and average GNDVI map
growth and produce soil degradation such as soil erosion, surface runoff from different seasons.
and nutrient leaching (Grashey-Jansen, 2014). The decision of irrigation – To compare the average corn yield production from the regional
depends on the crop water requirements, soil properties, environmental scale with the study field.
factors and the available amount of water from different resources.
Currently, soil moisture sensors plays an important role for irrigation
scheduling and can be used for an empirical site-specific adjustments of

2
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

2. Materials and methods (ECa) was collected by using the Automatic Resistivity Profiling (ARP©
GEOCARTA, Paris, France). ARP is an on-the-go measuring tool con­
This study was conducted in a 22-hectare field cultivated with corn sisting of four pairs of toothed wheels where the front pair inject elec­
for 14 years since 2004 except in the 2015 season. The study field, and trical current and the following three pairs measure the voltage drop
two neighboring fields, are irrigated by a center pivot system. This field which is related to the conductivity of the soil. The three receivers
is part of Porto Felloni agricultural company that is located in Lagosanto, distributed at different distances in order to measure soil resistivity at
Ferrara, North Italy. This field lies within the Mediterranean climatic different depths; 50, 100 and 200 cm then converting measurements to
zone which characterized by hot, dry summers and cool, wet winters soil ECa (Cillis et al., 2018; Tabbagh et al., 2000). Soil samples and ARP
between the latitudes of 44.690762 to 44.697934 and longitudes of data were collected in the winter season at a dry soil condition before
12.209291 to 12.216269 (Fig. 1). The total rainfall is between 530 and corn planting. The ARP was mounted to a small tractor and the soil
830 mm per year according to the farm weather station for the last 6 survey conducted in parallel swaths 15 m apart. The observation num­
years with an average rainfall of 270 mm during the corn growing ber exceeded 3000 points from the whole study field (around 135
season. The corn planting date is by end of March and harvesting during points/ha). The data was filtered by removing outliers over the range of
September with an average crop age of 160 days from planting date. mean ± 3 SD. Fig. 2 B, shows the ECa map for the study field. The cost for
soil analysis was between 50 and 70€/ha depending on the analyzed
2.1. Soil and weather data collection elements while the ARP scanning costs 17€/ha.

Weather data was collected through a weather station starting from 2.2. Irrigation practices
the 2004 season. The weather station collects data for maximum and
minimum air temperature, rainfall, humidity, solar radiation, wind The study field was irrigated by a centre pivot system since 2001 and
speed and soil temperature. In addition, archived weather data since followed by a new underground drainage system established in 2005. In
1985 are available from Lagosanto weather station which is within 10 2010, new soil moisture sensors were installed in the study field to
km from the study field and other governmental weather stations nearby support irrigation decisions. The moisture sensors were based on
the study field. For the purpose of this study, the total accumulated capacitance type (Sentek Co., Adelaide, Australia) and consist of 5
rainfall for three months between 1st of May till 31st of July were sensors that measure soil moisture every 10 cm with a total sensing
calculated by averaging rainfall measurements from the closest three depth of 50 cm from the soil surface. The soil moisture sensors were
weather stations for each season as this period of time is the most sen­ connected to a computer software that used to monitor the soil moisture
sitive for crop production. The three weather stations are located within content and tell farmers when and how much to irrigate. Most fields in
100 m (Guagnino), 12 km North-East (Volano) and 12 km South-West the study region depend on rain for covering crop water requirements
(Camse) the study field and averaging these measurements could with irregular complementary irrigation based on farmers experience
cover some missed data through the study period. according to crop need using a gun sprinkler or pivot irrigation systems.
Soil samples were collected from the study field in 2004, 2008, 2013 The common irrigation practices in the study field involve 3–4 appli­
and 2015 using a grid sampling strategy for the whole center pivot field. cations per corn season with an average 20 mm/irrigation except in
Soil analysis time was decided based on a four years’ time span, prior to 2014 where it was a rainy season and no irrigation was applied.
newly developed MZ maps (Section 2.5) and considering the lab analysis
cost. These samples were analysed for soil texture, EC, pH, N, P, K, Mg, 2.3. Yield data collection
Na and Ca. Soil analysis contributed to the delineation of nitrogen MZ
using the most recent soil chemical analysis and accumulated soil Yield data was collected by a grain yield monitoring system mounted
texture information corresponding to each season. Fig. 2 A, illustrates on a combine harvester and calibrated before the harvesting process.
the distribution of the soil samples through the study field in the The combine working width is 8 m and consists of 10 rows, where there
different sampling years. In 2012, soil apparent electrical conductivity is a sensor for each row to calculate the actual working width and

Fig. 1. Study field of 22 ha in a rectangular shape at Ferrara, Emilia-Romagna region, Italy.

3
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

Fig. 2. Study field with: (A) Soil samples positions collected at different years and (B) Soil ECa data in mS/m.

