LIBRARY '
Michigan State
University

 

 

 

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

fits/19% 502904

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DatoDuopBS—sz

SPATIAL AND TEMPORAL DYNAMICS OF NITROGEN-
WATER INTERACTIONS IN CORN IN MICHIGAN

by

Mohamed E. M. Elwadie

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Crop and Soil Sciences

2002

ABSTRACT

Spatial and Temporal Dynamics of Nitrogen-Water Interactions in
Corn in Michigan

Mohamed gyM. Elwadie

This study was initiated to evaluate sensor-based nitrogen and, water application
for corn (Zea mays L.) in Michigan. Our hypothesis was that sensor-based N and water
applications are more efficient for corn production than the current N fertilizer and
irrigation practices based on mass balance recommendations. Chapter I evaluates the
agronomic effects of N-water interactions in corn. The objectives were 1) to compare
conventional mass balance N treatments to sensor-based N management strategies based
on stress in plants, 2) compare dryland farming to irrigated corn, and 3) develop a crap.
coefficient for corn from remotely sensed data. Irrigation increased yield in 1999 but not
without supplemental N application. Nitrogen effectively increased corn grain yield, but
there was no significant difi‘erence between N applied at planting, or based on a pre-
sidedress nitrate test (PSNT), or sensor-based application. A nitrogen x water interaction
was present in the 1999 season, but was absent in the wet 2000 growing season. PSNT
values were obtained at V6 and at silking (R1) growth stages. EarleafN, stalk N, grain
N, and postharvest residual soil profile N were determined. A reflectance-based crop
coemcient (Ker) from green normalized vegetation index (GNDVI) is introduced for
estimating crop evapotranspiration. Chapter II evaluates the temporal dynamics and
biophysical variables estimation of corn mom in Michigan. High sensitivity to N

centered around 560 and 810 nm spectral bands. These two wavelengths provided the

provided the best separation between different N treatments. Corn grain yield was
correlated to different spectral vegetation indices (SVI) and chlorophyll meter readings.
The highest correlation between SVI and corn grain yield was achieved at R2, R3, R4,
and R5 growth stages. Green normalized vegetation index (GNDVI) performed better
than all other SVI when correlated to chlorophyll meter readings (R2 = 0.98). The results
suggest that GNDVI could be used to substitute for chlorophyll meter readings in N
scheduling. While NDVI and GNDVI reasonably estimated corn leaf area index (LAI),
SAVI overestimated LAI. All SVI performed very well in estimating fractional cover
(Fc) over the growing season. Chapter IH assesses the spatial variability of selected soil
properties and landscape attributes and their relationship to corn grain yield. Soil
properties such as length of the horizon, bulk density, soil water content at - 10, - 33, and
-1500 kPa, soil texture and landscape data such as elevation and slope magnitude were
collected from a transect 224 meters long at the study site. Variability for all properties
tested exhibited a strong spatial structure and 50 to 100 percent of the sample variance
was spatially dependent and autocorrelated over a range of 5 to 224 m. Stepwise multiple
regression analysis was applied to corn grain yields to identify the most important soil
properties in Ap and Btl horizons affecting yield variation. Ap depth, Ap silt, Btl bulk
density and Btl silt were the only significant factors in explaining yield variation. Corn
grain yield and soil properties in Ap and Btl horizon exhibited a very strong spatial cross
correlation. Fifty to 100 percent of the sample covariance was spatially dependent over a
range of 26 to 212 m. Cumulative probability and Spearman rank correlation coefficient,
the mean relative difference, and the standard deviation were used to characterize the

temporal stability of spatial soil water content and water storage over the growing season.

ACKNOWLEDGMENTS

I am grateful to Dr. Bernard Knezek who accepted the chair of my guidance
committee in a difficult time and for his support and encouragement. I thank Dr. Robert
von Bemuth who was always ready for my questions, Dr. J iaguo Qi who assisted me with
remote sensing analyses, and Dr. Kurt Thelen who helped me with editing the
manuscript; he was a valuable member of my committee. I thank Dr. Fran Pierce for his
guidance and his support of this study. I am grateful to Dr. James Jay who coordinated
the funding of this dissertation. I would like to thank the Crop and Soil Sciences
chairperson, Dr. Douglas Buhler, for his enthusiastic support and financial assistance to
complete this dissertation. A special thank you is owed to Dr. Richard Harwood who
introduced me to this department and Dr. Del Mokma who first accepted me in the Ph.D.
program. This work greatly benefitted from the knowledge and the vision of Dr. Ashton
Shortridge, Dr. Phil Robertson, and Dr. Sasha Kravchenko in the area of spatial
variability analysis. Special thanks to Betsy Bricker for her help and valuable input in
editing the manuscripts many times. I greatly appreciated the technical support of Brian
Long. Special thanks goes to Cal Bricker who taught me many technical and computer
skills, and he was always there ready to help and to support me with his friendship. I also
thank Greg Parker, Jose Sanchez, Jeanette Makries, and Ann Marie F ortuna for their
support and fi'iendship. Thank you to my father, mother, my brothers, and sisters who
encouraged me to pursue this endeavor. Above all I thank my wife Insaf, and my children

Aisha, Batran, Ala, and Ahmed for their suffering, patience and encouragement.

iv

TABLE OF CONTENTS

Page
LIST OF TABLES .................................................... vii
LIST OF FIGURES .................................................... ix
LIST OF ABBREVIATIONS ............................................ xi
INTRODUCTION ...................................................... 1
CHAPTERI PERFORMANCE OF SENSOR-BASEDN AND WATER
MANAGEMENT STRATEGIES FOR CORN IN MICHIGAN
ABSTRACT .................................................. 9
INTRODUCTION .............................................. 10
MATERIALS AND METHODS ................................... 17
RESULTS AND DISCUSSION ................................... 21
CONCLUSION . . . . . . ; ........................................ .. 29
REFERENCES ................................................ 43
CHAPTER II. REMOTE SENSING OF CORN CANOPY DYNAMICS AND
BIOPHYSICAL VARIABLES ESTIMATION IN MICHIGAN
ABSTRACT ..................... ' ............................ '47
INTRODUCTION .............................................. 48
MATERIALS AND METHODS ................................... 52
RESULTS AND DISCUSSION ............................. ' ...... 57
CONCLUSION ................................................ 61
REFERENCES ................................................ 73
CHAPTER III. ASSESSING THE SPATIAL VARIABILITY OF SELECTED
SOIL RESOURCES AND THEIR RELATIONSHIP TO CORN
GRAIN YIELD IN MICHIGAN
ABSTRACT ................................................ 1 . . . 75
INTRODUCTION ........................................ . ...... 76
MATERIALS AND METHODS ................................... 81
RESULTS AND DISCUSSION ................................... 85
CONCLUSION ........ g ........................................ 91
REFERENCES ............................................... 109

SUMMARY ........................................................ 1 13
APPENDIX ........................................................ l 18
REFERENCES ..................................................... 123

Table 1.1.

Table 1.2.
Table 1.3.

Table 1.4.
Table 1.5.
Table 1.6.

Table 1.7.

Table 1.8.
Table 1.9.

Table 2.1.

Table 2.2.

LIST OF TABLES

Page

Least square means for corn yield (Mg/ha) for two growing

seasons ................................................. 38
Nitrogen fertilizer applied and least square means of PSNT

(N 03—N) for different N treatments at V6 grth stage, silking

and afier harvesting for 2000 season ........................... 38
Analysis of variance table of KCl extract as soil NO3-N (kg/ha)

for PSNT at V6 leaf stage and at silking for 1999 and 2000

growing seasons .......................................... 39
Analysis Of variance table of KCl extract for soil profile NO3-N

after corn harvest for 1999 and 2000 growing seasons ............ 39
Least squares means for post-harvest soil profile NO3-N (kg/ha)

for 1999 and 2000 growing seasons .......................... 40'
Analysis of variance table for plant nitrogen for 1999 and 2000

growing seasons .......................................... 41
Least squares means for plant nitrogen for irrigated and

non-irrigated N treatments for a dry irrigated 1999 and a
relatively wet 2000 season .................................. 41
Average and effective vegetation indices, multiplier (a) and

Ofl‘set (b) from GNDVI, NDVI, and SAVI (2000) ............... 42
Coeflicient of determination (R2) and RMSE for ch and Kcr

for NDVI, SAVI and GNDVI for N Treatments (2000) ........... 42
Coefficients of determination (R2) and RMSE between corn

grain yield and wavelengths (510, 560, 610, 660, 710, 760

and 810 nm) at different dates for 1999 ....................... 62

Coefficient of deterrninatiOn (R2) and RMSE between corn

. grain yield and vegetation indices at different dates and growth

stages for 1999 and 2000. .................................. 63

vii

Table 2.3.

Table 3.1.

Table 3.2.

Table 3.3.

Table 3.4.

Table 3.5.

Table 3.6.

Table 3.7.

Coefficient of determination (R2) and RMSE between
measured and modeled LAI (from different vegetation indices)

for different N treatments (2000) ............................

Statistics for corn grain yield and selected soil physical properties
and landscape attributes for 48 sample locations along the main
transect. SD is the standard deviation and CV is the coefficient

of variation .............................................

Variogram model parameters and sample variance (5’) for

selected variables that investigated across the study site ..........

Variogram model parameters and sample variance (5’) for
soil water content and available water storage capacity at- 10
and- 33 kPa for 70 m range and the entire range of 224 In

across the study site ......................................

Stepwise multiple regression analysis of soil properties that

affect yield variation in the Ap and Btl horizons. ...............

The spatial relationship between corn grain yield (Mg/ha) and
selected soil properties that are significant in determining yield

variation in stepwise regression analysis. .....................

The spatial relationship between elevation (In) and Ap soil water
at difi‘erent pressures, Ap thickness and Ap silt over the entire

.64

94

.95

.95

.96

range of the study site ..................................... 97

The spatial relationship between slope (Va) and Ap soil water
at different pressures, Ap thickness, and Ap clay over the entire

viii

' range of the study site - ............................ p ......... 9 8

Figure 1.1.

Figure 1.2.

Figure 1.3.

Figure 1.4.
Figure 1.5.
Figure 1.6.

Figure 1.7.
Figure 2.1.
Figure 2.2.
Figure 2.3.

Figure 2.4a.

LIST OF FIGURES

Monthly average precipitation and potential evapotranspiration

for the years 1970 - 2000 at KBS. ............................

Nitrogen sufficiency index (N SI) for all N treatments.
3) an irrigated 1999 season, and b) a non-irrigated 2000

growing seasons .........................................

Effect of N treatments and irrigation on corn grain yield.
a) efiecm of irrigation and N treatment on corn grain yield in
The dry 1999 season, and b) efi‘ects of N treatments without

irrigation on corn grain yield for the wet 2000 growing season .....

Soil pre-sidedress nitrate test (PSNT) for N treatments at V6,
silking, and after harvesting for a) a dry irrigated 1999 growing

season, and b) a wet 2000 grOwing season without irrigation .......

Effect of N treatments on soil profile NO3-N distribution in
the effective root zone after harvesting for a) a dry 1999

growing season, and b) a wet 2000 growing season ..............

Basal crop coefficient (ch Wright) and crop coefficient (Ker)
from vegetation indices vs the day of the year for irrigated and

non-irrigatedN treatments (1999) .................... , ........ '

Temporal performance of ch (Wright) and Kcr from vegetation

indices for all N treatments (2000) vs the day of the year ..........

Nitrogen sufl'rciency index (N SI) for all N treatments. a) an.

irrigated 1999 season, and b) a non-irrigated 2000 seasons ........

Percentage reflectance for N treatments at different wavelengths,
with a) an irrigated 1999 season and b) a non-irrigated 2000 season

Relative reflectance for N treatments at all wavelengths

with a) irrigation in 1999 and b) without irrigation in 2000 ........ 67

Corn grain yield for all N treatments vs vegetation indices

31

32

33

34

35

36

37

65

66

on August 12, 1999 ....................................... 68

ix

Figure 2.4b.

Figure 2.5a.

Figure 2.5b.

Figure 2.6.

Figure 3.1a.

Figure 3.1b.

Figure 3.2.

Figure 3.3.

Figure 3.4.

Figure 3.5.

Figure 3.6.

Figure 3.7.

Figure 3.8.

Figure 3.9.

Figure 3.10.

Corn grain yield for all N treatments vs vegetation indices

on August 9, 2000 ........................................ 69
Modeled leaf area index (LAI) fi'om vegetation indices using

empirical coefficients for all N treatments (2000) ................. 70
Measured vs Modeled LAI fi'om vegetation indices for

at-planting N treatment (2000) ............................... 71
Modeled fractional cover from vegetation indices for

all N treatments (2000) ..................................... 72
Sample locations of the study site overlaid over the contour lines . . . . 99
Sample locations of the main transect overlaid over elevation ....... 99

Variograms of soil properties for Ap horizon, Ap thickness,
Ap texture, bulk density, elevation, slope, and corn grain ,
yield (from residuals) ..................................... 100

Variograms of soil water in Ap horizon. AW is available water
and AWSC is available water storage capacity ................. 101

Cross variograms of residual corn grain yield (Mg/ha) and
significant soil properties in the Ap and Btl horizons and
elevation. Dotted lines indicate sample covariance .............. 102

Cross variograms of elevation and soil water at different
pressures, Ap thickness, and Ap silt. ......................... 103

Cross variograms of slope and Soil water at different pressures,
Ap thickness and Ap clay. ................................. 104

Volumetric water content (TDR) for three different dates at
five different depths in the effective root zone .................. 105

Cumulative probability of soil water storage (0 - 90 cm)
for two different dates: the most dry storage on June 2, 2000 and
the most wet storage on August 16, 2000. ..................... 106

Rank of temporal relative deviation fi'om the mean spatial water
storage. Vertical bars associated with time standard deviations
and numbers are measuring locations or plots. .................. 107

Rank of temporal relative deviation from the mean of spatial
water content. Vertical bars are associated with time standard
deviations and numbers are measuring locations or plots .......... 108

X

Etc

Etp

Fc
GNDVI
GRVI
IRt

Kcr
KBS
ch (Wright)
LAI
NDVI
RVI .
SAVI
SVI .
TDR
Oij

Yo)

LIST OF ABBREVIATIONS

Potential evapotranspiration of the crop

Potential evapotranspiration

Fractional cover

Green normalized difference vegetation index
Green ratio vegetation index ,

Infrared thermometer

Reflectance based crop coefficient

W. K. Kellogg Biological Station

Wright basal crop coefficient

Leaf area index

Normalized diflerence vegetation index
Ratio vegetation index

Soil adjusted vegetation index

Spectral vegetation indices

Time Domain Reflectometry

The relative difference in soil water content

Semivariance for interval distance class h

xi

INTRODUCTION

This study was initiated to evaluate sensor-based nitrogen and water application
for corn (Zea mays L.) in Michigan. The study has three main Objectives relative to the
understanding of temporal and spatial variability of processes and properties that regulate
crop performance and crop yield within a field. First, this study tests the feasibility of
stress-based N and water management for corn production in terms of crop yield, N
leaching potential, and irrigation scheduling. Our hypothesis was that sensor-based N and
water applications are more emcient fOr corn production than current fertilizer and
irrigation practices based on mass balance recommendations. Therefore these sensor-
based strategies must be site-specific, maximize corn yield and N and water use
efliciency, and minimize N losses to the environment. Second, the study identifies
wavelengths that are most sensitive to N deficiency and facilitate corn yield estimation.

In addition, the investigation identifies the best growth stage for corn yield estimation and
estimates biophysical variables of corn such as the leaf area index (LAI) and fi'actional
cover (F c) fiom multi-spectral data acquired over a corn canOpy during the growing
season. Finally, this study asesses the spatial variability of selected soil physical
pr0perties and landscape attributes and their relationship to corn grain yield. 1
Chapter I Overview

Current N fertilizer recommendations are based on yield response data averaged
over a large landscape area resulting in a dilution of the accuracy of the response
relationships (Buchholz, 1983). Nitrogen response and water distribution are field and

season dependent and can vary greatly within the same field in the same season. Spatial

1

variability within the same field makes it difficult to develop variable rate N and water
recommendations for corn production that result in optimal N and water application and
maximum economic returns.

Nitrogen management strategies for corn can be approached in two basic ways:
mass balance strategies, whereby prescriptive applications of N inptrts are made prior to
or early in the N uptake phase of plant grth to avoid N deficiencies or as an interactive
strategy, whereby N inputs are applied to meet N requirements as determined by the
nutrient status of soils or plants during the rapid N uptake phase of corn.

Mass balance has been the most commonly used N management strategy. In this
system fertilizer recommendations are usually based on a combination of yield goal, N
requirement Of the crop, residual soil N, and N mineralized from soil or plant residues
and sometimes using soil organic matter as a proxy for N mineralization (Pierce and
Novak, 1999). Interactive strategies include foliar fertilization, delayed N applications
based on N content in plants, and chemigation strategies that assess crop needs based on
tissue sampling, sensing plant reflectance, or crop simulation.

Schepers et a1. (1996) suggested that strategies which sense the occurrence of N
deficiencies in plants during the uptake phase may be better adapted to precision N
management than strategies that use mass balance approaches that are, to a large extent,
soil sampling based. The interactive N strategies are based on the fact that N deficiency
in corn reduces chlorophyll content of leaves thereby increasing the amount of light
transmitted through a leaf. The idea is to 1) monitor plant N concentration by monitoring

plant or canopy reflectance of light or some measure of plant N content such as

chlorophyll content, 2) estimate N fertilizer requirement using relationships between
reflectance and plant N content, and then 3) fertilize the crop to the optimal N content for
maximum economic yield. Often, a portion of a field is fertilized to optimal levels and
the well-fertilized corn used as the standard for adjusting the N recommendations for the
remainder of a field (Schepers et al., 1996). Bausch et al. (1996) estimated plant N using
a previously developed index calculated fiom measured canopy reflectance

Water is essential to crop productivity since crop yields generally increase linearly
with water transpired by a crop (Howell, 1990). Improved water management
necessitates accurate estimation of a daily crop evapotranspiration (ETC). Improved
water management is essential to optimize water relations for plants because it has a
direct impact on the fate and transport of pollutants to surface and ground waters.

There are two areas of great interest in precision irrigation management. The
major part of research has been made with variable rate irrigation systems, mainly with
sprinkler irrigation provided by center pivot and linear move systems (Camp and Sadler,
1994; Evans et al., 1996; King et 51., 1995; King et al., 1996; and McCann and Stark,
1993). The second part of variable rate irrigation research has been focused on
developing irrigation scheduling programs based upon remote sensing vegetation indices
obtained either from plant canopy reflectance and/or from canopy temperature. The
introduction of a basal crop coefficient (ch) by Wright (1982) improved irrigation
scheduling. The problem with the Wright basal crop coefficient is that it is dependent on
time based parameters such as planting date and effective cover and the resistance to

water movement from soil and plant surfaces. Neale et al. (1989) developed a

reflectance-based crop coefficient (Ker) fiom a normalized difference vegetation index
(NDVI). The NDVI crop coefficient is independent of planting date and effective
fractional cover since NDVI is a measure of the photosynthetic size of the plant canOpy.
Therefore, Ker is a true representation of crop growth and development and is affected by
nutrient and water stress, and diseases. Bausch et al. (1996) have developed a Ker from a
soil adjusted vegetation index (SAVI) to reduce the effect of soil background on canopy
reflectance. Green normalized vegetation index (GNDVI) is another candidate parameter
that could be used to develop crop coefficients from plant reflectance to be used for
Scheduling irrigation. The canopy temperature approach for variable rate irrigation
obtains a measure of crop stress based on remotely sensed temperature differences
between the surface and air temperatures. The most widely used indices are the crop
water stress index (CWSI) which ranges from zero to one (Idso, 1982 and Jackson, 1982),
and the water deficit index (WDI) (Moran et al., 1994). Both WDI and CWSI are
commercially available to farmers and can be used as a basis for applying variable
irrigation water based on water stress in corn rather than on the current basis of demand
or weather-based water budgets which do not account for spatial variability within a field.
More importantly, it is possible to remotely monitor water stress in plants and apply
irrigation water based on the site-specific needs of corn within a field rather than
uniformly applying irrigation water over a field as is the current practice.
Chapter II Overview

Optical remote sensing provides a powerful tool for monitoring changes in a crop

over the growing season and can provide crop developmental information that is time-

critical for site-specific crop management. Leaf chlorophyll is positively correlated with
N concentration (Al-Abbas et al., 1974). Remote sensing has been used to characterize
properties of vegetation, estimate yield, estimate total biomass, and to monitor plant
health and plant stress (Jackson, 1986).

Many spectral vegetation indices (SVI) have been presented in the literature.
Jackson and Huete (1991) interpreted some of the vegetation indices. Spectral vegetation
indices (SVI) were used to measure the photosynthetic size of the plant canopy and to
estimate yield (W iegand et al., 1990). The most used indices are: ratio vegetation index
(RVI), NDVI, and soil adjusted vegetation index (SAVI). A major advantage of the SVI
is that they reduce multi-band observations into a single numerical index. Biophysical
variables that are of great interest and can be estimated from remote sensing data are
grain yield, leaf area index (LAI), and fractional cover (Fe).

Remote sensing has been used for yield estimation where grain yield was
correlated to a single measurement of NDVI. Tucker et al. (1980) used NDVI as a time
integral to estimate yield. Blackmer et al. (1996) correlated relative corn grain yield to
reflective radiation from a photometric cell centered around 550 nm. They concluded that
this wavelength was alSo more sensitive than other wavelengths to N stress in corn at dent
time. Blackmer et al. (1996) analyzed aerial photos for N stress and found that black and
white. photos centered around 536 nm also predicted yield at the R5 growth stage.
Therefore, identification of spectral wavelengths that are sensitive to N stress will aid in

Site-specific management of corn.

Leaf area index (LAI) is an important parameter for light interception by plants
and is used extensively as an input to predict crop response in many simulation models.
Remote sensing has been used to estimate biophysical variables such as LAI and Fe
(fractional cover). There are three approaches as reported by Qi et al. (2000a) to estimate
LAI. These include a simple regression that uses multi-band regression, a modeling
approach, and a vegetation index approach. The vegetation index approach is associated
with SVI such as NDVI. Here the LAI can be estimated using an empirical relationship.
Fraction cover (F c) is another very important property of vegetation. It has a direct role
in daily agronomic decisions such as irrigation. Qi et al. (2000a) used an empirical
approach to estimate LAI and Fe from imagery acquired from Landsat TM and from
aircraft sensors.

Measurements of a canopy multi-spectral reflectance can provide a quick, cost
effective , and non-destructive real time method of assessment of nutrient status and
biophysical variable estimation of the whole field. Otherwise, the current methods of
cr0p assessment are tedious, laborious, costly and time consuming.

Chapter III Overview

Precision Agriculture (PA) or site-specific crop management (SSCM) involves
the management of spatially variable factors according to localized conditions (Larson
and Robert, 1991). Precision agriculture is concerned with variability in the both space
and time. Soil properties and processes that regulate corn production and environmental

sensitivity vary in space and time (Lake et al., 1997; Pierce and Nowak, 1999; and Bell et

al., 1995). Therefore, understanding the relationship between crop yield and
environmental spatial variables is essential for SSCM.

Without variability, the concept of precision agriculture has no meaning. Mulla

and Schepers (1997) suggested that the most important factors in variability include

spatial patterns in pest infestations, plant available water, soil drainage, crop rooting
depth, nutrient availability, soil texture, organic matter content and pH. The challenge for
PA is to determine which major factors are responsible for variations in crOp yield.

Spatial data do not comply with the assumptions of classical Statistics, in
particular independence. Tobler’s first law of geography (Tobler, 1977) states that
“everything is related to everything else, but near things are more related than distant
things”. Soil properties fi'equently exhibit spatial dependence, i.e., samples collected
close to one another are often more similar in value than widely spaced samples
(Trangmar et al., 1985). There are many geostatistical methods available that have been
adopted for detecting the auto-correlation in environmental data for the characterization
of spatial dependence of soil properties. The most widespread is the semivariance
analysis. Semivariance analysis provides a versatile and unbiased means for examining
autocorrelation in environmental data (Robertson and Gross, 1994). A semivariograrn is
basically a plot of dissimilarity (semi-variance) between samples against distance
between samples (Issaks and Srivastava, 1989). The semi-variance ideally increases with
distance between sample location, to a more or less constant value (the sill or total semi-
variance or the population variance) at a given separation distance, called the range of the

spatial dependence (Trangmar et al., 1985). Attributes separated by distances closer than

the range are spatially related. Those separated by distances greater than the range are not
spatially related. The semi-variance, when the distance of sample separation tends to
zero, or y -intercept, is called the nugget variance or nugget effect (Webster, 1985). The
nugget effect gives an indication of variability at seales less than the data spacing and/or
due to analytical or random error. The difference between the sill and the nugget is the
structural variance. Moreover, the semivariograrn documents whether there are spatial
components to the variability and how distinct the patterns may be. The semivariogram
also reveals the spatial scale over which autocorrelation occurs.

Though variability may occur naturally as a result of complex geological and
pedological or manmade processes, knowing the degree of spatial dependence and its
extent is critical for evaluating its agricultural significance. Application of geostatistical
techniques may provide more power for analyzing spatial variability and semivariance
analysis gives a robust means for quantifying autocorrelation in spatial and temporal
dimensions.

To characterize soil water content for precise variable rate irrigation application,
the temporal stability of spatial soil water content variation has to be determined.
Vachaud et al. (1985) used cumulative probability function and a mean relative difference
technique to evaluate the temporal stability of spatial soil water storage. Cassel et al.
(2000) applied the same technique to evaluate the temporal stability of soil spatial water
' content. The method uses the Spearman rank correlation coefficient as a criteria to

evaluate the temporal stability of spatial soil water content.

