.3:

.1
5 .’

:rtnuwn ,

. .. t I

z. x . 2:1 . t

:1. a 3.... r a
”p.11: 3 8. incl...h paoiiavlv

t: :- .
v IL. 51:9

_ .1133.
{o .3 I“
35?: .

3.,

aka... .
.. V ‘ ‘ Ila ‘8;
. , :3... .
, I" find“.
, , . Rhea;
..
‘ , a... a
‘ a .1.» «.3
. . . , s s. n... .x.i.uk.....d ,
, .. .13.... . a.
. . .,u.....:AI1 , It ..
4 hand“... , L’s-I1...»
rt; .finasu..m.? In»...
. are? 44.?
l,.!:...fiu..i. fl.
.3. awn

1

)3.
2.511.?!
I! 3.1!. I
I!

. 3..

ill.

(4.: vi

. .103 x
l ‘7‘ 11“
2-. ‘

3.1} .
{<24};
“but!

2.

 

I :3:
f. u...-
3!: .5412!
1...... 5...!
, 11.4.21.
2... -1125...
.ftt
..
;:!. .3. 1.. .o {a}???
w .329: ”Lina-.132... I 7..
l?! . n 3 A» .

.4 . i. u»
‘ u ,I. "gum r ‘ 1b“... . Jo a A.

 

LIBRARY
Michigan State
University

This is to certify that the
thesis entitled

Managing Nitrogen For Quality Carrot
Tops Using Remote Sensing

presented by

Jeanette Leah Makries

has been accepted towards fulfillment
of the requirements for the

Master of Science degree in Crop and Soil Sciences

 

 

2 . (Q/VZ/I @3748

Mr Professor’s Signature
Q76 cl /5 2? £05"

 

Date

MSU is an Affirmative Action/Equal Opportunity Institution

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_
2/05 c:/ClRC/DateDue.indd—p. 1 5

 

 

MANAGING NITROGEN FOR QUALITY CARROT
TOPS USING REMOTE SENSING

By

Jeanette Leah Makries

A THESIS

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

MASTER OF SCIENCE
Department of Crop and Soil Sciences

2005

Abstract

MANAGING NITROGEN FOR QUALITY CARROT
TOPS USING REMOTE SENSING

By

Jeanette Leah Makries

Michigan carrots are mechanically harvested; therefore quality tops are essential
for a successful harvest. During the process of harvesting, the tops are grabbed by
mechanical arms that lift the loosened carrots and carry them up a conveyor. Site specific
management using remote sensing may be useful in maintaining healthy tops without
over fertilizing the roots. A two year study was conducted at the Montcalm Experiment
Station and Sandyland Farms; Montcalm County. Four replications of four N treatments,
45, 90, 135, and 180 kg ha", were arranged in a randomized complete block design at all
locations. N content of soil, petioles, and harvested plants was compared to individual
reflectance using narrow wavebands centered at 460, 510, 560, 610, 660, 710, 760, and
810 nm and selected vegetation indices, NDVI, SAVI, TSAVI, and GNDVI.

Visible wavebands centered at 560, 610, and 710 were the earliest and most
consistent to correlate with treatments, petiole-N, and selected harvest measurements
where r2 was as high as 0.90. NIR reflectance at 760 and 810 nm was weakly correlated
with plant N status where canopy coverage was affected by variables other than N
treatments and when the canopy reached full coverage affecting the sensitivity of indices
to N status. GNDVI out performed the other indices; soil adjustment did not enhance the

usefulness of the indices.

This Thesis is
Dedicated
T 0 my husband Jim
Who patiently supported my endeavors, and shared
Many hours with this project, and

T 0 my sons Jeffery and AJ
Who have also shared many hours with my studies
And who have taught me as much about courage and
Determination as I have taught them

And to my parents
Ernie and Eleanor, who taught me to will, to help myself
And persevere, and who have been ever
Vigilant in their watch over me

And to my brothers and sisters
Whose continuous encouragement
Helped me through
This process

iii

ACKNOWLEDGMENTS

This journey of discovery into research is not traveled alone, and for this I am
truly gratefiil. Without the expertise, experience, knowledge, and questions of others
one’s grasp of knowledge may be incomplete.

I would like to thank Dr. Darryl Wamcke who has been an ever patient mentor,
not only during my graduate studies but also while an undergraduate. He has provided
me with endless opportunities to learn, teaching me to appreciate science and to dig
beyond the surface, as well as research protocol and field technique. I would also like to
thank my remaining committee members Dr. David Lusch, Dr. Mary Hausbeck, and Dr.
William Northcott, who have offered their time, encouragement, and resources so I could
succeed in this endeavor.

I would also like to thank Dr. Francis Pierce, who originally convinced me to
pursue these graduate studies, and provided me with opportunities for undergraduate
work experience in this area.

Of course, location is everything, as they say. I would like to thank, Tim and
Todd Young of Sandyland Farms who graciously hosted my research plots in their
production fields. In addition, I am grateful to Dick Crawford, the farm manager at the
Montcalm Experiment Station, who prepared my plots, and applied the fungicides and
herbicides.

I cannot say enough about the three research assistants I have worked with both in
the laboratory and in the field, Gary Zehr, Cal Bricker, and Brian Long. They have been

good friends and have made many field hours both enjoyable and highly productive.

iv

Their expertise made my sampling more proficient. John Dahl, Vickie Smith and Rosie
Cabrera, of the Soil and Plant Nutrient Lab, have always been available to teach me lab
procedures and answer my endless questions; thank you.

I am truly grateful to the student technicians who spent many hours in the field
and in the laboratory with me doing many repetitious and tedious tasks characteristic of
research. Non of these students were majoring in Crop and Soil Science, and yet, they
were diligent, capable, and I really relied on the skills they learned. They are Melissa
Meyers, Leanne Parry, Jeremy Zehr, Mike Kozma, and my son, Jeffery Makries. Jeff
became my right hand and often kept me focused.

Volunteers set aside their own activities to help another. I would like to recognize
their efforts and say, thank you: to my son, AJ Makries and Elizabeth Webster for
volunteering for field work; Ryan Bounds, for scouting my fields for plant disease; Sarah
Marshall and Tracy Jenkins, for helping with harvest. My sister Pamela set aside her
own studies to help edit my thesis.

The completion of research projects would be difficult if it were not for the
funding that is provided through the various entities. Most of the funding for this project
was provided by the USDA Carrot Ramp Grant, with the remainder from the MSU

Agriculture Experiment Station.

Thank you everyone.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................. viii
CHAPTERI

EFFECT OF NITROGEN APPLICATIONS ON SOIL NO3' N
PLANT NITROGEN STATUS, AND BIOMASS

PRODUCTION IN CARROT ................................................................... 10
Introduction ............................................................................... 10
Literature Review ........................................................................ 11
Materials and Methods .................................................................. 18

Experimental Sites ............................................................. 18
Plot Design and Management Protocol ...................................... 19
Agronomic Measurements and Sample Analysis .......................... 22
Results and Discussion ................................................................. 25
Soil N Availability and Plant Uptake ....................................... 25
N Treatments vs Petiole Response ........................................... 29

Treatments vs N Uptake in Harvested Dry Matter
AndYreld 33

Conclusions .............................................................................. 38
References ............................................................................... 40
CHAPTER II

SPECTRAL MEASUREMENTS OF THE CARROT CANOPY AS
RELATED TO NITROGEN STATUS OF THE CROP 44

Introduction .............................................................................. 44
Literature Review ........................................................................ 47
Materials and Methods ................................................................. 61
Experimental Sites, Plot Design, Management Protocol,
And Agronomic Sampling... . . .. 61
Reflectance and Agronomic Measurements ................................ 62
Analysis of the Data ........................................................... 65
Results and Discussion ................................................................. 66
N Status of Carrot Canopy as Result of Soil-N Availability. 66
Individual Reflectance ..................................................... 66
Selected Indices ............................................................. 84
In Season N Management: Reflectance vs Total N and
Sap Nitrate ...................................................................... 93

vi

CHAPTER II (cont’d)

End of Season: Reflectance vs Selected Harvest Parameters ............ 99

Individual Reflectance ...................................................... 100

Selected Indices ............................................................. 110

Conclusions .............................................................................. 1 18

References ............................................................................... 122
CHAPTER III

SAVI DETERMINATION IN CARROTS: COMPARING CONSTANT AND
DYNAMIC SOIL ADJUSTMENT

FACTORS ......................................................................................... 127
Introduction .............................................................................. 127
Literature Review ....................................................................... 129
Materials and Methods ................................................................. 136

Experimental Sites, Plot Design, Management Protocol,

And Agronomic Sampling .................................................... 136

Reflectance and Agronomic Measurements ................................ 137

Image Processing ............................................................... 138

The fc Calculation ............................................................ 139
Results and Discussion ................................................................. 139
Conclusions ............................................................................ 149
References ............................................................................... 1 52

Vii

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

LIST OF TABLES

Chapter I

N fertilizer (urea) split applications broadcast on the indicated dates,
listed by treatment number at the four field locations .......................... 21

Rainfall recorded and irrigation amounts delivered at the Montcahn
Experiment Station during 2001 and 2002. Irrigation at Sandyland was
estimated at 3.0 to 3.8 cm per week ............................................... 23

Nitrogen activity in the soil including initial residual levels in the top 30

cm, applications, and ending residual levels compared to N uptake by

plants and carrot root yield. Mean separation of plant uptake of N as

compared to N treatments. Linear regression analysis as used to compare

plant uptake ofN with available N 26

Comparison of beginning of season residual N with end of season
residual N in the soil derived with KCl extractant .............................. 28

Mean separation between treatments of % N in carrot petioles sampled on
the indicated dates. Average yield per N treatment, fiom each specified
location, listed for comparison to % N in petioles .............................. 30

Linear regression results of petiole sap NO3‘ - N vs Total N (TKN) of

dried petioles sampled on various dates during the 2001 season. N

available was as of the petiole sampling date. Values shown were

averaged by treatment .............................................................. 31

Linear regression results of petiole sap NO3' - N vs Total N (TKN) of

dried petioles sampled on various dates during the 2002 season. N

available was as of the petiole sampling date. Values shown were

averaged by treatment .............................................................. 32

Mean N uptake in tops and roots compared to dry matter and root yield

using regression analysis. Mean available N compared to root:shoot ratio

using regression analysis. Significance of treatment differences in dry

matter, root:shoot ratio, and root yield determined using analysis of

variance .............................................................................. 34

2001 and 2002 carrot root yield by grade and percent of the total ........... 37

viii

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 11

Table 12

Chapter 11

Comparison of measurements taken August 20, 2001 by the SE590
spectroradiometer and CropScan multispectral radiometer. The canopy
represented grth at 104 days after planting at the Montcalm Experiment
Station and approximately 127 days after planting at Sandyland. . . . . . . .

Reflectance measurement protocol specific to individual field locations...

Linear regression coefficients of reflectance at individual wavelengths vs
N; where N is treatment, applied N, or available N at the Montcahn
Experiment Station, 2001, Diamond Cut variety .................................

Mean reflectance measurements as influenced by treatment or applied N
at the Montcalm Experiment Station, 2001, Diamond Cut variety. . .

Linear regression coefficients of reflectance at individual wavelengths vs
N; where N is treatment, applied N, or available N at Sandyland, 2001,
Asgrow B1 and Prime Cut 59 varieties ...........................................

Mean reflectance measurements as influenced by treatment or applied N
at Sandyland, 2001, Asgrow B1 and Prime Cut 59 varieties ..................

Linear regression coefficients of reflectance at individual wavelengths vs
N; where N is treatment, applied N, or available N at the Montcalm
Experiment Station, 2002, Diamond Cut variety ................................

Mean reflectance measurements as influenced by treatment or applied N
at Montcalm Experiment Station, 2002, Diamond Cut variety ...............

Linear regression coefficients of reflectance at individual wavelengths vs
N; where N is treatment, applied N, or available N at the Montcahn
Experiment Station, 2002, Goliath variety .......................................

Mean reflectance measurements as influenced by treatment or applied N
at the Montcalm Experiment Station, 2002, Goliath variety ..................

Linear regression coefficients of reflectance at individual wavelengths vs
N; where N is treatment, applied N, or available N at Sandyland, 2002,
Sugar Snax variety ..................................................................

Mean reflectance measurements as influenced by treatment or applied N
at Sandyland, 2002, Sugar Snax.variety ..........................................

ix

63

65

68

7O

71

72

74

76

77

79

80

81

Table 13

Table 14

Table 15

Table 16

Table 17

Table 18

Table 19

Table 20

Table 21

Table 22

Table 23

Chapter II (cont’d)

Linear regression coefficients of selected indices vs N; where N is
treatment, applied N, or available N at the Montcalm Experiment Station,
2001, Diamond Cut varrety

Linear regression coefficients of selected indices vs N; where Index = mN
+ b through 7/26 and N is treatment, applied N, or available N at
Sandyland, 2001, Asgrow B1 and Prime Cut 59 varieties. Beginning 8/2
Index = mN + mN2 + b .............................................................

Mean reflectance measurements as influenced by treatment or applied N
at Sandyand, 2001, Asgrow BI and Prime Cut 59 varieties ..................

Linear regression coefficients of selected indices vs N; where N is
treatment, applied N, or available N at the Montcalm Experiment Station,
2002, Diamond Cut varrety

Mean reflectance measurements as influenced by treatment or applied N
at the Montcalm Experiment Station, 2002, Diamond Cut variety ...........

Linear regression coefficients of selected indices vs N; where N is
treatment, applied N, or available N at Sandyland, 2002, Sugar Snax
variety .................................................................................

Mean reflectance measurements as influenced by treatment or applied N
at Sandyland, 2002, Sugar Snax variety ..........................................

Linear regression coefficients of reflectance at individual wavelengths vs
petiole-N; where petiole-N = total N content or petiole sap NO3' at the
Montcahn Experiment Station, 2002, Goliath variety ..........................

Linear regression coefficients of reflectance at individual wavelengths vs
petiole-N; where petiole-N = total N content or petiole sap NO3' at the
Montcalm Experiment Station, 2001, Diamond Cut variety ...................

Linear regression coefficients of reflectance at individual wavelengths vs
petiole-N; where petiole-N = total N content or petiole sap NO3' at the
Montcalm Experiment Station, 2002, Diamond Cut variety ......................

Linear regression coefficients of reflectance at individual wavelengths vs
petiole-N; where petiole-N = total N content or petiole sap NO3' at
Sandyland, 2001, Asgrow B1 and Prime Cut 59 varieties .....................

85

87

88

89

89

90

92

94

95

96

97

Table 24

Table 25

Table 26

Table 27

Table 28

Table 29

Table 30

Table 31

Table 32

Table 33

Table 34

Table 35

Table 36

Chapter II (cont’d)

Linear regression coefficients of reflectance at individual wavelengths vs
petiole-N; where petiole-N = total N content or petiole Sap NO3' at
Sandyland, 2002, Sugar Snax variety .............................................

Linear regression coefficients of reflectance at individual wavelengths vs
% N in harvested tops from selected locations .................................

Linear regression coefficients of reflectance at individual wavelengths vs
% N in harvested roots from selected locations .................................

Linear regression coefficients of reflectance at individual wavelengths vs
N uptake in harvested tops from selected locations ............................

Linear regression coefficients of reflectance at individual wavelengths vs
top biomass of harvested tops from selected locations .........................

Linear regression coefficients of reflectance at individual wavelengths vs
N uptake in harvested roots from selected locations ...........................

Linear regression coefficients of reflectance at individual wavelengths vs
root biomass of harvested roots from selected locations ......................

Linear regression coefficients of reflectance at individual wavelengths vs

root:shoot ratio of biomass at harvest fi'om selected locations. . . . . . . . .

Linear regression coefficients of indices calculated from reflectance at
individual wavelengths vs % N in harvested roots from selected
locations ...............................................................................

Linear regression coefficients of indices calculated from reflectance at
individual wavelengths vs yield as Mg ha'1 fresh weight from selected
locations ..............................................................................

Linear regression coefficients of indices calculated from reflectance at
individual wavelengths vs root biomass of harvested roots from selected
locations ..............................................................................

Linear regression coefficients of indices calculated from reflectance at
individual wavelengths vs N uptake in harvested tops from selected
locations ..............................................................................
Linear regression coefficients of indices calculated from reflectance at
individual wavelengths vs N uptake in harvested roots from selected
locations ..............................................................................

xi

98

100

102

104

105

107

108

109

111

112

113

113

115

Table 37

Table 38

Table 1

Table 2

Table 3

Table 4

Chapter II (cont’d)

Linear regression coefficients of indices calculated fi'om reflectance at
individual wavelengths vs top biomass of harvested tops from selected
locations .............................................................................. 1 16

Linear regression coefficients of indices calculated from reflectance at
individual wavelengths vs root:shoot ratio of biomass at harvest from
selected locations ..................................................................... 1 1 7

Chapter III

2001 Regression analysis of Percent Vegetation Coverage (PVC) vs
Calculated fc (PVC = a + b fc +cTreatment) where a is the intercept and

b and c are regression coefficients. Treatment did not significantly

influence correlation of PVC with fc at p < 0.05 ............................... 141

2002 Regression analysis of Percent Vegetation Coverage (PVC) vs.
Calculated fc (PVC = a + b fc + cTreatrnent) where a is the intercept and

b and c are regression coefficients. Treatment significantly influenced
correlation of PVC with fc on the dates indicated at p < 0.05. . 142

Mean Percent Vegetation Coverage as influenced by fc and treatment
differences on selected dates .................................................... 144

Results of regression analysis of SAVI comparing L = 0.5 and
L = (1- fc) ............................................................................ 150

xii

LIST OF FIGURES

Chapter III

Figure 1 Results of SAVIfc and SAVI L = 0.5 for the 2001 field season at the
Montcalm Experiment Station and Sandyland locations. . .. 146

Figure 2 Results of SAVI f c and SAVI L = 0.5 for the Diamond Cut and Goliath
varieties at the Montcalm Experiment Station for the 2002 field season. . .. 147

Figure 3 Results of SAVIfc and SAVI L = 0.5 at Sandyland for the 2002 field
season ................................................................................. 148

Figure 4 Example of the difference between SAVI f c with and without the (1+L)

multiplier. The Experiment Station 2002, Diamond Cut data is shown
here, the other locations exhibited similar differences ......................... 149

xiii

Introduction

Managing Nitrogen for Quality Carrot Tops Using Remote Sensing

Agricultural studies involving remote sensing are driven by the need for precision
agriculture for the purpose of improving crop performance and environmental quality
(Pierce and Nowak, 1999; Kutcher et al., 2005). As defined by Pierce and Nowak
(1999), precision agriculture is the application of technology and principles to manage
spatial and temporal variability associated with all aspects of agricultural production.
Soil and crops are managed by soilscapes, management zones, and the management of
non-crop periods (Pierce and Nowak, 1999; Lauzon et al., 2005; Allrnaras et al., 1998).
The fact that the soil supply of nutrients and plant demand, and nutrient loss through
leaching, erosion, and runoff vary in space and time indicates there are significant
opportunities for precision management of soil fertility (Pierce and Nowak, 1999).

Precision agriculture must fit the needs and capabilities of the farmer (Pierce and
Nowak, 1999). It will only be economically beneficial if questions pertaining to type of
variability present and potential management opportunities are addressed (O’Halloran,
2005), because the benefit is not derived from the technology itself, but from the
management decisions resulting from its use (Pierce and Nowak, 1999). Precision
management of crop production must be an improvement over whole field management
(Pierce and Nowak, 1999). If the parameter of interest is homogenous or random, then
the cost does not warrant its use; but as the degree of spatial and temporal dependence
increases so do the prospects for precision management (Pierce and Nowak, 1999). It is

necessary to report simultaneously on both spatial and temporal variables. Some spatial

patterns develop over time and the cause and effect may exist in time but not in space;
even though the degree of difficulty in achieving precision management increases with
temporal variance (Pierce and Nowak, 1999).

Soil productivity and spatial variability in crop growth and yield have always
been realities of farming (Pierce et al., 1995), and vary across fields as the result of the
interaction of topography, soil properties, and management practices (Kravchenko et al.,
2005). Water distribution is also a function of this dynamic trio, and along with weather
conditions, varies from year to year (Kravchenko et al., 2005). Therefore, crop
susceptibility to erosion, and surface and groundwater vulnerability to pollution exist in
specific spatial patterns (Nowak and Korsching, 1998; Allrnaras et al., 1998).
Assessment of variability is the first step in precision management (Pierce and Nowak,
1999). Lauzon et a1. (2005) found soil test results reasonably correlated with
topographic-position variables from site to site. Pierce et a1. (1995) found on the average,
each of three fields in southern Michigan showed optimum pH and medium to high soil
test results. Soil fertility, however, generally ranged fiom deficient to excessive for most
parameters measured. More specifically, Kutcher et a1. (2005) cited several studies that
found crop productivity varying between slope position as a result of differing conditions
particularly moisture and fertility. The soil physical properties or landscape undulation
may be more important than fertility in explaining yield variation; particularly in their
effect on water availability (Pierce et al., 1995). Soil moisture, as it changes across an
undulating landscape, affects the potential for N mineralization, immobilization,
denitrification, and leaching (Kutcher et al., 2005). There are many researchers whose

studies are dedicated to the identification of controllable variability, and determination of

the intensity of measurement at which profitable precision can be achieved. Among them
are authors featured in the following chapters including Schepers et a1. (1992, 1996),
Blackrner et a1. (1994, 1996a, 1996b), and Osborne et a1. (2002).

Site-specific management may improve economic returns and reduce
environmental contamination (O’Halloran, 2005), because it involves the variable
management of soils, crops, and pests according to conditions within a field (Pierce et al.,
1995). It provides farmers the potential to apply the exact requirement of nutrients at
each given location (Lauzon et al., 2005; Larson et al., 1998). In the previously cited
study in southern Michigan, the cost of over fertilizing a corn crop when fertilizer was
applied uniformly was only a few dollars per hectare; however, the estimated yield loss
from under fertilization could be greater than 2 Mg ha'1 (Pierce et al., 1995). Profitable
site-specific management includes the ability to accurately locate one’s self in the field;
vary input; have a reasonable understanding of how the nutrient or crop response will
vary across the field; and, the level of variability must be enough to make the investment
worthwhile and it must be manageable (O’Halloran, 2005).

Geostatistics has been adapted for use in site-specific management to assess
spatial variability. Spatial estimations are made using points and interpolation, but one
must decide on the appropriate scale (Pierce and Nowak, 1999; Lauzon et al., 2005). The
sample point unit, design, and map accuracy should result in a quality that has value for
management decisions and be appropriate for available equipment (Pierce and Nowak,
1999). Sampling intensity should be driven by field characteristics rather than cost, and
the distance between samples should be sufficiently small so that resulting data points are

spatially related (Pierce and Nowak, 1999, Lauzon et al., 2005). Researchers at

Oklahoma State University found that the field element size is seldom larger than 1 m2
(Dept. of Plant and Soil Sciences Oklahoma State University, 2004); however, even if
sampling intensity is reduced to a 30 m grid, there is economic concern (Lauzon et al.,
2005)

Precision agriculture includes five general groups of technology: computers, GPS,
GIS, sensors, and variable rate control (Pierce and Nowak, 1999). Sensors may provide
the cost effective solution by which sampling intensity can be driven by field
characteristics. They have a fixed initial cost that actually decreases as sampling
intensifies (Pierce and Nowak, 1999). This means that, the sampling scheme can be
determined by the sensors capability and the nature of sampled parameters independent
of cost or difficulty, in contrast to traditional sampling (Pierce and Nowak, 1999).
Remote sensing of a growing crop will reveal stresses that impact the crop during the
growing season, and intervention strategies can be applied to meet the demand during the
rapid uptake phase of growth (Pierce and Nowak, 1999). It is even more applicable
where the temporal component of spatial variability is medium to high as with N
management versus P, K and pH where temporal variability is low (Pierce and Nowak,
1999)

This study was focused on the use of sensors, specifically above canopy proximal
sensing of the spectral reflectance, one form of remote sensing. According to Bronson et
a1. (2005), estimation of crop N using proximal sensing has gained strong interest.
Height above the soil varies from study to study and examples range from directly over
the row (Dept. of Plant and Soil Sciences Oklahoma State University, 2004), to several

meters high (Bronson et al., 2005; Ma et al., 2001; Osborne et al., 2002). Remote sensing

holds real promise for precision agriculture because of its potential for monitoring spatial
variability over time at high resolution (Pierce and Nowak, 1999). Using sensing to
discover deficiencies may be better adapted to precision management than those that rely
on soil sampling (Pierce and Nowak, 1999), or may supplement preplant soil test data
(Bronson et al., 2005).

The amount of electromagnetic energy reflected or emitted from an object varies
by wavelength as determined by the object’s physical and chemical structure (Pierce and
Nowak, 1999). One sampling can measure many plants and monitor many conditions
(Blackrner et al., 1996). A single measurement can be used to construct different images
of a target using a single waveband or a combination of wavebands depending on the
parameter of interest (Pierce and Nowak, 1999). For example, spectra at 550 to 560 nm
and at 660 nm are associated with plant chlorophyll content. Wavebands at
approximately 900 nm are absorbed by iron oxide and may be associated with soil
characteristics (Fontes and Carvalho, Jr., 2005). Wavebands at 960, 1200, 1420, 1920,
and 2620 nm are water absorption bands (Lillesand and Kiefer, 2000; DP. Lusch,
personal communication, 2001). An image developed from these bands may reveal the
water content of the canopy in the field, and over time track the temporal trend of
seasonal moisture distribution. Plant litter can be discriminated fi'om soil using
wavebands at 1730, 2100, and 2300 nm which are primarily associated with N, cellulose,
and lignin, respectively (Daughtry et al., 2005). Plant residue lacks the spectral response
of green vegetation, but still retains alcoholic-OH groups such as sugar, starch and
cellulose which are absent in soils. Although researchers have studied the relationship of

spectral measurements to plant physiological and biochemical aspects for some time now,

new and more complex indices are under design using more precise hyperspectral
radiometers. Such physiological and biochemical aspects include chlorophyll,
carotenoids, and water content, as well as cellulose, lignin and dry matter (Zarco—Tejada
etal., 2005).

One approach to precision N management is to develop site-specific intervention
strategies based on crop monitoring of N status using remote sensing (Pierce and Nowak,
1999). There are many representative studies exemplifying this approach. Some of those
studies are summarized in the next chapters. The first step is to monitor N concentration
by measuring plant or canopy reflectance of light. The second step is to estimate N
fertilizer requirements using a relationship established between reflectance and N content
(a reference strip may be used as a standard). The final step is to fertilize the crop to
optimum N content (Pierce and Nowak, 1999).

To date, the bulk of agricultural research using remote sensing has concentrated
on agronomic crops. Since they are mechanized and precision agriculture really gained
momentum with the development of the yield monitor (Pierce and Nowak, 1999), it
follows that subsequent advances in agricultural remote sensing should gain popularity
first with mechanized crops. The field size of agronomic crops is generally large enough
to necessitate the ability to sample on a large scale. However, there are many vegetable
crops that may also benefit from remote sensing and precision agricultural management,
especially those that are mechanized such as carrot.

The following chapters present the results of a two year study in the N
management of carrot tops using remote sensing. Quality tops, even though they are not

the income producing part of the plant, are essential for a successful carrot harvest.

Michigan carrots are mechanically harvested. During the process of harvesting, the tops
are grabbed by mechanical arms that uproot the loosened carrots and carry them up a
conveyor. Weak tops will break off and leave the carrots in the ground resulting in lost
yield. Site specific management using remote sensing to monitor nutrient status may

provide the means to maintain healthy tops without over fertilizing the roots.

References

Alhnaras, R.R., D.E.Wilkins, O.C. Burnside, and DJ. Mulla.1998. Agricultural
technology and adoption of conservation practices. p. 99-158. In F.J. Pierce and

W.W. Frye (ed.) Advances in Soil and Water Conversation. Sleeping Bear Press,
Chelsea, MI.

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

Bronson, K.F., J .D. Booker, J.W. Keeling, R.K. Boman, T.A. Wheeler, R.J. Lascano, and
R.L. Nichols. 2005. Cotton canopy reflectance at landscape scale as affected by
nitrogen fertilization. Agron. J. 97:654-660.

Chang, K.-W., Y. Shen, and J.-C. Lo. 2005. Predicting rice yield using canopy
reflectance measured at booting stage. Agron. J. 97:872-878.

Daughtry, C.S.T., E.R. Hunt, Jr., P.C. Doraiswarny, and J .E. McMurtrey HI. 2005.
Remote sensing the spatial distribution of crop residues. Agron. J. 97:864-871.

Dept. of Plant and Soil Sciences Oklahoma State University. 2004. Indirect measures of
plant nutrients. [Online]. Available at

http://www.dasnr.okstate.edu/nitrogen_use/indirect measures_of_plant_nutri.htm
(verified 23 Mar. 2004.)

Fontes, M.P.F., I.A. Carvalho, Jr. 2005. Color attributes and mineralogical
characteristics, evaluated by radiometry, of highly weathered tropical soils. Soil
Sci. Soc. Am. J 69:1162-1172. .

Kravchenko, A.N., G.P. Robertson, K.D. Thelen, and RR. Harwood. 2005. Management,
topographical, and weather effects on spatial variability of crop grain yields.
Agron. J. 97:514-523.

Kutcher, H.R., S.S. Malhi, and KS. Gill. 2005. Topography and management of nitrogen
and fimgicide affects diseases and productivity of canola. Agron. J. 97:533-541.

Larson, W.E., M.J. Mausbach, B.L. Schmidt, and P. Crosson. 1998. Policy and
government programs in soil and water conservation. p. 195-218. In F .J . Pierce
and W.W. Frye (ed.) Advances in Soil and Water Conversation. Sleeping Bear
Press, Chelsea, MI.

Lauzon, J.D., I. P. O’Halloran, D.J. Fallow, A. P. von Bertoldi, and D. Aspinall. 2005.
Spatial variability of soil test phosphorus, potassium, and pH of Ontario soils.
Agron. J. 97:524-532.

Lillesand, T.M., and R.W. Kiefer. 2000. Remote sensing and image interpretation. 4th ed.
John Wiley & Sons, New York.

Ma, B.L., L.M. Dwyer, C.Costa, E.R. Cober, and M.J. Monison. 2001. Early prediction
of soybean yield from canopy reflectance measurements. Agron. J. 93:1227-1234.

Nowak, P., and RF. Korsching. 1998. The human dimension of soil and water
conservation: a historical and methodological perspective. p. 159-194. In F.J.
Pierce and W.W. Frye (ed.) Advances in Soil and Water Conversation. Sleeping
Bear Press, Chelsea, MI.

O’Halloran, I. Spatial variability and nutrient management. OCPA. [Online]. Available at
http://ontariocorn.org/ocpmag/aprar12.html (vertified 13 Sept. 2005.)

Osborne, S.L., J.S. Schepers, D.D. Francis, and MR. Schlemmer. 2002. Detection of
phosphorus and nitrogen deficiencies in corn using spectral radiance
measurements. Agron. J. 94:1215-1221.

Pierce, F.J., D.D. Warncke, and M.W. Everett. 1995. Yield and nutrient variability in
glacial soils of Michigan. In “Proceedings of the Second Intemational Conference
on Site Specific Management of Agricultural Systems, Bloomington/Minneapolis,
MN, 27-30 March 1994” (RC. Robert, R. H. Rust, and WE. Larson, Eds.), ASA
Miscellaneous Publications, pp. 133-151. ASA,CSSA, and SSSA, Madison, WI.

Pierce, F .J ., and P. Nowak. 1999. Aspects of precision agriculture. Adv. in Agron. 67:1-
85.

Zarco-Tejada, P.J., S.L. Ustin, and ML. Whiting. 2005. Temporal and spatial
relationships between within-field yield variability in cotton and high-spatial
hyperspectral remote sensing imagery. Agron J. 97:641-653.

Chapter I

Effect of Nitrogen Applications on Soil NO3-, Plant Nitrogen
Status, and Biomass Production in Carrot

Introduction

Carrot (Daucus carota L.) is harvested for market every month of the year in the
United States (McGiffin et al., 1997; Mills, 2001). Worldwide, carrot is a minor crop;
18.5 million tons of carrots were produced in 1998 on 794,000 ha (Suojala, 2000). The
National Agricultural Statistics Service (USDA NASS, 2002) reported that in 2001, the
United States produced 31.3 million cwt (1.4 million Mg) of carrot on over 100,000 acres
(41,000 ha) worth $545 million out of more than $10 billion in principal commercial
vegetable production. The carrot share of the United States market was just over 5%
(USDA NASS, 2002). California is the leading US producer of carrot where it is grown
year-round. Other states that are ranking producers are Texas, Georgia, Washington,
Michigan, and Wisconsin.

In 2001, Michigan ranked third in the US for fresh market production and fifth for
processing carrot (USDA NASS, 2002). Michigan carrot production is concentrated in
five counties: Muskegon, Newaygo, and Oceana growers produce carrots predominantly
for processing, of which one-third is used in baby food (Carrot Ramp, 2003); fresh
market carrot production is concentrated in Montcalm and Lapeer counties.

In addition to its commercial value, carrot provides an economical source for
seven to eight times the recommended daily allowance of vitamin C. It is also high in

fiber, potassium and vitamins A, B, D, and E. Carrot contains calcium, is rich in mineral

10

salts, high in beta-carotene, and contains smaller amounts of essential oils, carbohydrates,
and nitrogenous compounds. Carrot is well known for its sweetening, antianaemic,
healing, diuretic, remineralizing, and sedative properties (MDA, 2002).

Finally, carrot may help control excess soil NO3'. The deep fibrous root system
can effectively lower excessive accumulations that have leached deeper into the soil
profile (Warncke, 1996; White and Strandberg, 1978).

The objective of this study was to evaluate the field response of carrot to nitrogen

treatments at four locations over a two-year period (2001, 2002).

Literature Review

Carrot (Daucus carota L. ), shares the same family (Apiaceae) as celery, fennel,
parsnip, and parsley. It emerges from seed with two strap-like cotyledons followed by
rosettes of doubly compounded leaves rising from the crown. A taproot develops fiom
the hypocotyl (Mills, 2001), and the hypocotyl will eventually form about 2.54 cm of the
upper part of the storage root (Suojala, 2000). During the first 24 days after emergence,
early and rapid growth of the taproot in the temperature range of 16-24°C is striking, with
little secondary or tertiary root development and no visible secondary thickening (White
and Strandberg, 1978). Secondary thickening begins with initiation of the secondary
cambium. This enlarging causes cells of the cortex and endoderrnis to rupture, at which
point the orange color appears (Suojala, 2000). The size of individual roots increases
with maturity and is affected by plant population, which is a function of the crop end use.
For production purposes, root uniformity is a common demand of processors (Suoj ala,

2000). Although early top growth is slower than root growth, the tops generally produce

11

greater biomass (White and Strandberg, 1978). The length of the growing season varies
with end use and available markets, variety, geographical location, and time of planting
(Hipp, 1978). Michigan carrots are harvested 80 to 180 days after planting (USDA,

1999, Zandstra et al., 1986).

Carrot is a cool season crop that demands specific growing conditions for
successful commercial production and effective use of N applications. The crop is grown
best at 60 to 70° F (16 to 21°C) (G012 and Aakre, 1993; Mills, 2001; Fritz et al., 1998;
MDA, 2000; Zandstra et al., 1986). While young seedlings can withstand mild frost,
high temperatures can result in damage (USDA, 1999). High temperatures cause greater
respiration in the leaves reducing color development and sugar accumulation as the root
matures, which results in a strong unpleasant flavor (Mills, 2001; Zandstra et al., 1986).
Alaska, with its cool climate, boasts of a high quality carrot due to greater sugar
accumulation in the roots (Epps, 1970). It is this cool season requirement that allows
Florida to use carrot as a winter crop (Hochmuth et al., 1999; McCollum et al., 1986) and
Midwestern states such as Michigan to plant in early spring (USDA, 1999). California,
with its varied climate zones, high desert, southern desert, central coast, and central

valleys, grows carrot continuously (McGiffin et al., 1997).

For optimal carrot production, the soil should be warm, loose, deep, and well
drained (Epps, 1977; MDA, 2002; Zandstra et al., 1986). It is generally agreed that carrot
grows best on coarse mineral or organic soils (Fritz et al., 1998; Hanlon et al., 2002;
Mills, 2001; Hochmuth, 1999; Epps, 1970; G012 and Dwight, 1993; McGiffin et al.,
1997). Heavy soils are less desirable even if uniform moisture is maintained, because

carrot is very sensitive to soil compaction (Golz, 1993). Compact, cold, and poorly

12

drained soil causes crooked forked roots (Epps, 1970; MDA, 2002; USDA, 1999). The
ideal mineral soil is silt loam, according to McGiffin et a1. (1997), because it has the best
combination of water holding capacity and drainage. Other soil recommendations range
from loamy sand to sandy loam (Epps, 1970; Hipp, 1978; Hochmuth, 1999; MDA, 2002;
Sanderson, 1997; Wamcke 1996). In Michigan, carrot is primarily grown in deep, well-
drained muck with a pH range of 5.5 to 5.8, and in mineral soils with a pH range of 6.2 to

6.8 (USDA, 1999; Zandstra et al.,1986).

Carrot is directly seeded into the soil; transplanting disturbs the taproot and
prohibits proper hypocotyl development. Seeding population varies depending on the
purpose for which it is grown, and the planting density is selected to provide the greatest
number of carrots of the size required for the specific market (McCollum et al., 1986).
Varieties used in production of “baby carrots” are planted at the highest population, 80 to
100 seeds per bed foot [with 20 to 40 in. (51 to 102 cm) wide beds] (Fritz et al., 1998).
Fresh market varieties are planted 20 to 30 seeds per row foot and processing varieties
are generally planted 10 to 20 seeds per row foot (Mills, 2001; Fritz et al., 1998, G012 and
Aakre, 1993). Carrot may be sown in beds or nonbedded. Total yield, root size, and
uniformity at harvest are a function of stand establishment (Finch and Savage, 1987). In
Michigan, growers commonly interseed barley or other small grains with carrot to protect

emerging plants from wind damage (Zandstra and Warncke, 1993).

A uniform water supply, as well as good soil fertility, is critical for the
development of good color and formation of uniform root size (McGiffin et al., 1997;
Mills, 2001). The University of Alaska (Epps, 1997), the University of California (Fritz

et al., 1998), and Michigan State University recommend 2.5 to 3.8 cm of water per week,

13

with seasonal totals of 25.4 to 35.6 cm in Michigan (Zandstra et al., 1986) and 35.6 to
38.1 cm in California (Fritz et al., 1998). The soil should be soaked completely to avoid
separation between sub-soil and surface-soil moisture that may cause differential growth
and cracking (Zandstra etal., 1986). Irregular watering, such as significant wet/dry
patterns, may result in root splitting (McGiffin et al., 1997), or rough, lumpy carrots with
obvious grth rings (Fritz et al., 1998). Carrots are most sensitive to moisture stress
during seed germination and root enlargement promoting small, woody and poorly
flavored roots with grth cracks (Fritz et al., 1998). Excessive water discourages good
color and encourages soil borne diseases (Kelly, 1998; McGiffin et al., 1997).