maintain yield data accuracy. The company started to collect yield data 2011–2012, 2013, 2014, 2015 season cultivated with wheat only,
in 2002 and archived corn yield data for the study field are available 2016–2017 and finally 2018–2019. The first prescription maps between
since 2004, except 2005 and 2009 data which are not available. Yield 2007 and 2010 were quite simple due to a lack of knowledge, available
data was filtered by removing outliers over the range of mean ± 3 SD information, equipment technologies and Decision Support Software
and removing data that recorded during the harvester turnings. The (DSS). Delineating MZ was based on previous yield maps, most recent
original yield monitor could collect around 350 points/ha of yield data soil chemical analysis, accumulated soil texture information, expert
and starting from 2012 a more precise yield monitoring system of about agronomist consultation and using a farm management software for
1000 points/ha was used. Furthermore, the average corn yield from the whole field recommendations from JD link, (Deere & Company, Moline,
region of Emilia-Romagna were collected from the official regional re­ Illinois, USA). The JD link software was used for transferring data be­
cords in order to compare between the regional and the study field corn tween machine and farm head office and was used for documenting all
yield dynamics. The regional records are available on the Italian Na­ operations at the field that resulted in well-archived documentation
tional Statistics Institute website (https://www.istat.it/it/agricoltura? across field operations. Fig. 3, shows all prescription maps for nitrogen
dati). The region of Emilia-Romagna has an area of 22,400 km2 and application during the last 10 seasons. All of the prescription maps since
considered as one of the wealthiest and most developed regions in 2007 were applied in the field using a variable rate application system
Europe. Farmers in this region are well educated and mostly have effi­ (Dickey-John Corp., Auburn, IL, USA).
cient agricultural management practices with modern machinery, Prescription maps delineated the field into three MZ between 2007
highly productive seed varieties, and almost the same environmental till 2013 which was followed by two different MZ during the season of
conditions. 2014. At this stage, all MZ were delineated as a combination of different
regular rectangular shapes. Soil analysis from 2004 contributed in the
2.4. Calculating GNDVI maps nitrogen rate recommendations for 2007–2008 prescription maps while
soil analysis from 2008 were used for 2009–2012 nitrogen prescription
Sentinel-2 satellite images were downloaded from Google Earth maps. Prescription maps between 2007 and 2010 consists of three MZ
Engine platform and processed through the Sen2Cor Level-2A (L2A) where most of the field fell in one MZ while only field boundaries fell in
processor to obtain the bottom of atmosphere products. GNDVI maps other MZs. This delineation could be considered as the first steps to­
were calculated from Sentinel-2 (L2A) images using Eq. (1). A total of wards the application of PA practices in this field. In addition, these 4
three maps were calculated from 2016, 2017 and 2018 growing seasons seasons were prior to the soil moisture sensors installation and could be
from Sentinel-2 images acquired on 4/8/2016, 4/8/2017 and 17/8/ considered as the control seasons for evaluating the impact of applied
2018. These dates were selected based on a previous study by (Kayad technologies. The rate of nitrogen fertilizer was based on soil analysis
et al., 2019) which reported that GNDVI at crop stage between R4 to R6 results and considering the regional fertilizer recommendations for each
are the most suitable index and time to describe the within field vari­ crop and this approach is the main reference nitrogen management
ability of corn yield. The three maps were averaged and reported as the practices at the regional scale.
GNDVI map for further analysis. From 2011 to 2012 seasons, a nitrogen-rich strip test was applied in
order to examine the response of corn yield to different nitrogen rates
GNDVI =
NIR − G
(1) across different soil characteristics and yield patterns. The nitrogen-rich
NIR + G strip refers to applying sufficient rate of nitrogen fertilizer where crop
yield reaches a response plateau while another two other less rates
where G is the green band (560 nm) and NIR is the near-infrared band
corresponding the regional reference application where applied for the
(842 nm) of Sentinel-2 reflectance images.
rest of the field. The information from this test used to fine-tune the
nitrogen rate for further seasons.
2.5. Development of nitrogen management zones From 2013 to 2014 season, another level of MZ delineation was
applied. At these seasons, the field was delineated to different rectan­
The company started to delineate the study field to different MZ for gular shaped MZ according to the previous yield maps, soil analysis
nitrogen application in 2007. Each prescription map was repeated for information, ECa map. The decision of each MZ delineation boundaries
two years except 2013 and 2014 as follows: 2007–2008, 2009–2010,

4
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

Fig. 3. Nitrogen application prescription maps.

was evaluative with respect to the available data and agronomist relative similarity between maps, we compared the average yield and
consultation while the exact nitrogen rate recommendations were ECa maps with yield maps from all seasons by calculating the average
calculated for each MZ based on the nitrogen-rich strip test results from agreement (AA) as described by (Stępień et al., 2016) and (Ali et al.,
2011 to 2012 seasons. Starting from 2016, a farm management software, 2019). The agreement was estimated according to the number of grids
Farm works Office (Trimble Inc, Sunnyvale, USA), was used to delineate that have the same class (High agreement) and the number of grids that
the MZ providing the advantage of using multi-layers of soil analysis, deviate by one class (Medium agreement) using Eq. (2).
archived yield maps, Sentinel-2 satellite images and ECa data. Since
NH + (N2M )
2016, the field was delineated to five MZ with different nitrogen rate Average agreement (AA) = (2)
applications taking the advantage of all archived information across NT
different seasons.
where NH is the number of grids that have a high agreement, NM is the
number of grids that have a medium agreement, and NT is the number of
2.6. Data analysis total grids.
Furthermore, nitrogen prescription maps were converted to grids of
According to the available data and the improvements in utilized 10 m × 10 m spatial resolution to match the yield maps resolution by
technologies, field conditions and farming practices, the analysis using the conversion tool Polygon to Raster on ArcGIS software. All of
focused on the last 10 corn seasons starting from 2007 season and the the yield maps and Urea application maps were used to calculate the
evolution in the study field was divided to three phases as follows: i) Partial Factor Productivity for Nitrogen (PFP-N) as per Eq. (3). PFP-N is
Phase 1: between 2007 and 2010 seasons where the N prescription maps a simple and useful indicator of nutrient use efficiency in crop produc­
showed an almost constant rate of N except field boundaries. ii) Phase 2; tion and used in several studies (Li et al., 2020a; Li et al., 2020b; Nafi
between 2011 and 2014 seasons where soil moisture sensor utilized and et al., 2019). Urea fertilizer contains 46% of nitrogen, and this per­
more logical MZ delineation was applied. iii) Phase 3: between 2016 and centage was used in nitrogen calculations for PFP-N. In addition, in 2014
2018 seasons where the MZ was delineated based on Formworks soft­ the company applied a constant rate of poultry manure through the
ware taking the advantage of multi-layers of available information and study field which contains 75 kg nitrogen per hectare. This application
previous season’s experience. was also included in the calculations of PFP-N for the 2014 season. For
In order to compare the relative improvements in corn yield between each phase, an average PFP-N was calculated and ANOVA and t-test (5%
the study field and the regional records, the annual average yield was significance level) was performed to evaluate the significant difference
compared following the previously described phases after adding the between the three phases at the study field.
2006 record to Phase 1. Statistical analysis on average yield data (t-test,
5% significance level) was performed to evaluate the significant differ­ PFP − N =
Total applied nitrogen (kg)
(3)
ence between the three phases from both regional and study field Total grain yield (kg)
records.
Yield data after the filtering process was interpolated and resampled 3. Results
to 10 m × 10 m grids using the kriging and resample tools from ArcGIS
software. The 10 m resolution was considered as a suitable resolution for The corn yield maps between 2004 and 2018 growing seasons
such analysis taking into consideration the common positioning errors showed a clear improvement in total corn yield (Fig. 4). Average corn
from both yield maps and as-applied nitrogen maps. Furthermore, this yield for the phase 1 (2007–2010) was 10.9 t/ha, for phase 2
resolution matches with the calculated GNDVI spatial resolution from (2011–2014) was 12.91 ton/ha and reached 14.32 t/ha at phase 3 in the
Sentinel-2 images. For each phase, an average yield map was calculated last three years (2016–2018) which is a greater than 31% increase be­
and ANOVA and t-test (5% significance level) was performed to evaluate tween phase 1 and 3. Statistical analysis between average yield maps
the significant difference between the three phases at the study field. corresponding to phase 1, 2 and 3 proved a statistical difference between
Moreover, an average yield map for the ten studied seasons the three phases. Table 1, shows the ANOVA results and Fig. 5 shows a
(2007–2018) was then calculated from the grid maps. All the maps (ten box-plot for the average yield maps from the three phases. Starting from
seasons yield maps 2007–2018, one average yield map, three GNDVI 2009, corn yield increased year by year (except a drop in 2012) until
maps 2016–2018 and one ECa map 2012) were then normalised to be 2014 where the average yield reached a stable record of more than 14 t/
between 0 and 100 in order to compare between different maps. The ha.
average yield map and ECa were compared with all 10 seasonal yield Grain yield is a function of many environmental factors such as
maps and the average GNDVI map by calculating the absolute difference weather which varies between seasons. In order to compare yield
between maps. The deference was divided to 5 classes as follows: improvement in the study field with reference data, we used the official
0–20%, 20–40%, 40–60%, 60–80% and 80–100%. To investigate the corn yield data from the region of Emilia-Romagna. Fig. 6, shows the