CHAPTER I

PERFORMANCE OF SENSOR-BASED N AND WATER MANAGEMENT
STRATEGIES FOR CORN IN MICHIGAN

ABSTRACT

This study was initiated to evaluate sensor-based nitrogen and water application
for corn (Zea mays L.) in Michigan. Our hypothesis was that sensor-based N and water
applications are more efficient for com production than current fertilizer and irrigation
practices based on mass balance recommendations. Therefore, these sensor-based
strategies must be site-specific and maximize N and water use efficiency. The specific
objectives were 1) comparing at-planting and pre-sidedress nitrate test (PSNT) N
management with sensor-based N strategies based on stress in plants determined by leaf
chlorophyll readings, 2) comparing dryland farming to irrigated corn, and 3) developing a
crop coefficient from remotely sensed data that can be used in irrigation scheduling.
Intervention N management strategies through PSNT and leaf chlorophyll sufficiency
performed as well as at-planting N management strategies. There was a N x water
interaction in the 1999 season but none in the wet 2000 season. Irrigation increased yield
but not without N application. Nitrogen was the main factor determining yield in the wet
2000 season. We obtained PSNT values at V6 and at silking (R1) stages of growth, and
earleaf N, stalk N, grain N, and postharvest soil profile N in the effective rooting depth.
Canopy multispeetral reflectance was measured in the visible and near infrared spectral
bands. A real time reflectance-based crop coefficient (Ker) from the green normalized

vegetation index (GNDVI) is introduced for estimating crop evapotranspiration.

INTRODUCTION

Nitrogen Management Strategies

Current N fertilizer recommendations and practices are based on response data
averaged over a large landscape area which dilutes the accuracy of response relationships
(Buchholz, 1983). The N fertilizer recommendations that are based on an average
response function and yield goal are poorly correlated with actual economically optimum
N rates (Doerge, 2001). Nitrogen response and water distribution are field and season
dependent and can vary greatly within the same field in the same season. Spatial
variability within the same field makes it difficult to calculate variable rate N application
rates and water recommendations for corn production for optimal application and
maximal economic returns. Precision Agriculture (PA) was designed to target crop and
soil inputs according to a specific field requirement to optimize profitability and protect
the environment. Soil properties and processes that regulate corn production vary in both
space and time (Lake et al., 1997, and Pierce and Nowak, 1999). There is an increasing
interest in whether this variability implies that there could be an advantage in managing
field crops in a spatially variable way instead of treating them according to average
conditions (Robert et al., 1993). Therefore, understanding the relationship between crop
yield and environmental spatial variables is essential for site-specific crop management
(SSCM). Site-specific crop management involves the management of spatially variable
factors according to their localized conditions (Litrson and Robert, 1991, and Schueller,
1992). The ability to identify differences in crop N status within fields could lead to

efficiencies in N fertilizer application (Blackmer et al., 1996). Therefore, it becomes

10

essential to identify differences within the field for efficient variable rate N and water
application.

Variable rate N recommendations started as a designated or a prescripted approach
where N was applied at planting to prevent any nutrient deficiencies. Fields were divided
into management zones and recOmmendations using prescribed methods were applied to
individual zones. Ferguson et al. (1997) reported map units from grid soil sampling and
Kitchen et al. (1995) suggested crop productivity zones; This prescription N management
did not improve N use efficiency. The interactive N management approach relied heavily
on remote sensing data and assessed crop needs based on tissue sampling and sensing
plant reflectance. This approach depended on an interactive intervention of N
management where N was applied to meet real time crop requirements depending on
sensing N in plants. Schepers et a1. (1992) delayed N applications based on N content in
plants. They estimated N fertilizer requirements using relationships between reflectance
and plant N content, and fertilized the crop to the optimal N content for maximum I
economic yield. Schepers et al. (1996) suggested that using strategies which sense the
occurrence of N deficiencies in plants during the uptake phase may be better adapted to
precision N management than strategies that use mass balance approaches that are based
on soil sampling. The idea is to monitor plant N concentration through chlorophyll
content in the plant by monitoring plant or canopy reflectance of light. Often, some
portion of a field is fertilized to optimal levels and the well-fertilized corn used as the
standard for adjusting the N recommendations for the remainder of a field (Schepers et

al., 1996). Some researchers estimated plant N using a previously developed index

11

calculated fiom measured canopy reflectance (Bausch et al., 1996). Blackmer et al.
(1995) proposed that the images of canopy reflectance centered at 550 nm acquired late in
the season could be used to detect portions of the field that were nitrogen deficient.
Measuring light reflected from a plant canopy can be used to detect N deficiency in corn
(Blackmer et al., 1996) and to trigger N fertilizer applications based on plant need.

From an energy balance vieWpoint, all solar radiant flux incident upon any object
is either reflected, transmitted, or absorbed. Vegetation is unique in the three~segment
partitioning of solar irradiance. In the visible part ofthe spectrum (400 - 700 nm),
reflectance is very low, transmittance is nearly zero, and absorption is high. The
fundamental control of energy-matter interactions with vegetation in this part of the
spectrum is plant pigmentation. In the longer wavelengths of the near-infiared portion of
the spectrum (700 - 1350 nm), both reflectance and transmittance are high whereas
absorption is very low. Intervention N strategies are based on the fact that N deficiency
in corn reduces the chlorophyll content of leaves thereby increasing the amount of light
transmitted through a leaf. A typical reflectance curve shows a peak at around 550 nm
corresponding to green color due to reflection of light by chlorophyll and a dip in the
plant spectral signature around 690 nm corresponding to a red color due primarily to
chlorophyll absorption. Therefore, measuring multispectral reflectance of the plant
canopy is a non-destructive method, a whole field approach, and a real time assessment of
the crop.

Although technology is available to deliver variable N rates of fertilizer across

fields, this practice in not yet economically beneficial in most cases and the potential cost

12

savings are minimal (Doerge, 2001). Therefore, there is a need for a new N management
strategy that fertilizes the crop to the optimal N content for maximum economic yield and
protects the environment fi'om nitrogen leaching.
Water Management Strategies

Water is critical to crop productivity since crop yields generally increase linearly
with water transpired by a crop (Howell, 1990). Improved water management
necessitates an accurate estimation of daily crop evapotranspiration. Water management
is also critical to maintaining high water quality and to increasing water use efficiency.
Techniques that optimize water relations for plants have a direct impact on the fate and
transport of pollutants to surface and ground waters. Differences in water availability
within a field are favored by l) the occurrence of dissimilar soil types; 2) the presence of
degradation processes (e.g., erosion, compaction, salinity); and 3) variation in the
landscape which affects the horizontal distribution of water. Pierce et al. (1995)
suggested that soil physical properties or landscapes (especially their effect on plant water
relations) may be even more important than soil fertility in explaining yield variability.

There is a significant amount of research in the area of variable rate irrigation
management with two, well-connected areas of interest in precision irrigation
management. The major portion of research has focused on variable rate irrigation
systems, mainly with sprinkler irrigation provided by center pivot and linear move
machines (Camp and Sadler, 1994; Evans et al., 1996; King et al., 1995,1996; and

McCann and Stark, 1993). The second part of variable rate irrigation research has

13

focrlsed on developing irrigation scheduling programs from remote sensing vegetation
indices either from plant canopy reflectance or from canopy temperature.

The introduction of the basal crop coefficient (ch) by Wright (1982) successfully
improved irrigation scheduling. The problem with the Wright basal crop coefficient is
that it is dependent on time based parameters such as planting date and effective cover
and the resistance to water movement from soil and plant surfaces. Specific crop
evapotranspiration is a product of the crop coefficient and evapotranspiration rate from a
reference crop. Neale et al. (1989) developed a reflectance-based crop coefficient (Ker)
from a normalized difference vegetation index (NDVI). They developed an equation of
the form:

Kcr=a*NDVI+b (1.1)

where: a = multiplier, and b = offset

a = (chw

ective

_ Kw“) / (NDVLMM — NDW,,,.,) (1.2)

ch effective = 0.93 at effective cover and chsoil = 0.15 at dry soil conditions.

b = (Kebsotl — a) * NDWMII) (13)

Neale et al. (1989) reported two equations for two research sites:
Kcr = 1.092 * NDVI - 0.053 (1.4)
Ker = 1.181 * NDVI - 0.026 (1.5)

The NDVI crop coefficient is independent of planting date and effective fractional
cover since NDVI is the measure of photosynthetic size of plant canopy. Therefore, Ker

is a true representation of crop growth and development and is affected by nutrient and

14

water stress, and diseases. The NDVI is computed by ratioing spectral reflectance of
visible red light (R= 625 - 675 nm) and the near infrared light (NIR= 725 - 810 nm) as
(NIR - R)/(NTR + R) and ranges from -1 to 1. Bausch (1993) used the soil adjusted
vegetation index (SAVI) to predict evapotranspiration fi'om a reflectance-based crop
coefficient for corn that can be used as a basis for variable rate irrigation. Bausch et al.
(1995) used the same form of the equation developed by Neale et al. (1989) and
substituted SAVI for NDVI to reduce the effect of soil background on canopy reflectance.
Green normalized vegetation index (GNDVI), which is the ratioing of the near infiared
(NIR= 725 - 810 nm) and the green reflectance (560 nm)‘ as (NIR - G)/(NIR + G) is
another candidate that could be used to develop a crop coefficient for scheduling
irrigation from plant reflectance data

The mom temperature approach to variable rate irrigation is accomplished by
obtaining a measure of crop stress based on remotely sensed temperature differences ,
between the surface and air temperatures. The most widely used index is the crop water
stress index (CWSI) which ranges from 0 to l (Idso, 1982, and Jackson, 1982), and the
water deficit index (WDI) (Moran et al., 1994). Both the WDI and CWSI indices are
commercially available to farmers and can be used as a basis for applying variable rates
of irrigation water based on water stress in corn rather than on the current basis of crop
demand or weather-based water budgets which do not account for spatial variability
within a field. More importantly, it is possible to remotely monitor water stress in plants
and apply irrigation water based on site-specific nwds of corn within a field rather than

uniformly over a field as is the current practice.

15

Our hypothesis was that sensor-based N and water applications are more efficient
for corn production than current fertilizer and irrigation practices based on mass balance
recommendations. Therefore, sensor-based N and water management strategies can be
site-specific and maximize corn yield and N and water use efficiency and minimize N
losses to the environment.

Objectives

The overall objectives of this study were to determine if variable rate N fertilizer
application based on plant chlorophyll sensing and variable irrigation based on vegetation
indices such as NDVI, SAVI, and GNDVI are more efficient for corn production than the
current fertilizer and irrigation practices based on mass balance recommendations.

The specific objectives of this study were:
- To compare the effectiveness Of at-planting and PSNT N application strategies
with intervention N management strategies based on stress in plants determined

from remotely sensed chlorophyll measurements,

- To compare N fertilizer application strategies on dryland farming to irrigated com,
and
- To develop irrigation scheduling recommendations based on remotely sensed data.
MATERIALS AND METHODS
Site Description

The study was conducted at the W. K. Kellogg Biological Station (KBS), which is

located in Southwest Michigan (850 24' W longitude, 420 24' N latitude) in the northern

16

portion of the Midwest corn belt. The W. K. Kellogg Biological Station is on the pitted
outwash plain of the moriarric system lefi by the last retreat of the Wisconsin glaciation,
about 12000 years ago (Robertson et al., 1997). Soils in the area were developed on the
glacial outwash; soils at the site are Typic Hapludalfs, either fine-loamy, mixed, mesic
Kalamazoo series or coarse-loamy, mixed, mesic Oshtemo series (Mokrna and Doolittle,
1993). Mean annual temperature at KBS is 9.4°C; precipitation is approximately 920 mm
annually and spread evenly all throughout the year, and potential evaporation exceeds
precipitation (30 years mean) for three months per year (Crum et al., 1990). These soils
respond to irrigation in most years and are vulnerable to leaching because of the coarse
textured soil materials and shallow depth of the groundwater. The experimental area is
gently sloping, decreasing in elevation from west to north and decreasing from east to
west. It is well known from previous studies that the soils are variable over'short
distances making it possible to conduct spatial variability studies (Francis J. Pierce,
Personal communication).
Experimenml Design

The experiment evaluated two irrigation and four N treatments in a split-plot
experimental design. Irrigation was the main plot and N treatments were subplots with
four replications. The irrigation treatments evaluated were 1) none (control), and 2)
irrigation scheduled according to the Michigan irrigation scheduling program (Shayya et
al., 1990). The four N fertilizer treatments evaluated included; 1) none (control), 2) N
applied at planting, (200 and 145 kg/lra for irrigated and non-irrigated corn respectively),

3) N application based on PSNT levels plus yield at the V6 leaf stage of corn, and 4)

17

stress based N applied in 70 kg/ha increments using remotely sensed chlorophyll
measurements.
Measurements

Water was applied to N treatments using drip irrigation. Since N management and
soil parameters affect corn growth and development, each N plot within the irrigation
treatment was irrigated independently using a drip irrigation system devised to apply
water between each row. The irrigation scheme used in the 1999 season consisted of
applying 12.7 mm (0.5 inch) of water every 3 to 4 days; allowances were made for
rainfall. Irrigation was scheduled in 1999 with the MSU SCHED program when 50% of
plant available water was depleted from the crop root zone. The crop was exposed to
water stress early in 1999 before irrigation lines were installed. A weather station at KBS
provided input for irrigation scheduling calculations. A rain gauge was maintained at the
Site to ensure a site-specific rainfall record. Soil profile water content was measured
weekly from planting to maturity using Time Domain Reflectometry (TDR). Time
Domain Reflectometry access tubes were inserted in the middle of each plot prior to
planting to a depth of 90 cm. A tube access probe with a mobile moisture meter (Trime
Moisture Measuring System“) was used to measure volumetric soil water content at 18,
36, 53, 71, and 89 cm intervals on a weekly basis.

A multi-spectral ground based radiometer (CropScan, MSR87) was used to
measure the green, red, and NIR Spectral reflectance bands. The CropScan radiometer
contains a set of eight narrow band filters, centered at:

460 nm, 510 nm , 560 nm, 610 nm, 660 nm, 710 run, 760 nm, and 810 nm

18

The CropScan radiometer has a 28 degree field of view for reflected irradiation
and the sensor was mounted on a 2.62 In pole above the canopy at nadir viewing and a
cherry picker was used when the crOp was very high. Two measurements per plot were
taken for each sampling date in 1999* and three measurements per plot were taken in 2000
between the daylight hours of 10:30 am and 1:30 pm local time. The sensor was oriented
parallel to the corn rows resulting in an oval ground area of about 1.53 m2. Weekly
multi-spectral measurements Were taken using the CropScan radiometer fiom the V6
growth stage to the R5 (dent) growth stage. Leafarea was measured from V6 (2000
season only) to silking for each plot using the LAI-2000 plant canopy analyzer (Li Cor 0"
leaf area meter).

All fertilizer N was applied as 28% solution using a precise liquid fertilizer
applicator equipped with a capacity to apply N throughout the N uptake phase of corn
growth and development. Fertilizer N recommendations forcom following cornwas
based on MSU fertilizer recommendations (N recommendation = -27 + 1.36“ yield goal).
For PSNT N treatment, N fertilizer rates were reduced based on the PSNT values of
6*ppm'N in the surface 0- 30 cm (N recommendation = 27 + 1.36* yield goal - 6*ppm).
Nitrogen applied at planting time was 145 kg/ha for a yield goal of 7,800 kg/ha for
non-irrigated corn and 202 kg/ha was applied for a yield goal of 12,480 kg/ha for irrigated
corn. Nitrogen was applied in 70 kg/ha increments from V6 to V14 for sensor-based N
treatments based on leaf chlorophyll readings.

A Minolta SPAD-502 mobile chlorophyll meter (Spectrum Technologies Inc.)

was used to measure red (690 nm) and NIR (940 nm) canopy absorption and

19

transmittance in all plots. Leaf chlorophyll contents of 30 plants per plot were measured
using a SPAD chlorophyll meter from V6 to V14 stages of growth to evaluate changes in
sensor-based chlorophyll measurements. Nitrogen fertilizer application was triggered in
the stress-based N treatment when reflectance Of corn fell to 96% of the reflectance of
well-fertilized corn as this level has been Shown to maintain corn yield (Scheppers et al.,
1996).

The soil was chisel plowed in the fall of 1998 and was fit with a field cultivator
prior to planting in 1999 and in 2000. The site was mapped in the spring prior to planting
for elevation and soil conductivity using commercial mapping procedures. These data
were used in conjunction with soil profile measurements to establish the variability of
soils and landscape parameters on the experimental site. Pioneer 3730 corn variety was
planted on April 26, 1999 and April 29, 2000. The seeding rates were 64,250 seeds/ha
for non-irrigated corn and 86,500 seeds/ha for irrigated corn. Crop measurements
recorded included plant emergence, crop phenology, plant count and spacing, earleaf N at
silking, stalk N, grain N, grain moisture, and grain yield. Soil measurements included
water content at weekly intervals over the growing season using TDR, soil mineral N at
V6 and silking (O - 30 cm), and soil profile N after harvesting (0 - 90 cm). Corn grain
yields were determined by harvesting two rows from each plot 9.84 m (30 ft) by 1.64 In

(5 ft) for a total area of 16.1 m2 (150 ft"). Corn grain moisture was adjusted to 155 g/kg

moisture.

20

Statistical Analysis

Analysis of variance (ANOVA) and covariance were performed using the SAS
Mixed Procedure (SAS Institute, 2000). The Mixed Procedure in SAS uses a maximum
restricted likelihood estimation of linear statistical models involving both fixed and
random terms, such as the linear model underlying the split-plot analysis, and therefore is
well suited to this task. The Tukey-Kramer test was used for comparison between
different treatments at the 95% level of significance. Log transformations of the data
were performed when there were outliers and then back transformation was performed on
the original data.

RESULTS AND DISCUSSION

It is very clear from Figure 1.1 that evapotranspiration exceeded precipitation
during the growing season and that supplemental irrigation was necessary for consistent
high yield.

Chlorophyll meter readings were used to schedule sensor-based N treatment by
calculating a N sufficiency index (N SI). The Nitrogen sufficiency index is the ratio of
chlorophyl meter readings from different N treatments to the chlorophyl meter readings
from a reference N treatment which had sufficient N. In this study the reference N
treatment consisted of N applied at planting (145 kg/ha was applied for a yield goal of
7,800 kg/ha for non-irrigated corn and 202 kg/ha applied for a yield goal of 12,480 kg/ha
for irrigated corn). The at-planting N treatment was the reference N treatment and was

compared to different N treatments over the entire growing season. The nitrogen

21

sufficiency index for both 1999 and 2000 growing seasons is illustrated in Figure 1.2a
and Figure 1.2b. The control treatment showed a decline in leaf chlorophyll content as
the season progressed. Early season drought in 1999 resulted in lower leaf chlorophyll
content even after N was applied until irrigation or rainfall occurred in early July when
leaf chlorophyll recovered as shown in Figure 1.2a. Figure 1.2b Showed that leaf
chlorophyll for sensor-based and PSNT N treatments for the 2000 growing season
recovered immediately once the intervention management N was applied.

Interactive N management based upon the PSNT and leaf chlorophyll
measurements performed as well as at-planting N management for corn grain yield.
Irrigation increased corn grain yield but not without supplemental N application in 1999
(Figure 1.3a). There was a N x water interaction in the dry season of 1999 (P-value =
0.045). Figure 1.3a showed that under dryland farming conditions corn grain yield was
not increased whether N was applied or not. Irrigation significantly increased corn grain
yield over the control at all levels of N treatment. Sensor-based irrigation was not
scheduled because acquisition of infrared thermometer (IRt) data was not available in the
1999 season. There was no N x water interaction for the wet 2000 season (P-value =
0.1677). Comparison of least square means shows that com grain yield response to
irrigation was not significant in the 2000 growing season. Figure 1.3b compares corn
grain yield at different levels and strategies of N addition indicating that N is the main
factor determining yield in the wet 2000 growing season. However, there was no
significant difference in corn grain yield when N was applied at planting time as

compared to application based on PSNT or sensor-based measurements. Table 1.1 shows

22

the least square means for corn grain yields for the experiment in irrigated and non-
irrigated N treatments in 1999 and during the wet 2000 growing season without irrigation.
It is obvious fi'om this least squares means analysis that irrigation doubled corn yield in
1999, however, an excellent 2000 yield without irrigation exceeded that of an irrigated
1999 for the same N treatments. The 1999 irrigated corn grain yield was lower than the
2000 corn grain yield because irrigation was not possible until late July of the 1999
season. .

Table 1.2 shows total nitrogen fertilizer applied at different N treatments and least
squares means for PSNT levels at V6 leaf, at silking, and after harvesting for the 2000
growing season. It is obvious that the at-planting N treatment had the highest nitrate-N
levels at V6 on the first PSNT testing date (144.37 kg/ha), however, this high level
decreased to 36.76 kg/ha when the PSNT was performed at the second sampling date at
silking (7/23/2000). Therefore, much of the fertilizer applied at planting time was
probably leached down through the soil profile and not used for plant nutrition. A starter
N fertilizer was applied for PSNT N treatment and sensor-based N treatment at planting
time. Then N fertilizer was applied for PSNT N treatments according to the MSU
recommendation (N recommendation = -27 + 1.36" yield goal -6*ppm N in the 0 -30 cm
depth). The recommended rate was calculated per PSNT N treatrrrent at V6 and at silking
when the PSNT was performed. For sensor-based N treatments, there was a weekly
assessment of plant chlorophyll through chlorophyll meter readings that aided in the
calculation of the N reference index. When the N reference index was less than 96% of

the reference N treatment, N fertilizer was applied in 70 kg/ha increments from V6 to

23

silking. Table 1.2 also shows that PSNT and sensor-based N treatment have N available
for crop uptake at the rapid stage of growth at silking. Although the at-planting N
treatment has the highest amount of N fertilizer applied, by the time of rapid N plant
uptake there was significantly less soil NO3-N in the at-planting N treatment compared to
soil NO3-N in the intervention N treatments that were applied based upon PSNT or on the
sensor-based measurements. Therefore, intervention N management based upon PSNT
and chlorophyll measurement resulted in less N fertilizer addition and less N leaching
compared to at-planting N management. Table 1.3 illustrates the analysis of variance for
the PSNT nitrate-N levels for 1999 and 2000 at two different dates (V6 and silking).
Irrigation, irrigation x N, and irrigation x date were not significant in the PSNT N
treatment. However, N, date, and N x date were significant (p-value = 0.001). The
irrigation x N x date interaction was not significant in 1999 but was significant in the
2000 growing season. Although PSNT and sensor-based N treatments had smaller
amounts of N fertilizer applied, adequate NO3-N for crop nutrition was present at the two
different PSNT dates for 1999 and 2000 as illustrated in Figures 1.4a and 1.4b. These
results indicate that lower N addition based upon PSNT and sensor-based measurements
can result in optimal corn grain yields while leading to less nitrate leaching potential
when compared to the at-planting N treatment.

Table 1.4 shows the analysis of variance of KCl extract for soil profile NO3-N
after harvesting at three rooting depths for the two growing seasons. The rooting depth
was 91 cm divided into 3 measuring depths of 0 -30 cm, 30 - 61 cm, and 61 - 91 cm.

Figurel .5a illustrates that a dry 1999 irrigated season had a higher soil nitrate

24

accumulation compared to the wet season of 2000. Although irrigation was only
significant in the 2nd depth (30 - 61 cm), it is important to Show the impact of irrigation on
N leaching as summarized by least squares means (Table 1.5). Figure 1.5b shows the
effect of the wet 2000 season without irrigation on soil nitrate leaching using the different
N application strategies. Dryland farming has the highest accumulation of nitrate and a
very high amount of NO3-N was found in the first depth (0 - 30 cm) as compared to other
depths in the irrigated 1999 season. The sensor-based N treatment had the highest
residual accumulation of nitrate in the first depth in the 1999 irrigated season, probably
due to the late N application that makes this interactive N application more appealing in
reducing N potential for leaching to the ground water in theory. But, in a dry season like
1999, especially under dryland conditions and reduced growth and yield, the excess
nitrate applied later in the season may still move beyond the rooting zone to groundwater
after the crop is harvested. Therefore, sensor-based N application reduces nitrate leaching
beyond the rooting zone when there is a normal growing season but may not be effective
in dry years without irrigation. Nitrogen treatment based upon PSNT was the only N
treatment with Significant NO3-N levels at the 30 - 61 cm depth in 2000 growing season.
The NO3-N accumulation at 30 - 61 cm is likely due to the application of the
recommended dose that was not leached and yet was not taken up by the crop prior to
maturity and harvest. The pre-sidedress nitrate test (PSNT) N treatment provides
adequate N for crop production during a growing season but may be a source of nitrate

leaching during dry seasons with irrigation as is the case with sensor-based N application.

25

Soil NO3-N levels were not significantly different at the 61 - 91 cm depth regardless of N
application or application strategy used.

Table 1.6 summarizes the analysis of variance for plant nitrogen (earleaf, stalk N,
grain N, and corn grain yield). Irrigation was not significant in increasing earleaf N and
there was no irrigation x N interaction. Although N application was significant in
increasing earleaf N in 1999, it was not significant in 2000. Irrigation and N application
were very Significant when grain N was analyzed. While irrigation reduced grain N
concentration probably due to a growth dilution effect, N application increased grain N
concentration due to increased N uptake. There was an irrigation x N interaction in both
seasons. Recent studies by Blackmer and Mallorino (1996) at Iowa State University
(ISU) showed that N status of corn can be assessed by measuring nitrate concentration in
the lower portion of comstalks at the end of the season. Plants that have more N than
needed will accumulate nitrate in the lower stalk at the end of the season. Corn plants
that are deficient in N usually translocate N from lower comstalks and leaves to grains
during grain filling. While N fertilization has a Significant impact on lower stalk N build
up, irrigation reduced stalk N accumulation in the lower stalk and there was an irrigation
x N interaction. According to the ISU publication, stalk nitrate concentration can be
divided into four categories: low (<250 ppm), marginal (250 - 700 ppm), optimal (700 -
2000 ppm) and excess (>2000 ppm) NO3-N. Table 1.7 shows the stalk NO3-N levels with
different irrigation and N management strategies. Optimal levels were found only in the
at-planting N treatment. Sensor-based and PSNT N treatments represent a marginal

category according to the ISU test, however, these N management strategies produced

26

comparable corn grain yield to the at-planting N treatment. Care must be taken when
considering these categorizations because there are many factors affecting N uptake.
While the factors associated with time of application or irrigation affect N , generally,
under dry conditions, corn will have higher levels of nitrate in the lower comstalk than
irrigated corn (Table 1.7).

