Download Estimation of Crop Water Use for Different Cropping Systems in the

Document related concepts
no text concepts found
Transcript
Estimation of Crop Water Use for Different Cropping Systems in the Texas High Plains
Using Remote Sensing
by
Nithya Rajan, B. Sc., M. Sc.
A Dissertation
In
Agronomy
Submitted to the Graduate Faculty
of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of
DOCTOR OF PHILOSOPHY
Approved
Stephan Maas
Chair
Vivien Allen
Seiichi Nagihara
Steve Mauget
Fred Hartmeister
Dean of the Graduate School
December, 2007
Copyright 2007, Nithya Rajan, Texas Tech University
ACKNOWLEDGEMENTS
The author would like to express her sincere thanks and deep gratitude to her
major advisor, Dr. Stephen Maas, for his encouragement, direction, assistance, and
patience during the course of her studies. The author would also like to thank her doctoral
committee members, Dr. Vivien Allen, Dr. Seiichi Nagihara, and Dr. Steve Mauget for
their encouragement and support.
The author would like to acknowledge the timely help provided by the Texas
Alliance for Water Conservation Demonstration Project (TAWC) Director, Mr. Rick
Kellison, through out the period of research. The author would also like to thank all the
members of TAWC for their assistance. The author acknowledges the funding provided
by TAWC for carrying out her dissertation research.
The author would like to thank Dr. Wenxuan Guo and Mr. Jerry Brightbill (South
Plains Precision Ag., Inc, Plainview, TX) for their assistance in collecting aerial imagery.
The author thanks her colleagues, especially Mr. Shyam. S. Nair and Ms. Jessica
Torrion, for their support and assistance.
ii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS................................................................................................ii
ABSTRACT.......................................................................................................................iv
LIST OF TABLES.............................................................................................................v
LIST OF FIGURES ..........................................................................................................vi
LIST OF ABBREVIATION............................................................................................xiii
INTRODUCTION.............................................................................................................1
LITERATUTE TEVIEW...................................................................................................6
MATERIALS AND METHODS.....................................................................................17
RESULTS AND DISCUSSION......................................................................................42
CONCLUSION..............................................................................................................147
LITERATURE CITED...................................................................................................150
iii
ABSTRACT
The spectral crop coefficient (Ksc) is a novel approach for estimating the water use
of field crops. In this study, Ksc is evaluated from remote sensing observations (satellite
or aircraft imagery) of the field in question and, thus, is specific to the crop growth
characteristics in the field. This approach assumes that the crop is acclimated to its
environment and determines crop water use (CWU) based on the product of potential
evapotranspiration and remotely sensed crop ground cover (GC). Because the remotely
sensed measurements of GC are infrequent over the growing season, these measurements
are used in a crop model to simulate values of GC for each day of the growing season,
resulting in a crop coefficient curve (known as the spectral crop coefficient – Ksc) that is
specific to the field, crop, and growing conditions. The method used for estimating the
GC from remote sensing data involves the Perpendicular Vegetation Index (PVI). GC is
calculated by dividing the average PVI for a field by the value of PVI for full canopy
point. Statistical analysis of estimated and field-measured GC from a large number of
fields indicates that the procedure for estimating crop GC from remote sensing imagery is
accurate so that, on average, estimates of GC determined using this procedure should be
within 6 percent of their true values. The seasonal CWU estimated by this method
showed differences in water utilization by individual fields. Comparison of these
seasonal CWU values among the fields in the study was effective in showing differences
related to year, crop, and irrigation type. Comparing daily values of CWU estimated
using the Ksc method and the regular crop coefficient method recommended for crops in
the Texas High Plains with actual measurements of evapotranspiration made using the
eddy covariance method showed that the Ksc method was consistently more accurate than
the regular crop coefficient method.
iv
LIST OF TABLES
Number
3.1.1
3.5.1
3.5.2
4.1.1
4.1.2
4.2.1
4.2.2
4.3.1
4.3.2
Title
Page No
Field number, section number, crop and irrigation type for fields in
the study in 2006 and 2007.
19
Acquisition dates of images containing the study area from
Landsat-5 Thematic Mapper (TM) and Landsat-7 Enhanced
Thematic Mapper (ETM+) satellite sensors in 2006 and 2007.
27
Image acquisition dates for fields in the study using the Texas Tech
Airborne Multispectral Remote Sensing System (TTAMRSS).
30
Slope (a1) and intercept (a0) of the bare soil line, DC values in the
red (DCFC,RED) and NIR (DCFC,NIR) spectral bands for full canopy,
and the PVI value (PVIFC) associated with full canopy for the four
Landsat-5 image acquisitions.
45
Slope (a1) and intercept (a0) of the bare soil line, DC values in the
red (DCFC,RED) and NIR (DCFC,NIR) spectral bands for full canopy,
and the PVI value (PVIFC) associated with full canopy for all the
Landsat-5 and Landsat-7 image acquisitions.
47
Percent Ground Cover (GC) estimated for the fields in the study by
the Perpendicular Vegetation Index (PVI) method using
TTAMRSS, Landsat-5 and Landsat-7 images in 2007.
56
Percent Ground Cover (GC) estimated for the fields in the study by
the Perpendicular Vegetation Index (PVI) method using Landsat-5
and Landsat-7 images for all the fields in the study in 2006.
57
Seasonal PET and ET0 values for all the fields in the study in
2006....................................................................................................
80
Seasonal PET and ET0 values for all the fields in the study in
2007..................................................................................................... 81
v
LIST OF FIGURES
Number
Title
Page No
3.1.1
Locations of study fields in Hale and Floyd counties of Texas.
18
3.4.1
Mobile eddy covariance system located in Field No. 2.
25
3.4.2
LI-7500 calibration unit built on a hand dolly for user calibration of
LI-7500 Infrared Gas Analyzer.
25
3.5.1
Various parts of Texas Tech Air-borne Multispectral Remote
Sensing System (TTAMRSS) and the Cessna Model 172 plane used
to acquire aerial images of fields in the study.
30
Shooting over-head pictures in a corn field for making groundbased observations of ground cover.
32
Pixel digital count (DC) values in the near-infrared (NIR) spectral
band plotted vs. corresponding DC values in the red spectral band
for a portion of a Landsat-5................................................................
37
Example of mosaic of aerial images that is used to determine the
soil line and 100% ground cover (GC) point for evaluating the GC
from aerial images acquired on 8 June 2007......................................
39
Results of plotting the pixel digital count (DC) values in the nearinfrared (NIR) spectral band vs. the corresponding DC values in the
red spectral band ................................................................................
44
Values of ground cover (GC) estimated for the 31 locations in the
study using the satellite image data plotted vs. the corresponding
ground-based field observations of GC...........................................
46
Results of plotting the pixel digital count (DC) values in the nearinfrared (NIR) spectral band vs. the corresponding DC values in the
red spectral band for the (A) 30 June..........................................
50
Results of plotting the pixel digital count (DC) values in the nearinfrared (NIR) spectral band vs. the corresponding DC values in the
red spectral band for a mosaic of six aerial images.....................
52
3.6.1
3.7.1
3.7.2
4.1.1
4.1.2
4.1.3
4.1.4
4.1.5
Ground cover (GC) maps produced by the Perpendicular Vegetation
Index (PVI) method using the aerial images acquired on 8 June
2007 using.........................................................................
53
vi
4.2.1
Spectral crop coefficient curves (Ksc) for (A) corn for silage (Field
Nos. 20-1 and 27) and (B) corn for grain (Field No. 24 and 26-2)
fields for the 2007 growing season.....................................................
58
4.2.2
Spectral crop coefficient curves (Ksc) for (A) corn for silage (Field
No. 20-2) and (B) corn for grain (Field No. 24-1 and 26-1) fields for
the 2006 growing season...............................................................
59
4.2.3
Spectral crop coefficient curves (Ksc) for drip irrigated cotton (Field
Nos. 1-1, 1-2, and 2) fields for the (A) 2007 and (B) 2006 growing
seasons.................................................................................................
60
Spectral crop coefficient curves (Ksc) for center-pivot irrigated
cotton (Field Nos. 3-1 and 6) fields for the (A) 2007 and (B) 2006
growing seasons..................................................................................
61
Spectral crop coefficient curves (Ksc) for furrow irrigated cotton
fields for the (A) 2007 (Field Nos. 15-1 and 15-4) and (B) 2006
(Field Nos. 15-1 and 15-3) growing seasons......................................
62
Spectral crop coefficient curves (Ksc) for dryland cotton fields for
the (A) 2007 (Field No. 12-1) and (B) 2006 (Field No. 13-1)
growing seasons..................................................................................
63
Spectral crop coefficient curves (Ksc) for forage sorghum (Field No.
20-2) for the 2007 growing season......................................................
64
4.2.4
4.2.5
4.2.6
4.2.7
4.2.8
Spectral crop coefficient curves (Ksc) for grain sorghum fields for
(A) 2007 (Field Nos. 15-3 and 18-2) and (B) 2006 (Field No 154).......................................................................................................... 65
4.2.9
Spectral crop coefficient curves (Ksc) for forage sorghum fields (A)
Field No. 20-1 and (B) Field No 4-2 for the 2006 growing
season................................................................................................... 66
4.2.10
Spectral crop coefficient curves (Ksc) for pearl millet for the (A)
2007 (Field No. 26-1) and (B) 2006 (Field No. 19-3) growing
seasons.................................................................................................
67
4.2.11
Regular crop coefficient (Kc) curve for corn developed for the Texas
High Plains from lysimeter studies at Bushland, TX.
69
4.2.12
Regular crop coefficient (Kc) curve for cotton developed for the
Texas High Plains from lysimeter studies at Bushland, TX.
vii
69
4.2.13
Regular crop coefficient (Kc) curve for grain sorghum developed for
the Texas High Plains from lysimeter studies at Bushland, TX.
70
4.2.14
Comparison of the spectral crop coefficient curve (Ksc) generated
using remotely sensed ground cover (GC) and the regular crop
coefficient curve (Kc)........................................................................... 71
4.2.15
Comparison of the spectral crop coefficient curve (Ksc) generated
using remotely sensed ground cover (GC) and the regular crop
coefficient curve (Kc)........................................................................... 72
4.2.16
Comparison of the spectral crop coefficient curve (Ksc) generated
using remotely sensed ground cover (GC) and the regular crop
coefficient curve (Kc)........................................................................... 73
4.3.1
Comparison of potential evapotranspiration (PET) and reference
evapotranspiration (ET0) calculated using the FAO-56 guidelines
for a center-pivot irrigated corn field (Field No. 24) in 2007.............. 75
4.3.2
Comparison of potential evapotranspiration (PET) and reference
evapotranspiration (ET0) calculated using the FAO-56 guidelines
for a drip irrigated cotton field (Field No. 2) in 2007.......................... 76
4.3.3
Comparison of potential evapotranspiration (PET) and reference
evapotranspiration (ET0) calculated using the FAO-56 guidelines
for a furrow irrigated grain sorghum field........................................... 77
4.3.4
Comparison of potential evapotranspiration (PET) and reference
evapotranspiration (ET0) calculated using the FAO-56 guidelines
for a center-pivot irrigated forage sorghum field (Field No. 20-2) in
2007.....................................................................................................
78
Comparison of potential evapotranspiration (PET) and reference
evapotranspiration (ET0) calculated using the FAO-56 guidelines of
a center-pivot irrigated pearl millet field.............................................
79
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 20-1..................................................................
83
4.3.5
4.4.1
4.4.2
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 24...................................................................... 84
viii
4.4.3
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 26-2..................................................................
85
4.4.4
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 27...................................................................... 86
4.4.5
Seasonal Crop Water Use (CWU) in mm estimated by the spectral
crop coefficient (Ksc ) and regular crop coefficient (Kc ) methods for
corn fields in 2007...............................................................................
87
4.4.6
Daily Crop Water Use (CWU) estimated by the spectral crop
coefficient (Ksc), regular crop coefficient (Kc), and eddy covariance
(EC) methods for Field No. 20-1 in 2007............................................ 90
4.4.7
Daily Crop Water Use (CWU) estimated by the spectral crop
coefficient (Ksc), regular crop coefficient (Kc), and eddy covariance
(EC) methods for Field No. 24 in 2007...............................................
91
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 1-1....................................................................
93
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 1-2....................................................................
94
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 2.......................................................................
95
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 6.......................................................................
96
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 11-1..................................................................
97
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 12-1..................................................................
98
4.4.8
4.4.9
4.4.10
4.4.11
4.4.12
4.4.13
ix
4.4.14
4.4.15
4.4.16
4.4.17
4.4.18
4.4.19
4.4.20
4.4.21
4.4.22
4.4.23
4.4.24
Seasonal Crop Water Use (CWU) in mm estimated by the spectral
crop coefficient (Ksc ) and regular crop coefficient (Kc ) methods for
cotton fields in 2007............................................................................
100
Daily Crop Water Use (CWU) estimated by the spectral crop
coefficient (Ksc), regular crop coefficient (Kc), and eddy covariance
(EC) methods for Field No. 12-1 (dryland cotton) in 2007................
102
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 20-2..................................................................
104
Daily Crop Water Use (CWU) estimated by the spectral crop
coefficient (Ksc) and eddy covariance (EC) methods for Field No.
20-2 (center-pivot irrigated forage sorghum) in 2007.......................
105
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method are plotted versus the day
of the year for Field No. 26-1.............................................................
107
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 12-1..................................................................
108
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 15-3..................................................................
109
Daily estimates of Crop Water Use (CWU) in 2007 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 18-2..................................................................
110
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 20-2.................................................................
112
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 22-2.................................................................
113
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 24-1..................................................................
114
x
4.4.25
4.4.26
4.4.27
4.4.28
4.4.29
4.4.30
4.4.31
4.4.32
4.4.33
4.4.34
4.4.35
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 26-2..................................................................
115
Seasonal Crop Water Use (CWU) in mm estimated by the spectral
crop coefficient (Ksc ) and regular crop coefficient (Kc ) methods for
corn fields............................................................................................
117
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 1-1....................................................................
119
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 1-2....................................................................
120
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 2.......................................................................
121
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 3-1....................................................................
122
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 15-1..................................................................
123
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 13-1.................................................................
124
Seasonal Crop Water Use (CWU) in mm estimated by the spectral
crop coefficient (Ksc ) and regular crop coefficient (Kc ) methods for
cotton fields in 2006............................................................................
126
Daily Crop Water Use (CWU) estimated by the spectral crop
coefficient (Ksc), regular crop coefficient (Kc), and eddy covariance
(EC) methods for Field No. 13-1 (dryland cotton) in 2006................
129
Daily Crop Water Use (CWU) estimated by the spectral crop
coefficient (Ksc), regular crop coefficient (Kc), and eddy covariance
(EC) methods for Field No. 2 (drip irrigated cotton) in 2006.............