5
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

Fig. 4. Corn yield maps from the study field for all studied seasons.

was observed in both phase 2 and 3. The difference in corn grain yield
Table 1
between the study field and regional records were in narrow range in the
ANOVA results for average yield maps corresponding to the three studied
first five years and almost the same in 2008. Then the difference be­
phases.
tween yield values increased year by year and reached a peak in 2017
ANOVA with a difference of more than 6 t/ha. In addition, in the 2012 growing
Source of Variation SS df MS F P-value F crit
season, both values had the same drop because of environmental con­
Between Groups 12,553 2 6276 3210 0 2.997 ditions in this year. Starting from 2014, corn yield in the study field was
Within Groups 12,441 6363 1.955
high and stable with values over 14 t/ha while the regional yield fluc­
Total 24,994 6365
tuated between 8.1 and 10.6 t/ha. Moreover, yield data were almost
proportional to total rainfall data, especially at the regional level. The
average total rainfall between May to July is almost over 150 mm except
a significant reduction of just 84 mm in 2012 season which matches with
the lowest regional yield record of 6.4 ton/ha in this season.
The average agreement (AA) between each season yield map and the
average yield map showed a high agreement with AA values of around
90% for all seasons except 2012 season (Fig. 7). This value means that
90% of the study field has the same yield spatial variability, within
almost ± 20% of yield, regardless of the different nitrogen application
rates and growing season conditions. In addition, the AA between each
of the 10 years of yield maps and the map of GNDVI averaged over 2016
– 2018 ranged between 80 and 90%. Meanwhile, , the AA between each
of the 10 years of yield maps and the ECa map collected in 2012 ranged
between 45 and 80%.
To compare the spatial variability of the average yield map with the
ECa and average GNDVI maps, the normalised average yield map (%)
was subtracted from the normalised ECa and GNDVI maps (%) to pro­
vide the range of difference between these maps. The AA between yield
Fig. 5. Box-plot for the average yield maps corresponding to the three stud­ and GNDVI was 95% and with ECa was 55% while AA between GNDVI
ied phases. and ECa was 49%. According to the difference maps (Fig. 8), most of the
study field area is within ≤ 20% for the difference between average yield
improvement of yield data in both the study field and the region of and GNDVI which indicates the ability of the GNDVI maps to define the
Emilia-Romagna and the total rainfall between May to July for each yield potential variability through the study field even from previous
season. It is clear that regional and study field values for each year have seasons compared to the ECa map.
the same trend while the study field yield was improving more than the The average applied nitrogen amount was 187, 234, 175, 214, 188,
regional yield. Statistical analysis (t-test, 5% significance level) showed 158 and 169 kg of nitrogen/ha for the 2007–2008, 2009–2010,
that between 2006 and 2010 there was no significant difference between 2011–2012, 2013, 2014, 2016–2017 and 2018 prescription maps
regional and study field annual yield records while significant difference respectively. Considering the three defined phases, the average applied

6
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

Fig. 6. Corn grain yield data from the study field and the region of Emilia-Romagna and the total rainfall between May to July for each growing season.

Fig. 7. The average agreement (AA) between yield maps of each season with average yield, GNDVI and soil ECa maps.

nitrogen amount was 211, 188 and 162 kg of nitrogen/ha for phase 1, 2 4. Discussion
and 3 respectively. In general, the amount of applied nitrogen reduced
through the studied ten seasons while 2009–2010 prescription maps Corn yield analysis of the results of this study shows that average
showed the highest amount of nitrogen application by 234 kg/ha. yield was fluctuating significantly every year with an overall increasing
Compared to the lowest amount of applied nitrogen in 2016–2017 of trend till 2013. Starting from 2014, average corn yield reached a stable
158 kg/ha, the reduction is over 32% while corn grain yield was stage over 14 ton/ha every season. Archived data and management
increasing. The difference between phase 1 and 3 shows a reduction of practices from the study field could be classified to three phases based on
over 23% in total applied nitrogen. the available PA tools and management decisions by the farmer as
In contrary, the Partial Factor Productivity for Nitrogen (PFP-N) described in Section 2.6.
increased through studied seasons from 65 and 43 kg grain/ kg nitrogen Phase 1, from 2007 up to 2010 seasons where management zones
in 2007 and 2009 to reach 90 and 81 kg grain/ kg nitrogen in 2017 and depended on previous seasons yield maps and limited soil analysis. At
2018 respectively. Statistical analysis showed a significant difference this stage, the farmer starts to take his first steps towards site-specific
between the studied phases where the average PFP-N was 54, 69 and 87 nitrogen application where management zones were delineated
kg grain/kg nitrogen in phase 1, 2 and 3 respectively. Table 2, shows the around field boundaries to three regular rectangular shapes. The ma­
ANOVA results and Fig. 9, shows a box-plot for the average PFP-N maps jority of the study field fell in one MZ with a constant nitrogen appli­
from the three phases. Comparing 2009 and 2017 seasons that represent cation rate which cannot be considered as an efficient and practical MZ.
the highest and lowest PFP-N values, the increase is over 109% while In addition, this phase was prior to the installation of soil moisture
comparing phase 1 and 3 shows 60% increase which prove the sensors and irrigation decisions was based on farmer’s experience. Sta­
improvement in prescription maps efficiency from both economical and tistical analysis proved that there was no significant difference between
environmental points of view. Fig. 10, illustrates the increase of PFP-N regional and study field yield records, thus, this phase could be
and the decrease of total nitrogen applied year by year. Furthermore, considered as a control phase to evaluate the impact of applied PA
the PFP-N maps for each season in Fig. 11 show the improvement in practices in following phases. Although, nitrogen prescription maps
nitrogen application efficiency for each year. were repeated for two following years between 2007 and 2012, the
corresponding yield maps showed a clear difference between each two