A water stress sensing method needed for sensor-based irrigation control was not
successfully developed in either season. Although we acquired infrared thermometer data
(IRt) in the 2000 season, there were no data differences due to the wet season.

The difficulty with current irrigation scheduling programs, such as J enson (1969),
is that they require the estimation of daily evapotranspiration (Etr) from a reference crop
and a defined crop coefficient (Kc). Wright’s crop coefficient (ch) has the difficulty of
needing to estimate the time interval from planting time to effective cover. The
advantages of the reflectance-based crop coefficient (Ker) over traditional crop
coefficients is its lack of dependence on a time based variable such as planting date to
effective cover (Bausch, 1993). In addition, the effective cover is reached when the
NDVI reaches its maximum (Neale et al., 1989). The previously introduced reflectance-
based crop coefficient (Ker) for corn estimated from NDVI has been shown to
overestimate the basal crop coefficient for corn by 24% (Bausch et al., 1993). We
investigated GNDVI to develop a reflectance-based crop coefficient (Ker) over two
growing seasons. We used the same procedure developed by Neale et al. (1989) and
substituted GNDVI for NDVI after calculating the multiplier and the offset from multi-

spectral reflectance. Table 1.8 shows a comparison between the multiplier and the offset

27

values (2000 growing season) for the three vegetation indices (GNDVI, NDVI, and
SAVI). From this analysis we developed the following equation for Ker from GNDVI for
corn in the temperate region:

Kcr = 1.256 * GNDVI - 0.1699 (1.6)

Figures 1.6a and 1.6b show the temporal progress of the crop coefficient estimated
from planting date and effective cover for ch (Wright) and the Ker calculated crop
coefficient from canopy reflectance for NDVI, SAVI, and GNDVI for irrigated and non-
irrigated N treatments for 1999. All three vegetation indices were similar in mimicking
ch (Wright) for the full growing season. Reflectance-based crop coefficients (Kcr) for
all vegetation indices, whether irrigated or non-irrigated, had not under- or overestimated
evapotranspiration for the 1999 growing season compared to ch (Wright) as illustrated
in Figure 1.6. Because irrigation was not significant in the analysis of variance in the
2000 growing season, Figure 1.7 only shows Ker (2000) data from NDVI, SAVI, and
GNDVI for N treatments without irrigation. Figure 1.7 also illustrates that NDVI
overestimated the corn crop coefficient compared to ch (Wright), and therefore
overestimated the amount of irrigation needed especially in times of higher demand for
water by the crop. The soil adjusted vegetation index (SAVI) crop coefficient follows
ch (Wright) and it does not overestimate irrigation water at the time of demand because
there is no soil background effect. Green normalized difference vegetation index
(GNDVI) crop coefficient, however, follows ch and SAVI curves because we used the
green reflectance in calculating Ker instead of the red reflectance. Green reflectance

represents the amount of chlorophyll present in the crop canopy and hence the activity

28

and transpiration of the whole leaf canopy. When regression analysis was performed
between ch (Wright) and Ker obtained from NDVI, SAVI, and GNDVI, there was a
very high correlation between ch (Wright) and the Kcr from NDVI, SAVI, and GNDVI,
as can be seen in Table 1.9. These correlations illustrate that all vegetation indices
studied can be used to develop an instantaneous crop coefficient from crop growth and

development independent of the planting date and effective cover.

CONCLUSION

Early season drought in 1999 resulted in lower leaf chlorophyll even after
intervention N application until irrigation or rainfall occurred in early July at which time
leaf chlorophyll recovered. Starter N application was necessary to avoid early season N
stresses and subsequent yield losses. Interactive N management through PSNT and
sensor-based strategies resulted in less N application than prescription at-planting N
treatment based on yield goals while producing comparable corn grain yields. There was
significant N x water interaction when irrigation water was applied, but these interactions
were absent when irrigation was not needed in the wet season of 2000. Chlorophyll
meter readings could be used to schedule N fertilizer by predicting N stress in corn.
Nonetheless, there are still unanswered questions of how, when, and what rate of N
fertilizer to apply for sensor-based N application. This study would have benefitted from
an analysis of total soil N prior to planting to more accurately determine how much starter

was needed for intervention N treatments at planting time to prevent N stress in the crop.

29

Vegetation indices can be used to develop an instantaneous crop coefficient from
crop growth and development data independent of the planting date and effective cover.
Each vegetation index has advantages and disadvantages. While NDVI is sensitive to soil
backgrounds interference, SAVI is more susceptible tO interference from Sky
illumination. Green normalized difference vegetation index (GNDVI) is sensitive to soil
background but is not affected by sky illumination. Green normalized difference
vegetation index (GNDVI) was transformed into a reflectance-based crop coefficient.

The crop coefficient from GNDVI performed well when compared to Ker from SAVI and
ch (Wright) and did not under- or overestimate evapotranspiration of corn. An equation
(Kcr = I. 256 x GND VI - 0.1699) was derived that represented all field environments for
estimating a corn crop coefficient. Reflectance-based crop coefficients can be used to
apply variable irrigation water based on water stress in corn rather than on the current
basis of demand or weather-based water budgets which do not account for Spatial
variability within a field. In addition, it is possible to remotely monitor water stress in
plants and apply irrigation water based on site-Specific needs of corn within a field rather

than to apply water uniformly over a field as is the current practice.

30

 

 

582
m < A A 2<2a A a 20

 

- cm
i ov
.. ow
r ow
oo—
- cg
o3

. o2
fiesta; I41 I
ABBEamam new

of

 

 

 

com

max a 88 - 22 was»
05 Se commafimqmboago 33:38 wee eoufifimoee owfiog 55:02 .2 23me

(mm) runourV uorrmdroard

 

 

31

Figure 1.2. Nitrogen sufficiency index (N SI) for all N treatments.
a) an irrigated 1999 season, and b) a non-irrigated 2000 season

 

a. Nitrogen sufficiency index for irrigated and non-irrigated N
treatments (1999)

     

i +Control non-irrigated (none)
1 +Ar-planting non-irrigated (I45 kg/ha))
40 .. +PSNT non-irrigated (142 kg/ha) - ~ f—
l +Sensor non-irrigated (134 kg/ha)
—-—Control Irrigated (none)

Nitrogen suffiency index (%)
8
l

 

 

    

20 i +At-plantinglrrigated (202 kg/ha) ‘ “T'T ”W‘ ' “ "‘
. +PSNTlrrigatcd(l77 kg/ha)
o 1,--_-!:‘Se'1§9rIflsmdtl34.Ks/ha)_-, _,
160 167 174 181 I88 195 208 223 238
Day of the Year

b. Nitrogen sufficiency index for all N treatments (2000)

,3 120 I

e\/ l

>< 100 + ,

d)

-o

.E I

>. 80 .

U

5

.g 60 .

a; ,. L H

m 40 9 +Control(nonc)

§D +At—planting (290 kg/ha) ,

g 20 +PSNT270(kg/ha) f' “ T T

E 0 +Sensor(134kg/ha)

168 175 182 189 194 202 229 238 250 260
Day of the Year

 

 

32

 

Figure 1.3. Effect of N treatments and irrigation on corn grain yield.
a) effects of irrigation and N treatments on corn grain yield in the
dry irrigated 1999 season, and b) effects of N treatment without
irrigation on corn grain yield for the wet 2000 growing season

, . If, _,,_ LL A, L J, , 7,,L_imwv,ifim 7‘

a. Interaction between irrigation and N treatments on corn ’

grain yield (1999) Q
l
14 , I
A INO Irrigation l
g 12 Ilrrigation '
2 10 l
g 8 1
.5: ‘
.E 6‘
l e i
i go 4. i
o
c.) 2 l
0 .

 

control at-planting sensor

i

N Treatments 1
W V l

l

b. Effect of N treatments on corn grain yield (2000)

   

Corn graln yreld (Mg/ha)

   

"7

control at-planting PSNT sensor

1
i
N Treatments 1
i

33

Figure 1.4.

Soil pre-sidedress nitrate test (PSNT) for N treatments at V6,
silking, and after harvesting for a) a dry irrigated 1999 season,
and b) a wet 2000 growing season without irrigation

 

 

 

 

 

  

 

 

 

 

 

  

 
   
     
 

 

 

 

 

 

 

 

a. PSNT NO3-N for N treatments at V6, silking and after
harvesting (1999)
120 ‘ I06/09/l999
100 007/22/1999
910/23/1999
80 .
i
'5, 60 .
Zia 40
0
Z
20
0 ,
control at-planting PSNT sensor
-20
N treatments
b. PSNT NO3-N for N treatments at V6, silking, and after
harvesting (2000)
180 ~
I06/l 1/2000
160 -
D07/23/2000
140‘ a 10/29/200
120 ~
g) 100 -
:5 80 -
Z,“ 60 .
g 40 a
20 - .
0 ‘ z I " r
-20 . co ol at-planting
-40
N treatments

 

34

 

Figure 1.5. Effects of N treatments on soil profile NO3-N distribution in the
effective root zone after harvesting for a) a dry irrigated 1999
growing season with values ranging from 3 to 58 kg/ka and
b) a wet 2000 growing season without irrigation with nitrate values
ranging from 4 to 20 kg/ha

 

a. Soil profile NO3-N for N treatments after harvesting
(1999)

I “lama—o1 ,
l I at-planting 1‘
‘IPSNT
Pianist

NO3 (kg/ha)

 

 

 

 

 

0-30 cm 30-61 cm 61 -91 cm
Depth (cm)

 

b. Soil profile NO3-N for N treatments after harvesting
(2000)

l I COEOI
i I at-planting
i I PSNT

‘ El sensorm

    

NO,-N (kg/ha)

 

 

 

 

35

 

ch (Wright), Ker

Figure 1.6. Basal crop coefficient (ch Wright) and crop coefficients (Ker)
from vegetation indices vs the day of the year for irrigated and
non-irrigated N treatments (1999)

a. Non-irrigated N treatments 1). Irrigated N treatments
COHUOI Control

 

1.2
1.0 r
0.8 -
0.6 -
0.4 -
0.2 -
0.0
1.0 -
0.8 -
0.6 i '
0.4

0.2 T

0.0 t
1,0 _ PSNT PSNT

0.8 -
0.6 ~
0.4 -
0.2 -

0.0
1.0 . Sensor Sensor

 

 

it’d-planting At-planting

 

 

  
  
  
 
 
 

0.8 -
0.6 -
0.4 -
0.2

 

    
  

+ ch
-v— NDVI
+ SAVI
-<>- GNDVI

0.0 I r | I I I r I r T I 'I T I I I

120 140 160 180 200 220 240 260 120 140 160 180 200 220 240 260 280

 

 

 

 

 

Day of the Year
36

ch (Wright), Kcr

Figure 1.7. Temporal performance of ch (Wright) and Kcr from vegetation
indices for all N treatments (2000) vs day of the year

 

Control

 
   

 

 

 

 

 

+ ch Wright
—0— Ker NDVI
+ Kcr SAVl

-<>— Kcr GNDVI

0.0 I l I l I l
140 160 180 200 220 240 260

 
   
     

 

 

 

 

 

 

Day of the Year

37

Table 1.1. Least square means for corn grain yield (Mg/ha) for two growing seasons

 

 

Treatment Control Pre-Plant PSNT Sensor
No Irrigation (1999) 4.8161 a 5.6867 b 6.1705 b 6.2041 b
Irrigated (1999) 5.8477 a 10.6141 b 10.2826 b 10.5010 b
N treatments (2000) 4.899a 12.037b 12.1551b 1 1.433b

 

Note: comparison and significance between N treatments at a given irrigation treatment.

Table 1.2. Nitrogen fertilizer applied and least square means of PSNT (N O3-N ) for
different N treatments at V6 growth stage, silking and after harvesting for 2000 season

 

 

 

2000
Total N Fertilizer PSNT (kg/ha) PSNT (kg/ha) PSNT (kg/ha)
Applied V6 Silking After Harvest

N Treatment kg/ha 6/19/2000 7/23/2000 10/29/2000
Control None 26.34 a 5.15 a 10.59 a
At planting 291 144.37 b 36.76 b 15.56 b
PSNT (MSU) 270 33.16 a 51.39 c 17.63 b
Sensor N 134 29.06 a 52.46 c 14.53 ab

 

Note: comparison and significance between N treatments at a given date. 202 kg of 291 kg/ha N was
applied at planting and 84 kg/ha was applied on 6/20/2000 . For PSNT treatment the 270 kg/ha N was
applied as 70 kg/ha (as starter N fertilizer for 2000 season only) and then after the PSNT was conducted at
V6 (6/19/2000) and applied on 6/20/2000 according to the MSU N fertilizer recommendation. Then N
fertilizer was not applied for PSNT treatment at silking afier PSNT was conducted on 7/23/2000. For
Sensor-based N treatment the 134 kg/ha was applied in 70 kg/ha increments. The first dose was applied at
planting then from V6 to silking according to the nitrogen sufficiency index (NSI). If NSl of the N treatment
was <96% of the reference treatment then N was applied.

38

Table 1.3. Analysis of variance table of KCl extract as soil NO3-N (kg/ha) for PSNT at
V6 leaf stage and at silking for 1999 and 2000 growing seasons

 

 

Source of Variation 1999 2000

Replication NS NS
Irrigation NS NS
N Treatment * * * * * *
Irrigation X N Treatment NS NS
Date * i: :1: a: at:
Irrigation X Date NS NS
N Treatment X Date * * * * * *
Irrig. x N Treatment x Date NS **

 

H, ***, **** significant at 0.05, 0.01, and 0.001 probability levels respectively

Table 1.4. Analysis of variance table of KCl extract for soil profile NO3-N after corn
harvest for 1999 and 2000 growing seasons

 

 

 

1999 2000
Source of Variation Depth Depth Depth Depth Depth Depth 3
1 2 3 l 2
Replication NS * * "‘ NS "‘ * NS * *
Irrigation NS **** NS NS NS NS
NTreatment am rum tum u **** ****
Irrigation x N NS NS NS NS NS NS

treatments

 

**, ***, **** significant at 0.05, 0.01, and 0.001 probability levels respectively

39

Table 1.5.

Least squares means for post-harvest soil profile NO3-N (kg/ha) for 1999 and
2000 growing seasons

 

 

 

 

 

1999
0-30cm 30-61cm 61-91cm
Non-irrig Irrigated Non-irrig Irrigated Non-irrig Irrigated
control 20.26 8.95 6.42 1.68 8.51 3.66
At 26.34 21.41 10.61 6.90 10.67 5.69
planting
PSNT 45.06 21.41 15.35 9.27 11.04 10.74
Sensor 44.67 56.39 14.85 10.82 10.10 11.38
2000
control 13.97 a 7.26 a 3.80 a
At 19.32 ab 11.49 ab 9.62 b
planting
PSNT 21.15b 15.47b 11.13b
Sensor 17.86 ab 9.55 ab 6.18 ab

 

Note significance at 0.05 probability levels across treatment at each depth. Irrig = irrigated.

40

 

 

Table 1 .6. Analysis of variance table for plant nitrogen for 1999 and 2000 growing
seasons
1999 2000
Source of Earleaf Grain Stalk Yield Earleaf Grain Stalk Yield
Variation
Replication NS NS NS NS NS NS NS NS
Irrigation NS ** ** **** NS NS NS NS
NTreatment **** **** ** **** NS NS **** ****
Irrigation x N NS NS NS **** NS NS ** NS

 

 

 

**, ***, **** significant at 0.05, 0.01, and 0.001 probability levels respectively

Table 1.7. Least squares means for plant nitrogen for irrigated and non-irrigated N

treatments for a dry irrigated 1999 and a relatively wet 2000 season

 

 

 

1999 2000
Source of Ear- Grain % Stalk (ppm NO3-N) Ear- Grain Stalk
leaf leaf
Variation % Non- Irrig Non-Irrig Irrig % % (ppm)
Irrig
Control 2.06a 1.30 a 1.09a 44a 223 2.96a 1.33a 23a
At- 3.01b 1.57 b 1.33b 1335b 839b 3.09a 1.35a 1849b
planting
PSNT 2.94b 1.53 b 1.2% 647C 416C 3.30a 1.39a 1282b
Sensor 2.90b 1.56 b 1.32b 722C 498C 2.98a 1.30 a 360a

 

 

Note: significance at 0.05 probability levels across treatment at each plant N parameter. Irrig = irrigated.

Nitrogen expressed as percentage for earleaf and grain and as ppm for stalk.

41

Table 1.8. Average and effective vegetation indices, multiplier (a) and offset (b) from
GNDVI, NDVI, and SAVI (2000)

 

Source of . Multiplier (a) Offset (b)

Variation Avg Effective Avg Stdev CV Avg Stdev CV
NDVI 0.7055 0.82288 1.167 0.0094 0.008 -0.1563 0.0014 -0.009
SAVI 0.7675 0.81869 1.173 0.0095 0.008 -0.1995 0.0019 -0.009
GNDVI 0.6445 0.76540 1.275 0.0132 0.010 -0.188 0.0023 -0.012

Avg is average, stdev is the standard deviation and CV is the coefficient of variation.

Table 1.9. Coefficient of determination (R2) and RMSE for ch and Ker for NDVI,
SAVI and GNDVI for N treatments (2000)

 

 

 

Source of NDVI SAVI GNDVI
Variation RMSE R2 RMSE R2 RMSE R2
Control 0.068 0.94 0.075 0.93 0.073 0.93
At-planting 0.081 0.91 0.098 0.87 0.077 0.92
PSNT 0.064 0.95 0.077 0.92 0.059 0.95
Sensor 0.064 0.95 0.077 0.92 0.060 0.95

 

42

REFERENCES

Bausch, W. C. 1993. Soil background effects on reflectance-based crop coefficients for
corn. Remote Sensing and the Environment 461213-22.

Bausch, W. C. 1995. Remote sensing of crop coefficient for improving irrigation
scheduling of corn. Agric. Water Management 27:55-68.

Bausch, W. C., H. R. Duke, and J. C. Iremonger. 1996. Assessment of plant nitrogen in
irrigated corn, p. 23 - 32. In P. C. Robert et al. (eds.) Proceedings of the Third
International Conference on Precision Agriculture. Minneapolis, MN, 23 - 26 June,
1996. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Blackmer, A. M. and A. P. Mallarino. 1996. Comstalk testing to evaluate nitrogen
management. In Iowa State University Extension Publication, PM 1584, August,
1 996.

Blackmer, A. M., J. S. Schepers, and G. E. Myer. 1995. Remote sensing to detect
nitrogen deficiency in corn. p. 505-512. In Proceedings of Site-specific Management
for Agric. Systems, Minneapolis, MN 23 - 26 June, 1996. ASA Misc. Publ. ASA-
CSSA- SSSA, Madison, WI.

Blackmer, A.M., J.S. Schepers, G.E. Varvel, and E. A. Walter-Shea. 1996. Nitrogen
deficiency detection using reflected shortwave radiation from irrigated corn canopies.
Agron. J. 88:1 - 5.

Blackmer, T. M., and S. E. White. 1996. Remote sensing to identify spatial patterns in
optimal rates of nitrogen fertilization. p. 33-41. In P.C. Robert et al. (ed.) Proceedings
of the Third International Conference on Precision Agriculture. Minneapolis, MN, 23
- 26 June, 1996. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Camp, C. R. and E. J. Sadler. 1994. Center pivot irrigatiOn system for site-specific water
and water management. ASAE Paper No. 94 - 1586. ASAE, St. Joseph, MI.

Crum, J. R., G. P. Robertson and F. Nurenberger. 1990. Long-term climate trends, and
agricultural productivity in southwestern Michigan. pp 53-58 In D. Greenland and L.
W. vaift (eds.) Climate variability and ecosystem response. US. Forest Service, US.
Dept. of Agriculture, Washington, DC.

Doerge, T. 2001. Variable rate N management for corn production: suCcess proves
elusive. Crop Insights 11(11):1 - 4. Pioneer Hi-bred International.

43

Evans, R. G., S. Han, MAW. Kroeger, and S. M. Schneider. 1996. Precision center pivot
irrigation for efficient use of water and nitrogen. p. 75-84. In P.C. Robert et al. (eds)
Proceedings of the Third International Conference on Precision Agriculture.
Minneapolis, MN, 23-26 June 1996. ASA Misc. Publ. ASA, CSSA, and SSSA,
Madison, WI.

Ferguson, R. B., C. A. Gotway, G. W. Hergert, and T. A. Peterson. 1996. Soil sampling
for site-specific nitrogen management p. 13 - 22. In P.C. Robert et al. (eds.)
Proceedings of the Third International Conference on Precision Agriculture.
Minneapolis, MN, 23-26 June 1996. ASA Misc. Publ. ASA, CSSA, and SSSA,
Madison, WI.

Howell, T. 1990. Relationships between crop production and transpiration,
evapotranspiration, and irrigation. p. 391-434. In B.A. Stewart and D. R. Nielsen (eds)

Irrigation Of Agricultural Crops. Agron. Monograph 30, ASA, CSSA, and SSSA,
Madison, WI.

Idso, S. B. 1982. Non-water-stressed baseline: a key to measuring and interpreting plant
water stress. Agric. Meteorology 27:59-70.

Jackson, R. D. 1982. Canopy temperature and crop water stress, Adv. Irrigation 1:43-85.

Jensen, M. E. 1968. Water consumption by agricultural plants. p. 1-22 In Kozlowki, T.
T. (ed.) Water Deficit and Plant Growth. Academic Press, New York. Vol. 2.

King, B. A., R. A. Brady, 1. R. McCann, and J. C. Stark. 1995. Variable rate water
application through sprinkler irrigation. p. 485-493. In P.C. Robert et al. (eds.)
Proceedings of the 2“d International Conference on Site Specific Management for

Agricultural Systems. Minneapolis, MN, 27-30 March 1994 ASA Misc. Publ. ASA,
CSSA, and SSSA, Madison, WI.

King, B. A., J. C. Stark, I. R. McCann, and D. T. Westermann. 1996. Spatially varied
nitrogen application through a center pivot irrigation system. p. 85-94. In P.C. Robert
et al. (eds.) Proceedings of the of the Third International Conference on Precision
Agriculture. Minneapolis, MN, 23-26 June 1996. ASA Misc. Publ. ASA, CSSA, and
SSSA, Madison, WI.

Kitchen, N. R., D. F. Hughes, K. A. Sudduth, and S. J. Birrel. 1995. Comparison of
variable rate to Single rate nitrogen fertilizer application: corn production and residual
soil NO3-N. p. 427 - 441. In: P.C. Robert et al. (eds) Proceedings of the 2nd
International Conference on Site Specific Management for Agricultural Systems.
Minneapolis, MN, 27-30 March 1994, ASA Misc. Publ. ASA, CSSA, and SSSA,
Madison, WI.

44

Lake, J. V., G. R. Beck, and J. A. Goode (ed.). 1997. Precision Agriculture: spatial and
temporal variability of environmental quality. John Wiley & Sons, New York.

Larson, W. E. and P. C. Robert. 1991. Farming by soil. p. 103 - 112. In: R. Lal and F. J.
Pierce (Eds). Soil management for sustainability. SWCS, Ankeny, Iowa.

McCarm, I. R., and J. C. Stark. 1993. Method and apparatus for variable application of
irrigation water and chemicals. US. Patent No. 5 246 164 Date issued: September 21,
1993.

Mokma, D. L. and J. A. Doolittle. 1993. Mapping some loamy Alfisols in southwestern
Michigan using ground-penetrating radar. Soil Survey Horizons 34:71-77.

Moran, M. S., Y. Inoue, and E. M. Barnes. 1997. Opportunities and limitations for

image-based remote sensing in precision crop management. Remote Sensing and the
Environ. 61 :319-346.

Moran, M. S., T. R. Clarke, Y. Inoue, and A. Vidal. 1994. Estimating crop water deficit
using relation between surface-air temperature and spectral vegetation. Remote
Sensing and the Environ. 49:246-253.

Neale, C. M. U, W. C. Bausch, and D. F. Heerrnann. 1989. Development of reflectance-
based crop coefficient for corn. Transactions, ASAE 32:1891-1899.

Pierce, F. J. and P. Nowak. 1999. Aspects of precision agriculture. Adv. Agronomy: 67.
Academic Press, 1999.

Pierce, F. J., D. D. Wamke, and M. W. Everett. 1995. Field and nutrient variability in
glacial soils of Michigan. p. 133-151. In P.C. Robert et al. (eds.) Proceedings of the 2nd
International Conference on Site Specific Management for Agricultural Systems.

Minneapolis, MN, 27-30 March 1994 ASA Misc. Publ. ASA, CSSA, and SSSA,
Madison, WI.

Robert, P. C., R. H. Rust, and W. E. Larson. 1993. Proceedings of soil specific crop
management: A workshop on research and development issues. SSSA Spec. Publ.
SSSA, Madison, WI.

Robertson, G. P., K. M. Klingensmith, M. J. Klug, E. A. Paul, J. R. Crurn and B. G. Ellis.
1997. Soil resource, microbial activity, and primary production across an agricultural

ecosystem. Ecological Application 7(1): 158-170.