130
xi
4.4.36
4.4.37
4.4.38
4.4.39
4.5.1
4.5.2
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 19-3..................................................................
132
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 20-1..................................................................
133
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 4-2....................................................................
134
Daily estimates of Crop Water Use (CWU) in 2006 determined by
the spectral crop coefficient (Ksc ) method plotted versus the day of
the year for Field No. 15-4..................................................................
135
Comparison of Crop Water Use (CWU) determined by the spectral
crop coefficient (Ksc) method for fields that were planted to cotton
in both 2006 and 2007.........................................................................
137
Monthly average rainfall data for 2006 and 2007 recorded at the
mesonet weather station in Plainview, TX.........................................
138
4.5.3
4.5.4
4.5.5
4.6.1
4.6.2
Comparison of seasonal Crop Water Use (CWU) determined by the
spectral crop coefficient (Ksc) and regular crop coefficient (Kc)
methods averaged for all cotton fields in the study.............................
140
Comparison of seasonal Crop Water Use (CWU) determined by the
spectral crop coefficient (Ksc) and regular crop coefficient (Kc)
methods averaged...............................................................................
141
Comparison of seasonal Crop Water Use (CWU) determined by the
spectral crop coefficient (Ksc) method averaged for all fields in the
study by crop in 2006 and 2007..........................................................
143
Daily Crop Water Use (CWU) estimated by the spectral crop
coefficient (Ksc) method plotted versus corresponding values of
daily CWU measured using eddy covariance....................................
145
Calculated values of the stress factor Fs plotted versus
corresponding values of measured daily CWU using eddy
covariance. Horizontal solid line represents Fs = 1............................
146
xii
LIST OF ABBREVIATIONS
2D
AAE
CWU
DC
DCFC,NIR
DCFC,RED
EC
ENVI
ETM+
ET0
FC
Fs
GC
GC100
GPS
Kc
Ksc
LAI
NDVI
NIR
PET
RED
PVI
SAVI
TM
TAWC
TTAMRSS
USDA
WUE
Two Dimensional
Average Absolute Error
Crop Water Use
Digital Counts
Digital Counts of full canopy in the RED wavelength
Digital Counts of full canopy in the NIR wavelength
Eddy Covariance
Environment for Visualizing Images
Enhanced Thematic Mapper Plus
Reference Evapotranspiration
Full Canopy
Stress Factor
Ground cover
100 percent Ground Cover
Global Positioning System
Regular crop coefficient
Spectral crop coefficient
Leaf Area Index
Normalized Difference Vegetation Index
Near Infrared
Potential Evapotranspiration
Red spectral band
Perpendicular Vegetation Index
Soil Adjusted Vegetation Index
Thematic Mapper
Texas Alliance for Water Conservation
Texas Tech Airborne Multispectral Remote Sensing
System
United States Department of Agriculture
Water Use Efficiency
xiii
Chapter I
Introduction
Depleting water resources and diminishing crop production are the central topics of
many ongoing research projects in the Southern High Plains. Due to a lack of sufficient
rainfall in 2006, dryland crop production in the Southern High Plains faced a major
setback, and those growing irrigated crops had to pump considerably more water than usual
from the Ogallala Aquifer. In most of the Southern High Plains, the Ogallala Aquifer is
being continually depleted. Although it is debatable the number of years this aquifer can
continue to support agriculture in the High Plains, most of the reports suggest that it will be
unable to support extensive irrigated agriculture within a few decades. Hence, to sustain
agriculture in the Southern High Plains, it is important to use the water from this aquifer
judiciously.
Another factor contributing to the difficulty of sustaining crop production in the
Southern High Plains is the semi-arid climate. The efficiency with which a crop can utilize
irrigation depends on the climate. Plants growing in dry weather are in a high evaporative
demand condition. This situation can reduce the irrigation efficiency, as there can be
increased evaporative loss of water from the soil surface (Kreig, 2000). To truly assess
irrigation efficiency, one must have an estimate of the amount of water actually used in
growing the crop. This is called the crop water use (CWU), and is essentially equal to the
transpiration of the crop. Knowing CWU, one can determine the water use efficiency
(WUE) of the crop (in terms of the biomass produced per unit of water transpired), along
with the efficiency of applied irrigation (in terms of CWU per unit of irrigation applied to
1
the crop). By providing accurate information on CWU, particularly at the field and
regional scales, one can hope to influence farming practices that potentially improve
irrigation water management and conservation.
Many procedures have been proposed for estimating CWU. The most common
approach for estimating daily CWU involves multiplying a crop coefficient Kc by the daily
value of reference evapotranspiration (ET0) for a well-watered reference vegetation (Allen,
2003),
CWU = Kc x ET0
[Eq.1.1]
Here, ET0 is calculated from ambient weather conditions, and Kc is determined empirically
for a specific crop. The value of the crop coefficient normally varies over the duration of
the growing season, increasing from a value near zero early in the season to a value near 1
in mid-season. As the crop matures, the value of the crop coefficient starts declining. Crop
coefficients are used to adjust the potential evapotranspiration (PET) of the reference crop
to match the PET of agricultural crops. These crop coefficients are empirically determined
by comparing the actual water use of crops grown in precision weighing lysimeters with
the water use of reference crops such as short grass or alfalfa. Researchers have developed
crop coefficients for different crops that are applicable to different climatic regions. These
are published by organizations such as the United Nations Food and Agriculture
Organization (FAO). The FAO guidelines for computing crop evapotranspiration are used
world-wide to estimate crop water use and schedule irrigation. However, the crop
coefficients determined by the above-described method may not reflect conditions in a
given field. So, this study was undertaken in an attempt to develop crop coefficients that
are real-time and are specific to the crop and the particular set of growing conditions.
2
The variation in the crop coefficient over the growing season tends to follow the
variation in crop canopy density. Thus, it has been suggested that remotely sensed
measures of crop canopy density, such as vegetation indices, can be used to approximate
the crop coefficient in estimating CWU (Jackson et al., 1980). In the past two decades,
numerous researchers came up with different methodologies to quantify the crop
coefficient from remote sensing data. Most of these researchers have come up with
empirical relationships between the regular crop coefficients and some type of vegetation
index, such as the Normalized Difference Vegetation Index or NDVI (Hunsaker et al.,
2005). These reflectance-based crop coefficients are sensitive to the actual field conditions
(Neale et al., 1989), but still rely on the regular crop coefficients and reference
evapotranspiration measurements.
In this study, we make use of the concept of Potential Evapotranspiration (PET) to
estimate CWU. PET is the maximum evapotranspiration possible from a homogenous,
horizontally uniform crop canopy. The Penman-Monteith equation can be used to calculate
the PET of the crop by assuming that the plant canopy is represented as a “big leaf” that
completely covers the soil surface (Raupach and Finnigan, 1988). For a crop with
incomplete ground cover, it is hypothesized that the CWU can be approximated by
multiplying the PET for a uniform crop by the observed crop ground cover (GC). Ground
cover measures the degree to which a crop canopy covers the soil surface. Thus, GC is
numerically similar to Kc in that it also varies from near zero early in the growing season to
1 at maximum canopy development. In general, CWU for a crop may be estimated,
[Eq.1.2]
CWU = GC x PET x Fs
where Fs is a factor ranging from 0 to 1 that represents the degree to which the leaf stomata
3
are open. Thus, Fs represents a “stress factor” that reduces CWU as the stomata close in
response to reduced soil moisture. For crops acclimated to their environment, however,
water loss from the plants is more effectively controlled by limiting the leaf area on the
plants as opposed to limiting the opening of stomata on the leaves. Because photosynthesis
is also affected by stomatal opening, it is better under conditions of limited soil moisture
for the plants to have a relatively small amount of leaf area while maintaining relatively
open stomata than to have a relatively large amount of leaf area with closed stomata.
Therefore, for a crop acclimated to its environment, it is hypothesized that Fs should be
approximately 1. In this case, [Eq.1.2] reduces to
CWU = GC x PET
[Eq.1.3]
CWU in [Eq.1.3] is different from actual evapotranspiration, ETa, in that it does not
include soil evaporation. [Eq.1.3] provides a means of estimating daily CWU using remote
sensing, since GC can be easily estimated from remote sensing observations. The use of
GC in place of the standard empirically determined Kc allows the estimation of CWU to be
specific for a given field.
The use of remote sensing to estimate the crop coefficient has several advantages over
the conventional crop coefficient method. Remote sensing observations reflect the actual
growing conditions in a field, so that these measurements implicitly include the influences
of weather and other growth-limiting factors that may uniquely be present in that field.
Unlike the conventional crop coefficient, a crop coefficient that could be based on remote
sensing could successfully capture the spatial variability within the field.
The main objective of my research is to develop a spectral crop coefficient that
can be used in estimating regional crop water use. Specific objectives are:
4
1. To derive a coefficient related to the ratio of actual to potential crop
evapotranspiration through vegetation ground cover quantified from remote
sensing (the spectral crop coefficient, Ksc).
2. To compare the crop water use of different crops estimated by the spectral crop
coefficient and regular crop coefficient methods.
3. To compare estimates of daily crop water use determined using this methodology
against actual field measurements of crop evapotranspiration.
5
Chapter II
Literature Review
2.1. Evapotranspiration
Evapotranspiration is the combined process of evaporation from the soil and
transpiration from the plants (Thornthwaite, 1948). The concept of potential
evapotranspiration was first introduced by Thorthwaite in 1948. According to him,
potential evapotranspiration is the maximum evapotranspiration from a vegetation
completely covering the ground surface that has unlimited water supply to its roots.
Although Thorthwaite came up with a temperature-based empirical equation to estimate
evapotranspiration, the values did not match the actual measurements (Fuchs, 2003).
Penman in 1948 introduced another method to estimate the evapotranspiration based on
the energy balance of the surface. This method, popularly known as the Penman method,
is one of the most discussed methods to measure evapotranspiration. Some differences
were found in the evapotranspiration estimated by the Penman method when compared
with actual measurements. Monteith (1965) modified the Penman equation and
incorporated an additional surface (vegetation) resistance term. The modified Penman
equation is known as the combination equation because it involves principles of energy
balance and resistance to water vapor movement. When applied to a plant canopy, this
combination equation assumes the evaporating surface as a single big leaf (Raupach and
Finnigan, 1988). Allen et al. (1989) found that the Penman- Monteith method produced
more accurate estimates of evapotranspiration when compared with other forms of
Penman’s equation.
6
The Penman-Monteith equation is widely used by researchers to estimate crop
evapotranspiration. To estimate crop evapotranspiration, the potential evapotranspiration
was modified to measure the evapotranspiration from a reference crop surface such as
short grass (Example: Tall fescue – Festuca arundinacea) or alfalfa (Medicago Sativa L.)
and multiplied by a factor called the crop coefficient (Doorenbos and Pruitt, 1977,
Wright, 1982). The crop coefficient is the ratio of reference evapotranspiration to crop
evapotranspiration. The crop coefficient incorporates the effects due to the difference
between the hypothetical reference crop and various field crops in terms of crop height,
surface resistance, and albedo (Allen, 2000). The value of crop coefficient varies
depending on the crop growth stage, from a value near zero during the early growing
season to a value near 1 during the mid-growing season. The value of the crop
coefficient declines in the late growing period as the crop matures (Jensen et al., 1990,
Allen et al., 1998).
The United Nations Food and Agriculture Organization (FAO) played an
important role in popularizing the crop coefficient approach for estimating crop
evapotranspiration. FAO published the details of estimating crop evapotranspiration and
crop coefficients in the publication FAO-24 (Doorenbos and Pruitt, 1977) based on the
Penman equation. Researchers found that the methodology published in FAO-24 tended
to overestimate the reference evapotranspiration and suggested the use of PenmanMonteith equation in place of the Penman equation (Allen et al., 1994). The FAO
irrigation and drainage paper 56 (Allen et al., 1998) included the revised procedures for
estimating reference evapotranspiration. In this publication, the reference crop is a
hypothetical grass surface growing under ideal conditions with height 0.12 m, fixed
7
surface resistance of 70 s m-1, and an albedo of 0.23. This publication also presents two
types of crop coefficients: a single crop coefficient for use when the soil surface is dry,
and a dual crop coefficient (a basal crop coefficient and a soil evaporation factor) when
the soil surface is wet as proposed by Wright (1982).
The crop coefficient approach to estimating evapotranspiration is the most
commonly used method world-wide in irrigation applications. Researchers have
developed crop coefficients for different agricultural crops by growing plants in precision
weighing lysimeters and collecting data for several years. Wright (1982) developed crop
coefficient curves for various Pacific Northwest irrigated crops such as, alfalfa
(Medicago sativa L.), potatoes (Solanum tuberosum L.), peas (Pisum sativum L.), and
corn (Zea mays L.), based on the alfalfa reference evapotranspiration. Wright and
Hanson (1990) developed crop coefficients for rangeland from lysimeter studies
conducted in three northern states in the USA and found that these crop coefficients were
the same.
Allen et al. (2000) compared the crop evapotranspiration of three agricultural
crops, snap beans (Pisum sativum L.), sugarbeets (Beta vulgaris L.), and sweet corn,
grown in lysimeters in Kimberly, Idaho, with the crop evapotranspiration calculated by
the FAO-56 method. They found that both methods produced similar results. Brown et
al., (2001) developed crop coefficients for turf grass growing in a desert climate. They
observed that the crop coefficients varied during cloudy condition in winter and
suggested that irrigation scheduling based on weather data may be less reliable under
those environmental conditions. Benli et al. (2005) developed basal crop coefficients for
alfalfa along with a weighing lysimeter for estimating the reference evapotranspiration in
8
the semi-arid region around Ankara, Turkey. Howell et al. (2006) developed crop
coefficients for major irrigated crops in the Texas High Plains by measuring
evapotranspiration with large precision weighing lysimeters. The crops were corn, wheat
(Triticum aestivum L.), sorghum (Sorghum bicolor L.), soybean (Glycine max L.), cotton
(Gossypium hirsutum L.), and alfalfa. These crop coefficients, known as the Bushland
crop coefficients, are used in the evapotranspiration networks in Texas and surrounding
states.
Hunsaker (1999) developed crop coefficients for a short season cotton variety in
Arizona and found that these crop coefficients were larger than the published FAO-56
crop coefficients for cotton. Howell et al. (2004) applied the FAO-56 method to estimate
the evapotranspiration for well-watered, deficit-irrigated, and dryland cotton on the
Northern Texas High Plains. They found that the FAO-56 method performed better for
the fully irrigated cotton than for the deficit-irrigated and dryland cotton. However,
Sueleiman et al. (2007) found that in humid- areas the FAO-56 method worked
accurately for estimating the cotton evapotranspiration under deficit irrigation conditions.