7
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

significant difference in phase 2 and 3, this proves that the improvement

in corn yield was due to the improved practices and not to environ­
mental conditions.
In Phase 2, seasons between 2011 and 2014 where more soil analysis
and yield data were available, soil moisture sensors from 2010 were
installed through the field and the ARP scan from 2012. Another level of
variability management was applied for nitrogen rates and irrigation
decisions, the farmer still delineated his field to regular shapes following
an evaluative approach, while in 2011 and 2012 prescription maps were
delineated to different nitrogen rates strips with high application rate in
just one strip and after the ARP scan in 2012, the field was delineated to
more efficient MZ that proved a clear improvement in corn yield of ˃13
ton/ha. The nitrogen-rich strip test between 2011 and 2012, supported
the nitrogen rate decisions by giving a reference information to nitrogen
rate response across different soil characteristics and yield patterns. This
phase considered to have successful results where yield records were
promising and encouraged the farmer to invest in DSS considering the
accumulated experience and archived data. Statistical analysis showed a
significant difference between Phase 1 (control) and phase 2 which
proves that the utilized soil moisture sensors and nitrogen prescription
maps improved corn yield in the study field. In this phase, corn yield
improved by >18% compared to the control phase in average.
In 2012, the season was very dry with low rainfall that led to a very
low average yield record of 6.4 ton/ha from the region of Emilia-
Fig. 8. Relative average yield, GNDVI and ECa maps and the absolute defer­ Romagna while the study field maintained the relatively high yield re­
ence between them. cord of ˃11 ton/ha thanks to the information from the soil moisture
sensors that were installed in 2010. This claim could be supported by the
huge difference between regional and study field yield in both 2011 and
Table 2
2012 seasons where the applied nitrogen prescription maps were the
ANOVA results for average PFP-N maps corresponding to the three studied
same. In contrast, the rainy season of 2014 led to one of the highest yield
phases.
seasons for both study field and at the region scale as well. Farmers in
ANOVA
the region of Emilia-Romagna depend on rainfall to cover crop water
Source of Variation SS df MS F P-value F crit
requirements with irregular supplementary irrigations. Usually, farmers
Between Groups 853,561 2 426,780 2197 0 2.997 decide to irrigate when the drought stress symptoms appear on the crop
Within Groups 1,196,349 6159 194.2
Total 2,049,909 6161
canopy, at this stage the crop is already affected negatively by drought
and in sensitive stages such as flowering and grain filling, the grain re­
duces dramatically (Jiang et al., 2020). The benefit of soil moisture
sensors was very clear at 2012 season because of its early alert before
reaching the sensitive drought stress levels which support farmers in
efficient irrigation decisions and reduces the risk of drought stress.
(Bonfante et al., 2019) developed an irrigation DSS based on three
methodologies; in-situ soil sensors, remote sensing and simulation
models for soil, plant and atmosphere system. Their results showed that
in situ soil sensors supplied 40% more water compared to the other two
methods which represents an opportunity for improving water use ef­
ficiency in the study field in case of integrating the remote sensing and
simulation models for irrigation decisions.
Phase 3, is from 2016 to 2018 seasons where the impact of decision
support software was clear. The integration between the well archived
data from accumulated soil analysis, ECa and yield with the presence of
soil moisture sensors improved corn yield to the highest and continued
stable record of >14 ton/ha regardless of yield spatial patterns that
reduced the risk in such investment. Compared to the control phase and
Fig. 9. Box-plot for PFP-N maps corresponding to the three studied phases.
phase 2, this phase showed an increase in corn yield by >31% and 11%
respectively. The difference between phase 2 and phase 3 was the
seasons that have the same nitrogen prescription map. This result is due improved delineation of MZ approach where irrigation decisions based
to the variation in environmental conditions between seasons, specif­ on soil moisture sensors have the same contribution to both phases. This
ically the rainfall amount before installing the soil moisture sensors. could highlight the impact of improved MZ delineation at this phase.
Several studies reported that extreme weather events such as high However, the observed stability in corn yield records at this phase add
temperature and drought severely threaten corn rain-fed yield between more value to the utilized management practices and technologies.
different growing seasons (Alexander et al., 2006; Jiang et al., 2020; These results agree with the findings described by (Yost et al., 2017) for
Madugundu et al., 2017a; Porter and Semenov, 2005). Linking regional a long-term impact of PA on grain crop. This result could lead to a more
yield records to this study, add value to such long term analysis by confident prediction of total crop yield and better commercial man­
eliminating the impact of environmental conditions that has a clear ef­ agement for the study field. Certainly, there are other limiting factors in
fect on each season yield. As there was no significant difference between the agricultural system that affect the final yield such as insects, dis­
study field and regional yield records in phase 1 (control) compared to eases, extreme weather events (Cogato et al., 2019) etc., however if the

8
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

Fig. 10. Total nitrogen applied and PFP-N for each season through the study field.

Fig. 11. The PFP-N map for all studied seasons in kg of corn grain per kg of applied nitrogen. 4.

results from this study field can be projected to the regional scale, then supported by the AA between ECa and yield maps where there is a
the risk of agricultural investment and subsequently the insurance relatively high agreement through early seasons (up to 2012). In the
subscription may be reduced for farms that apply PA systems. early seasons, applied nitrogen prescription maps were quite simple and
The high average agreement between each season yield map and the were mostly delineated around field boundaries and almost constant
average yield map (except the 2012 season that has extreme weather rate for other parts in the field which makes the impact of soil variability
event) prove that yield patterns maintained the relative spatial variation clearer compared to later seasons. Starting from 2013, the efficient MZ
regardless the different prescription maps for applied nitrogen. This delineation could manage such spatial variability in soil leading to
result may lead to the conclusion that the yield spatial variability related higher yield. On the other hand, the higher AA between average GNDVI
to a constant factor over all seasons (Blackmore et al., 2003; Vrindts compared with ECa with yield highlights the usefulness of using
et al., 2003) which in our study related to soil variability. This is archived satellite images for investigating within-field spatial variability