SAS Institute 2000. SAS/STAT user’s guide. Version 8.1. SAS Inst., Cary, NC.

45

Schepers, J. S., D. D. Francis, M. Vigfl, and F. E. Below. 1992. Comparison of corn leaf
N concentration and chlorophyll meter readings. Comm. Soil Sci. and Plant Anal.
23:2173-2187.

Schepers, J. S., T. M. Blackmer, T. Shah, and N. Christensen. 1996. Remote sensing tools
for site-specific management. p. 315-319. In P.C. Robert et al. (eds) Proceedings of
the Third International Conference on Precision Agriculture. Minneapolis, MN, 23-26
June 1996. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Schueller, J. K. 1992. A review and integrating analysis of spatially-variable control of
crop production. Fertilizer Research 33:1 - 34.

Shayya, W. H., V. F. Bralts, and T. R Olmsted. 1990. General irrigation scheduling
package for micro-computers. Computer Electronics in Agric. 5:197 - 212. Elsevier
Sciences Publishers B. V., Amsterdam.

Stone, M. L., J. B. Solie, W. R. Raun, R. W. Whitney, S. L. Taylor, and J. D. Ringer.
1996. Use of spectral radiance for correcting in-season fertilizer nitrogen deficiencies

in winter wheat. Trans. ASAE 39:1623-1631.

Tsuji, G. Y., G. Uehara, and S. Balas (ed.). 1994. DSSAT v3. University of Hawaii,
Honolulu.

USDA-ERS. 1987. Farm-drainage in the United States: history, status, and prospects.
Misc Publ 1455. GA. Pavelis (ed.). USDA-ARS, Washington, DC.

Wesseling, J. 1974. Drainage and crop production. p.7-37 In: J. Van Schilfgaarde (ed.)
Drainage for Agriculture. Agron. Monograph17. ASA, Madison, WI.

Wright J. L. 1982. New evapotranspiration crop coefficients. Journal of Irrigation and
Drainage Div. ASCE 108(IR2):57-74.

46

CHAPTER II
REMOTE SENSING OF CORN CANOPY DYNAMICS AND BIOPHYSICAL
VARIABLES ESTIMATION IN MICHIGAN
ABSTRACT
This study was initiated to evaluate sensor-based nitrogen and water application

for corn (Zea mays L.) in Michigan. The specific objectives of this study were: 1) to
identify wavelengths that are more sensitive to N deficiency in corn, 2) to determine when
to predict corn grain yield from spectral remote sensing data, and 3) to estimate
biophysical variables of corn such as leaf area index (LAI) and fractional cover (Fe) from
spectral vegetation indices (SVI) obtained from radiometric measurements over corn
canopy. Nitrogen treatments included control, at-planting, pre-sidedress nitrate test
(PSNT), and sensor-based N treatment based upon chlorophyll meter readings. Corn was
either irrigated or non-irrigated. Sensitivity to N status centered around 560 and 810 nm.
These two wavelengths provided the best separation between different N treatments.
Corn grain yield was correlated with the NDVI (normalized difference vegetation index),
the GNDVI (green normalized difference vegetation index), and with chlorophyll meter
readings. The highest correlation between SVI and corn grain yield was achieved at the
R2, R3, R4, and R5 growth stages. There was a very high correlation between the
GNDVI and chlorophyll meter readings. These results suggested that GNDVI could be
used to substitute for chlorophyll meter readings in N scheduling. While NDVI and
GNDVI reasonably estimated the corn LAI, SAVI overestimated the LAI for all
N treatments. All spectral vegetation indices performed very well in estimating Fc over

the growing season.

47

INTRODUCTION

Optical remote sensing provides a powerful tool for monitoring changes in the
crop canopy over the growing season and can provide crop developmental information
that is time-critical for site-specific crop management. Radiometric measurements in the
solar spectral domain contain important information pertinent to the health of vegetation.
Remote sensing has been used to characterize properties of vegetation, to estimate yield,
to estimate total biomass, and to monitor plant health and plant stress (Jackson and Pinter,
1986)

From an energy balance viewpoint, all solar radiant flux incident upon an object is
either reflected, transmitted, or absorbed. Vegetation is unique in its three-segment
partitioning of solar irradiance. In the visible part of the spectrum (400 - 700 nm),
reflectance is very low, transmittance is nearly zero, and absorption is high. The
fundamental control of energy-matter interactions with vegetation in this part of the
spectrum is plant pigmentation. In this spectral range light reflectance increases with N
deficiency (Blackmer et al , 1994). In the longer wavelengths of the near-infrared portion
of the spectrum (700 - 1350 nm), both reflectance and transmittance are high whereas
absorption is very low. Sensor-based strategies rely on the fact that N deficiency in corn
reduces chlorophyll content of leaves thereby increasing the amount of light transmitted
through a leaf. Because chlorophyll content affects the amount of light absorbed or
reflected (Al-Abbas et al., 1974), leaf chlorophyll has been positively correlated with N
concentration. A typical reflectance curve shows a peak at around 550 nm corresponding

to green color due to reflection of light by chlorophyll and a dip in the plant spectral

48

signature around 690 nm corresponding to a red color due primarily to chlorophyll
absorption.

Many spectral vegetation indices (SVI) have been presented in the literature.
Jackson and Huete (1991) interpreted the utility of some of the vegetation indices.
Spectral vegetation indices (SVI) have been used to measure the photosynthetic capacity
of a plant canopy and to estimate yield (Wiegand et al., 1991). Spectral vegetation
indices reduce multiband observations into a single numerical index. The most widely
used indices are:

1) Ratio Vegetation Index (RVI) - reflectance in the near infrared (NIR) spectral

region divided by the reflectance in the red spectral region (red);

2) Green Ratio Vegetation Index (GRVI) - reflectance in the near infrared (N IR)

spectral region divided by the reflectance in the green Spectral region (green);

3) NDVI - normalized difference vegetation index calculated as (N IR-red)/(NRI +

red);

4) GNDVI - green normalized difference vegetation index calculated as (NIR-

green)/(NRI + green); and

5) SAVI - soil adjusted vegetation index - a modified NDVI using a constant (L)

value of approximately 0.5 to block interference from soil background
reflectance.

Biophysical variables that are of great interest and can be estimated from remote

sensing data are grain yield, the leaf area index (LAI), and fractional cover (Fe).

49

Remote sensing has been used for yield estimation where grain yield has been
correlated to a single measurement of NDVI. Tucker et al. (1980) used NDVI as a time
integral to estimate yield. Blackmer et al. (1996) correlated relative corn grain yield to a
reflective radiation from a photometric cell centered around 550 nm. They concluded that
this wavelength was also more sensitive than other wavelengths to N stress in corn.
Blackmer et al. (1996) analyzed aerial photos for N stress and found that black and white
photos centered around 536 nm also predicted corn grain yield at the R5 growth stage.
Therefore, identification of spectral wavelengths sensitive to N stress can aid in Site-
specific management of corn.

The leaf area index (LAI) is an important parameter in light interception by plants
and was used extensively as an input to predict crop response in many simulation models.
Remote sensing has been used to estimate biophysical variables such as LAI and Fe.
There are three approaches as reported by Q1 et al. (2000a) to estimate LAI. These
include a simple regression that uses multi-band regression, a modeling approach and a
vegetation index approach. The vegetation index approach is associated with SVI such as
NDVI. Here the LAI can be estimated using an empirical relationship.

The LAI-SVI relationship has the following forms:

LA]: aX3+bX2 +cX+d (2.1)
LAI = a + bX" (22)
LA] = l/2aln(l - X) (2.3)

where X is either vegetation indices or reflectance derived from remote sensing
data. Coefficients a, b, c and d vary with vegetation type. This equation can be applied to

50

remote sensing data to map spatial and temporal dynamics of vegetation (Qi et al.,
2000a). The approach is simple and easy to compute. However, this approach varies
with the form of the equation used (polynomial, power, or exponential), and the
coefficients which depend on the vegetation types. Radiative transfer modeling is an
alternative to the empirical approach. Here LAI is an input parameter to the model and
reflectance is the output of the model. There are many models in the remote sensing
literature (Suita, 1972, and Strahlar, 1994). The major advantage of modeling is that it is
independent of vegetation types. Carlson et al. (1997) used a simple transfer model to
illustrate that NDVI, LAI, and Fe are dependent on each other. Qi et al. (2000a) used an
empirical approach to estimate LAI and Fe from imagery acquired from Landsat TM,
SPOT VEGETATION, and aircraft sensors. Recently Qi et al. (2000b) used the Neural
fuzzy inference system to estimate LAI from remote sensed data. Fraction cover (Fe) is a
very important property of vegetation. It can be used on a daily basis when considering
variable rate application of agricultural inputs. Qi et al (2000a) used the object void
vegetation technique to estimate Fc from NDVI as follows:

Fc = (ND VI — NDVIW) / (ND VIM, - NDVIW) (2.4)

where NDVI is the NDVI at any time while NDVIso” is the NDVI value of bare
soil, and NDVIveg is the NDVI maximum of a pure pixel of vegetation.
Measurements of canopy multispectral reflectance can provide a quick, in-

expensive, and non-destructive real time assessment of the corn crop status on a whole

field basis.

51

Objectives

The objectives of this study were:

1) to identify spectral wavelengths that are more sensitive to corn N deficiency detection
and yield estimation, 2) to identify the optimum growth stage for corn grain yield
estimation, and 3) to estimate biophysical variables of corn such as LAT and Fe from

multi-spectral data acquired over a corn canopy during two growing seasons.

MATERIALS AND METHODS

Site Description

The study was conducted at the W. K. Kellogg Biological Station (KBS), which is
located in Southwest Michigan (85° 24' W longitude, 42° 24' N latitude) in the northern
portion of the Midwest corn belt. The W. K. Kellogg Biological Station is on the pitted
outwash plain of the morianic system left by the last retreat of the Wisconsin glaciation,
about 12000 years ago (Robertson et al., 1997). Soils in the area were developed on the
glacial outwash; soils at the site are Typic Hapludalfs, either fine-loamy, mixed, mesic
Kalamazoo series or coarse-loamy, mixed, mesic Oshtemo series (Mokma and Doolittle,
1993). Mean annual temperature at KBS is 9.4°C; precipitation is approximately 920 mm
annually and spread evenly all throughout the year, and potential evaporation exceeds
precipitation (30 years mean) for three growing months per year (Crum et al., 1990).
These soils respond to irrigation in most years and are vulnerable to leaching because of
the coarse textured soil materials and shallow depth of the groundwater. The

experimental area is gently sloping, decreasing in elevation from west to north and

52

decreasing in elevation from east to west. It is well known from previous studies that the
soils are variable over short distances making it possible to conduct Spatial variability
studies (Francis J. Pierce, Personal communication).
Experimental Design

The experiment evaluated two irrigations and four N treatments in a split-plot
experimental design. Irrigation was the main plot and N treatments were subplots with
four replications. The irrigation treatments evaluated were 1) none (control), and 2)
irrigation scheduled according to the Michigan irrigation scheduling program (Shayya et
al., 1990). The four N fertilizer treatments evaluated included none (control), N applied
at planting, (202 and 145 kg/ha for irrigated and non-irrigated corn respectively), N
application based on PSNT values plus yield at V6 leaf stage of corn, and 4) stress-based
N applied in 70 kg/ha increments using remotely sensed chlorophyll measurements.
Measurements

Water was applied to N treatments using drip irrigation system. Since N
management and soil parameters affect corn growth and development, each N plot within
the irrigation treatment was irrigated independently using a drip irrigation system devised
to apply water between each row. The irrigation scheme used in the 1999 season
consisted of applying 12.7 mm (0.5 inch) of water every 3 to 4 days; allowances were
made for rainfall. Irrigation was scheduled in 1999 with the MSU SCHED program when
50% of plant available water was depleted within the crop root zone. The crop was
exposed to water stress early in 1999 before irrigation lines were installed. A weather

station at KBS provided input for irrigation scheduling calculations. A rain gauge was

53

maintained at the site to ensure a site-specific rainfall record. Soil profile water content
was measured weekly from planting to maturity using Time Domain Reflectometry
(TDR). Time Domain Reflectometry access tubes were inserted in the middle of each
plot prior to planting to a depth of 90 cm. A tube access probe with a mobile moisture
meter (T rime Moisture Measuring System)TM was used to measure volumetric soil water
content at 18, 36, 53, 71, and 89 cm intervals on a weekly basis.

A multi-specnal ground based radiometer (CropScan, MSR87) was used to
measure the green, red, and NIR spectral reflectance bands. The CropScan radiometer
contains a set of eight narrow band filters, centered at:

460 nm, 510 nm , 560 nm, 610 nm, 660 nm, 710 nm, 760 nm, and 810 nm.

The CropScan radiometer has a 28 degree field of view for reflected irradiation.
The sensor was mounted on 2.62 m pole above the canopy at nadir viewing and a cherry
picker was used when the crop was very high. Two measurements per plot were taken in
1999 between 10:30 am and 1:30 pm local time for the following dates: (day of the year)
162, 167, 175, 177, 181, 188, 195, 217, 224, and 238. Three measurements per plot were
taken in 2000 on day of the year 161 , 168, 172, 179, 188, 202, 208, 217, 222, 229, 238,

245, and 250. The sensor was oriented parallel to the corn rows resulting in an oval

' ground area of about 1.53 m2. Weekly multi-spectral measurements were taken using the -
CropScan radiometer from the V6 growth stage to the R5 (dent) growth stage. Leaf area
was measured from V6 (2000 season only) to silking for each plot using the LAT-2000

plant canOpy analyzer (Li Cor 0‘) leaf area meter).

54

All fertilizer N was applied as 28% solution using a precise liquid fertilizer
applicator equipped with a capacity to apply N throughout the N uptake phase of corn
growth and development. Fertilizer N recommendations for corn following corn was
based on MSU fertilizer recommendations (N recommendation = -27 + 1.36* yield goal).
Pre-sidedress nitrate test (PSNT) fertilizer N rates were reduced based on the PSNT
values (6*ppm N in the surface 0-30 cm). Nitrogen applied at planting time was 145
kg/ha for a yield goal of 7,800 kg/ha for non-irrigated corn and 202 kg/ha was applied for
a yield goal of 12,480 kg/ha for irrigated corn. Nitrogen was applied in 70 kg/ha
increments from V6 to V14 for sensor-based N treatments based on leaf chlorophyll
readings.

A Minolta SPAD-502 mobile chlorophyll meter (Spectrum Technologies Inc.)
was used to measure red (690 nm) and NIR (940 nm) canopy absorbance and
transmittance in all plots. Leaf chlorophyll contents of 30 plants per plot were measured
using a SPAD chlorophyll meter from V6 to V14 stages of grth to evaluate sensor-
based chlorophyll readings treatment. Nitrogen fertilizer application was triggered in the
stress-based N treatment when reflectance of corn reached 96% of the reflectance of well-
fertilized corn as this level has been shown to maintain corn yield (Scheppers et al.,
1996).

The soil was chisel plowed in the fall of 1998 and was fit with a field cultivator
prior to planting in 1999 and in 2000. The site was mapped in the spring prior to planting
for elevation and soil conductivity using commercial mapping procedures. These data

were used in conjunction with soil profile measurements to establish the variability of

55

soils and landscape parameters on the experimental site. Pioneer 3730 corn variety was
planted on April 26, 1999 and April 29, 2000. The seeding rates were 64,250 seeds/ha
for non-irrigated corn and 86,500 seeds/ha for irrigated corn. Crop measurements
recorded included plant emergence, crop phenology, plant count and spacing, earleaf N at
silking, stalk N, grain N, grain moisture, and grain yield. Soil measurements included
water content at weekly intervals over the growing season using TDR, soil mineral N at
V6 and silking (0 - 30 cm), and soil profile N after harvesting (0 - 90 cm). Corn grain
yields were determined by harvesting two rows from each plot 9.84 m (30 ft) by 1.64 m
(5 ft) for a total area of 16.1 m2 (150 ftz). Corn grain moisture was adjusted to 155 g/kg.
Statistical Analysis

An analysis of variance (ANOVA) and covariance was performed using the SAS
Mixed Procedure (SAS Institute, 2000). The Mixed Procedure in SAS uses a maximum
restricted likelihood estimation of linear statistical models involving both fixed and
random terms, such as the linear model underlying the split-plot analysis and therefore
was well-suited to this task. Because there was a significant N effect and irrigation x N
interaction, regression analysis was performed on grain yield for each plot for all N
treatments at different wavelengths and against selected vegetation indices such as NDVI,
SAVI, RVI, GNDVI, and GRVI using the SAS PROC REG procedure (SAS Institute,
2000). Coefficients of determination and RMSE were reported in these results to
compare the results at different wavelengths and the performance of spectral vegetation

indices (SVI).

56

RESULTS AND DISCUSSION

Chlorophyll meter readings were used to schedule sensor-based N treatment by
calculating a N sufficiency index (N SI). Nitrogen sufficiency index is the ratio of
chlorophyl meter readings from different N treatments to the chlorophyl meter reading
from a reference N treatment that has sufficient N. In this study the reference N treatment
consisted of N applied at planting (145 kg/ha was applied for a yield goal of 7,800 kg/ha
for non-irrigated corn and 202 kg/ha applied for a yield goal of 12,480 kg/ha for irrigated
corn). The at-planting N treatment was the reference N treatment and compared to
different N treatments over the entire growing season. The nitrogen sufficiency index for
both 1999 and 2000 growing seasons is illustrated in Figure 2.1a and Figure 2.1b. The
control treatment Showed a decline in leaf chlorophyll content as the season progressed.
Early season drought in 1999 resulted in a lower leaf chlorophyll content even after N
was applied until irrigation or rainfall occurred in early July when leaf chlorophyll
recovered as shown in Figure 1.2a. Figure 1.2b showed that leaf chlorophyll for sensor-
based and PSNT N treatments for the 2000 growing season recovered immediately once
the intervention management N was applied.

Canopy reflectance values for two years (percentage reflectance) Showed
sensitivity to N treatments centered around 560 nm, 710 nm, and 810 nm (Figure 2.2).
The greatest separation between different N treatments occurred at wavelengths Of 560,
710 and 810nm. The sensitivity centered around 560 nm where N strongly reflects light.
When linear regression was performed on a single wavelength reflectance, it was clear

that 560, 610, 710, and 810 nm have a better correlation with grain yield (Table 2.1). The

57

correlation between a single wavelength and corn grain yield started very weak early in
the season and peaked on day of the year 224 (August 12, 1999) that coincided with the
growth stage of R5 (dent). Wavelength 560 nm (R2 = 0.90 and RMSE = 875) was
superior to all other wavelengths followed by 710 nm (R2 = 0.88 and RMSE = 957), then
810 nm (R2 = 0.78 and RMSE = 1298).

The relative reflectance as defined by Blackmer et al. (1996) is the ratio of
reflected radiation from different N treatments to reflected radiation from a reference
canopy treatment. The purpose of the relative reflectance is to eliminate non N-based
illumination differences between different N treatments. Relative reflectance showed
important information about wavelength sensitivity to N treatments at 560, 660, 710, and
810 nm at silking. However, after N was applied to PSNT and sensor-based N treatments
the difference was only apparent between the reference N treatment and the control
treatment and between irrigated and non irrigated treatments in 1999 (Figure 2.3).

Vegetation indices were extensively used in the past for yield prediction. The
most widely used indices were RV] and the NDVI. Both indices use NIR and red
reflectance values. Because corn N sensitivity centers around 560 um (which is the line
between the green and the yellow) GNDVI and GRVI were tested here for corn grain

yield prediction. GNDVI and GRVI are defined as follows:

GNDVI : (pNIR — pgreen) / (leR + pgrecn) (25)

GR VI : ION/R / pgrccn (2'6)

where p is the reflectance

58

Table 2.2 shows that there is a trend of low correlation between grain yield and
vegetation indices prior to tasseling and after grain dent time late in the season. A 50%
correlation was observed at silking. This increased at R2, R3 and reaches its maxima at
R5 (dent). After R5 there was no correlation between corn grain yield and SVI. There
was a significant difference between different SVI for corn grain yield prediction in the
2000 growing season. All SVI were very efficient in predicting corn grain yield (Figure
2.4a). GNDVI and GRVI were more superior in corn yield prediction than NDVI or RVI
for 1999 (Figure 2.4b). This is attributed to the sensitivity of the green reflectance at 560
nm due to chlorophyll. This study suggests that corn grain yield is best predicted at
reproductive stages of grth from milk to dent time. In addition, vegetation indices that
use green reflectance values such as GNDVI and GRVI are better suited for corn grain
yield prediction than vegetation indices that use red reflectance values. Improved
correlation in 2000 growing season was attributed to a very wet growing season that did
not require irrigation. Figure 2.4a shows a very high correlation between corn grain yield
and chlorophyll readings acquired using the SPAD meter. This high correlation was only
apparent at reproductive stages of growth. In addition, Figures 2.4a and 2.4b also
illustrate the high correlation between GNDVI and the chlorophyll meter readings. This
suggests that GNDVI or GRVI values could be used to replace chlorophyll meter readings
in N scheduling. Green normalized vegetation index (GNDVI) is relatively easy to
calculate and quick to acquire compared to a chlorophyll meter readings that require the

measurement Of 30 plant leaves.

59

Remote sensing can be used to estimate biophysical variable such as LAI and F c.
To examine the temporal dynamics of corn vegetation, LAI was measured for the 2000
growing season using a Li Cor 2000 leaf area meter. The data acquisition was not
possible in the 1999 growing season. Leaf area index (LAI) is a variable describing the
density of green vegetation and is defined as the total single-area of green leaves per unit
ground area. We modeled LAI from NDVI, SAVI, and GNDVI for different N treatments

(Figure 2.5a) using the equation developed by Qi et al. (2000) as follows:
GLAI = 18.99X3 - 15.24)(2 + 6.124X — 0.352 (2.7)
LAI = 18.99X3 —- 15.24X2 + 6.124X— 0.7 (2.8)

The equation was previously developed using data from grassland in the
Southwest US. To apply this equation for corn we changed the coefficient that is not
associated with the vegetation index (0.3 52) in the original equation to 0.7. When
validating the modeled LAI with the measured LAI the results Show that GNDVI is more
accurate in estimating LAI than and SAVI (Figure 2.5b). Both NDVI and GNDVI did not
predict LAI very well early in the season. However, they are consistent with published
corn LAI values (Westgate et al., 1997). Conversely, SAVI over-predicted LAI because
of soil background blockage inherent in the index. Green normalized vegetation index
(GNDVI) performed better than NDVI in modeling LAI as shown in Table 2.3. This is
due to the use of green reflectance values instead of red reflectance in developing the
GNDVI.

In order to assess the temporal dynamics of the corn canopy, F c was modeled

using the equation 2.4. All vegetation indices including NDVI, SAVI, and GNDVI are

60

very useful in modeling Fc. There was no significant difference between these indices in
modeling F c as shown in Figure 1.6. All indices predict Fc very reasonably except SAVI
which exceeds 100% estimation of fractional cover probably due to human or

experimental error.

CONCLUSION

Nitrogen deficiency in com was detected using a radiometer in the short
wavelength radiation in the range of 460 to 810 nm. A spectral region centered around
560 nm provided the best separation between different N treatments. These results
suggest that a future corn N sensor could be developed that uses a Simple combination of
limited spectral bands centered around 560 and 810 nm for variable N rate application.
Spectral vegetation indices such as GNDVI and GRVI performed better for corn grain
yield prediction than RV] and NDVI. This was due to sensitivity to N deficiency in the
green region of the spectrum that centers around 560 nm. Spectral vegetation indices
(SVI) such as NDVI and GNDVI performed very well in estimating biophysical variables
such LAI and F c. Extensive ground truth data were still needed to verify the validity of
these results. Finally, because remote sensing measurements in the temperate region are
hampered by a very high percentage of cloud cover, care must be taken when making

these measurements.

61

Table 2.1. Coefficients of determination (R2) and RMSE between corn grain yield and
wavelengths (510, 560, 610, 660, 710, 760, and 810 nm) at different dates

 

 

for 1999
Wavelength, nm Day of the Year 188 195 217 224 238
510 R2 0.42 0.32 0.51 0.68 0.005
RMSE 2089 2263 1923 1559 2730
560 R2 0.38 0.39 0.794 0.90 0.09
RMSE 2163 2141 1242 875 2615
610 R2 0.47 0.38 0.67 0.82 0.10
RMSE 1985 2153 1566 1164 2600
660 R2 0.45 0.34 0.56 0.68 0.16
RMSE 2023 2217 1822 1549 2503
710 R2 0.43 0.37 0.75 0.88 0.15
RMSE 2061 2167 1370 957 2527
760 R2 0.12 0.02 0.49 0.68 0.12
RMSE 2562 2709 1963 1557 2568
810 R2 0.20 0.0 0.63 0.78 0.12
RMSE 2445 2737 1670 1298 2561

 

62

Table 2.2.

Coefficient of determination (R2) and RMSE between corn grain yield and

vegetation indices at different dates and growth stages for 1999 and 2000.