Marek et al. (2006) compared the evapotranspiration of cotton, grain sorghum, and
soybeans by the FAO-56 and Bushland crop coefficient methods. They used the
ASCE/EWRI reference evapotranspiration equation (Allen et al., 2005) to calculate the
reference evapotranspiration. They found major differences in crop evapotranspiration
estimated by these methods in the late growing season for corn and grain sorghum.
Several scientists have observed that the FAO-56 method has had problems in
accurately predicting crop evapotranspiration. Allen (1999) concluded that these
differences in crop evapotranspiration may be due to the fact that the crop growing
9
conditions in other studies are not representative of the conditions used to develop the
crop coefficients in FAO-56.
2.2. Reflectance-based crop coefficient
The rapid technological developments in the past three decades have facilitated
the use of remote sensing as a tool for agricultural crop management (Pinter et al., 2003).
Remotely sensed measures of spectral reflectance in the form of vegetation indices can be
used to indirectly estimate the crop coefficient, because evapotranspiration is related to
Leaf Area Index (LAI) and fractional vegetation cover (Glenn et al., 2007). Jackson et
al. (1980) first suggested that reflectance measurements of the crop-soil scene could be
used to evaluate the crop coefficient, since both the crop coefficient and canopy-soil
reflectance are closely related to plant growth. They found that the crop coefficient curve
and the curve describing the ratio of the Perpendicular Vegetation Index (PVI) to the
maximum value of PVI for a crop were similar for small grains, indicating the possibility
of evaluating the seasonal change in the crop coefficient from the reflectance data.
Heilman et al. (1982) reported statistically significant linear relationships between
the crop coefficient and percent ground cover, and PVI and percent ground cover, for
irrigated alfalfa. This suggests that the spectral measurements could be used to evaluate
the crop coefficient. Bausch and Neale (1987, 1989) and Neale et al. (1989) developed
crop coefficient curves similar to the basal crop coefficient curve for corn using spectral
measurements. They used a linear transformation of Normalized Difference Vegetation
Index (NDVI) to derive the reflectance-based crop coefficient and found that this
modified basal crop coefficient allowed proper timing of irrigation. Unlike the traditional
crop coefficient method, the reflectance-based crop coefficient is sensitive to the actual
10
condition of the crop resulting from weather conditions. This ability of the reflectancebased crop coefficient to represent actual crop growth and water needs helped to improve
irrigation scheduling in corn (Neale et al, 1989).
Choudhury et al. (1994) found a statistically significant relationship between the
transpiration coefficient and LAI, and between the transpiration coefficient and
vegetation indices such as Soil Adjusted Vegetation Index or SAVI (Huete, 1988) and
NDVI for an unstressed wheat crop. Their results also showed that the modeled
evapotranspiration measurements using vegetation indices agreed with lysimeter data.
Bausch (1993) pointed out that the NDVI-based crop coefficient curves could be affected
by soil background effects, and hence recommended the use of SAVI-based crop
coefficient curves for irrigation scheduling of corn. Bausch (1995) compared three basal
crop coefficient curves. These were SCHED (the USDA-ARS irrigation scheduling
program), the tabular crop coefficient as proposed by Wright (1982), and the reflectancebased crop coefficient (using SAVI). He found that irrigation scheduled using the
reflectance-based crop coefficient was more appropriately timed. The reason was that the
variation in this reflectance-based crop coefficient was synchronized with crop
development; and thus irrigation was applied when it was needed. This avoided over- or
under-irrigation. Neale et al. (1996) developed SAVI-based crop coefficients for
irrigation scheduling of cotton.
Ray and Dadhwal (2001) derived crop coefficients by constructing a linear
relationship between monthly averaged SAVI and crop coefficients. The remote sensing
data was obtained from the wide-field sensor onboard the IRS-IC satellite. The
regression equation was used to derive pixel-specific crop coefficient values.
11
Humsaker et al. (2003) developed seasonal basal crop coefficient curves for a
full-season cotton cultivar using the regression relationships between NDVI and basal
crop coefficient. The crop evapotranspiration estimations made using the NDVI-based
crop coefficient model closely matched the actual evapotranspiration observations made
for two other cotton cultivars grown under different conditions. Hunsaker et al. (2005)
developed wheat basal crop coefficients using NDVI and found that the measured and
estimated ET was in agreement. They suggested that remotely sensed vegetation indices
such as NDVI can be used to determine a real-time crop coefficient for irrigation
scheduling. Neale et al. (2005) developed SAVI-based crop coefficients for beans and
potato grown in southern Idaho using radiometer-derived reflectance measurements.
They suggested that high resolution aerial images could be used to capture the in-field
variability in crop growth, and these crop coefficients could be useful for irrigation
scheduling.
Bashir et al. (2006) derived a seasonal crop coefficient curve for irrigated
sorghum using Landsat ETM+ images by dividing the actual evapotranspiration
computed for each pixel using SEBAL (Surface Energy Balance Algorithm for Land) by
the reference evapotranspiration estimated by the FAO-56 method. They concluded that
the satellite-based energy balance models such as SEBAL can be used to update and
verify the existing crop coefficients for a region. Duchemin et al. (2006) found that the
relationship between the basal crop coefficient and NDVI estimated from a hand-held
radiometer was linear for winter wheat grown in Central Morocco. To show the
application of this in irrigation scheduling, they used two Landsat-7 ETM+ images
12
containing the study area and converted them into transpiration requirement maps. They
suggested that these maps could be used to decide how water should be applied spatially.
Er-Raki et al. (2007) compared three methods for estimating the crop coefficient
for winter wheat grown under different irrigations in Central Morocco and compared the
crop evapotranspiration with actual evapotranspiration collected using the eddy
covariance method. The results showed that the FAO-56 method was unable to estimate
the crop evapotranspiration accurately. Hence, the FAO-56 crop coefficients were
locally calibrated and then found to give good results. They also suggested that the basal
crop coefficients could be determined using NDVI since NDVI derived from ground
measurements of reflectance and basal crop coefficient had similar seasonal patterns.
Although numerous papers are available on directly estimating evapotranspiration
using remote sensing data, most of them use remote sensing data to evaluate terms in an
energy balance to estimate the latent heat flux. Few researchers have come up with a
novel method for estimating the crop coefficient curve using remote sensing since
Jackson et al. (1980) suggested it two decades ago.
2.3. Vegetation indices and ground cover
The red and near-infrared reflectance of agricultural fields obtained using
multispectral imagers aboard satellites and aircraft are widely used to estimate crop
growth-related parameters such as LAI or GC (Barnes et al., 1996). This is usually
achieved by deriving empirical relationships between multispectral vegetation indices
such as the NDVI and LAI or GC (Carlson and Ripley, 1997; Turner et al., 1999).
Although there are numerous studies done on the use of remote sensing data to estimate
13
LAI of different vegetation, few works has been done to quantify the GC of field crops
from remote sensing data.
Ormsby et al. (1987) reported strong linear relationships between NDVI and the
Simple Ratio (SR) vegetation index, and fractional vegetation cover. The error in
determining the fractional cover was less than 12.7 %. Their study showed that an NDVI
value of 0.3 or less corresponded to areas with fractional vegetation cover less than 5%,
while an NDVI of more than 0.7 indicated 80% or more ground cover. Carlson et al.
(1990) calculated the fractional vegetation cover for an agricultural region with remotely
sensed measurements of surface temperature and NDVI. Bouman et al. (1992) estimated
GC of potato from the Weighted Difference Vegetation index (WDVI). These WDVIderived GC estimates was in good agreement with the GC estimated by visual inspection
by trained experts. Pickup et al. (1993) described a perpendicular difference index
(PD54) similar to the Perpendicular Vegetation Index (PVI) to estimate vegetation cover
using Landsat data. They found that this index could produce reasonably good estimates
of ground cover for rangeland. Wittich and Hansing (1995) found that the relationship
between NDVI and fractional cover was linear, and suggested that an area-average NDVI
approach with suitable corrections would be ideal to estimate fractional ground cover.
Carlson and Ripley (1997) observed that NDVI was sensitive to fractional vegetation
cover until full ground cover was reached. After the attainment of full ground cover,
NDVI became insensitive to increasing vegetation amount and the regression relationship
between NDVI and fractional cover was non-linear. Purevdorj et al. (1998) described the
relationship between several vegetation indices and percent ground cover for grasslands.
Their results showed that NDVI and TSAVI (Tranformed Soil Adjusted Vegetation
14
Index) provided the best estimates of vegetation cover. Choudhury et al. (1994) used the
data from Heute et al. (1985) and showed that the relationship between SAVI and
fractional vegetation cover of cotton was linear. Goel and Grier (1986) reported that a
canopy reflectance model could be used to accurately determine the ground cover of row
crops such as corn and soybean.
Maas (1998) used a linear mixture model to estimate ground cover (GC) of cotton
from ground-based measurements of red and near-infrared scene reflectance. Maas
(2000) applied this approach to estimating GC of cotton fields using Landsat
multispectral imagery. Using ATSR-2 imagery, North (2002) found that a linear mixture
model based on a library of spectral signatures was a better method to estimate fractional
vegetation cover than using regular vegetation indices.
White et al. (2000) used digital cameras to estimate vegetation cover in shrub
land. Zeng et al. (2000) used NDVI to estimate global fractional vegetation cover using
Advanced Very High Resolution Radiometer (AVHRR) data. They used the annual
maximum value of NDVI for each pixel and the NDVI value corresponding to 100% GC
to get an estimate of fractional vegetation cover. Ringersam and Sikking (2001)
estimated the GC of vegetation barriers by observing the shaded areas at noon in an
attempt to determine transpiration coefficients. Gitelson et al. (2002) developed
algorithms based on spectral values in the visible spectral range for a wheat canopy to
estimate the vegetation cover fraction. They suggested that these newly proposed
algorithms could predict vegetation cover fraction with less than 10% error, and could
replace other popular indices such as NDVI and SAVI.
15
Wanjura et al. (2003) reported a linear relationship between NDVI and ground
cover of seedlings of cotton and corn. Hirano et al. (2004) found a linear relationship
between factional vegetation cover obtained from aerial photographs and NDVI. Xiao
and Moddy (2005) reported a linear relationship between NDVI and fractional vegetation
cover. They concluded that simple NDVI-based methods performed well for regional
estimation of fractional green vegetation cover. They found that these methods
overestimated fractional vegetation cover in areas with sparse vegetation with bright soils
and senesced vegetation. Moredorf et al. (2006) discussed a methodology for estimating
fractional vegetation cover from airborne laser scanning (LIDAR) data. The qualitative
comparison of fractional cover maps with equivalent maps based on imaging
spectrometry showed similar ranges of values. Qi et al. (2006) mapped the spatial and
temporal fractional vegetation cover using a modified form of NDVI and found
satisfactory relationship between NDVI and fractional ground cover. They suggested that
this method should be improved to get accurate results with different vegetation types.
NDVI has been the most popular vegetation index to estimate plant canopy
characteristics such as LAI and GC. However, several scientists (described above) have
observed both linear and non-linear relationships between NDVI and these plant
characteristics. Small (2001) concluded that, in the absence of a consistent regression
relationship between NDVI and GC, use of NDVI may not be a good method for
estimating GC.
16
Chapter III
Materials and Methods
3.1. Study area
The study was conducted in 16 agricultural fields in Hale and Floyd counties in
the Texas High Plains (Fig. 3.1.1) that are part of a large demonstration project called the
Texas Alliance for Water Conservation (TAWC). These fields were planted to different
crops under different management systems. Based on hectares planted, the dominant
field crop was cotton (Gossypium hirsutum L.), followed by corn (Zea mays L. – for grain
and silage), sorghum (Sorghum bicolor L. – for grain and silage), wheat (Triticum
aestivum L. – for grain and as a winter cover crop), pearlmillet (Pennisetum glaucum L.)
and alfalfa (Medicago sativa L.). Among the 16 fields in study, 9 were center pivot
irrigated fields, of which 8 were full circles. Other irrigation systems used were
subsurface drip (3 fields) and furrow (2 fields). Two fields were not irrigated (dryland).
Some of the center pivot, furrow, and dryland fields had more than one crop planted in
different sections. The details of each field, including field number, section number,
crop, and irrigation type are given in Table 3.1.1. Among the different crops, corn was
planted the earliest, during the third and fourth weeks of April. Cotton and other crops
were planted in May. Harvest dates varied from field to field depending on the crop type.
Predominant soils in the study area are noncalcareous clay loams and loams in the
Pullman and Pullman-Olton associations (NRCS, 1974, 1978). The climate is semi-arid
and the topography is nearly level to gently sloping. Detailed results will be presented in
this dissertation for a subset of fields indicated in Table 3.1.1.
17
Fig. 3.1.1. Locations of study fields in Hale and Floyd counties of Texas.
Fig. 3.1. Locations of Study Fields in Hale and Floyd
Counties of Texas
Legend
Study Fields
Primary Roads
County Boundary
Secondary
Roads
18
Table 3.1.1 Field number, section number, irrigation type and crop for fields in the study
in 2006 and 2007
Crop
Field
No
Section
Irrigation
type
2006
2007
1
1
Drip
Cotton*
Cotton*
1
2
Drip
Cotton*
Cotton*
2
..
Drip
Cotton*
Cotton*
3
1
Center-pivot
Cotton*
Cotton
3
2
Center-pivot
Cotton*
Grain Sorghum*
4
2
Center-pivot
F. Sorghum*
Cotton
6
..
Center-pivot
Cotton*
Cotton*
11
1
Furrow
Cotton*
Cotton*
12
1
Dryland
F. Sorghum
Cotton*
12
2
Dryland
Cotton*
Grain Sorghum*
13
1
Dryland
Cotton*
Wheat
15
1
Furrow
Cotton*
Cotton*
15
3
Furrow
Cotton*
Grain Sorghum*
15
4
Furrow
Grain Sorghum*
Cotton*
18
2
Center-pivot
Oats
G. Sorghum*
19
3
Center-pivot
Pearlmillet*
Cotton
20
1
Center-pivot
F. Sorghum*
Corn*
20
2
Center-pivot
Corn*
F. Sorghum*
Center-pivot
Cotton*
Cotton
22
24
1
Center-pivot
Corn*
Corn*
24
2
Center-pivot
Cotton
Corn*
26
1
Center-pivot
Corn*
Millet*
26
2
Center-pivot
Cotton
Corn*
27
Drip
Cotton
* indicates fields used in the present study
Corn*
19
3.2. Meteorological data
The weather data used in the study were obtained from the West Texas Mesonet
stations at Plainview and Floydada, Texas, and from the Texas High Plains
Evapotranspiration Network weather station at Lockney, Texas.