9
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

(Kayad et al., 2019). Considering the high AA across seasons and 5. Conclusions
accumulated yield map from the study field. This could lead us to claim
that the average GNDVI map from the most sensitive period R4-R6 for Ten years of archived data from a 22 ha corn field in north Italy
three seasons was an accurate indication to crop yield spatial variability between 2007 and 2018 including archived yield maps, as applied ni­
within field scale even from previous seasons. trogen maps, soil analysis, and improved irrigation decisions based on
The long-term evaluation for the site-specific nitrogen application soil moisture sensors were used to assess the evolution of precision
combined with soil moisture sensors based irrigation decisions showed a agriculture applications in this field. The delineation of MZ for nitrogen
clear improvement in the total corn yield, reduction for total nitrogen application improved through the studied seasons and these were
applied and subsequently a dramatic improvement in the nitrogen use compared with the corresponding yield maps to understand the impact.
efficiency presented in PFP-N (Fig. 9). Given the reduction in applied In addition, average yield from the study field per season was compared
nitrogen by 23% and the increase of corn grain yield by 31% between with the regional yield to evaluate the impact of applied PA techniques.
the control phase and phase 3, it appears that there was an over appli­ Moreover, ECa map and average GNDVI maps from the last 3 seasons
cation of nitrogen through this field in early seasons with no real benefit. were compared with yield maps to evaluate within-field variability. The
This result highlights the benefit of site-specific nitrogen application and studied seasons were divided to three phases considering the evolution
soil moisture sensor based irrigation decisions in this field and could be of applied techniques. The main findings can be summarized as follows:
considered as one of the success stories for precision agriculture over
business as usual practices. The over application of nitrogen in corn – Average corn grain yield increased by ˃31% between phase 1 and
fields is more prevalent than other cereals as reported by (Luce et al., phase 3 and reached a stable record of ˃14 ton/ha since 2014.
2011) which highlights the importance of site-specific nitrogen appli­ – Average nitrogen application reduced by around 23% between phase
cation in corn fields to preserve the environment taking into account 1 and phase 3 nitrogen applied prescription maps at the study field.
that nitrogen is the most limiting nutrient in crop production. An – The average corn grain yield from the study field performed better
average of 10 to 30% of applied nitrogen fertilizer is typically lost due to than the regional average yield in phases 2 and 3 while there was no
nitrate leaching leading to environmental contamination (Li et al., 2007; significant difference in phase 1.
Schepers and Raun, 2008). A long term experiment with corn crop under – Soil moisture sensors prevented the crop from the extreme weather
supplementary irrigation showed an increase in nitrogen losses as the event in 2012 and maintained a high record of ˃11 ton/ha in the
rate of nitrogen applied increases (Aparicio et al., 2008). Furthermore, study field compared to the quite low regional average yield record
(Muschietti-Piana et al., 2018) reported a reduction of 18% in soil re­ of 6.4 ton/ha in the same year.
sidual nitrogen under site-specific management. Achieving long-term – Average agreement between the average yield map out of ten years
sustainable use of resources such as eco-efficient agriculture is an with each year yield map was around 90% which proves the con­
essential solution to face the rising food demand considering the envi­ sistency of yield patterns regardless of the different applied nitrogen
ronmental and economic dimensions of sustainability. Eco-efficient prescription maps.
agriculture means to produce more agricultural outputs with less use – Average agreement between average yield map and average GNDVI
of resources (Keating et al., 2010). from the last three years was higher than the ECa map which high­
An interview with the farm manager was done and he reported that light the possibility to use archived satellite images to investigate the
profit from this field considering all related costs increased by around historical yield variability in corn fields.
50€/ha which is more than the average reported profit by (Colaço and – The Partial Factor Productivity for Nitrogen (PFP-N) increased by
Bramley, 2018) of 33$/ha for site-specific nitrogen application in grain 60% between phase 1 and phase 3.
crops. In this study field the farmer had an irrigation system supported
by soil moisture sensors which would help explain this difference. Funding
(Tozer, 2009) reported that relative size of MZ has an influence on the
generated returns which could increase by taking out marginally pro­ This research was financially supported by the Land Environment
ductive zones. (Capmourteres et al., 2018) used different yield maps to Resources and Health (L.E.R.H.) doctoral course (http://www.tesaf.
calculate profit maps and found that up to 14% of farmland lost money unipd.it/en/research/doctoral-degrees-phd-lerh-program) and by the
and more than 50% might not meet minimum revenue expectations. CARIPARO foundation (AGRIGNSSVeneto - Precision positioning for
Such profit maps could encourage farmers to invest in PA tools if the precision agriculture project).
farmer’s field suffers from significant variations. In 1990 s PA applica­
tions and related technologies had a strong growing trend then the CRediT authorship contribution statement
market flattened through 2000 s because of the expensive costs
compared to expected benefits. Currently, PA related technology costs Ahmed Kayad: Conceptualization, Formal analysis, Data curation,
have been reduced and more profitable results at farm levels have been Writing - original draft, Validation. Marco Sozzi: Data curation. Simone
proved (Balafoutis et al., 2017). The global market of PA reached about Gatto: Data curation. Brett Whelan: Supervision, Validation. Luigi
2.3 Billion € in 2014 and expected an annual growing by 12% up to 2020 Sartori: Supervision. Francesco Marinello: Conceptualization, Formal
(Michalopoulos, 2015). analysis, Supervision, Validation.
Further, it is noted that yield spatial variability was large in the early
seasons and reduced year by year thanks to the efficient MZ delineation Declaration of Competing Interest
for nitrogen application and better irrigation decisions. Out of this study,
it is clear that accumulated data records from different seasons can The authors declare that they have no known competing financial
improve the fertilizer application, management and improve produc­ interests or personal relationships that could have appeared to influence
tivity year by year. Additionally, the use of archived satellite images has the work reported in this paper.
a promising rule for other farmers who do not have access to their his­
torical yield maps and delineating MZ. Currently, there is a growing Acknowledgments
number for PA services providers, especially the cloud-based companies
offering affordable subscriptions to archived remotely sensed data, and The cooperation and support extended by Massimo Salvagnin (Porto
their use is that expected to grow more in the near future (Paccioretti Felloni Co.) in providing the archived data is gratefully acknowledged.
et al., 2020).