 

Coefficient of determination R2 and RMSE (1999)

 

 

 

 

 

 

 

 

 

DOY NDVI RVI GNDVI GRVI Growth
R2 RMSE R2 RMSE R2 RMSE R2 RMSE Stage
162 0.10 2593 0.11 2589 0.07 2694 0.07 2709 V6 leaf
167 0.1 2601 0.1 2610 0.09 2613 0.09 2617 V7 leaf
175 0.49 1947 0.38 1956 0.34 2225 0.38 2236 V8 leaf
181 0.48 1934 0.48 1978 0.32 2258 0.48 2292 Tasseling
188 0.48 1966 0.49 1858 0.48 1973 0.50 1944 Tasseling
195 0.36 2195 0.40 2129 0.44 2046 0.45 2026 Silking
217 0.61 1719 0.74 1382 0.87 1002 0.90 853 R2 -R3
224 0.71 1487 0.82 1167 0.92 788 0.95 607 R5 -Dent
23 8 0.20 2456 0.18 2473 0.19 2468 0.16 251 l Maturity
Coefficient of determination R2 and RMSE (2000)
161 0.005 3258 0.002 3262 0.11 3074 0.11 3080 V6 leaf
168 0.12 3056 0.13 3055 0.12 3056 0.11 3088 V7-V8
172 0.27 2842 0.23 2878 0.29 2755 0.27 2783 V10
179 0.23 2861 0.23 2865 0.18 2960 0.18 2962 V12-V14
188 0.31 2601 0.27 2790 0.43 2357 0.37 2590 Tasseling
202 0.48 2345 0.48 2365 0.49 3232 0.49 2337 Tasseling
217 0.90 1002 0.86 1218 0.96 667 0.96 683 Silking
222 0.94 81 1 0.98 719 0.97 578 0.98 503 R2 -R3
229 0.91 992 0.96 760 0.96 561 0.97 554 R4
238 0.93 875 0.96 690 0.97 572 0.97 532 R5
245 0.70 1797 0.68 1840 0.74 1674 0.72 1743 R5
250 0.05 3182 0.04 3196 0.04 3203 0.03 3214 Maturity

 

 

 

 

 

 

63

Table 2.3. Coefficient of determination (R2) and RMSE between measured and
modeled LAI (from different vegetation indices) for different N treatments

 

 

 

(2000)
Vegetation Indices NDVI SAVI GNDVI
N Treatment R2 RMSE R2 RMSE R2 RMSE
Control 0.89 0.296 0.65 0.383 0.70 0.357
At-planting 0.94 0.244 0.90 0.316 0.95 0.211
PSNT 0.92 0.273 0.84 0.388 0.94 0.244
Sensor 0.92 0.286 0.88 0.363 0.94 0.246

 

64

Figure 2.1. Nitrogen sufficiency index (NSI) for all N treatments.
a) an irrigated 1999 season, and b) a non-irrigated 2000 season

 

a. Nitrogen sufficiency index for irrigated and non-irrigated N
treatments (1999)

120. , ’ - , 7 a - L A- A I
100 I
80 .

60-

    
 

1 +At-planting non-irrigated (145 kg/ha»
40 -. i +PSNT non-irrigated (I42 kg/ha)
‘ +Sensor non-irrigated (134 kg/ha)
( 1 Control Irrigated (none)
20 *1 I +At-planting Irrigated (202 kg/ha) ; W 'W , * W 'W W
‘ +PSNT Irrigated (177 kg/ha)
L—t Sensor Irrigated (Dim/Ila)

 

Nitrogen suffiency index (%)

 

 

 

0 74—? _ , , , ,, Y I, 7 7
160 167 174 181 188 195 208 223 238
Day of the Year
b. Nitrogen sufficiency index for all N treatments (2000)
120 - - i

 

40 ;-O—Control (none)
+At-planting (290 kg/ha)
20 ' ' +PSNT 27o (kg/ha) ‘ W

Nitrogen sufficiency index (%)
8

+ Sensor (134 kg/ha)

 

I68 175 182 189 194 202 229 238 250 260
Day of the Year

 

    

 

65

 

Percentage Reflectance

Figure 2.2. Percentage reflectance of N treatments at different wavelengths

50

40

30

50

40

30

20

10

with a) irrigation in 1999 and b) without irrigation in 2000

 

 

. . . a Percenta e reflectance on Au rust 5. 1999
Control no Irrigation ) g i: -

Control irrigated
At-planting no irrigation
At-planting irrigated
PSNT no irrigation
PSNT Irrigated

Sensor no irrigation
Sensor irrigated

   
 
 
 
 
 

1

iiitttii

 

 

 

Control a) Percentage reflectance on August 4, 2000

At-planting
PSNT
Sensor

  

tilt

 

 

l

 

 

 

I I I I

400 500 600 700 800

Wavelength, nm

66

Figure 2.3. Relative reflectance for N treatments at all wavelengths
with a) irrigation in 1999 and b) without irrigation in 2000

Relative Reflectance

1.94

1.74

1.54

1.34

1.14

0.94

0.74

l .04

1.02

1.00

0.98

0.96

0.94

 

 

+ Control no irrigation
+ Control irrigated

-I- At-planting no irrigation
« -6- At-planting irrigated
+ PSNT no irrigation
-e— PSNT irrigated

+ Sensor no irrigation

1 -V- Sensor irrigated

 
 
 
 

 

 

 

 

+ Control
r -'6'— Preplant
-I— PSNT
-9- Sensor

  
  
 

 

 

 

 

 

b) relative reflectance on August 5, 2000

 

 

400

I I I I

500 600 700 800 900

Wavelength, nm

67

Corn grain yield (Mg/ha)

Corn grain yield (Mg/ha)

Figure 2.4a. Corn grain yield for all N treatments vs vegetation indices

 

 

    

 

on August 12, 1999
14
NDVl . GNDVI
O
12 ~
R2 = 0.7] i R2 = 0.92
10 . RMSE = 1487 0 RMSE = 788
8 ~ 0 ‘
O
6 - .. .0 . .0
4 « . " ..
2 RVI - GRVI
12 * R2 = 0.82 :3. R2 = 0.95
RMSE = 1167 o RMSE = 607 :

 

 

 

 

 

 

 

 

 

 

 

 

3 4 5 6 7 8
14
12 R’=0.87 .3; ==o.923
RMSE = 990 RMSE = 0.0176
10 T
C.
8 '-
O
. O
6 0;. o ..
4 .3 '0'
2 I I I I I I I I l I I I
010 20 3o 40 50 6001020 3040 so 60 7o

Chlorophyll

68

‘ 0.775

0.725

' 0.675

‘ 0.625

‘ 0.575

GNDVI

Figure 2.4b. Corn grain yield for all N treatments vs vegetation indices

Corn Yield (Mg/ha)

Corn Yield (Mg/ha)

I4
12

ONAONOO

60.72 0.74 0.76 0.78 0.80 0.82 0.84 0.86 0.88
l

14

on August 9, 2000

 

R1 = 0.938
‘ RMSE = 81]

  

NDVl

  

R2 = 0.969
RMSE = 578

 

GNDVI

 

 

 

I I I I I I

I I

0.60 0.65 0.70 0.75 0.80 0.85

 

- R’ = 0.95
RMSE =7l9

 

 

  

RVl
R1 = 0.976
RMSE = 503

  

 

 

 

GRVl

  

 

 

 

_ R2 = 0.959
RMSE = 663

 

R’ = 0.982

 

RMSE = 0.0095

 

    

r 0.80

r 0.75

r 0.70

- 0.65

- 0.60

 

30

 

I l I I

40 50 6O 20 30
Chlorophyll (%)

69

40

50

60

 

0.55
70

GNDVI

Modeled LAI

Figure 2.5a. Modeled leaf area index (LAI) from vegetation indices using
empirical coefficients for N treatments (2000)

 

d Control

 

coo—NwAUiOxxl
1

 

At-planting

  

 

PSNT

 

 

Sensor

 

    
 

+ NDVI
-Vh-GNDVI
-%l—- SAA/I

I I I I I

140 160 180 200 220 240 260

 

 

 

 

 

Day of the Year

70

Modeled LAI

Figure 2.5b. Measured vs modeled LAI from vegetation indices
for at-planting N treatment (2000)

 

 

NDVI

   
  

2 = 0.939
RMSE = 0.244

 
 

NDVI LAI = - 0.96 + 2.35 measured LAI

 

 

SAVI

R2 = 0.90 O
RMSE = 0.316 0

 

 

 

    
   
  

 

 

GNDVI .
4 ‘ R2 = 0.954
RMSE = 0.211

3 1
2 _
1 .4
0 GNDVI L A1 = 0.263 + 1.273 measured LAI

0 1 2 3

Measured LAI

71

4

Fractional Cover (%)

Figure 2.6. Modeled fractional cover from vegetation indices for

120

100 -

80
60
40
20

120

100 -

80
60
40
20

120

100 -

80
60
40
20

120
100
80
60
40

 

 

 

 

 

 

 

 

 

 

 

N treatments (2000)
Control
.1
_ Sensor
_ —v— GNDVI
‘ + SAVI
140 160 180 200 220 240
Day of the Year

72

 

REFERENCES

Al-Abbas, A. H., R. Barr, J. D. Hall, F. L. Crane, and M.F. Baumgardner. 1974. Spectra
of normal and nutrient—deficient maize leaves. Agronomy J. 66:16-20.

Bausch, W. C ., H. R. Duke, and J. C. Iremonger. 1996. Assessment of plant nitrogen in
irrigated corn. In P. C. Robert et al. (ed.) Proceedings of the Third International
Conference on Precision Agriculture. Minneapolis, MN, 23 - 26 June, 1996.
ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Blackmer, A. M., J. S. Schepers, and G. E.Varvel. 1994. Light reflectance compared to
other nitrogen stress measurements in corn leaves. Agron. J. 86:934 - 938.

Blackmer, A. M., J. S. Schepers, and G. E. Myer. 1995. Remote sensing to detect
nitrogen deficiency in corn. p. 505 -512. In: Proceedings of site—specific
management for Agric. Systems 27 -30 March 1994, Minneapolis, MN, ASA
Misc. Publ. ASA- CSSA- SSSA, Madison, WI.

Blackmer, A. M., J. S. Schepers, G. E. Varvel, and E. A. Walter-Shea. 1996. Nitrogen
deficienCy detection using reflected shortwave radiation from irrigated corn
canopies. Agron. J. 88:1- 5.

Blackmer, A. M., and S. E. White. 1996. Remote sensing to identify spatial patterns in
optimal rates of nitrogen fertilization. p. 33 ~41. In: P.C. Robert, R. H. Rust, W.
E. Larson (Eds). Proceedings of the Third International Conference on Precision
Agriculture. Minneapolis, MN, 23 - 26 June, 1996. ASA Misc. Publ. ASA,
CSSA, and SSSA, Madison, WI.

Carlson, N. T. and D. A. Ripley. 1997. On the relationship between NDVI, fractional
vegetation cover, and leaf area index. Remote Sens. and the Environ. 62:241-252.

Crum, J. R., G. P. Robertson and F. Nurenberger. 1990. Long-term climate trends, and
agricultural productivity in southwestern Michigan. p 53-58 In: D. Greenland and
L. W. Swift (eds) Climate variability and ecosystem response. US. Forest
Service, US. Dept. of Agriculture, Washington, DC.

Jackson, R. D. and Alfredo R. Huete. 1991. Interpreting vegetation indices. Preventive
veterinary medicine ll(1991):185 - 2000.

Jackson, R. D. and Printer Jr., P. J. 1986. Spectral response of architecturally different
wheat canopies. Remote Sensing and the Environ. 20:43 - 56.

73

 

Mokma, D. L. and J. A. Doolittle. 1993. Mapping some loamy Alfisols in southwestern
Michigan using ground-penetrating radar. Soil Survey Horizons, 34:71-7 7.

Qi, J., Y. H. Kerr, M. S. Moran, M. Weltz, A. R. Huete, S. Sorooshian and R. Bryant.
2000a. Leaf area index estimates using remotely sensed data and BRDF models
in a semiarid region. Remote Sensing of the Environ. 73: 18-30 (2000). Elsevier
Science Inc. New York, New York, 1995.

Qi, J ., R. C. Marsett, M. S. Moran, D. C. Goodrich, P. Heilman, Y. H. Kerr, G. Dedieu,
A. Chehbouni, and X. X. Zhang. 2000b. Spatial and temporal dynamics of
vegetation in San Pedro River Basin area. Agric. And Forest Meteorology
105:55-68 (2000). Elsevier Science Inc. New York, New York, 2000.

Robertson, G. P., K. M. Klingensmith, M. J. Klug, E. A. Paul, J. R. Crum and B. G. Ellis.
1997. Soil resource, microbial activity, and primary production across an
agricultural ecosystem. Ecological Society of America. Ecological Application
7(1): 158170.

SAS Institute 2000. SAS/STAT user’s guide. Version 8.1. SAS Inst., Cary, NC.

Shayya, W. H., V. F .Bralts, and T. R Olmsted 1990. General irrigation scheduling
package for micro-computers. Computer electronics in Agric. 5:197 - 212.

Strahler, A. 1994. Measuring and modeling the bidirectional reflectance of plant
' canopies. In 6th international symposium on physical measurements and
signatures in remote sensing, Val d’Isere, France, 17 - 24 January 1994.

Suits, G. H. 1972. The calculation of the directional reflectance of a vegetation canopy.
Remote Sensing and the Environ. 2:117 - 125.

Tucker, C. T. 1979. Red and photographic infrared linear combination for monitoring
vegetation. Remote Sensing and the Environ. 8:127-150.

Westgate, M. E., F. Forvella, D. C. Reicosky, and J. Sonsen. 1997. Rapid canopy
closure for maize production in the northern US corn belt: Radiation-use
efficiency and grain yield. Field Cr0ps Research 49:249 - 258.

Wiegand, c. L. A. J. Richardson, D. E. Escobar, and A. H. Gerbermann. 1991.

Vegetation indices in crop assessments. Remote Sensing and the Environ. 35: 105
-l 19. Elsevier Science Inc. New York, New York, 1991.

74

CHAPTER III

ASSESSING THE SPATIAL VARIABILITY OF SELECTED SOIL RESOURCES

AND THEIR RELATIONSHIP TO CORN GRAIN YIELD IN MICHIGAN

ABSTRACT
Understanding the structure of soil variability is imperative for the precise

management of agricultural production systems. The objective of this study was to
evaluate the spatial variability of selected physical soil properties and landscape attributes
and their relationship to corn grain yield (Zea mays L). Soil properties, such as horizon
depth, bulk density, soil water content at - 10, - 33, and -1500 kPa, soil texture, as well as
elevation and slope were collected from a 224 m long transect in a 48 plot field in
Southwest Michigan. These properties exhibited a wide range of variability. Ap depth
and Btl depth had coefficients of variation (CV) of 27.5% and 26.4% respectively. Soil
texture in the Ap horizon was also highly variable. Variability of soil water content
surprisingly was similar in the plowed layer and Btl horizon. Slope had the highest
variation with a CV of 66%, while elevation exhibited the lowest variation with a CV of
0.3%. For all properties tested, variability exhibited strong spatial structure, and 50 to
100 percent of the sample variance was spatially dependent and autocorrelated over a
range of 5 to 224 m. Stepwise multiple regression analysis showed that Ap depth, Ap
silt, Btl bulk density, and Btl silt were significant in explaining corn yield variation.
Corn grain yield and these soil properties exhibited a very strong spatial cross correlation
and 50 -100% of the sample covariance was spatially dependent over a range of 26 to 212
m. Cumulative probability and Spearman rank correlation coefficient were used to

evaluate the temporal stability of the spatial soil water storage over the growing season.

75

INTRODUCTION

Precision Agriculture (PA) or site-specific crop management (SSCM) involves the
management of spatially variable factors according to localized conditions (Larson and
Robert, 1991 , and Pierce and Sadler, 1997). Pierce and Novak (1999) defined PA as the
application of technologies to manage the spatial and temporal variability within all
aspects of agricultural production for the purpose of improving crop performance and
environmental quality. There is an increasing interest in whether this variability implies
that there could be an advantage in managing field crops in a spatially variable way
instead of treating them according to average conditions (Robert, 1993). Precision
agriculture is concerned with variability in the two dimensions of space and time. Soil
properties and processes that regulate corn production and environmental sensitivity vary
in both space and time (Lake etal., 1997; Pierce and Nowak, 1999; Bell et al., 1995).
Therefore, understanding the relationship between crop yield and environmental spatial
variables is essential for SSCM.

Pierce and Nowak (1999) postulated that prospects for precision management
increase as the degree of spatial dependence increases, but the degree of difficulty in
achieving precision management increases with temporal variance. Without variability,
the concept of precision agriculture has no meaning according to Mulla and Schepers
(1997). They suggested that the most important factors in variability included spatial
patterns in pest infestations, plant available water, soil drainage, crop rooting depth,
nutrient availability, soil texture, organic matter content, and pH. Knowledge of soil

variability within a field is a required first step in the development of PA strategies.

76

Variability of soil properties results in spatially varying crop yields. A preliminary report
by Pierce and Warncke (1994) suggested that neither yield maps nor soil tests by
themselves were sufficient to explain the crop yield variability. They also reported that
there was little correlation between soil test levels and yield. Pierce et al. (1995)
concluded that the grid sampling method had little effect on estimating accuracy of
average field fertility. Corn yield was highly variable and generally not correlated to soil
fertility. They added that plant population and soil physical properties, particularly their
effect on water relations, were more important in explaining yield variability than
fertility. One challenge for PA management is to determine which major factors are
responsible for variations in crop yield. Beckett and Webster (1971) found that most
within-field variability was within an area of one hectare.

Although the emphasis in PA had been on nutrient management, the correlation
between soil nutrient levels and yield is not always strong (Pierce et al., 1995; Everett and
Pierce, 1996). Plant available water, soil physical and morphological properties, and
landscape attributes may be more important than soil fertility in explaining crop yield
variability. Soil microbial activity (Robertson et al., 1988; Robertson et al., 1993;
Robertson and Gross, 1994; and Robertson et al., 1997), soil moisture (Jaynes et al.,
1995), and soil and landscape attributes (Khakural et al., 1996), have an evident effect on
soil variability and hence on site-specific crop management.

Spatial data do not comply with the assumptions of classical statistics, in
particular independence. Tobler’s first law of geography (Tobler, 1970) states that

“everything is related to everything else, but near things are more related than distant

77

things”. Soil properties frequently exhibit spatial dependence, i.e., samples collected
close to one another are often more similar in value than widely spaced samples
(Trangmar et al., 1985). Mallants et al. (1996) found spatial structure for water retention
data in a multi-layered soil profile. There are many geostatistical methods available that
have been adopted for detecting spatial dependence of soil properties. The most
widespread is the semivariance analysis. Semivariance analysis provides a versatile and
unbiased means for examining autocorrelation in environmental data (Robertson and
Gross, 1994). This involves calculating the semivariance for all possible distance
intervals within an area yielding a semivariograrn or a variogram that describes the
dependence among samples as a function of direction and separation distance, then fitting
the appropriate model to the semivariograrn (Issaks and Srivastava,]989). A variogram is
basically a plot of dissimilarity (semi-variance) between samples against distance

between samples (Issaks and Srivastava, 1989). The omnidirectional semivariance is

defined as:
7(h) = [1 / 2mm]: [2,. — 2...]z (3.1)

where:
Y(h) = semivariance for interval distance class h;
z,- = measured sample value at point i;
zi+h = measured sample value at point i + h; and
N(h) = total number of samples for the lag interval h.

The semi-variance ideally increases with distance between sample locations, to a

more or less constant value (the sill or total semi-variance or the population variance) at a

given separation distance, called the range of spatial dependence (Trangmar et al., 1985).

78

Attributes separated by distances closer than the range are spatially related. Those
separated by distances greater than the range are not spatially related. The semi-variance,
when the distance of sample separation tends to zero, or y -intercept, is called the nugget
variance or nugget effect (Webster, 1985). The nugget effect gives an indication of
variability at scales less than the data spacing and/or due to analytical or random error.
The difference between the sill and the nugget is the structural variance. The proportion
of the total variance accounted for by the structural or spatially dependent variance
(Structural/(Structural + nugget)) is a useful index of the spatially dependent
predictability of the attribute. As an exploratory tool, the semivariogram documents any
spatial component to the variability and the strength of the component. The
semivariograrn also reveals the spatial scale over which autocorrelation occurs. Using
geostatistical procedures, Mallants et al. (1996) concluded that most hydraulic
parameters, including water retention, at different profile layers fit semivariograms that
could be described by means of spherical model functions with a spatial range from 4 to 7
m. Variograms for bulk density have ranges reported to vary from 18 to 56 m (Boyer et
al., 1996; and Chung et al., 1995) and can differ between horizons of a given profile, e.g.
35 m for Ap horizon and 51 m for the A1 horizon, as documented by Poier and Richter
(1992). Other soil physical characteristics reported to exhibit spatial dependence include
saturated hydraulic conductivity (range varied from 10.2 and 13.8 m) as reported by
Reinert (1990); sand content (range = 40 m), silt content (range = 78 m), clay content
(range = 43 m) as documented by Boyer et al. (1996); and organic carbon content (range
varied from 47 to 56 m) reported by Boyer et a1. (1996). The semivariograrn documents

79

 

whether there are spatial components to the variability and how distinct the patterns may
be. The semivariogram also reveals the spatial scale over which autocorrelation occurs.
Though variability may occur naturally as a result of complex geological and
pedological or manmade processes, knowing the degree of spatial dependence and its
extent is critical for evaluating its agricultural significance. Semivariance analysis can
quantify the degree of spatial dependence for an attribute and explicitly define the scale
over which the dependence is expressed. That makes the analysis more valuable for
inferences about the attributes and possible effects on associated plants (Robertson and
Gross, 1994). Application of geostatistical techniques may provide more power for
analyzing spatial variability and semivariance analysis gives a robust means for

quantifying autocorrelation in spatial and temporal dimensions.

Objectives

The objective of this study was to assess the spatial variability of selected soil
physical properties and landscape attributes and their relationship to corn grain yield and

additionally, evaluate the temporal stability of spatial soil water storage.

80

MATERIALS AND METHODS

The study was conducted at the W. K. Kellogg Biological Station (KBS), which is
located in Southwest Michigan (850 24' W longitude, 420 24' N latitude) in the northern
portion of the Midwest corn belt. The W. K. Kellogg Biological Station is on the pitted
outwash plain of the morianic system left by the last retreat of the Wisconsin glaciation,
about 12000 years ago (Robertson et al., 1997). Soils in the area were developed on the
glacial outwash; soils at the site are Typic Hapludalfs, either fme-loamy, mixed, mesic
Kalamazoo series or coarse-loamy, mixed, mesic Oshtemo series (Mokma and Doolittle,
1993). Mean annual temperature at KBS is 9.4°C; precipitation is approximately 920 mm
annually and spread evenly throughout the year, and potential evaporation exceeds
precipitation (30 years mean) for three months per year (Crum et al., 1990). These soils
respond to irrigation in most years and are vulnerable to leaching because of the coarse
textured soil materials and shallow depth of the groundwater. The experimental area is
gently sloping, decreasing in elevation from west to north and decreasing from east to
west. It is well known from previous studies that the soils are variable over short
distances making it possible to conduct spatial variability studies (Francis J. Pierce,
Personal communication).
Measurements

The experiment evaluated two irrigation and four N treatments in a split-plot
experimental design. Water was applied to N treatments using drip irrigation. Since N
management and soil parameters affect corn growth and development, each N plot within
the irrigation treatment was irrigated independently using a drip irrigation system devised

81

to apply water between each row. The irrigation scheme used in the 1999 season
consisted of applying 12.7 mm (0.5 inch) of water every 3 to 4 days; allowances were
made for rainfall. Irrigation was scheduled in 1999 with the MSU SCHED program when
50% of plant available water was depleted from the crop root zone. The crop was
exposed to water stress early in 1999 before irrigation lines were installed. A weather
station at KBS provided input for irrigation scheduling calculations. A rain gauge was
maintained at the site to ensure a site-specific rainfall record. Soil profile water content
was measured weekly from planting to maturity using Time Domain Reflectometry
(TDR). Time Domain Reflectometry (TDR) access tubes were inserted in the middle of
each plot prior to planting to a depth of 90 cm. A tube access probe with a mobile
moisture meter (Trime moisture measuring system)TM was used to measure volumetric
soil water content at 18, 36, 53, 71 and 89 cm intervals on a weekly basis.

In 2000, intact soil profile cores were obtained from the center of each plot
immediately after harvesting. Plastic sleeve tubes (4.127 cm internal diameter and 91 cm
depth) were inserted into the hydraulic truck probe from the surface to a depth of 91 cm.
The samples were taken away from the wheel track and 20 cm away from the corn row.
The sampled tubes were capped at both ends, transported to the laboratory and stored at
4°C. The soil samples were classified according to different horizons and each horizon
thickness was then recorded. Duplicates of undisturbed small sections (5 cm depth) were
cut from each soil horizon and water retention was measured at —10, -33 and -1500 kPa
(Klute, 1986). The horizons were identified as Ap, Btl, Bt2, and 2bt2 (C). The section
was then oven dried at 104 °C and used for bulk density determination (Blake and Hartge,

82

1986). The remainder of the samples were carefully ground and used to determine
particle size by the standard hydrometer method (Gee and Bauder, 1986). Water retention
for Ap and Btl horizons was determined at -10, -33 and -1500 kPa (Klute, 1986) on
weight basis, then converted to volumetric water content by multiplying by the bulk
density. Plant available water was computed as the difference between water content at
field capacity (water content at - 10 or - 33 kPa) and water content at - 1500 kPa.
Available water storage capacity (AWSC) for each horizon was computed by multiplying
the depth of the horizon by plant available water. Soil properties for each horizon are
recorded in the appendix, Tables 1- 4.

The experimental plots were chisel plowed in the fall of 1998 and fit with a field
cultivator prior to planting in 1999 and in 2000. The site was mapped in the spring prior
to planting for elevation using as global positioning system (GPS) and soil electrical
conductivity using electromagnetic induction (EM3 8). These data were used in
conjunction with soil profile measurements to establish the variability of soil and
landscape parameters on the experimental site. Pioneer 3730 corn variety was planted on
April 26, 1999 and April 29, 2000. Corn grain yields were determined by harvesting two
rows from each plot 9.84 m (30 ft) by 1.64 m (5 ft) for a total area of 16.1 m2 (150 ftz).
Corn grain moisture was adjusted to 155 g/kg moisture.