3.3. Potential Evapotranspiration
The standard Penman-Monteith combination equation (Allen et al., 1998) used to
calculate the Potential Evapotranspiration (PET) from homogenous areas of vegetation
can be expressed as follows:
( es − ea )
ra

r 
∆ + γ 1 + s 
ra 

∆( Rn − G ) + ρ a c p
λET =
[Eq.3.1]
where λET is the latent heat flux (MJ m-2 d-1), Rn is the net radiation (MJ m-2 d-1), G is
the soil heat flux (MJ m-2 d-1), ( es − ea ) is the vapor pressure deficit between the ambient
air and the evaporating surface (k Pa oC-1), ρ a is the mean air density (kg m-3), c p is the
specific heat of air at constant pressure (MJ kg-1 oC -1), ∆ is the slope of the saturation
vapor pressure curve at the ambient air temperature (k Pa oC-1), γ is the psychrometric
constant (k Pa oC-1) , ra is the aerodynamic resistance (s m-1), and rs is the surface
resistance (s m-1). The above equation can be applied to different vegetation surfaces
with appropriate parameterization of the equation (Allen, 2005). For well-watered
vegetation surfaces, the aerodynamic resistance term can be calculated as follows (Allen
et al., 1998, p.20):
20
 z − d   zh − d 
 ln 

ln m
zom   zoh 

ra =
k 2uz
[Eq.3.2]
where ra is the aerodynamic resistance (s m-1), zm is the height of wind measurements
(m), zh is the height of humid-ity measurements (m), d is the zero-plane displacement
height (m), zom is the roughness length governing momentum transfer (m), zoh is the
roughness length governing transfer of heat and vapor (m), k is von Karman's constant
(0.41), and uz is the wind speed at height z (m s-1). The weather data used in the study
contained wind and humid-ity measurements made at a height of 2 m. The other
parameters (d, zom and zoh) are dependent on crop height h (m) and can be estimated using
the following equations (Allen et al., 1998, p. 21).
d = 2/3 h
[Eq.3.3]
zom = 0.123 h
[Eq.3.4]
zoh = 0.1 zom
[Eq.3.5]
By substituting [Eq.3.3], [Eq.3.4], and [Eq.3.5] in the equation to calculate ra [Eq.3.2],
the aerodynamic resistance for a well watered crop surface can be expressed as follows,
 2 − 0.67h   2 − 0.67h 
ln
 ln

0.123h   0.0123h 

ra =
0.1681u 2
[Eq.3.6]
Surface resistance rs is a function of effective leaf area index (LAI) for densely vegetated
crops (Allen et al., 2006) and is computed as:
rs =
rl
LAI eff
[Eq.3.7]
where rl is the bulk stomatal resistance in s m-1 and LAIeff is the effective LAI, which is
the leaf area that is actively contributing to PET. Previous studies have shown that, for
21
well watered agricultural crops, rl is approximately 100 s m-1 (Monteith, 1965; Allen et
al., 1989, and Allen et al., 2006). This value is widely accepted by FAO and ASCE for
reference ET calculation. For agricultural crops with dense canopies, only 50 percent of
the canopy in the upper part is active in heat and vapor transport (Choudhury and Idso,
1985., FAO 56 p. 22), hence LAIeff can be computed as:
LAIeff = 0.5 LAI
[Eq.3.8]
In the standard Penman-Monteith equation, the energy terms (Rn and G) expressed as
flux densities (MJ m-2 day -1) can be converted to equivalent water depths by dividing the
energy terms by the latent heat of vaporization, 2.45 kJ g-1. Using the Ideal Gas Law
Equation, the air density can be computed:
ρa =
P
1.01(T + 273) R
[Eq.3.9]
where P is the atmospheric pressure in kPa, T is the air temperature in oC, and R is the
Universal Gas Constant (0.287 kJ kg-1 K-1).
The specific heat at constant pressure is computed as:
Cp =
γελ
[Eq.3.10]
P
where γ is the psychrometric constant, ε is the ratio of the molecular weight of water
vapor density to that of dry air (0.622), and λ is the latent heat of vaporization. After
substituting [Eq.3.6], [Eq.3.8], [Eq.3.9], and [Eq.3.10] into [Eq.3.1], the PET of a fully
irrigated agricultural crop can be calculated using a modified form of [Eq.3.1] as follows:
185396γ ( es − ea )
(T + 273) ra

200 

∆ + γ 1 +
ra LAI 

0.408∆( Rn − G ) +
PET =
22
[Eq.3.11]
Equation 3.11 was used to calculate the PET of cotton, corn, sorghum (grain and forage),
and millet using weather data and LAI values at 100 per cent GC, which is approximately
3 (Glenn et al., 2007). The h in [Eq.3.11] is the estimated height of the crop during the
growing season.
3.4. Actual Evapotranspiration
Mobile Eddy Covariance System
Two mobile eddy covariance systems (called Systems 1 and 2 – Fig. 3.4.1) were
built in 2005 to measure evapotranspiration from the study fields. Each system consisted
of a CSAT-3 sonic anemometer (Campbell Scientific, Inc.) and a LI-7500 Infrared gas
analyzer (LI-COR Biosciences) attached to a mast mounted on a trailer. Electrical power
was supplied by two MSX64 solar panels (Campbell Scientific, Inc.) connected to an
external 12-V lead-acid battery, which in turn supplied power to the CR23X.
Calibration Unit
The LI-7500 was calibrated on a regular basis to ensure accurate measurement of
water vapor density. Before assembling the mobile eddy covariance systems, both LI7500 gas analyzers were sent to LI-COR Biosciences in Lincoln, Nebraska, for factory
calibration. For user calibration, a calibration unit was built at the USDA Plant Stress
Laboratory in Lubbock, Texas (Fig. 3.4.2). This unit was built onto a hand dolly and
consisted of a LI-610 Portable Dew Point Generator (LI-COR Biosciences, Lincoln,
Nebraska) for water vapor calibration and a regulated source of carbon dioxide gas (500
ppm in N) for CO2 calibration. The gas was supplied by a 34-L cylinder with a 400
series regulator (Calibration Gas, http://www.calibration-gas.com/index.htm). An LI-670
23
Flow Control Unit (LI-COR Biosciences, Lincoln, Nebraska) was used to provide CO2free air and dry air for zeroing the LI-7500.
The initial calibration was done at the U. S. Department of Agriculture (USDA)
laboratory at the beginning of each growing season before the mobile eddy covariance
systems were taken to the field. Subsequent calibrations were done in the field. Standard
procedures published in the LI-7500 User’s Manual were used to calibrate the LI-7500.
Although the CO2 concentration was calibrated each time, the CO2 flux data were not
used in the study reported in this document.
Latent Heat Flux
The eddy covariance program supplied by Campbell Scientific, Inc., was used to make
measurements of latent heat flux using a Campbell Scientific CR23X datalogger. The
measurements were taken every 0.1 seconds (10Hz) and averaged every half hour. To
meet the fetch requirements, the sensors were placed 1.5 m above the crop canopy.
24
Fig. 3.4.1. Mobile eddy covariance system located in Field No.2.
Fig. 3.4.2. LI-7500 calibration unit built on a hand dolly for user calibration of LI-7500
Infrared Gas Analyzer
25
3.5. Remote sensing data
Satellite data
Landsat-5 Thematic Mapper (TM) and Landsat-7 Enhanced Thematic Mapper
(ETM+) imagery containing the study site was acquired on several dates during the 2006,
and 2007 growing seasons (Table 3.5.1). Images containing the study area were located
along Path 30 at Row 36, according to the Landsat World Reference System (WRS-2).
Imagery was purchased on CD-ROM from the U.S. Geological Survey (USGS)
EarthExplorer website (http://edcsns17.cr.usgs.gov/EarthExplorer/). The images received
Level 1-G processing by USGS prior to delivery and had a pixel size of 30 m and a
radiometric resolution of 8 bits (256 gray levels). This processing included systematic
correction to rotate, align, and project the image to the World Geodetic Survey 1984
(WGS84) datum, georeferencing to the Universal Transverse Mercator (UTM) coordinate
system, and radiometric correction based on characteristics of the TM sensor (Chander
and Markham, 2003).
26
Table 3.5.1. Acquisition dates of images containing the study area from Landsat-5
Thematic Mapper (TM) and Landsat-7 Enhanced Thematic Mapper
(ETM+) satellite sensors in 2006 and 2007.
Year
Landsat-5 TM
Landsat-7 ETM+
13 May
29 May
8 July
30 June
09August
16 July
25August
2006
01August
10 September
18 September
26 September
04 October
20 October
22 April
2007
29 March
24 May
19 July
27 July
5 September
12 August
28 August
13 September
27
Aerial image data
Aerial imagery was acquired during the 2006 and 2007 growing seasons using the
Texas Tech Airborne Multispectral Remote Sensing System (TTAMRSS). This system
contained two Dalsa 1M30 digital CCD cameras, a portable computer, a separate monitor
for use by the pilot, and a Global Positioning System (GPS) receiver. Narrow-bandpass
filters were mounted on the lenses of the cameras to obtain images in the red (660 nm)
and near-infrared (850 nm) spectral bands. The filters were used to block radiation from
other wavelengths so that the camera sensor is sensitive to only red (RED) or nearinfrared (NIR) radiation. Prior to mounting the system in the aircraft, the focus and
exposure of the two cameras were set at the USDA Plant Stress Laboratory in Lubbock,
Texas. Focus was adjusted by viewing an object several miles away. The exposure was
adjusted by exposing the cameras to white and black calibration targets.
TTAMRSS was mounted in a Cessna Model 172 airplane (Fig. 3.5.1). Images
were acquired at an altitude of 9000 to 9500 feet above ground level. The altitude of
9000 to 9500 feet was chosen to limit bidirectional reflectance effects on the imagery,
and also to include the whole field in a single image. Unlike Landsat, TTAMRSS had the
capability of acquiring imagery with 12-bit radiometric resolution (4096 gray levels).
The system had a fixed gain and offset which allowed comparison of images of different
fields on the same date without radiometric correction. The image size was 1024 x 1024
pixels with an individual pixel size of approximately 2 m. Fig. 3.5.1 show various
components of TTAMRSS.
Tracker software version 2.0 (Tetracam Inc.) was used for GPS-guided camera
triggering. The AgGPS 132 receiver from Trimble was upgraded to receive the Wide
28
Area Augmentation System (WAAS) differential correction for greater accuracy based on
National Marine Electronic Association (NEMA) transmissions. Prior to each flight, a
flight plan was made which contained the latitude and longitude coordinates of each
target field. The software triggered the cameras when the aircraft flew over each field
and the image files were stored in the computer memory with the geographical
coordinates incorporated into their file names. Table 3.5.2 summarizes the TTAMRSS
acquisition dates in 2006 and 2007.
29
Figure 3.5.1. Various parts of Texas Tech Air-borne Multispectral Remote Sensing
System (TTAMRSS) and the Cessna Model 172 plane used to acquire
aerial images of fields in the study.
Table 3.5.2. Image acquisition dates for fields in the study using the Texas Tech
Airborne Multispectral Remote Sensing System (TTAMRSS)
Year
Date
2006
28 Aug
8 June
21 June
10 July
9 Aug
14 Aug
2007
30
3.6. Ground-based Observations
Ground-based observations of ground cover (GC) were made several times during
the growing season. Some of these observations were made around the dates of a satellite
image or aerial image acquisition to compare the remote sensing-based observations of
GC with the actual ground-based measurements of GC. Three methods were used for
making ground-based observations of GC. The method selected at any given time
depended on the crop type, stage of crop growth, and the time of data collection. For row
crops with open canopy structure, such as corn, sorghum, and millet, overhead
photographs of the canopy were taken using a standard digital still camera mounted on a
long pole (Fig. 3.6.1). The camera at the end of the pole was positioned approximately 3
m above the ground pointing directly down at the plant canopy. This method was also
used in the early stages of cotton. The second method was used for mature rows of
cotton plants. In this case, a meter stick was used to obtain the approximate width of the
plant canopy perpendicular to the row direction. Typically, 20 to 30 measurements were
made within the sampled portion of the field and the average of these measurements
provided the average canopy width. GC was determined by dividing the average leaf
canopy width by the row spacing of the field. The third method utilized an AccuPAR
Model PAR-80 Linear PAR Ceptometer (Decagon Devices, Pullman, WA) and was used
for measuring the GC of alfalfa and corn. Before taking measurements, the ceptometer
was calibrated to the ambient irradiance level by holding the instrument horizontally
above the plant canopy. Measurements were then taken by placing the instrument under
the canopy at the soil surface. A set of individual measurements (typically 20 to 30) were
made with in the sampling area in the field and averaged to estimate the GC.
31
Fig 3.6.1.
Shooting over-head pictures in a corn field for making ground-based
observations of ground cover
32
3.7. Image Processing
Image processing was done using the Environment for Visualizing Images
(ENVI) software package (ITT, Boulder, CO) and Adobe Photoshop (Adobe Systems,
San Jose, CA).
Overhead photographs
Overhead digital images were imported into Photoshop for estimating GC. For
row crops (cotton, corn, and sorghum), the image was cropped to include two or four
rows of plants directly below the camera. For alfalfa, the image was cropped to include
only the central portion of the image. Cropping was done to minimize the effects of
optical distortions of the plant canopy present near the edges of the image. After
cropping, the portions containing the leaf canopy was delineated using the “lasso” tool.
The “histogram” function was then used to determine the number of pixels in the
delineated portions. Dividing the number of pixels in the delineated portions by the total
number of pixels in the cropped image provided an estimate of GC. This method could
be used for any crop, but the delineation of the leaf canopy in the images tended to be a
tedious operation for open plant canopies, like corn or sorghum.
Satellite imagery
The satellite image data were used to estimate GC for each field in the study. The
TM data for each acquisition date were imported into ENVI and a subset containing the
study area was extracted from TM Band 2 (green spectral band), TM Band 3 (red spectral
band), and TM Band 4 (NIR spectral band) of each image. Another subset of smaller size
was extracted from the center of the study area that contained primarily agricultural fields
33
and bare soil without any urban structures. This smaller subset was used to identify the
“bare soil line” and the “100 percent GC point (GC100)”.
When values of reflectance or pixel DC in the red wavelengths are plotted versus
comparable values in the NIR wavelengths for bare soil targets, the values tend to lie
along a straight line. This line is called the “bare soil line” (Fox and Metla, 2005; Fox et
al., 2004; Richardson and Weigand, 1977), and is expressed as:
NIR = a1 RED + a0
where a1 is slope and a0 is the intercept of the soil line. To construct the soil line, a subset
of a Landsat image in Band 2, Band 3, and Band 4 was extracted for the study area that
primarily contained agricultural fields and bare soils without urban structures. Some
images were cloud-free (for example, the TM image on 1 August 2006), but some had
scattered clouds (for example, the TM image on 30 June 2006). The clouds and playa
lakes were masked using the build mask function in ENVI and excluded from any
analysis. The DC values in the NIR spectral band (TM Band 4) were plotted versus the
corresponding DC values in the red spectral band (TM Band 3) for pixel data of the
subset using the 2 dimensional (2D) Scatterplot function in ENVI. The resulting
scatterplot had a distribution of points similar to that in Fig. 3.7.1.a. Fig. 3.7.1.b is a
diagrammatic representation of the distribution of points in Fig. 3.7.1.a. The straight
edge of this distribution represents pixels containing bare soil. For a single soil type,
point a might correspond to pixels containing only dry soil, while point b might
correspond to pixels containing only wet soil. Points along the soil line between a and b
would correspond to pixels with intermediate levels of soil wetness, or pixels with
varying mixtures of wet and dry soil. To determine slope (a1) and intercept (a0) of the
34
soil line, the scatterplot was exported as a JPEG image and brought into Photoshop. A
straight line was placed by visual inspection through the edge, and a1 and a0 were
calculated.