10
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

References cropping system. Agron. J. 95 (2), 303–315. https://doi.org/10.2134/

agronj2003.3030.
Jones, J.W., Hoogenboom, G., Porter, C.H., Boote, K.J., Batchelor, W.D., Hunt, L.A.,
Alexander, L.V., Zhang, X., Peterson, T.C., Caesar, J., Gleason, B., Klein Tank, A.M.G.,
Wilkens, P.W., Singh, U., Gijsman, A.J., Ritchie, J.T., 2003. The DSSAT cropping
Haylock, M., Collins, D., Trewin, B., Rahimzadeh, F., Tagipour, A., Rupa Kumar, K.,
system model. Eur. J. Agron. 18 (3-4), 235–265. https://doi.org/10.1016/S1161-
Revadekar, J., Griffiths, G., Vincent, L., Stephenson, D.B., Burn, J., Aguilar, E.,
0301(02)00107-7.
Brunet, M., Taylor, M., New, M., Zhai, P., Rusticucci, M., Vazquez-Aguirre, J.L.,
Kayad, A., Paraforos, D.S., Marinello, F., Fountas, S., 2020. Latest advances in sensor
2006. Global observed changes in daily climate extremes of temperature and
applications in agriculture. Agriculture 10 (8), 362. https://doi.org/10.3390/
precipitation. J. Geophys. Res. Atmos. 111 (D5) https://doi.org/10.1029/
agriculture10080362.
2005JD006290.
Kayad, A., Sozzi, M., Gatto, S., Marinello, F., Pirotti, F., 2019. Monitoring within-field
Ali, A., Martelli, R., Lupia, F., Barbanti, L., 2019. Assessing multiple years’ spatial
variability of corn yield using sentinel-2 and machine learning techniques. Remote
variability of crop yields using satellite vegetation indices. Remote Sens. 2019, 11,
Sens. 11, 2873. https://doi.org/10.3390/rs11232873.
Page 2384 11, 2384. https://doi.org/10.3390/RS11202384.
Kayad, A.G., Al-Gaadi, K.A., Tola, E., Madugundu, R., Zeyada, A.M., Kalaitzidis, C., 2016.
Aparicio, V., Costa, J.L., Zamora, M., 2008. Nitrate leaching assessment in a long-term
Assessing the spatial variability of alfalfa yield using satellite imagery and ground-
experiment under supplementary irrigation in humid Argentina. Agric. Water
based data. PLoS One 11. https://doi.org/10.1371/journal.pone.0157166.
Manage. 95 (12), 1361–1372. https://doi.org/10.1016/j.agwat.2008.06.003.
Keating, B.A., Carberry, P.S., Bindraban, P.S., Asseng, S., Meinke, H., Dixon, J., 2010.
Balafoutis, A., Beck, B., Fountas, S., Vangeyte, J., Van Der Wal, T., Soto, I., Gómez-
Eco-efficient agriculture: Concepts, Challenges, And opportunities. Crop Sci. 50, S-
Barbero, M., Barnes, A., Eory, V., 2017. Precision agriculture technologies positively
109–S-119. https://doi.org/10.2135/cropsci2009.10.0594.
contributing to ghg emissions mitigation, farm productivity and economics. Sustain.
Keating, B.A., Carberry, P.S., Hammer, G.L., Probert, M.E., Robertson, M.J.,
9, 1–28. https://doi.org/10.3390/su9081339.
Holzworth, D., Huth, N.I., Hargreaves, J.N.G., Meinke, H., Hochman, Z., McLean, G.,
Blackmore, S., 2003. The role of yield maps in Precision Farming. PhD thesis, Cranfiled
Verburg, K., Snow, V., Dimes, J.P., Silburn, M., Wang, E., Brown, S., Bristow, K.L.,
University at Silsoe, UK.
Asseng, S., Chapman, S., McCown, R.L., Freebairn, D.M., Smith, C.J., 2003. An
Blackmore, S., Godwin, R.J., Fountas, S., 2003. The analysis of spatial and temporal
overview of APSIM, a model designed for farming systems simulation. Eur. J. Agron.
trends in yield map data over six years. Biosyst. Eng. 84 (4), 455–466. https://doi.
18 (3-4), 267–288. https://doi.org/10.1016/S1161-0301(02)00108-9.
org/10.1016/S1537-5110(03)00038-2.
Khosla, R., Fleming, K., Delgado, J.A., Shaver, T.M., Westfall, D.G., 2002. Use of site-
Bolfe, É.L., Jorge, L.A.d.C., Sanches, I.D., Luchiari Júnior, A., da Costa, C.C., Victoria, D.
specific management zones to improve nitrogen management for precision
d.C., Inamasu, R.Y., Grego, C.R., Ferreira, V.R., Ramirez, A.R., 2020. Precision and
agriculture. J. Soil Water Conserv. 57, 513–518.
digital agriculture: adoption of technologies and perception of Brazilian farmers.
Kizito, F., Campbell, C.S., Campbell, G.S., Cobos, D.R., Teare, B.L., Carter, B.,
Agriculture 10 (12), 653. https://doi.org/10.3390/agriculture10120653.
Hopmans, J.W., 2008. Frequency, electrical conductivity and temperature analysis
Bonfante, A., Monaco, E., Manna, P., De Mascellis, R., Basile, A., Buonanno, M.,
of a low-cost capacitance soil moisture sensor. J. Hydrol. 352 (3-4), 367–378.
Cantilena, G., Esposito, A., Tedeschi, A., De Michele, C., Belfiore, O., Catapano, I.,
https://doi.org/10.1016/j.jhydrol.2008.01.021.
Ludeno, G., Salinas, K., Brook, A., 2019. LCIS DSS—An irrigation supporting system
Li, C., Xiong, Y., Cui, Z., Huang, Q., Xu, X.u., Han, W., Huang, G., 2020. Effect of
for water use efficiency improvement in precision agriculture: A maize case study.
irrigation and fertilization regimes on grain yield, water and nitrogen productivity of
Agric. Syst. 176, 102646. https://doi.org/10.1016/j.agsy.2019.102646.
mulching cultivated maize (Zea mays L.) in the Hetao Irrigation District of China.
Bramley, R.G.V., Lawes, R.A., Cook, S.E., 2013. Spatially distributed experimentation.
Agric. Water Manag. 232, 106065. https://doi.org/10.1016/j.agwat.2020.106065.
Precision Agric. Sustain. Environ. Protection. 205–218. https://doi.org/10.4324/
Li, H., Liu, H., Gong, X., Li, S., Pang, J., Chen, Z., Sun, J., 2020. Optimizing irrigation and
9780203128329.
nitrogen management strategy to trade off yield, crop water productivity, nitrogen
Campbell, J.E., 1990. Dielectric properties and influence of conductivity in soils at one to
use efficiency and fruit quality of greenhouse grown tomato. Agric. Water Manag.
Fifty Megahertz. Soil Sci. Soc. Am. J. 54 (2), 332–341. https://doi.org/10.2136/
245, 106570. https://doi.org/10.1016/j.agwat.2020.106570.
sssaj1990.03615995005400020006x.
Li, X., Hu, C., Delgado, J.A., Zhang, Y., Ouyang, Z., 2007. Increased nitrogen use
Capmourteres, V., Adams, J., Berg, A., Fraser, E., Swanton, C., Anand, M., 2018.
efficiencies as a key mitigation alternative to reduce nitrate leaching in north china
Precision conservation meets precision agriculture: A case study from southern
plain. Agric. Water Manag. 89 (1-2), 137–147. https://doi.org/10.1016/j.
Ontario. Agric. Syst. 167, 176–185. https://doi.org/10.1016/j.agsy.2018.09.011.
agwat.2006.12.012.