Carrot is especially susceptible to weed competition because emergence and early
top growth is slow (White and Strandberg, 1978; MSU and MDA, 2000). Since
mechanical cultivation may injure roots, chemical herbicides are recommended (Epps,
1977; MSU and MDA, 2000). Linuron has been shown to be most effective (Epps, 1997;
Bell, 2000), and preplant applications may prove particularly successful in controlling
weeds (Kelly, 1998). Altemaria (Alternaria dauci) and Cercospora (Cercospora
carotae), both foliar blights, can cause serious damage to top quality. In some areas,
including Michigan, where carrot is harvested using the tops to lift the plant out of the
soil, weakened tops result in yield reduction. Pressure from these diseases cause
fungicides to be the primary pesticide applied to carrot.

Nitrogen (N) is a component of chlorophyll, all proteins and many other
compounds in plants, and is an essential nutrient for plant health (Carrot Ramp, 2003;
Marshner, 1998). It is the most limiting nutrient for crop production because of the large

need for it by plants and the limited ability of soils to supply available N. Only about 1%

l4

or less of the total N in soils is available to plants and microorganisms as N03' or
exchangeable NHX, which is rapidly consumed or susceptible to leaching. It must be
replaced by fertilizer applications or by mineralization (Foth and Ellis, 1997). The
following studies describe past experiences with carrot response to various N limiting

situations due to reduced applications and excessive rains.

Sanderson (1997) conducted a five-year study on Prince Edward Island, on soils
of loamy sand to sandy loam texture [Orthic Humo-Ferric Podzols (Orthic Podzol in
FAO system)], to determine whether a reduction in N rate had any effect on carrot yield.
During a three year period at six locations, N applications were reduced by as much as
40% from 72 to 44 lb A'1 (80.6 to 49 kg ha'l) with no effect on yield. During an
additional two-year study, at four locations, 132 lb A‘1 (148 kg ha") applications were
reduced by 67% when the 88 lb A'I (98.6 kg ha '1) preplant application was eliminated.
No reduction of yield was observed, and the root weight, diameter, and length were not
affected by the reduction in N. There were no differences between any of the 10
locations based on preplant or split applications. Baseline N levels were not included in
the information given.

Hemphill and Jackson (1982) also reported no effect on carrot yield with N
treatments ranging from 0 to 240 lb A" (0 to 269 kg ha '1). Their study on Williarnette
silt loam (fine-silty, mixed, mesic, Pachic Ultic Argixeroll) focused on pH levels as a
limiting factor and reported that generally higher yield was associated with a pH of 5.1 to
5.7. Baseline N levels were not included.

Hipp (1978) conducted a study at the Weslaco Texas Agricultural Experiment

Station on soil of sandy loam texture, and related the N requirement of carrot to the

15

length of the growing season. He found that of the four N treatments, 0, 56, 112, and 168
kg ha", applied preplant, maximum yield was obtained from the 112 kg ha", but not
before 128 days or more afler planting. Extending the season another 15 to 33 days
promoted higher yield and more definitive differences between treatments. Baseline N
level was reported at 65 kg ha'1 in the 0 to 120 cm profile.

In a study on coarse textured soils [McBride sandy loam (coarse-loamy, mixed,
frigid, Oxyaquic Fragiorthods) and Montcahn loamy sand (coarse-loamy, mixed, fiigid,
Alfie Haplorthods)] where leaching of NO3' is a concern, Warncke (1996) reported that
where residual N from the previous corn crop, applied manure and preplant N at the rate
of 45 kg ha’1 totaled as much as 150 1kg ha'1 by June 25, additional applications did not
affect yield, when harvested 135 days after planting. In another study in which residual
or baseline N was 44 kg ha'1 and significant rainfall leached N into the soil profile beyond
the normal 30 cm sampling depth, root and shoots were still significantly affected by
treatments (Warncke, 1996). This study also revealed that a single N application of 90 kg
ha'1 produced root yields and top growth comparable to yields from plots receiving higher
and more frequent applications.

A study on early carrot grth showed that carrot is capable of making use of
indigenous N that has leached below the normal sampling depth of 30 cm. After just 24
days fiom emergence the maximum average length of the taproot was 38.5 cm, although
a few reached a length of 43 cm (White, 1978). Even though the study was conducted in
pots it shows that carrot is capable of reaching N whether it has leached due to rain or
over time from the previous season. Warncke (1996) provides field evidence that carrot is

capable of accessing N deeper in the profile. A combination of soil and petiole N03"

16

testing may be the best approach to managing N for carrot (Warncke, 1996).

Reduction in N applications helps prevent leaching and reduces cost to growers; it
is also important because carrot is capable of accumulating and storing excessive N in the
storage root, though the excess does not contribute to the yield (Warncke, 1996). Excess
storage can be a food consumption concern especially in baby food (Warncke, 1996;
Carrot Ramp, 2003). In 1989, Evers found that excess N also led to a reduced
concentration of sugar in roots (Evers, 1989). Marshner (1998) indicated that as the N
supply increases so does the soluble N, especially in leaves and storage organs with high
water content. As the level of N increases sucrose, polyfructosan and starch decrease. It
is important to consider subsoil sampling for baseline measurements of N, similar to Hipp
(1978), since the taproot will reach the subsoil before it is time to sample for
sidedressing.

While there is concern about excessive N in the roots, the quality of carrot tops is
equally important. In addition to providing photosynthates to the plant, many growers
use the tops to lift the roots out of the soil at harvest. For this reason many growers use
frequent N applications to keep tops healthy, but timing of the frequent applications is
essential. An application too close to harvest may contribute to residual soil nitrate and to
increased N content in the roots rather than contributing to healthy tops (Warncke, 1996).
Sufficient N also reduces the potential for disease infection. N deficiency in carrot tops
can be difficult to detect, often the leaves have a healthy green appearance but the height
of tops throughout the field may be irregular (McGiffin et al., 1997). Twenty tons per
acre (44.8 Mg ha'l) of carrot removes about 100 lb A" N (112 kg ha'l) (Zandstra et al.,

1986), and Michigan State University’s recommendation is based on replacement only at

17

100 lb A'l (Warncke et al., 1992). Other recommendations range from 60 to 150 lb A'1
(67 to 168 kg ha'l), depending on the location and soil type (Walworth, 1998; G012 and
Aakre, 1993; Fritz et al., 1998, McGiffin et al., 1997, Mills, 2001). Split applications are
recommended to help avoid leaching and runoff (Warncke, 1996). They are
recommended as a sidedress four to six weeks after planting to prevent early excessive N-
uptake, which promotes excessive vegetation, and delays root development (Mills, 2001,
Warncke et al., 1992). In addition, too much preplant N may cause forking (McGiffin et
aL,1997)

This chapter presents results of the field response of carrot to N treatments in a _
study at four locations in Montcalm County, Michigan over a two-year period, where the

response to N applications is the focus of this study.

Materials and Methods

Experimental Sites

Field studies were conducted in 2001 and 2002 at four locations in Montcalm
County, Michigan with one site split between two varieties in 2002. The soils in this
county developed from glacial debris left upon the final retreat of the Wisconsin glacial
age approximately 15,000 years ago. Soil differences throughout the county are due to
differences in texture, mineralogical composition of the parent material, and drainage.
Glacial deposits ranged from 30 to 91m thick; therefore, bedrock did not directly affect
the development of the soil. The county at large was originally forested (Soil Survey,
Montcalm Co., 1960).

Plots in 2001 and 2002 were located at the Michigan State University Montcalm

18

Experiment Station, Douglass Township, in the southern 1/2 of Section 8, T1 1N, R7W.
The experimental plots were located in Range 1 SW in 2001 and in Range 15 SE in 2002.
The soil is a well drained to moderately well drained loamy sand to sandy loam,
moderately low in organic matter, of the Hillsdale-Spinks map unit (Hillsdale: coarse-
loamy, mixed, mesic Typic Hapludalfs; Spinks: sandy, mixed, mesic Psammentic
Hapludalfs). This soil was formed on till plains from loamy sand to sandy loam parent
material. A 2 to 6% slope declines from north to south at these ranges (Soil Survey,
Montcalm County, 1960; D.L. Mokrna, personal communication, 2003). The soil surface
at the Experiment Station is coarse gravelly to cobbly.

In 2001, the second field site was located along the south side of Deaner Rd, in
the NW ‘/4 of Section 36, T12N, R9W (Winfield Township). This site belongs to
Sandyland Farms, and consists of Plainfield Sand, loamy substratum, (mixed, mesic,
Typic Udipsamments) formed on old lake plains in sand over glaciofluvial materials.
Organic matter content is low and the slope is generally 0 to 2 % (Soil Survey, Montcalm
County, 1960; D.L. Mokma, personal communication, 2003).

In 2002, the second field site was located west of Masters Rd. on Sandyland
Farms at the mid-point of the eastern V2 of Section 27, T12N, R9W (Winfield Township).
The soil is a Plainfield Sand (mixed mesic Typic Udipsamments) sloping 2 to 6% from
west to east at this location, and developed from well-drained sand (Soil Survey,

Montcahn County, 1960; D.L. Mokrna, personal communication, 2003).

Plot Design and Management Protocol

Four replications of each of four N treatments were arranged in a randomized

19

complete block design at all four locations. Plots were situated so as to minimize
heterogeneity across the treatments.

At the Experiment Station, carrots were planted in beds, May 8, 2001 on Range 1,
and May 7, 2002 on Range 15 in an east to west direction. Each bed consisted of three
rows with three lines to a row and each 4.6 x 15.0 m plot contained three beds. Sixteen
plots were planted with Diamond Cut (XPH18006, Seminis/Asgrow), a fresh market
cultivar, in both seasons. Goliath (PS30489, Petoseed), a processing cultivar, was not
replicated in 2001; only four plots, each representing a treatment, were planted to provide
another cultivar-specific coloration to contrast with the Diamond Cut. The contrast was
intended to address the potential need for a field-specific reference strip. In 2002, 16
plots were also planted with Goliath; that became the fifth site location. Granular urea
[(NH2)2CO, 46-0-0] was broadcast in three applications according to target season totals
0145, 90, 135, and 180 kg N ha" (Table 1).

The Sandyland fields, in both seasons, were already established when plots were
set up in four replications of the four N treatments. Carrots were planted in mid-April on
raised beds. Each bed had three rows with three lines in each row. Barley was planted
between rows to protect emerging carrots and was killed off with fluazifop-P-butyl once
the carrot plants were established. The Deaner Rd. field (2001) was planted in a north to
south orientation with a combination of Asgrow Bl (Asgrow) and Prime Cut 59
(Sunseeds) grown for “out and peel” production. Plots were located between the second
and fourth towers of the center pivot irrigation system. N treatments were broadcast in
three applications with granular urea (Table 1). The Masters Rd. field (2002) was planted

in Sugar Snax 54 (Sunseeds) in an east to west direction. Plots were located between the

20

first and third irrigation towers. Urea was applied in two applications totaling 20, 59, 98

and 136 kg ha" N (Table 1).

Table 1. N fertilizer (urea) split applications broadcast on the indicated dates, listed by
treatment number at the four field locations.

 

N fertilizer applications kg ha’1

 

 

 

 

 

 

 

2001 2002
Experiment Station Experiment Station

Trt+ 6/ l 3 7/ l 1 8/9 Total Trt 6/ 18 7/29 8/24 Total
1 45 -- -- 45 l -- 22 23 45
2 45 22 23 90 2 34 28 28 90
3 45 45 45 135 3 66 34 35 135
4 45 67 68 180 4 101 39 40 180

Sandyland (Deaner Rd) Sandyland (Masters Rd)

Trt 6/ l 3 7/6 8/ 1 Total Trt 7/ 3 7/29 Total
1 45 -- -- 45 1 -- 20 -- 20
2 45 34 11 90 2 34 25 -- 59
3 45 56 34 135 3 66 32 -- 98
4 45 78 57 180 4 100 36 -- 136

 

T
Trt = Treatment

Weeds at the Experiment Station were controlled with linuron (3-(3, 9-
dichlorophenyl)-1-methoxy-l methylurea) and hand weeding. Chlorothalonil
(tetrachloroisophthalonitrile) was used to control Alternaria blight (Alternaria dauci) and
Cercospora leaf spot (Cercospora carotae), serious fungal diseases for Michigan carrots
that affect top quality (Center for Integrated Pest Management MSU, 1999; Michigan
FQPA Residue Report, 2000). Chlorothalonil was supplemented with copper hydroxide
in 2002 upon the discovery of Bacterial Blight (Xanthomonas carotae) in the Goliath
cultivar. Weeds in the Sandyland fields were controlled with linuron and fluazifop-P-
butyl. Chlorothalonil was used to control fungal diseases and cyfluthrin (Cyano (4-fluoro-
3-phenoxyphenyl) methyl 3-(2, 2-dichloroethyl)-2, 2-dimethylcyclopropanecarboxylate)

was applied to control insects.

21

Rainfall, recorded by the Automated Montcahn Research Farm Weather Station,
and irrigation records for the Experiment Station plots are provided in Table 2. During
2001, irrigation was delivered with a “big gun” irrigator, and in 2002, by stationary
overhead sprinklers. The Sandyland Farms locations were irrigated with a center-pivot
system during both seasons at the rate of 3.0 to 3.8 cm per week delivered in multiple

applications.

Agronomic Measurements and Sample Analysis

Baseline soil samples were taken at planting at the Experiment Station and before
the first N treatment at the Sandyland locations. Additional in-season soil samples were
taken prior to fertilizer applications and at harvest. Eight cores per sample, approximately
30 cm deep, were pulled from alternating sides of the middle row of the middle bed of
each plot. Soil samples were dried at 60°C, ground, and analyzed for N03' and NH;
with a 1N KCl extractant.

Carrot petioles were sampled for petiole sap N03” periodically throughout the
season. Ten to twenty petioles of the youngest fully extended leaves were chosen from
the middle row of the middle bed of each plot. The leaves were discarded in the field and
the petioles transported in a cooler. When necessary the petioles were refrigerated prior
to NO3’ determination. A small segment cut from the middle of the petioles was
squeezed through a garlic press, and a few drops of the sap were placed on the electrode
surface of a Cardy Nitrate Meter (Horiba Group, Japan) to measure the N03‘
concentration (Warncke, 1996). Subsequently, the remaining tissue was dried at 60°C,

ground, and analyzed for total N concentration using the Kj eldahl Method.

22

Table 2. Rainfall recorded and irrigation amounts delivered at the Montcalm Experiment Station

during 2001 and 2002. Irrigation at Sandyland was estimated at 3.0 to 3.8 cm per week

 

 

 

 

 

2001 2002
en: en:
Week of Rainfall+ Irrigation Total Week of Rainfall+ Irrigation Total
Apr 1 3.99 0.00 3.99 Apr 1 0.48 0.00 0.48
Apr 8 0.91 0.00 0.91 Apr 7 1.93 0.00 1.93
Apr 15 1.80 0.00 1.80 Apr 14 1.52 0.00 1.52
Apr 22 1.63 0.00 1.63 Apr 21 2.34 0.00 2.34
Apr 29 0.00 0.00 0.00 April 28 2.10 0.00 2.10
May 1 0.05 0.00 0.05 May 5 4.17 0.00 4.17
May 6 1.93 0.00 1.93 May 12 3.28 0.00 3.28
May 13 6.32 0.00 6.32 May 19 0.61 0.00 0.61
May 20 6.60 0.00 6.60 May 26 1.45 0.00 1.45
May 27 5.84 0.00 5.84 June 2 2.31 0.00 2.31
June 3 0.10 0.00 0.10 June 9 1.22 0.00 1.22
June 10 1.96 0.00 1.96 June 16 4.42 0.00 4.42
June 17 1.52 0.00 1.52 June 23 0.38 3.81 4.19
June 24 0.15 0.00 0.15 June 30 0.00 3.81 3.81
July 1 0.41 1.91 2.32 July 1 0.05 0.00 0.05
July8 0.05 1.27 1.32 July 7 1.93 1.91 3.84
July 15 1.14 1.91 3.05 July 14 0.00 3.18 3.18
July 22 0.28 0.00 0.28 July 21 4.62 1.91 6.53
July 29 4.44 0.00 4.44 July 28 4.88 0.00 4.88
Aug 5 2.79 1.91 4.70 Aug 4 0.66 0.00 0.66
Aug 12 3.76 0.00 3.76 Aug 11 8.48 0.00 8.48
Aug19 4.93 0.00 4.93 Aug 18 6.63 0.00 6.63
Aug 26 3.10 0.00 3.10 Aug 25 0.02 0.00 0.02
Sept 2 2.46 0.00 2.46 Sept 1 0.69 0.00 0.69
Sept 9 2.51 0.00 2.51 Sept 8 0.00 1.91 1.91
Total 58.67 7.00 65.67 Total 54.17 16.53 70.70

 

 

1hAutornated Montcalm Research Farm Weather Station

Ancillary soil moisture information was obtained with a TDR (Time Domain

Reflectometry, Trime, Irnko) 3-rod, 160 mm probe, that uses an electromagnetic pulse to

23

determine soil moisture content. Four measurements per plot, two perpendicular and two
parallel to the row, were taken weekly throughout the 2001 field season. Calibration was
performed against the volumetric water content of samples taken fiom the Experiment
Station, Range 1 site, and the Deaner Rd. (2001) field. Four cores per plot were pulled
and divided into four equal depths of 5.0 cm each. Gravirnetric water content and bulk
density were determined from which the volumetric water content was derived. A simple
linear regression model (SAS Inst., version 8.2) revealed that the TDR measurements
were reliable 64% of the time at the Experiment Station and 84% of the time at the
Deaner Rd (2001) field when three depths, 5 to 10, 10 to 15, and 15 to 20 cm, were
combined. The cobbly nature of the soil at the Experiment Station may have caused
wave interruption which reduced reliability at that location. In 2002, measurements of
soil moisture were reduced to two parallel measurements per plot. Sampling was
terminated in July due to mechanical problems with the equipment.

Harvest at the Experiment Station took place on September 13, in both years.
Deaner Rd. and Masters Rd. fields were harvested on August 23, 2001 and August 20,
2002, respectively. Carrots, at all locations, were dug by hand from the center 3.0 In of
the middle row of the middle bed of each plot. Whole plants were harvested from each
plot and weighed in bulk. The tops were separated from the roots using a portable
squeeze-roll topper, and subsarnpled. The roots were graded according to marketable
size: #1 > 5/8 inch to 1% inch diameter at the shoulder, jumbo > 1% inch, and small < 5/8
inch. Culls were roots that were misshaped, cracked or infected. Graded roots were
weighed, counted, and subsarnpled. Top and root subsarnples were subsequently

weighed before drying at 60°C. Dried samples were weighed again, ground and analyzed

24

for total N concentration using the Kjeldahl Method. Subsarnples were used to determine
the total dry matter of tops and roots. Certain errors in recording dry weight required the
use of estimated moisture content to determine dry matter in tops at the Experiment
Station in 2001, and to determine dry matter in tops and roots at Sandyland 2002.
Michigan State University Plant and Soil Nutrient Laboratory performed the chemical
analyses.

Statistical analysis was performed using regression models and analysis of

variance (SAS Inst., version 8.2).

Results and Discussion

Soil N Availability and Plant Uptake

All data were normally distributed as evidenced by the Shapiro-Wilk test and
residual plots. Extreme outliers, defined by SAS Univariate procedure (SAS Inst.,
version 8.2), were eliminated. Over 6000 measurements were analyzed and 12 were
removed as outliers. The deep fibrous root system of the carrot crop (Warncke, 1996)
used not only the N applied but generally drew down the residual N03" - N from the
previous potato crop, when N was applied early enough before harvest. Total plant
uptake resulted in accumulation of more N than was thought available in the soil. Only N
applied at the rate of 180 kg ha'1 resulted in available N in excess of plant uptake.
Nitrogen uptake by the plants responded to treatments, and as the amount of applied N
increased the amount of unaccounted for N decreased (Table 3). The high level of
indigenous N in 2001 and other unexpected sources of N, described below, may in part

explain the reason that results did not show more separation between treatments.

25

Table 3. Nitrogen activity in the soil including initial residual levels in the top 30 cm, applications, and
ending residual levels compared to N uptake by plants and carrot root yield. Mean separation of plant
uptake of N as conrpared to N treatments. Linear regression analysis as used to conrpare plant uptake of
N with available N.

 

 

 

 

 

N03' Intake by Plants Avfilljble —
Initial Urea fromf Ending (Uptake + Root
Residual Applied water Available Residual Tops Roots Total residual) Yield
Nitrogen kg ha'1 Mg ha'I
Montcalm Experiment Station 2001/Diamond Cut
45.7 45 14.4 105.1 34.9 60% 67.4 128.3b -58.1a 45.2
49.2 90 14.4 153.6 27.5 80.7ab 84.2 164.9ab -38.8ab 53.3
52.4 135 14.4 201.8 40.7 99.5a 81.3 180.8a -19.7ab 47.5
44.3 180 14.4 238.7 60.0 84.7ab 81.3 166.0ab 12.7b 46.2
p-value 0.03 ns 0.01 0.01
0.38‘ 0.22 0.37‘ 0.52“
Sandyland (Deaner Rd) 2001/Asgrow Bl, Prime Cut 59
18.1 45 n/a 63.1 26.8 49.3b 50.3b 99.6b -63.3a 47.8
18.5 90 n/a 108.5 25.3 61 .7ab 63.5ab 125.2ab -42.0ab 51.7
21.9 135 n/a 156.9 26.7 71 .5ab 71 .7ab 143.2a -l3.0bc 53.4
17.7 180 n/a 197.7 24.5 80.2a 73.6a 153.8a 19.4c 49.9
P'value 0.02 0.03 0.007 0.0004
r2 0.50” 0.49” 0.59'” 0.78‘”
Montcahn Experiment Station 2002/Diamond Cut
15.2 45 34.0 94.2 24.1 74.1 79.2 153.3 -83.2a 54.4
16.2 90 34.0 140.2 24.2 99.5 78.6 178.1 -62.1a 50.1
18.0 135 34.0 187.0 30.5 93.2 88.9 182.1 -25.6ab 54.2
16.9 180 34.0 230.9 34.0 103.3 84.9 188.2 8.7b 52.2
p-value ns ns ns 0.003
0.14 0.09 0.19 .67‘”
Montcalm Experiment Station 2002/Goliath
15.9 45 34.0 94.9 19.3 77.8 73.6 151.4 -75.8a 59.7
13.1 90 34.0 137.1 23.5 87.9 83.2 171.1 -57.5ab 55.6
20.4 135 34.0 189.4 29.9 82.2 99.7 181.9 -22.4bc 61.1
18.4 180 34.0 232.4 37.0 99.3 93.5 192.8 2.6c 58.5
p-value ns ns ns .002
0.18 0.29’ 0.35 0.85’”
Sandyland (Masters Rd) 2002/Sugar Snax
16.2 20.4 n/a 36.6 16.5 70.3 39.9 110.2 -90. la 34.8
12.1 59.1 n/a 71.2 15.0 95.1 41.3 136.4 -80.2ab 30.4
14.0 97.9 n/a 111.9 16.8 81.4 48.3 129.7 -34.6ab 40.3
13.1 136.5 n/a 149.6 21.8 106.6 50.6 157.2 -29.4b 37.0
p-value ns ns ns 0.02
0.21 0.15 0.28‘ 0.49"

 

T N03- - N from irrigation water was calculated based on approximate rate of 20.8mg N03- - N L'I .
Mean values with the same letters are not significantly different at p $0.05.

ns = Overall F -value is not significant.

r2: Correlation of Available N to plant uptake resulting from regression analysis.

0 O. 0..

Significance of overall F-values at p< 0.05, 0.01, 0.001, respectively.

26

In 2001, at the Experiment Station, several confounding factors contributed to the
somewhat unpredictable responses by carrot plants to N treatments. Residual N in 2001
from a previous crop was relatively high at approximately 45 to 53 kg ha'1 (Table 3),
predominantly in the form of N03“ (Table 4). At least 10 of the 16 plots (2 1/2 reps) of the
Diamond Cut variety were subjected to additional inigation from an adjacent grower’s
field that may have contained unknown quantities of NO;'. Experiment Station wells
contained from 17 to 21 mg L‘1 NO3' - N that added approximately 14.4 kg ha'I NO3' - N
to the crop in 2001 and 34.0 kg ha'I NOg‘ - N in 2002 through the irrigation water. The
unscheduled additional N raised even the lower treatments, at the Experiment Station,
close to the seasonal MSU recommendation for carrot of 112 kg ha’1 (Warncke et al.,
1992), and the remaining treatments well above planned N levels. Nitrogen content of
the irrigation water at Sandyland was unknown; however, Table 3 shows plant uptake in
excess of known availability.

In 2001, N uptake was divided fairly equally between tops and roots at both the
Experiment Station and the Sandyland location. Although tops generally produce greater
biomass than roots (White and Strandberg, 1978), only treatments 3 (135 kg ha") and 4
(180 kg ha") at the Experiment Station and treatment 4 at Sandyland showed greater N
accumulation in the tops than the roots. At the Experiment Station, only the tops showed
significant correlation to the N treatments, while at Sandyland, both the tops and roots
showed significance at p $0.05. In 2002 tops generally accumulated more N than roots
(Table 3), but N uptake was not significantly correlated to treatments in either the tops or
the roots.

When N was applied at least 35 days before harvest (Table 4, 2001 Exp. Stn.),

27

residual NO3' - N in the soil, and total residual N as well, was reduced below elevated
residual levels from the previous crop. Only where N was applied at the rate of 180 kg

ha'1 did the residual N exceed the levels at planting across the four replications (Table 4).

Table 4. Comparison of beginning of season residual N with end of season residual N in the
soil derived with KCl extractant.

 

 

 

 

Urea Initial Residual N from previous crop Residual N following carrot harvest
Applied N03' - N NH,“ - N Total NO3' - N NH,+ - N Total
kg ha" kg ba‘I

Montcalm Experiment Station 2001/Diamond Cut
45 39.1 6.6 45.7 7.6 27.3 34.9
90 42.3 6.8 49.1 3.7 23.7 27.5
135 44.8 7.6 52.4 15.2 25.3 40.7
180 38.9 5.5 44.4 34.4 25.6 60.0
Sandyland (Deaner Rd) 2001/Asgrow Bl Prime Cut 59
45 8.1 10.1 18.2 3.6 23.2 26.8
90 7.4 11.1 18.5 3.7 21.6 25.3
135 8.1 13.9 22.0 3.7 22.8 26.7
180 7.9 9.9 17.8 2.2 22.3 24.5
Montcalm Experiment Station 2002/Diamond Cut
45 12.4 2.8 15.2 9.9 14.2 24.1
90 11.6 4.6 16.2 11.9 12.3 24.2
135 12.9 5.1 18.0 16.9 13.5 30.5
180 12.2 4.7 16.9 22.8 11.1 34.0
Montcalm Experiment Station 2002/Goliath
45 11.6 4.3 15.9 3.6 15.7 19.3
90 9.5 3.4 12.9 9.0 14.6 23.5
135 14.4 5.9 20.3 15.2 14.7 29.9
180 12.5 5.8 18.3 22.2 14.8 37.0
Sandyland (Masters Rd) 2002/Sugar Snax 54
20 13.7 2.6 16.3 1.2 15.2 16.5
59 9.6 2.5 12.1 1.1 13.9 15.0
98 12.0 2.0 14.0 1.3 15.5 16.8
136 10.6 2.5 13.1 2.4 19.4 21.8

 

In 2001, at Sandyland, N was applied only 23 days before harvest and even though
residual N03' - N was reduced, the total residual N was higher at harvest than at planting
(Table 4). In 2002, at all locations, the last treatment was applied 20 days before harvest.

Here too, while residual NOg' -N was generally reduced in all but the plots representing

28

rates of 135 and 180 kg ha”, the total residual N remained higher than initial levels.
According to Wamcke (1996) N applications too close to harvest may contribute to
excess residual N such as experienced here. When N was applied at least 62 days before
harvest, available N levels were reduced to near background levels (Warncke, 1996). In
this study, available N would have been reduced to below initial levels if the last
application had occurred earlier in the season or the carrots had been allowed to continue

development for a later harvest date.

N Treatments vs Petiole Response

The % N content of the petioles sampled throughout the growing season is shown
in Table 5. In both 2001 and 2002, % N generally responded to the amount of N applied
as of each date petiole samples were taken at both the Experiment Station and Sandyland;
however, not all resulted in significant differences between treatments.

In addition, petiole sap N03' generally responded to the N treatments, including N
uptake in addition to that required for maximum yield (Tables 6 and 7). When compared
to % N content, petiole sap NO; was significantly correlated at r2 > 0.40 on nine of the
15 sampling dates covering the two year period (Tables 6 and 7), even when neither
parameter reflected positive response to N treatments. In a previous study Warncke
(1996) indicated that petiole sap NO3' content is an apparent good indicator of the N
status of the carrot plant. Petiole sap N03' using the Cardy Meter, a quick in-field NO3'
test, was shown to reflect the N status of the carrot plant in this study as well as earlier
studies by Warncke (1996) and Hochmuth (1994). Just prior to harvest, at the

Experiment Station, % N and petiole sap NO3' generally indicated that N uptake and

29

Table 5. Mean separation between treatments of % N in carrot petioles sampled on the indicated dates.
Average yield per N treatment, from each specified location, listed for corrrparison to % N in petioles.

 

Montcalm Experiment Station 2001/Diamond Cut
Planting date = 5/8/01 Harvest date = 9/13/01

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Trt)r Root Yield July 20 (73): July 25 (78) Aug 15 (99) Sept 11 (126)
kg ha‘1 Mg ha" % N (TKN)
45 45.2 1.56 1.26 0.65b 0.73c
90 53 .3 1.62 1.32 0.84ab 0.93bc
135 47.5 1.57 1.35 0.88ab 1.03ab
180 46.2 1.72 1.30 1.00a 1.20a
p-value ns ns 0.01 0.0002
Sandyland (Deaner Rd) 2001/Asgrow Bl Prime Cut 59
Planting date ~ 4/20/01 Harvest date = 8/23/01
Trt Root Yield July 11 (82) July 19 (90) Aug 1 (103)
kg ha1 Mg ha" % N (TKN)
45 47.8 0.44 0.25c 0.27
90 51.7 0.54 0.45bc 0.35
135 53.4 0.65 0.58ab 0.32
180 49.9 0.61 0.803 0.41
p-value ns 0.0002 ns
Montcalm Experiment Station 2002/Diamond Cut
Planting date = 5/7/02 Harvest date = 9/13/02
Trt Root Yield July 25 (76) Aug 22 (104) Sept 9 (122)
kg ba'T Mg ha‘T % N (TKN)
45 54.4 0.76b 0.50c 0.62
90 50.1 0.88ab 0.56bc 0.69
135 54.2 0.89ab 0.66ab 0.76
180 52.2 0.96a 0.72a 0.77
p-value 0.02 0.001 ns
Montcalm Experiment Station 2002/Goliath
Planting date = 5/7/02 Harvest date = 9/13/02
Trt Root Yield July 25 (76) Aug 22 (104) Sept 9 (122)
kg ha'I Mg ha" % N (TKN)
45 59.7 0.83 0.42b 0.77b
90 55.6 0.87 0.57a 1.02ab
135 61 . 1 0.90 0.59a 0.96ab
180 58.5 1.02 0.66a 1.19a
p-value ns 0.002 0.009
Sandyland (Masters Rd) 2002/sugar Snax 54
Planting date ~ 4/20/02 Harvest date = 8/20/02
Trt Root Yield July 25 (96) Aug 19 (121)
kg ha" Mg ha“ % N (TKN)
20.4 34.8 0.58b 0.63
59.1 30.4 0.72b 0.64
97.9 40.3 0.6% 0.59
136.5 37.0 0.99a 0.58
p-value 0.002 ns

 

TTrt = Treatment
tNumber in parenthesis represents days afier planting.

Mean values with the same letters are not significantly different at p $0.05.

30

Table 6. Linear regression results of petiole sap NO3' - N vs Total N (TKN) of dried petioles
sanrpled on various dates during the 2001 season. N available was as of the petiole sampling date.
Values shown were averaged by treatment.

 

Montcalm Experiment Station 2001/Diamond Cut

 

 

Sap Sap Sap
N“ Avail NO3' Total N N Avail NO3' Total N N Avail NO3' Total N
------------- July 20------------- -------------July 25 Aug 15-------------
Kg ha"l mg L'1 % Kg ha'1 mg L'1 % Kg ha'I mg L'1 %
101.0 2600 1.56 101.0 2250 1.26 105.1 720 0.65

126.9 2600 1.62 126.9 3175 1.32 153.6 1240 0.84
152.5 3150 1.57 152.5 2475 1.35 201.8 1725 0.88
166.9 3075 1.72 166.9 2475 1.30 238.7 1975 1.00
t2 = 0.06 r2 = 0.31‘ r2 = 0.84‘“
Montcalm Experiment Station 2001/Diamond Cut

 

Sap
N Avail NO3' Total N
------------- Sept 11--------------
Kg ha'l mg L'1 %
105.1 640 0.73

153.6 1220 0.93
201.8 2200 1.03
238.7 2800 1.20

 

 

 

r2 = 0.75‘”
Sandyland (Deaner Rd) 2001/Asgrow Bl, Prime Cut 59
Sap Sap Sap

N Avail NO3' Total N N Avail NO3' Total N N Avail NO3' Total N
------------- July 11------------ --------------July 19 Aug 1------------
Kg ha'1 mg L'I % Kg ha'l mg L'1 % Kg ha'l mg L'1 %

63.1 294 0.44 63.1 242 0.25 63.1 300 0.27

96.9 327 0.54 96.9 380 0.45 108.5 287 0.35

122.7 502 0.65 122.7 595 0.58 156.9 262 0.32

140.9 487 0.61 140.9 762 0.80 197.7 330 0.41

2 = 0.69‘” r2 = 0.72‘” r2 = 0.004

 

IEstimated N available at the specific sarrrpling date including residual from the previous crop,
treatment applications, and an estimate of N03- - N from the irrigation water.

O O. 0..

Significance of overall F-values at p $0.05, 0.01, 0.001, respectively.

storage in petioles responded to N treatments. Results were less favorable at the

Sandyland locations.
In 2001, N uptake in harvested tops generally agreed with % N in petiole and

petiole sap NO3', except at the Experiment Station where treatment 3 (135 kg ha") had

31

Table 7. Linear regression results of petiole sap N03- - N vs Total N (TKN) of dried petioles

sampled on various dates during the 2002 season. N available was as of the petiole sampling date.
Values shown were averaged by treatment.

 

Montcalm Experiment Station 2002/Diamond Cut

 

 

 

 

 

 

 

 

 

 

Sap Sap Sap
Nf Avail NO3' Total N N Avail NO3° Total N N Avail NOg' Total N
-- -------- ---July 25 Aug 22 Sept 9 -------------
Kg ha'l mg L'1 % Kg ha'l mg L'1 % Kg ha'l mg L'1 %
45.2 1470 0.75 67.2 530 0.50 94.2 410 0.62
80.0 1550 0.88 108.0 648 0.56 140.2 790 0.69
115.4 1975 0.89 149.0 960 0.66 187.0 1074 0.76
147.8 1875 0.96 187.0 1233 0.72 230.9 1375 0.77
r2 = 0.06 r2 = 0.43“ r2 = 0.23
Montcalm Experiment Station 2002/Goliath
Sap Sap Sap
N+ Avail NO3' Total N N Avail NO3' Total N N Avail NO3' Total N
-------------- July 25 Aug 22 Sept 9-------------
Kg ha'l mg L'1 % Kg ha'1 mg L'1 % Kg ha'l mg L'1 %
46.0 2225 0.83 68.4 290 0.42 94.9 437 0.77
76.8 2600 0.87 104.8 845 0.57 137.1 797 1.02
117.7 2550 0.90 151.3 815 0.59 189.4 670 0.96
149.3 3550 1.02 188.5 1575 0.66 232.4 1310 1.19
r2 = 0.0 r2 = 0.70‘” r2 = 0.50”
Sandyland (Masters Rd) 2002/Sugar Snax 54
Sap Sap
N Avail NO3' Total N N Avail NOg' Total N
-------------- July 25 Aug 19------------
Kg ha’I mg L'1 % Kg ha'l mg L'1 %
16.2 453 0.58 36.6 230 0.63
45.7 378 0.72 71.2 313 0.64
81.2 383 0.69 112.9 273 0.59
1 13.9 943 0.99 149.6 315 0.58
r2 = 0.51” r2 = 0.14

 

1Estimated N available at the specific sampling date including residual from the previous crop,
treatment applications, and an estimate of NO3' - N in the irrigation water.

0 O. 0..

Significance of overall F -values at p $0.05, 0.01, 0.001, respectively.

the highest accumulation of N instead of the treatment 4 as reflected in the petiole results
(Tables 6 and 8). At Sandyland, N accumulation in tops reflected the amount of N
applied; however, results of late season petiole analyses were mixed. In 2002 treatment 1

(45 kg ha'1 N) resulted in the lowest accumulation of N, while treatment 4 (180 kg ha")

32

resulted in the highest, but intermediate treatment levels varied (Table 8).

Treatments vs N Uptake in Harvested Dry Matter and Yield

Dry matter and N uptake in tops were significantly correlated (Table 8) at ahnost
all locations. At the Experiment Station 2001, dry matter and N uptake increased with
treatment, and treatment 3 resulted in the highest amount of dry matter accumulation.
Even though % N and petiole sap NO3' showed that carrot continued to accumulate more
N through treatment 4 (Table 6), it did not increase dry matter indicating excessive
uptake of N by the plant (Table 8). Dry matter and N uptake in tops as well as % N and
petiole sap NO3' at Sandyland increased with treatment amounts through treatment 4.
Available N in treatment 4 at Sandyland was about the same as treatment 3 at the
Experiment Station, where higher amounts of N had been available at the beginning of
the season. There, the carrot crop generally reduced available soil N below initial levels.
With the 2002 season end at Sandyland, dry matter content of tops related to N uptake
but was not significantly correlated. Top dry matter and the N uptake of Goliath carrots
at the Experiment Station were significantly correlated, but with an inverse trend that
may have been due, in part, to foliar blight in the Goliath variety. N accumulation in tops
generally increased with treatment but permanent damage from the blight reduced the
amount of dry matter and affected the positive correlation to N uptake. Generally, the N
treatments seemed to influence top grth to a greater extent than roots, as indicated by
Warncke (1996).

Only at the Experiment Station in 2001 was dry matter (Table 8) in roots

nominally influenced by N treatment. There, and in the Goliath variety and at Sandyland

33

Table 8. Mean N uptake in tops and roots compared to dry matter and root yield using

regression analysis. Mean available N compared to root:shoot ratio using regression analysis.
Significance of treatment differences in dry matter, root: shoot ratio, and root yield determined

using analysis of variance.