As vegetation starts growing in the field, the crop canopy gradually obscures the
soil surface. Since green vegetation absorbs more red light than NIR, the red DC starts
decreasing while the NIR DC starts increasing. In Fig. 3.7.1a, the body of the
distribution represents pixels containing varying amounts of vegetation (Curran, 1983).
At full canopy, the vegetation completely obscures the soil surface, so DC at full canopy
(DCFC,RED and DCFC,NIR) would represent a single point (point c) in Fig. 3.7.1b. In this
study, the point GC100 was identified visually at the top of each distribution at the location
corresponding to full canopy (point c), and the DC values of GC100 (DCFC,RED, DCFC,NIR)
were recorded.
The Perpendicular Vegetation Index (PVI) was used to estimate the GC from
satellite and aerial image data. PVI is the perpendicular distance from any point in the
sactterplot (Fig. 3.7.1a) to the bare soil line and is calculated using the following
equation,
½
PVI = [DCPIXEL,NIR – a1 (DCPIXEL,RED) – a0] / (1 + a12)
[Eq.3.12]
in which a1 and a0 are the slope and intercept, respectively, of the bare soil line
(Richardson and Wiegand, 1977). The PVI of full canopy represents GC100, the point
identified visually at the top of each scatterplot. Hence, an approximate value of GC for
any point in this distribution could be obtained by dividing the PVI of the corresponding
pixel by the PVI for the full canopy point,
GC = PVI ANY PIXEL/ PVIFC
[Eq.3.13]
35
For each TM and ETM+ image, the red and NIR DC of the point corresponding to
GC100 were used in [Eq.3.12] with the corresponding bare soil line slope and intercept
values to calculate a PVI value of full canopy (GC100 or PVIFC). Average DC values in
the red and NIR spectral bands were determined by selecting each of the 16 agricultural
fields in the study (DCPIXEL,RED, DCPIXEL,NIR) as a region of interest (ROI). These values,
along with the slope and intercept of the bare soil line, were used to calculate PVI values
of each pixel in the field using the Band Math function in ENVI and [Eq.3.12]. The
resulting image was saved as a PVI image. This PVI image was then converted to a
ground cover map using the Band Math function and [Eq.3.13]. The average GC of each
field was determined by taking the GC values of all the pixels in a field.
In 2006, estimates of GC determined for a set of fields from Lansat-5 imagery
were compared to corresponding ground based observations of GC. The accuracy of GC
estimates from satellite data was evaluated by statistical analysis of these data.
36
Fig. 3.7.1. Pixel digital count (DC) values in the near-infrared (NIR) spectral band
plotted vs. corresponding DC values in the red spectral band for a portion of a
Landsat-5 image of an agricultural region. (A) Actual distribution of DC
values; (B) diagrammatic representation of features of the distribution of DC
values
A
B
37
Aerial imagery
For the aerial imagery, the red and NIR images for each field were registered to
each other in ENVI by selecting a number of ground control points in each image. For
each field, the red and NIR images were displayed as an RGB with near infrared data
displayed using the red channel and red data displayed in both the green and blue
channels. This resulted in an image that looked similar to a standard false-color NIR
composite image. The advantage of displaying red and NIR data in this manner was that
this helped to visually identifying areas with vegetation and bare soil. All the non
agricultural features were masked using the build mask function in ENVI and excluded
from any analysis. Since a single aerial image contained only one field, the bare soil line
and full canopy point were identified in the scatterplot of a mosaic of several aerial
images. Fig. 3.7.2a and 3.7.2b show the mosaic of six fields (Field Nos. 1, 2, 3, 4, 24 and
26) that was used to estimate the GC on 8 June 2007.
The GC was determined for each field following the same methodology described
for satellite data analysis. For better visualization, the GC map was classified into 10
different classes using the decision tree classification function in ENVI. A different color
was assigned to each class to allow pseudocoloring of the GC maps.
38
Fig. 3.7.2
Example of mosaic of aerial images that is used to determine the soil line and
100% ground cover (GC100) point for evaluating the GC from aerial images
acquired on 8 June 2007. Non-agricultural features were masked (appear as
black in the mosaic) and excluded form any analysis.
A: Mosaic of six aerial images acquired on 8 June 2007 in the Red band.
B: Mosaic of six aerial images acquired on 8 June 2007 in the NIR band.
39
3.8. Ground Cover Modeling
Because the satellite and aerial image data provided only infrequent
measurements of GC, the TAWC version of the Yield Tracker model (Ko et al., 2006 and
2005, Maas et al., 2004, 2003 and 2002; Maas, 2001) was used to simulate daily values
of GC. The model used a weather data file containing the day of the year, average daily
air temperature (oC), average daily photosynthetically active radiation (PAR), and daily
rainfall (mm). Another input file used to run the model was the GC file, which contained
the day of the year and remotely sensed GC data on that day.
3.9. Data Analysis
Ground cover values estimated for each field from the satellite image data were
plotted versus the corresponding values of GC obtained from ground-based field
measurements. Simple linear regression (Ostle and Mensing, 1975, p. 169) was used to
fit a straight line to these pairs of GC values. Student’s t test was then used to test if this
regression line was significantly different from the 1:1 line (Ostle and Mensing, 1975, p.
177). Finally, a paired Student’s t was used to test if the average GC value estimated
from the satellite imagery was significantly different from the average GC value
determined from ground-based field observations (Ostle and Mensing, 1975, p. 120).
The average absolute error (AAE) was calculated between the values of GC estimated
using the satellite image data (GCEST) and the corresponding GC values determined from
the ground-based field observations (GCOBS) using the following formula,
AAE = ∑ | GCEST – GCOBS | / n
[Eq.3.14]
40
in which n is the number of observations (in this study, n = 51). The value of AAE
indicates, on average, how close the estimated and observed GC values were, and
provides an estimate of the overall accuracy of the procedure.
A paired Student’s t test was also used to test if the CWU estimates by the
spectral crop coefficient and regular crop coefficient methods were significantly different
from the actual CWU estimates from the eddy covariance measurements.
41
Chapter IV
Results and Discussion
4.1. Ground Cover (Spectral Crop Coefficient or Ksc) using Perpendicular Vegetation
Index
Satellite Imagery
The 2 dimensional (2D) scatterplots generated by plotting the DC values in the
near-infrared (NIR) spectral band versus the corresponding DC values in the red (RED)
spectral band for the 30 June, 16 July, 1 August and 18 September 2006 Landsat-5 image
acquisitions are presented in Fig. 4.1.1. The distribution of points in these scatterplots
resembled the example presented in Fig. 3.7.1a, so the locations of bare soil line and full
canopy (GC100) could be identified for each distribution by visual inspection. The
locations of bare soil lines and GC100 points determined for each of the distribution are
shown in Fig. 4.1.1. Table 4.1.1 presents the slope (a1) and intercept (a0) of the bare soil
line, DC values in the red (DCFC,RED) and NIR (DCFC,NIR) spectral bands for full canopy,
and the PVI value (PVIFC) calculated using [Eq.3.12] associated with full canopy for the
four Landsat-5 image acquisitions (30 June 2006, 16 July 2006, 1 August 2006 and 18
September 2006).
Results of GC estimated using the Landsat-5 image data on these days for 31
locations in the study area are plotted in Fig. 4.1.2 versus the corresponding ground based
observations of GC. Regression analysis of the data shows that the points tend to lie
along the 1:1 line. The slope and intercept of the least-square linear regression fitted line
is 0.929 and 3.22 respectively. Results of the Studtent’s t-test of the slope and intercept
of this regression indicated that the slope and intercept were not significantly different
42
from 1 (t = -1.546, 49 df, α = 0.05) and 0 (t = 0.107, 49 df, α = 0.05) respectively. Thus,
the regression line through these points is not significantly different from the 1:1 line.
The average value of satellite based estimates of GC was 43.95 percent, while the
average value of ground based observations of GC was 43.82 percent. The Student’s ttest of the pairs of estimated and observed GC values indicates that the average value of
satellite based estimates of GC is not significantly different from the average value of
ground based observations of GC (t = -0.110, 50 df, α = 0.05). The calculated value of
AAE for this data set was 5.76 percent. This suggests that, on average, estimates of GC
determined using this procedure should be within 6 percent of their true values.
The 2D scatterplots generated for all the image acquisitions dates in the study for
2006 and 2007 showed similar pattern in distribution of points as in Fig. 3.7.1a or Fig.
4.1.1. The slope (a1) and intercept (a0) of the bare soil line, DC values in the red
(DCFC,RED) and NIR (DCFC,NIR) spectral bands for full canopy and the associated PVI
values (PVIFC) for full canopy calculated using [Eq.3.12] are summarized in Table 4.1.2.
43
Fig. 4.1.1. Results of plotting the pixel digital count (DC) values in the near-infrared
(NIR) spectral band vs. the corresponding DC values in the red spectral band
for the 30 June 2006, 16 July 2006, 1 August 2006, and 18 September 2006
Landsat-5 image acquisitions. The location of the bare soil line is indicated
for each distribution of points, along with the location of the point
representing full vegetation canopy (“FC”).
44
Table 4.1.1. Slope (a1) and intercept (a0) of the bare soil line, DC values in the red
(DCFC,RED) and NIR (DCFC,NIR) spectral bands for full canopy, and the PVI
value (PVIFC) associated with full canopy for the four Landsat-5 image
acquisitions in Fig. 4.1.1.
Date
Slope
(a1)
Intercept
(a0)
DCFC,RED
DCFC,NIR
PVIFC
30 Jun 2006
1.17
-5.91
25
168
94.0
16 Jul 2006
1.04
5.00
20
166
97.2
1 Aug 2006
1.04
5.00
25
180
103.3
18 Sep 2006
0.98
10.00
21
156
89.8
45
Fig. 4.1.2. Values of ground cover (GC) estimated for the 31 locations in the study using
the satellite image data plotted vs. the corresponding ground-based field
observations of GC. The solid diagonal line represents the 1:1 line, while the
dashed line represents the least-squares linear regression line fit to these
points. The slope of the regression line is 0.929 and the intercept is 3.22, and
the regression line is not significantly different from the 1:1 line.
46
Table 4.1.2. Slope (a1) and intercept (a0) of the bare soil line, DC values in the red
(DCFC,RED) and NIR (DCFC,NIR) spectral bands for full canopy, and the PVI
value (PVIFC) associated with full canopy for all the Landsat-5 and Landsat7 image acquisitions used in this study.
Year
Satellite
Landsat7
2007
Landsat5
Landsat7
2006
Landsat5
Slope
Intercept
(a1)
(a0)
22-Apr
0.56
5
22
163
127.1
24-May
0.53
10
43
148
101.8
27-Jul
0.5
20
48
165
108.2
12-Aug
0.53
12
42
162
112.9
28-Aug
0.53
12
44
152
103.1
13-Sep
0.53
10
43
148
101.8
19-Jul
1
8
26
169
90.8
5-Sep
1
8
24
154
86.0
21-Sep
1
8
21
149
84.9
8-July
0.53
8
56
165
112.5
9-Aug
0.56
4
44
152
107.6
25-Aug
0.53
9
46
159
111.0
10-Sep
0.48
11
41
147
104.9
26-Sep
0.54
6
36
143
103.4
30-Jun
1.1
6
30
175
91.5
16-July
0.9
2
27
173
109.0
1-Aug
0.95
4
28
181
109.0
18-Sep
1.1
4
20
157
88.1
4-Oct
1
3
21
141
82.6
20-Oct
1.1
4
16
128
72.0
DCFC,RED DCFC,NIR
Date
47
PVIFC
As explained above, the procedure to estimate the ground cover of various
agricultural crops using the satellite data appears to be accurate. The accuracy of this
procedure depends upon the identification of the soil line and point corresponding to
100% GC in the distribution of red and NIR pixel DC values. The 2D scatterplots of all
the satellite image data used in this study resembled the theoretical distribution shown in
Fig. 3.7.1a. This was achieved because the area selected to make the 2D scatterplots
were primarily agricultural regions dominated by vegetation and bare soils. The nonagricultural targets (water bodies, clouds, cloud shadows, paved roads and buildings)
were masked out from any analysis. The inclusion of non-agricultural targets can
confound the identification of the bare soil line and point corresponding to 100% GC.
This can be explained by Fig. 4.1.3a, which shows the distribution of points in the RED
and NIR band for the 30 June Lansat-5 image. The same portion of the Landsat-5 image
used to construct this distribution was used to construct the distribution in Fig. 4.1.1,
except that in this case the areas containing clouds and cloud shadows were not masked.
In Fig. 4.1.3a, it is difficult to identify the bare soil line visually since the edge of the
distribution is now dominated by pixels containing clouds and cloud shadows.
Similarly, Fig. 4.1.3b shows the distribution of points corresponding to the 18 September
Landsat-5 image shown in Fig. 4.1.1, except that in this case the pixels containing lakes
were not masked out before constructing the distribution. Again, the idealized form of
the distribution expected from Fig. 4.1.3b is confounded by the points corresponding to
pixels affected by the lakes. Therefore, proper screening of the medium-resolution
multispectral satellite imagery is important to remove non-agricultural targets and
facilitate identification of the features necessary for applying this procedure.
48
The masking out of non-agricultural features does not affect the identification of
the bare soil line as there is always some bare soil surfaces in most agricultural regions.
This procedure also needs a 100% GC point. In the region where this study was
conducted, it was possible to identify points of vegetation with full GC in the satellite
image at practically any time during the growing season. Even in the early spring, winter
wheat canopies are sufficiently dense to allow this determination. However, it is
conceivable that in some agricultural regions there might be periods at the start or end of
the growing season for which there is no vegetation with a density approaching full
canopy, thereby making identification of the full canopy point difficult. In such cases,
one could use an average value of PVIFC determined at other times during the growing
season in calculating GC. For example, the average of the values of PVIFC presented in
Table 4.1.1 is 96.1. Using this average value in the calculations of GC for each of the
four acquisition dates in this study results in an average GC of 48.06 percent, which is
less than 5 percent different from the corresponding value (GCEST = 43.95 percent)
obtained using the values of PVIFC from Table 4.1.1 that are specific to each acquisition
date. Therefore, in the absence of adequate data to identify the point corresponding to
full canopy, it may be feasible to use a previously determined average value of PVIFC in
the calculations of GC.