Statistical Analysis

Stepwise multiple linear regression was performed on corn grain yield for selected
soil physical properties to determine what factors affect yield the most using the SAS
Proc Reg Procedure (SAS Institute, 2000). Spatial statistics (S-Plus) software was used

83

_"
'11s.

to develop the scatter plots for auto- and cross-correlation plots and calculate the slope for
the spatial data. Geostatistical analysis (Robertson and Gross, 1994) was performed
using GS + (Gamma Design 5.1, 2002). Gamma design software (GS +) was used to
develop semivariograrn and cross variograms for auto-and cross-correlation plots. The
proportion of the sample variance (the variogram sill or C0 + C ) explained by the
structural variance C (C/C0 +C) was used as a measure of spatial dependence for all
variables. If this proportion (C/Co + C ) approaches 1, spatial dependence is high: a large
proportion of a sample variance, $2, is spatially dependent (Robertson et al., 1997). When
this proportion approaches zero, apparent spatial dependence is low indicating that
measurement error is high or spatial dependence occurs primarily at scales smaller than
the average distance of the first lag separation distance. The range is the separation
distance over which sample locations are autocorrelated, i.e. over which there is spatial
dependence among sample locations. For some models there is a difference between the
range parameter A0 and the effective range. The range parameter is the value Ao that is
used in the formula that defines the best-fit line. The effective range is the separation
distance at which the spatial dependence is apparent. For spherical and linear-to-sill
models the effective range is defined as A0. For exponential models the effective range is
defined as 3 * A0, or the distance at which the sill is within 5% of the asymptote. For
gaussian models the effective range is (30'5)* A0. Reduced sum of squares (RSS) were
used here as a criterion to choose the model that best fit the data.

To characterize soil water content and soil water storage for precise variable rate

irrigation application, the temporal stability of spatial soil water content variation has to

84

 

C2
In
to:

C02

1111

cm
c01
plt
M,-

001'

be determined. Vachaud et al. (1985) used cumulative probability function and the
Spearman correlation coefficient with mean relative difference and the standard deviation
of the time mean technique to evaluate the temporal stability of spatial soil water storage.
Cassel et al. (2000) applied the Spearman correlation coefficient technique to evaluate the
temporal stability of soil spatial water content. The method uses the Spearman rank
correlation coefficient as a criteria to evaluate the temporal stability of spatial soil water

content. The relative difference in soil water content is defined by:

5,}. = (6,]. - ism/0, (3.2)

where:
j is time of the measurement
60‘ is soil water content measurement at location i
3]. is the average soil water content

RESULTS AND DISCUSSION

Figure 3. la and 3.1b illustrate sample locations for the 48 plots overlain over the
contour lines and the elevation of the study site. Table 3.1 summarized the statistics of
corn grain yield and selected physical soil properties and landscape attributes for the 48
plots across the transect of the study area. Corn grain yield ranged from 3.504 to 13.565
Mg/ha due mainly to N treatments effects. Ap depth ranged from 20 to 54 cm with a
coefficient of variation (CV) of 27.5 percent and a mean of 30.75 cm above the mean of
Ap horizon depth (25 cm) for a typical sandy loam soil (Cassel et al., 2000). Water
content ranged from 0.144 to 0.262, from 0.128 to 0.247, and from 0.027 to 0.081 m3/m3

for 10, - 33, and - 1500 kPa with CV of 14, 15, and 19.5% respectively. Plant available

85

 

 

water for Ap horizon ranged from 0.117 to 0.231 m3/ m3 for - 10 kPa and from 0.086 to
0.214 m3/m3 for - 33 kPa. Available water storage capacity (AWSC) ranged from 3.120
to 11.628 cm for - 10 kPa and from 2.559 to 10.514 cm for - 33 kPa with CV of 34 and
37 percent respectively. The Ap horizon bulk density ranged from 1.11 to 1.47 g/cm3
with very little variation along the transect. The Btl bulk density ranged from 1.18 to
1.87 g/cm3 with a CV of 14% as shown in Table 3.1. The Ap soil texture varied widely.
Sand content ranged from 24.6% to 76.35% sand with a CV of 23.8%, and a mean of
46.37%, silt content ranged from 10.05 to 65.02% with a CV of 26.7% and a mean of
39%, and clay content ranged from 7.7 to 24.3% with a CV of 23.3% and a mean of
14.61%. Slope ranged from - 3.4% to 0 with a mean of 1.6% and exhibited the highest
variation among all variables measured in this study with a CV of 66.1%. The elevation
ranged from 283.45 to 285.44 m with an average of 284.4 m and exhibited the minium
variation among all variables measured with a CV of 0.3%.

Variograms revealed that all studied soil properties showed spatial dependence
although to different degrees. Ap thickness, sand, silt and clay contents, bulk density,
elevation, and slope all exhibited strong auto-correlation within the range from 5 to 224
m with spatial structural variance accounting for more than 95% of the total model
variance with the exception of clay content (53%) and bulk density (66%) as summarized
in Table 3.2. Ap texture was well structured except for clay which had the lowest spatial
structure as illustrated in Figure 3.2. Because N treatments had a significant effect on
corn grain yield, corn grain yield residuals were used in the geostatistical analysis and

showed a spatial dependence over a range of 31.5 m. The spatial dependence of the

86

Am
..
4

112

11'

210
110
111'
to

C0

3111

501'

DUI}

elevation and slope exceeded the range of 224 m we examined (Table 3.2). Ap depth,
soil texture, bulk density, and corn grain yield all had a spatial range of less than 75 m
proving that these soil are highly variable and changed over short distances. Therefore,
when making precision crop management decision one must take into account the size of
the research plots or management zones.

Variograms of Ap soil water at different pressures showed that the auto-
correlation of soil water content and available soil water storage capacity had a very
narrow range (Figure3.3). A 50 m range was used for comparative reasons; however,
when soil water was autocorrelated over the entire range the correlation was very weak
(Table 3.3). Ap soil water content was autocorrelated over a range of 19 and 18 m for
- 10 and - 33 kPa pressure respectively, and available soil water storage capacity
autocorrelated over a range of 32 and 30 n1 for - 10 and - 33 kPa respectively. There was
no significant difference between the range of autocorrelation of soil water content at
different pressures whether we used 50 m range or the entire range. Nonetheless, the
correlation coefficient indicated by the R2 is very low for the entire range for soil water
content at both pressures. When the entire range was used for available water storage
capacity, it was autocorrelated over a range of 203 and 197.6 m for -10 and - 33 kPa
respectively. In addition, the coefficient of determination was very high for both the
entire range and the 50 m range. These results illustrate that soil water content in these
soils were autocorrelated only over very short distances. Therefore, a management
decision has to take into account this variability in managing these soils for research
purposes and precision water management decisions.

87

 

(IO

I:J

In order to study the spatial relationship between corn grain yield and soil
properties, the factors that affect yield the most need to be identified. The stepwise
multiple regression analysis was used to identify the most important soil properties
affecting the variation observed on crop yield among the 48 studied plots. Stepwise
model selection was applied to yields (after adjusting the effects of nitrogen treatment
using a dummy variable). Among all available soil measurements, only those with a

variance inflation factor below 10 were used as independent variables. This was carried

out to minimize multi-colinearity among soil properties (Freund and Littell, 2000). For
the Ap horizon, properties that satisfied this condition were thickness, bulk density,
available water storage capacity at — lO kPa(AWSC10), silt content, slope, and elevation.
All of them were able to explain collectively only 29% of the variation observed on yield
after adjusting for the effect of nitrogen treatments. The results for the stepwise model
selection were presented in Table 3.4. The only two factors accounting for significant
variation were Ap depth and Ap silt. Ap depth accounted for the most variation; however
it explained only 20% of the total variation of corn grain yield.

For the Btl horizon, the soil properties used in the stepwise model selection were
depth, bulk density, available water storage capacity at 10 kPa(AWSC10), sand content,
silt content, slope, and elevation (after excluding properties with variance inflation factor
greater than 10). These soil properties collectively explained 38% of the variation on
corn grain yield. Two factors accounted for significant variation of 37%: Btl bulk
density and Btl silt (Table 3.4). The bulk density accounted for the most variation at
24% and Btl silt explained the additional 13%.

88

The spatial relationship between these significant variables and corn grain yield
over a 100 m range was illustrated in Figure 3.4. Although elevation was not significant
in the stepwise regression analysis, it is shown here in Figure 3.4 to illustrate the spatial
relationship between corn grain yield and elevation. All these properties expressed strong
spatial relationship with corn grain yield (Table 3.5). The structural cross correlation
variance explained more than 88% of the sample variance. One exception was Btl bulk
density for which the cross correlation variance explained only 65% of the sample
variance.

In these glaciated, rolling soils, elevation and slope are the major factors
determining many soil properties such as soil water and soil texture (Robertson et al.,
1997). Figure 3.5 shows the spatial relationship between elevation and Ap soil water at
different pressures, Ap thickness, and Ap silt content. With the exception of Ap soil
water content, all properties showed a nested periodic pattern and therefore, fit to a
gaussian model as shown by the model parameters in Table 3.6. All the properties
showed strong cross correlation with elevation, because the structural cross correlation
variance explained more than 99% of the cross sample variance. The only difference was
the range over which this cross correlation occurred. The smallest distance was water
content that occurred over a range of 122 m and Ap silt that ranges over 150 m.
Available water had the highest range of 711 and 570 for -10 and - 33 kPa respectively.

The spatial cross correlations between slope and Ap soil water, Ap thickness and
Ap clay showed similar patterns of elevation for these properties (Figure 3.6) with the

exception that these properties were negatively correlated to slope. There was a strong

89

spatial cross correlation where the cross structural variance explained almost all the cross
sample variance (Table 3.7). Again, with the exception of the Ap soil water content, all
the properties showed a nested periodic pattern that was fitted to a Gaussian model.

In order to characterize spatial soil water content and storage, and reduce the
number of measurements or samples for site-specific crop management, the temporal
stability of spatial water content variation was evaluated as described by Vauchad (1985).
Volumetric soil water content was measured on a weekly basis using TDR access tubes
for five intervals segments to 90 cm depth. The five depth intervals were 18, 36, 53, 71,
and 89 cm for 17 measuring dates throughout 2000 growing season. Figure 3.7 illustrates
the dynamics at three measuring dates (at the beginning of the season, silking, and at
physiological maturity) for the five depth intervals. It was clear the lowest measurements
occurred on July 22“d because of the plant root actions and the high demand for water by
the crop. To characterize the temporal stability of spatial water content and spatial water
storage, cumulative probability function and Spearman rank correlation coefficient were
used. Spearman rank correlation coefficient used in conjunction with measurements
mean relative difference and the time standard deviation.

The data for soil water storage was ranked from smallest to largest and the
cumulative probability function was identified as normal. The 50% probability was
selected as field average storage. In addition, one standard deviation, 0 ,and 20 could be
selected for extreme locations. This method depends on the probability of a certain
location to maintain its rank in the cumulative probability function for different
measuring dates as illustrated in Figure 3.8. The methods were defined by Vachuad et al.

90

1‘” .fl

(1985) as time stability and they suggested that the wet locations would remain wet and
the dry locations would remain dry because of soil texture. Spearman rank correlation
coefficient, the mean relative difference and the standard deviation were also used to
characterize the temporal stability of spatial soil water storage (Vachuad et al., 1985).
Spearman rank correlation coefficient, the mean relative difference and the standard
deviation were computed for soil water content and soil water storage measurements
using SAS software (SAS institute 2000). The criteria used for significance of the rank is
that, “The smaller the standard deviation the more significant is the rank”. The location
that their respective rank was relatively stable could be used as a management zone for
homogeneous water application and no more soil sampling would be needed. Locations
with unstable rankings need to be sampled frequently and treated differently when using
variable rate irrigation. Figure 3.9 presents the temporal stability of soil water storage
using the Spearman correlation coefficient, the mean relative difference, and the standard
deviation. Figure 3.10 shows a stable temporal stability in soil water content measured
by the mean relative difference in the upper depths compared to the lower depths due to
the actions of plant roots. Our results showed that the rank was different for the same
plot at different depths because of soil textural differences. The mean relative difference
increased with depth because of the stability of the parent material in deeper layers as

reported by Cassel et a1. (2000)

CONCLUSION
Soil properties such as depth of the horizon, bulk density, soil water content

91

r

'I._

at - 10, - 33, and -1500 kPa, soil texture, elevation, and slope from a 224 m long transect
exhibited a wide variability. For all properties tested, variability exhibited strong spatial
structure, and 50 to 100 percent of the sample variance was spatially dependent and
autocorrelated over a range of 5 to 224 m. Stepwise multiple regression analysis was
applied on yields to identify the most important soil properties affecting the variation. Ap
depth and Ap silt were significant in explaining yield variation in Ap horizon. Btl bulk
density and Btl silt were significant in explaining corn yield variation on Btl horizon.
Surprisingly, soil water content was not significant in the stepwise multiple regression to
identify the most important variable that affected corn grain yield. That may be due to the
fact that samples for this study were taken in a relatively wet year (2000) which did not
require irrigation. Corn grain yield and those soil properties in the Ap and Btl horizon
exhibited a very strong spatial cross correlation. Fifty-100% of the sample covariance
was spatially dependent over a range of 26 to 212 m. Spearman rank correlation
coefficient of the soil water content showed a relatively temporal stability in the upper
depth of soil and a relative temporal instability in deeper depths. It is evident from these
results that the soil resources tested were highly variable and very heterogeneous. In
addition, these soil properties were spatially structured. The degree and scale of spatial
dependence found in this study was similar to a previous experiment at KBS by
Robertson et al. (1988). This structure should be considered when designing an
experiment and making management decisions. Geostatistical analysis provided a
substantial and a robust technique for determining the scale over which spatial
dependence occurs.

92

 

 

Cumulative probability function, Spearman rank correlation coefficient, the mean
relative difference, and the standard deviation values were used to characterize the
temporal stability of the spatial water content and the spatial soil water storage. The
smaller the standard deviation, the more significant the rank. The locations at which their
respective ranks are relatively stable can be used as a management zone for homogeneous
water application, and no more soil sampling would be needed except low, medium and
high rank locations. There was a stable temporal stability in soil water content in the

upper depths compared to the lower depths due to the actions of plant roots.

93

 

 

 

 

 

 

 

 

 

 

Table 3.1. Statistics for corn grain yield and selected physical soil properties and
landscape attributes for 48 sample locations along the main transect. SD is
the standard deviation and CV is the coefficient of variation

Variable 1 mean (SD) 1 Range CV

Number of samplesw _4L , - LL ,,-_f18 _

Min Max %

Corn grain yield (Mg/ha) 10.10 (3.23) 3.504 13.565 32

EM38 13.94 (0.62) 12.80 15.20 4.5

Slope % 0.016 (0.01) -0.004 0.034 66.1

Elevation (m) 284.40 (0.78) 283.45 285.44 0.30

An Horizon

Depth (cm) 30.75 (8.46) 20.00 54.000 27.5

Water content - 10 kPa m3/ m 3 0.196 (0.028) 0.144 0.262 14.2

Water content - 33 kPa m3 / m 3 0.177 (0.028) 0.128 0.247 15.6

Water content - 1500 kPa m3 /m 3 0.062 (0.012) 0.027 0.081 19.4

Available water at - 10 kPa m3/m3 0.17 (0.027) 0.117 0.231 16.1

Available water at - 33 kPa m3/m3 0.146 (0.027) 0.086 0.214 18.5

Available water storage capacity at - 10 kPa 5.22 (1.8) 3.12 11.628 34.4

Available water storage capacity at - 33 kPa 4.52 (1.71) 2.560 10.514 37.80

Bulk density (g/cm3) 1.27 (0.09) 1.109 1.465 7.3

Sand % 46.37(11.03) 24.60 76.350 23.8

Silt % 39.03 (10.41) 10.05 65.015 26.7

Clay % 14.61 (3.4) 1.100 24.30 23.3

Btl Horizon

Depth (cm) 24.73 (6.53) 12.00 40.00 26.4

Water content - 10 kPa m3/ m 3 0.195 (0.033) 0.150 0.281 16.8

Water content - 33 kPa m3 / m 3 0.176 (0.028) 0.131 0.236 15.7

Water content - 1500 kPa m3 /m 3 0.059 (0.012) 0.025 0.075 19.5

Available water at - 10 kPa m3/m3 0.19 (0.028) 0.142 0.256 14.9

Available water at - 33 kPa m3/m3 0.164 (0.022) 0.120 0.207 13.3

Available water storage capacity at - 10 kPa 4.72 (1.59) 2.180 8.621 33.7

Available water storage capacity at - 33 kPa 4.07 (1.338) 1.836 7.840 32.8

Bulk density (g/cm3) 1.42 (0.205) 1.184 1.871 14.4

Sand % 51.417(14.33) 21.16 77.050 27.9

Silt % 28.354(10.70) 2.78 52.09 37.7

Clay % 20.228 (7.81) 5.20 24.30 38.6

 

94

 

 

 

 

Table 3.2. Variogram model parameters and sample variance ($2) for selected
variables that were investigated across the study site
Model Parameters

Property Range (m) C/(C+Co) R2 82
Ap thickness (cm)? 56.1 0.993 0.959 71.6
Ap sand %'1' 36.9 0.999 0.936 123.5
Ap silt %i 36 0.999 0.969 103
Ap clay %I 43 0.534 0.452 11.5
Ap bulk density (g/cm3)i 71.1 0.658 0.95 0.0088
Elevation (m)§ 265.87 1.00 0.99 0.613
Slope (%)6 352.64 0.874 0.892 0.0041
Corn grain yield (Mg/ha)'1' 46.5 0.649 0.611 0.605

 

 

 

Notes Texponential model, I spherical model, and 6 Gaussian model

 

 

 

 

 

Table 3.3. Variogram model parameters and sample variance (32) for soil water
content and available water storage capacity at - 10 and - 33 kPa for 70 m
range and the entire range of 224 m across the study site

Model Parameters
Property 50 m Range
(Entire Range)
(m) C/(C+Co) R2 52

Ap soil WC at - 10 kPa 19.6 (21.0) 0.999 (1.00) 0.68 (0.069) 0.023
Ap soil WC at - 33 kPa 18.3 (22.8) 1.00 (1.00) 0.63 (0.077) 0.024
Ap soil AW at - 10 kPa 16.44 (19.7) 1.00 (1.00) 0.71 (0.072) 0.025
Ap soil AW at - 33 kPa 17.64 (18.9) 0.998 (1.00) 0.67 (0.017) 0.038
Ap AWSC at - 10 kPa 32.4(203.0) 0.88 (0.966) 0.88 (0.947) 0.097
Ap AWSC at - 10 kPa 30.3(197.6) 0.95 (0.99) 0.82 (0.969) 0.108

 

Note: The brackets indicate the values for entire range parameters. WC is water content, AW is the
available water, and AWSC is available water storage capacity.

95

 

 

 

Table 3.4. Stepwise multiple regression analysis of soil properties that affect yield
variation in the Ap and Btl horizons.
Predictors Predicted yield Cumulative P-value
,2
A9
Ap depth (cm) Yield = 48.99 + 0.0525 Ap depth 0.2017 0.0014'
Ap silt (%) Yield = 48.99 + 0.02475 Ap silt + 0.2641 0.0569”
0.0525 Ap depth
Ap AWSC - 10 kPa 0.2784 0.356 ‘-
Elevation (m) 0.2796 0.791
Slope (%) 0.2849 0.578
Btl
Btl bulk density (g/cm3) Yield = - 4.99 + 2.43 Btl bulk 0.2404 0.0004‘
density
Btl silt (%) Yield = - 4.99 + 0.035 Btl silt + 0.3716 0.0037'
2.43 Btl bulk density
Slope (%) 0.3767 0.553
Btl depth (cm) 0.3803 0.621
Btl sand (%) 0.3832 0.658

 

 

 

*, **, significant at 0.05 and 0.10 probability levels respectively, AWSC is available water storage.

 

 

 

Table 3.5. The spatial relationship between corn grain yield (Mg/ha) and selected soil
properties that are significant in determining yield variation in stepwise
regression analysis.

7 7 Model Parameters
Variable Range (m) (c768 + a R211 52 f 7
Corn grain yield and Ap depth 1' 176 0.875 0.609 0.0048
Corn grain yield and Ap silt'l' 31.5 0.999 0.109 0.0045
Corn grain yield and Btl bulk densityI 157.9 0.653 0.556 0.0038
Corn grain yield and Btl silt '1' 26.10 0.934 0.021 0.0045
Corn grain yield and elevation (1 211.83 0.998 0.782 0.0051

 

Notes: 1‘ is an exponential model, I is a spherical model while 6 is a gaussian model

96

Table 3.6.

The spatial relationship between elevation (m) and Ap soil water at

different pressures, Ap thickness and Ap silt over the entire range of the

study site

 

Model Parameter

 

 

Property Range (C/Co + R2 52

(In) C)
Ap water content at - 10 kPa and elevation '1' 121.76 1.00 0.512 0.0197
Ap water content at - 33 kPa and elevation ‘1' 121.07 0.998 0.491 0.0241
Ap available water at - 10 kPa and elevation § 711.70 0.998 0.938 0.0281
Ap available water at - 33 kPa and elevation 6 570.88 0.999 0.962 0.031
Ap water storage capacity at - 10 kPa and 231.92 0.998 0.968 0.091
elevation
Ap water storage capacity at - 10 kPa and 233.83 0.998 0.963 0.141
elevation é
Ap thickness and elevation § 548.02 1.00 0.998 0.059
Ap silt and elevation 6 149.65 0.999 0.833 0.052

 

Note: 7‘ is an exponential model and 6 is a gaussian model

97

 

 

 

Table 3.7.

The spatial relationship between slope (%) and Ap soil water at different
pressures, Ap thickness and Ap clay over the entire range of the study site

 

Model Parameter

 

Property

Ap water content at - 10 kPa and slope '1

Ap water content at - 33 kPa and slope 1'

Ap available water at - 10 kPa and slope é

Ap available water at - 33 kPa and slope é

Ap water storage capacity at - 10 kPa and

slope é

Ap water storage capacity at - 10 kPa and

slope é

Ap thickness and slope é

Ap clay and slope é

Range (m) (C/Co + C) R2

150.90
132.60
711.70
668.18
270.89

264.31

260.00
325.8

Note: 1‘ is an exponential model and 6 is a gaussian model

98

1.00
1.00
0.998
0.974
0.999

0.998

0.999
1.00

0.249
0.223
0.925
0.961
0.962

0.957

0.973
0.941

U)

0.0004
0.0006
0.0003
0.0004
0.0022

0.0023

0.063
0.0009

 

 

285 "

Elevation (m)

1~1

(191W

   
 

“ -’-- "‘T m;“”‘”"‘"‘;; ””“T "“"—“;7 '" ,,_,_____ ---— I-“ ~ 1
1 I 1 ,1 / {2 i
“011.. 00000000 00 3'1

‘“%$IW 2004mm

Northing (m)

 

Figure 3.1a Sample locations of the study site overlaid over the contour lines

 

Figure 3.1b. Sample locations on the rmin transect overlaid over elevation

99

 

148

 

   

 

 

 

      

 

o U
11 1
74
I79 Ap Thickness 37 Ap Sand Content
0.0' ‘l"

127 O O
8 0.61 1 ' " """ 96 .....
5
g 0.458 64
.E 0305 3, Ap Silt Content
cu
m 0.153 Elevation Ofi

0.000r—v

 

 

 

Ap Clay Content

 

 

 

 

 

   

 

 

. . o
O
0.213 Corn grain yield from residuals 0.0025 Ap Bulk Density
0.00()' ‘ ' i i 3 f ‘ ; fl ' ' ' 0.0000. . t . r v t v , V e r 1
O 25 50 75 100 0 25 50 75 100

Separation Distance (in) Separation Distance (111)

Figure 3.2. Variograms of soil properties for Ap horizon, Ap thickness, Ap texture
bulk density, elevation, slope, and corn grain yield (from residuals)

100

Figure 3.3.

 

 

 

000024 Ap wc at - 10 kPa
0.00000'
0.0323: 0 o 0 9 O
3 ’ . .. 0
.5; 0.0242: 0
a t
.5. 00162
E 0
{jg .
0'0081 - Ap AW at - 10 kPa
0.0000

 

 

 

 

 

0.09521: . - . .. ........
0.0714:
0.0476’
,
”-0238: Ap AWSC at - 10 kPa
0.1)(l()()'?9-tt¢;:;.;
0 13 25 38 50
Separation Distance (m)

 

   

Ap WC at - 33 kPa

 

0.0000 ,7

 

Ap AW at - 33 kPa

 

    

Ap AWSC at - 33 kPa

. . A A A A

0 15 30 44 59

Separation Distance (m)

Variograms of soil water in Ap horizon. WC is water content, AW is

available water, and AWSC is available water storage capacity.

 

 

Figure 3.4.

Cross Semivariance

 

-0.96

0.830 '

0.610"

0.389"

0.169"

 

 

 

 

0.052

0.180

0.103

0.025

-0.053

0 0 c o

0
Cl

0

Com grain yield and Btl silt

 

-0.l30'

0.120"
1»

0.029 ’
1 Com grain yield and Btl bulk density

-0.001

 

O

 

 

 
   

Com grain yield and elevation

 

0 25 50 75 100
Separation Distance (m)

Cross variograms of residual corn grain yield (Mg/ha) and significant soil
properties in the Ap and Btl horizons and elevation. Dotted lines indicate
sample covariance.