 

 

 

 

 

 

Root: Fresh
Available ------- N Uptake ------------ Dry Matter---- Shoot Root
trtf N Tops Roots Tops Roots Ratio Yield
kg ha’1 Mg ha'1
Montcalm Experiment Station (2001)/Diamond Cut (128 daysi)
45 105.1 60.9 67.4 3.4b 4.9a 1.5a 45.2
90 153.6 80.7 84.2 4.3a 5.8a 1.3a 53.3
135 201.8 99.5 81.3 4.3a 5.1a 1.2a 47.5
180 238.7 84.7 81.3 4.1ab 5.0a 1.2a 46.2
p-value 0.026 0.045 0.04 0.06
Regression (r2) 0.48“ 0.59‘“ 0.38“ 0.51“
Sandyland (Deaner Rd.) 2001/ Asgrow B1, Prime Cut 59 (125 days)
45 63.1 49.3 50.3 4.6b 6.1 1.3a 47.8
90 108.5 61.7 63.5 5.1ab 6.8 1.3a 51.7
135 156.9 71.5 71.7 6.2ab 6.3 1.0b 53.4
180 197.7 80.2 73.6 6.4a 5.8 0% 49.9
p-value 0.029 ns 0.001 ns
Regression (r2) 0.36. 0.10 0.63.” 0.37.
Montcalm Experiment Station (2002)/Diamond Cut (129 days)
45 94.2 74.1 79.2 4.7 5.5 1.2 54.4
90 140.2 99.5 78.6 6.0 5.3 1.0 50.1
135 187.0 93.2 88.9 5.0 5.2 1.0 54.2
180 230.9 103.2 84.9 5.6 5.2 0.9 52.2
p-value ns ns ns ns
Regression (r2) 0.73‘” 0.19 0.06 0.27‘
Montcalm Experiment Station (2002)/Goliath (129 days)
45 94.9 77.8 73.6 5.8 5.8 1.0 59.7
90 137.1 87.9 83.2 5.7 5.3 0.9 55.6
135 189.4 82.2 99.7 5.5 6.2 1.2 61.1
180 232.4 99.3 93.5 5.5 5.4 1.0 58.5
p-value ns ns ns ns
Regression (r2) 0.37‘ 0.29‘ 0.02 0.14
Sandyland (Masters Rd) 2002/ Sugar Snax 54 (122 days)
45 36.6 70.3 39.9 6.2 4.3 0.7 34.8
90 71.2 95.1 41.3 7.7 3.8 0.5 30.4
135 111.9 81.4 48.3 7.4 5.1 0.7 40.3
180 149.6 106.6 50.6 7.8 4.5 0.6 37.0
p-value ns ns ns ns
Regression (r2) 0.23 0.47” 0.004 0.37"

 

f tn = treatment

1 Number of days from planting to harvest.
Mean values with the same letters are not significantly different at p $0.05.
ns = Overall F-value is not significant.

O O. .0.

Significance of overall F-values at p s 0.05, 0.01, 0.001, respectively.

34

in 2002, correlation was significant between N uptake and dry matter accumulation in
roots. At the Experiment Station, in 2001and 2002 the Diamond Cut and Goliath
varieties maximized N uptake in the roots at between 153 and 189 kg ha'1 available N.
The large range is due to the inclusion of estimated N03' - N fiorn irrigation water of up
to 34.0 kg ha". Root yield increased with N applied and the maximum yield, which was
significantly correlated to N uptake, also ranged between 153 to 189 kg ha'1 N. In both
years, the Sandyland carrot crops were planted at a high population rate for the “cut-and-
peel” market. While in 2001, yield was maximized at 157 kg ha”, similar to the
Experiment Station, N accumulation in the roots continued to increase with total
available N, up to almost 198 kg M”. In 2002, the carrots were harvested before the last
planned N application occurred so that the highest available N was 150 kg ha". It is
believed that the 2002 yield at Sandyland would have been sirrrilar to the others if the
carrots had not been harvested until after the last scheduled N application, and the season
extended to match the season length of the other locations (Table 8). At both Sandyland
locations the roots accumulated more N than was used for biomass production, similar to
the Warncke (1996) study. Results at Sandyland reflect treatments applied to research
plots within the field; they are not indicative of N applied to the commercial areas.

The tops continued to accumulate more N at the excessive 180 kg ha'1 rate
(treatment 4) at all locations except the Experiment Station 2001, even though it was not
reflected in the yield as previously shown by Warncke (1996) and Hochmuth et a1.(1999).
At Sandyland, 2002, maximum yield occurred at 112 kg ha'1 due, at least in part, to
reduced applications. The amount of N applied in both treatments 3 and 4 were higher

than the MSU recommendation of 112 kg ha'1 (Warncke et al., 1992). According to

35

Hochmuth (1999) additions of N in excess of that required for maximum root production
only served to increase shoot growth. Generally, when N is the limiting nutrient,
increasing the N supply enhances both root and shoot, but mostly shoot growth that
results in a decreasing root:shoot ratio (Marshner, 1998). Hochmuth (1999) noted that in
carrot once maximum yield has been attained, a decreasing root:shoot ratio resulting from
additional N applications, could result in potential reduction in yield and profit from
excess N fertilization. Table 8 shows where the excess N accumulated in the shoots, it
resulted in excessive top growth at three of the four locations affected. At the Sandyland
locations, top growth was maximized in treatment 4, with a decreasing root:shoot ratio.
The root:shoot ratio also fell in the 2002 Experiment Station plots, as the result of either a
decreased root dry matter or increased shoot grth (dry matter).

Carrot yield did not significantly respond to N treatments (Table 9). Maximum
yield was attained with treatment 3 at all locations except the Experiment Station in 2001.
At that site, treatment 2, which corresponded to the amount of available N in treatment 3
at the other locations, produced the highest yield. Diamond Cut and Prime Cut 59, both
Irnperator type varieties, maximized yield at about 53 to 54 Mg ha”. If Sugar Snax 54,
also an Irnperator type variety, would have had a comparable season length and N
treatments, it too may have yielded as much as the other Irnperator type varieties. The
Danvers type variety, Goliath, yielded 61 Mg ha". The majority of jumbo roots, greater
than 1% inches in diameter at the shoulder, were yielded by Goliath at 35-45% of the
total yield. Diamond Cut yielded 3-15% jumbos. The Prime Cut 59 and Sugar Snax 54
varieties were planted at a high population rate for the cut and peel market, and as

expected, did not yield any jumbo roots. Yield of Diamond Cut and Sugar Snax was

36

Table 9. 2001 and 2002 carrot root yield by grade and percent of the total.

 

 

 

 

 

 

N Grade Total ---------- “Grade ------------
Applied Jumbo“ #11 Small“ Can“ “61“ Jumbo #1 Small Cull
Mg ha" ----------- % of total yield ---------

Montcalm Experiment Station (2001)/ Diamond Cut Seeding rate: 204,000/A

45 6.6 28.5 1.0 9.1 45.2 0.15 0.63 0.02 0.20
90 6.9 35.9 1.0 9.5 53.3 0.13 0.67 0.02 0.18
135 7.1 27.6 0.9 11.9 47.5 0.15 0.58 0.02 0.25
180 7.1 29.6 1.4 8.1 46.2 0.15 0.64 0.03 0.18

Sandyland (Deaner Rd) 2001/Asgrow BI, Prime Cut 59 Seeding rate 870,000/A

45 0.00 6.0 41.8 0.00 47.8 0.00 0.13 0.87 0.00
90 0.00 12.9 38.8 0.00 51.7 0.00 0.25 0.75 0.00
135 0.00 11.1 42.3 0.00 53.4 0.00 0.21 0.79 0.00

180 0.00 7.2 42.7 0.00 49.9 0.00 0.14 0.86 0.00
Montcalm Experiment Station (2002)fDiamond Cut Seeding rate: 450,000/A

45 1.5 39.1 8.9 4.9 54.4 0.03 0.72 0.16 0.09

90 2.7 32.4 8.8 6.2 50.1 0.05 0.65 0.18 0.12

135 1.6 36.4 10.8 5.4 54.2 0.03 0.67 0.20 0.10

180 1.8 35.1 9.8 5.5 52.2 0.03 0.67 0.19 0.11

Montcalm Experiment Station (2002)/Goliath Seeding rate: 450,000/A
45 20.9 20.5 14.7 3.6 59.7 0.35 0.34 0.25 0.06

90 19.8 20.5 11.7 3.6 55.6 0.36 0.37 0.21 0.06
135 26.7 17.9 12.4 4.1 61.1 0.44 0.29 0.20 0.07
180 23.8 17.3 12.6 4.8 58.5 0.41 0.30 0.22 0.08

Sandyland (Masters Rd) 2002/Sugar Snax Seeding rate: 750,000/A

45 0.00 23.4 10.4 1.0 34.8 0.00 0.67 0.30 0.03

90 0.00 17.0 12.3 1.1 30.4 0.00 0.56 0.40 0.04
135 0.00 25.2 13.1 2.0 40.3 0.00 0.63 0.33 0.05
180 0.00 24.9 10.9 1.2 37.0 0.00 0.67 0.29 0.03

 

fJumbo = >1 V2 in at the shoulder, #1 = > 5/8 in to 1V2, small = 5/8 in and smaller,
culls = rrrisshaped, cracked or infected roots.

dominated by the #1 grade, between 5/8 inch and 1‘/2 inch, ranging from 56 to 72% of the
total yield. The Goliath variety had a fairly equal split between jumbo and #1. The
Sandyland 2001 location was planted with an excessively high population rate; most of
the roots were small at 5/8 inch and smaller. The culls were a higher percentage of total

yield in Diamond Cut than other varieties, as high as 25% in treatment 3, 2001. Culls

37

consisted primarily of miSshaped roots, possibly resulting from cobbly soil conditions at

the Experiment Station and disturbance of the roots caused by intense early season rains.

Conclusions

Carrot response to the planned N treatments as applied was limited and
confounded by extraneous variables discussed throughout the chapter. Only in 2001 did
total N uptake in the plants respond to treatments. Petiole samples provided the best in-
season correlation to treatments. However, the carrot crop did respond to the conditions
and certain significant correlations and lessions were provided in the outcome.

Maximum N uptake in the storage roots correlated with maximum yield at the
Experiment Station. Only in the experimental plots at Sandyland did N continue to
accumulate beyond maximum yield, even though it did not contribute to biomass
production. Timing of the last N application could have contributed to the excessive
storage of N and points out the importance of applying N early enough to avoid excess
accumulation in the roots, as shown in Warncke (1996). At the Experiment Station 2001,
where N had been applied at least 35 days before harvest, dry matter content significantly
correlate to N uptake. Better timing of N applications may have promoted better tissue
growth with realized economic benefits at harvest.

Total plant uptake exceeded amounts estimated as available to the crop at all
treatment levels except treatment four. Nitrogen that was unaccounted for may have in
part become available through mineralization of past crop residue. Initial soil sampling
below 30 cm prior to N applications may have revealed additional N that became

available to the carrot crop once the deep fibrous root system extended beyond the 30 cm.

38

Although tops are not considered part of yield, their health is important in certain
harvesting operations where tops are used to help lift the roots out of the soil. If N
applications are necessary toward season end, foliar applications may be enough to boost
the health of the tops without adding nutrients to the soil. In 2002 tops at the Experiment
Station received an estimated 34.0 kg ha’1 N through inigation compared to 14.4 kg ha'1
N in 2001. Comparison of the root:shoot ratio for the two year study for the Diamond
Cut variety indicated a positive influence fiom the foliar applications. Foliar applications
of N have been known to aid plant vigor during times of stress (Warncke, 2000), and the
N is absorbed very quickly (Tremblay et al., 2001).

Total petiole N and petiole sap NO3' have been shown to be good indicators of the
in-season N status of the carrot crop (Warncke, 1996). Results of analyses on samples
collected several times during the two seasons indicated that petiole sampling
significantly reflected the amount of N applied in about 60% of the samples, even where
foliar diseases were a problem. The remaining samples, although lacking significance,
generally reflected the amount of N applied. N applied through irrigation water may
have compromised differences between planned applications. Percent N was still a good
indicator of N status in this study, and petiole sap NO3' , using the Cardy Meter, a quick
in-field NO3' test performed as well as total petiole N analysis.

In this study, carrot generally drew down N in the profile when N was applied at
least 35 days prior to harvest. If treatments were applied closer to harvest, the carrot crop
did not have adequate time to take up the N applied. Tremblay et a1. (2001)
experimentally determined for carrot that residual N in excess of 30 kg ha'1 was

excessive.

39

References

Bell, C.E., B.E. Boutwell, E.J. Ogbuchiekwe, and ME. McGiffen, Jr.
2000. Weed control in carrots: the efficacy and economic value of Linuron.
HortScience. 35(6): 1089-1091.

Bunn, J.H. Ltd. Bunns Fertilizer Online: Bulk densities. Bunns Lane, Great Yarrnouth,
Norfolk, NR 31 OJD, UK. (Available on-line at
http://www4hbunn.co.uk/techical/Bulkdensity.html.) (Verified 1 May 2001.)

Carrot Ramp. Hausbeck, M.K., Project Director. 2003. Risk Avoidance and Mitigation
Program (RAMP). [Online]. Available
http:/@lantpathology.msu.edu/labs/hausbeck/CarrotRAMP/. (verified 03 January
2004).

CropScan, Inc. 1995. User’s manual technical reference.

Epps, A.C. 1970. W. Vandre (ed.) 1997. Carrots in Alaska. Bull. HGA-00025. Alaska
Coop. Ext., Univ. of Alaska, Fairbanks. (Available on-line at
http://www.uaf.edu/coop-ext/publications/freepubs/HGA-00025.html.) (Verified
28 April 2003.)

Evers, A.-M.1989. The role of fertilization practices in the yield and quality of carrot
(Daucus carota L.) J. Agr. Sci. Finland. 61:329-360.

Finch-Savage, W.E. 1987. The potential for seed, sowing, and seedbed preparation
treatments to improve the production of uniformly sized carrot roots for
processing. Acta Hortic. 220: 181-1 86.

Fotlr, H.D., and B.G. Ellis. 1997. Soil fertility. 2nd ed. CRC Press, Inc. Boca Raton, Fl.

Fritz, V., C. Tong, C. Rosen, and J. Wright. 1998. Vegetable crop
management: Carrots Daucus carota. Bull. WW-7196-GO. Univ. of Minnesota
Ext. Serv., Univ. of Minnesota. (Available on-line at
http://www.extension.umn.edu/distribution/horticulture/DG7l96.htm1.) (Verified
28 April 2003.)

Golz, T., and D. Aakre (ed.) 1993. Carrots. Alternative agriculture series,
No.14. North Dakota State Univ. Ext. Serv., North Dakota State Univ. of Agric.
And Applied Science, Fargo. (Available on-line at
http://www.ext.nodak.edu/extpubs/alt-ag/carrotshtm.) (Verified 28 April 2003.)

Hanlon, E.A., G.J. Hochmouth, and CR. Campbell (ed.) 2002. Reference sufficiency

ranges vegetable crops: carrot. Sufficiency Ranges for Plant Analysis. Regional
Project SERA-IEG-6. Bull. SCSB #394. (Available on-line at

40

http:l/www.ncagr.com/agronomi/saaesd/.) (Verified 28 April 2003.)

Hemphill, Jr., DD, and T.L. Jackson. 1982. Effect of soil acidity and nitrogen on yield
and elemental concentration of bush bean, carrot, and lettuce. J. Amer. Soc. Hort.
Sci. 107:740-744.

Hipp, B.W. 1978. Response by carrots to nitrogen and assessment of nitrogen status by
plant analysis. HortScience. 13(1):43-44.

Hipp, B.W., and OJ. Gerard. 1971. Influence of previous crop and nitrogen
mineralization on crop response to applied nitrogen. Agron. J. 63:583-586.

Hochmuth, G.J. 1994. Efficiency ranges for nitrate-nitrogen and potassium for vegetable
petiole sap quick tests. 4:218-222.

Hochmuth, G.J., J .K. Brecht, and M.J. Bassett. 1999. Nitrogen fertilization to maximize
carrot yield and quality on a sandy soil. HortScience 34(4):641-645.

Kelley, W.T. (ed). 1998. Carrot tops & tips. Extension & Research News About
Georgia Carrots. Univ. of Georgia, College of Agric and Environ. Sci. Coop. Ext.
Serv. & Agric. Res. Serv. December 1998. Vol. 1 No. 2.

Marshner, H. 1998. Mineral nutrition of higher plants. 2nd. ed. Academic Press, San
Diego, CA.

McCollum, T.G., S.J. Locascio, and J.M. White. 1986. Plant density and row
arrangement effects on carrot yields. J. Amer. Soc. Hort. Sci. lll(5):648-651.

McGiffrn, M., J. Nunez, T. Suslow, and K. Mayberry. 1997. Carrot production
in California. Vegetable Research and Information Center: Vegetable Production
Series. Publ. 7226. Division of Agric. & Nat. Res., Univ. of California, Oakland.
(Available on-line at http://anrcatalog.ucdavis.edu/spicials.ihtml.) (Verified 28
April 2003.)

Michigan Dept of Agric. 2002. Michigan carrots. [Online]. Available at
http:// www.michigan.gov/mda/0,l607.7-125-15 70-13066--,00.htm1 (verified 29
April 2003).

Michigan State University and Michigan Dept. of Agric. 2000. 2000 FQPA-targeted
pesticide residue study presented to the EPA. (Available on-line at

http://www.cips.msu.edu/cips/residuereports/residue.htm.) (Verified 28 April
2003.)

Mills, HA. 2001. Carrot Daucus carota var sativus. College of Agric. & Environ.
Sci., Dep. of Horticulture Univ. Georgia. [Online]. Available at
httpzl/www.uga.edu/vegetable/carrot.htm (verified 28 April 2003).

 

41

Sanderson, K.R., and J .A. Ivany. 1997. Carrot yield response to nitrogen rate. J. Prod.
Agric. 10:336-339.

Soil Survey Montcalm County. 1960.

Stevenson, A.B., and J. Chaput. 1998. Carrot Insects. Factsheet. Publ. 93-077. Ministry
of Agriculture, Food and Rural Affairs. Ontario, Canada. (Available on-line at
http://www.gov.on.ca/OMAFRA/english/crops/facts/93-077.htm.) (Verified 29
April 2003.)

Suojala, T. 2000. Pre- and postharvest development of carrot yield and quality. Ph.D.
diss. Univ. of Helsinki, Finland (Diss. Fin-21500 Piikkio). (Available on-line at
http://ethesis.helsinki.fi/iulkaisut/maa/kastu/vk/suoiala/geandpohtm.) (Verified
28 April 2003.)

Tremblay, N., H. Scharpf, U. Weier, H. Laurence, and J. Owen. 2001. Nitrogen
management in field vegetables: a guide to efficient fertilization. [Online].
Available at http://re52.agr.ca/stiean/publication/bulletin/nitrogen-azote e.htm
(verified 12 Feb 2004).

USDA, NASS.2002. USDA-NASS Agricultural Statistics 2002. (Available on-line at
http://www.usda.gov/nass/pubs/agrOZ/acr002.htm.) (Verified 28 April 2003.)

USDA. NASS. 2002. Michigan Agricultural Statistics 2001-2002. Lansing, Mi.

USDA Pest Management Centers. 1999. Carrots in Michigan: Overview. USDA Crop
Profiles, Michigan. [Online]. Available at
http://pestdata.ncsu.edu/cropprofiles/docs/micarrots.htm1 (verified 28 April 2003).

Walworth, J .L. 1998. Crop production and soil management series: Field crop
fertilizer recommendations for Alaska: vegetables. Bull. FGV-00643. Alaska
Coop. Ext., Univ. of Alaska, Fairbanks. (Available on-line at
http://www.uafedu/coop-ext/publications/freepubs/.) (Verified 28 April 2003.)

 

Warncke, DD. 2000. Nutrient management for wet soil conditions. Vegetable Advisory
Team Alert. 15(5):4. (Available on-line at
http://www.ipm.msu.edu/CAT00 veg/V05-24-00.htm.) (Verified 6 Feb 2004.)

Warncke, DD. 1996. Soil and plant tissue testing for nitrogen management in carrots.
Commun. Soil Sci. Plant Anal. 27(3&4):597-605.

Warncke, D.D., D.R. Christenson, L.W. Jacobs, M.L.Vitosh, and EH. Zandstra. 1992.
Fertilizer recommendations for vegetable crops in Michigan. Michigan State

University Ext. Bull. E550B. Michigan State Univ., East Lansing.

Wehner, T.C., C. Barrett, and P.W. Simon (ed). 2003. Vegetable cultivar

42

descriptions for North America: carrot, lists 1-26 combined. Dep. of Horticultural
Sci. North Carolina State Univ., Raleigh. [Online]. Available at
http://cuke.hort.ncsu.edu/cucurbit/wehner/vegcult/carrot.html (verified 29 April
2003)

White, J .M., and J .O. Strandberg. 1978. Early root grth of carrots in organic soil. J.
Amer. Soc. Hort. Sci. 103(3):344-347.

Zandstra, B.H., and DD. Warncke. 1993. Interplanted barley and rye in carrots and
onions. HortTechnology 3(2):214-218.

Zandstra, B.H., D.D. Warncke, E.J. Grafuis, and ML. Lacy. 1986. Carrots commercial
recommendations. Michigan State University Ext. Bull. E1437. Michigan State
Univ., East Lansing. (Available on-line at
http://www.msu.edu/~zandstra/extbult/carrots1986.html.) (Verified 03 January
2004)

43

Chapter II

Spectral Measurements of the Carrot Canopy as Related
to Nitrogen Status of the Crop

Introduction

As early as 1952, Moss and Loorrris measured absorptance spectra of individual
leaves using an integrating sphere (Moss and Loomis, 1952). By 1956, the pioneering
research of Colwell was detecting field-wide loss of plant vigor to disease from the lofty
view of aerial photography (Shanahan et al., 2001). In 2002, Oklahoma State University
(OSU) scientists perfected the Greenseeker, an integrated sensing and application system.
This field-scale variable rate applicator, built in cooperation with Ntech Industries Inc.,
calculates N rates in fractions of a second, using remote sensing, and variably applies N
as it travels across the field. It has already been shown to increase yield and decrease N
applications (Dept. of Plant and soil Sciences, OSU, 2004).

Reduction of N applications for the purpose of protecting groundwater from
leaching nitrogen (N) (Blackmer and Schepers, 1994; Flowers et al., 2003a) and
promoting economic stability through efficient use (Flowers et al., 2001; Flowers et al.,
2003b) has been the focus of both research and equipment development. Soil properties,
landscape position, and disposition of previous N applications cause available N for plant
uptake to vary spatially (Flowers et al., 2003b). To address the question of efficient use,
knowing that conditions vary spatially and temporally across fields, suggests that
intensive sampling is necessary. The fundamental field element “defined as the area

which provides the most precise measurement of the available nutrient and where the

44

level of that nutrient changes with distance” is seldom larger than 1 m2 (Dep. of Plant and
Soil Sciences Oklahoma State University, 2004). Physical sampling of soil and tissue at
such an intensive level is impractical (Aparicio et al., 2000; Blackrner et al., 1994) and
compromises repeat sampling. Technological advancements, such as variable rate
fertilizer applicators, made precision management on a large-scale possible, but created
the need for better methods of assessing within-field variability. Remote sensing
technology is making tremendous strides and seems to be the answer to the need for
intensive sampling as described above. Remote sensing has made it possible to measure
many plants at once for a variety of parameters (Blackmer et al., 1996a). Sensors such as
the OSU Greenseeker, attached to variable rate applicators, have made real-time
assessment of nutrient requirements with almost simultaneous application a reality. In
addition to nutrient management, remote sensing may be used to monitor diseases and
crop damage (Flowers et al., 2001). Automating measurements by mounting sensors on
mobile overhead sprinkler systems may facilitate continuous monitoring and detection of
changes in nutrient as well as water sufficiency. Early estimates of yield in wheat and

corn have also been successful using remote sensing (Flowers et al., 2003a, 2003b).

Lillesand and Kiefer (2000) define remote sensing as both a science and an art. It
is the gathering of information about an object, area, or phenomenon using a device not
in contact with the target, and includes both the collection and analysis of the data
(Lillesand and Kiefer, 2000). The discussion, herein, is limited to spectral reflectance

data obtained from portable (hand-held) radiometers, and aerial photography.

Remote sensing is a spatial and temporal measurement. Spectral radiance

collected at different dates provides different information about the system. Early season

45

reflectance is primarily influenced by soil characteristics with different soils having
different spectral characteristics. As the season progresses, plant characteristics
increasingly dominate spectral information (Chang et al., 2003). Osborne et a1. (2002)
found that important wavebands used for predicting N content, biomass and grain yield
change with sampling date and that such changes may be attributed to temporal variations
in percent ground cover and growth stages. The energy incident on a crop canopy varies
as well over the day and through the season as a function of the solar zenith angle

(Epiphanio and Huete, 1994).

Surface reflectance, internal scattering, and attenuation of sun light by a leaf is
greatly affected by its physical and chemical characteristics (Al-Abbas et al., 1974; Maas
and Dunlap, 1989). Nutrient deficiencies cause visible abnormalities in pigmentation as a
result of the reduction of leaf chlorophyll content, the size and shape of leaves, and the
photosynthetic rate linked to the amount of absorbed radiation (Al-Abbas et al., 1974;
Maas and Dunlap, 1989; Masoni et al., 1996), resulting in measurable changes in
reflectance, absorptance and transmittance. For example, N treatment effects are
attributed to differences in leaf area, crop biomass, soil cover, plant height, and
chlorophyll concentration (Blackmer et al., 1996a), and interpretation of the information
gathered is based on the knowledge of the interaction of electromagnetic radiation with
the plant leaves and canopy (Maas and Dunlap, 1989). Chlorophyll and N concentration
influence reflectance in the visible blue, green and red wavelengths, where a reduction in
chlorophyll content results in increased reflectance. For example, N treatment at 45 kg
ha'1 should result in higher visible canopy reflectance compared to a treatment at 135 kg

ha". Vegetative cover and vigor directly influence reflectance in the NIR (Flowers et al.,

46

2003b). An increase in NIR reflectance is usually associated with increasing N. External
and internal reflectance and pigment content (mostly chlorophyll) affect the extent of
absorptance (Maas and Dunlap, 1989).

Many factors challenge the interpretation of spectral data at the ground and aerial
level: weeds, diseases, insect damage, water stress, varietal differences, plant nutrition,
soil background, sun angle, bidirectional information, and equipment irregularities
(Flowers et al., 2003a; Al-Abbas et al., 1974). Interpreting the interaction of irradiance
with canopy characteristics has been the focus of extensive research. Researchers have
developed vegetation indices to account for or eliminate factors that confound parameters
of interest by taking advantage of the high absorption of red wavelengths and the strong
reflectance of the NIR portion of the spectrum by photoactive tissue in plants which is
distinctive from soil and water (Wiegand et al., 1991). Several of the studies are
described in the following section.

The first objective of this study was to determine which reflectance measurements
correlated to various physical parameters typically used to evaluate the health of the
carrot crop. The second objective was to determine if reflectance measurements could be

used for in-season N management of healthy carrot tops.

Literature Review

Plant pigments are of particular interest, in remote sensing, because it may be
possible to detect nutrient deficiency, salinity, stress, and other parameters at wavelengths
influenced by plant pigments (Maas and Dunlap, 1989). The greatest differences in

pigmentation are detected between 380 and 750 nm (Blackmer et al., 1994; Maas and

47

Dunlap, 1989). Maas and Dunlap (1989) using the spectral differences between normal,
etiolated, and albino corn (Zea mays L.) leaves, identified the individual spectral effects
of chlorophyll and carotenoid pigments, and the background cellular structure and water
content. They observed large concentrations of chlorophyll and carotenoids in normal
leaves that were totally absent from the albino leaves. The etiolated leaves contained
intermediate levels of B-carotene, the most abundant carotenoid pigment in higher plants
(Maas and Dunlap, 1989). Their study showed that chlorophyll and carotenoid pigments
controlled the visible optical properties in normal leaves, and that leaf reflectance in the
visible band is controlled by absorption by the chlorophylls. The greatest absorptance
(low reflectance) was observed at 430 and 670nm wavelengths similar to the absorptance
peaks of extracted chlorophyll, which exhibits a sharp peak at 670nm (Maas and Dunlap,
1989). A gradual decrease of absorptance (increase in reflectance) between 482 and 550
nm is associated with carotenoids and was observed in both normal and etiolated leaves.
The increase in absorptance (decrease in reflectance) between 550-670nm is attributed to
“biological forms of chlorophyll” (Brown, 1972). Maas and Dunlap (1989) noted that
wavelengths at 550 and 670nm were affected by a combination of carotenoid and
chlorophyll pi grnents in corn. The optical properties of the etiolated leaves were
dominated by carotenoids along with the background optical characteristics associated
with the albino leaves. B-carotene was the influencing factor in the visible band. The
optical properties of albino leaves were dominated by the cell structure and water content
(Maas and Dunlap, 1989). Without pigments to absorb the irradiance, albino leaves
showed leaf reflectance similar to that observed in the NIR (800-1200nm) region: low

absorptance resulting in high reflectance (Maas and Dunlap, 1989).

48

Thomas and Gausman (1977) reported that 550 nm, where there was relatively
little absorption, was superior to 450 and 670 nm in relating leaf reflectance to either
chlorophyll or carotenoid concentration for eight different crops. Blackrner et a1. (1994),
Blackrner et al. (1996a), and Masoni et a1. (1996) also reported that the measure of
reflectance at wavelengths showing relatively little absorption (550 nm) provided the
most sensitive assessment of N status. N deficiencies showed little or no effect on
reflectance from any of the corn hybrids at 450 or 650 nm, which suggests that either
light was equally absorbed (even with deficiencies) or that some of the light was
transmitted through the leaf (Blackmer et al., 1994). Reflectance of the green
wavelengths peaks at 550 nm and is generally recognized as an indication of N status for
many agronomic crops (Blackmer et al., 1994). The greatest reflectance consistently
occurred with the lowest N rates because N deficiencies result in decreased amounts of
leaf chlorophyll that absorbs less light and results in greater reflectance (Blackrner et al.,
1994)

Other nutrients as well as N affect chlorophyll content and, therefore, the leaf
spectra (Al-Abbas et al., 1974; Masoni et al., 1996). It is important to summarize some
of these characteristics, especially in relation to N deficiency similarities. In corn,
chlorophyll-a was greatly affected, in order, by Fe, Mg, and Mn deficiencies, and to a
lesser extent by S deficiency. The order of severity of deficiency symptoms affecting
chlorophyll content varies with species (Masoni et al., 1996). While Fe, S, Mg, and Mn
all contributed to the reduction in chlorophyll that in turn resulted in decreased absorption
and increased reflectance, there was no correlation between the chlorophyll content and

mineral content. Lack of correlation may be due to the depleted uptake of other nutrients

49

in addition to treatments (Masoni et al., 1996). However, chlorophyll content is highly
correlated with leaf -N content (Wolfe et al., 1988; and Schepers et al., 1992). For each
mineral deficiency, there was first a reduction of leaf chlorophyll concentration and then
a decrease in spectral absorptance with a synergistic increase in reflectance. The best
correlation between leaf chlorophyll concentration and reflectance, transmittance and
absorptance was found at 555 and 700 run where Fe, Mg, Mn, and S were deficient
(Masoni et al., 1996). Blackrner et al. (1996a) found 550nm and 710nm better for
detecting N deficiency than other wavebands.

Al-Abbas et a1. (1974) studied the spectral effects of deficiencies of six nutrients
including N. He found that in corn, N deficiency resulted in the lowest chlorophyll
content followed in increasing order by Mg, S, K, Ca, and P, although, the highest
reflectance was displayed by K followed in order by Mg, N, S, P, and Ca. Potassium
deficient leaves had the lowest moisture content and were among the thinnest indicating
that reflectance may be closely related to leaf thickness and moisture content in addition
to pigment abnormalities. Maas and Dunlap (1989) noted that knowledge of leaf
thickness and water content is essential for determination of pigment concentration from
leaf reflectance at visible wavelengths. Walburg et a1. (1982) and Maas and Dunlap
(1989) also found that changes in external as well as cellular leaf structure, along with
pigment concentration, affected the spectral reflectance resulting from N treatments. In a
field study, Osborne (2002) noted that where P was deficient in corn there was an
increase in anthocyanin production causing purpling at leaf margins. Anthocyanin
strongly absorbs in the blue to green spectral region compared to the red spectral region.

However, in a greenhouse study, Milton et al. (1991) noted that P deficient leaves of

50

soybean plants grown in hydroponic solutions had higher reflectance in the green and
yellow portion of the visible band. Along with the presence of anthocyanin, reflectance
measurements indicated that NIR reflectance was important for predicting P stress during
the early season driven by the internal cell structure, but N concentrations could be
predicted throughout the growing season (Osborne et al., 2002). Al-Abbas et a1. (1974)
found that regardless of the deficiency, all deficient plants contained less chlorophyll than
the control (normal). The results indicate that chlorophyll has a dominant influence on
the spectral variation in the visible region of the spectrum. Variation at 550 nm seems to
best indicate N status, but other nutrient deficiencies are also expressed in this region.
Near infrared (N IR), reflectance and transmittance fi'om 750 to 1300 nm is
generally associated with leaf structure and morphology (Al-Abbas et al., 1974). At 780
to 810 nm, NIR is particularly sensitive to the presence of amino acids (R-NHZ), the
building blocks of protein, the presence or absence of which largely determine the N
content of the plant (Dep. of Plant and Soil Sciences Oklahoma State University, 2004).
As vegetative cover increases, NIR reflectance increases because multiple leaf layers
increase light scattering and reflectance (Walburg et al., 1982). Absorption in the NIR
region is characteristically lower than in the visible region of the spectrum. NIR
reflectance has been used to predict nutrient concentration, yield and crop density (A1-
Abbas et al., 1974; Chang et al., 2003; Osborne et al., 2002; Senay, 1998; Blackrner et
al., 1996a; Flowers et al., 2001; Flowers et al., 2003a). Chang et a1. (2003) noted that NIR
is inversely correlated to corn yield when measurements were taken before the third leaf
because soil moisture influenced the measurements. High reflectance at that time

indicated low soil moisture and eventual low yields. Beyond the second leaf, NIR

51

reflectance was positively correlated to yield. Osborne et a1. (2002) noted that green, red,
and NIR could predict N concentration in corn in June, but that in July N was better
estimated by NIR associated with canopy biomass. Al-Abbas et a1. (1974) found that
NIR and middle infrared (MIR) spectra significantly varied with treatment in corn. Leaf
age did not contribute to MIR reflectance variations as it did in the visible 530 and 640
nm wavebands, where pi grnent concentration covaried with leaf age. Maas and Dunlap
(1989) reported no qualitative differences in the NIR or MIR spectra among normal,
etiolated, or albino leaves; however, quantitative differences were notable at 1000 nm.
This is expected, since absorption by chlorophyll is very low in NIR and MIR regions.
Al-Abbas et al. (1974) associated low absorption (high reflectance) at 830, 940, and 1100
nm with high chlorophyll content. Lower absorption in NIR may protect plant pigments
from denaturation (Gates et al., 1965; Al-Abbas et al., 1974). Absorptance in this range,
with the same efficiency as in the visible region, would fiequently over heat plants and
irreversibly denature the proteins (Gates et al., 1965). Transmittance through normal
leaves was significantly less than through either the etiolated or albino leaves with
corresponding increase in reflectance and absorptance (Maas and Dunlap, 1989). Carlson
et a1. (1971) and Woolley (1971) noted that normal leaves are thicker than N deficient
leaves and quantitative differences at infrared as well as visible wavelengths can be
related to leaf thickness and water content.

Other nutrients besides N also can affect NIR reflectance. Al-Abbas et a1. (1974)
in their study of several nutrient deficiencies noted that at 830, 940, and 1100 nm, P and
Ca deficient corn leaves absorb less (reflect more) than normal leaves. Marshner (1998)

noted that P deficiency may result in higher chlorophyll concentration. P and Ca

52

 

deficiencies affected the chlorophyll concentration to a lesser extent than the other
deficiencies studied (Al-Abbas et al., 1974). Deficiency in S, Mg, K and N resulted in
much higher absorptance (lower reflectance) than the normal leaf. Higher than normal
absorption (lower than normal reflectance) is attributed to above normal heat content
within the leaves (Al-Abbas et al., 1974). Beyond the NIR, is the middle infrared region
(MIR 1350 to 2500 nm). Characteristically, with increasing N treatments, reflectance
decreases in the visible where radiation is absorbed by plant pigments and in the MIR
where radiation is absorbed by plant water. Reflectance increases in the NIR (Gates et
al., 1965; Al-Abbas et al., 1974).

Spectral measurements are often recorded by equipment as digital counts that are
proportional to the amount of reflected radiation from the target as in the case of
spectroradiometers. The same is true for aerial photography where the image, a
recording of reflected radiation, is digital or digitized and the digital numbers (DNs) of
each pixel can be enumerated. However, raw counts are difficult to use because
instrument response is typically not uniform over all wavebands, and the absolute scale is
dependent on factors such as sensor, illumination angles, and canopy arch (Blackmer et
al., 1996a). These inconsistancies can usually be avoided by referencing data to incident
or incoming radiation acquired using a reference panel (Blackmer et al., 1996a; Bausch,
1993; Williams et al., 2001; Shanahan et al., 2001; Walburg et al., 1982; Osborne et al.,
2002) or invariant object resulting in percent reflectance (Chang et al., 2003). The raw
counts have also been used directly in vegetation indices (Flowers et al., 2001; Flowers et
al., 2003a, 2003b). However, if vegetation indices are not calculated from percentages,

the reflectance results may not be consistent when images are compared over time (U .S.