49
Fig. 4.1.3. Results of plotting the pixel digital count (DC) values in the near-infrared
(NIR) spectral band vs. the corresponding DC values in the red spectral band
for the (A) 30 June and (B) 18 September Landsat-5 image acquisitions
without masking the image data to remove non-agricultural targets such as
clouds, cloud shadows, and lakes.
50
Aerial Imagery
Fig. 4.1.4 shows the distribution of points obtained by plotting the DC values in
the NIR spectral band versus the corresponding DC values in the red spectral band for an
image mosaic of six fields in the study (Nos. 1, 2, 3, 4, 24 and 26) acquired using
TTAMRSS on 8 June 2007. The characteristic shape of this distribution also resembled
the example presented in Fig. 3.7.1a. The locations of bare soil line and full canopy
(GC100) identified for this distribution by visual inspection are shown in Fig. 4.1.4. The
equation of the bare soil line obtained for this distribution has the following equation.
NIR = 66 + 0.8 RED
[4.1]
The DC values in the red (DCFC,RED) and NIR (DCFC,NIR) spectral bands for full canopy
location determined in Fig. 4.1.4 are 1476 and 3069 respectively, and the associated PVI
value calculated using [Eq.3.12] is 1423. Fig. 4.1.5 show examples of GC maps
generated after classification for two fields in the study using TTAMRSS image (Field
Nos. 24 and 26). Although the PVI method can be used to produce a spatial GC map as
in Fig. 4.1.4, an average value of GC for the entire field was used in this study.
51
Fig. 4.1.4. Results of plotting the pixel digital count (DC) values in the near-infrared
(NIR) spectral band vs. the corresponding DC values in the red spectral band
for a mosaic of six aerial images acquired on 8 June 2007 using the Texas
Tech Airborne Multispectral Remote Sensing System (TTAMRSS). The
location of the bare soil line is indicated by the straight line, along with the
location of the point representing full vegetation canopy.
8 June 2007
TTAMRSS NIR DC
Full Canopy
Bare Soil Line
TTAMRSS Red DC
52
Fig. 4.1.5. Ground cover (GC) maps produced by the Perpendicular Vegetation Index
(PVI) method using the aerial images acquired on 8 June 2007 using the
Texas Tech Airborne Multispectral Remote Sensing System (TTAMRSS) for
Field No. 24 and 26. Field No. 24 is a center-pivot corn field. Field No. 26
has half the area planted to corn and the other half is bare soil. Each color
indicates a range of percent GC as indicated in the legend.
53
The estimation of GC using aerial image data follows the same procedure as in
the case of satellite data, but has several aspects to be addressed in more detail. Since the
accuracy of this method depends on the identification of soil line and 100% GC point in
the distribution of red and NIR pixel DC values, one aerial image may not be sufficient to
extract all the information needed to estimate crop GC. As described in the Materials and
Methods section, this difficulty is overcome by mosaicing several aerial images. This
produces an “image” that contains many different surfaces, usually including some bare
soil and full crop canopy. Thus, the mosaiced aerial images are similar to a portion of a
Landsat image. Again, masking out of non-agricultural features is very important for the
identification of soil line and 100% GC point. Unlike satellite data, which have medium
spatial resolution (30 x 30 m), aerial images taken in this study have high surface
resolution (2 x 2 m). Hence, the aerial images contain a lot more detail than the satellite
data. The areas needed to be masked in aerial image data are buildings, equipment in the
field, paved roads, playa lakes, other water bodies, clouds, and cloud shadows. Masking
out of these non-agricultural features and making a composite of several aerial images
offers a way to build a distribution of points that are closer to the theoretical distribution
(Fig. 3.7.1a).
54
4.2. Spectral Crop Coefficient (Ksc)
Table 4.2.1 and Table 4.2.2 summarize the GC estimated for the fields in the
study using TTAMRSS, Landsat-5 and Landsat-7 images acquired in 2007 and 2006
respectively. Because values of GC were needed for each day of the growing season, and
the remote sensing values were only for certain days within the growing season, daily
values of GC were simulated using the TAWC version of Yield Tracker model (Ko et al.,
2006 and 2005, Maas et al., 2004, 2003 and 2002, and Maas, 2001). The results of the
simulation for few fields are presented in Fig. 4.2.1 through 4.2.10. The simulation
resulted in a continuous curve of GC for the entire growing season and is called the
spectral crop coefficient curve (Ksc curve). The values of Ksc can range from 0 to 1
depending on how much GC the crop is attaining during the growing season. The Ksc
curves illustrate the change in GC values, starting at a low value in the beginning of the
growing and increases to a maximum value in the mid- growing season when the crop
attains the maximum GC. The value of Ksc decrease during the later part of the growing
season as the canopy starts senescence. The Ksc curves of those crops used as silage
(corn and forage sorghum) ended abruptly during the mid- growing season at the time of
harvest (Fig. 4.2.1A and 4.2.2A, Fig. 4.2.8 and 4.2.9). For grain crops (Fig. 4.2.1B and
4.2.2B), the ksc curve followed the general crop growth pattern. For indeterminate cotton
crop (Fig.4.2.3 through 4.2.6), the ksc curves look similar to the growth pattern of
determinate crops as the cotton was sprayed with defoliants two to three weeks prior to
harvest.
55
Table 4.2.1. Percent Ground Cover (GC) estimated for the fields in the study by the Perpendicular Vegetation Index (PVI) method
using TTAMRSS, Landsat-5 and Landsat-7 images in 2007.
Field
No
1-1
1-2
2
3-1
6
11-1
12-1
12-2
15-1
15-3
15-4
18-2
20-1
20-2
24 (1 & 2)
26-1
26-2
27
Date of image acquisitions
22 Apr
..
..
..
..
..
..
..
..
..
..
..
..
..
..
1
..
..
3
08 Jun
7
6
9
9
4.4
5
4
8
5
13
7
..
29
..
67
11
60
64
19 July
28
18
39
47
*
*
*
*
32
68
32
74
72
25
74
85
74
76
27 July
*
*
75
59
*
*
*
*
43
68
43
*
70
64
71
98
72
75
12 Aug
69
72
100
62
72
37
59
46
60
48
46
69
67
83
64
100
64
28 Aug
85
84
105
82
93
51
62
36
69
40
55
60
05 Sep
84
83
99
83
91
52
63
29
72
39
54
55
13 Sep
79
71
91
79
84
48
61
21
70
33
53
46
21 Sep
77
63
85
78
82
51
62
18
70
33
54
43
77
0
64
14
72
0
55
8
**
97
29
88
44
85
18
76
31
**
.. Indicates the fields were not planted, * indicates the field was cloudy in the image, and ** indicates the field was harvested.
56
Table 4.2.2. Percent Ground Cover (GC) estimated for the fields in the study by the Perpendicular Vegetation Index (PVI) method
using Landsat-5 and Landsat-7 images for all the fields in the study in 2006.
Field
No
Date of image acquisitions
30 Jun
08 Jul
16 Jul
01 Aug
09 Aug
25 Aug
10 Sep
1-1
28
46
60
*
73
70
69
1-2
28
45
59
*
85
75
73
2
35
47
55
75
89
75
69
3-1
29
37
40
*
49
48
49
4-2
89
96
107
..
54
87
97
6
11
32
52
60
67
63
56
12-2
9
19
27
18
16
14
21
13-1
13
25
34
*
21
16
19
15-1
*
25
50
*
55
57
61
15-3
*
38
50
*
64
54
58
15-4
*
40
45
*
43
24
18
19-3
*
34
60
66
67
51
44
20-1
*
80
79
64
20-2
*
63
83
91
99
78
76
22-2
82
83
83
*
51
19
7
24-1
85
84
83
*
26-1
*
72
74
64
55
30
15
* indicates the field was cloudy in the image, and ** indicates the field was harvested.
57
18 Sep
26 Sep
4 Oct
20 Oct
63
62
65
44
91
53
18
17
58
54
19
40
**
65
4
**
8
54
50
52
36
..
40
25
18
54
46
20
40
45
44
34
22
39
31
15
49
41
25
33
**
**
3
**
**
12
23
9
28
**
23
13
64
9
58
4
**
**
10
9
**
Fig 4.2.1.
Spectral crop coefficient curves (Ksc) for (A) corn for silage (Field Nos. 20-1
and 27) and (B) corn for grain (Field No. 24 and 26-2) fields for the 2007
growing season. The crop coefficient curve was produced by simulating the
daily vales of ground cover (GC) for the entire growing season using the
remotely sensed GC and weather data. The crop model used for simulation
was the TAWC version of Yield Tracker model.
A. Corn for silage
1
0.9
Ground Cover (Ksc )
0.8
Series1
Field No. 27
(Drip)
Field No. 20-1
Series2
(Drip)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
240
260
280
220
240
260
280
Day of the Year
B. Corn for grain
1
0.9
Ground Cover (Ksc )
0.8
Series1
Field No. 24
(Drip)
Field No. 26-2
Series2
(Drip)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
Day of the Year
58
Fig. 4.2.2. Spectral crop coefficient curves (Ksc) for (A) corn for silage (Field No. 20-2)
and (B) corn for grain (Field No. 24-1 and 26-1) fields for the 2006 growing
season. The spectral crop coefficient curve was produced by simulating the
daily vales of ground cover (GC) for the entire growing season using the
remotely sensed GC and weather data. The crop model used for simulation
was the TAWC version of Yield Tracker model.
A. Corn for silage
1
Field No. 20-2
Series1
(Drip)
0.9
Ground Cover (Ksc )
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
80
100 120 140 160 180 200 220 240 260 280 300
Day of the Year
B. Corn for grain
1
Series1
Field No. 24-1
(Drip)
Field No. 26-1
Series2
(Drip)
0.9
Ground Cover (Ksc )
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
80
100
120
140
160
180
200
Day of the Year
59
220
240
260
280
Fig. 4.2.3. Spectral crop coefficient curves (Ksc) for drip irrigated cotton (Field Nos. 1-1,
1-2, and 2) fields for the (A) 2007 and (B) 2006 growing seasons. The
spectral crop coefficient curve was produced by simulating the daily vales of
ground cover (GC) for the entire growing season using the remotely sensed
GC and weather data. The crop model used for simulation was the TAWC
version of Yield Tracker model.
A. 2007
1.1
1
Ground Cover (Ksc )
0.9
0.8
Series1
Field
No. 1-1
Field
No. 1-2
Series2
(Drip)
Series3
Field
No. 2
(Drip)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
240
260
280
280
300
Day of the Year
B. 2006
0.9
0.8
Ground Cover (Ksc )
0.7
0.6
Series1
Field
No. 1-1
Field
No. 1-2
Series2
(Drip)
Series3
Field
No. 2
(Drip)
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
Day of the Year
60
240
260
Fig. 4.2.4. Spectral crop coefficient curves (Ksc) for center-pivot irrigated cotton (Field
Nos. 3-1 and 6) fields for the (A) 2007 and (B) 2006 growing seasons. The
spectral crop coefficient curve was produced by simulating the daily vales of
ground cover (GC) for the entire growing season using the remotely sensed
GC and weather data. The crop model used for simulation was the TAWC
version of Yield Tracker model.
A. 2007
Ground Cover (Ksc )
1
0.9
Field
No. 3-1
Series1
0.8
Series2
Field No. 6
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
240
260
280
300
240
260
280
300
Day of the Year
B. 2006
0.7
Series1
Field
No. 3-1
0.6
Ground Cover (Ksc )
Field
No. 6
Series2
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
Day of the Year
61
Fig. 4.2.5. Spectral crop coefficient curves (Ksc) for furrow irrigated cotton fields for the
(A) 2007 (Field Nos. 15-1 and 15-4) and (B) 2006 (Field Nos. 15-1 and 15-3)
growing seasons. The spectral crop coefficient curve was produced by
simulating the daily vales of ground cover (GC) for the entire growing season
using the remotely sensed GC and weather data. The crop model used for
simulation was the TAWC version of Yield Tracker model.
A. 2007
Ground Cover (Ksc )
0.9
0.8
Series1
Field
No. 15-1
0.7
Field
No. 15-4
Series2
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
240
260
280
300
240
260
280
300
Day of the Year
B. 2006
Ground Cover (Ksc )
0.9
0.8
Series1
Field
No. 15-1
0.7
Field
No. 15-3
Series2
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
Day of the Year
62
Fig. 4.2.6. Spectral crop coefficient curves (Ksc) for dryland cotton fields for the (A)
2007 (Field No. 12-1) and (B) 2006 (Field No. 13-1) growing seasons. The
spectral crop coefficient curve was produced by simulating the daily vales of
ground cover (GC) for the entire growing season using the remotely sensed
GC and weather data. The crop model used for simulation was the TAWC
version of Yield Tracker model.
A. 2007
0.7
Ground Cover (Ksc )
0.6
Field
No. 12-1
Series1
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
240
260
280
300
240
260
280
300
Day of the Year
B. 2006
0.4
Ground Cover (Ksc )
Field
No. 13-1
Series1
0.3
0.2
0.1
0
100
120
140
160
180
200
220
Day of the Year
63
Fig. 4.2.7. Spectral crop coefficient curves (Ksc) for forage sorghum (Field No. 20-2) for
the 2007 growing season. The spectral crop coefficient curve was produced
by simulating the daily vales of ground cover (GC) for the entire growing
season using the remotely sensed GC and weather data. The crop model used
for simulation was the TAWC version of Yield Tracker model.
1
0.9
Field No. 20-2
Series1
Ground Cover (Ksc )
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
Day of the Year
64
240
260
280
300
Fig. 4.2.8. Spectral crop coefficient curves (Ksc) for grain sorghum fields for (A) 2007
(Field Nos. 15-3 and 18-2) and (B) 2006 (Field No. 15-4). The spectral crop
coefficient curve was produced by simulating the daily vales of ground cover
(GC) for the entire growing season using the remotely sensed GC and
weather data. The crop model used for simulation was the TAWC version of
Yield Tracker model.
A. 2007
0.8
Field No. 15-3
Series1
Series2
Field No. 18-2
Ground Cover (Ksc )
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
80
100 120 140 160 180 200 220 240 260 280 300
Day of the Year
B. 2006
0.5
Field No. 15-4
Series1
Ground Cover (Ksc )
0.4
0.3
0.2
0.1
0
120
140
160
180
200
Day of the Year
65
220
240
260
Fig. 4.2.9. Spectral crop coefficient curves (Ksc) for forage sorghum fields (A) Field No.
20-1 and (B) Field No. 4-2 for the 2006 growing season. Field No. 4-2 was
harvested twice. The spectral crop coefficient curve was produced by
simulating the daily vales of ground cover (GC) for the entire growing season
using the remotely sensed GC and weather data. The crop model used for
simulation was the TAWC version of Yield Tracker model.