102

 

 

O
O Ap WC at - 33 kPa and elevation

 

 

AA
7 v

0.086 ’

0.054“

 

0022‘:

Cross Semivariance

 

Ap AW at - 33 kPa and elevation

 

-0.010

 

 

   
     

Ap AWSC at - 33 kPa and elevation

 

 

 

0.663

  
 

0.497 ‘ _
Ap Tl lickness and elevation

 

 

0.332 0.068
“-166 0'034 Ap silt and elevation
0000. v.- v v e v c . Y . . . r -0_()()|- v. v v 4' r . . v V v . f
0 50 100 150 200 0 50 100 150 200
Separation Distance (m) Separation Distance (m)

Figure 3.5. Cross variograms of elevation and soil water at different pressures, Ap
thickness, and Ap silt. Dotted lines, if present, indicate sample
covariance, and if not present, the sample covariance is out of range of y-
axis. WC is water content, AW is available water, and AWSC is available
water storage capacity.

103

   

Ap we at - 10 kPa and slope ’ Ap we at - 33 kPa and slope

 

 

        

 

 

 

 

 

      

 

        

 

 

 

 

0.0015v 0 0.0016 '
0.0011" 0.00124
1* .
0.0006“ 000071
0.0002» 0 0 0.0003:
‘ 0 AP A“, 31 _ 10 kPa and SlOpC ‘ t ApAW at - 33 kPa and SlOpC
-00003 00002
D
a 0.0075 0 o
.2 Ap AWSC at - 10 kPa and slope
7;? 0.0056
25
at 0.0037
U)
U)
5 0.0019
0.0000W v
0.00291.
0.0059r .
0,0042 * Ap thickness and slope 00020:
“ 0.0012
0.0025» .
’ 0.00034-
0.0007
’,,,L, 00006 %¢;:Afi;e.j..f
“mm C I ' ' ' I F ' ' ' 0 50 100 150 200

0 50 100 150 200
. Separation Distance (in)
Separation Distance (n1)

Figure 3.6. Cross variograms of slope and soil water at different pressures, Ap
thickness and Ap clay. WC is water content, AW is available water, and
AWSC is available water storage capacity.

104

 

Volumetric Water Content %

Figure 3.7. Volumetric water content (T DR) for three different
dates at five different depths in the effective rootzone

 

35
30 r
25 ~
20 4 i
15 .
10 .

Depth 0-18 cm

 

30 .Depth18-37
25 i
20 r
15 *
10 *

   

 

30 qDepth37-53 cm

I

. .1
'5‘ . 1W. v ‘ 11“ ‘
10-1 NV ’y’ 5“: 7

35 .Dept1153-71 m
30. ‘
25- .
20« 1‘1,
“’1 ‘ ' ' "it“! 1" 1'

0 + 5/30/2000
40 . Depth 71 ' 89 —v— 707/2000

-I— 9f29/2000

 

 

 

 

 

30-
20:
10‘

 

 

 

 

 

0 50 100 150 200 250
Distance (m)

105

Figure 3.8. Cumulative probability of soil water storage(0 - 90) for
two different dates: The most dry on June 2, 2000 and
the most wet storage conditions on Agust 16, 2000.

 

 

 

 

 

 

1.0 -
June 2, 2000 20580
112640.194332g3648194
11 '
28’ .- 126
:1 10’48 9 48
:5 30 o 41 e 4
14 10
.8 40 I 9 3
E 0.6 — 17 . 7344 30 {443
a. 39032 6, 4o
.3 43 6 3; 17
f; 0.4 . 2&3 24, 39
3 45 43 13, 31
S 210.33 27, 4‘62
0 36 ‘47 45 43
0.2 . 15 2f33
36 47
354:6 255'6 .16 A 16 2000
1 35 u ust ,
.20 $318 2 .658 g
0.0 a 2522
0 50 100 150 200 250 300

Water storage, mm

 

106

 

Figure 3.9. Rank of temporal relative deviation from the mean spatial
water storage. Vertical bars are associated time standard
deviations and numbers are measuring locations or plots.

 

(l6

OA-e

 

OJZ~
01)~

412 -

E5
25 }
-0.4 - £322

0 10 20 30 40 50 60
Rank

Mean relative difference

 

 

 

 

107

Mean Relative Difference (%)

Figure 3.10. Rank of temporal relative deviation from the mean spatial
water content. Vertical bars are associated time standard
deviations and numbers are measuring locations or plots.

 

0-l8cm

boo
NON
“ES“?
qF—O—l
“£33;
0‘—
1 a
O
a—+—c
5%
e+$.
2%
zi—o—i

 

18 -36cm

booo
NON-b
[Ill
:53;
131:
\O
#1
Pro—4.:
1";
£5
\l
w
t
:3
w
G
A \O
w
w
&
N
on
ool——O—‘l
i—e—i

 

O6 36 - 53 cm 2
0.4 ~ 17 36 34 8 ii
0.2 ‘ 4 3 4

b o
N c
l I
$13
~1—e—1
.5
“LEI”
\O
1-0-12
Si—O—L
%
O‘

33.33.
ON-h
ll
mN

 

.—:<'>.
O00

53-71cm

grit

1111111111
$5
\t
—
as
to
N
N
\l
w
—
06
_
”w
M —n
w
\l
_
0101
A

rcpcpppppp
OOOON-bNON-RQOO

 

H 0‘
o -5
1 1

N

71 - 89 cm
' 4
. ,,f
0.6 " 3724§¥é§3l

27
0.2 " 53234 01% 7 l3 8
-02 4 329;? gfifiifig @331}?
-0.6 7 2&5313 21

 

 

 

0 10 20 30 40 50
108 Rank

60

REFERENCES

Beckett, P. H. T. and R. Webster. 1971. Soil variability: a review. Soil and Fertil. 34: 1-
15.

Blake, G. R. and K. H. Hartge. 1986. Bulk density. p. 363-375. In: A. Klute (ed.)
Methods of soil analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA, Madison, WI.

Bell, J. C., C. A. Butler, and J. A Thomas. 1995. Soil terrain modeling for site-specific
management. p. 207- 227: In P. C. Robert, R. H. Rust, and W. E. Larson, )Eds).
Proceedings of the second international Conference on Site-Specific. Management
for Agric. Systems, Bloomington/Minneapolis, MN, 27 - 30 March 1994", ASA
Miscellaneous Pub, ASA, CSSA, and SSSA, Madison, WI.

Boyer, D. G., R. J. Wright, C. M. Feldhake, and D. P. Bligh. 1996. Soil spatial variability
relationships in a steeply sloping acid soil environment. Soil Sci. 161:278—287.

Cassel, D. K., O. Wendroth, and D. R. Nielson. 2000. Assessing spatial variability in an
agricultural experiment station field. Opportunities arising from spatial
dependence. Agron. J. 92:706 -714.

Chung, C. K., S. Chong, and E. C. Varsa. 1995. Sampling strategies for fertility on a
stony silt loam soil. Commun. Soil Sci. Plant Anal. 26:741-763.

Crum, J. R., G. P. Robertson and F. Nurenberger. 1990. Long-term climate trends, and
agricultural productivity in southwestern Michigan. p 53-58 In D. Greenland and
L. W. Swift (Eds). Climate variability and ecosystem response. US. Forest
Service, US. Dept. of Agriculture, Washington, DC, USA.

Everett, M. W. and F. J. Pierce. 1996. Variability of com yield and soil profile nitrates in
relation to site-specific management. p. 43-53. MPG. Robert et al. (eds)
Precision Agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Freund, R. J. and R. C. Littel. 2000. SAS” System for Regression, Third Edition. SAS
Institute Inc., Cary, NC, USA.

Gee, G. E. and J. W Bauder. 1986. Particle-size analysis. p.383-410. In: A. Klute (ed.)
Methods of soil analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA, Madison, WI.

Han, S., R. G. Evans, T. Hodges, and S. L. Rawlins. 1995. Linking a geographic

information system with a potato simulation model for site-specific management.
J. Environ. Qua]. 24:772-777.

109

 

Han, S., S. M. Schneider, R. G. Evans, and S. L. Rawlins. 1996. Spatial variability of soil
properties on two center pivot irrigated fields. p. 97-106. In:P. C. Robert et al.
(eds.) Precision Agriculture. ASA Misc. Publ., ASA, CSSA, and SSSA, Madison,
WI.

Issaks, E. H. and R. M. Srivastava. 1989. An introduction to applied geostatistics. Oxford
Univ. Press, Oxford.

J aynes, D. B., T. S. Colvin, and J. Ambuel. 1995. Yield mapping by electromagnetic
induction. p. 383-394. In P. C. Robert et al. (eds.) Site-specific management for
agricultural systems. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Khakural, B. R., P. C. Robert, and D. J. Mulla. 1996. Relating corn/soybean yield to
variability in soil and landscape characteristics. p. 117-128. In: P. C. Robert et al.
(eds.) Precision Agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison,
WI.

Klute, A. 1986. Water retention: Laboratory methods. p. 635-662. In: A. Klute (ed.)
Methods of soil analysis. Part 1. 2“d ed. Agron. Monogr. 9. ASA, Madison, WI.

Lake, J. V., G. R. Bock, and J. A. Goode (eds.). 1997. Precision Agriculture: Spatial and
temporal variability of environmental quality. John Wiley & Sons, New York.

Larson, W. E. and P. C. Robert. 1991. Farming by soil. p. 103-112. In: R. Lai and F. J.
Pierce (eds.) Soil management for sustainability. SWCS, Ankeny, Iowa.

Mallants, D., B. P. Mohanty, D. Jacques, and J. Feyen. 1996. Spatial variability of
hydraulic properties in a multi-layered soil profile. Soil Sci. 161:167-181.

Mokma, D. L. and J. A. Doolittle. 1993. Mapping some loamy Alfisols in southwestern
Michigan using ground-penetrating radar. Soil Survey Horizons 34:71-77.

Mulla, D. J ., and J. S. Schepers. 1997. Key processes and properties for site-specific soil
and crop management. p. 1- 18. In: F. J. Pierce and E. J. Sadler (eds.) The state of
site-specific management for agriculture. ASA Misc. Publ. ASA, CSSA, and
SSSA, Madison, WI.

Pierce, F. J. and P. Nowak. 1999. Aspects of precision agriculture. Adv. Agronomy 67:1
- 85. Academic Press.

Pierce, F. J. and D. D. Wamcke. 1994. Site-specific management: An overview. p.3-6.

In: Site-specific management in agriculture. Proc. Workshop. Kellogg Center,
MSU, East Lansing, MI, March 7,1994.

110

 

Pierce, F. J ., D. D. Warncke, and M. W. Everett. 1995. Yield and nutrient variability in
glacial soils of Michigan. p. 133-154. In P.C. Robert et al. (eds.) Site-specific
management for agricultural systems. ASA Misc. Publ. ASA, CSSA, and SSSA,
Madison, WI.

Pierce, F. J. and E. J. Sadler (eds.). 1997. The state of site-specific management for
agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Pierce, F. J ., M. C. Fortin, and M. J. Staton. 1994. Periodic plowing effects on soil
properties in a no-till farming system. Soil Sci. Soc. Am. J. 58:1782-1787.

Poier, K. R. and J. Richter. 1992. Spatial distribution of earthworms and soil properties in
an arable loess soil. Soil Biol. Biochem. 24:1601-1608.

Reinert, D. J. 1990. Soil structural form and stability induced by tillage in a Typic
Hapludalf. Ph.D. Diss. Michigan State University, East Lansing, MI.

Robert, P. 1993. Characterization of soil conditions at the field level for soil specific
management. Geoderma 60:57-72.

Robertson, G. P., M. Huston, F. Evans, and J. Tiedje. 1988. Spatial variability in

successional plant community: patterns of nitrogen availability. Ecology 69: 15 1 7 -
1524.

Robertson, G. P., J. R. Crum and B. G. Ellis. 1993. The spatial variability of soil
resources following long-term disturbance. Oecologia 96:451-456.

Robertson, G. P., K. M. Klingensmith, M. J. Klung, E. A. Paul, J. R. Crum and B. G.
Ellis. 1997. Soil resource, microbial activity, and primary production across an
agricultural ecosystem. Ecological Application: 7(1) p. 158-170.

Robertson, G. P. and K. L. Gross. 1994. Assessing the heterogeneity of below -ground
resources: qualifying pattern and scale.p237 - 253 In M. M. Caldwell and R. W.

Pearcy (eds.) Plant exploitation of environmental heterogeneity. Academic Press,
New York, New York, USA.

SAS Institute. 2000. SAS/STAT user's guide. Version 8.1. SAS Inst., Cary, NC.

Tobler, W. 1970. A computer movie simulating urban growth in Detroit region.
Economic Geography 46(2):234 - 240.

Trangmar, B. B., R. S. Yost, and G. Uehara. 1985. Application of geostatistics to spatial
studies of soil properties. Advances in Agronomy 38:45-94.

111

 

Vachaud, G., A. P de Silans, P. Balabanis, and M. Vauclin. 1985. Temporal stability of
spatially measured soil water probability density function. Soil Science Soc. Am.
J. 492822 - 828.

Webster, R. 1985. Quantitative spatial analysis of soil in the field. Advances in Soil Sci.
3:1-70.

Webster, R. and M. A. Oliver. 1990. Statistical methods in soil and land resource survey.
Oxford Univ. Press, New York.

112

 

SUMMARY

Early season drought in 1999 resulted in lower leaf chlorophyll even after
intervention N application until irrigation or rainfall occurred in early July which resulted
in increased leaf chlorophyll. Starter N application at planting time was necessary to
avoid early season N stresses and subsequent yield losses. Interactive N management
based upon PSNT or leaf chlorophyll readings resulted in less N applied and produced a
comparable yield. There was a significant N x water interaction when irrigation water
was applied in 1999, but the interaction was absent when irrigation was not needed in the
wet 2000 growing season. Irrigation increased corn grain yield but not without N
application. Nitrogen was the main factor determining corn grain yield in the relatively
wet 2000 season. Soil NO3-N measurements (though PSNT) revealed that most of the N
applied at planting was lost even before the crop reached silking. Irrigation had no effect
on earleaf N content. Nitrogen application increased leaf N in the dry 1999 season but
had no effect in the wet 2000 growing season. Irrigation and N application had very
significant effects on grain N content. While irrigation reduced grain N content on one
hand, N addition to the crop increased grain N content on the other hand. There was no
irrigation x N interaction in either seasons. End of the season assessment of nitrate
concentrations in the lower portion of comstalk showed that N fertilization had a
significant impact on lower stalk NO3-N build up. Irrigation reduced stalk NO3-N
accumulation in the lower stalk and there was an irrigation x N interaction. When

comstalk NO3-N levels were compared to levels in an Iowa State University test, optimal

113

 

stalk NO3-N levels were found only in the at-planting N treatment. Sensor-based and
PSNT N treatments represented a marginal NO3-N category based on the ISU data,
however, the lower PSNT and sensor-based applications produced comparable yields to
those of the at-planting N treatment. Care must be taken when considering these
categorizations because there are many factors affecting N uptake. Generally under dry
conditions, corn will have higher levels of nitrate accumulated in the lower comstalk than
will the lower stalks of irrigated corn.

Vegetation indices can be used to develop instantaneous crop coefficients from
crop growth and development that are independent of the planting date and effective
cover. While NDVI is sensitive to soil backgrounds, SAVI is more susceptible to sky
illumination. Green normalized difference vegetation index (GNDVI) is sensitive to soil
background but is not affected by sky illumination. The green normalized difference
vegetation index (GNDVI) was transformed into a reflectance-based crop coefficient.
Kcr GNDVI performed well when compared to Kcr from SAVI and ch (Wright) and did
not under- or overestimate evapotranspiration of corn. An equation (Kcr = 1.256 x
GND VI - 0.1699) was derived that represented all field environments for estimating corn
crop coefficient and did not require any more calibration. Reflectance-based crop
coefficients can be used to apply variable irrigation rates based on water stress in corn
rather than on the current basis of demand or weather-based water budgets which do not
account for spatial variability within a field. In addition, it is possible to remotely
monitor water stress in plants and apply irrigation water based on site-specific needs of

corn within a field rather than to apply water unifome over a field as is the current

114

 

 

 

practice

Nitrogen deficiency in corn was detected using a radiometer in the short
wavelength range of 460 to 810 nm. A spectral region centered around 560 nm provided
the best separation between N treatments. These results suggest that a future corn N
status sensor could be developed that uses a simple combination of limited spectral bands
centered around 560 and 800 nm for variable N rate application. Spectral vegetation
indices such as GNDVI and GRVI performed better for corn grain yield prediction than
RVI and NDVI. This was due to the sensitivity to N deficiency in the green region of the
spectrum that centers around 560 nm. Vegetation indices such as GNDVI could be used
to replace chlorophyll meter readings to schedule N fertilizer and to predict N stress in
corn. Nonetheless, there are questions to be addressed such as how, when, and what rate
of N fertilizer to apply for sensor-based N application. Spectral vegetation indices (SVI)
such as NDVI and GNDVI performed very well in estimating biophysical variables such
LAI and Fe. Extensive data are still needed to verify the validity of these results. Finally,
because remote sensing measurements in the temperate region are hampered by a very
high percentage of cloud cover, suitable atmospheric conditions must be present to obtain
high quality measurements.

Geostatistical analysis provides a substantial and a robust technique for
determining the scale over which spatial dependence occurs. Soil properties such as
length of the horizon, bulk density, soil water content at - 10, - 33, and - 1500 kPa, soil

texture, elevation, and slope from a 224 m long transect in the experimental area

115

 

 

 

exhibited a wide range of variability. For all properties tested, variability exhibited strong
spatial structure, and 50 to 100 percent of the sample variance was spatially dependent
and autocorrelated over a range of 5 to 224 m. Stepwise multiple regression analysis was
applied on corn grain yields to identify the most important soil properties affecting yield
variation in the Ap and Btl horizons. Ap depth and Ap silt were significant in explaining
corn grain yield variation in Ap horizon. Btl bulk density and Btl silt were significant in
explaining corn yield variation in Btl horizon. Surprisingly, soil water content was not
significant in the stepwise multiple regression conducted to identify the variables that
most affected corn grain yield. This may be due to the fact that the 2000 season was
relatively wet and did not require irrigation. Corn grain yield and the soil properties in
the Ap and Btl horizon exhibited a very strong spatial cross correlation. Fifty to 100% of
the sample covariance was spatially dependent over a range of 26 to 212 m. It is evident
from these results that the soil resources tested are highly variable and very
heterogeneous. In addition, properties such as Ap texture, Ap bulk density, Ap depth, Ap
water content at different pressures, Btl bulk density, and Btl silt are spatially structured.
Landscape properties such as elevation and slope are spatially well-structured and nested
or cyclic. The degree and scale of spatial dependence found in this study is similar to an
experiment conducted at KBS by Robertson et al. (1988). This variable soil structure
should be considered when designing experiments and making management decisions.
Cumulative probability function, Spearman rank correlation coefficient, the mean

relative difference, and the standard deviation were used to characterize the temporal

116

 

stability of spatial water storage and spatial water content. The smaller the standard
deviation the more significant was the rank. The location whose respective rank is
relatively stable can be used as the management zone for homogeneous water application,
and no more soil sampling is needed except at low, medium and high rank locations.
There is a temporal stability in soil water content in the upper soil depths compared to the

lower depths due to the action of plant root.

117

r;-
I

 

APPENDIX

The following tables provide supplementary soil profile data collected fiom the study site

after harvesting for Ap, Btl, Bt2 and 2Bt2 (C) horizons.

118

 

 

 

 

J. Soil Prope ' '
Length Water Water Water Bulk
of Horizon Content Content Content Density Sand Silt Clay
Plot # (cm) 0.10 Bar 0.33 Bar 15 Bar fim‘ % % %

101 38 0.17 0.15 0.12 1.14 60.13 30.75 9.13
102 37 0.23 0.19 0.14 0.95 52.95 33.90 13.15
103 27 0.20 0.18 0.13 0.99 51.40 35.63 12.97
104 35 0.20 0.18 0.10 1.11 46.79 39.36 13.85
105 37 0.17 0.15 0.12 1.02 47.45 37.08 15.47
106 30 0.26 0.25 0.17 0.94 45.90 38.36 15.74
107 45 0.22 0.21 0.18 0.90 40.08 41.69 18.24
108 48 0.23 0.21 0.18 0.90 28.93 52.93 18.15
109 54 0.32 0.25 0.11 0.92 24.79 65.02 10.20
110 45 0.23 0.21 0.16 0.97 24.60 60.11 15.29
111 52 0.22 0.25 0.16 0.92 34.23 53.26 12.52
112 34 0.21 0.19 0.16 0.94 44.85 42.80 12.35
201 44 0.17 0.16 0.12 1.00 54.33 31.73 13.95
202 43 0.18 0.16 0.11 1.02 60.64 31.22 8.15
203 41 0.15 0.13 0.09 1.12 63.80 26.53 9.67
204 37 0.18 0.15 0.12 1.11 62.17 27.63 10.20
205 34 0.16 0.15 0.12 1.08 54.05 32.53 13.42
206 29 0.21 0.18 0.17 1.05 76.35 10.05 13.60
207 25 0.24 0.23 0.17 1.01 42.31 49.99 7.70
208 24 0.25 0.22 0.18 1.06 39.87 35.83 24.30
209 30 0.20 0.18 0.16 1.09 73.38 10.53 16.10
210 24 0.18 0.17 0.13 0.98 45.18 39.63 15.20
211 30 0.17 0.15 0.13 1.01 55.53 31.15 13.33
212 25 0.15 0.14 0.09 1.19 62.28 22.53 15.20
301 25 0.16 0.15 0.10 1.16 55.50 29.12 15.38
302 29 0.19 0.16 0.13 1.08 54.10 35.70 10.20
303 27 0.18 0.17 0.13 1.15 44.43 39.66 15.92
304 29 0.24 0.20 0.17 1.13 35.32 56.18 8.50
305 28 0.19 0.18 0.17 1.07 44.25 40.10 15.65
306 26 0.18 0.15 0.08 1.12 55.38 31.92 12.70
307 25 0.15 0.14 0.11 1.06 54.80 32.14 13.06
308 32 0.17 0.15 0.12 1.12 45.83 41.82 12.35
309 24 0.19 0.17 0.11 0.99 43.83 39.72 16.45
310 23 0.19 0.17 0.14 1.12 42.32 42.48 15.20
311 25 0.14 0.13 0.10 1.08 42.38 38.68 18.95
312 22 0.18 0.16 0.15 0.96 42.85 41.50 15.65
401 25 0.21 0.18 0.12 1.04 40.03 41.38 18.60
402 27 0.21 0.19 0.16 1.02 34.10 47.30 18.60
403 25 0.21 0.19 0.17 1.06 35.65 47.45 16.90
404 24 0.21 0.19 0.17 1.05 38.03 44.72 17.26
405 25 0.20 0.18 0.15 0.95 40.95 42.60 16.45
406 20 0.18 0.17 0.14 1.08 40.38 43.98 15.65
407 24 0.19 0.17 0.14 0.95 38.83 45.08 16.10
408 26 0.19 0.18 0.14 0.97 39.75 44.60 15.65
409 26 0.20 0.18 0.12 1.04 41.00 42.90 16.10
410 25 0.20 0.18 0.15 0.99 41.90 44.77 13.33
411 22 0.21 0.18 0.14 0.95 41.03 42.88 16.10
412 24 0.23 0.20 0.16 1.01 40.99 36.50 22.52

 

119

afii

 

 

2. Soil Properties in Btl Horizon

 

 

Length Water Water Water Bulk
of Horizon Content Content Content Density Sand Silt Clay
Plot # (cm) 0.10 Bar 0.33 Bar 15 Bar g/cm3 % % %
101 19 0.15 0.14 0.12 1.18 64.58 19.96 15.47
102 18 0.16 0.13 0.08 1.12 71.38 12.71 15.92
103 16 0.12 0.10 0.07 1.28 77.05 14.62 8.33
104 25 0.25 0.21 0.12 1.03 37.77 39.99 22.25
105 14 0.13 0.09 0.06 1.50 76.80 13.89 9.31
106 36 0.12 0.11 0.07 1.44 74.85 16.82 8.33
107 15 0.10 0.09 0.04 1.50 74.25 14.74 11.01
108 28 0.14 0.13 0.11 1.48 61.45 27.90 10.65
109 32 0.18 0.18 0.10 1.25 69.33 25.47 5.20
110 35 0.10 0.10 0.09 1.50 69.50 20.21 10.29
111 35 0.20 0.19 0.12 1.22 48.50 33.17 18.33
112 20 0.15 0.14 0.12 1.44 45.33 43.58 11.10
201 29 0.20 0.19 0.17 1.09 44.79 30.36 24.85
202 19 0.16 0.14 0.12 1.52 52.63 32.18 15.20
203 16 0.12 0.11 0.07 1.38 69.78 21.63 8.60
204 31 0.23 0.19 0.12 1.22 44.74 43.81 11.45
205 32 0.23 0.21 0.19 0.96 31.03 39.76 29.22
206 20 0.22 0.21 0.18 1.00 48.35 23.05 28.60
207 40 0.25 0.24 0.20 0.99 21.16 45.70 33.15
208 21 0.24 0.22 0.16 1.03 46.21 44.05 9.75
209 33 0.21 0.19 0.17 1.12 48.80 25.28 25.92
210 28 0.19 0.17 0.13 1.09 56.93 19.93 23.15
211 21 0.13 0.12 0.10 1.36 72.40 16.95 10.65
212 15 0.13 0.13 0.08 1.18 74.10 10.16 15.74
301 12 0.11 0.09 0.04 1.28 55.43 21.88 22.70
302 26 0.18 0.16 0.14 1.06 52.44 31.12 16.45
303 32 0.17 0.15 0.12 1.19 56.90 22.45 20.65
304 27 0.28 0.23 0.20 0.99 38.92 30.43 30.65
305 31 0.21 0.19 0.17 0.99 41.68 30.09 28.24
306 24 0.19 0.17 0.14 1.00 50.02 29.79 20.20
307 27 0.21 0.20 0.18 1.06 43 .68 30.68 25.65
308 26 0.20 0.19 0.16 1.11 73.97 2.78 23.25
309 13 0.09 0.08 0.06 1.37 33.26 44.39 22.35
310 16 0.21 0.20 0.15 1.05 40.21 52.09 7.70
311 23 0.19 0.17 0.15 1.08 55.88 17.23 26.90
312 20 0.20 0.18 0.18 1.12 46.40 28.04 25.56
401 22 0.19 0.18 0.13 1.11 36.43 41.68 21.90
402 26 0.21 0.20 0.16 1.10 33.50 39.15 27.35
403 27 0.20 0.19 0.16 1.10 42.18 32.18 25.65
404 19 0.21 0.19 0.17 1.03 50.25 21.60 28.15
405 25 0.21 0.19 0.15 0.97 36.88 34.18 28.95
406 30 0.20 0.18 0.17 1.02 38.18 33.68 28.15
407 24 0.18 0.16 0.13 1.05 43.10 29.55 27.35
408 25 0.19 0.18 0.15 1.03 49.69 22.06 28.25
409 24 0.21 0.20 0.17 1.03 37.95 34.70 27.35
410 33 0.12 0.11 0.08 1.07 41.05 35.62 23.33
411 25 0.26 0.19 0.19 1.02 37.60 36.30 26.10
E412 20 0.23 0.22 0.17 1.10 50. 23 2 .