53

 

Water Conservation Laboratory, 2004). Reflectance is the hardest value to obtain, but the
most valuable since it is characteristic of the surface itself and not affected by the
intensity of light shining on it (U .S. Water Conservation Laboratory, 2004). Rather than
use incident radiation or an invariant object, Blackrner et al. (1996a) standardized the raw
counts of reflected radiation using the highest-N-rate plots within a hybrid to give values

of relative reflectance where:

Digital Count
Reference Digital Count

 

Relative Reflectanc e =

[1]

Reflected radiation expressed as relative reflectance did not alter the interpretive
importance of the 550 nm and 710 nm wavebands. However, comparisons to NIR
wavebands resulted in inverse relationships rather than the expected positive relationship
(Blackmer et al., 1996a, 1996b). Most of the single-wavelength reflectance
measurements had significant hybrid effects and hybrid x N treatment effects. Blackrner
et al. (1996a, 1996b) found relative reflectance able to account for differences in
conditions between years, hybrids, soil fertility level, and instrumentation. Relative
reflectance can be used with aerial photography, as well as radiometric measurements.
When relative yield was also calculated in a similar manner, it was possible to evaluate
management areas with more than one hybrid where relative reflectance explained 94%
of the differences (Blackmer et al., 1996b). Use of relative reflectance with non-limiting
reference plots makes it possible to use less expensive equipment by internally calibrating
to a field situation (Blackrner et al., 1996a). Flowers et al. (2003a) also found that weeds,
variety and soil type confounded the relationship between GS-25 tiller density (TD) in
wheat and NIR digital counts. They used an approach similar to Blackrner et al. (1996a)

to determine the likelihood of predicting tiller density in wheat at GS-25 and GS-30

54

developmental stages critical to in-season N applications. Using digital counts obtained
from aerial photography, and modifying the relative reflectance model used by Blackrner

et al. (1996a, 1996b); Flowers et al. (2003a) developed the following relationship where:

NIR — NIR

lowest density

NIR

Relative Reflectanc e =

 

NIR [2]

highest density ‘ lowest density

Highest density and lowest density represent the NIR digital counts for the highest and
lowest tiller densities at the particular location. This relationship was applied to NIR
measurements within hybrid and location. Like Blackrner et al. (1996a, 1996b), when
NIR digital counts alone were regressed against tiller density at each location, significant
varietal and environmental differences were apparent (Flowers et al., 2001; Flowers et al.,
2003a). However, when relative tiller density was regressed against relative NIR
reflectance (Eq. 2), the slopes and intercepts of the equations were not significantly
different. Nor was the slope and intercept of the equation resulting from the regression of
data combined across all locations and hybrids significantly different from the equations

of the individual locations. Flowers et a1. (2001, 2003a) rearranged the linear regression

equation to produce the following equation that was used to predict tiller density:

TD — [(TDmax —TDm,n)x(NrR,,, —.07)/1.04]+TDm,.n [3]

predicted ‘-
TD means tiller density, and max and min represent the highest and lowest tiller densities
at the particular location. In their 2003 study, Eq. [3] correctly recommended N
applications, relative to tiller density, across hybrids and locations 85.5% of the time.
The TDmax, TDmin, NIRmax, and NIRmin must be determined for each soil type or variety.
Weed populations continue to be a problem and cannot be corrected with relative

parameters. They have to be physically kept to a minimum. Fields with good weed

55

control may be candidates for remote sensing while those with weeds are not (Flowers et
al., 2003a).

Vegetation indices reduce multiband observations (radiometric and digital image)
to a single numerical index (W iegand et al., 1991). This use of ratios has also been
shown to minimize some multiplicative effects while enhancing small increases in
vegetation coverage (Epiphanio and Huete, 1994), but they are influenced by sensor
calibration, sun and view angle, canopy variation, leaf optical properties, and canopy
background (Yoshioka et al., 2000). One of the earliest vegetation indices was simply
NR reflectance divided by red reflectance (Jordan, 1969). This Simple Vegetation Index
took advantage of the contrast between the NR low absorption (high reflectance) and the
high absorptance (low reflectance) of the red waveband by chlorophyll (Epiphanio and
Huete, 1994; Shanahan et al., 2001; US Water Conservation Laboratory, 2004). Areas of
dense vegetation will appear very bright in NR and very dark in red because only about
4% of the red waveband is reflected (US Water Conservation Laboratory, 2004). A
yellow leaf will appear much brighter than a healthy leaf in the red waveband where there
is little chlorophyll content to absorb the light. Both the healthy and yellowed leaves will
reflect light similarly in the NR (US Water Conservation Laboratory, 2004). 8-bit
digital images present brightness and darkness on a scale of gray shades between 0 for
black and 255 for white. For indices that subtract red from NR, materials with similar
NR and red brightness becomes dark. The soil, which usually reflects about the same
for both red and NR, becomes dark. If NR is brighter than red, the ratio will be larger
(i.e., brighter). With increased red absorptance, a smaller amount of red reflectance is

subtracted from NR, leaving a difference.

56

NDVI (Normalized Difference Vegetation Index), developed by Tucker (1979)

was intended to estimate green biomass where:

(NIR - red)

NDVI =
(NIR + red)

 

[4]

The difference divided by the sum compensates for differing amounts of incoming light

and is ideally suited for detecting subtle coverage differences in early crop stages or crops

 

h
under stress conditions. NDVI ranges from 0.0 to 1.0 with soil producing an NDVI value
of approximately 0.1 while dense vegetation gives a value of about 0.9. Aparicio et a1.
(2000), Epiphanio and Huete (1994), and Huete (1988) found that NDVI is highly
sensitive where the leaf area index (LAI) is between 0 and 2. At LAI greater than 3, 1:-

sensitivity to environmental changes diminishes (Aparicio et al., 2000; Epiphanio and
Huete, 1994), because once ground coverage by vegetation is complete red absorptance
saturates while NR reflectance gradually increases with increasing canopy density.
Flowers et al. (2003b) found that whole plant N at GS-30 in wheat had a relatively strong
relationship with individual bands or, spectral indices, especially NDVI, where there was
high biomass. NDVI produced an r2 = 0.69. At low biomass, they found a poor
relationship between whole plant N and spectral information because the amount of
canopy coverage did not relate spectral information to the whole plant N concentration
(Flowers et al., 2003b). NDVI could predict N rate 64% of the time where there was high
biomass. According to the Flowers et al. (2003b) study, NDVI was still among the best
estimators at all sites (12 = 0.61) when correlated to N uptake (whole plant N x biomass).
The index saturated at high GS-30 biomass (high N uptake) values making differentiation
by NDVI difficult, which may limit its usefirlness in predicting GS-30 N uptake (Flowers

et al., 2003b).

57

NDVI is affected by view angle, increasing as the angle moves from antisolar to
forward scattering. In addition, NDVI also increases as the solar zenith angle increases,
adding more depth to the canopy (Epiphanio and Huete, 1994), because there is more
absorptance of red and less reflectance to subtract from NR (U .S. Water Conservation
Laboratory, 2004). Aparicio et a1. (2000) found that NDVI is limited for use as a crop-
area indicator. Epiphanio and Huete (1994) found NDVI to be a sensitive grth index
for early crops or other sparse canopies, but it was influenced by factors other than
vegetation and angle, such as soil. However, Chang et a1. (2003) noted that including soil
data early on provides information about drainage, organic matter, and texture, which
later impacts yield (Chang et al., 2003). Bausch (1993) found that soil background color
significantly altered NDVI values throughout the vegetative growth period in com.
NDVI values are greater where vegetation covers dark soils than where the soil is light in
color (Chang et al., 2003). Soil type differences present a problem for remote sensing
because they commonly occur within a field (Flowers et al., 2003a).

In answer to limitations surrounding NDVI, Huete (1988) introduced the “L”

factor into NDVI to create the Soil Adjusted Vegetation Index where:

SAVI=[ NIR—red ](1+L) [5]

 

NIR + red + L

“L” adjusts for the different brightrresses of the background soil. The factor “L” made
SAVI less sensitive to red reflectance changes and more sensitive to NR changes,

especially for high amounts of vegetation (Epiphanio and Huete, 1994). By definition,
“L” varies from 0 to l; 1 represents low vegetation coverage and the adjustment factor
diminishes as the vegetation grows denser. However, 0.5 is often used as a reasonable

approximation when the amount of soil in the scene is unknown (U .5. Water

58

 

Conservaiton Laboratory, 2004). SAVI is a better estimator of LAI and biomass than
NDVI at high vegetation density, and has the opposite response of NDVI with regard to
view angle (Epiphanio and Huete, 1994). In contrast to NDVI, values start higher in
antisolar viewing and decrease as the angle moves to forward scattering. However, SAVI
is more sensitive to NR variation caused by sensor and sun geometry. NR has much
more interaction with the canopy due to scattering and transmission compared to the red
band. At high vegetation densities, SAVI is expected to be better correlated to NR-
related environmental variables, because in this range SAVI is more sensitive to NR
without saturation. This sensitivity also made SAVI more sensitive to view angle
variation induced by changes in sensor and sun geometry in medium to high density
alfalfa. NDVI tends to saturate at LAI greater than 3 (Aparicio el al., 2000; Epiphanio
and Huete, 1994).

A year after Huete (1988) developed the SAVI, Baret et a1. (1989) published

modifications to it, the Transformed Soil Adjusted Vegetation Index where:

a[NlR — (a . red)— b]

TSAVI = [red+(aaNIR)-(a*b)]

 

[6]

where a = the slope and b = the intercept of an equation fitted through a plot of NR vs.
red reflectance data for a variety of bare soil conditions: dry, wet, smooth, and rough
(Shanahan et al., 2001; Wiegand et al., 1991). Representation by only one condition,
such as dry soil, would account for only a short segment of the line, skewing the slope
(Wiegand et al., 1991). For this reason, Wiegand et a1. (1991) took periodic soil spectral
measurements throughout the season under various conditions. Using the parameters a
and b, the index value becomes exactly zero for all points on the soil line (Y oshioka et

al., 2000). The differences in the reflectance contributions from bare soil areas are exactly

59

proportional to the differences in soil brightness in both the red and NR bands. The
changing rate of NR reflectance to red reflectance is exactly the same as that of
background brightness, which is the slope of the soil line, a (Yoshioka et al., 2000).
The Green Normalized Difference Vegetation Index (GNDVI), developed by
Gitelson et a1. (1996), is still another vegetation index, but it makes use of the green

waveband in place of the red waveband:

(NIR - green)
(NIR + green)

 

GNDVI = [7]

GNDVI was correlated to corn yield more consistently throughout the season than NDVI
or TSAVI. Values obtained during rrrid grain filling stage would have the greatest
potential for estimating final grain yields over other indices (Shanahan et al., 2001).
GNDVI could prove useful in producing relative yield maps that depict spatial variability
in the field before harvest while there is still time to improve conditions (Shanahan et al.,
2001). Blackrner et a1. (1994), Schepers et a1. (1992), and Schepers et a1. (1996) found
that the green band, together with NR (GNDVI), is better at showing the variability in
leaf chlorophyll, N content, and grain yield compared to indices using the red band
(NDVI, SAVI, TSAVI). The green band where there is less absorptance provides the
most sensitive assessment of N status. NDVI values have been associated with crop
biomass accumulation, LAI, chlorophyll concentration in leaves, PAR absorbed by the
canopy, and crop yield. However, when chlorophyll content, fractional coverage, or LAI
reach moderate to high values, NDVI is apparently less sensitive to these parameters,
whereas GNDVI consistently exhibited the highest correlation.

Several wavelengths, relative reflectance, and vegetation indices have been

successful in estimating crop parameters such as nutrient status, crop coverage, and yield.

60

Success appears dependent on strict attention to spectral measurement protocol, an
understanding of the spectral measurements and relationship to plant structure, and the
limitations of the vegetation indices. While much of the field research has focused on
field crops, greenhouse research performed by Thomas and Gausmarm (1977) also
included several fi'uit and vegetable crops: cantaloupe (Cucumis melo L. cv reticulatus
Naud), cucumber (Cucumis sativus L.), lettuce (Lactura sativa L. cv capitata L.), and
spinach (Spinacia oleracea L.).

Guided by the research discussed above, this chapter presents the results of

 

correlating the field response to N treatments using conventional sampling to spectral
measurements to determine whether the N requirement for quality carrot tops is
manageable using remote sensing. The conventional parameters used to measure the
field response to N treatments were compared to individual wavelengths, NDVI, SAVI,

TSAVI and GNDVI.

Materials and Methods

Experimental Sites, Plot Design, Management Protocol, and Agronomic Sampling
Field studies were conducted at four locations during 2001 and 2002, in
Montcalm County, Michigan. In both years, plots were located at the Michigan State
University Montcalm Experiment Station on moderately well drained loamy sand to
sandy loam soil, of the Hillsdale-Spinks map unit (Hillsdale: coarse-loamy, mixed, mesic
Typic Hapludalfs, Spinks: sandy, mixed, mesic Psammentic Hapludalfs) (D.L. Mokrna,
personal communication, 2003). Diamond Cut and Goliath varieties were planted in both

years on flat beds in early May and harvested in mid-September. Each year plots were

61

also established on commercial carrot fields, at Sandyland Farms, on Plainfield Sand,
including a loamy substratum at the 2001 site, (mixed mesic Typic Udipsamments) (D.L.
Mokrna, personal communication, 2003). Asgrow B1 and Prime Cut 59 varieties were
planted at the 2001 site, and Sugar Snax 54 was planted at the 2002 site. These fields
were planted in mid-April on raised beds and harvested in mid-August. Barley was
planted between rows to protect emerging plants and killed off once the carrots were
established. Four replications of each of four N-treatrnents (45, 90, 135, and 180 kg M”)
were arranged in a randomized complete block design at all locations. Weeds were
controlled with linuron, and foliar blight was controlled with Chlorothalonil. A detailed

description is given in the Materials and Methods section of Chapter 1.

Reflectance and Agronomic Measurements

Plant and soil reflectance measurements were made using a MSR87, multispectral
radiometer (CropScan, Rochester, MN) equipped with the standard eight narrowband
interference filters centered at 460, 510, 560, 610, 660, 710, 760 and 810 nm. The
MSR87 is equipped with 8 up- and 8 down-facing channels designed for near
simultaneous measurements of incident irradiance and reflected radiance. This feature
ensures the accuracy of percent reflectance calculations by eliminating any variability of
sun angle or light conditions between the target and reference panel measurements.
Flashed opal glass, a cosine diffuser, covered the incident irradiance (up-facing) sensors,
while clear glass covered the reflected radiance (down-facing) sensors. The field of
view (FOV) was 28°. All measurements were viewed at nadir. The standard vegetation

filters of the MSR87 were evaluated for use in carrot by comparing them to

62

 

measurements from an SE590 spectroradiometer (Spectrum Eng., Denver, CO). The
SE590 is equipped with a silicon photodiode array measuring wavebands ranging from
365.7 to 1125.1 nm, each about 2.7 nm wide. Dual spectral measurements were taken at
both 2001 field locations. Peak responses of the spectroradiometer were comparable to
the waveband centers of the multispectral radiometer (Table l). Reflectance was similar
between the MSR87 and the SE590 in the visible spectra, but discrepancies were greater
in the NR spectra. Bandwidths of the MSR87 were wider in the NR region than in the
visible region, and therefore the average reflectance spanned a wider range. In addition,

the spectroradiometer measurements may have contained unknown error, because

Table 1. Comparison of measurements taken August 20, 2001 by the SE590 spectroradiometer and
CropScan multispectral radiometer. The canopy represented growth at 104 days after planting at the
Montcalm Experiment Station and approximately 127 days after planting at Sandyland.

 

 

 

 

 

MSR87 $12590 Reflectaneei
*
Comparable
Color Centered Range Peak Range MSR87 SE590
nm (%)
Montcalm Experiment Station (2001)

Green 560.1 556 - 564 556.2 556.2 - 564.8 7.7 8.1
Red 659.2 654 - 665 673 655.2 - 664.1 2.7 3.2
NIR 813.2 797 - 829 828.6 797.5 - 828.6 47.6 64.5

Sandyland (Deaner Rd.) (2001)

Green 560.1 556 - 564 556.2 556.2 - 564.8 11.7 12.3
Red 659.2 654 - 665 676 655.2 - 664.1 4.1 4.3
NIR 813.2 797 - 829 813 797.5 - 828.6 65.3 79.2

 

1 Range is based on the MSR87 that has wider bandwidths
* Reflectance % is the average of the range

downwheling irradiance and radiance from the canopy were measured separately. Any
wisp of cloud cover between readings could have induced error. Since the MSR87
measures both irradiance and radiance simultaneously, its reflectance measurements were

VICWCd as more accurate.

63

In the MSR87, incident irradiance and reflected radiance (W m'z) passing through
the filters is converted to electrical current by the detectors, amplified to millivolts (mV)
and quantized to 8-bit radiometric resolution by the analog-to-digital converter of the
Data Logger Controller (DLC). Additional software was used to apply temperature and
sun angle cosine corrections to the digital output and perform percent reflectance
calculations (CropScan, Inc., 1995). The viewing height was 2.55 m from the soil
surface providing an effective ground resolution diameter of 1.27 m, (i.e. a 1.27 m
composite measurement of plant canopy and soil reflectance).

In 2001, at the experiment station, reflectance measurements were taken weekly
beginning at planting and continuing until harvest. The intensive scanning schedule was
intended to track canopy development and determine a feasible future starting date when
the partial canopy would be large enough that the reflectance measurements provided
useable information. Measurements taken in 2002 were delayed until July 11, when the
carrot canopy was partially developed, and readings per plot were increased from 2 to 4
to account for canopy variability. The Deaner Rd. (2001) and Masters Rd. (2002) fields
were scanned weekly using the same protocol as at the experiment station. Table 2
describes site-specific information pertinent to reflectance measurements. The direction
of scanning was determined based on minimizing shadows by plants and the operator.

A Canon Powershot G1 digital camera was mounted alongside and at the same
height as the radiometer at sampling. Digital images were taken of at least one scanned
site per plot for a visual record of radiometric measurements and to determine percent
vegetation coverage. Ground resolution of the digital camera at a height of 2.55 m is 2.4

x 1.83 m. The images were cropped to the same ground resolution as the radiometer and

64

reclassified to quantify pixels as soil or vegetation using Erdas Imagine 8.5.

Table 2. Reflectance measurement protocol specific to individual field locations.

 

T

T

 

Experiment Stn Experiment Stn Deaner Rd. Masters Rd.
Range 1, 2001 Range 15, 2002 2001 2002
Row orientation East - West East - West North - South East - West
Scanning Dir. East East South West
Samples per plot 2 4 2 4
Sun Angle Range 204° - 39.4° 21.4° - 51° 27.1° - 43° 23° - 41.3°
Viewing Plane Principal Principal Orthogonal Principal
Planting Date May 8 May 7 Mid Apr Mid Apr
Harvest Date September 13 September 13 August 23 August 20
Begin Scanning June 13 July 11 June 13 July 11
End Scanning September 6 September 6 August 9 August 15

 

T Experiment Stn. refers to the Montcalm Experiment Station in Montcalm County.

Analysis of the Data

Relationships between the N treatments, petiole samples, and yield versus
reflectance measurements were evaluated using regression and general linear models
(SAS Inst. Inc., Release 8.2/2004). Data points of reflectance that were consistantly
recorded as outliers according to SAS were eliminated. Vegetation indices were applied
according to the equations set forth previously. The factor “L” in SAVI, Eq. 5, was
defined as 0.50 as a reasonable approximation of vegetation cover (U .8. Water
Conservation Laboratory, 2004), and in SAVISL as the soil line slope. The soil line
model required by SAVISL and TSAVI was developed using regression of the NR and
red reflectance measurements of the soil. The soil line was calculated separately for each
location, where possible. In 2001, spectral measurements of the soil were taken only once
at the Montcalm Experiment Station. Since the Sandyland location was already

established when the plots were staked, a large enough area of bare soil was no longer

65

available. The soil line equation used for Sandyland was the same line used with
Experiment Station spectra. Soils fi'om the two locations were similar in color, organic
matter content, and water holding capacity. The soil line for both the Montcalm
Experiment Station and Sandyland in 2002 was derived from on-site bare soil
measurements taken throughout the season. During the season conditions developed at
the Experiment Station that caused the soil line slope to shift. Generally differences in
soil type will alter the slope (Rondeaux et al., 1995); while wet and dry surface
conditions generate the full range of the regression line (Wiegand et al., 1991). During
rain events at the Montcalm Experiment Station, soils that had been disked were washed
and settled by the rain resulting in exposed sand granuals and stones. The settled soils
with exposed sand grains could have altered the appearance enough to change the slope
of the soil line. Therefore, the 2002 soil line regression equations were calculated using

weekly soil reflectance measurements taken up to peak canopy coverage.

Results and Discussion

N Status of Carrot Canopy as the Result of Soil-N Availability.

Individual Reflectance: Individual wavebands centered at 460, 510, 560, 610,
660, 710, 760, and 810 were compared to three variations of the N treatments using
regression analysis to determine which bands had the strongest relationship to the N
status of the carrot canopy, as a result of the soil N available. Each spectral measurement
was compared to N Treatment (the total of seasonal applications) to determine if there
was a point at which the canopy reflected the seasonal outcome. In addition spectral

measurements were compared to Applied N (the total amount of N applied as of the date

66

of the particular spectral measurement) and to Available N (the total amount of the
residual N fi'om the previous crop added to Applied N) to determine whether the temporal
nature of spectral measurements could characterize existing field conditions. The
purpose of comparing the three different variations of soil-N was to understand the nature
of the predictability of reflectance measurements for use as an in-season management
tool. Regression analysis showed the relationship of individual reflectance measurements
to Treatment, Applied N, and Available N on any specific date was best described by a
first order equation, similar to results of Flowers et al. (2003b).

In 2001, the earliest reflectance measurements were taken at the Montcalm
Experiment Station on May 18, ten days after planting; plants were barely emerging
(Table 3). Measurements in all wavebands were significantly correlated to Available N;
depicting a spurious relationship to soil rather than plant canopy. Alter May 18,
correlation of the reflectance measurements at 760 and 810 nm were no longer
significant. Visible bands continued to correlate with Available N through July 5, but
only accounted for 28 to 35% of the variation among reflectance measurements. At that
time, canopy coverage averaged 19%, and reflectance depicted the plant response to N
available from the previous season and the first application of 45 kg ha'1 N applied to all
plots. The coefficient of determination was low probably because soil was an important
component of the spectral measurements during early season. Table 3 also illustrates,
beginning with July 12 and lasting throughout most of the remaining season, an
unexpected lack of correlation of the reflectance measurements to any aspect of soil-N.
Average canopy coverage was 35% on July 12, after the second urea application had

been broadcasted at different treatment rates. Results from the variable plant emergence

67

Table 3. Linear regression coefficients of reflectance at individual wavelengths vs N; where N is
treatrrrent, applied N, or available N at the Montcalm Experiment Station, 2001, Diamond Cut variety.

 

 

 

 

 

 

 

 

 

 

 

Ave + + . r . .
Scan ‘ Veg trt app avarl trt app avarl trt app avarl
Date DAP’ Coy 460 nm 510 nm 560 nm
% :2
5/18 10 0 ns ------ 0.46” ns ...... 0.47“ ns ------ 0.48”
6/13 36 0 ns ------ ns ns ------ ns ns ------ ns
6/22 45 7 ns ...... 0.23’ ns ...... 0.31' ns ------ 0.28‘
6/28 51 11 ns ------ 0.31‘ ns ...... 0.32‘ ns ...... 0.33‘
7/5 58 19 ns ...... 0.30‘ ns ...... 0.33‘ ns ...... 0.35‘
7/ 12 65 35 ns ns ns ns ns ns ns ns ns
7/20 73 64 ns ns ns ns ns ns ns ns ns
7/26 79 85§ ns ns ns ns ns ns ns ns ns
8/2 86 94 ns ns ns ns ns ns ns ns ns
8/9 93 97 ns ns ns ns ns ns ns ns ns
8/17 101 96 ------------------ 0.27‘ 0.27‘ ns ns ns ns
9/6 121 95§ ------------------ ns ns ns 0.3 0.35‘ 0.34‘
610 nm 660 nm 710 nm
% :2
5/18 10 0 ns ...... 0.49“ ns ------ 0.49“ ns ...... 0.39‘
6/ 13 36 0 ns ------ ns ns ------ ns ns ------ ns
6/22 45 7 ns ...... 0.29‘ ns ...... 0.32‘ ns ------ ns
6/28 51 1 1 ns ...... 0.34‘ ns ...... 0.35‘ ns ...... 0.33’
7/5 53 19 ns ...... 0.35' ns ...... 0.35‘ ns ...... 0.34‘
7/ 12 65 35 ns ns ns ns ns ns ns ns ns
7/20 73 64 ns ns ns ns ns ns ns ns ns
7/26 79 85§ ns ns ns ns ns ns ns ns ns
8/2 86 94 ns ns ns ns ns ns ns ns ns
8/9 93 97 ns ns ns ns ns ns ns ns ns
8/17 101 96 0.47” 0.47“ 0.40“ ns ns ns ns ns ns
9/6 121 959' 0.45" 0.45“ 0.44“ 0.38‘ 0.38‘ 0.35‘ 0.3 0.37‘ 0.37‘
760 nm 810 nm
°/o :2
5/18 10 0 ns ------ 0.48“ ns ..... 0.48”
6/13 36 0 ns ------ ns ns ------ ns
6/22 45 7 ns ------ ns ns ------ ns
6/28 51 11 ns ------ ns ns ------ ns
7/5 58 19 ns ------ ns ns ------ ns
7/ 12 65 35 ns ns ns ns ns ns
7/20 73 64 ns ns ns ns ns ns
7/26 79 85§ ns ns ns ns ns ns
8/2 86 94 ns ns ns ns ns ns
8/9 93 97 ns ns ns ns ns ns
8/ 17 101 96 ns ns ns ns ns ns
9/6 121 959’ 0.25‘ 0.25‘ 0.25‘ ns ns ns
ftrt = Treatment, app = Applied N, avail = Available N. Urea was broadcasted: June 13, July 11, and

August 9.

O .0 0..

*DAP = Days after planting
§ Calculated approximation using NDVI rather than approximation developed from images

68

Significance of overall F -ratio at p $0.05, 0.01, 0.001. ns = Overall F-value is not significant.

due to heavy rains in May (20.7 cm) and equipment wheel damage to some beds became
apparent as the developing canopy began to dominate spectral measurements. The early
conditions resulted in gaps in the rows that may have interrupted the developing
correlation between N treatments and reflectance both in the visible and NR wavebands.
In addition, high residual NO3' from the previous crop and irrigation water, and sporadic
irrigation from the adjacent field may have skewed the planned in-season N availability.
Mean separation of the reflectance measurements confirmed that blocking was a
significant factor throughout most of the growing season. A week after the last N
application was broadcast August 9 at peak canopy coverage, correlation between
Treatment, Applied N, and Available N was significant. Although, regression analysis
typically resulted in low coefficient of determination, mean separation between
reflectance measurements did show significant differences between treatments for
reflectance measurements of wavebands centered at 610, 660, and 710 nm, and the
significant differences were in order of treatment (Table 4). Results for Applied N were
identical to Treatment since incremental applications were broadcast in proportions
similar to the seasonal treatment increments.

The 2001 Sandyland location (Table 5), planted for “cut and peel”, was seeded at
approximately four times the rate used at the experiment station. The mid-April planting
date and the barley cover facilitated plant establishment before the heavy rains in May,
resulting in a better stand than at the Montcalm Experiment Station. Until July 5, when
vegetation coverage was approximately 63%, results of the regression model reflected the
N from the previous season available at the time of planting (Chapter 1, Table 3), and

subsequent application of 45 kg ha'1 to all plots. Predictability was low much like the

69

Experiment Station early results. Soil and the senesced barley were important factors.

Table 4. Mean reflectance measurements as influenced by treatment or
applied N at the Montcalm Experiment Station, 2001, Diamond Cut

 

 

variety.
Treatment Wavelength August 17 September 6
Kg ha.I (11m) ---------- Reflectance % ----------
45 610 5.11‘ 5.40“
90 5.04“” 5.01ab
135 4.97“ 491°
180 4.96b 4.88"
p-value 0.03 0.02
45 660 ns 3.383
90 ns 3.17““
135 ns 3.13ab
180 ns 3.10b
p-value 0.04
45 710 ns 10.90'
90 ns 10.07°b
135 ns 9.98“
180 ns 9.89b
p-value 0.04

 

Mean values with the same letters are not significantly different at p = 0.05
based on HSD. Results for Treatment and Applied N are the same.
p-value is of the overall F -ratio. ns = Overall F-ratio is not significant.

It was too early to expect significant correlation to Treatment and Applied N since
differing rates of N had not yet been applied. Reflectance in the visible part of the
spectrum, at wavelength 560 nm, was the first to show significance with Available N on
June 22 at 43% vegetative coverage (Table 5). On July 12, approximately six days after
the first broadcast of urea at differing treatment rates, the regression model for the
waveband at 560 nm was significantly correlated to Treatment, and Applied N, as well as
Available N, and reflectance at 710 nm was significantly correlated to Applied N and
Available N. The mean separation of reflectance measurements (Table 6) showed

significant separation of treatment at 560 and 710 nm; however, predictability was less

70

 

Table 5. Linear regression coefficients of reflectance at individual wavelengths vs N; where N is '

treatment, applied N, or available N at Sandyland, 2001, Asgrow Bl and Prime Cut 59 varieties.

 

Ave

 

 

 

 

 

 

 

 

 

Scan Veg trtlr app+ avail:r tlt app avail trt app avail
Date DARi COV 460 nm 510 nm 560 nm
% 1'2
6/13 54 21 ns ------ ns ns ------ ns ns ------ ns
6/22 63 43 ns ------ ns ns ...... ns ns ...... 0.26‘
6/28 69 63 ns ...... 0.37' ns ...... 0.36‘ ns 0.34‘
7/5 76 63 ns ------ ns ns ------ ns ns ------ ns
7/12 83 89 ns ns ns ns ns ns 0.35‘ 0.40“ 0.46“
7/19 90 93 ns ns 0.26‘ ns ns ns 0.38‘ 0.36' 0.28‘
7/26 97 98° ns ns ns 0.71‘“ 0.75‘” 0.72’” 0.54” 0.56’” 0.54“
8/2 104 99° ns ns ns 0.64‘“ 0.68‘” 0.64’” 0.66‘” 0.70‘” 0.66‘”
8/9 111 99 ns ns ns 0.70‘“ 0.76‘” 0.75‘” 0.81‘“ 0.84‘” 0.83’”
8/17 119 99 ns ns ns 0.85‘" 0.85‘” 0.85‘” 0.90‘“ 0.90‘“ 0.90‘”
610nm 660nm 710nm
% 1'2
6/13 54 21 ns ------ ns ns ------ ns ns ------ ns
6/22 63 43 ns ------ ns ns ------ ns ns ------ ns
6/28 69 63 ns ...... 0.36‘ ns ...... 0.36‘ ns ...... 0.35‘
7/5 76 63 ns ------ ns ns ------ ns ns ------ ns
7/12 83 89 ns ns ns ns ns ns ns 0.25. 0.31.
7/19 90 93 0.48” 0.48“ 0.42“ ns ns ns 0.27‘ 0.25‘ ns
7/26 97 98° 0.74'” 0.78‘” 0.75‘“ 0.65‘” 0.70‘“ 0.68‘” 0.60‘“ 0.64‘" 0.62‘”
8/2 104 99° 0.74‘” 0.79‘“ 0.75‘” 0.64’” 0.68‘" 0.64‘” 0.60'” 0.65‘" 0.61‘”
8/9 111 99 0.82‘” 0.87‘“ 0.86'” 0.72‘” 0.78‘” 0.77‘” 0.77‘“ 0.82‘” 0.81‘“
8/17 119 99 0.89’” 0.89’” 0.89'“ 0.84‘" 0.84‘” 0.84‘” 0.89‘” 0.89‘“ 0.89‘”
760nm 810nm
% rz
6/13 54 21 ns ------ ns ns ------ ns
6/22 63 43 ns ...... 0.28‘ ns ...... 0.29‘
6/28 69 63 ns ...... 0.49“ ns ...... 0.49”
7/5 76 63 ns ------ 0.45“ ns ------ 0.46“
7/12 83 89 ns ns ns ns ns ns
7/19 90 - 93 0.32' 0.36‘ 0.42“ 0.33‘ 0.37' 0.43"
7/26 97 98° 0.43“ 0.48“ 0.52“ 0.44“ 0.49” 0.53"
8/2 104 99° 0.30‘ 0.34‘ 0.40“ 0.38' 0.42" 0.48“
8/9 111 99 ns ns ns ns ns ns
8/17 119 99 ns ns ns ns ns ns

 

Ttrt = treatrrrent, app = applied N, avail = available N. Urea was broadcasted: June 13, July 6, and
August 1.
:DAP = Days after planting
§ Calculated approximation using NDVI rather than approximation developed from images.
Significance of overall F -ratio at p $0.05, 0.01, 0.001. ns= Overall F-ratio is not significant.

0 O. .0.

71

" ’VG‘I‘I_*

lflT- ~

Table 6. Mean reflectance measurements as influenced by treatment or applied N at Sandyland, 2001,
Asgrow B1 and Prime Cut 59 varieties.

 

 

 

 

Treatment Wavelength July 12 July 19 JulyZ6 August 2 August 9 August 17
Kg ha'I (nm) Reflectance %
45 510 ns ns 4.10a 4.03a 4.40a 4.27a
90 ns ns 3.87b 3.85ab 4.01b 3.95b
135 ns ns 3.73b 3.68b 3.72c 3.57c
180 ns ns 3.72b 3.71b 3.78bc 3.55c
p-value 0.0001 0.0005 <0.0001 <0.0001
45 560 9.61b ns 11.49a 11.59a 12.5 la 12.74a
90 10.27ab ns 11.1 lab 11.01b 11.50b 11.86b
135 10.67a ns 10.60b 10.58b 10.530 10.55c
180 10.313 ns 10.6% 10.62b 10.52c 10.27c
p-value 0.004 0.0043 0.0004 <0.0001 <0.0001
45 610 ns 6.17a 6.99a 6.69a 7.553 7.82a
90 ns 5.98ab 6.53b 6.21b 6.74b 7.02b
135 ns 5.82ab 6.14c 5.91c 6.11c 6.18c
180 ns 5.70b 6.17c 5.92bc 6.10c 6.03c
p-value 0.013’ <0.0001 <0.0001 <0.0001 <0.0001
45 660 ns ns 3.75a 3.36a 3.82a 3.89a
90 ns ns 3.42b 3.13ab 3.42b 3.51b
135 ns ns 3.260 2.9% 3.16c 3.11c
180 ns ns 3.26c 2.97b 3.22c 3.10c
p-value <0.0001* 0.0012 <0.0001 <0.0001
45 710 12.45b ns 14.68a 14.46a 16.27a 16.60a
90 12.98ab ns 14.03ab 13.65b 14.77b 15.26b
135 13.71a ns 13.44b 13.19b 13.62c 13.62c
180 12.99ab ns 13.54b 13.25b 13.68c 13.37c
p-value 0.008 0.001 1 0.0008 <0.0001 <0.0001
45 760 ns 56.57b 59.34b ns ns ns
90 ns 62.01ab 65.92ab ns ns us
135 ns 66.92a 67.78a ns ns us
180 ns 63.52ab 67.52a ns ns ns
p-value 0.020 0.0128
45 810 ns 56.64b 59.24b 63.22b ns ns
90 ns 62.56ab 66.53ab 66.13ab ns us
135 ns 68.1 la 68.73a 67.93a ns ns
180 ns 64.22ab 68.37a 67.23ab ns ns
p-value 0.015 0.012 0.029

 

I Blocking was significant July 19: 610 nm p-value = 0.052; July 26: 660 nm p-value = 0.003.
Mean values with the same letters are not significantly different at p= 0.05 based on HSD.
p—value is that generated from overall F-ratio. ns = Overall F-ratio is not significant.

Results for Treatment and Applied N are the same.

72

than 50%. Correlation generally improved throughout the season as the canopy
developed, evidenced by the steady increase in the coefficient of determination to 0.90.
Reflectance measured at 610 nm also showed significant correlation with Available N,
Applied N, and Treatment as early as July 19, with increasing improvement over time. It
was not until July 26, when canopy coverage averaged 98%, that reflectance at 510 and
660 nm significantly correlated to Soil-N. NR reflectance measured at 760 and 810 nm
was significantly correlated with Available N as early as June 22 and to Treatment and
Applied N on July 19 at 93% average canopy coverage. Coefficient of determination was
highest on July 26 at 98% coverage. Thereafter, it decreased because the NR
wavebands could not detect N differences in biomass at full canopy coverage, and
healthy and yellow leaves appear the same. Mean separation of reflectance
measurements indicated only two incidences of significant blocking interference between
replications: July 19 at 610 nm and July 26 at 660 nm.

In 2002, reflectance measurements were delayed until July 11, based on 2001
early season low correlation results. At the Montcahn Experiment Station, the first
application of N, in differing rates, was broadcast on June 8, and early season canopy
development was normal. The Diamond Cut variety (Table 7) again showed unexpected
lack of significant correlation between reflectance measurements in any waveband and
Treatment, Applied N, or Available N until August 9. It was discovered that the
irrigation nozzles produced uneven spray patterns, and within treatments the canopy
height varied throughout the study. The uneven irrigation during June and July with
water containing elevated NO3' concentrations may have contributed to the varied canopy

height. The varied canopy height could have resulted in uncharacteristic and uneven

73

Table 7. Linear regression coefficients of reflectance at individual wavelengths vs N; where N is
treatment, applied N, or available N at the Montcalm Experiment Station, 2002, Diamond Cut variety.

 

Ave

 

 

 

 

 

 

 

 

 

t '1' .+ . .
Scan + Veg trt app avarl trt gap avarl trt app avarl
Date DAP‘ Cov 460nm 510nm 560nm
% :2
7/1 1 65 5 1§ ns ns ns ns ns ns ns ns ns
7/17 71 68 ns ns ns ns ns ns ns ns ns
7/24 78 86 ns ns ns ns ns ns ns ns ns
8/1 86 96 ns ns ns ns ns ns ns ns ns
8/9 94 94 ns ns ns ns ns ns 0.51” 0.51" 0.50“
8/15 100 98 ns ns ns 0.40“ 0.40“ 0.35' 0.75‘” 0.75‘” 0.73‘”
8/21 106 97 ns ns ns 0.34‘ 0.34‘ 0.34‘ 0.70‘” 0.70‘“ 0.70‘”
8/30 115 88 ns ns ns 0.50" 0.50“ 0.50“ 0.75‘” 0.75‘" 0.76‘”
9/6 122 93 ns ns ns 0.49“ 0.49“ 0.51” 0.77‘” 0.77‘“ 0.80‘”
610nm 660nm 710nm
% :2
7/11 65 51§ ns ns ns ns ns ns ns ns ns
7/17 71 68 ns ns ns ns ns ns ns ns ns
7/24 78 86 ns ns ns ns ns ns ns ns ns
8/1 86 96 ns ns ns ns ns ns ns ns ns
8/9 94 94 ns ns ns ns ns ns 0.40” 0.40” 0.40"
8/15 100 98 0.61‘” 0.61'" 0.61‘“ 0.27‘ 0.27‘ 0.26‘ 0.48” 0.48" 0.52”
8/21 106 97 0.48“ 0.48“ 0.51“ 0.23 0.23 0.25‘ 0.62‘” 0.62’” 0.67'”
8/30 115 88 0.64‘” 0.64‘” 0.68‘” 0.38" 0.38" 0.40” 0.70‘” 0.70'” 0.77'”
9/6 122 93 0.63‘“ 0.63‘” 0.68‘” 0.38‘ 0.38' 0.42“ 0.70'“ 0.70‘” 0.70‘”
760nm 810nm
% 1'2
7/11 65 51° ns ns ns ns ns ns
7/17 71 68 ns ns ns ns ns ns
7/24 78 86 ns ns ns ns ns ns
8/1 86 96 ns ns ns ns ns ns
8/9 94 94 ns ns ns ns ns ns
8/15 100 98 ns ns ns ns ns ns
8/21 106 97 ns ns ns ns ns ns
8/30 115 88 ns ns ns ns ns ns
9/6 122 93 ns ns ns ns ns ns

 

Ttrt = treatment, app = applied N, avail = available N. Urea was broadcasted on June 8, July 29,
August 24.