Chlingaryan, A., Sukkarieh, S., Whelan, B., 2018. Machine learning approaches for crop
Luce, M. St., Whalen, J.K., Ziadi, N., Zebarth, B.J., 2011. Nitrogen dynamics and indices
yield prediction and nitrogen status estimation in precision agriculture: A review.
to predict soil nitrogen supply in humid temperate soils, 1st ed, Advances in
Comput. Electron. Agric. 151, 61–69. https://doi.org/10.1016/j.
Agronomy. Elsevier Inc. https://doi.org/10.1016/B978-0-12-385538-1.00002-0.
compag.2018.05.012.
Madugundu, R., Al-Gaadi, K.A., Tola, E., Kayad, A.G., Hassaballa, A.A., Patil, V.C.,
Cillis, D., Pezzuolo, A., Marinello, F., Sartori, L., 2018. Field-scale electrical resistivity
2017a. Seasonal dynamics of surface energy fluxes over a center-pivot irrigated
profiling mapping for delineating soil condition in a nitrate vulnerable zone. Appl.
cropland in Saudi Arabia. J. Environ. Biol. 38 (5), 743–751.
Soil Ecol. 123, 780–786. https://doi.org/10.1016/j.apsoil.2017.06.025.
Madugundu, R., Al-Gaadi, K.A., Tola, ElKamil, Kayad, A.G., Jha, C.S., 2017b. Estimation
Cogato, A., Pagay, V., Marinello, F., Meggio, F., Grace, P., De Antoni Migliorati, M.,
of gross primary production of irrigated maize using Landsat-8 imagery and Eddy
2019. Assessing the feasibility of using medium-resolution imagery information to
Covariance data. Saudi J. Biol. Sci. 24 (2), 410–420. https://doi.org/10.1016/j.
quantify the impact of the heatwaves on irrigated vineyards. Remote Sens. 11, 1–19.
sjbs.2016.10.003.
https://doi.org/10.3390/rs11232869.
McBratney, A., Whelan, B., Ancev, T., Bouma, J., 2005. Future directions of precision
Colaço, A.F., Bramley, R.G.V., 2018. Do crop sensors promote improved nitrogen
agriculture. Precis. Agric. 6 (1), 7–23. https://doi.org/10.1007/s11119-005-0681-8.
management in grain crops? F. Crop. Res. 218, 126–140. https://doi.org/10.1016/j.
Miao, Y., Mulla, D.J., Robert, P.C., 2018. An integrated approach to site-specific
fcr.2018.01.007.
management zone delineation. Front. Agric. Sci. Eng. 5, 432–441. https://doi.org/1
Delgado, J.A., Khosla, R., Bausch, W.C., Westfall, D.G., Inman, D.J., 2005. Nitrogen
0.15302/J-FASE-2018230.
fertilizer management based on site-specific management zones reduces potential for
Michalopoulos, S., 2015. Europe entering the era of ‘precision agriculture’ [WWW
nitrate leaching. J. Soil Water Conserv. 60, 402–410.
Document]. Euractiv.Com. URL http://www.euractiv.com/section/science-policyma
Domínguez-Niño, J.M., Oliver-Manera, J., Girona, J., Casadesús, J., 2020. Differential
king/news/europe-entering-the-era-of-precision-agriculture/ (accessed 6.23.20).
irrigation scheduling by an automated algorithm of water balance tuned by
Moral, F.J., Terrón, J.M., Rebollo, F.J., 2011. Site-specific management zones based on
capacitance-type soil moisture sensors. Agric. Water Manag. 228, 105880. https://
the Rasch model and geostatistical techniques. Comput. Electron. Agric. 75 (2),
doi.org/10.1016/j.agwat.2019.105880.
223–230. https://doi.org/10.1016/j.compag.2010.10.014.
EUROSTAT, 2019. European statistics on agriculture, forestry and fisheries [WWW
Muangprathub, J., Boonnam, N., Kajornkasirat, S., Lekbangpong, N., Wanichsombat, A.,
Document]. URL https://ec.europa.eu/eurostat/data/database (accessed 8.17.20).
Nillaor, P., 2019. IoT and agriculture data analysis for smart farm. Comput. Electron.
FAOSTAT, 2020. Food and Agriculture Organization, Statistics Data [WWW Document].
Agric. 156, 467–474. https://doi.org/10.1016/j.compag.2018.12.011.
URL http://www.fao.org/faostat/en/#data/QC (accessed 8.17.20).
Muschietti-Piana, M.D.P., Cipriotti, P.A., Urricariet, S., Peralta, N.R., Niborski, M., 2018.
Fleming, K.L., Westfall, D.G., Wiens, D.W., Brodahl, M.C., 2000. Evaluating farmer
Using site-specific nitrogen management in rainfed corn to reduce the risk of nitrate
defined management zone maps for variable rate fertilizer application. Precis. Agric.
leaching. Agric. Water Manage. 199, 61–70. https://doi.org/10.1016/j.
2, 201–215. https://doi.org/10.1023/A:1011481832064.
agwat.2017.12.002.
Grashey-Jansen, S., 2014. Optimizing irrigation efficiency through the consideration of
Nafi, E., Webber, H., Danso, I., Naab, J.B., Frei, M., Gaiser, T., 2019. Soil tillage, residue
soil hydrological properties - Examples and simulation approaches. Erdkunde 68 (1),
management and site interactions affecting nitrogen use efficiency in maize and
33–48. https://doi.org/10.3112/erdkunde.2014.01.04.
cotton in the Sudan Savanna of Africa. F. Crop. Res. 244, 107629. https://doi.org/
Higgins, S., Schellberg, J., Bailey, J.S., 2019. Improving productivity and increasing the
10.1016/j.fcr.2019.107629.
efficiency of soil nutrient management on grassland farms in the UK and Ireland
Nawar, S., Corstanje, R., Halcro, G., Mulla, D., Mouazen, A.M., 2017. Delineation of soil
using precision agriculture technology. Eur. J. Agron. 106, 67–74. https://doi.org/
management zones for variable-rate fertilization: A review. Adv. Agron. 175–245.
10.1016/j.eja.2019.04.001.
https://doi.org/10.1016/bs.agron.2017.01.003.
Jiang, H., Hu, H., Wang, S., Ying, Y., Lin, T., 2020. Understanding the impact of sub-
Paccioretti, P., Córdoba, M., Balzarini, M., 2020. FastMapping: Software to create field
seasonal meteorological variability on corn yield in the U.S. Corn Belt. Sci. Total
maps and identify management zones in precision agriculture. Comput. Electron.
Environ. 724, 138235. https://doi.org/10.1016/j.scitotenv.2020.138235.
Agric. 175, 105556. https://doi.org/10.1016/j.compag.2020.105556.
Jin, X., Kumar, L., Li, Z., Feng, H., Xu, X., Yang, G., Wang, J., 2018. A review of data
Pampolino, M.F., Witt, C., Pasuquin, J.M., Johnston, A., Fisher, M.J., 2012. Development
assimilation of remote sensing and crop models. Eur. J. Agron. 92, 141–152. https://
approach and evaluation of the Nutrient Expert software for nutrient management in
doi.org/10.1016/j.eja.2017.11.002.
cereal crops. Comput. Electron. Agric. 88, 103–110. https://doi.org/10.1016/j.
Johnson, C.K., Mortensen, D.A., Wienhold, B.J., Shanahan, J.F., Doran, J.W., 2003. Site-
compag.2012.07.007.
specific management zones based on soil electrical conductivity in a semiarid