A. Field No. 20-1
1.1
1
Field No. 20-1
Series1
Ground Cover (Ksc )
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
140
160
180
200
220
240
Day of the Year
B. Field No. 4-2
Ground Cover (Ksc )
1.2
1.1
Field No. 4-2 (1)
Series2
1
Field No. 4-2 (2)
Series3
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
Day of the Year
66
240
260
280
300
Fig. 4.2.10. Spectral crop coefficient curves (Ksc) for pearl millet for the (A) 2007 (Field
No. 26-1) and (B) 2006 (Field No. 19-3) growing seasons. The spectral
crop coefficient curve was produced by simulating the daily vales of ground
cover (GC) for the entire growing season using the remotely sensed GC and
weather data. The crop model used for simulation was the TAWC version
of Yield Tracker model.
A: 2007
1
0.9
Field
SeriesNo. 26-1
1
Ground Cover (Ksc )
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
240
260
280
300
240
260
280
300
Day of the Year
B. 2006
0.8
Ground Cover (Ksc )
0.7
Field
SeriesNo. 19-3
1
0.6
0.5
0.4
0.3
0.2
0.1
0
100
120
140
160
180
200
220
Day of the Year
67
Fig. 4.2.11 through 4.2.13 illustrates the regular crop coefficient curves (Kc) for
corn, cotton and sorghum developed for the Texas High Plains. These crop coefficients
were developed from lysimeter studies at Bushland and included field observations over
several years (Mareck et al., 2006). Comparison of the spectral crop coefficient and
regular crop coefficient values for the same crop and field show that they are different.
Fig. 4.2.14 presents an example comparing the spectral crop coefficient and regular crop
coefficient curves for two corn fields, Field Nos. 24 and 27, in 2007. Field No. 24 was
center-pivot irrigated and Field No. 27 was drip irrigated. The Ksc curve follows the realtime growth pattern of corn observed in these fields from the remote sensing data. The
Kc values represent average crop growth for this region. The Kc curve follow the same
pattern for both fields and the shift in Kc values is only due to the difference in planting
dates. The Kc and Ksc curves for Field No. 27 end earlier than Field No. 24 because the
crop was harvested for silage.
Fig. 4.2.15 compares the Ksc and Kc values for two cotton fields that had the same
planting date. The Ksc values are different for both fields as noted previously in the case
of corn fields. The Kc values are the same for both fields. Fig 4.2.16 compares the Ksc
and Kc values for two grain sorghum fields (Field Nos. 12-2 and 15-3) in 2007. The Ksc
values are smaller than the Kc values for both fields. The Ksc values for Field No. 12.2 is
smaller than the Ksc values for Field No. 15-3 as this field is dryland and had a poor crop
stand compared to the furrow irrigated grain sorghum field (Field No. 15-3).
68
Fig. 4.2.11. Regular crop coefficient (Kc) curve for corn developed for the Texas High
Plains from lysimeter studies at Bushland, TX.
1.4
1.2
1.3
1.25
1.2
1.2
1.25
Crop coefficient (Kc )
1.15
1
1
1
0.9
0.85
0.8
0.7
0.7
0.6
0.45
0.4
0.35
0.25
0.2
0
0
20
40
60
80
100
120
140
Days after planting
Fig. 4.2.12. Regular crop coefficient (Kc) curve for cotton developed for the Texas High
Plains from lysimeter studies at Bushland, TX.
1.4
Crop coefficient (Kc )
1.2
1.1
1.1
1
0.83
0.8
0.6
0.2
0.44
0.44
0.4
0.22
0.07
0.1
0
0
20
40
60
80
100
Days after planting
69
120
140
160
Fig. 4.2.13.
Regular crop coefficient (Kc) curve for grain sorghum developed for the
Texas High Plains from lysimeter studies at Bushland, TX.
1.4
1.2
1.1
Crop coefficient (Kc )
1
0.95
1
0.95
0.8
0.9
0.8
0.85
0.7
0.6
0.55
0.4
0.6
0.4
0.4
0.2
0
0
20
40
60
80
100
Days after planting
70
120
140
160
Fig. 4.2.14. Comparison of the spectral crop coefficient curve (Ksc) generated using
remotely sensed ground cover (GC) and the regular crop coefficient curve
(Kc) recommended for the Texas High Plains. Examples are presented for
two corn fields in 2007: (A) center-pivot irrigated corn (Field No. 24) and
(B) drip irrigated corn (Field No. 27).
A. Field No. 24 (Center-pivot)
1.4
Kc
Series1
K sc
Series2
Crop Coefficient
1.2
1
0.8
0.6
0.4
0.2
0
80
100
120
140
160
180
200
220
240
260
280
Day of the Year
B. Field No. 27 (Drip)
1.4
Kc
Series1
K sc
Series2
Crop Coefficient
1.2
1
0.8
0.6
0.4
0.2
0
80
100
120
140
160
180
Day of the Year
71
200
220
240
Fig. 4.2.15. Comparison of the spectral crop coefficient curve (Ksc) generated using
remotely sensed ground cover (GC) and the regular crop coefficient curve
(Kc) recommended for the Texas High Plains. Examples are presented for
two cotton fields in 2007. The Kc curve is the same for both fields.
1.2
Crop Coefficient
1
0.8
Series1
Kc
K sc (Field No. 15-1)
Series2
K sc (Field No. 15-4)
Series3
0.6
0.4
0.2
0
120
140
160
180
200
220
Day of the Year
72
240
260
280
Fig. 4.2.16. Comparison of the spectral crop coefficient curve (Ksc) generated using
remotely sensed ground cover (GC) and the regular crop coefficient curve
(Kc) recommended for the Texas High Plains. Examples are presented for
two grain sorghum fields: (A) dryland (Field No. 12-2) and (B) furrow
irrigated (Field No. 15-3)
A. Field No. 12-2 (Dryland)
1.2
Series1
Kc
Crop Coefficient
1
K sc
Series2
0.8
0.6
0.4
0.2
0
120
140
160
180
200
220
240
260
280
Day of the Year
B. Field No. 15-3 (Furrow)
1.2
Series1
Kc
Crop Coefficient
1.0
K sc
Series2
0.8
0.6
0.4
0.2
0.0
100
120
140
160
180
200
Day of the Year
73
220
240
260
280
4.3. Potential Evapotranspiration
PET calculated for each day of the growing season using [Eq.3.11] are presented
in Fig. 4.3.1 through 4.3.5 for selected fields in the study. Also presented in these figures
are the daily values of reference crop evapotranspiration (ET0) calculated using the FAO56 guidelines (Allen, 2005). Germination is assumed to be one week after planting and
from planting to germination PET is assumed equal to be zero. The daily fluctuations in
PET and ET0 curves are due to the difference in daily weather parameters. For wellwatered crops, daily values of PET are greater than ET0 as a function of crop height since
height of the crop used to determine the aerodynamic resistance term is greater than the
height of the reference crop (0.12 m). The increase in daily values of PET for corn (Fig.
4.3.1) during the growing season is higher than the increasing trend seen for cotton (Fig.
4.3.2). Daily values of PET and ET0 were summed to get seasonal values for all the
crops. The seasonal PET and ET0 for all the fields in the study for the 2006 and 2007
growing seasons are presented in Table 4.3.1 and Table 4.3.2 respectively.
74
Fig. 4.3.1. Comparison of potential evapotranspiration (PET) and reference evapotranspiration (ET0) calculated using the FAO-56
guidelines for a center-pivot irrigated corn field (Field No. 24) in 2007.
15
ET (mm/day)
PET
ETo
10
5
0
80
100
120
140
160
180
Day of the Year
75
200
220
240
260
280
Fig. 4.3.2. Comparison of potential evapotranspiration (PET) and reference evapotranspiration (ET0) calculated using the FAO-56
guidelines for a drip irrigated cotton field (Field No. 2) in 2007.
15
ET (mm/day)
PET
ETo
10
5
0
120
140
160
180
200
Day of the Year
76
220
240
260
280
Fig. 4.3.3. Comparison of potential evapotranspiration (PET) and reference evapotranspiration (ET0) calculated using the FAO-56
guidelines for a furrow irrigated grain sorghum field (Field No. 15-3) in 2007.
15
ET (mm/day)
PET
ETo
10
5
0
100
120
140
160
180
200
Day of the Year
77
220
240
260
280
Fig. 4.3.4.
Comparison of potential evapotranspiration (PET) and reference evapotranspiration (ET0) calculated using the FAO-56
guidelines for a center-pivot irrigated forage sorghum field (Field No. 20-2) in 2007.
15
ET (mm/day)
PET
ETo
10
5
0
160
180
200
220
Day of the Year
78
240
260
280
Fig. 4.3.5.
Comparison of potential evapotranspiration (PET) and reference evapotranspiration (ET0) calculated using the FAO-56
guidelines of a center-pivot irrigated pearl millet field (Field No. 26-1) in 2007.
15
ET (mm/day)
PET
ETo
10
5
0
160
180
200
220
Day of the Year
79
240
260
280
Table 4.3.1. Seasonal PET and ET0 values for all the fields in the study in 2006. The
seasonal PET and ET0 values for each field are computed by summing the
daily values of PET and ET0 for all the days in the growing season.
Crop
Field No
PET (mm)
PET (in)
ET0 (mm)
ET0 (in)
20-2
1209
48
845
33
22-2
1545
61
1040
41
24-1
1294
51
894
35
26-1
1607
63
1059
42
1-1
1098
43
1011
40
1-2
1098
43
1011
40
2
1028
40
962
38
3-1
1078
42
968
38
3-2
1078
42
968
38
6
986
39
917
36
12-2
935
37
909
36
13-1
1008
40
955
38
15-1
861
34
962
38
15-3
861
34
962
38
15-4
895
35
805
32
4-2(1)
766
30
622
24
4-2(2)
404
16
332
13
20-1
252
10
405
16
19-3
1248
49
1050
41
Corn
Cotton
Grain
Sorghum
Forage
Sorghum
Pearl Millet
80
Table 4.3.2. Seasonal PET and ET0 values for all the fields in the study in 2007. The
seasonal PET and ET0 values for each field are computed by summing the
daily values of PET and ET0 for all the days in the growing season.
Crop
Field No
PET (mm)
PET (in)
ET0 (mm)
ET0 (in)
20-1
849
33
785
31
24
1175
46
881
35
26-2
1086
43
799
31
27
829
33
659
26
1-1
914
36
846
33
1-2
914
36
846
33
2
910
36
838
33
3-2
888
35
789
31
6
845
33
778
31
11-1
865
34
778
31
12-1
839
33
778
31
15-1
860
34
793
31
15-4
861
34
793
31
12-2
736
29
717
28
15-3
902
36
821
32
18-2
851
33
785
31
Forage
Sorghum
20-2
754
29
615
24
Pearl Millet
26-1
724
29
654
26
Corn
Cotton
Grain
Sorghum
81
4.4. Crop Water Use
2007 growing season
Corn
In 2007 CWU was estimated for four corn fields in the study area. Three of the
fields were center-pivot irrigated and one was drip irrigated. Values of the daily CWU
estimated by the spectral crop coefficient method using [Eq.1.3] are presented in Fig.
4.4.1 through 4.4.4. Also presented in these figures are the CWU estimated by the
regular crop coefficient method using [Eq.1.1]. During the early part of the growing
season, CWU estimated by the Kc method was higher than the CWU estimated by the Ksc
method. The difference in CWU estimated by these two methods decreased during the
mid- and late growing seasons. For Field No. 20-1 (Fig. 4.4.1), the Ksc-based CWU was
lower than the Kc-based CWU for the entire growing season. For Field Nos. 26-2 (Fig.
4.4.3) and 27 (Fig. 4.4.4), CWU estimates by both methods were approximately equal
during the mid- and late growing seasons. For Field No. 24 (Fig. 4.4.2), the Ksc-based
CWU was higher than the Kc-based CWU during the late growing season.
The seasonal CWU estimated by Ksc and Kc methods for the corn fields in 2007
are presented in Fig. 4.4.5. As expected, the seasonal CWU for corn harvested for silage
(Field Nos. 20-1 and 27) is smaller than the seasonal CWU of corn harvested for grain
(Field Nos. 24 and 26-2). The average seasonal CWU for corn harvested for grain is 555
mm by the Ksc method and 723 mm by the Kc method. For corn harvested for silage, the
average seasonal CWU by Ksc method is 473 mm and by Kc method is 669 mm.
82
Fig. 4.4.1. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 20-1 (center-pivot irrigated corn). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
80
100
120
140
160
180
Day of the Year
83
200
220
240
260
Fig. 4.4.2. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 24 (center-pivot irrigated corn). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
80
100
120
140
160
180
Day of the Year
84
200
220
240
260
280
Fig. 4.4.3. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 26-2 (center-pivot irrigated corn). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
80
100
120
140
160
180
Day of the Year
85
200
220
240
260
280
Fig. 4.4.4. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 27 (drip irrigated corn). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
80
100
120
140
160
Day of the Year
86
180
200
220
240
Fig. 4.4.5. Seasonal Crop Water Use (CWU) in mm estimated by the spectral crop
coefficient (Ksc ) and regular crop coefficient (Kc ) methods for corn fields in
2007. Seasonal CWU is calculated by summing the daily values of CWU.
Field Nos. 20-1 and 27 are harvested for silage. Field Nos. 24 and 26-2 are
harvested for grain.
1000
Seasonal CWU (mm)
K sc
Ksc
800
648
560
549
600
Kc
Kc
724
722
689
477
468
400
200
0
20-1
24
26-2
Field Number
87
27
In 2007, eddy covariance measurements were made for two corn fields in the
study (Field No. 24 and Field No. 20-1). The eddy covariance measurements (corrected
for soil evaporation) are considered as the actual values of CWU. Although eddy
covariance measurements were collected for 2- to 3-week time periods from each field,
the measurements were not used for those days when the wind direction was not from the
south. Fig. 4.4.6 and Fig. 4.4.7 presents the CWU estimated by the Ksc, Kc, and eddy
covariance methods for Field No. 20-1 and Field No. 24 respectively.
Results of the Student’s t-test comparing the CWU by the Ksc method and eddy
covariance measurements for Field No. 20-1 indicate that the Ksc-based estimates are not
significantly different from the actual measurements of CWU (t = -0.137, 5 df, α = 0.05,
p = 0.896). Since eddy covariance measurements include both transpiration and soil
evaporation, a value of 1 mm/day was added to the Ksc-based CWU estimates to account
for soil evaporation. This value was based on prior observations of soil evaporation in a
center-pivot corn field. The results indicate that the CWU estimates by the spectral crop
coefficient method are the same as the actual measurements of CWU. The CWU
estimates by the regular crop coefficient method were compared against the actual field
measurements of CWU from the eddy covariance method. The Student’s t test of the
pairs of observations of CWU by these two methods indicates that the CWU estimates by
these methods were significantly different (t = -3.478, 5 df, α = 0.05, p = 0.018). This
suggests that the estimates of CWU by the Kc method are different from the actual field
measurements of CWU for this field.