: DAP = Days after planting
§ Calculated approximation using NDVI rather than approximation developed from images.
Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant.

and

shadowing, measured as part of the reflectance that resulted in the lack of significant

74

 

correlation. On August 9, when canopy coverage averaged 94%, it was the visible bands
centered at 560 and 710 nm that first became significantly correlated to Treatment,
Applied N, and Available N. The second application of N at differing rates was 11 days
old and rain as of July 21(Table 2, Ch. 1) had eliminated the need for inigation with NO;;'
high water. On August 15, at peak canopy coverage averaging 98%, correlation between
the visible bands at 510, 610 and 660 nm and Treatment, Applied N and Available N was
first significant. Correlation remained significant in the visible bands for the remainder
of the season. NR wavebands centered at 760 and 810 nm were never significantly
correlated to soil-N. It may have resulted from the uneven canopy height that was further
complicated by early senescence from late season development of Aster Yellows and
Alternaria leaf blight. Correlation of spectral measurements to N treatments was best
described by reflectance measured in the visible bands at 560 and 710 nm with r2 values
ranging fiom 0.70 to 0.80. Mean separation of reflectance measurements (Table 8)
indicated that wavebands at 560 and 710 nm had the greatest separation between
treatments. Further, both the regression model and the mean separation of reflectance
measurements indicated that reflectance at 660 nm, where chlorophyll strongly absorbs
irradiance, was not a good indicator of differences in plant response to soil-N (1'2 values
ranged fi'om 0.27 to 0.42). Generally, where the regression model resulted in significant
correlation of less than 40%, the difference between treatments exhibited by the mean
separation of reflectance measurements was insignificant.

The Goliath variety, also planted at the experiment station and subject to the same
water regime, exhibited the same irregular canopy height. In addition, in early August

the plants were diagnosed with bacterial blight and Cercospora leaf spot which resulted in

75

Table 8. Mean reflectance measurements as influenced by treatment or applied N at Montcalm
Experiment Station, 2002, Diamond Cut variety.

 

 

 

 

Treatment Wavelength August 9 August 15 Augrgst 21 August 30 September 9
Kg ha’I (nm) Reflectance %
45 510 ns 2.463 ns 2.58a 2.64a
90 ns 2.26ab ns 2.40ab 2.45ab
135 ns 2.23ab ns 2.36b 2.39b
180 ns 2.1% ns 2.32b 2.35b
p-value 0.04 0.01 0.018
45 560 5.98a 5.62a 5.71a 5.56a 5.54a
90 5.81ab 5.30b 5.42b 5.25b 5.20b
135 5.76b 5.18bc 5.35b 5.16bc 5.09bc
180 5.72b 5.00c 5.26b 5.01c 4.91c
p-value 0.014 0.0003 0.0002 <0.0001* <0.0001*
45 610 ns 3.95a 4.208 4.22a 4.34a
90 ns 3.72ab 3.92ab 3.93b 4.0lab
135 ns 3.60b 3.83b 3.83b 3.90b
180 ns 3.48b 3.80b 3.74b 3.79b
p-value 0.006 0.015 0.0017 0.003
45 660 ns ns ns ns ns
90 ns ns ns ns ns
135 ns ns ns ns us
180 ns ns ns ns ns
45 710 8.96a 8.87a 9.06a 8.883 9.03a
90 8.69ab 8.4Sab 8.65ab 8.42b 8 45b
135 8.67ab 8.28ab 8.55b 8.29b 8.30bc
180 8.54b 7.86b 8.38b 8.04c 8.07c
p-value 0.053 0.009’ 0.003 <0.0001’ <0.0001*
45 760 ns ns ns ns ns
90 ns ns ns ns ns
135 ns ns ns ns us
180 ns ns ns ns ns
45 810 ns ns ns ns ns
90 ns ns ns ns us
135 ns ns ns ns ns
180 ns ns ns ns ns

 

I Blocking was significant August 15: 710 nm p-value = 0.04; August 30: 560 nm p-value = 0.03,

710 nm p-value = 0.004; September 6: 560 nm p-value = 0.03, 710 nm p-value = 0.02.
Mean values with the same letters are not significantly different at p= 0.05 based on HSD.

p-value is that generated from overall F -ratio. ns = Overall F -ratio is not significant.
Results for Treatment and Applied N are the same.

76

considerable damage to the petioles and leaves. Table 9 illustrates the limited

significance, in the visible wavebands centered at 560, 610, 660, and 710 nm, and only

Table 9. Linear regression coefficients of reflectance at individual wavelengths vs N; where N is
treatment, applied N, or available N at the Montcalm Experiment Station, 2002, Goliath variety.

 

 

 

 

 

 

 

 

 

 

Ave + t . + . .
Scan + Veg trt app avarl trt app avarl trt app avarl
Date DAP+ Cov 460 nm 510 nm 560 nm
% r2
7/11 65 46§ ns ns ns ns ns ns ns ns ns
7/17 71 66 ns ns ns ns ns ns ns ns ns
7/24 78 80 ns ns ns ns ns ns ns ns as
8/1 86 96 ns ns ns ns ns ns ns ns ns
8/9 94 93 ns ns ns ns ns ns ns ns ns
8/ 15 100 96 ns ns ns ns ns ns ns ns ns
8/21 106 94 ns ns ns ns ns ns 0.38" 0.38” ns
8/30 1 15 87 ns ns ns ns ns ns 0.46” 0.46“ ns
9/6 122 90 ns ns ns ns ns ns ns ns ns
610 nm 660 nm 710 nm
% 1'2
7/11 65 46§ ns ns ns ns ns ns ns ns ns
7/17 71 66 ns ns ns ns ns ns ns ns ns
7/24 78 80 ns ns ns ns ns ns ns ns ns
8/1 86 96 ns ns ns ns ns ns ns ns ns
8/9 94 93 ns ns ns ns ns ns ns ns ns
8/15 100 96 ns ns ns ns ns ns 0.27. 0.27. ns
8/21 106 94 0.34‘ 0.34‘ ns ns ns ns 0.47” 0.47” ns
8/30 115 87 0.55‘” 0.55‘” ns 0.32‘ 0.32‘ ns 0 51" 0.51” ns
9/6 122 90 ns ns ns ns ns ns ns ns ns
760 nm 810 nm
% :2
7/11 65 46§ ns ns ns ns ns ns
7/ 17 71 66 ns ns ns ns ns ns
7/24 78 80 ns ns ns ns ns ns
8/ 1 86 96 ns ns ns ns ns ns
8/9 94 93 ns ns ns ns ns ns
8/ 15 100 96 ns ns ns ns ns ns
8/21 106 94 ns ns ns ns ns ns
8/30 115 87 ns ns ns ns ns ns
9/6 122 90 ns ns ns ns ns ns

 

Ttrt = treatment, app = applied N, avail = available N. Urea was broadcasted on June 8, July 29, and
August 24. 3: DAP = Days after planting
3 Calculated approximation using NDVI rather than the approximation deve10ped from images.

” ... Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant.

77

significantly correlated to Treatment and Applied N. Reflectance at 710 nm was
significantly correlated on August 15, August 21, and August 30. Reflectance at 560 and
610 nm was significantly correlated on August 21 and August 30 and at 660 nm only on
August 30. Even though correlation was significant on the indicated dates in the visible
wavebands, the predictability was marginal, possibly due to the now diminished canopy
where soil reflectance once again was a significant part of the signal. The mean
separation of reflectance measurements (Table 10) indicated that significant treatment
differences could be detected only where the regression model could explain 40% of
reflectance variability. While increased use of fungicide promoted recovery and new
growth toward season end, the canopy never fully recovered. Structure and morphology
(Al-Abbas et al., 1974) of the canopy were affected, resulting in the irregular canopy
coverage as evidenced by the lack of correlation in the regression model and the
unordered separation of treatments (Table 10).

In 2002 the Sandyland location (Table l 1), again, was planted at a high
population rate for the “cut and peel” market. The seeding rate was almost double that at
the Experiment Station and the canopy was visibly denser. The mean separation of
reflectance measurements (Table 12) indicated blocking was not a significant factor any
time during the season. Regression analysis (Table l 1) resulted in significant correlation
first exhibited on July 17, when canopy coverage averaged 92%, and the first broadcast
of urea, at differing rates, was about two weeks old. Visible wavebands centered at 610
and 710 nm, and NR wavebands centered at 760 and 810 nm, were significantly
correlated with Treatment, Applied N, and Available N. In 2001, the visible band at 560

nm showed significance a week earlier at 89% coverage. As of July 24, visible

78

Table 10. Mean reflectance measurements as influenced by treatment or applied N at the Montcalm

Experiment Station, 2002, Goliath variety.

 

 

 

 

Treatment Wavelength August 9 August 15 August 21 August 30 September 6
Kg ha'I (nm) Reflectance %
45 510 ns ns ns ns ns
90 ns ns ns ns us
135 ns ns ns ns ns
180 ns ns ns ns ns
45 560 ns ns 4.813 4.703 ns
90 ns ns 4.59ab 4.4lab us
135 ns ns 4.623b 4.47ab ns
180 ns ns 4.48b 4.2% ns
p-value 0.051 0.01 1
45 610 ns ns ns 3.57a ns
90 ns ns ns 3.43ab us
135 ns ns ns 3.36ab us
180 ns ns ns 3.23b ns
p-value 0.017
45 660 ns ns ns ns ns
90 ns ns ns ns ns
135 ns ns ns ns us
180 ns ns ns ns ns
45 710 ns ns 7.83a 7.63a ns
90 ns ns 7.63ab 7.24ab ns
135 ns ns 7.503b 7.213b ns
180 ns ns 7.30b 6.96b ns
p-value 0.046 0.015
45 760 37.18ab ns ns 31.103b ns
90 33.81b ns ns 27.1% us
135 39.143 ns ns 32.383 us
180 37.023b ns ns 29.7lab ns
p-value 0.044 0017*
45 810 38.86ab ns ns 32.96ab ns
90 35.56b ns ns 29.05b us
135 40.993 ns ns 34.523 ns
180 38.84ab ns ns 31.73ab ns
p-value 0.045 0.0161

 

T Blocking was significant August 30: 760 nm p-value = 0.017, 810 nm p-value = 0.015.
Mean values with the same letters are not significantly different at p= 0.05 based on HSD.

p-value is that generated from overall F-ratio. ns = Overall F-ratio is not significant.

Results for Treatment and Applied N are the same.

Table 11. Linear regression coefficients of reflectance at individual wavelengths vs N; where N is
treatment, applied N, or available N at Sandyland, 2002, Sugar Snax variety.

 

 

 

 

AVG + r .t , ,
Scan * Veg trt app avarl trt app avarl trt app avarl
Date DAP“ Cov 460nm 510nm 560nm

% 1‘2
7/11 82 91§ ns ns ns ns ns ns ns ns as
W” 88 92§ ns ns ns ns ns ns ns ns us
704 95 92 0.62‘” 0.62‘“ 0.64‘” 0.76‘” 0.76‘” 0.78‘” 0.72’” 0.72‘” 0.70‘”

8/1 103 97 0.51“ 0.51“ 0.49“ 0.54” 0.54 0.54“ 0.34‘ 0.34‘ 0.33‘
8/9 111 96 0.44" 0.44“ 0.43“ 0.47” 0.47 0.47“ 0.39‘ 0.39‘ 0.38'

 

 

 

8/15 117 99 0.67‘” 0.67'” 0.66‘“ 0.60‘” 0.60 0.59‘" 0.40" 0.40“ 0.40“
610nm 660nm 710nm
% 1'2
7/11 82 91§ ns ns ns ns ns ns ns ns ns
7/17 88 92§ 0.30‘ 0.30‘ 0.30‘ ns ns ns 0.34‘ 0.34‘ 0.34‘
7/24 95 92 0.81’" 0.81‘” 0.82’” 0.78‘“ 0.78'“ 0.80'” 0.77‘“ 0.78‘” 0.78‘”

8/1 103 97 0.54“ 0.54” 0.54" 0.61“ 0.61
8/9 111 96 0.47‘ 0.47‘ 0.47’ 0.58‘“ 0.58

0.61’” 0.46“ 0.46“ 0.46“
0.58‘" 0.31‘ 0.31‘ 0.30‘

 

 

 

8/15 117 99 0.49“ 0.49“ 0.48“ 0.62‘” 0.62’“ 0.62’” 0.37‘ 0.37" 0.36‘
760nm 810nm
% :7
7/11 82 91§ ns ns ns ns ns ns
7/17 88 92° 0.43“ 0.43“ 0.42“ 0.45“ 0.45“ 0.45“
7/24 95 92 0.67‘” 0.66‘” 0.70‘“ 0.67‘" 0.67‘” 0.70‘”
8/1 103 97 0.54” 0.54“ 0.57" 0.64'“ 0.64‘” 0.67'”
8/9 111 96 0.29‘ 0.29‘ 0.33‘ 0.41” 0.41” 0.46”
8/15 117 99 0.24' 0.24' 0.25' 0.32‘ 0.32‘ 0.33‘

 

Ttrt = treatment, app = applied N, avail = available N. Urea was broadcasted on July 3 and July 29.
: DAP = Days after planting
fgalculated approximation using NDVI rather than approximation developed from images
.. Significance of overall F -values at p $0.05, 0.01, 0.001. as = Overall F-ratio is not significant.

wavebands centered at 460, 510, 560, and 660 nm were first significantly correlated with
all three variations of soil-N. In fact, the coefficient of determination peaked across
wavebands on July 24, unlike Sandyland in 2001 where r2 steadily increased, for the
visible wavebands, throughout the season as the canopy developed. Reflectance
measurements in 2002 appeared to indicate that the canopy was less responsive to N

applications compared to 2001. Results shown in Table 12 confirmed the condition,

80

 

Table 12. Mean reflectance measurements as influenced by treatment or applied N at Sandyland, 2002,
Sugar Snax variety.

 

 

 

 

Treatment Wavelength July 17 July 24 Agust l Arggsn 9 August 15
Kg ha'I (11m) Reflectance %
45 510 ns 3.213 3.133 2.983 3.093
90 ns 2.77b 2.923b 2.863b 2.973
135 ns 2.66b 2.94ab 2.933 2.973
180 ns 2.54b 2.70b 2.67b 2.74b
p-value 0.0004 0.007 0.004 0.002
45 560 6.243 6.963 ns 7.07ab 7.193
90 5.88b 6.43b ns 6.97ab 7.013b
135 6.123b 6.3% ns 7.113 7.08ab
180 5.97ab 6.20b ns 6.60b 6.59b
p-value 0.025 <0.0001 0.031 0.019
45 610 4.523 5.343 5.003 4.963 4.973
90 4.06b 4.60b 4.723b 4.73ab 4.7lab
135 4.17ab 4.35b 4.723b 4.843 4.79ab
180 4.1% 4.14b 4.2% 4.3% 4.35b
p-value 0.010 0.0003 0.008 0.004 0.007
45 660 ns 3.723 3.093 2.923 2.823
90 ns 2.94ab 2.7lab 2.69ab 2.633
135 ns 2.5% 2.63b 2.723b 2.623
180 ns 2.42b 2.38b 2.45b 2.36b
p-value 0.002 0.004 0.003 0.002
45 710 9.713 10.963 10.753 ns 10.383
90 9.01b 9.91b 10.49ab ns 10.053b
135 9.3 lb 9.67b 10.603 ns 10.253
180 9.05b 9.33b 9.7% ns 9.44b
p-value 0.0006 <0.0001 0.014 0.016
45 760 39.443 43.13b 43.96b ns ns
90 41 .653b 45 .053b 46.523b ns us
135 46.003 49.303 49.363 ns ns
180 45.123b 50.043 49.283 ns ns
p-value 0.027 0.004 0.017
45 810 39.58b 43.77c 44.74b ns ns
90 41 .89ab 45.89bc 47.433b ns ns
135 46.583 50.54ab 50.813 ns ns
180 45.643b 51.333 50.963 ns ns
p-value 0.019 0.003 0.006

 

Mean values with the same letters are not significantly different at p= 0.05 based on HSD.
p—value is that generated from overall F-ratio. ns = Overall F-ratio is not significant.
Results for Treatment and Applied N are the same.

81

where separation of treatments was less significant and not in order. In both years the
last two applications were 26 days apart. The 2002 field (Table 11) had a 2 to 6% slope
and fiequent inigation may have resulted in some drainage away from the plots resulting
in runoff of broadcasted N. Instead of the coefficient of determination continuing to
increase until season end, it decreased as if soil-N was “used up” and the canopy stressed
across all treatments before the last application on July 29, as evidenced by the drop in
correlation to spectral measurements. Approximately two weeks following the final
application, r2 in the visible bands increased showing the effects of the N applied on July
29. Likewise, the mean separation of reflectance measurements (Table 12) indicated
significant and orderly separation of treatments. Correlation of NR reflectance at 760
and 810 nm continued to decrease in significance at full canopy coverage similar to 2001.
Overall canopy condition was better at Sandyland during the two-year study than
at the Experiment Station. Seeding rate affected density of the canopies and disease
reduced the plant vigor at the Experiment Station. Canopy condition affected the ability
of NR wavebands to correlate with soil-N when variables other than N treatments
reduced plant N uptake or when full coverage eliminated the use of biomass as a means
of evaluating treatments. There was little correlation of reflectance at 760 and 810 nm to
soil-N at the Experiment Station while at Sandyland correlation was significant until full
canopy coverage. NR is unable to distinguish between coloring due to chlorophyll
concentration (U .S. Water Conservation Laboratory, 2004); and at full canopy coverage
biomass was no longer variable. However, the visible wavebands, even with blight at the
experiment station, were able to provide some predictability about the plant response to

soil-N. The visible bands centered at 560 and 710 nm were the earliest to correlate with

82

soil-N and generally remained significant throughout the season explaining as much as
89-90% of the difference between treatments. The consistent results in the 560 and 710
nm wavebands at the various locations, and in both years, are in agreement with Thomas
and Gausman (1977), Blackrner et al. (1994), Blackrner et al. (1996a), and Masoni et al.
(1996) who found wavebands centered at 550 and 710 nm better for detecting N
deficiency than other wavebands. At 550nm there was relatively little absorption by
plants; therefore, a greater percentage of the irradiance was reflected providing the most
sensitive assessment of N which depicted a larger separation between treatments. In
addition, bands centered at 510 and 610 were also significantly correlated, although
significance generally lagged by one to two weeks. The bluer green at 510 nm and the
orange-red at 610 nm exhibited significance equal to or surpassing the wavelengths at
560 and 710 nm on certain dates, and may prove to be important to carrot. Plots of the
individual wavebands over time revealed a curvilinear relationship. However, the
prevailing conditions encountered over the season at most locations made statistical
analysis difficult.

Where soil-N was significantly correlated to canOpy reflectance during the early
season, the correlation was generally low or sporadic because soil was such a dominant
feature of target radiation. This was true across all wavebands. As split N applications in
differing rates were applied, and the canopy developed, soil-N showed significant
correlation within about two weeks of application (Tables 5, 7, 11) and remained constant
or decreased in varying degrees of significance depending on waveband as the N
application was “used-up” in about 25 days. Regression analysis indicated Applied N,

defined as the total amount of N applied to date, performed slightly better in nearly all

83

wavebands than Treatment and Available N at Sandyland 2001 (Table 5), but did not
show the same distinction at other locations. Reflectance was compared to Treatment to
deterrrrine if there was a point in canopy development that seasonal outcome could be
predicted. Predictability was best at 560 and 710 nm at approximately 89 to 94%
vegetative cover, 93 to 94 days after planting and generally after the second application.
Ninety-three to 94 days after planting was at least 35 days before harvest (Chapter 1,
Table 8). That is enough time to fertilize, realize results, and not accumulate excess N
before harvest. Only reflectance at 460 nm failed to correlate with soil-N at most
locations. Correlation of the NR wavebands to soil-N depended on the condition of the
canopy more than the visible bands. The results varied from location to location.

Selected Indices: Many factors affect reflectance measurements such as sun
angle, time of day, variety, and wetness of soil surface. A number of indices have been
developed to address some of these factors. Treatment, Applied N, and Available N were
compared to four indices: NDVI, SAVI, TSAVI, and GNDVI chosen because they have
returned a measure of success in other studies. SAVISL, is 3 variation of SAVI where L
= soil line slope.

Table 13 shows that at the Experiment Station, in 2001 the indices followed the
same trend in correlation as the individual reflectance measurements in Table 3. Indices
significantly correlated to Available N from the first measurement date through July 5.
The response was similar among indices, but unlike the individual wavebands, the
coefficient of determination decreased approaching July 5. The decreasing response may
be influenced by the lack of correlation of the reflectance at 810 run, one of the terms of

the indices equations. Thereafter, the indices showed less correlation with soil N than the

84

individual reflectance measurements. It should be noted, that on June 13, while the
components of the indices (560, 660, 810 nm) were not significant all the indices were

significantly correlated to Treatment, Applied N and Available N. Again on August 17,

Table 13. Linear regression coefficients of selected indices vs N; where N is treatment, applied N, or
available N at the Montcalm Experiment Station, 2001, Diamond Cut variety.

 

 

 

 

 

 

 

Ave t t .t . .
Scan Veg trt app avarl trt app avarl trt app avarl
Date DAPi Cov NDVI SAVISL SAVI

% 1'2
5/18 10 0 ns ........ 0.45“ ns ........ 0.45“ ns -------- 0.45“
6/13 36 0 ns ........ 0.29‘ ns ........ 029‘ ns 0.29‘
6/22 45 7 ns ........ 0.37" ns -------- 0.37“ ns ........ 0.37‘
6/28 51 11 ns ........ 0.31‘ ns -------- 0.31‘ ns -------- 0.31’
7/5 58 19 ns ........ 0.29‘ ns ........ 0 29‘ ns ........ 0.29'
7/12 65 35 ns ns ns ns ns ns ns ns ns
7/20 73 64 ns ns ns ns ns ns ns ns ns
7/26 79 85§ ns ns ns ns ns ns ns ns ns
8/2 86 94 ns ns ns ns ns ns ns ns ns
8/9 93 97 ns ns ns ns ns ns ns ns ns
8/17 101 96 ns ns ns ns ns ns ns ns ns
9/6 121 95 ns ns ns ns ns ns ns ns ns

TSAVI GNDVI

% 1'2
5/18 10 0 ns ........ ns ns 0.41“
6/13 36 0 ns ........ 0.34‘ ns ....... 0.29‘
6/22 45 7 ns ........ 0.33' ns -------- 0.33‘
6/28 51 11 ns -------- 0.30‘ ns -------- 0.31‘
7/5 58 19 ns ........ 0.28‘ ns 0.29'
7/12 65 35 ns ns ns ns ns ns
7/20 73 64 ns ns ns ns ns ns
7/26 79 85§ ns ns ns ns ns ns
8/2 86 94 ns ns ns ns ns ns
8/9 93 97 ns ns ns ns ns ns
8/17 101 96 ns ns ns 0.31‘ 0.31’ 0.28'
9/6 121 95 ns ns ns ns ns ns

 

:m = treatment, app = applied N, avail = available N
” DAP = Days after planting
§Calculated approximation using NDVI rather than approximation developed from images

Significance of overall F-values at p $0.05, 0.01, 0.001.

the situation was similar with GNDVI. The mean separation of index values indicated

85

 

the indices at this location (not shown) could not distinguish between treatments while
individual reflectance at 610, 660, and 710 could distinguish between treatments 1 and 4
(Table 4).

Similar to the individual wavebands, correlation of indices to soil-N was better at
Sandyland (Table 14) than at the Experiment Station. GNDVI was the first index to
significantly correlate with Available N on June 22 similar to reflectance at 560 and 810
nm. On June 28 and July 5 NDVI, SAVISL, SAVI, and TSAVI were significantly
correlated to Available N while the associated reflectance measurement at 660 nm was
only significant on June 28 and the NR reflectance at 810 nm was significant on both
dates; demonstrating the sensitivity of the indices to NR (Epiphanio nad Huete, 1994) at
low coverage. In contrast, on July 12, when NR reflectance measurements were not
significantly correlated to N, neither were the indices. Similar to individual reflectance
measurements, all five indices were significantly correlated to Treatment, Applied N, and
Available N from July 19 to season end. By August 9 the indices based on red
reflectance, NDVI, SAVI, SAVISL and TSAVI, exhibited predictability equal to GNDVI
in estimating N status with the coefficient of determination at 0.93 to 0.96 across all
indices. NDVI and SAVI were expected to saturate at full canopy coverage, but a
comparison between Table 5 and Table 14 indicates that not only did they not saturate,
but the predictability of N status represented as Treatment, Applied N, or Available N,
was better explained by the indices enhancing the information provided by reflectance
measurements (Yoshioka et al., 2000). The indices that corrected for soil background
“noise”, SAVI, SAVISL, and TSAVI by rendering red less sensitive and NR more

sensitive to canopy coverage did not perform better than NDVI or GNDVI. The mean

86

 

separation of index values (Table 15) indicated that all five indices differentiated between

at least two treatments. GNDVI detected the difference between three treatments as early

as August 2, approximately 104 days after planting.

Table 14. Linear regression coefficients of selected indices vs N; where Index = mN + b through 7/26

and N is treatment, applied N, or available N at Sandyland, 2001, Asgrow B1 and Prime Cut 59
varieties. Beginning 8/2 Index = mN + mN2 + b.

 

Ave

 

 

 

 

 

 

t .+ . .
Scan Veg trt app avarl trt app avarl trt app avarl
Date DAPi Cov NDVI SAVISL SAVI

% 1'2
6/13 54 21 ns -------- ns ns -------- ns ns -------- ns
6/22 63 43 ns -------- ns ns -------- ns ns -------- ns
6/28 69 63 ns ........ 0.38" ns -------- 0.38” ns 0.38“
7/5 76 63 ns ........ 0.26‘ ns ........ 0.26‘ ns 0.26'
7/12 83 89 ns ns ns ns ns ns ns ns us
7719 90 93 0.30‘ 0.33‘ 0.36” 0.32' 0.35‘ 0.38‘ 0.31‘ 0.34‘ 0.37‘
7/26 97 98° 0.62‘” 0.68‘“ 0.70'" 0.61‘” 0.68’” 0.70‘” 0.61‘“ 0.68‘“ 0.70‘”
8/2 104 99° 0.82‘“ 0.81’” 0.82’“ 0.82‘“ 0.81‘” 0.82‘” 0.82‘” 0.81‘” 0.82‘“
8/9 111 99 0.96‘” 0.95‘“ 0.94‘” 0.95‘” 0.94‘" 0.93‘” 0.96‘” 0.95‘” 0.93‘"
8/17 119 99 0.96‘" 0.96‘“ 0.95‘” 0.95‘” 0.95’” 0.95‘” 0.95'” 0.95‘” 0.95‘”
TSAVI GNDVI
% 1'2

6/13 54 21 ns ........ ns ns ........ ns
6/22 63 43 ns ........ ns ns ........ 0.39“
6/28 69 63 ns -------- 0.38” ns ........ 0.31‘
7/5 76 63 ns ........ 0.26‘ ns -------- 0.31‘
7/12 83 89 ns ns ns ns ns ns
7/19 90 93 0.31‘ 0.34‘ 0.37‘ 0.52“ 0.56‘” 0.60‘”
7/26 97 98° 0.61‘“ 0.68‘” 0.70‘” 0.67‘“ 0.74‘” 0.76‘”
8/2 104 99§ 0.82‘” 0.81‘” 0.82‘” 0.91‘” 0.89‘“ 0.91‘“
8/9 111 99 0.96‘” 0.95‘“ 0.94‘“ 0.96‘” 0.95‘” 0.95‘“
8/17 119 99 0.95‘” 0.95‘“ 0.95’” 0.96‘” 0.96'” 0.96‘”

 

1Ltrt = treatment, app = applied N, avail = available N
: DAP = Days after planting
§Calculated approximation using NDVI rather than approximation developed from images

Significance of overall F-values at p $0.05, 0.01, 0.001.

O O. .0.

In 2002, at the Montcahn Experiment Station (Table 16), the indices were not

well correlated with soil-N. All five indices studied include the waveband centered at

87

810 nm that was never significant at this location anytime during the season. The
equations for NDVI, SAVI, SAVISL and TSAVI that include reflectance at 660 nm; even
though significantly correlated to soil-N, explained at most 42% of the N differences and

were apparently not significant enough to overcome the insignificance of the NR.

Table 15. Mean reflectance measurements as influenced by treatment or applied N at Sandyland, 2001,
Asgrow B1 and Prime Cut 59 varieties.

 

 

 

 

Treatment Index J try 1 9 July 26 August 2 August 9 Algust 17
Kg ha'I (nm) Index Value
45 NDVI 0.88b 0.88b 0.90b 0.89c 0.89c
90 0.89ab 0.903 0.913 0.90b 0.90b
135 0.913 0.913 0.92a 0.913 0.913
180 0.90a 0.913 0.913 0.913 0.91a
p-value 0004’r <0.0001“ <0.0001 <0.0001 <0.0001
45 SAVISL 1.79b 1.77b 1.81b 1.79c 1.79c
90 1.8lab 1.82a 1.84a 1.82b 1.82b
135 1.843 1.843 1.853 1.833 1.833
180 1.823 1.843 1.853 1.83a 1.84a
p-value 0.005f 0.0002 0.0001 <0.0001 <0.0001
45 SAVI 1.32b 1.31b 1.34b 1.32c 1.326
90 1.343b 1.343 1.353 1.34b 1.34b
135 1.35a 1.353 1.36a 1.35a 1.353
180 1.343 1.35a 1.363 1.353 1.36a
p-value 0004* <0.0001“ <0.0001 <0.0001 <0.0001
45 TSAVI 0.88b 0.87b 0.89b 0.88c 0.88c
90 0.89ab 0.893 0.903 0.90b 0.8%
135 0.903 0.90a 0.913 0.903 0.903
180 0.90a 0.903 0.913 0.903 0.90a
p-value 0.004'r 0.0002 <0.0001 <0.0001 <0.0001
45 GNDVI 0.77b 0.76b 0.77c 0.76e 0.76e
90 0.79ab 0.79a 0.80b 0.79b 0.78b
135 0.813 0.813 0.813 0.803 0.80a
180 0.803 0.803 0.813 0.803 0.813
p-value 0.0025 <0.0001 <0.0001 <0.0001 <0.0001

 

T Blocking was significant July 19: NDVI p-value = 0.006, SAVISL p-value = 0.01, SAVI p-value = 0.008,
TSAVI p-value = 0.008; July 26: NDVI p-value = 0.04, SAVI p-value = 0.05.

Mean values with the same letters are not significantly different at p = 0.05 based on HSD.

p-value is that generated from overall F -ratio.

Results for Treatment and Applied N are the same.

88

However, the waveband centered at 560 nm (Table 7), as of August 15, could explain as
much as 75% of N differences which may have been enough to dominate the

insigrrificance of 810 nm. GNDVI was the only index to make a significant contribution,

Table 16. Linear regression coefficients of selected indices vs N; where N
is treatment, applied N, or available N at the Montcalm Experiment
Station 2002, Diamond Cut variety.

 

 

 

 

Ave 1- f .1 .
Scan + Veg trt app avarl trt app avarl
Date DAP+ Cov All other indices GNDVI

% :2
7/11 65 51§ ns ns ns ns ns ns
7/17 71 68 ns ns ns ns ns ns
7/24 78 86 ns ns ns ns ns ns
8/1 86 96 ns ns ns ns ns ns
8/9 94 94 ns ns ns ns ns us
8715 100 98 ns ns ns 0.30' 0.30‘ 0.29‘
8/21 106 97 ns ns ns 0.51" 0.51" 0.52“
8/30 115 88 ns ns ns 0.46” 0.46” 0.49“
9/6 122 93 ns ns ns 0.39" 0.39“ 0.44“

 

Tut = treatment, app = applied N, avail = available N
* DAP = Days after planting
§Calculated approximation using NDVI rather than approximation

developed from images
’ " ... Significance of overall F-values at p $0.05. 0.01. 0-001-

Table 17. Mean reflectance measurements as influenced by treatment or
applied N at the Montcalm Experiment Station, 2002, Diamond Cut variety.

 

 

 

Treatment Index August 21 August 30 Sgrtember 6
Kg ha'I (nm) Index Value ------------- -
45 GNDVI 0.81b 0.78b 0.77b
90 0.823b 0.79ab 0.783b
135 0.823 0.813 0.793
180 0.823b 0.803 0.79ab
p-value 0.03 0.01 0.03

 

Mean values with the same letters are not significantly different at p = 0.05
based on HSD. p-value is that generated from overall F-ratio.
Results for Treatment and Applied are the same.

89

 

qn'

and explained nearly 50% of soil N differences. Mean separation of index values,

prepared for GNDVI only (Table 17), showed differences between treatments 1 and 3.

None of the indices calculated from reflectance measurements of the Goliath canopy

were significant. Canopy health at this location greatly affected the use of indices.

At the 2002 Sandyland location (Table 18) SAVISL and GNDVI showed

significant correlation to soil-N even before the visible wavebands. GNDVI at 92% of

canopy coverage explained as much as 53% of the differences in N on July 17. While the

correlation of individual wavebands to soil-N seemed to peak on July 24 (Table 11), the
indices peaked the following week, but did not attain the same level of predictability as

the individual reflectance (Table 11) where reflectance at 560 nm explained 72%, and

 

Table 18. Linear regression coefficients of selected indices vs N; where N is treatment, applied N, or
available N at Sandyland, 2002, Sugar Snax variety.

 

 

 

 

 

 

 

AVE 1* + .t . .
Scan Veg trt app avarl trt app avarl trt app avarl
Date DAP‘ Cov NDVI SAVISL SAVI

% :2
7/11 82 91° ns ns ns ns ns ns ns ns ns
7/17 88 92° ns ns ns 0.27‘ 0.27‘ 0.28' ns ns 0.27‘
7/24 95 92 0.30‘ 0.30‘ 0.30‘ 0.31‘ 0.32‘ 0.31‘ 0.31‘ 0.31’ 0.31'
8/1 103 97 0.62'” 0.62‘” 0.63‘” 0.63‘” 0.63‘” 0.65’” 0.63‘” 0.63’“ 0.64‘”
8/9 111 96 ns ns ns ns ns ns ns ns ns
8/15 1 17 99 0.51" 0.51“ 0.51” 0.50“ 0.50“ 0.50” 0.50” 0.50" 0.50"

TSAVI GNDVI

% ,2
7/11 82 91° ns ns ns ns ns ns
7/17 88 92° ns ns ns 0.52” 0.52” 0.53“
7/24 95 92 0.31' 0.31’ 0.31‘ 0.41“ 0.41“ 0.40“
8/1 103 97 0.62‘” 0.62'” 0.64‘” 0.70‘“ 0.70‘” 0.72‘"
8/9 111 96 ns ns ns ns ns ns
8/15 117 99 0.50” 0.50" 0.50“ 0.56‘” 0.56’” 0.56‘"

 

7‘trt = treatment, app = applied N, avail = available N
° DAP = Days after planting
§Calculated approximation using NDVI rather than approximation developed from images

Significance of overall F-values at p $0.05, 0.01, 0.001.

O .0 CO.

90

660 nm explained 80% of soil-N differences. Like the individual wavebands, all five
indices showed a decrease in the coefficient of determination and subsequent increase as
the plants reflected N uptake from the July 29 application. Again, GNDVI responded the
best and was the most sensitive to change in Treatment, Applied N, and Available N.
While individual reflectance measurements showed significant separation of treatments
(Table 12), the order did not necessarily follow treatment. The indices, beginning on July
17, seemed to minimize some of the confounding effects (Epiphanio and Huete, 1994)
exhibiting better separation of treatment differences and in treatment order (Table 19).
Just as with 2001, GNDVI detected the difference between three treatments as early as
August 1, approximately 103 days following planting. Michigan carrots are generally
harvested 80 to 180 days (USDA, 1999; Zandstra et al., 1986) after planting. On August
1 there would be plenty of time to correct nutrient deficiencies.

The 2001 early season measurements at the Experiment Station and Sandyland
indicated all five indices performed equally according to the regression models. Though
indices were generally significantly correlated with Available N, predictability of the
models was too low to be relied on for N management, when canopy coverage averaged
19% at the Experiment Station and 63% at Sandyland by July 5. Mid and late season
results at Sandyland, 2001 and 2002, indicated where canopies were healthy all five
indices performed well. NDVI lagged slightly behind SAVI, SAVISL, and TSAVI as
they all increased in significance over the season. GNDVI out performed the other
indices in its assessment of soil-N both in the regression models and the mean separation
of index values. Blackrner et 31. (1994), Schepers et al (1992), and Schepers et 31. (1996)

also found that the “green band” together with NR is better at showing the variability in

91

leaf chlorophyll. At Sandyland, the only locations that attained full canopy, none of the
indices plateaued as evidenced by the very high coefficients of determination. This
phenomenon may be attributed to the lacy nature of the carrot canopy that may result in

micro-views of the soil and shadows.

Table 19. Mean reflectance measurements as influenced by treatment or applied N at Sandyland, 2002,
Sugar Snax variety.

 

 

 

 

Treatment Index J ulyl 7 July 24 August 1 August 9 August 15 E
Kg ha'I (nm) Index Value 'I

45 NDVI ns 0.84b 0.87b 0.88b 0.8%

90 ns 0.88ab 0.89ab 0.89ab 0.903b

135 ns 0.903 0.903 0.903b 0.903b
180 ns 0.913 0.913 0.913 0.913
p-value 0.005 0.005 0.01 0.002
45 SAVISL ns 1.83b 1.90b 1.92b 1.94b

90 ns 1.923b 1.953b 1.953b 1.96ab
135 ns 1.973 1.973 1.96ab 1.973
180 as 1.993 1.993 1.983 1.993
p-value 0.005 0.005 0.01 0.003
45 SAVI ns 1.25b 1.29b 1.31b 1.32b

90 ns 1.303b 1.323b 1.33ab 1.33ab
135 ns 1.343 1.343 1.33ab 1.343
180 as 1.353 1.353 1.353 1.353
p-value 0.005 0.005 0.01 0.003
45 TSAVI ns 0.83b 0.86b 0.87b 0.88b

90 ns 0.87ab 0.883b 0.89ab 0.89ab
135 ns 0.89a 0.89a 0.89ab 0.893
180 ns 0.903 0.903 0.903 0.903
p-value 0.006 0.005 0.01 0.003
45 GNDVI 0.78b 0.77b 0.79c 0.79b 0.80c

90 0.8lab 0.8lab 0.81bc 0.8lab 0.81bc

135 0.823 0.833 0.823b 0.8lab 0.823b
180 0.823 0.843 0.833 0.833 0.833
p-value 0.024 0.002 0.001 0.003 0.001

 

 

Mean values with the same letters are not significantly different at p= 0.05 based on HSD.
p-value is that generated from overall F-ratio.
Results for Treatment and Applied N were the same.