11
A. Kayad et al. Computers and Electronics in Agriculture 185 (2021) 106126

Peralta, N.R., Barbieri, P., Gowland, T., Aparicio, V., Costa, J.L., 2015a. Precision Taylor, J.A., McBratney, A.B., Whelan, B.M., 2007. Establishing management classes for
agriculture: Variable rate nitrogen in Barley. Cienc. del Suelo 33, 131–137. broadacre agricultural production. Agron. J. 99 (5), 1366–1376. https://doi.org/
Peralta, N.R., Costa, J.L., 2013. Delineation of management zones with soil apparent 10.2134/agronj2007.0070.
electrical conductivity to improve nutrient management. Comput. Electron. Agric. Thompson, L.J., Ferguson, R.B., Kitchen, N., Frazen, D.W., Mamo, M., Yang, H.,
99, 218–226. https://doi.org/10.1016/j.compag.2013.09.014. Schepers, J.S., 2015. Model and sensor-based recommendation approaches for in-
Peralta, N.R., Costa, J.L., Balzarini, M., Castro Franco, M., Córdoba, M., Bullock, D., season nitrogen management in corn. Agron. J. 107 (6), 2020–2030. https://doi.
2015b. Delineation of management zones to improve nitrogen management of org/10.2134/agronj15.0116.
wheat. Comput. Electron. Agric. 110, 103–113. https://doi.org/10.1016/j. Tola, E., Al-Gaadi, K.A., Madugundu, R., Zeyada, A.M., Kayad, A.G., Biradar, C.M., 2017.
compag.2014.10.017. Characterization of spatial variability of soil physicochemical properties and its
Porter, J.R., Semenov, M.A., 2005. Crop responses to climatic variation. Philos. Trans. R. impact on Rhodes grass productivity. Saudi J. Biol. Sci. 24 (2), 421–429. https://doi.
Soc. B Biol. Sci. 360 (1463), 2021–2035. https://doi.org/10.1098/rstb.2005.1752. org/10.1016/j.sjbs.2016.04.013.
Rubio, V., Más, F.R., 2020. From smart farming towards agriculture 5.0: a review on crop Tozer, P.R., 2009. Uncertainty and investment in precision agriculture - Is it worth the
data management. Agronomy 10, 1–21. https://doi.org/10.3390/ money? Agric. Syst. 100 (1-3), 80–87. https://doi.org/10.1016/j.agsy.2009.02.001.
agronomy10020207. Visconti, F., de Paz, J.M., Martínez, D., Molina, M.J., 2014. Laboratory and field
Sawyer, J., Nafziger, E., Randall, G., Bundy, L., Rehm, G., Joern, B., 2006. Concepts and assessment of the capacitance sensors Decagon 10HS and 5TE for estimating the
rationale for regional nitrogen rate guidelines for corn concepts and rationale for water content of irrigated soils. Agric. Water Manag. 132, 111–119. https://doi.org/
regional nitrogen rate guidelines for corn. Iowa State Univ. Univ. Ext. 1–28. 10.1016/j.agwat.2013.10.005.
Schepers, J.S., Raun, W.R., 2008. Nitrogen in agricultural systems. In: Agron. Monogr. Vrindts, E., Mouazen, A.M., Reyniers, M., Maertens, K., Maleki, M.R., Ramon, H., De
ASA, CSSA, SSSA, Madison, Wisconsin. https://doi.org/10.2134/agronmonogr49. Baerdemaeker, J., 2005. Management zones based on correlation between soil
c12. compaction, yield and crop data. Biosyst. Eng. 92 (4), 419–428. https://doi.org/
Schwalbert, R.A., Amado, T.J.C., Nieto, L., Varela, S., Corassa, G.M., Horbe, T.A.N., 10.1016/j.biosystemseng.2005.08.010.
Rice, C.W., Peralta, N.R., Ciampitti, I.A., 2018. Forecasting maize yield at field scale Vrindts, E., Reyniers, M., Darius, P., De baerdemaeker, J., Gilot, M., Sadaoui, Y.,
based on high-resolution satellite imagery. Biosyst. Eng. 171, 179–192. https://doi. Frankinet, M., Hanquet, B., Destain, M.-F., 2003. Analysis of soil and crop properties
org/10.1016/j.biosystemseng.2018.04.020. for precision agriculture for winter wheat. Biosyst. Eng. 85 (2), 141–152. https://
Shapiro, C.A., Ferguson, R.B., Hergert, G.W., Wortmann, C.S., Walters, D.T., 2008. doi.org/10.1016/S1537-5110(03)00040-0.
Fertilizer suggestions for corn. Extension Circular EC117. Wang, X., Miao, Y., Dong, R., Zha, H., Xia, T., Chen, Z., Kusnierek, K., Mi, G., Sun, H.,
Solie, J.B., Monroe, A.D., Raun, W.R., Stone, M.L., 2012. Generalized algorithm for Li, M., 2021. Machine learning-based in-season nitrogen status diagnosis and side-
variable-rate nitrogen application in cereal grains. Agron. J. 104 (2), 378–387. dress nitrogen recommendation for corn. Eur. J. Agron. 123, 126193. https://doi.
https://doi.org/10.2134/agronj2011.0249. org/10.1016/j.eja.2020.126193.
Stanford, G., 1973. Rationale for optimum nitrogen fertilization in corn production. West, P.C., Gerber, J.S., Engstrom, P.M., Mueller, N.D., Brauman, K.A., Carlson, K.M.,
J. Environ. Qual. 2 (2), 159–166. https://doi.org/10.2134/ Cassidy, E.S., Johnston, M., MacDonald, G.K., Ray, D.K., Siebert, S., 2014. Leverage
jeq1973.00472425000200020001x. points for improving global food security and the environment. Science 345 (6194),
Stępień, M., Gozdowski, D., Samborski, S., Dobers, E.S., Szatyłowicz, J., Chormański, J., 325–328. https://doi.org/10.1126/science:1246067.
2016. Validation of topsoil texture derived from agricultural soil maps by current Yost, M.A., Kitchen, N.R., Sudduth, K.A., Sadler, E.J., Drummond, S.T., Volkmann, M.R.,
dense soil sampling. J. Plant Nutr. Soil Sci. 179 (5), 618–629. https://doi.org/ 2017. Long-term impact of a precision agriculture system on grain crop production.
10.1002/jpln.201500566. Precis. Agric. 18 (5), 823–842. https://doi.org/10.1007/s11119-016-9490-5.
Tabbagh, A., Dabas, M., Hesse, A., Panissod, C., 2000. Soil resistivity: A non-invasive tool Zhai, Z., Martínez, J.F., Beltran, V., Martínez, N.L., 2020. Decision support systems for
to map soil structure horizonation. Geoderma 97 (3-4), 393–404. https://doi.org/ agriculture 4.0: Survey and challenges. Comput. Electron. Agric. 170, 105256.
10.1016/S0016-7061(00)00047-1. https://doi.org/10.1016/j.compag.2020.105256.

12