 

120

 

 

 

3. Soil Properties in Bt2 Horizon

 

 

Length Water Water Water Bulk
of Horizon Content Content Content Density Sand Silt Clay
Plot # (cm) 0.10 Bar 0.33 Ba_r 15 Bar g/cm3 % % °/o

101

102

103

104 28 0.20 0.1 0.15 1.15 37.19 36.46 26.35
105 13 0.12 0.11 0.07 1.59 78.23 12.29 9.49
106 19 0.15 0 1 0.12 1.48 66.98 17.47 15.56
107 12 0.11 0.06 1.33 80.48 6.11 13.42
108

109 16 0.16 0.11 0.10 1.19 74.06 13.25 12.70
110

111 4 0.18 0.17 0.13 1.31 55.65 23.43 20.92
112 12 0.14 0.14 0.12 1.54 55.75 28.33 15.92
201 16 0.17 0.16 0.14 1.14 67.27 7.89 24.85
202 1 1 0.11 0.10 0.06 1.43

203 14 0.08 0.06 0.04 1.33 87.38 6.53 6.10
204 9 0.23 0.20 0.15 1.19 62.58 12.23 25.20
205 15 0.19 0.17 0.14 1.07 78.83 5.26 15.92
206 20 0.20 0.18 0.16 1.09 55.13 22.18 22.70
207 23 0.26 0.24 0.16 1.02 49.91 24.44 25.65
208 26 0.1 1 0.10 0.05 1.27 86.23 1.08 12.70
209 9 0.15 0.13 0.11 1.24 74.40 12.90 12.70
210 22 0.09 0.06 0.05 1.29 87.65 1.52 10.83
211 8 0.09 0.08 0.05 1.23 79.43 5.38 15.20
212 16 0.12 0.11 0.07 1.23 78.73 5.63 15.65
301 10 0.08 0.07 0.04 1.26 86.25 5.78 7.97
302 23 0.23 0.20 0.16 1.01 51.29 27.61 21.10
303 20 0.18 0.16 0.14 1.27 56.25 20.60 23.15
304 13 0.19 0.14 0.11 1.22 76.69 5.61 17.70
305 10 0.17 0.16 0.14 1.23 73.93 10.34 15.74
306 14 0.14 0.12 0.09 1.17 65.67 14.49 19.85
307 12 0.13 0.12 0.10 1.15 76.35 8.18 15.47
308 14 0.13 0.12 0.09 1.26 54.07 23.58 22.35
309 15 0.21 0.18 0.14 1.04 67.05 11.05 21.90
310 24 0.21 0.15 0.14 1.08 64.43 26.63 8.95
311 11 0.14 0.13 0.11 1.28 76.10 6.20 17.70
312 18 0.17 0.15 0.13 1.20 68.85 5.50 25.65
401 17 0.22 0.18 0.14 1.02 49.85 22.80 27.35
402 12 0.20 0.18 0.14 1.07 57.33 16.58 26.10
403 17 0.15 0.14 0.12 1.31 69.88 9.48 20.65
404 20 0.08 0.07 0.05 1.28 72.58 6.51 20.92
405 15 0.11 0.10 0.07 1.41 83.63 3.68 12.70
406 24 0.16 0.15 0.12 1.15 72.85 9.00 18.15
407 12 0.16 0.15 0.12 1.17 67.95 18.45 13.60
408 18 0.15 0.14 0.11 1.13 74.33 13.78 11.90
409 15 0.16 0.13 0.08 1.12 85.85 3.05 11.10
410 11 0.14 0.13 0.09 1.30 75.70 8.56 15.74
411 17 0.13 0.12 0.09 1.17 66.90 9.50 23.60
412 12 0.20 0.18 0.11 1.14 75.90 8.01 16.10

121

 

 

 

4. Soil Properties in 2Bt2 (C) Horizon

 

Length Water Water Water Bulk
of Horizon Content Content Content Density Sand Silt Clay
Plot # (cm) 0.10 Bar 0.33 Bar 15 Ba_L g/cm3 % % %

101 18 0.05 0.04 0.03 1.27 92.28 2.17 5.56
102 20 0.08 0.07 0.05 1.39 86.48 3.86 9.67
103 21 0.12 0.11 0.09 1.45 88.03 3.92 8.06
1 10

l l l

201

202 81.01 7.89 11.10
203

204 13 0.24 0.19 0.17 1.27 76.35 7.95 15.70
205 7 0.24 0.08 0.06 1.33 89.65 0.59 9.76
206 17 0.08 0.06 0.04 1.32 91.70 3.10 5.20
207

208

209 17 0.06 0.05 0.04 1.33 91.23 1.79 6.99
210

21 1

212 17 0.12 0.11 0.07 1.50 78.53 5.65 15.83
301 9 0.06 0.06 0.03 1.18 78.75 5.60 15.65
302 11 0.18 0.17 0.11 1.19 75.02 18.95 6.03
303 10 0.14 0.13 0.09 1.47 56.05 20.53 23.42
304 12 0.11 0.09 0.06 1.31 86.54 0.76 12.70
305 14 0.07 0.06 0.05 1.33 87.83 2.60 9.58
306 18 0.10 0.09 0.07 1.32 80.83 6.83 12.35
307 13 0.06 0.05 0.03 1.31 91.73 0.22 8.06
308 13 0.09 0.08 0.06 1.37 81.23 3.58 15.20
309 23 0.10 0.08 0.05 1.21 88.32 3.44 8.25
310 14 0.17 0.16 0.12 1.39 56.80 33.00 10.20
31 1 16 0.08 0.07 0.05 1.34 88.88 0.48 10.65
312 20 0.10 0.09 0.07 1.34 63.18 26.00 10.83
401 26 0.17 0.14 0.10 1.24 80.63 3.73 15.65
402 17 0.12 0.11 0.01 1.32 85.85 1.81 12.35
403 10 0.06 0.05 0.04 1.15 89.80 3.30 6.90
404 13 0.14 0.13 0.10 1.09 89.55 2.30 8.15
405 10 0.13 0.13 0.10 1.52 80.30 4.23 15.47
406 14 0.08 0.07 0.06 1.38 89.83 2.03 8.15
407 22 0.12 0.11 0.08 1.37 85.85 1.80 12.35
408 16 0.09 0.07 0.05 1.26 88.55 1.60 9.85
409 12 0.07 0.07 0.05 1.32 84.93 3.98 1 1.10
410 1 I 0.08 0.08 0.05 1.35 86.50 5.26 8.24
41 1 16 0.24 0.03 0.03 1.27 78.30 9.35 12.35
412 15 0.13 0.08 0.05 1.27 87.73 3.32 8.95

122

 

 

REFERENCES

Al-Abbas, A. H., R. Barr, J. D. Hall, F. L. Crane, and M.F. Baumgardner. 1974. Spectra
of normal and nutrient-deficient maize leaves. Agronomy J. 66:16-20.

Bausch, W. C. 1993. Soil background effects on reflectance-based crop coefficients for
corn. Remote Sensing and the Environment 46:213-22.

Bausch, W. C. 1995. Remote sensing of crop coefficient for improving irrigation
scheduling of corn. Agric. Water Management 27:55-68.

Bausch, W. C., H. R. Duke, and J. C. Iremonger. 1996. Assessment of plant nitrogen in
irrigated corn. In P. C. Robert et al. (ed.) Proceedings of the Third International
Conference on Precision Agriculture. Minneapolis, MN, 23 - 26 June, 1996.
ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Beckett, P. H. T. and R. Webster. 1971. Soil variability: a review. Soil and Fertil. 34:1-
15.

Bell, J. C., C. A. Butler, and J. A Thomas. 1995. Soil terrain modeling for site-specific
management. p. 207- 227: In P. C. Robert, R. H. Rust, and W. E. Larson (Eds).
Proceedings of the second international Conference on Site-Specific. Management
for Agric. Systems, Bloomington/Minneapolis, MN, 27 - 30 March 1994", ASA
Miscellaneous Pub, ASA, CSSA, and SSSA, Madison, WI.

Blackmer, A. M., J. S. Schepers, and G. E. Myer. 1995. Remote sensing to detect
nitrogen deficiency in corn. p. 505 -512. In: Proceedings of site-specific
management for Agric. Systems 27 -30 March 1994, Minneapolis, MN, ASA
Misc. Publ. ASA- CSSA- SSSA, Madison, WI.

Blackmer, A. M., J. S. Schepers, G. E. Varvel, and E. A. Walter-Shea. 1996. Nitrogen
deficiency detection using reflected shortwave radiation from irrigated corn
canopies. Agron. J. 88:1 - 5.

Blackmer, A. M. and A. P. Mallarino. 1996. Comstalk testing to evaluate nitrogen

management. In Iowa State University Extension Publication, PM 1584, August,
1996.

123

"J

F“-
l

Blackmer, T. M., and S. E. White. 1996. Remote sensing to identify spatial patterns in
optimal rates of nitrogen fertilization. p. 33-41. In P.C. Robert et al. (ed.)
Proceedings of the Third International Conference on Precision Agriculture.
Minneapolis, MN, 23 - 26 June, 1996. ASA Misc. Publ. ASA, CSSA, and SSSA,
Madison, WI.

Blackmer, A. M., J. S. Schepers, and G. E.Varve1. 1994. Light reflectance compared to
other nitrogen stress measurements in corn leaves. Agron. J. 86:934 - 938.

Blake, G. R. and K. H. Hartge. 1986. Bulk density. p. 363-375. In: A. Klute (ed.) ,1.
Methods of soil analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA, Madison, WI. ‘3

Boyer, D- G., R. J. Wright, C. M. Feldhake, and D. P. Bligh. 1996. Soil spatial variability '-'-
relationships in a steeply sloping acid soil environment. Soil Sci. 161:278-287.

Camp, C. R. and E. J. Sadler. 1994. Center pivot irrigation system for site-specific water
and water management. ASAE Paper No. 94 - 1586. ASAE, St. Joseph, MI.

Carlson, N. T. and D. A. Ripley. 1997. On the relationship between NDVI, fractional
vegetation cover, and leaf area index. Remote Sens. and the Environ. 62:241-252.

Cassel, D. K., O. Wendroth, and D. R. Nielson. 2000. Assessing spatial variability in an
agricultural experiment station field. Opportunities arising from spatial
dependence. Agron. J. 92:706 -714.

Chung, C. K., S. Chong, and E. C. Varsa. 1995. Sampling strategies for fertility on a
stony silt loam soil. Commun. Soil Sci. Plant Anal. 26:741-763.

Crum, J. R., G. P. Robertson and F. Nurenberger. 1990. Long-term climate trends, and
agricultural productivity in southwestern Michigan. pp 53-58 In D. Greenland
and L. W. Swift (eds.) Climate variability and ecosystem response. US. Forest
Service, US. Dept. of Agriculture, Washington, DC.

Doerge, T. 2001. Variable rate N management for corn production: success proves
elusive. Crop Insights 11(11):1 - 4. Pioneer Hi-bred International.

Evans, R. .G., S. Han, M. W. Kroeger, and S. M. Schneider. 1996. Precision center pivot
irrigation for efficient use of water and nitrogen. p. 75-84. In P.C. Robert et al.
(eds.) Proceedings of the Third International Conference on Precision Agriculture.

Minneapolis, MN, 23-26 June 1996. ASA Misc. Publ. ASA, CSSA, and SSSA,
Madison, WI.

124

 

 

Everett, M. W. and F. J. Pierce. 1996. Variability of com yield and soil profile nitrates in
relation to site-specific management. p. 43-53. In:P.C. Robert et al. (eds)
Precision Agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Ferguson, R. B., C. A. Gotway, G. W. Hergert, and T. A. Peterson. 1996. Soil sampling
for site-specific nitrogen management p. 13 - 22. In P.C. Robert et al. (eds.)
Proceedings of the Third International Conference on Precision Agriculture.
Minneapolis, MN, 23-26 June 1996. ASA Misc. Publ. ASA, CSSA, and SSSA,
Madison, WI.

Freund, R. J. and R. C. Littel. 2000. SAS“ System for Regression, Third Edition. SAS
Institute Inc., Cary, NC, USA.

Gee, G. E. and J. W Bauder. 1986. Particle-size analysis. p.383-410. In: A. Klute (ed.)
Methods of soil analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA, Madison, WI.

Han, S., S. M. Schneider, R. G. Evans, and S. L. Rawlins. 1996. Spatial variability of soil
properties on two center pivot irrigated fields. p. 97-106. InsP. C. Robert et al.

(eds.) Precision Agriculture. ASA Misc. Publ., ASA, CSSA, and SSSA, Madison,
WI.

Han, S., R. G. Evans, T. Hodges, and S. L. Rawlins. 1995. Linking a geographic
information system with a potato simulation model for site-specific management.
J. Environ. Qual. 24:772-777.

Howell, T. 1990. Relationships between crop production and transpiration,
evapotranspiration, and irrigation. p. 391-434. In B.A. Stewart and D. R. Nielsen

(eds.) Irrigation of Agricultural Crops. Agron. Monograph 30, ASA, CSSA, and
SSSA, Madison, WI.

Idso, S. B. 1982. Non-water-stressed baseline: a key to measuring and interpreting plant
water stress. Agric. Meteorology 27:59-70.

Issaks, E. H. and R. M. Srivastava. 1989. An introduction to applied geostatistics. Oxford
Univ. Press, Oxford.

Jackson, R. D. and Alfredo R. Huete. 1991. Interpreting vegetation indices. Preventive
veterinary medicine 11(1991): 1 85 - 2000.

Jackson, R. D. and Printer Jr., P. J. 1986. Spectral response of architecturally different
wheat canopies. Remote Sensing and the Environ. 20:43 - 56.

125

 

 

Jackson, R. D. 1982. Canopy temperature and crop water stress, Adv. Irrigation 1:43-85.

J aynes, D. B., T. S. Colvin, and J. Ambuel. 1995. Yield mapping by electromagnetic
induction. p. 383-394. In P. C. Robert et al. (eds.) Site-specific management for
agricultural systems. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Jensen, M. E. 1968. Water consumption by agricultural plants. p. 1-22 In Kozlowki, T.
T. (ed.) Water Deficit and Plant Growth. Academic Press, New York. Vol. 2.

Khakural, B. R., P. C. Robert, and D. J. Mulla. 1996. Relating corn/soybean yield to
variability in soil and landscape characteristics. p. 117-128. In: P. C. Robert et al.
(eds.) Precision Agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison,
WI.

King, B. A., R. A. Brady, 1. R. McCann, and J. C. Stark. 1995. Variable rate water
application through sprinkler irrigation. p. 485-493. In P.C. Robert et al. (eds.)
Proceedings of the 2"d International Conference on Site Specific Management for
Agricultural Systems. Minneapolis, MN, 27-30 March 1994 ASA Misc. Publ.
ASA, CSSA, and SSSA, Madison, WI.

King, B. A., J. C. Stark, I. R. McCann, and D. T. Westermann. 1996. Spatially varied
nitrogen application through a center pivot irrigation system. p. 85-94. In P.C.
Robert et al. (eds.) Proceedings of the of the Third International Conference on
Precision Agriculture. Minneapolis, MN, 23-26 June 1996. ASA Misc. Publ.
ASA, CSSA, and SSSA, Madison, WI.

Kitchen, N. R., D. F. Hughes, K. A. Sudduth, and S. J. Birrel. 1995. Comparison of
variable rate to single rate nitrogen fertilizer application: corn production and
residual soil NO3-N. p. 427 - 441. In: P.C. Robert et al. (eds.) Proceedings of the
2"d International Conference on Site Specific Management for Agricultural
Systems. Minneapolis, MN, 27-30 March 1994, ASA Misc. Publ. ASA, CSSA,
and SSSA, Madison, WI.

Klute, A. 1986. Water retention: Laboratory methods. p. 635-662. In: A. Klute (ed.)
Methods of soil analysis. Part 1. 2"d ed. Agron. Monogr. 9. ASA, Madison, WI.

Lake, J. V., G. R. Bock, and J. A. Goode (eds.). 1997. Precision Agriculture: Spatial and
temporal variability of environmental quality. John Wiley & Sons, New York.

Larson, W. E. and P. C. Robert. 1991. Farming by soil. p. 103 - 112. In: R. Lal and F. J.
Pierce (Eds). Soil management for sustainability. SWCS, Ankeny, Iowa.

126

 

 

Mallants, D., B. P. Mohanty, D. Jacques, and J. Feyen. 1996. Spatial variability of
hydraulic properties in a multi-layered soil profile. Soil Sci. 161:167-181.

McCann, I. R., and J. C. Stark. 1993. Method and apparatus for variable application of
irrigation water and chemicals. U.S. Patent No. 5 246 164 Date issued:
September 21, 1993.

Mokma, D. L. and J. A. Doolittle. 1993. Mapping some loamy Alfisols in southwestern
Michigan using ground-penetrating radar. Soil Survey Horizons 34:71-77.

Moran, M. S., T. R. Clarke, Y. Inoue, and A. Vidal. 1994. Estimating crop water deficit
using relation between surface-air temperature and spectral vegetation. Remote
Sensing and the Environ. 49:246-253.

Moran, M. S., Y. Inoue, and E. M. Barnes. 1997. Opportunities and limitations for
image-based remote sensing in precision crop management. Remote Sensing and
the Environ. 61 :3 19-346.

Mulla, D. J ., and J. S. Schepers. 1997. Key processes and properties for site-specific soil
and crop management. p. 1- 18. In: F. J. Pierce and E. J. Sadler (eds.) The state of
site-specific management for agriculture. ASA Misc. Publ. ASA, CSSA, and
SSSA, Madison, WI.

Neale, C. M. U, W. C. Bausch, and D. F. Heermann. 1989. Development of reflectance-
based crop coefficient for corn. Transactions, ASAE 32: 1891-1899.

Pierce, F. J ., D. D. Wamke, and M. W. Everett. 1995. Field and nutrient variability in
glacial soils of Michigan. p. 133-151. In P.C. Robert et al. (eds.) Proceedings of
the 2nd International Conference on Site Specific Management for Agricultural
Systems. Minneapolis, MN, 27-30 March 1994 ASA Misc. Publ. ASA, CSSA,
and SSSA, Madison, WI.

Pierce, F. J ., M. C. F ortin, and M. J. Staton. 1994. Periodic plowing effects on soil
properties in a no-till farming system. Soil Sci. Soc. Am. J. 58:17 82-1 787.

Pierce, F. J. and E. J. Sadler (eds.). 1997. The state of site-specific management for
agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Pierce, F. J. and D. D. Warncke. 1994. Site-specific management: An overview. p.3-6.

In: Site-specific management in agriculture. Proc. Workshop. Kellogg Center,
MSU, East Lansing, MI, March 7,1994.

127

 

Pierce, F. J. and P. Nowak. 1999. Aspects of precision agriculture. Adv. Agronomy 67:1
- 85. Academic Press.

Poier, K. R. and J. Richter. 1992. Spatial distribution of earthworms and soil properties in
an arable loess soil. Soil Biol. Biochem. 24:1601-1608.

Qi, J ., R. C. Marsett, M. S. Moran, D. C. Goodrich, P. Heilman, Y. H. Kerr, G. Dedieu,
A. Chehbouni, and X. X. Zhang. 2000b. Spatial and temporal dynamics of
vegetation in San Pedro River Basin area. Agric. And Forest Meteorology
105:55-68 (2000). Elsevier Science Inc. New York, New York, 2000.

Qi, J ., Y. H. Kerr, M. S. Moran, M. Weltz, A. R. Huete, S. Sorooshian and R. Bryant.
2000a. Leaf area index estimates using remotely sensed data and BRDF models
in a semiarid region. Remote Sensing of the Environ. 73:18-30 (2000). Elsevier
Science Inc. New York, New York, 1995.

Reinert, D. J. 1990. Soil structural form and stability induced by tillage in a Typic
Hapludalf. Ph.D. Diss. Michigan State University, East Lansing, MI.

Robert, P. C., R. H. Rust, and W. E. Larson. 1993. Proceedings of soil specific crop

management: A workshop on research and development issues. SSSA Spec. Publ.
SSSA, Madison, WI.

Robert, P. 1993. Characterization of soil conditions at the field level for soil specific
management. Geoderrna 60:57-72.

Robertson, G. P., J. R. Crum and B. G. Ellis. 1993. The spatial variability of soil
resources following long-term disturbance. Oecologia 962451-456.

Robertson, G. P., K. M. Klingensmith, M. J. Klug, E. A. Paul, J. R. Crum and B. G. Ellis.
1997. Soil resource, microbial activity, and primary production across an
agricultural ecosystem. Ecological Application 7(1): 158-170.

Robertson, G. P. and K. L. Gross. 1994. Assessing the heterogeneity of below -ground
resources: qualifying pattern and scale.p237 - 253 In M. M. Caldwell and R. W.

Pearcy (eds.) Plant exploitation of environmental heterogeneity. Academic Press,
New York, New York, USA.

Robertson, G. P., M. Huston, F. Evans, and J. Tiedje. 1988. Spatial variability in

successional plant community: patterns of nitrogen availability. Ecology 69:1517 -
1524.

128

SAS Institute. 2000. SAS/STAT user's guide. Version 8.1. SAS Inst., Cary, NC.

Schepers, J. S., D. D. Francis, M. Vigfl, and F. E. Below. 1992. Comparison of corn leaf
N concentration and chlorophyll meter readings. Comm. Soil Sci. and Plant Anal.
23:2173-2187.

Schepers, J. S., T. M. Blackmer, T. Shah, and N. Christensen. 1996. Remote sensing tools
for site-specific management. p. 315-319. In P.C. Robert et al. (eds.) Proceedings
of the Third International Conference on Precision Agriculture. Minneapolis, MN,
23-26 June 1996. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI.

Schueller, J. K. 1992. A review and integrating analysis of spatially-variable control of
crop production. Fertilizer Research 33:1 - 34.

Shayya, W. H., V. F.Bralts, and T. R. Olmsted 1990. General irrigation scheduling
package for micro-computers. Computer electronics in Agric. 5:197 - 212.

Stone, M. L., J. B. Solie, W. R. Raun, R. W. Whitney, S. L. Taylor, and J. D. Ringer.
1996. Use of spectral radiance for correcting in—season fertilizer nitrogen
deficiencies in winter wheat. Trans. ASAE 39:1623-1631.

Strahler, A. 1994. Measuring and modeling the bidirectional reflectance of plant
canopies. In 6lh international symposium on physical measurements and

signatures in remote sensing, Val d’Isere, France, 17 - 24 January 1994.

Suits, G. H. 1972. The calculation of the directional reflectance of a vegetation canopy.
Remote Sensing and the Environ. 2:117 - 125.

Tobler, W. 1970. A computer movie simulating urban growth in Detroit region.
Economic Geography 46(2):234 - 240.

Trangmar, B. B., R. S. Yost, and G. Uehara. 1985. Application of geostatistics to spatial
studies of soil properties. Advances in Agronomy 38:45-94.

Tsuji, G. Y., G. Uehara, and S. Balas (ed.). 1994. DSSAT v3. University of Hawaii,
Honolulu.

Tucker, C. T. 1979. Red and photographic infrared linear combination for monitoring
vegetation. Remote Sensing and the Environ. 8:127-150.

129

USDA-ERS. 1987. Farm-drainage in the United States: history, status, and prospects.
Misc Publ 1455. GA. Pavelis (ed.). USDA-ARS, Washington, DC.

Vachaud, G., A. P de Silans, P. Balabanis, and M. Vauclin. 1985. Temporal stability of

spatially measured soil water probability density function. Soil Science Soc. Am.
J. 49:822 - 828.

Webster, R. and M. A. Oliver. 1990. Statistical methods in soil and land resource survey.
Oxford Univ. Press, New York.

Webster, R. 1985. Quantitative spatial analysis of soil in the field. Advances in Soil Sci.
3:1-70.

Wesseling, J. 1974. Drainage and crop production. p.7-37 In: J. Van Schilfgaarde (ed.)
Drainage for Agriculture. Agron. Monograph17. ASA, Madison, WI.

Westgate, M. E., F. Forvella, D. C. Reicosky, and J. Sonsen. 1997. Rapid canopy closure
for maize production in the northern US corn belt: Radiation-use efficiency and
grain yield. Field Crops Research 49:249 - 258.

Wiegand, C. L. A. J. Richardson, D. E. Escobar, and A. H. Gerbermann. 1991.
Vegetation indices in crop assessments. Remote Sensing and the Environ. 35:105

-119. Elsevier Science Inc. New York, New York, 1991.

Wright J. L. 1982. New evapotranspiration crop coefficients. Journal of Irrigation and
Drainage Div. ASCE 108(IR2):57-74.

130

   

lrill)?'lljlflllilflliflllllflllil