92

In Season N Management: Reflectance vs Total N and Sap Nitrate

Carrot petioles were sampled at varied intervals during the two seasons and
analyzed for total N and petiole sap NO'3 content (Petiole-N) (Chapter 1, Table 5). These
results representing the conventional measurement of plant response to N treatments were
regressed against individual reflectance measurements and indices to determine if
reflectance measurements depicted similar information about N treatment results as the
conventional measurements. Spectral measurements taken on the three consecutive dates
surrounding each petiole sampling date were compared to petiole analysis results. The
purpose of comparing the three measurement dates was to find out when reflectance
measurements were best correlated with the conventional samples: whether prior to
physical sampling, at the same time, or following physical sampling.

Overall correlation of petiole samples to individual reflectance measurements and
indices was insignificant until canopy coverage was greater than 90%. Only the Goliath
variety showed earlier significance at 80 % canopy coverage (Table 20). It is
distinguished by a darker green and more robust canopy than any of the other varieties
included in the study. These results were comparable to the findings of Flowers et al.
(2003b) who attributed such late correlation to the poor relationship between whole plant-
N, without consideration for biomass, and spectral measurements. The extent of canopy
coverage did not relate spectral measurements to N concentration. In this study, the N
concentration, although not derived from whole plant tissue, appears to convey the same
lack of correlation at lower canopy coverage. Statistical analysis indicated that blocking
significantly impacted measured parameters. Canopy coverage of the Diamond Cut

variety, during August and September was greater than 90 % and correlation between

93

 

Table 20. Linear regression coefficients of reflectance at individual wavelengths vs petiole-N;
where petiole-N = total N content or petiole sap N03- at the Montcalm Experiment Station, 2002,

 

 

 

 

 

 

 

 

 

Goliath variety.
Petiole Previous Date)r Sample Datef Following Date+
Sampling ‘
Date Wavelength TKN Sag TKN Sap TKN Sap
(80% cov.) (nm) rT
7/25 510 0.24‘ ns 0.24‘ 0.36‘ ns 0.42"
560 0.25‘ ns 0.26‘ 0.40" ns 0.49“
610 0.24‘ ns 0.24‘ 0.38" ns 0.57‘”
660 ns ns 0.24‘ 0.36‘ ns 0.46“
710 0.27‘ ns 0.25‘ 0.39“ ns 0.54“
760 0.24‘ ns ns 0.37‘ ns ns
810 0.28’ ns ns 0.41“ ns ns
NDVI 0.25‘ ns ns 0.37‘ ns 0.47"
SAVISL 0.25‘ ns ns 0.37‘ ns 0.46“
SAVI 0.25’ ns ns 0.37‘ ns 0.47"
TSAVI 0.26‘ ns ns 0.37‘ ns 0.46”
GNDVI 0.27‘ ns ns 0.39” ns 0.46“ t
(94% COV.) (nm) :2
8/22 510 ns ns ns 0.27‘ ns 0.43”
560 ns 0.48” 0.27‘ 0.58‘“ 0.43" 0.70‘“
610 ns 0.25’ ns 0.39” ns 0.53'”
660 ns ns ns ns ns us
710 ns 0.45“ ns 0.44“ 0.33‘ 0.65‘“
760 ns ns ns ns ns ns
810 ns ns ns ns ns ns
NDVI ns ns ns ns ns ns
SAVISL ns ns ns ns ns ns
SAVI ns ns ns ns ns ns
TSAVI ns ns ns ns ns ns
GNDVI ns ns ns ns ns ns
(90% cov.) (am) 1'2
9/9 510 0.48” ns ns us
560 0.72”. 0.46” ns ns
610 0.43“ ns ns ns
660 ns ns ns us
710 0.60m 0.41" ns us
760 0.38” ns 0.24‘ ns
810 0.36" ns 0.24‘ ns
NDVI ns ns ns ns
SAVISL ns ns ns ns
SAVI ns ns ns ns
TSAVI ns ns ns ns
GNDVI ns ns ns ns

 

T Previous = Reflectance measurements one week before petiole sampling. Sampling date =
Reflectance measurements on the same day as petiole sampling. Following = Reflectance
measurements one week following petiole sampling.

' ” Significance of overall F-values at p $0.05, 0.01, 0.001.

94

Table 21. Linear regression coefficients of reflectance at individual wavelengths vs petiole-N;
where petiole-N = total N content or petiole sap NO3' at the Montcalm Experiment Station, 2001,
Diamond Cut variety.

 

 

 

 

 

 

PCthlC Previous Date)r Sarrrple Date,r Following Date)r
Sampling —
Date Wavelength TKN Sap TKN Sap TKN Sap
(96% cov.) (11m) r
8/15 510 ns ns ns ns ns us
560 ns 0.31‘ ns ns ns 0.29‘
610 ns ns 0.37“ 0.33' 0.28‘ 0.38“
660 ns ns ns ns 0.28‘ 0.34“
710 0.32‘ 0.36“ ns 0.38" ns 0.31‘
760 ns ns ns ns ns us
810 ns ns ns ns ns ns !
NDVI ns ns 0.30. ns ns ns :
SAVISL ns ns 0.30. ns ns ns I
SAVI ns ns 0.30. ns ns ns i
TSAVI ns ns 0.30. ns ns ns
GNDVI ns ns 0.51” 0.47" ns ns
(95% COV.) (nm) 1'2
9/11 510 ns 0.32’ 0.27‘ ns
560 ns ns 0.42“ 0 33'
610 0.41“ 0.46” 0.51” 0.44“
660 ns 0.26‘ 0.48“ 0.43”
710 ns ns 0.45” 0.34‘
760 ns ns ns 0.26.
810 ns ns ns ns
NDv1 ns 0.27‘ 0.39‘ 0.37‘
SAVISL ns 0.27‘ 0.37“ 0.36‘
SAVI ns 0.27‘ 0.38“ 0.37‘
TSAVI ns 0.27‘ 0.38” 0.37‘

GNDVI 0.41" 0.55“ 0.43“ 0.36‘

T Previous = Reflectance measurements one week before petiole sampling. Sampling date =
Reflectance measurements on the same day as petiole sampling. Following = Reflectance
measurements one week following petiole sampling.

' .. ... Significance of overall F-values at p $0.05, 0.01, 0.001.

 

reflectance and petiole samples increased over time until end of season (Table 21 and
Table 22). Correlation was best at the September sampling date. At Sandyland
correlation of reflectance measurements with Petiole-N was significant on only one date

each year, July 19, 2001 and July 25, 2002 (Table 23 and Table 24). At that time

95

coverage was 93% and 92%, respectively. Coverage was less than 90%, at earlier

measurements and the last measurements were made when coverage was approximately

99%. Full canopy coverage was attained at Sandyland and reflectance appears to have

saturated since none of the parameters at full coverage were significant at p 50.05.

Table 22. Linear regression coefficients of reflectance at individual wavelengths vs petiole-N;
where petiole-N = total N content or petiole sap N03- at the Montcalm Experiment Station, 2002,

Diamond Cut variety.

 

 

 

 

 

 

 

Petiole Previous Date)r Sample Date+ Following Date+ F
Sampllng
Date Wavelength TKN Sap TKN Sap TKN Sap
(97% cov.) (nm) 2'2
8/22 510 0.29‘ ns 0.36‘ ns 0.56‘” 0.36“
560 0.71'” 0.45" 0.71'” 0.52“ 0.77'“ 0.64‘”
610 0.73‘” 0.44” 0.66’” 0.43“ 0.69'” 0.67‘” '
660 ns ns ns ns 0.38“ 0.28‘
710 0.48“ 0.64‘” 0.51" 0.60‘” 0.65'” 0.76‘”
760 0.33‘ 0.24‘ ns ns ns us
810 0.23' 0.24‘ ns ns ns ns
NDVI ns nS ns ns ns ns
SAVISL ns ns ns ns ns ns
SAVI ns ns ns ns ns ns
TSAVI ns ns ns ns ns ns
GNDVI ns ns 0.49” 0.32‘ 0.36‘ 0.36”
(93% COV.) (nm) 2'2
9/9 510 ns 0.47" ns 0.56‘”
560 ns 0.69‘” ns 0.74‘“
610 ns 0.72‘” 0.25‘ 0.73‘”
660 ns 0.43“ ns 0.44”
710 0.33‘ 0.72’” 0.37" 0.73‘"
760 ns ns ns us
810 ns ns ns ns
NDVI ns 0.25‘ ns ns
SAVISL ns ns ns ns
SAVI ns ns ns ns
TSAVI ns ns ns ns
GNDVI ns 0.47“ ns 0.41“

 

I Previous = Reflectance measurements one week before petiole sampling. Sampling date =

Reflectance measurements on the same day as petiole sampling. Following = Reflectance
.measurements one week following petiole sampling.
‘ ‘ ' Significance ofoverall F-values atp $0.05, 0.01, 0.001.

Petiole-N of the more robust Goliath variety correlated with reflectance on July 25, 2002,

96

at 80% coverage. However, following July 25, further comparison to Petiole-N

diminished in significance as the canopy senesced prematurely due to disease that

affected biomass correlation to the NR wavebands and all of the indices. Reflectance

provided a small window of time where correlation with petiole-N was significant at p S

0.05; where canopy coverage was greater than 90% and less than 99% coverage.

Many of the studies referenced herein used leaf or whole plant samples as the

physical comparison to reflectance. In this study, petioles were sampled because petiole

nitrate content has been found to be 3 good indicator of the N status in carrot (Warncke,

1996). Petioles are often used as a reliable indicator of the N status in many crops

(Hemphill and Jackson, 1982; Warncke, 1996; Walworth, 1998). However, petiole-N

-' '1-1

1—

was better correlated with the reflectance measurements taken the week following

Table 23. Linear regression coefficients of reflectance at individual wavelengths vs petiole-N;
where petiole-N = total N content or petiole sap NO3' at Sandyland, 2001, Asgrow B1 and Prime

Cut 59 varieties.

 

 

 

 

Petiole Previous Date+ Sample Date+ Following Date)r
Sampling ‘
Date Wavelength TKN Sap TKN Sap TKN Sap
(93% cov.) (nm) 1'2
7/19 510 ns ns ns ns 0.46“ 0.52“
560 ns ns 0.29‘ 0.31‘ 0.39“ 0.52“
610 ns ns 0.30‘ 0.33‘ 0.58‘” 0.65‘”
660 ns ns ns ns 0.44” 0.46”
710 ns ns 0.27' ns 0.50“ 0.52‘”
760 ns ns ns 0.26‘ 0.35‘ 0.32‘
810 ns ns ns 0.28‘ 0.36" 0.33‘
NDVI ns ns nS ns 0.45” 0.45"
SAVISL ns ns ns ns 0.45” 0.44“
SAVI ns ns ns ns 0.45” 0.45“
TSAVI ns ns ns ns 0.45” 0.44”
GNDVI ns ns 0.33‘ 0.41” 0.51“ 0.53‘”

 

T Previous = Reflectance measurements one week before petiole sampling. Sampling date =
Reflectance measurements on the same day as petiole sampling. Following = Reflectance
measurements one week following petiole sampling.

Significance of overall F-values at p 50.05, 0.01, 0.001.

O O. .0.

97

physical sampling, at all five locations. It appears that petiole samples may reflect the
firture condition of the canopy and that leaf samples may have been a more timely

comparison to reflectance observations.

Table 24. Linear regression coefficients of reflectance at individual wavelengths vs petiole-N;
where petiole-N = total N content or petiole sap N03- at Sandyland, 2002, Sugar Snax variety.

 

 

 

 

 

Petiole Previous DateI Sample Date+ Following Date+
Sarrrpling ' 7
Date Wavelength TKN Sap TKN Sap TKN Sap -.
(92% cov.) (nm) ,2
7/25 510 ns ns 0.47” ns 0.83‘” 0.55"
560 0.39" ns 0.53‘” ns 0.71‘“ 0.62‘“
610 0.43” ns 0.50" ns 0.80‘“ 0.56‘”
660 ns ns 0.38“ ns 0.62‘” 0.28‘
710 0.51” ns 0.51“ ns 0.84‘“ 0.62‘” i
760 ns ns ns ns ns ns
810 ns ns ns ns ns ns
NDVI ns ns 0.33‘ ns 0.46“ ns
SAVISL ns ns 0.32‘ ns 0.43” ns
SAVI ns ns 0.32' ns 0.45” ns
TSAVI ns ns 0.32‘ ns 0.45“ ns
GNDVI ns ns 0.35‘ ns 0.55“ ns

 

f Previous = Reflectance measurements one week before petiole sampling. Sampling date =
Reflectance measurements on the same day as petiole sampling. Following = Reflectance
measurements one week following petiole sampling.

’ " ... Significance of overall F -values at p $0.05, 0.01, 0.001.

If a reflectance measurement is to provide in-season nutrient management, it
should explain more than 50% of the variation between parameters. Reflectance
measurements at 560, 610 and 710 nm exhibited the strongest correlation to Petiole-N in
mid-to-late season. Of the three wavebands, reflectance at 560 and 610 nm were more
often the best correlated with petiole-N, explaining as much as 77% of the differences in
Diamond Cut (Table 22), and 80% of the differences in Sugar Snax (Table 24).
Reflectance at 710 nm explained as much as 76 % of the differences in Diamond Cut

(Table 22) and 84% Sugar Snax (Table 24).

98

NR reflectance at 760 and 810 nm, associated with biomass, was weakly
correlated with Petiole-N which also had an impact on the significance of the indices.
GNDVI was the only index that, on occasion, could explain differences in Petiole-N
where the coefficient of determination was greater than 50%. It was the significance of
560 nm in the formula that contributed to the strength of GNDVI. The remaining indices
which rely on red reflectance at 660 nm, exhibited weak significance, but were unreliable
for nutrient management. None of the indices performed as well as reflectance at
individual wavebands, specifically, 560, 610, and 710 nm attributable to the weakness of
reflectance at 810 nm.

It appears that in August at the Experiment Station and in July at Sandyland,
when coverage was less than 99%, but greater than 90%, that the reflectance
measurements explained similar information about canopy health as did Petiole-N. The
respective timing represented mid-season when, if necessary, additional N amendments

would have time to be taken up and utilized in the plant before harvest.

End of Season: Reflectance vs Selected Harvest Parameters
Reflectance measurements taken throughout the two seasons were compared to
eight physical parameters measured at harvest. Those parameters were:

% N in Tops N Uptake in Tops Dry top Biomass Yield
% N in Roots N Uptake in Roots Dry root Biomass Root:Shoot

Not all parameters were significantly correlated to reflectance measurements or the
indices calculated from reflectance; therefore, only the parameters showing more than
sporadic significance were included in the following tables. % N in tops, N Uptake in

Tops, Dry Top Biomass and Root:Shoot are the parameters concerned with healthy tops.

99

1‘."

Individual Reflectance: Reflectance was not well correlated with % N in Tops

(Table 25). Only in 2002, was reflectance of the Diamond Cut variety at the Experiment

Station and Sugar Snax at Sandyland significantly correlated to % N in Tops, albeit low.

Significance was focused at 560, 610 and 710 nm but it was not high enough to be

predictable. Correlation to % N in the Tops which included the leaves as well as petioles

was strongest in the visible bands. NR reflectance at 760 and 810 nm, associated with

biomass, was not significantly correlated with % N in Tops similar to the results for in-

season Petiole-N.

Table 25. Linear regression coefficients of reflectance at individual wavelengths vs % N in harvested
tops from selected locations.

 

 

 

 

 

 

Ave 510nm 560nm 610nm 660nm 710nm 760nm 810nm
Scan 1 Veg
Date DAP Cov 2002 Montcalm Experiment Station, Diamond Cut variety

% r2
7/11 65 51§ ns ns ns ns ns ns ns
7/ 17 71 68 ns ns ns ns ns ns ns
7/24 78 86 ns ns ns ns ns ns ns
8/1 86 96 ns ns ns nS ns ns ns
8/9 94 94 ns 0.27. ns ns ns ns ns
8/15 100 98 ns 0.37‘ 0.27‘ ns 0.54” 0.48“ 0.50”
8/21 106 97 ns 0.47“ 0.26‘ ns 0.39“ ns ns
8/30 115 88 ns 0.41" 0.35‘ ns 0.40“ ns ns
9/6 122 93 ns 0.42" 0.33‘ ns 0.34‘ ns ns

2002 Sandyland, Sugar Snax variety

% ,2
7/11 82 91§ ns ns ns ns ns ns ns
7/17 88 92° 0.33‘ 0.37‘ 0.44“ ns 0.35‘ ns ns
7/24 95 92 ns ns ns ns ns ns ns
8/1 103 97 0.46" 0.54" 0.37' ns 0.35‘ ns ns
8/9 111 96 0.31‘ 0.26‘ 0.26‘ ns ns ns ns
8/15 117 99 ns ns ns 0.31. ns ns ns

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.

Significance of overall F -values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

100

 

I- “I'- .s. 4.1.? -1_

The % N in Roots (Table 26) was significantly correlated to reflectance at 560
and 610 nm in mid July, or the first of August. Preharvest predictability was evident at r2
> 0.50 almost a month before harvest at the two Sandyland sites. A week before harvest
correlation was strongly significant in all visible wavebands and reached a maximum
correlation at wavebands 560 and 610 nm of r2 = 0.86 followed by 710, 510, and 660
nm. It is interesting to note that 660 nm was not significantly correlated to % N in Tops
where its response to chlorophyll was expected.

Reflectance correlated significantly with N Uptake in Tops (Table 27) at
Sandyland as early as the first of August 2001 and mid July 2002. Predictability of the
reflectance at 610, 560, and 510 nm was greater than 50 %. Here the combination of %
N in Tops and Dry Top Biomass was more representative of the canopy than each of the
two parameters separately (Tables 25 and 28). Dry Top Biomass was the dominant
parameter for N Uptake in Tops. Reflectance from the Diamond Cut canopy; however,
was better correlated to each of the separate parameters, % N and Dry Top Biomass. The
resulting combination was Sporadic in significance in N Uptake. The Goliath variety
correlated best with mid-season Dry Top Biomass rather than % N. Where the canopy is
healthy N Uptake in Tops may be a better in-season comparison than Petiole-N alone; the
% N segment includes leaf -N and Petiole-N, and the biomass segment incorporates the
coverage dimension into the measurement. Reflectance at 560, 610 and 710 nm provided
the best overall correlation. Reflectance in the NR wavebands was not well correlated.

While N Uptake in Roots, a combination of % N in Roots and Dry Root Biomass,
is not a healthy tops issue, the strong significance with reflectance exhibited at Sandyland

(Table 29) may be useful for monitoring the N content of storage roots used in food

101

Table 26. Linear regression coefficients of reflectance at individual wavelengths vs % N in harvested
roots from selected locations.

 

Ave 510nm 560nm 610nm 660nm 710nm 760nm 810nm

 

Scan Veg

 

 

 

Date DAP‘r Cov 2001 Montcalm Experiment Station, Diamond Cut variety

% r2
7/ 12 65 35 ns ns ns ns ns ns ns
7/20 73 64 ns ns ns ns ns ns ns
7/26 79 85° ns ns ns ns 0.33. ns ns
8/2 86 94 ns ns ns ns ns ns ns
8/9 93 97 ns ns 0.30. ns ns ns ns
8/17 101 96 0.39’ ns 0.47” 0.30‘ ns ns ns
9/6 121 95 ns 0.32‘ 0.36‘ 0.39‘ 0.31’ 0.41" ns

2001 Sandyland, Asgrow B1 and Prime Cut 59 varieties

% :2
7/12 83 89 0.29‘ 0.28' 0.26‘ ns ns ns ns
7/19 90 93 ns 0.28‘ 0.33‘ ns ns 0.32‘ 0.33‘

7/26 97 98° 0.41” 0.37‘ 0.45“ 0.35' 0.42” 0.38“ 0.39
8/2 104 99§ 0.43” 0.51" 0.58‘” 0.41“ 0.44" 0.36‘ 0.44"

 

 

 

 

 

8/9 11 l 99 0.65’” 0.77‘“ 0.76‘” 0.65‘“ 0.75‘” ns ns
8/17 1 19 99 0.80‘“ 0.86‘“ 0.86‘” 0.78’” 0.84‘“ ns us
2002 Montcalm Experiment Station, Diamond Cut variety
% r2
7/ 1 1 65 51° ns ns ns ns ns ns ns
7/17 71 68 0.32‘ 0.33‘ 0.30‘ 0.30‘ ns ns ns
7/24 78 86 ns ns ns ns ns ns ns
8/1 86 96 ns ns ns ns ns ns ns
8/9 94 94 ns 0.27‘ ns ns ns ns ns
8/15 100 98 ns 0.30‘ 0.34‘ ns 0.46” ns ns
8/21 106 97 ns 0.30‘ 0.42” ns 0.25‘ ns ns
8/30 115 88 0.26‘ 0.27‘ 0.39‘ 0.33‘ 0.32‘ ns ns
9/6 122 93 0.31' 0.27‘ 0.35‘ 0.44“ 0.32‘ ns ns
2002 Montcalm Experiment Station, Goliath variety
% :2
7/11 65 46° ns ns ns ns ns ns ns
7/17 71 66 ns nS ns ns ns ns ns
7/24 78 80 ns ns ns ns ns ns ns
8/1 86 96 0.30‘ 0.37‘ ns 0.30‘ ns ns ns
8/9 94 93 0.46“ 0.51“ 0.52“ 0.46“ 0.34‘ ns ns
8/15 100 96 ns ns ns ns ns ns ns
8/21 106 94 ns ns ns ns 0.28. ns ns
8/30 115 87 ns 0.30‘ 0.25‘ ns 0 27‘ ns ns
9/6 1 22 90 ns ns ns ns ns ns ns

 

102

Table 26 (cont’d)

 

 

 

Scan 3:; 510nm 560nm 610nm 660nm 710nm 760nm 810nm
Date DAPT Cov 2002 Sandyland, Sugar Snax variety
% :2

7/11 82 91° ns ns ns ns ns ns ns

7/17 88 92§ ns 0.26‘ 0.27‘ ns 0.27‘ ns ns

7/24 95 92 ns ns ns ns ns ns ns

8/1 103 97 0.53” 0.66‘” 0.49" 0.28‘ 0.53" ns ns

8/9 1 1 1 96 0.46" 0.47” 0.46” 0.38‘ ns ns ns

8/15 117 99 0.35‘ 0.34‘ 0.39" 0.44“ ns ns ns

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.
' .. ... Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

products. It may be an important topic for future studies. Dry Root Biomass (Table 30)

 

showed very little correlation at any location, whereas % N in Roots (Table 26) and the
resulting N Uptake in Roots (Table 28) was strongly correlated with reflectance in ahnost
all wavebands. As early as August in 2001 and July in 2002, reflectance could explain
greater than 50 % of the differences for N Uptake in Roots among N treatments. By
harvest 12 > 0.70 at 610 nm and r2 > 0.60 at 510, 560, 660, and 710 nm.

Dry Top Biomass (Table 28) was best correlated where there was a healthy
canopy as with the other parameters measured. While other parameters were better
correlated with the visual wavebands, Dry Top Biomass as expected was also correlated
to the NR wavebands at 760 and 810 nm but saturated at 100 % coverage in 2001
Sandyland. In 2002 biomass of the Goliath variety at harvest exhibited significance to
mid-season reflectance as though the recovering canopy was responding to mid season
grth patterns. In early August, the Goliath variety had lost foliage due to disease, but
later recovered.

While significant correlation to Dry Top Biomass at the Experiment Station in

103

Table 27. Linear regression coefficients of reflectance at individual wavelengths vs N uptake in
harvested tops from selected locations.

 

 

 

 

 

 

 

 

 

Ave 510nm 560nm 610nm 660nm 710nm 760nm 810nm
Scan ‘1 Veg
Date DAP Cov 2001 Sandyland, Asgrow B1 and Prime Cut 59 varieties

% r2
7/ 12 83 89 ns ns ns ns ns ns ns
7/19 90 93 ns 0.38‘ 0.45” ns 0.28‘ ns ns
7/26 97 98° 0.41" 0.31‘ 0.43“ 0.40” 0.37‘ ns ns
8/2 104 99° 0.52" 0.53" 0.56’” 0.48” 0.48" ns ns
8/9 11 1 99 0.55" 0.61‘” 0.59‘“ 0.56‘“ 0.61 ns ns
8/17 1 19 99 0.61‘“ 0.62‘“ 0.61‘” 0.58‘” 0.61‘” ns ns

2002 Montcalm Experiment Station, Diamond Cut variety

% 1'2
7/1 1 65 51° ns ns ns ns ns ns ns
7/17 71 68 ns ns ns ns ns 0.33. 0.35.
7/24 78 86 ns ns ns ns ns ns ns
8/1 86 96 ns ns ns ns ns ns ns
8/9 94 94 ns 0.40' ns ns ns ns ns
8/15 100 98 0.30‘ 0.43" 0.31‘ ns 0.25’ 0.31’ ns
8/21 106 97 ns ns ns ns ns ns ns
8/30 115 88 ns 0.27. ns ns ns ns ns
9/6 122 93 ns ns ns ns ns ns ns

2002 Montcalm Experiment Station, Goliath variety

% r2
7/1 1 65 46° ns ns ns ns ns ns ns
7/17 71 66 ns ns ns ns ns 0.30‘ 0.31‘
7/24 78 80 ns ns ns ns ns 0.29‘ 0.28‘
8/1 86 96 ns 0.29‘ 0.30‘ 0.29‘ ns 0.27‘ 0.30‘
8/9 94 93 ns ns 0.33‘ 0.35’ 0.28' ns ns
8/ 15 100 96 ns ns ns ns ns ns ns
8/21 106 94 ns ns ns ns ns ns ns
8/30 115 87 ns ns ns ns ns ns ns
9/6 122 90 ns ns ns ns ns 0.25‘ 0.25‘

2002 Sandyland, Sugar Snax variety

% r'2
7/11 82 91° ns ns ns ns ns ns ns
7/17 88 92° 0.32‘ 0.38‘ 0.46” ns 0.50“ ns ns
7/24 95 92 ns ns ns ns ns ns ns
8/1 103 97 0.65’” 0.48“ 0.54“ 0.35‘ 0.44" ns ns
8/9 111 96 0.45“ ns 0.37' 0.29‘ ns ns ns
8/ 15 117 99 ns ns ns 0.36. ns ns ns

 

T DAP = Days after planting

° Calculated approximation using NDVI rather than approximation developed from images.
Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F -ratio is not significant

0 O. 0..

104

Table 28. Linear regression coefficients of reflectance at individual wavelengths vs top biomass of
harvested tops from selected locations.

 

Ave

 

 

 

 

 

 

 

 

 

510nm 560nm 610nm 660nm 710nm 760nm 810nm

Scan Veg
Date DAPT Cov 2001 Montcalm Experiment Station, Diamond Cut variety

% r2
7/ 12 65 35 ns ns ns ns ns ns ns
7/20 73 64 ns ns ns ns ns 0.27. ns
7/26 79 85° ns ns ns ns ns ns ns
8/2 86 94 0.28' ns ns ns ns 0.25‘ 0.27'
8/9 93 97 ns ns ns ns ns ns ns
8/17 101 96 ns 0.41" ns ns 0.33' ns ns
9/6 121 95 0.37‘ 0.58‘” 0.61'“ 0.48“ 0.65‘” 0.33‘ ns

2001 Sandyland, Asgrow BI and Prime Cut 59 varieties

% :2
7/5 76 63 ns ns ns ns ns 0.30‘ 0.28‘ 1
7/12 83 89 ns 0.27‘ ns . ns 0.26‘ 0.35:. 0.36:. i;
7/19 90 93 ns ns 0.27 ns ns 0.43 0.45 ~
7/26 97 98° 0.42" 0.32’ 0.47” 0.35‘ 0.38‘ 0.44" 0.47“
8/2 104 99° 0.31‘ 0.40‘ 0.41” ns 0.43“ 0.47" 0.54“
8/9 1 1 1 99 0.40‘ 0.48“ 0.48“ 0.45” 0.47“ ns ns
8/17 119 99 0.62‘” 0.62‘“ 0.63‘” 0.65‘” 0.58‘” ns ns

2002 Montcalm Experiment Station, Diamond Cut variety

% 1'2
7/1 1 65 51° ns ns ns ns ns 0.36‘ 0.36‘
7/17 71 68 ns ns ns ns ns 0.38‘ 0.42‘
7/24 78 86 ns ns ns ns ns 0.32‘ 0.32'
8/1 86 96 ns ns ns ns ns 0.36‘ 0.33'
8/9 94 94 ns 0.29. ns ns ns ns ns
8/15 100 98 0.41“ ns nS 0.32' ns ns ns
8/21 106 97 ns ns ns ns ns ns ns
8/30 115 88 ns ns ns ns ns ns ns
9/6 122 93 ns ns ns ns ns ns ns

2002 Montcahn Experiment Station, Goliath variety

% 1'2
7/11 65 46° 0.52“ 0.52” 0.52“ 0.52“ 0.51” ns ns
7/17 71 66 0.41” 0.41" 0.40” 0.39" 0.34‘ 0.45” 0.45“
7/24 78 80 0.45“ 0.48" 0.46“ 0.45" 0.47" 0.37‘ 0.36‘
8/1 86 96 0.39" 0.39“ 0.46“ 0.41“ 0.41“ 0.27' ns
8/9 94 93 ns ns ns ns ns ns ns
8/ 15 100 96 ns ns ns ns ns ns ns
8/21 106 94 0.29. ns ns ns ns ns ns
8/30 1 15 87 ns ns ns ns ns ns ns
9/6 122 90 ns ns ns ns ns 0.47” 0.48"

 

105

Table 28 (cont’d)

 

 

 

Scan + 3:; 510nm 560nm 610nm 660nm 710nm 760nm 810nm
Date DAP Cov 2002 Sandyland, Sugar Snax variety

% 1'2
7/11 82 91° ns ns ns ns ns ns ns
7/17 88 92° ns ns ns ns 0.27. ns ns
7/24 95 92 0.39" ns 0.36' 0.38‘ 0.35’ ns ns
8/1 103 97 0.30‘ ns 0.27‘ 0.51” ns 0.38‘ 0.36‘
8/9 1 1 1 96 ns ns ns 0.33‘ ns 0.46” 0.45“
8/15 117 99 ns ns ns ns ns 0.51“ 0.51"

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.
' .. ... Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

2001 was sporadic at best, and correlation to Dry Root Biomass nonexistent, the
combination of the two parameters in the form of Root:Shoot ratio (Table 31) were
significantly correlated with reflectance at a level that could predict the partitioning of N
treatments about a month before harvest. Reflectance at 560 and 710 nm, as it related to
Root:Shoot, peaked about a week before harvest at 1’2 = 0.87 and at 610 nm r2 = 0.83. At
Sandyland in 2001, the coefficient of determination could explain greater than 50 % of
the differences as early as July 26, and even though it fluctuated, by season end r2 = 0.71
at 510 nm and 0.70 at 560 nm. The Goliath variety (2002) exhibited a split correlation.
Root:Shoot at harvest correlated to mid-season visible reflectance and to late season NR
reflectance.

The outcome of every site was different, and the relationship between the in-
season reflectance measurements and the various harvest parameters revealed the
importance of a healthy canopy throughout the season. In 2001 and 2002 various
setbacks at the Experiment Station affected the correlation between reflectance and the

harvest parameters. Correlation to reflectance was significant only for the Root:Shoot

106

 

Table 29. Linear regression coefficients of reflectance at individual wavelengths vs N uptake in
harvested roots from selected locations.

 

Ave 510nm 560nm 610nm 660nm 710nm 760nm 810nm

 

Scan Veg

 

 

 

 

 

 

 

 

Date DAPT Cov 2001 Sandyland, Asgrow BI and Prime Cut 59 varieties

% :2
7/12 83 89 ns 0.28' ns ns 0.26’ 0.25‘ 0.25‘
7/19 90 93 . ns ns ns ns ns 0.41" 0.41"
7/26 97 98° 0.25‘ 0.25‘ 0.31‘ ns 0.30’ 0.49” 0.50”
8/2 104 99° 0.27‘ 0.42“ 0.44“ ns 0.40” 0.50” 0.52‘“
8/9 1 1 1 99 0.51“ 0.62‘” 0.62'“ 0.54“ 0.61‘” ns 0.29‘
8/17 119 99 0.63‘” 0.68‘“ 0.71‘“ 0.67‘” 0.67‘” ns us

2002 Montcalm Experiment Station, Diamond Cut variety

% :2
7/11 65 51° 0.35' 0.40“ 0.39” 0.38‘ 0.41” ns ns
7/17 71 68 0.48" 0.50“ 0.49" 0.50" ns 0.36‘ 0.32‘
7/24 78 86 0.28‘ 0.39‘ 0.33‘ 0.30’ 0.40‘ ns ns
8/1 86 96 ns ns ns ns ns ns ns
8/9 94 94 ns ns ns ns ns ns ns
8/15 100 98 ns ns ns ns 0.41‘ ns ns
8/21 106 97 ns ns ns ns ns ns ns
8/30 115 88 ns ns ns 0.28. ns ns ns
9/6 122 93 0.33‘ ns 0.29‘ 0.51” ns ns ns

2002 Montcalm Experiment Station, Goliath variety

% :2
7/11 65 46° ns ns ns ns ns ns ns
7/17 71 66 0.34‘ 0.36' 0.35' 0.34‘ 0.38‘ ns ns
7/24 78 80 ns ns ns ns ns ns ns
8/ 1 86 96 ns ns ns ns ns ns ns
8/9 94 93 ns ns ns ns ns ns ns
8/15 100 96 ns ns ns ns ns 0.39‘ 0.39'
8/21 106 94 ns ns ns ns 0.27‘ ns ns
8/30 115 87 ns ns 0.25‘ 0.29' ns ns ns
9/6 122 90 ns ns ns ns ns ns ns

2002 Sandyland, Sugar Snax variety

% 1'2
7/1 1 82 91° 0.39” ns 0.58'“ 0.54“ 0.25' 0.31‘ 0.30'
7/17 88 92° 0.34‘ ns 0.46“ 0.51" ns 0.43“ 0.41"
7/24 95 92 0.31‘ ns 0.33’ 0.36‘ ns 0.38‘ 0.35’
8/1 103 97 0.34‘ 0.28‘ 0.36‘ 0.37‘ 0.28‘ ns 0.26‘
8/9 11 1 96 ns ns ns 0.36’ ns 0.34' 0.35‘
8/15 117 99 ns ns 0.29‘ 0.49“ ns 0.30‘ 0.32‘

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.

Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

107

Table 30. Linear regression coefficients of reflectance at individual wavelengths vs root biomass of
harvested roots from selected locations.

 

Ave 510nm 560nm 610nm 660nm 710nm 760nm 810nm

 

 

 

 

 

Scan 1' Veg
Date DAP Cov 2002 Montcalm Experiment Station, Goliath variety

% :2
7/11 65 46° ns ns ns ns ns ns ns
7/17 71 66 ns ns ns nS ns ns ns
7/24 78 80 ns ns ns ns ns ns ns
8/1 86 96 nS ns ns ns ns ns ns
8/9 94 93 ns ns ns ns ns 0.32‘ 0.31‘
8/15 100 96 nS ns ns ns ns 0.34‘ 0.33‘ E
8/21 106 94 ns ns ns ns ns 0.25‘ 0.25' _
8/30 115 87 ns ns ns ns ns ns ns
9/6 122 90 ns ns ns ns ns ns ns

2002 Sandyland, Sugar Snax variety

% 1'2
7/1 1 82 91° ns ns 0.40“ 0.58‘“ ns 0.63'” 0.63‘" ,
7/17 88 92° nS ns ns 0.40" ns 0.56‘” 0.54“ '
7/24 95 92 ns ns ns ns ns 0.46" 0.44”
8/1 103 97 ns ns ns ns ns 0.40' 0.40‘
8/9 1 l 1 96 ns ns ns ns ns 0.35‘ 0.37‘
8/15 1 17 99 ns ns ns ns ns ns ns

 

GAP = Days after planting

° Calculated approximation using NDVI rather than approximation developed from images.
' .. ... Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

ratio in 2001. During 2002, in-season reflectance measurements were only loosely
predictive of the harvest parameters for either variety. Consequently, % N in Tops and %
N in Roots for Diamond Cut were the only parameters that were consistently significant
from August 9 to harvest. Dry Top Biomass and Root:Shoot ratio calculated for the
Goliath variety significantly correlated with early reflectance through August 9 across all
wavelengths. On August 9 reflectance could explain more than 50 % of the differences
in % N in Roots. Following August 9, correlation diminished as the canopy senesced.
Harvest was satisfactory but the canopy was unable to reflect that condition; therefore, it

could not predict it.

108

Table 31. Linear regression coefficients of reflectance at individual wavelengths vs root:shoot ratio of
biomass at harvest from selected locations.

 

 

 

 

 

 

 

 

Ave 510nm 560nm 610nm 660nm 710nm 760nm 810nm
Scan Veg
Date DAP‘r Cov 2001 Montcalm Experiment Station, Diamond Cut variety

% :2
6/13 36 0 ns ns ns ns ns ns nS
6/22 45 7 ns nS ns ns ns ns us
608 51 11 ns nS ns ns ns ns ns
7/5 58 19 0.28‘ ns ns 0.28‘ ns ns ns
7/12 65 35 0.42“ 0.39‘ 0.42“ 0.41" 0.37‘ ns ns
7/20 73 64 ns ns ns ns ns 0.26‘ 0.26' E
7/26 79 85° ns ns ns ns ns 0.28‘ 0 31‘ ~'
8/2 86 94 ns ns ns ns ns 0.30' 0 32‘
8/9 93 97 ns 0.49“ ns ns 0.27‘ ns ns
8/17 101 96 0.45" 0.77'” 0.68‘” ns 0.59‘” ns ns
9/6 121 95 0.56‘” 0.87‘” 0.83‘” 0.67‘” 0.87'” ns ns t

2001 Sandyland, Asgrow B1 and Prime Cut 59 varieties

% r2
6/ 13 54 21 ns ns ns ns ns ns ns
6/22 63 43 ns ns ns ns ns ns ns
6/28 69 63 ns ns ns ns ns ns ns
7/5 76 63 nS ns ns ns ns ns ns
7/12 83 89 0.33‘ 0.29' 0.27‘ ns ns ns ns
7/19 90 93 ns 0.36‘ 0 45“ ns 0.30‘ 0.27‘ 0.28'
7/26 97 98° 0.52“ 0.39" 0.52“ 0.56“ 0.45” 0.42“ 0.43”
8/2 104 99° 0.45“ 0.47” 0.49” 0.46“ 0.45” 0.29' 0.40‘
8/9 111 99 0.48“ 0.55“ 0.54" 0.50“ 0.53" ns ns
8/17 119 99 0.71‘” 0.70‘" 0.66‘” 0.67‘” 0.66‘“ ns us

2002 Montcalm Experiment Station, Goliath variety

% 1'2
7/1 1 65 46° 0.43” 0.41” 0.42“ 0.43“ 0.41” ns ns
7/17 71 66 0.60‘” 0.58’” 0.56‘” 0.56‘” 0.57'” ns ns
7/24 78 80 0.46“ 0.46" 0.45“ 0.45“ 0.45" 0.28’ 0.27‘
8/1 86 96 0.45“ 0.39‘ 0.50“ 0.43” 0.39‘ ns ns
8/9 94 93 ns ns ns ns ns ns ns
8/15 100 96 ns ns ns ns ns 0.32. 0.30.
8/21 106 94 ns ns ns ns ns 0.29. 0.30.
8/30 1 15 87 ns ns ns ns ns 0.45" 0.46"
9/6 122 90 ns ns ns ns ns 0.58m 0.58m

 

T DAP = Days after planting

° Calculated approximation using NDVI rather than approximation developed from images.
Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

109

In contrast, the 2001 and 2002 Sandyland sites showed better and more consistent
correlation to reflectance for several of the harvest parameters. In 2001 reflectance was
strongly correlated to % N in Roots, N Uptake in Tops and Roots, Dry Top Biomass, and
Root:Shoot ratio. In 2002 correlation between reflectance and harvest parameters was
similar but did not have the level of predictability experienced in 2001. Percent N in
Tops, Dry Top Biomass and the combination of the two parameters, N Uptake in Tops,
showed less consistency than in 2001. As discussed earlier, the effect of the N fertilizer
applications appeared to be depleted before subsequent application and the canopy
experienced temporary stress.

Selected Indices: The final test of the reflectance measurements was to correlate
the indices against harvest parameters. At the Experiment Station in 2001, the individual
reflectance measurements resulted in little correlation to harvest parameters; the indices
also exhibited the lack of correlation. Results from the Experiment Station in 2001 only
appear in Table 32.

None of the five locations exhibited significant correlation between the indices
and % N in Tops. Due to the nature of the ratio-based indices and the wavebands
sensitive to nutrient detection; this may be the expected outcome. These ratio-based
indices depend in part on a visible band and in part on 3 NR band. The absence of
significant correlation to nutrient content in the NIR wavebands weakened the indices.

Consistent, significant correlation between the indices and % N in Roots (Table
32) was established only at the Sandyland 2001 location. Timing was about the same as
individual wavebands, and continued to increase until harvest. However, none of the

indices surpassed the individual wavebands for level of significance. NDVI and the soil

110

 

adjusted variations performed about the same. GNDVI was the best correlated to % N in

Roots.

Table 32. Linear regression coefficients of indices calculated from reflectance at individual
wavelengths vs % N in harvested roots from selected locations.

 

 

 

 

 

Ave NDVI SAVISL SAVI TSAVI GNDVI
Scan Veg
Date DAPT Cov 2001 Montcalm Experiment Station, Diamond Cut variety
% 1'2
7/20 73 64. ns ns ns ns ns
7/26 79 85° ns ns ns ns ns
8/2 86 94 ns ns ns ns ns
8/9 93 97 ns ns ns ns ns
8/17 101 96 0.26‘ 0.26’ 0.26‘ 0.26" 0.45“
9/6 121 95 0.36‘ 0.36‘ 0.36‘ 0.36‘ ns
2001 Sandyland, Asgrow B1 and Prime Cut 59 varieties
% 1'2
7/19 90 93 . 0.27‘ 0.29' 0.28‘ 0.28‘ 0.48“
7/26 97 98° 0.45“ 0.45" 0.45” 0.45” 0.54”
8/2 104 99° 0.52“ 0.54“ 0.53“ 0.52" 0.58‘”
8/9 1 1 1 99 0.64‘“ 0.63‘” 0.63’“ 0.63‘“ 0.74‘”
8/17 119 99 0.73'” 0.71‘“ 0.72‘” 0.72’” 0.80‘”

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.
° .. ... Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

Yield, defined as harvest fresh weight, is similar to Dry Root Biomass, except for the
water content. The results of regression analysis between reflectance and Yield at
individual wavebands are not shown. Only NR reflectance at 760 and 810 nm showed
significant correlation that peaked in mid-season and then decreased until harvest. The
indices exhibited more consistency (Table 33), but significant correlation diminished
following mid-season. Notable is the difference in results between Yield and Dry Root
Biomass (Table 34) and the apparent influence of water content on reflectance
measurements. Only at Sandyland in 2002 were the results similar for both Yield and

Dry Root Biomass. The remaining sites differed to such extent that none of the other

111

 

locations are represented in both tables due to lack of significance.

Table 33. Linear regression coefficients of indices calculated from reflectance at individual

wavelengths vs yield as Mg ha '1 fresh weight from selected locations.

 

 

 

 

 

 

 

 

 

Ave NDVI SAVISL SAVI TSAVI GNDVI
Scan Veg
Date DAPJr Cov 2001 Sandyland, Asgrow B1 and Prime Cut 59 varieties

% 1'2
6/13 54 21 ns ns ns ns ns
6/22 63 43 0.37‘ 0.37’ 0.37‘ 0.39“ 0.39“
6/28 69 63 0.47" 0.47" 0.47" 0.47" 0.47"
7/5 76 63 0.50“ 0.50" 0.50" 0.50" 0.51”
7/12 83 89 0.34‘ 0.35‘ 0.35' 0.35‘ 0.50"
7/19 90 93 ‘ ns ns ns ns 0.28’
7/26 97 98° ns ns ns ns 0.29‘
8/2 104 99° ns ns ns ns ns
8/9 111 99 ns nS ns ns ns
8/17 119 99 ns ns ns ns us

2002 Montcalm Experiment Station, Diamond Cut variety

% 1'2
7/1 1 65 51° 0.42” 0.42” 0.42“ 0.42“ 0.42“
7/17 71 68 0.33‘ 0.34‘ 0.33‘ 0.34‘ 0.33‘
7/24 78 86 0.33' 0.33‘ 0.33‘ 0.33‘ 0.34‘
8/1 86 96 ns ns ns ns ns
8/9 94 94 ns ns ns ns ns
8/ 15 100 98 ns ns ns ns ns
8/21 106 97 ns ns ns ns ns
8/30 115 88 ns ns nS ns ns
9/6 122 93 ns ns ns ns ns

2002 Sandyland, Sugar Snax variety

% r2
7/1 1 82 91° 0.63‘” 0.63‘“ 0.63‘” 0.63‘” 0.63‘”
7/17 88 92° 0.45” 0.46“ 0.46“ 0.45" 0.40“
7/24 95 92 0.27‘ 0.28‘ 0.28‘ 0.28‘ 0.27‘
8/1 103 97 ns ns ns ns ns
8/9 1 1 1 96 ns ns ns ns ns
8/ 15 117 99 ns ns nS ns ns

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from irrrages.

O .0 0..

112

Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

Table 34. Linear regression coefficients of indices calculated fiom reflectance at individual
wavelengths vs root biomass of harvested roots from selected locations.

 

 

 

 

 

 

Ave NDVI SAVISL SAVI TSAVI GNDVI
Scan Veg
Date DAPT Cov 2002 Montcalm Experiment Station, Goliath variety

% 1'2
8/15 100 96 ns 0.25‘ ns ns ns
8/21 106 94 0.27‘ 0.28‘ 0.28‘ 0.27‘ 0.25‘
8/30 115 87 0.25‘ 0.25‘ 0.25‘ ns ns
9/6 122 90 ns ns ns ns ns

2002 Sandyland, Sugar Snax variety

% 1'2
7/11 82 91° 0.64‘” 0.64’” 0.64'” 0.64‘" 0.64‘"
7/17 88 92° 0.49” 0.50” 0.49” 0.49" 0.44”
7/24 95 92 0.30' 0.31‘ 0.30' 0.31’ 0.30'
8/1 103 97 ns ns ns ns ns
8/9 111 96 ns ns ns nS ns
8/15 1 17 99 ns ns ns ns ns

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.
. .. ... Significance of overall F-values at p 5005, 0.01, 0.001. ns = Overall F-ratio is not significant

Individual wavebands out performed the indices when compared to N Uptake in
Tops (Table 27 vs Table 35) at the Sandyland 2001 location. Sandyland was the only

location to exhibit the Significance and consistent correlation in the visible wavebands

Table 35. Linear regression coefficients of indices calculated from reflectance at individual
wavelengths vs N uptake in harvested tops from selected locations.

 

 

 

Ave NDVI SAVISL SAVI TSAVI GNDVI
Scan T Veg
Date DAP Cov 2001 Sandyland, Asgrow B1 and Prime Cut 59 varieties

% r2
7/ 12 83 89 ns ns ns ns ns
7/19 90 93 ns ns ns ns 0.35‘
7/26 97 98° 0.41' 0.40‘ 0.40’ 0.40‘ 0.39‘
8/2 104 99° 0.51“ 0.50“ 0.51“ 0.51“ 0.48“
8/9 111 99 0.48“ 0.47" 0.48” 0.48“ 0.52“
8/17 119 99 0.49” 0.48” 0.49” 0.49“ 0.53“

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.
. .. ... Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

113

(Table 27) to result in the performance shown in Table 35. Again, visible reflectance was
significantly correlated while NR reflectance was insignificant or loosely correlated
ultimately due to canopy biomass conditions.

When compared to N Uptake in Roots (Table 36), the indices out performed
individual reflectance (Table 28), specifically evident at both the 2001 and 2002
Sandyland locations. The synergy provided by the significance of both the visible and
NR wavebands was apparent, especially when compared to the results of the Experiment
Station. In addition, the indices exhibited more consistency throughout the season as Z

multiplicative effects were minimized (Epiphanio and Huete, 1994).

 

In 2001 GNDVI out performed the red reflectance indices; however, in 2002 all indices é
performed equally well. Significance was evident at mid-season when canopy coverage
was approximately 90 %, and continued to increase in significance until harvest.

Indices, when compared to Dry Top Biomass (Table 37) of the harvested tops,
Showed the same synergy as N Uptake in Roots. Indices performed slightly better and
significance was more consistent over time than individual reflectance but followed the
same trends. The indices were not well correlated to Dry Root Biomass (Table 34). The
table is included because of the function Dry Root Biomass has as part of N Uptake in
Roots and the following discussion of Root:Shoot ratio. In addition, it Shows the strength
of the NR (Table 30) wavebands as they relate to the indices.

Correlation of the indices to the Root:Shoot ratio (Table 38) offered no new
surprises. Overall, significance mirrored the trends set by the individual reflectance.
Where NR reflectance was significant along with the applicable visible bands, the

indices were better correlated and the coefficient of determination could explain more of

114

Table 36. Linear regression coefficients of indices calculated from reflectance at individual
wavelengths vs N uptake in harvested roots from selected locations.

 

 

 

 

 

 

 

 

 

Ave NDVI SAVISL SAVI TSAVI GNDVI
Scan Veg
Date DAP'r Cov 2001 Sandyland, Asgrow BI and Prime Cut 59 varieties

% r2
7/ 12 83 89 ns ns ns ns ns
7/19 90 93 ns 0.27‘ 0.26‘ 0.26‘ 0.47“
7/26 97 98° 0.43” 0.44“ 0.43" 0.44” 0.57‘”
8/2 104 99§ 0.38‘ 0.40“ 0.39“ 0.39" 0.55“
8/9 11 1 99 0.65‘” 0.65‘” 0.65‘” 0.65‘” 0.73‘”
8/17 119 99 0.71'” 0.71’” 0.71‘” 0.71‘” 0.73‘“

2002 Montcalm Experiment Station, Diamond Cut variety

% 1'2
7/1 1 65 51° 0.26‘ 0.26‘ 0.26‘ 0.25‘ 0.25‘
7/ 17 71 68 ns ns ns 0.25. ns
7/24 78 86 ns ns ns ns ns
8/1 86 96 ns ns ns ns ns
8/9 94 94 ns ns ns ns ns
8/ 15 100 98 ns nS ns ns ns
8/21 106 97 0.29‘ 0.29‘ 0.29’ 0.29‘ 0.31‘
8/30 115 88 ns ns ns ns ns
9/6 122 93 0.36‘ 0.31‘ 0.34' 0.32’ 0.29’

2002 Montcalm Experiment Station, Goliath variety

% 1'2
7/1 1 65 46° ns ns ns ns ns
7/ 17 71 66 ns ns ns ns ns
7/24 78 80 ns ns nS ns ns
8/1 86 96 ns ns ns ns ns
8/9 94 93 0.29‘ 0.28‘ 0.28‘ 0.28‘ ns
8/15 100 96 0.31‘ 0.34‘ 0.32‘ 0.32‘ 0.37‘
8/21 106 94 0.29‘ 0.29‘ 0.29‘ 0.29‘ 0.30‘
8/30 115 87 ns ns ns ns ns
9/6 122 90 ns ns ns ns ns

2002 Sandyland, Sugar Snax variety

% 1'2
7/1 1 82 91° 0.47“ 0.47” 0.47“ 0.47“ 0.47”
7/17 88 92° 0.50“ 0.50" 0.50” 0.49“ 0.50”
7/24 95 92 0.38' 0.38‘ 0.38' 0.38‘ 0.37‘
8/1 103 97 0.38‘ 0.37“ 0.37' 0.37’ 0.40”
8/9 1 1 l 96 0.42“ 0.42" 0.42” 0.42” 0.42”
8/15 117 99 0.50“ 0.50“ 0.50“ 0.50“ 0.46"

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.
° .. ... Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

115

 

Table 37. Linear regression coefficients of indices calculated fi'om reflectance at individual
wavelengths vs top biomass of harvested tops from selected locations.

 

 

 

 

 

 

 

 

Ave NDVI SAVISL SAVI TSAVI GNDVI
Scan Veg
Date DARr Cov 2001 Sandyland, Asgrow B1 and Prime Cut 59 varieties

% :2
7/12 83 89 ns 0.25‘ 0.25‘ 0.25‘ 0.32‘
7/19 90 93 0.37‘ 0.40' 0.38' 0.38' 0.58'”
7/26 97 98§ 0.54“ 0.55” 0.54” 0.54“ 0.60‘“
8/2 104 99° 0.43“ 0.45” 0.44” 0.44“ 0.53“
8/9 111 99 0.52" 0.52” 0.52” 0.57” 0.56“
8/17 119 99 0.63’” 0.62‘” 0.63‘” 0.63'” 0.62’”

2002 Montcalm Experiment Station, Goliath variety

0/0 1‘2
7/11 65 46° 0.51" 0.51“ 0.51" 0.51” 0.51"
7/17 71 66 0.41“ 0.41” 0.41“ 0.41" 0.42“
7/24 78 80 0.45” 0.45“ 0.45“ 0.44” 0.46“
8/1 86 96 0.42" 0.42“ 0.42" 0.42" 0.42” l;
8/9 94 93 ns nS ns ns ns
8/15 100 96 ns ns ns ns ns
8/21 106 94 ns ns ns ns ns
8/30 115 87 ns ns ns ns ns
9/6 122 90 0.39” 0.42” 0.40" 0.42” 0.49”

2002 Sandyland, Sugar Snax variety

% :2
7/1 1 82 91° ns ns ns ns ns
7/17 88 92° ns ns ns ns us
7724 95 92 0.37‘ 0.36' 0.37‘ 0.37' 0.34’
8/1 103 97 0.41” 0.41” 0.41” 0.41" 0.33‘
8/9 11 1 96 0.37‘ 0.38‘ 0.38’ 0.38‘ 0.30‘
8/15 117 99 ns ns ns ns ns

 

T DAP = Days after planting
° Calculated approximation using NDVI rather than approximation developed from images.
° .. ... Significance of overall F-values at p $0.05, 0.01, 0.001. ns = Overall F-ratio is not significant

the differences.

The importance of a healthy canopy throughout the season is supported by the
varied results obtained when Spectral measurements were compared to the various
parameters measured at the five study locations. Harvest parameters indicative of the N

status of the tops were % N in Tops, N Uptake in Tops, Dry Top Biomass, and

116

Table 38. Linear regression coefficients of indices calculated fiom reflectance at individual
wavelengths vs root:shoot ratio of biomass at harvest from selected locations.

 

 

 

 

 

 

 

Ave NDVI SAVISL SAVI TSAVI GNDVI
Scan Veg
Date DAPT Cov 2001 Montcahn Experiment Station, Diamond Cut variety
% :2
7/5 58 19 0.39‘ 0.40‘ 0.40‘ 0.40‘ 0.38‘
7/12 65 35 0.59“ 0.59” 0.59“ 0.59” 0.60“
7/20 73 64‘ ns ns ns ns ns
7/26 79 85° ns ns ns ns ns
8/2 86 94 ns ns ns ns ns
8/9 93 97 ns ns ns ns ns
8/17 101 96 ns ns ns ns 0.26‘
9/6 121 95 0.46" 0.42“ 0.44“ 0.43“ 0.58‘”
2001 Sandyland, Asgrow B1 and Prime Cut 59 varieties
% :2
7/5 76 63 ns ns ns ns ns
7/12 83 89 ns ns ns ns ns
7/19 90 93 ‘ 0.47” 0.48“ 0.48" 0.48“ 0.63‘”
7/26 97 98° 0.61‘” 0.61‘” 0.61‘” 0.61‘“ 0.59‘”
8/2 104 99° 0.65‘“ 0.66‘“ 0.65‘” 0.65'” 0.59‘“
8/9 1 11 99 0.55“ 0.55“ 0.55“ 0.55“ 0.59‘“
8/17 1 19 99 0.64‘” 0.62‘” 0.63’” 0.63‘” 0.66‘”
2002 Montcalm Experiment Station, Goliath variety
% 1'2
7/1 1 65 46° 0.43“ 0.43“ 0.43“ 0.43" 0.42“
7/17 71 66 0.52“ 0.52" 0.52“ 0.51” 0.52“
7/24 78 80 0.44" 0.44" 0.44" 0.44" 0.44”
8/1 86 96 0.39‘ 0.38‘ 0.38‘ 0.38‘ 0.32‘
8/9 94 93 ns ns ns ns ns
8/15 100 96 ns ns ns ns ns
8/21 106 94 ns ns ns ns ns
8/30 1 15 87 0.29’ 0.32‘ 0.30‘ 0.30‘ 0.37‘
9/6 122 90 0.35‘ 0.38‘ 0.36‘ 0.36‘ 0.45”

 

T DAP = Days after planting

° Calculated approximation using NDVI rather than approximation developed from images.

Significance of overall F-values at p $0.05, 0.01, 0.001. as = Overall F-ratio is not significant.

Root:Shoot ratio. The biomass segment of N Uptake made an important contribution and

where the canopies suffered problems, biomass was not consistently indicative of plant

response to N. The % N portion, although important, did not appear to affect correlation

to the same extent as biomass. The Root:Shoot ratio showed Significant correlation

 

where canopies were healthy or recovering. Again, the most important factor appeared to
be the top biomass. Overall correlation to those parameters important to healthy tops was
best measured at 510, 560, 610, and 710 nm.

The canopy also showed promise as a predictive tool for monitoring N content of
roots. Reflectance was correlated to % N in Roots at least sporadically at all locations.
At the 2001 Sandyland site, reflectance at 560 and 610 nm explained as much at 86% of
the differences in N content followed by reflectance at 710 nm where 1‘2 = 0.84. Future
studies focused on the storage roots may Show that canopy reflectance can be an

important tool for managing N in the roots as well as tops.

Conclusion

This study has shown that remote sensing may work for the vegetable industry,
especially for crops that are typically mechanized, such as carrot. It was intended to
explore the possibility that in-season N management for quality carrot tops could be
successful using remote sensing. To that end, the first objective was to determine which
wavelengths correlate to the physical parameters typically used to evaluate the health of
the carrot crop.

The visible wavebands, even where there was plant disease, were able to provide
insight about the crop response to N availability. The visible bands centered at 560 and
710 nm were the earliest to correlate with Soil-N and generally remained Significant
throughout the season explaining as much as 90% of the difference between treatments
where the canopy was healthy. These results are in agreement with the findings of

Thomas and Gausman (1977), Blackrner et 31. (1994), Blackrner et al. (1996a, 1996b),

118

 

and Masoni et al. (1996) in agronomic crops. In carrot, wavebands centered at 510 and
610 were also significantly correlated to the physical parameters analyzed and may prove
to be important to future studies in carrot.

NR reflectance at 760 and 810 nm was weakly correlated with plant N status
when the plants were affected by variables other than N treatments and when the canopy
reached full coverage. This phenomenon also had an impact on the usefulness of the
indices which were out performed by visible reflectance when NR reflectance was not E
significantly correlated to the parameter of interest and correlation at 560 or 660 nm was

weak (r2 $0.40).

 

Indices covered in this study are ratio based, subject to the significance exhibited
by two wavebands, a visible and a near infrared. NDVI, SAVI, SAVISL, and TSAVI use
the same two wavebands: the visible band at 660 nm and the near infiared band at 810
nm. The only difference between the four indices is the handling of the soil background.
NDVI has no adjustment for soil background, and the three SAVI-type indices are
adjusted according to equations 5 and 6 (pages 58 and 59). Soil adjusted indices were
designed to be more sensitive to changes in NR (Epiphanio and Huete, 1994) influenced
by vegetative cover and vigor (Flowers et al., 2003b). While NDVI lagged slightly
behind the others in carrot, it was no less sensitive to canopy coverage than the SAVI-
type indices. The typical density of carrot planting and thickness of the maturing canopy
may hamper the NR indicator of vegetative cover and vigor, where biomass differences
may be nonexistent. The sensitivity of any of the indices in carrot seemed to be
dependant on two factors: 1) the amount of soil in the pixel; and 2) the dominance Shown

by the visible wavelength, the indicator of nutrient status. GNDVI, like NDVI, does not

119

include a soil adjustment, and the visible waveband was replaced with reflectance at 560
nm, where reflectance was generally more dominant as a nutrient status indicator and
therefore, out performed the other indices in its assessment of N status. GNDVI was the
only index that on occasion could explain differences in Petiole-N where the coefficient
of determination was greater than 50%. These results are in agreement with Blackrner et
al. (1994), Schepers et al (1992), and Schepers et al. (1996).

Where full canopy existed, none of the indices plateaued indicated by associated
high coefficients of determination. This relationship may be attributed to the nature of the

canopy, made up of layers of lacy structured leaves resulting in multilayer Shadowing and

 

possible micro-views of the soil. In addition, it may be due to the manner in which the 9.-
“L” adjustment in certain indices was used. The maximum value of NDVI was 0.91, as
expected in dense canopy (Aparicio et al., 2000; Epiphanio and Huete, 1994). TSAVI
also stayed within the range of 0 to 1 at a maximum value of 0.91 in full canopy similar
to NDVI. SAVISL and SAVI exceeded their expected range of 0 tol at 1.99 and 1.35,
respectively. Apparently, the use of a static “L” in dense canopies influences the range
restrictions built into the equation, but allows those indices to continue to increase where
they may have saturated or plateaued.

The second objective was to determine if reflectance could be used as an in-
season management tool. Soil-N status was significantly correlated to 560 and 710 nm at
approximately 89 to 94% vegetative cover, 93 to 94 days after planting and generally
after the second N fertilizer application until harvest. This was at least 35 days before
harvest, which was enough time to use intervention strategies to fertilize without

accumulating excess N in the storage root. In addition, the influence of soil N treatments

120

on the crop canopy showed significant correlation to N treatments in about two weeks
after fertilizer application and remained constant or decreased as the N application was
“used-up” by the plants in about 25 days.

In carrot, remote sensing research is in the early stage. Correlation between
reflectance in certain wavebands has been established, but critical reflectance values used
to estimate N fertilizer requirements, whether as individual wavebands or indices, have
not been determined. For example, while 560 nm seems to be the best individual
waveband, critical values may be linked to relative reflectance measurements due to the
numerous shades of healthy green reflected by the many varieties of carrot used in
production. It is important to note that remote sensing used to manage N will be useful in

well managed fields where the canopy is healthy and the fields are free of weeds.

121

 

References

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

Aparicio, N., D. Villegas, J. Casadesus, J .L. Araus, and C. Royo. 2000. Spectral
vegetation indices as nondestructive tools for determining durum wheat yield.
Agron. J. 92:83-91.

Baret, F., G. Guyot, and D]. Major. 1989. TSAVI: A vegetation index which minimizes
soil brightness effects on LAI and APAR estimation. p. 1355-1358171 Proc. .
IGARRS ’89 Can. Symp. Remote Sensing, 12th, Vancouver, BC, Canada 10-14 i
July 1989.

Bausch, WC. 1993. Soil background effects on reflectance-based crop coefficients for
corn. Remote Sens.Environ. 46:213-222.

 

Blackrner, T.M., and J .S. Schepers. 1994. Techniques for monitoring crop nitrogen status
in corn. Commun. Soil. Sci. Plant Anal. 25(9&10):1791-1800.

Blackrner, T.M., J .S. Schepers, and GE. Varvel. 1994. Light reflectance compared with
other nitrogen stress meansurements in corn leaves. Agron. J. 86:934-938.

Blackrner. T.M., J .S. Schepers, G.E. Varvel, and EA. Walter-Shea. 1996a. Nitogen
deficiency detection using reflected shortwave radiation from irrigated corn
canopies. Agron. J. 88:1-5.

Blackrner, T.M., J .S. Schepers, G.E. Varvel, and GE. Meyer. 1996b. Analysis of aerial
photography for nitrogen stress with corn fields. Agron. J. 88:729-733.

Brown, J.S. 1972. Forms of chlorophyll in vivo. Ann. Rev. Plant Physiol. 23:73-86.

Carlson, R.E., D.N. Yarger, and RH. Shaw. 1971. Factors Affecting the spectral
properties of leaves with special emphasis on leaf water status. Agron. J. 63:486-
489.

Chang, J., D.E. Clay, K. Dalsted, S. Clay, and M. O’Neill. 2003. Corn (Zea mays L.)
yield prediction using multispectral and multidate reflectance. Agron. J. 95: 1447-
1453.

Clarke, T.R., E.M. Barnes, and RI. Pinter. 2001. Remote sensing of chlorophyll content
in broccoli using a 3-band radiometer. In Annual Meetings Abstracts [CD-ROM].
ASA, CSSA, and SSSA, Madison, WI.

122

Clarke, T.R., E.M. Barnes, P.J. Pinter, B.A. Kimball, J .A. Haberland, E.A. Riley, and
T.L. Thompson. 2000. A three-band, two dimensial reflectance index for wheat

and cotton canopy chlorophyll content. p. 27. In 2000 Agronomy abstracts. ASA,
Madison, WI.

Daughtry, C.S.T., K.P. Gallo, and ME. Bauer. 1983. Spectral estimates of solar radiation
intercepted by com canopies. Agron. J. 75:527-531.

Dep.of Plant and Soil Sciences Oklahoma State University. 2004. Indirect measures of
plant nutrients. [Online]. Available at
http://www.dasnr.okstate.edu/nitrogen_use/indirect measures of plant_nutri.htm
(verified 23 Mar. 2004.)

Elvidge, CD. and R.J. Lyon. 1985. Influence of rock-soil spectral variation on the
assessment of green biomass. Remote Sens. Environ. 17:265-279.

Epiphanio, J .C., and AR. Huete. 1994. Dependence of NDVI and SAVI on sun/sensor
geometry and its effect on fAPAR relationships in alfalfa. Remote Sens. Environ.
5 l :3 5 1 -3 60.

Flowers, M., R. Weisz, and R. Heiniger. 2001. Remote sension of winter wheat tiller
density for early nitrogen application decisions. Agron. J. 93:783-789.

Flowers, M., R. Weisz, R. Heiniger, B. Tarleton, and A. Meijer. 2003. Field validation of
a remote sensing technique for early nitrogen application decisions in wheat.
Agron. J. 95:167-176.

Flowers, M., R. Weisz, and R. Heiniger. 2003. Quantitative approaches for using color
infrared photography for assessing in-season nitrogen status in winter wheat.
Agron. J. 95 :1 189-1200.

Gates, D.M., H.J. Keegan, J.C. Schleter, and V.R. Weidner. 1965. Spectral properties of
plants. Appl.Optics 4:1 1-20.

Gitelson, A.A., Y.J. Kaufman, and MN. Merzlyak. 1996. Use of a green channel in
remote sensing of global vegetation from EOS-MODIS. Remote Sens. Environ.
58:289-298.

Hemphill, Jr., DD, and T .L. Jackson. 1982. Effect of soil acidity and nitrogen on yield
and elemental concentration of bush bean, carrot, and lettuce. J. Amer. Soc. Hort.
Sci. 107:740-744.

Huete, AR. 1988. A Soil-adjusted vegetation index (SAVI). Remote Sens. Environ.
25:295-309.

123

Jordan, CF. 1969. Derivation of leaf area index from quality of light on the forest floor.
Ecology 50:663-666.

Lillesand, T.M., and R.W. Kiefer. 2000. Remote sensing and image interpretation. 4th ed.
John Wiley & Sons, New York.

Maas, S.J., and JR. Dunlap. 1989. Reflectance, transmittance, and absorptance of light
by normal, etiolated, and albino corn leaves. Agron. J. 81:105-110.

Marshner, H. 1998. Mineral nutrition of higher plants. 2nd. Ed. Academic Press, San
Diego, CA.

Masoni, A., L. Ercoli, and M. Mariotti. 1996. Spectral properties of leaves deficient in
iron, sulfur, magnesium, and manganese. Agron. J. 88:937-943.

Milton, N.M., B.A. Eiswerth, and CM. Ager., 1991. Effect of phosphorus deficiency on
spectral reflectance and morphology of soybean plants. Remote Sens. Environ.
36:121-127.

 

Moss, R.A., and W. E. Loomis. 1952. Absorption spectra of leaves. I The visible
spectrum. Plant Physiol. 27:370-391.

Osborne, S.L., J.S. Schepers, D.D. Francis, and MR. Schlemmer. 2002. Detection of
phosphorus and nitrogen deficiencies in corn using spectral radiance
measurements. Agron. J. 94:1215-1221.

Pierce, F .J ., D.D. Warncke, and M.W. Everett. 1995. Yield and nutrient variability in
glacial soils of Michigan. In “Proceedings of the Second International Conference
on Site Specific Management of Agricultural Systems, Bloomington/Minneapolis,
MN, 27-30 March 1994” (RC. Robert, R. H. Rust, and WE. Larson, Eds.), ASA
Miscellaneous Publications, pp. 133-151. ASA,CSSA, and SSSA, Madison, WI.

Rondeaux, G., M. Steven, and F. Barret. 1995. Optimization of soil-adjusted vegetation
indices. Remote Sens. Environ. 55:95-107.

Sembiring, H., H.L. Lees, W.R. Raun, G.V. Johnson, J .B. Solie, M.L. Stone, M.J.
DeLeon, E.V. Lukina, D.A. Cossey, J.M. LaRuffa, C.W. Woolfolk, S.B. Phillips,
and WE. Thomason. 2000. Effect of growth stage and variety on spectral
radiance in winter wheat. J. Plant Nutr. 23(1): 141-149.

Senay, G.B., A.D., Ward, J .G. Lyon, N.R. Fausey, and SE. Nokes. 1998. Manipulation

of high spatial resolution aircraft remote sensing data for use in site-Specific
farming. Trans. ASAE 41:489-495.

124

Schepers, J.S., D.D. Francis, M.F. Vigil, and RE. Below. 1992. Comparison of corn leaf
nitrogen concentration and chlorophyll meter readings. Commun. Soil Sci. Plant
Anal. 23:2173-2187. -

Schepers, J .S., T.M. Blackrner, W.W. Wilhelm, and M. Resende. 1996. Transmittance
and reflectance measurements of corn leaves from plants differing in nitrogen and
water supply. J. Plant Physiol. 148:523-529.

Shanahan, J.F., J .S. Schepers, D.D. Francis, G.E. Varvel, W.W. Wilhelm, J .M. Tringe,
M.R. Schlemmer, and DJ. Major. 2001. Use of remote-sensing imagery to
estimate corn grain yield. Agron. J. 93:583-589.

Thomas, J .R. and H.W. Gausman. 1977. Leaf reflectance vs. leaf chlorophyll and
carotenoid concentrations for eight crops. Agron. J. 69:799-802.

Tucker, C.J. 1979. Red and photographic infrared linear combinations for monitoring
vegetation. Remote Sens. Environ. 82127-150.

USDA Pest Management Centers. 1999. Carrots in Michigan: Overview. USDA Crop
Profiles, Michigan. [Online]. Available at
http://pestdata.ncsu.edu/cropprofiles/docs/micarrots.html (verified 28 April 2003).

US. Water Conservation Laboratory, Remote Sensing Research Program. 2004. How a
vegetation index works. [Online]. Available at
http://www.uswcl.arS.gov/epd/remsen/ViNiworkShtm (verified 24 Mar. 2004).

Walburg, G., M.E. Bauer, C.S.T. Daughtry, and T.L. Housley. 1982. Effects of nitrogen
nutrition on the growth, yield, and reflectance characteristics of corn canopies.
Agron. J. 74:677-683.

Walworth, J .L. 1998. Crop production and soil management series: Field crop
fertilizer recommendations for Alaska: vegetables. Bull. FGV-00643. Alaska
Coop. Ext., Univ. of Alaska, Fairbanks. (Available on-line at
http://www.uaf.edu/coop-ext/publications/freepubsL) (Verified 28 April 2003.)

Warncke, DD. 1996. Soil and plant tissue testing for nitrogen management in carrots.
Commun. Soil Sci. Plant Anal. 27(3&4):597-605.

White, J .M., and JD. Strandberg. 1978. Early root grth of carrots in organic soil. J.
Amer. Soc. Hort. Sci. 103(3):344-347.

Wiegand, C.L., A.J. Richardson, D.E. Escobar, and A. H. Gerbermann. 1991. Vegetation
indices in crop assessments. Remote Sens. Environ. 35:105-119.

Williams, J .D., N.R. Kitchen, and RC. Scharf. 2001. Detecting spatially variable corn
nitrogen needs using green reflectance from 35mm photographs. p. 183-190. In

125

 

Proc. North Central Extension — Industry Soil Fertility Conf.,Ames, IA, 14-15
Nov. 2001.

Wolfe, D.W., D.W. Henderson, T.C. Hsiao, and A. Alvino. 1988. Interactive Water and
Nitrogen Effects on sensescence of Maize. II. Photosynthetic decline and
longevity of individual leaves. Agron. J. 80:865-870.

Woolley, J.T. 1971. Reflectance and transmittance of light by leaves. Plant Physiol.
47:656-662.

Yoshioka, H., R. Miura, A.R. Huete, and B.D. Ganapol. 2000. Analysis of vegetation
isolines in red-NR reflectance space. Remote Sens. Environ. 74:313-326.

Zandstra, B.H., D.D. Warncke, E.J. Grafuis, and ML. Lacy. 1986. Carrots commercial
recommendations. Michigan State University Ext. Bull. E1437. Michigan State
Univ., East Lansing. (Available on-line at
http://www.msu.edu/~zandstra/extbult/carrotsl986.html.) (Verified 03 January
2004.)

126

Chapter III

SAVI Determination in Carrots: Comparing Constant and
Dynamic Soil Adjustment Factors

Introduction

Reliable interpretation of reflectance measurements of vegetation in incomplete
canopies is confounded by the influence of the soil background. The sun angle, view
angle and atmospheric conditions that alter remote sensing spectral signatures are
increasingly corrected by improvements in atmospheric models. However, the canopy
background “brightness” that affects the vegetation indices (VI) is not easily corrected
and must be handled within the VI equation itself (Gao et al., 2000).

Soil background has an impact on vegetation parameters because the spectral
response of soil generally rises gradually from the blue wavebands across the visible and
near infrared (NR) part of the spectrum. The spectral response of vegetation in the
visible bands is punctuated with peaks and valleys, as the reflectance measurements
reveal Si grrature absorption bands of chlorophyll pigments. In NR, vegetative responses
rise above the spectral response of soil. When a pixel contains both soil and vegetation
information, the spectral responses associated with differences in vegetative parameters,
such as crop development or in-season water and N management, are diluted. In
addition, the soil background is variable and sensitive to soil type and wetting and drying
cycles (Huete, 1987a; Li et al., 2001). For example, Stoner and Baumgardner (1981)
examined a sample of 485 soils, each with a distinct spectral Signature, and identified five

distinct soil reflectance curve patterns classified by the curve shape and absorption bands.

127

The five curves had certain characteristics pertaining primarily to organic matter and iron
oxide content. They found the spectral signatures for all 485 soils reflected the specific
Spectral properties of some of the same traits that identify the taxonomic suborders to
which these soils belong, as defined by the Soil Taxonomy (1975). In addition,
Rondeaux et al. (1995) found that organic matter adds important variation to the spectral
Signature and generally reduces reflectance measurements throughout the measured
spectrum. In general, soil type has a greater impact on the spectral properties of the soil A
background than either moisture or soil roughness by as much as one order of magnitude .-

(Rondeaux et al., 1995). Changes in moisture or roughness are revealed as movement up

 

and down the same soil line: the plot of NR reflectance versus red reflectance. The soil é
type will alter the slope of the soil line. For example, where the crop represented a 37%
soil cover, overall reflectance was almost three times greater on light colored soils than
on darker soils (Ma et al., 2001).

A major goal in remote sensing research of vegetation canopies is the separation
of spectral changes due to vegetative response from those attributed to soil background;
especially where studies involve spatial and temporal changes (Huete, 1987a). Soil
adjusted vegetation indices, developed for the purpose of correcting the background
“brightness” have produced varying degrees of success depending on the canopy density.
The objective of this study is to 1) determine whether the fc model (Qi, 2000), described
below, is a reasonable estimate of canopy coverage; and 2) determine whether fc can be
successfully substituted in SAVI as L = (1- fc) as a dynamic soil adjustment factor when

the amount of vegetation is unknown.

128

Literature Review
Huete (1987a), and later Gao et al. (2000) modeled the relationship between
canopy response and underlying bare soil reflectance using the following equation where:
dm =E0rstc2+E0rc [1]
0 dm = the composite spectra of the soil-canopy mixture,
0 Eol‘stc2 = the soil dependent component: the product of E0 (global irradiance), 1'S

(soil spectra), and tc2 (the slope that represents the upward and downward
transmittance of irradiance through the canopy),
o Eorc = the vegetative component.

Huete (1987a) found that the relationship was linear (Eq. 1) for each waveband,
indicating that first order soil-vegetation interaction was sufficient to explain measured
spectral response. A variation of Eq. 1 was set forth in another study by Huete (1987b)

where:

r. = r. + bir. [21
o 1'c = composite canopy reflectance,

- rv = vegetative component reflectance,

0 1'5 = bare soil reflectance, and

0 bl = slope (transmittance).

Gao et al. (2000), and Huete (1987a, 1987b), comparing the usefulness of various
vegetation indices in eliminating soil background contamination, used Eq: 1 and 2 to

separate the composite reflectance measurements. They found this is only possible when

129

the soil reflectance is known. In their methodology, canopy spectra were measured
repeatedly over a developing cotton canopy to measure the influence of progressively
greater coverage over four different soil types. On each measurement occasion, the four
soils on trays were interchanged under the canopy using 3 support frame. Using Eq.l and
2, canopy responses were plotted against the bare soil spectra, separately for each
wavelength. When the canopy coverage is held constant against various soil
backgrounds, the response is linear. As canopy coverage increases, the slope changes,
ranging from 1 when vegetation coverage is zero to 0 at 100% canopy coverage. The y-

intercept represents the point at which influence from the soil background is zero. The

slope of the line represents the two-way global canopy transmittance, tc2 (Huete, 19873).

If the Slope, tcz, of each measurement is plotted against the measured spectrum,

the appearance of the resulting spectral signatures increasingly resembles that of a
vegetation signature as the canopy coverage increases. In Huete (19873), the likeness
was most pronounced at 90% canopy coverage. At canopy coverage denser than 90%,
transmittance values approach zero for all wavebands, producing zero slope (Huete,

1987a). Therefore, the soil component is equal to the soil reflectance multiplied by the
slope, tcz, and has the appearance of both underlying bare soil as well as the overlying

transmitted vegetation signature. It consists of all radiant flux reaching the sensor above,
that has interacted with the soil background including that which is reflected fiom the soil
and has been scattered by the plant canopy before reaching the sensor (Huete, 19873).
By subtracting the soil component from the composite reflectance, the derived plant
spectra represent only the vegetative component, free of soil and the backseatter from the

canopy.

130

 

Red and NR wavebands are usually associated with vegetative growth. Of the
two, NR is most sensitive to increases in canopy coverage (Gao et al., 2000; Huete,
19873), but response to changes in vegetation is dependent on underlying soil brightness.
Bright soils tend to decrease modeled canopy reflectance, while darker soils increase it
(Gao et al., 2000). NR (760 to 900 nm) response represents several layers of canopy as a
result of low absorption by leaf elements and high reflectance and transmittance. Soil
and plant spectral interaction is strong in this area because of high NR flux scattering
within the vegetative canopy (Huete, 19873). The red waveband (630-690 nm)
reflectance from the canopy generally decreases with increasing vegetative coverage
(Gao et al., 2000), and may only represent the uppermost leaf layers of the canopy due to
the intense absorption by chlorophyll pigments. The red waveband is relatively
insensitive to changes in vegetative coverage amounts (Huete, 1987a) compared to NR.
In fact, the soil spectral contribution in red is primarily fiom exposed soil surfaces
reflecting direct solar radiation and diffused skylight (Huete, 1987a). Gao et 31. (2000)
also notes that visible bands, in general, provide very little discrimination and most
changes in visible bands are associated with soil background differences instead of
vegetation.

The use of NR or red spectral bands alone does not account for seasonal sun
‘angle differences (Ma et al., 2001). Vegetation indices have been developed to account
for spectral and temporal changes; the choice and suitability of which is generally
determined by the sensitivity to the characteristics of interest (Gao et al., 2000). The
optimal vegetation index should be invariant to the soil dependent component and yet

sensitive to spectral differences attributed to the vegetation component (Huete, 19873).

131

Two well-known indices featured in this study are discussed herein.

The Normalized Difference Vegetation Index (NDVI), first developed in 1979 by
Compton J.Tucker, a NASA researcher, is a measure of the green, leafy density of
vegetation (NASA, 2003). NDVI utilizes the NR and red wavebands in the following

equation where:
NDVI = (NIR — red )/(NIR + red) [3] 1
Ma et 31. (2001) regarded NDVI as one of the best indices at predicting yield and

midseason fertilizer amendments. Rondeaux et al. (1995) found NDVI well correlated to

vegetation amount until it saturates at full canopy coverage. It is useful for yielding

 

biophysical relationships applicable across varying canopy types; however, NDVI
sensitivity to soil optical properties affects these relationships and requires knowledge of
the soil reflectance for use in interpretation of measurements (Gao et al., 2000; Rondeaux
et 31., 1995). When Gao et 31. (2000) removed the soil background according to Eq.l;
NDVI exhibited very little sensitivity to vegetation, approaching saturation throughout
the entire range of canopy leaf area index (LAI). The inclusion of a soil background
restored an exponential dynamic range of NDVI, but in a manner dependent on the
background optical properties (Gao et al., 2000). Gao et al. (2000) firrther stated NDVI is
not only background sensitive, but most of its dynamic range occurs only with the
presence of a soil background: the brighter the background the greater the dynamic range.
They found little variation between the measurements of broadleaf crops and grasses.
The index is more sensitive to soil background than canopy type (Gao et al., 2000).
Huete (1987b) found that NDVI of the vegetation component (zero soil) achieved the

necessary invariance to differences in the solar sun angle, but once again, it was the soil

132

component contribution that limited its usefulness as a vegetation index and induced
strong anisotropic (directional) canopy behavior.

A number of soil adjusted vegetation indices have been developed: many are
variations of the Soil Adjusted Vegetation Index (SAVI) developed by Huete (1988)

where:
SAVI = (1+L)*(NIR —R)/(NIR +R+L) [4]

L = a soil adjustment factor that diminishes as the vegetation grows denser. According to
Rondeaux et 31. (1995), the term (1+L) is used to maintain the dynamic range of the index
between -1.0 and 1.0; however, the term was eliminated in their use of the equation.
SAVI is a Si grrificant improvement over earlier models. It is more reliable and less noisy
than NDVI. Rondeaux et 31. (1995) tested SAVI and several of its variations, the
Modified Soil Adjusted Vegetation Index (MSAVI), the Transformed Soil Adjusted
Vegetation Index (TSAVI), and the Two-axis Vegetation Index, by putting different soil
optical properties into a vegetation bidirectional reflectance model and examining the
sensitivity of the various indices to the soil. They found that SAVI has one of the lowest
standard deviations when vegetation coverage is low, remains quite constant over the mid
range of canopy coverage, and improves above 80% vegetation coverage. SAVI was less
definitive between 50% and 80% canopy coverage than other indices in the study.

Most of the present indices are related to the soil line. The optical properties of
the 26 soils used in the study by Rondeaux et al. (1995) were representative of five basic
types: fine sand, clay, peat, pozzolana, and pebbles. Even with the improvements of
SAVI, they found that using one universal soil line to account for 311 soil types rendered

an inadequate depiction of the vegetative canopy. Separating the 26 soils into mineral

133

 

and organic categories revealed that the slope of the organic soil line was twice that of the
mineral soils. Applying the appropriate soil line improved the outcome of the indices.
Rondeaux et al. (1995) tested several values for L in the SAVI index and found 0.16 or
0.20 best at minimizing the standard deviation over the firll canopy range, and proposed
that one of these values be adopted for agricultural applications. However, when used
where vegetative coverage was less than 50%, variances were somewhat higher than
when L was defined as 0.5. The choice of L in SAVI-type indices appears to be critical I
in minimizing the soil background effect (Rondeaux et al., 1995). '

Huete (1988) noted that L, as used in SAVI, should diminish as canopy density

 

increases. Therefore, when measurements are taken throughout the growing season the H‘-
definition of L should change as the canopy changes. Instead, L is typically assigned the
value of 0.5, which is a reasonable approximation when the amount of soil in the scene is
unknown (US Water Conservation Laboratory, 2003). A dynamic L would be more
attractive if quantitative determination of changing canopy coverage did not require
additional measurements such as Leaf Area Index (LAI). If canopy coverage could be
estimated using the model described below, and the estimate substituted for L as (1- fc) in
Eq. 4, it would eliminate additional measurements required for accurate coverage
assessment.

A fractional vegetation coverage model developed by Qi et al. (2000) was
intended to be used as an alternative processing technique to circumvent atmospheric
effects of satellite images in arriving at biophysical properties of land surfaces.
Atmospheric and bidirectional correction procedures are available, but often the ancillary

data about the concurrent atmospheric conditions are limited. A practical technique in

134

resolving atmospheric problems has been to subtract from all digital numbers (DN) the
minimum pixel values of a dark object found in the scene. Often, however, there are no
dark objects large enough to be identified. Again, an alternative is to use a pseudo
invariant object (PIO) within the scene. P10 is a surface such as a parking lot or bare soil
whose reflectance is known and remains constant over time. It is used to convert digital
numbers into reflectance values and in this manner circumvent atmospheric effects.
However, the reflectance properties do vary over time. Soil reflectance, for example,
varies with moisture content and surface roughness which changes due to rainfall events.
This invalidates the assumption of the invariant nature of such objects and results in
uncertain conversions to reflectance values from which products such as fractional
coverage are derived.

A physical property that does not vary with surface conditions is fractional green
vegetation cover (fc). An object void of vegetation (OW), such as soil, is located in the
image. By definition, an OVV has 0% vegetative coverage. Atmospheric corrections can
then be computed using the OW in terms of vegetation cover. In this way, the bare soil
as an OVV, is defined by numerically invariant properties, while the same bare soil, as 3
P10, is defined in terms of reflectance properties which are variant (Qi et al., 2000).

Each pixel normally contains a mixture of both soil and vegetation; the following “linear
mixing” model of the resulting remote sensor signal, S, describes this relationship

between the two physical characteristics where:
S=chv+(1_fc)Ss [5]
0 f C = fractional green cover,

0 l- fc = fractional soil cover,

135

o Sv = vegetation reflectance,
0 SS = soil reflectance, and,

o S = the remote sensing signal.

In accordance with Qi et al. (2000), the NDVI was substituted for S in Equation 5 and

algebraically rearranged to solve for fc where:
fc = (NDVIany —NDVIs0i1)/(ND VIngax —ND V130”) [a] i

The vegetation maximum (veg max) indicates the highest vegetation NDVI fi'om peak

vegetation coverage. NDVI of the soil should be constant throughout the season and

 

close to zero, but actually varies substantially with time and from location to location. g“
Therefore, in their study, soil NDVI was calculated from the reflectance of each image.
Qi et al. (2000) found that fc estimates agreed reasonably well with in situ measurements
and seasonal trends of f c agreed reasonably well with field observations.
In the 2001 and 2002 field study, fractional canopy coverage (Eq. 6) of the carrot
canopy was determined using NDVI derived from reflectance measurements taken

throughout the two seasons and substituted for L in calculating SAVI, where L = (1- f c).

Materials and Methods

Experimental Sites, Plot Design, Management Protocol and Agronomic Sampling
Field studies were conducted at four locations during 2001 and 2002, in

Montcalm County, Michigan. In both years plots were located at the Michigan State

University Montcahn Experiment Station on moderately well drained loamy sand to

sandy loam soil, of the Hillsdale-Spinks map unit (Hillsdale: coarse-loamy, mixed, mesic

136

Typic Hapludalfs, Spinks: sandy, mixed, mesic Psammentic Hapludalfs) (D.L. Mokma,
personal communication, 2003). In both years Diamond Cut and Goliath varieties were
planted on flat beds in early May and harvested in mid-September. Each year plots were
also established on commercial carrot fields, at Sandyland Farms, on Plainfield Sand,
including a loamy substratum at the 2001 site, (mixed mesic Typic Udipsamments) (D.L.
Mokrna, personal communication, 2003). Asgrow B1 and Prime Cut 59 varieties were
planted at the 2001 site, and Sugar Snax 54 was planted at the 2002 site. The fields were
planted in mid-April on raised beds and harvested in mid-August. Barley was planted
between rows to protect emerging plants and killed off once the carrots were established.
Four replications of each of four N-treatments, 45, 90, 135, 180 kg ha'1 were arranged in
a randomized complete block design at all locations. Weeds were controlled with
linuron, and foliar blight was controlled with Chlorothalonil. A detailed description is

given in the Materials and Methods section of Chapter 2.

Reflectance and Agronomic Measurements

Plant and soil reflectance measurements were made using a MSR87 multispectral
radiometer (CropScan, Rochester, MN) equipped with the standard eight narrowband
interference filters centered at 460, 510, 560, 610, 660, 710, 760, and 810 nm. Scanning
direction was with the row, to minimize shadows by plants and the operator. The field of
view was 28°, and measurements were viewed at nadir from a height of 2.55 m with a
ground resolution diameter of 1.27 m. Additional information describing the equipment
and the scanning protocol are given in the Materials and Methods section of Chapter 2.

A Canon Powershot G1 digital camera was mounted alongside and at the same height as

137

 

the radiometer at sampling. Digital images were taken of at least one scanned site per
plot for a visual record of radiometric measurements. Ground resolution of the digital
camera at a height of 2.55 m was 2.4 x 1.8 m. The images were used to determine

percent vegetation coverage and verify the validity of the f c calculation as used in carrot.

Image Processing

Digital images were cropped to match the area viewed by the radiometer with an
image processing application (PhotoImpact 7, Ulead, Taipei, Taiwan). Using the image
pixel count, a circle was generated from the image center outward, equal in size to the
diameter of the ground resolution of the radiometer. The supervised classification tools
of Erdas Imagine 8.5 were used to redefine the pixels of the cropped image into soil and
vegetation. Polygons, representative of the various elements of the image, were drawn
and assigned signature definitions. The classification process then tested the definitions
against each pixel in the image using maximum likelihood parameters to reclassify and
recolor the pixels under these new definitions. This process made it possible to quantify
the number of pixels attributed to soil and vegetation. The percent coverage was derived
from the pixel count defined as vegetation. It was necessary to develop two signature
files, one file covering the Montcalm Experiment Station in 2001 and 2002 with eight
signature definitions, the other file covering the Sandyland location in 2001 and 2002
with nine signature definitions. Each signature file was used in the classification process
of the appropriate group of images. The classified images were used to verify the

accuracy of the fc calculation.

138

The fc Calculation

Fractional cover (fc) was calculated according to Eq. 6. NDVI for each plot was
calculated from the averaged NR waveband centered at 810 nm, and red waveband
centered at 660 nm according to Eq. 3. In applying Eq. 6, NDVI so“ for the Montcahn
Experiment Station in 2001 was derived from one set of soil reflectance measurements
taken on May 18. Since the Sandyland location in 2001 was already established when
plots were staked and a large enough area of bare soil was no longer available, the
Experiment Station soil data was used in the model for Sandyland as an OVV, object
void of vegetation. Soils from the two locations were Similar in color, organic matter
content, and water holding capacity. NDVI so“ for both the Montcalm Experiment Station
and the Sandyland location in 2002 was derived from on-site bare soil measurements
taken throughout the season. NDVI veg m was derived from the seasonal peak canopy
reflectance measurements obtained from each location. Linear regression was used to
evaluate the relationship between percent vegetation coverage determined from the
classified images, and f c calculated from Eq. 6 according to multiple regression and

general linear models (SAS Inst. Inc., Release 8.2/2003).

Results and Discussion

All data were normally distributed as evidenced by the Shapiro-Wilk test and
residual plots. An outlier was defined as a viewing combination in which the camera and
radiometer viewed different amounts of vegatation coverage as a result of gaps due to
incomplete canopy coverage across the bed. Approximately 650 measurements were

tested, and 12 were removed as outliers.

139

 

In 2001, reflectance measurements at the experiment station began with plant
emergence and showed that at about 45 days following planting, the plant canopy was
large enough to produce a usable comparison between the percent foliar coverage derived
from the classified digital images and the calculated fc; (r2 = 0.54). Measurements taken
earlier resulted in coverage values so small that they were effectively zero. From days 51
and 54 to mid-July, fc (Eq. 6) correlated well with the amount of canopy coverage
derived from digital images, with r2 = 0.69 - 0.91 at both locations (Table 1). As the
canopy reached closure, correlation between the two parameters varied. At the
Experiment Station, correlation appeared to diminish at about 86 days of development
when canopy coverage ranged from 87 to 96%, according to the classified digital images.
In contrast, at Sandyland correlation with fc lasted until the carrot crop was 91 days old,
at which time canopy closure was at 99%. The successfirl Sandyland results also
indicated that the Experiment Station soil reflectance could be used as an object void of
vegetation (OVV) in the Sandyland data for the sole purpose of calculating fc.

During the 2002 season (Table 2) reflectance measurements were delayed until
later in the season and continued beyond peak canopy closure until harvest, since early
measurements in 2001 resulted in effectively zero coverage values. Most of the
measurements were taken during the last 45 days before harvest, over canopies with 90 to
99% closure, according to the classified digital images, and also revealed that once the
canopy reached peak closure, the correlation with fc diminished. This was especially
notable in the Goliath data. On July 17 and 24, f c correlated with percent vegetation
coverage with r2 = 0.80 and 0.82, but dropped sharply thereafter.

In 2002 at Sandyland, data collection was interupted by a number of sensor

140

 

Table 1. 2001 Regression analysis of Percent Vegetation Coverage (PVC) vs
Calculated fc (PVC = a + b f c +cTreatment) where a is the intercept and b and c are
regression coefficients. Treatment did not significantly influence correlation of PVC
with [C at p < 0.05.

 

 

 

 

 

c
Date DAPT y Intercept Coefficientf(b) r2 p-value
Montcalm Experiment Station
5/18/01 10 0 0 0 0
6/13/01 36 0 0 0 0
6/22/01 45 -0.0194 0.9657 0.54 .0012
6/28/01 51 -0.0157 1-1561 0.78 <-0001
7/5/01 58 -0.0897 12464 0.89 <0001
7/12/01 65 -0.0751 1.0856 0.73 <.0001
7/20/01 73 -0.1038 10933 0.91 <-0001
8/2/01 86 0.0945 03903 0.60 0004
8/9/01 93 -0.0565 1.0513 0.68 <-0001
8/17/01 101 -1.1406 2.1257 0.39 .0091
Sandyland (Deaner Rd)

6/13/01 54 0.0671 0.8485 0.75 <.0001
6/22/01 63 0.0427 07393 0.69 <-0001
6/28/01 69 -0.0643 19200 0.88 <.0001
7/5/01 76 -0.3450 13534 0.88 <-0001
7/12/01 83 -1.0679 2- 1098 0.78 <.0001
7/19/01 91 -1.2507 22358 0.78 <.0001
8/9/01 1 1 1 0 0 0 0
8/17/01 119 0.9041 00902 0.04 -4500

 

1Days after planting

equipment mishaps resulting in only three successful sampling dates. On August 9 and
15, the canopy was at full coverage and also exhibited the same late-season lack of
correlation between the percent canopy coverage derived from digital images and fc.

A full canopy, whether defined by fc or percent vegetation coverage derived fiom
classified digital images, is equal to 1.0, with the linear regression model resulting in zero
or at least very low correlation due to clustering of points. Tables 1 and 2 indicate a late
season drop in correlation at all four locations; however, the time at which the clustering

appeared varied between the locations. Population and varietal differences such as leaf

141

 

Table 2. 2002 Regression analysis of Percent Vegetation Coverage (PVC) vs. Calculated fc
(PVC = a + b f c + cTreatrnent) where a is the intercept and b and c are regression coefficients.
Treatment significantly influenced correlation of PVC with f c on the dates indicated at p < 0.05.

 

 

 

 

 

 

 

 

c Trt
Date DAP;r y Intercept Coefficient (Ir) Coefficient (c) r2 p-value
Montcalm Experiment Station Diamond Cut
5/21/02 14 0 0 0 0 0
7/17/02 71 0.2325 0.7450 -0.0006 0.66 .0008
7/24/02 78 0.3467 0.6432 0.45 .0040
8/1/02 86 0.8221 0.1634 0.0043‘ 0.39 .0400
8/9/02 94 0.6757 0.2852 0.27 .0400
8/15/02 100 0.7867 0.2119 -0.0001 0.70 .0014
8/21/02 106 0.3912 0.6151 -0.0001 0.54 .0066
8/30/02 1 15 0.4771 0.4281 .1 1 .2100
9/6/02 122 0.0732 0.9615 -0.0002 .74 .0003
Montcalm Experiment Station Goliath
5/21/02 14 0 0 0 0 0
7/17/02 71 -0.1192 1.1238 0.80 <.0001
7/24/02 78 0.0242 0.9565 0.82 <.0001
8/1/02 86 0.5801 0.4052 0.62 .0003
8/9/02 94 0.5372 0.4161 0.40 .0089
8/15/02 100 0.4414 0.5360 0.44 .0048
8/21/02 106 -0.3951 1.4147 -0.0003 .64 .0010
8/30/02 1 15 0.7422 0.1372 .005 .8000
9/6/02 122 0.9102 -0.0098 .0001 .9680
Sandyland (Masters Rd)
7/24/02 89 0.3606 05369 0.30 .0300
8/9/02 105 0.3016 (167“ 0.22 -0700
8/15/02 1 1 1 0.9214 0.0695 0.04 -4700
T Days after planting

°p-value 0.08

orientation, leaf size, canopy fullness, and developmental rate can contribute to timing of
full canopy. Differences in late season results may have been affected, in part, by the
manner in which the digital images were classified. Shadows that represented either
small pockets of soil or Shaded leaves nestled in the canopy were difficult to distinguish

in the images. The Shadows, which also represented a decrease in light spectra, affected

142

the resulting reflectance measurements. Incorrect interpretation of the shadows in the
digital images undoubtedly influenced the correlation of the images versus fc.

In 2002, the Goliath variety, at the Experiment Station, was affected by foliar
blight. The canopy coverage was reduced during the latter part of the season and it was
expected that the percent coverage derived fi'om classified digital images versus fc would
return to a more linear relationship; however, that was not the outcome (Table 2). In
addition to the Shadows, leaf discoloration also made interpretation of the digital images
in relation to the reflectance measurements difficult.

N treatments did not significantly influence the percent canopy coverage in 2001
(data not shown) according to multiple linear regression. However, on three occasions at
the experiment station, certain treatments differed significantly or nearly significantly at p
$.05, but did not necessarily vary sequentially (Table 3). On three occasions at
Sandyland, treatment differences were notable but not significant and percent vegetation
coverage was not significantly influenced by treatment.

During 2002, multiple linear regression, comparing percent vegetation coverage
to fc and treatments, showed that treatment significantly influenced the percent canopy
coverage derived from digital images in the Diamond Cut variety on four occasions July
17, August 15, August 21, and September 6 (Table 2). Table 3 indicates significant
differences at p $0.05, but treatments significantly influenced percent vegetation
coverage on less than half of the sampling dates shown; not necessarily sequentially. In
the Goliath variety Table 2 shows that treatment was significant only on August 21, but
only on September 6 was there significant separation of treatments according to the

general linear model.

143

 

Lack of signficant correlation between canopy coverage derived fiom the images
and the fc calculation versus treatments was not unexpected. The typical density of
carrot planting and thickness of the maturing canopy may dilute the distinction between
treatments where biomass differences may be nonexistent. Overall, fc determined fiom
NR and red reflectance measurements correlated reasonably well with percent vegetative

coverage derived from digital images throughout most of the season. Earliest correlation

Table 3. Mean Percent Vegetation Coverage as influence by f c and treatment differences
on selected dates.

 

 

 

 

 

 

 

 

 

 

Montcalm Experiment Station 2001 Sandyland (Deaner Rd) 2001
trt July 5 July 12 Aug 9 trt July 5 July 12 July 20
kg ha'1 --Percent Canopy Coverage---- kg ha'1 «Percent Canopy Coverage-«-
45 213 393 983 45 74b 88b 92b
90 183 33a 97a 90 76ab 903 953
135 193 333 963 135 823 90ab 953
180 19a 33a 983 180 70b 88ab 933
pT .004 .014 .115 pT .375 .741 .430
Montcalm Experiment Station (Diamond Cut) 2002
trt July 17° Aug 1 Aug 9 Aug 15* Aug 21* Sept 6°
kg ha'1 Percent Canopy Coverage
45 723 973 953 99a 97a 93a
90 713 96ab 93a 99a 97a 943
135 67a 95b 94a 983b 973 933
1 80 643 96ab 953 98b 96a 94a
JT .146 .026 .039 .009 .151 .047
Montcalm Experiment Station (Goliath) 2002
trt July 17 July 24 Aug 9 Aug 21° Sept 6
kg ha'1 Percent Canopy Coverage
45 643b 803b 93a 973 923
90 723 853 923 923 88b
135 57b 75b 94a 94a 9lab
180 693 Slab 94a 93a 893b
pT .274 .286 .399 .110 .017

 

Mean values with the same letter are not significantly different at p < 0.05.
t p = p-value of overall treatment response according to the analysis of variance.

“ Treatment response significantly influenced correlation of percent canopy coverage with f c
according to regression analysis on these dates. (See Table 2).

144

was possible at about 45 days at the experiment station in 2001. Other varieties may vary
according to grth patterns and climate. Saturation of f c occurred at peak canopy
coverage when fc = 1.0. L, the soil adjustment factor of SAVI, is defined as zero at full
coverage and therefore (1- fc) satisfies the soil adj ustrnent factor at saturation.

SAVI was derived for all reflectance measurements using fc as the soil
adjustment factor L = (1- fc) to determine whether it improved the accuracy of the
vegetation index. For comparison, L was also held constant at 0.5 as is typically done
when the amount of coverage is unknown (US Water Conservation Laboratory, 2003).
Rondeaux et al. (1995) and Gao et 31. (2000) also used L = 0.5 as a basis for their
comparative model testing. When L was held constant at 0.5, treatment differences
remained separated and resembled the growth curve, however, the index exceeded its
expected range when canopy coverage was dense. When L = (1-fc) was substituted,
SAVI was held to its dynamic range of -1 .0 to 1.0, the curves then plateaued at peak
canopy coverage and treatment differences were no longer distinguishable. Figures 1
through 3 depict the comparison between L = 0.5 and L = (1- fc).

SAVI, where L = (1- fc) [SAVIfc], was also derived for all reflectance
measurements without the multiplier (1+L) as Rondeaux et al. (1995) had done. The
resulting curves plotted over time were within the expected range for SAVI and
resembled the grth curve. Without the multiplier, SAVIfc did not plateau at peak
canopy, and the treatment effect remained separated. Figure 4 is an example taken fi'om
the Diamond Cut variety, Experiment Station, 2002 with and without the multiplier (1+
(1-fc)). The data from the other locations exhibited similar differences. Without the

multiplier (1+L), SAVIfc did not measurably change the outcome of SAVI in the carrot

145

 

1.60 .
+ 45 kg/ha SAVIfc

SAVI Montcalm Exp. Stn. 2001

 

1.40 ._ + 90 kg/ha SAVIfc

135 kg/ha SAVIfc

 

 

1.20

.4»—

—-x— 180 kg/ha SAVIfc

SAVI L = 0.5

 

1,00 ,_ ——)t(—— 45 kg/ha SAVI L=0.5
+ 90 kg/ha SAVI L=0.5
_ —4— 135 kg/ha SAVI L=0.5
0.60 1* —-— 180 kg/ha SAVI L=0.5

0.80 -

W—

SAVIfE’

 

Index Value

0.40 4

0.20 i

 

g ..

 

 

 

 

0

0.00 i? r

T 7 T j

r

j

714 2128 35 42 49 56 63 70 77 84 9198105112119126
Days After Planting

 

 

1.60 -
1.40 l
1.20 4»
1.00 4
0.80 ->

0.60 ->

Index Value

0.40 -

 

0.20 l

0.00 i

 

SAVI Sandyland 2001

+ 45 kg/ha SAVIfc
+ 90 kg/ha SAVIfc
135 kg/ha SAVIfc
—x— 180 kg/ha SAVIfc
—-11t— 45 kg/ha SAVI L=0.5
+ 90 kg/ha SAVI L=0.5
—+— 135 kg/ha SAVI L=0.5
—-—- 180 kg/ha SAVI L=0.5

 

 

SAVI L = 0.5

 

 

 

 

I R

T I 7 7 m 7 T r

0 7 14 21 28 35 42 49 56 63 70 77 84 91 98105112119126

Days After Planting

 

Fig. 1 Results of SAVIfc and SAVI L = 0.5 for the 2001 field season at the Montcalm

 

Experiment Station and Sandyland locations.
Images in this thesis are presented in color.

146

 

 

Index Value

1.60

1 .40

1.20

SAVI Montcalm Exp Stn Diamond Cut 2002

+ 45 kglha SAVIfc
+ 90 kglha SAVIfc

135 kg/ha SAVIfc
——x~— 180 kg/ha SAVIfc

1.00 1— + 45 kglha SAVI L=0.5

0.80 «—

+ 90 kg/ha SAVI L=0.5
-—+— 135 kglha SAVI L=0.5

0-60 *— —-— 180 kg/ha SAVI L=0.5

0.40 4

0.20 J

 

 

 

 

SAVI L = 0.5

 

 

 

SAVIfc

 

 

0.00

fif

0 7 14 21 28 35 42 49 56 63 70 77 84 91 98105112119126

7 T T T T *1

Days After Planting

 

 

 

Index Value

1.60 —

1.40 «— + 90 kg/ha SAVIfc
1.20 +—
1.00 L + 45 kg/ha SAVI L=0.5
0.80 -—

0.60 T— —-— 180 kg/ha SAVI L=0.5

0.40 —

0.20 «

 

0.00

SAVI Montcalm Exp Stn Goliath 2002

+ 45 kg/ha SAVIfc

 

135 kg/ha SAVIfc

 

 

 

——x—— 180 kg/ha SAVIfc

 

SAVI L = 0.5

 

+ 90 kglha SAVI L=0.5

   

 

 

—+— 135 kglha SAVIL=0.5

 

 

T f

ji T f

0 7 14 2128 35 42 49 56 63 70 77 84 9198105112119126

Days After Planting

 

Fig. 2 Results of SAVIfc and SAVI L = 0.5 for the Diamond Cut and Goliath varieties at the
Montcalm Experiment Station for the 2002 field season.
Images in this thesis are presented in color.

147

 

 

 

SAVI Sandyland 2002
1.60 -
—0— 45 kg/ha SAVIfc
” + 90 kg/ha SAVIfc
1.20 «F 135 kg/ha SAVIfc
—)e—— 180 kg/ha SAVIfc
+ 45 kg/ha SAVI L=0.5 Wax—”+58
0.80 ~~ —o— 90 kg/ha SAVI L=0.5
__ —+-— 135 kglha SAVI L=0.5 SAV'fC

—-—— 180 kg/ha SAVI L=0.5
0.404~ *9 7- 9 -- - -, ,, 4-.-- d L-..

1.40 «

 

 

SAVI L = 0.5

 

l

1.00 1

 

0.60 «

 

Index Value

0.204 , 7, 7 ,2 . .4 .7 L "L , , ..L,

 

0.00 T r T . T r . r T . . . . . . . T .
0 7 14 2128 35 42 49 56 63 70 77 84 9198105112119126

 

Days After Planting

 

 

 

 

Fig. 3 Results of SAVIfc and SAVI L = 0.5 at Sandyland for the 2002 field season.
Images in this thesis are presented in color.

crop. In fact, regression analysis (Table 4) revealed a significant relationship; (r2 = 0.99
to 1.0) for every date throughout the growing season at all four locations.

Differences between SAVI and SAVIfc were expected to occur during early and
late developmental stages of the carrot crop when the canopy coverage, and therefore fc,
differed from the previously defined L = 0.5. Figures 1 through 3 show that SAVIfc
preserved the dynamic range of SAVI even in dense canopy coverage, while L held
constant at 0.5 exceeded the expected range by as much as 30%. In addition, the curves
crossed each other at the point where L = 0.5 under both definitions of L, as expected,
approximately 63 to 65 days after planting (Table 3) at about 50% canopy coverage

(Figures 1 through 3). Where L was held constant, SAVI, while a reasonable estimation

148

(US Water Conservation Laboratory), was understated in low canopy coverage and

overstated in dense canopy conditions compared to SAVI f c.

 

 

 

 

 

 

 

 

SAVIfc With and Without the (1 +L) Multiplier
Montcalm Exp Stn 2002
1.00 «—- ' 45 mm
-—-— 90kg/ha
0.80 4 135 We ,
g —«x—- 180 We '
g 0.60 ._ —-)l(— 45 kg/haw/o
x + 90 kg/ha w/o I
0
g 0.40 4_ —+— 135 kglha w/o
- —-— 180 kglha w/o
0.20 « w , -L if f- i
0000 j I I T 7 T fl T l T j I T I I I F 1
0 7 14 21 28 35 42 49 56 63 70 77 84 91 98105112119126
Days After Planting

 

 

 

Fig. 4 Example of the difference between SAVIfc with and without the (1+L) multiplier. The
Experiment Station 2002, Diamond Cut data is shown here, the other locations exhibited
similar differences.

Images in this thesis are presented in color.

Conclusion

Images were used to assess the reliability of f 0 (Eq. 6) in estimating canopy
coverage. As the canopy neared closure f c tended to saturate. Late season images
presented challenges to the interpretation of the shadows created by the sun angle
reflecting off soil, shaded leaves, and leaf discoloration. Despite this, fc could be used
to predict percent vegetation coverage. When f c was used as the soil adjustment factor L

= (l - f c) to calculate SAVI, it was determined that for the carrot studies in 2001 and 2002,

149

L = (1- fc) held SAVI to its dynamic range of -1 .0 to 1.0 even when the canopy was
dense. However, differences between treatments were best viewed when SAVI f c was

determined without the multiplier (1+L) in Equation 6.

Table 4. Results of regression analysis of SAVI comparing L = 0.5 and L = (l-fc).

 

 

 

 

 

 

 

Average Average
Date Days+ l-fc+ r2 Date Days)f 1-fc” rz
Montcalm Exp Stn Sandyland (Deaner Rd )
5/18/01 10 1.00 1.00'"
6/13/01 36 1.00 1.00‘” 6/13/01 54 0.84 1.00‘”
6/22/01 45 0.91 1.00‘" 6/22/01 63 0.51 1.00‘“
6/28/01 51 0.89 1.00‘” 6/28/01 69 0.31 100‘”
7/5/01 58 0.77 100’“ 7/5/01 76 0.19 1.00‘“
7/12/01 65 0.61 1.00‘” 7/12/01 83 0.07 0.99‘“
7/20/01 73 0.32 1.00‘” 7/20/01 91 0.02 0.99‘“
7/26/01 79 0.15 1.00‘" 7/26/01 97 0.02 099'”
8/2/01 86 0.05 099‘“ mm 104 0.01 099‘”
8/9/01 93 0.03 0.99‘“ 8/09/01 1 1 1 0.02 0.99‘“
8/17/01 101 0.01 0.99’“ 8/17/01 1 19 0.02 099‘”
9/6/01 121 0.05 0.99‘”
Montcalm Exp Stn, Diamond Cut Montcalm Exp Stn, Goliath
5/21/02 14 1.00 1.00'" 5/21/02 14 1.00 1.00'"
7/1 1/02 65 0.49 1.00‘“ 7/11/02 65 0.54 1.00‘”
7/17/02 71 0.31 1.00‘“ 7/17/02 71 0.31 1.00’”
7/24/02 78 0.20 0.99’” 7/24/02 78 0.18 100‘”
8/1/02 86 0.09 099‘” 8/1/02 86 0.07 099‘”
8/9/02 94 0.06 099‘” 8/9/02 94 0.05 0.99‘“
8/15/02 100 0.02 0.99‘“ 8/15/02 100 0.02 0.99‘”
8/21/02 106 0.04 0.99‘” 8/21/02 106 0.03 0.99'"
8/30/02 1 15 0.07 0.99‘“ 8/30/02 1 15 0.06 099’”
9/6/02 122 0.09 099'“ 9/6/02 122 0.09 0.99‘”
Sandyland (Masters Rd)
7/1 1/02 71 0.09 0.99'"
7/17/02 82 0.05 0.99‘”
7/24/02 89 0.04 099‘"
8/1/02 97 0.03 099‘“
8/9/02 105 0.03 0.99‘"
8/15/02 1 1 1 0.02 0.99’“

 

T Days means number of days since planting.

1All treatments were combined to show general coverage at the specified days after planting.
mp—value <.0001

The choice of L in SAVI-type indices, while critical in minimizing the soil

background effect (Rondeaux et al., 1995), should also be simple to apply, especially if

150

these indices will become integral to production agriculture. fc is easy to apply as the
definition of the soil background adjustment factor because it is obtained from reflectance

measurements, which would already be available.

151

References

Gao, X., A.L. Huete, W. Ni, and T. Miura. 2000. Optical- biophysical relationships of

vegetation spectra without background contamination. Remote Sens. Environ.
74:609-620.

Huete, A. R. 1987a. Soil-dependent spectral response in a developing plant canopy.
Agron. J. 79:61-68.

Huete, A. R. 1987b. Soil and sun angle interactions on partial canopy spectra. Int. J.
Remote Sensing. 8(9): 1307-13 1 7.

Huete, AR. 1988. A soil-adjusted vegetation index (SAVI). Remote Sens. Environ.
25:295-309.

Huete, A.R., G. Hua, J. Qi, A. Chehbouni, and W.J.D. van Leeuwen.l992. Normalization
of multispectral red and NIR reflectances with the SAVI. Remote Sens. Environ.
41:143-154.

Li, H., R.J. Lascano, E.M. Barnes, J. Booker, L.T. Wilson, K.F. Bronson, and E. Segarra.
2001. Multispectral reflectance of cotton related to plant growth, soil water, and
texture, and site elevation. Agron. J. 93: 1327-1337.

Ma, B.L., L.M. Dwyer, C. Costa, E.R. Cober, and M.J. Morrison. 2001. Early prediction
of soybean yield from canopy reflectance measurements. Agron. J. 93:1227-1234.

NASA. 2003. Earth observatory: How plants can change our climate [Online]. Available
at http://earthobservatory.nasa.gov/Studv/LAWAH.html (verified 17 Sept. 2003)

Ott, R.L., and M. Longnecker. 2001. An introduction to statistical methods and data
analysis. 5th. ed. Duxbury, Pacific Grove, CA.

Qi, J ., R.C. Marsett, M.S. Moran, D.C. Goodrich, P. Heilman, Y.H. Kerr, G.Dedieu, A.
Chehbouni, and X.X. Zhang. 2000. Spatial and temporal dynamics of vegetation
in the San Pedro river basin area. Agric. For. Meteorol. 105:55-68.

Rondeaux, G., M. Steven, and F. Barret. 1995. Optimization of soil-adjusted vegetation
indices. Remote Sens. Environ. 55:95-107.

Soil Survey Staff. 1975. Soil taxonomy — a basic system of soil classification for making
and interpreting soil survey. Handb. No.436, SCS-USDA. US. Government
Printing Office, Washington, DC.

Stoner, E.R., and M.F. Baumgardner. 1981. Characteristic variations in reflectance of
surface soils. Soil Sci. Soc. Am. J. 45:1161-1165.

152

Tucker, C .J . 1979. Red and photographic infrared linear combinations for monitoring
vegetation. Remote Sens. Environ. 8: 127-150.

US Water Conservation Laboratory. 2003. How a vegetation index works [Online].

Available at http://www.uswcl.ars.ag.gov/epd/remsen/ViNIworkshtm (verified 7
Oct. 2003)

153

     
  

S

11111111111111111111111111111