Results of the Student’s t test comparing the CWU by the Ksc method and eddy
covariance measurements for Field No. 24 indicate that the Ksc-based estimates are not
88
significantly different from the actual measurements of CWU (t = 2.169, 4 df, α = 0.05, p
= 0.096). For Field No. 24, the measurements were made on days when the field had not
been irrigated. Hence, the soil evaporation was considered negligible. The CWU
estimates by the regular crop coefficient method were compared against the actual field
measurements of CWU from the eddy covariance method. The Student’s t test of the
pairs of observations of CWU by these two methods indicates that the CWU estimates by
these methods were significantly different (t = -6.700, 4 df, α = 0.05, p = 0.003). The
results indicate that the CWU estimates by the spectral crop coefficient method are the
same as the actual measurements of CWU and the Kc- based CWU are different from the
actual field measurements of CWU for this field.
89
Fig. 4.4.6. Daily Crop Water Use (CWU) estimated by the spectral crop coefficient
(Ksc), regular crop coefficient (Kc), and eddy covariance (EC) methods for
Field No. 20-1 in 2007.
12
Daily CWU (mm)
K sc
Ksc
10
Kc
Kc
8
EC
EC
6
4
2
0
200
201
202
203
Day of the Year
90
204
205
Fig. 4.4.7. Daily Crop Water Use (CWU) estimated by the spectral crop coefficient
(Ksc), regular crop coefficient (Kc), and eddy covariance (EC) methods for
Field No. 24 in 2007.
12
Daily CWU (mm)
K sc
Ksc
10
Kc
Kc
8
EC
EC
6
4
2
0
172
173
176
Day of the Year
91
177
189
Cotton
Nine cotton fields under different irrigation management systems were selected
for estimating CWU in 2007. Among these nine fields, three fields were drip irrigated
(Field Nos. 1-1, 1-2 and 2) and three were furrow irrigated (Field Nos. 11-1, 15-1 and 154). Two fields were center-pivot irrigated (Field Nos.3-2 and 6), and one was dryland
(Field No. 12-1). Examples of values of the daily CWU estimated by the spectral crop
coefficient method using [Eq.1.3] and the regular crop coefficient method using [Eq.1.1]
are presented in Fig. 4.4.8 through 4.4.13. As noticed in the case for corn, during the
early part of the growing season, CWU estimated by the Kc method was higher than the
CWU estimated by the Ksc method. During the late growing season, the CWU estimated
by the Ksc method was higher than the CWU estimates by the Kc method. During the
mid- season, for center-pivot and drip irrigated fields the difference in estimates of CWU
by these methods was small. For the furrow and dryland fields, the Kc-based CWU was
higher than the Ksc-based estimates of CWU. For fields that had the same planting dates
(Field Nos. 1-1 and 1-2), the estimates of the CWU by the Kc method are the same
irrespective of the particular growing conditions. Fields such as 1-1 and 1-2 are different
sections of the same field. The daily estimates of CWU by the Kc method are the same
for these fields, but Ksc method provided CWU values that were specific to each section
of the field. The Ksc-based CWU estimates are different for these fields for the same day
because the Ksc is site-specific.
92
Fig. 4.4.8. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 1-1 (drip irrigated cotton). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
93
240
260
280
300
Fig. 4.4.9. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 1-2 (drip irrigated cotton). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
94
240
260
280
300
Fig. 4.4.10. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 2 (drip irrigated cotton). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
95
240
260
280
300
Fig. 4.4.11. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 6 (center-pivot irrigated cotton). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
96
240
260
280
300
Fig. 4.4.12. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 11-1 (furrow irrigated cotton). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
97
240
260
280
300
Fig. 4.4.13. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 12-1 (dryland cotton). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
98
240
260
280
300
Fig. 4.4.14 presents the seasonal CWU for all the cotton fields in the study in
2007. In 2007, the seasonal of CWU for the drip irrigated fields (Field Nos. 1-1, 1-2, and
2) was approximately the same for all the fields by Kc method (479 mm average). The
seasonal CWU of the three drip fields were different by the Ksc method. The highest
CWU was calculated for Field No. 2 (490 mm). The average seasonal CWU for the drip
irrigated fields was 414 mm by the Ksc method. For the furrow and center-pivot irrigated
fields, the seasonal CWU by the Kc method was approximately the same (average 455
mm) for both. The Ksc-based CWU was approximately the same for both center-pivot
irrigated fields (average 400 mm). Furrow irrigated fields showed large variations in
seasonal CWU by the Ksc method. The average seasonal CWU for furrow irrigated fields
was 266 mm. The seasonal CWU of the dryland field (272 mm) was higher than two
furrow irrigated fields (11-1 and 15-4).
99
Fig.4.4.14. Seasonal Crop Water Use (CWU) in mm estimated by the spectral crop coefficient (Ksc ) and regular crop coefficient (Kc )
methods for cotton fields in 2007. Seasonal CWU is calculated by summing the daily values of CWU. Field Nos. 1-1, 1-2,
and 2 are drip irrigated. Field Nos. 3-1 and 6 are center-pivot irrigated. Field Nos. 11-1, 15-1, and 15-4 are furrow
irrigated. Field No. 12-1 is dryland.
600
Seasonal CWU (mm)
Ks
Ksc
Kc
Kc
400
200
0
1-1
1-2
2
3-1
6
Field Number
100
11-1
15-1
15-4
12-1
In 2007, eddy covariance measurements were made for the dryland cotton field,
Field No. 12-1. Fig. 4.4.15 summarizes the CWU estimated by the Ksc, Kc, and eddy
covariance methods for this field. The eddy covariance measurements were not used for
those days when the wind direction was not from the south. Results of the Student’s t test
comparing the CWU by the Ksc method and eddy covariance measurements for Field No.
12-1 indicate that the Ksc-based estimates are not significantly different from the actual
measurements of CWU (t = -2.181, 9 df, α = 0.05, p = 0.06). The results indicate that the
CWU estimates by the spectral crop coefficient method are the same as the actual
measurements of CWU. The CWU estimates by the regular crop coefficient method
were compared against the actual field measurements of CWU from the eddy covariance
method. The Student’s t test of the pairs of observations of CWU by these two methods
indicates that the CWU estimates by these methods are not significantly different (t = 0.750, 9 df, α = 0.05, p = 0.472). This suggests that the estimates of CWU by the Kc
method are also same as the actual field measurements of CWU.
101
Fig. 4.4.15. Daily Crop Water Use (CWU) estimated by the spectral crop coefficient (Ksc), regular crop coefficient (Kc), and eddy
covariance (EC) methods for Field No. 12-1 (dryland cotton) in 2007.
6
K
sc
Ksc
Daily CWU (mm)
Kc
Kc
EC
EC
4
2
0
259
260
261
262
263
264
Day of the Year
102
265
266
271
272
Forage Sorghum
The only forage sorghum field in the 2007 growing season was Field No. 20-2,
which was center-pivot irrigated. Values of the CWU estimated by the Ksc method for
this field are presented in Fig. 4.4.16. Since there were no published regular crop
coefficients specific to forage sorghum, CWU was not estimated by the Kc method.
Actual measurements of CWU were collected from this field by the eddy covariance
method (Fig. 4.4.17). Result of the Student’s t test comparing the CWU by the Ksc
method and eddy covariance measurements for this field indicates that the Ksc-based
estimates are not significantly different from the actual measurements of CWU (t = 0.178, 9 df, α = 0.05, p = 0.863). This suggests that the Ksc method was accurate in
estimating the CWU. The seasonal CWU estimated for this field by the Ksc method was
447 mm.
103
Fig. 4.4.16. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 20-2 (center-pivot irrigated forage sorghum).
Crop Water Use (mm/day)
12
Ksc
10
8
6
4
2
0
160
180
200
220
Day of the Year
104
240
260
280
Fig.4.4.17.
Daily Crop Water Use (CWU) estimated by the spectral crop coefficient (Ksc) and eddy covariance (EC) methods for
Field No. 20-2 (center-pivot irrigated forage sorghum) in 2007.
6
Ksc
K sc
Daily CWU (mm)
EC
EC
4
2
0
199
202
203
207
209
211
Day of the Year
105
212
213
215
216
Pearl Millet
Field No. 26-1 was a center-pivot irrigated pearl millet field. Values of the daily
CWU estimates by the Ksc method are presented in Fig. 4.4.18. This field was planted on
8 June 2007 and attained more than 90% GC in the month of August. This resulted in
high Ksc values during the mid-growing season and CWU estimates close to PET. The
seasonal CWU estimated for this field by the Ksc method was 464 mm.
Grain Sorghum
In 2007, CWU estimates were made for three grain sorghum fields, Field Nos. 122, 15-3 and 18-2. Field No. 18-2 was center-pivot irrigated, Field No. 15-3 was furrow
irrigated, and Field No. 12 -2 was dryland. Fig. 4.4.19 through 4.4.21 shows the CWU
estimates for these fields by the Ksc and Kc methods. For all the fields, daily CWU
estimates by the Kc method were higher than corresponding CWU estimates by the Ksc
method for most part of the growing season. The difference in CWU estimates by these
two methods was large for the dryland field. Since the Ksc was specific to a field, the
CWU estimates by the Ksc method for the dryland field were significantly lower as
compared to other fields due to the sparse GC. The average difference in seasonal CWU
estimated by the Kc method for these fields was only about 25 mm. The average
difference in seasonal CWU estimated by the Ksc method was approximately 125 mm.
The lowest seasonal CWU estimated by the Ksc method was for the dryland field (261
mm).
106
Fig. 4.4.18. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method are plotted
versus the day of the year for Field No. 26-1 (center-pivot irrigated pearl millet).
Crop Water Use (mm/day)
12
Ksc
10
8
6
4
2
0
160
180
200
220
Day of the Year
107
240
260
280
Fig. 4.4.19. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 12-1 (grain sorghum-dryland). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
10
Ksc
Kc
8
6
4
2
0
120
140
160
180
200
Day of the Year
108
220
240
260
280
Fig. 4.4.20.
Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 15-3 (grain sorghum-furrow irrigated). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
10
Ksc
Kc
8
6
4
2
0
100
120
140
160
180
200
Day of the Year
109
220
240
260
280
Fig. 4.4.21. Daily estimates of Crop Water Use (CWU) in 2007 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 18-2 (center-pivot irrigated grain sorghum). Also presented in this figure are the
daily estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
10
Ksc
Kc
8
6
4
2
0
100
120
140
160
180
200
Day of the Year
110
220
240
260
280
2006 growing season
Corn
In 2006, CWU was estimated for four center-pivot irrigated corn fields (Field
Nos. 20-2, 22-2, 24-1, and 26-2). Results of the daily CWU estimated by the spectral
crop coefficient and regular crop coefficient methods are presented in Fig. 4.4.22. As
noted for the 2007 season, during the early and late parts of the growing season, CWU
estimated by the Kc method was higher than the CWU estimated by the Ksc method. The
difference in CWU estimated by these two methods decreased during the mid- and late
growing seasons. For all the fields, the CWU values estimated by both methods were
approximately equal during the mid-growing season. For Field No. 26-2, the Ksc-based
CWU produced positive values of CWU for the last portion of the growing season as
compared to the Kc- based CWU estimates, which were zero. For Field No. 22-2, both
Ksc and Kc-based CWU estimates were zero during the last portion of the growing season.
111
Crop Water Use (mm/day)
Fig. 4.4.22. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 20-2 (center-pivot irrigated corn). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
20
18
16
14
12
10
8
6
4
2
0
100
Ksc
Kc
120
140
160
180
Day of the Year
112
200
220
240
Crop Water Use (mm/day)
Fig. 4.4.23. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 22-2 (center-pivot irrigated corn). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
20
18
16
14
12
10
8
6
4
2
0
100
Ksc
Kc
120
140
160
180
200
Day of the Year
113
220
240
260
280
Crop Water Use (mm/day)
Fig. 4.4.24. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 24-1 (center-pivot irrigated corn). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
20
18
16
14
12
10
8
6
4
2
0
100
Ksc
Kc
120
140
160
180
Day of the Year
114
200
220
240
260
Crop Water Use (mm/day)
Fig. 4.4.25. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 26-2 (center-pivot irrigated corn). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
20
18
16
14
12
10
8
6
4
2
0
100
Ksc
Kc
120
140
160
180
200
Day of the Year
115
220
240
260
280
Fig. 4.4.26 presents the seasonal CWU estimated by Ksc and Kc methods for the 2006
growing season. Among these 4 fields, two (Field Nos. 20-2 and 24-1) were harvested
early for silage. Hence, the seasonal CWU for those fields were smaller than the CWU of
other fields. The average seasonal CWU for the fields harvested for silage was 691 mm
by the Ksc method and 852 mm inches by the Kc method. The seasonal CWU calculated
for the two corn fields harvested for grain (Field Nos. 22-2 and 26-2) were approximately
the same (average 922 mm) same by the Kc method. The Ksc-based estimate of seasonal
CWU of Field No. 26-2 was 98 mm more than Field No. 22-2, which shows the ability of
this method to distinguish between seemingly similar fields. The average seasonal CWU
for corn harvested for grain was 753 mm and 896 mm by the Ksc and Kc methods,
respectively.
116
Fig.4.4.26.
Seasonal Crop Water Use (CWU) in mm estimated by the spectral crop
coefficient (Ksc ) and regular crop coefficient (Kc ) methods for corn fields in
2006. Seasonal CWU is calculated by summing the daily values of CWU.
Field Nos. 20-2 and 24-1 were harvested for silage. Field Nos. 22-2 and 262 are harvested for grain.
1200
Seasonal CWU (mm)
K sc
Ksc
1000
918
830
800
873
786
874
926
719
663
600
400
200
0
20-2
22-2
24-1
Field Number
117
26-2
Kc
Kc
Cotton
Ten cotton fields under different irrigation management systems were selected for
estimating CWU in 2006. Among these ten fields, three fields were drip irrigated (Field
Nos. 1-1, 1-2 and 2) and three were center-pivot irrigated (Field Nos. 3-1, 3-2 and 6).
Two fields were furrow irrigated (Field Nos.15-1 and 15-3), and the remaining two fields
were dryland (Field No. 12-1 and 13-1). Values of the daily CWU estimated by the
spectral crop coefficient method using [Eq.1.3] and the regular crop coefficient method
using [Eq.1.1] are presented in Fig. 4.4.27 through 4.4.32. Except for the drip irrigated
fields, Kc-based estimates of CWU for the cotton fields were larger than the Ksc-based
estimates. The difference between the Ksc-based CWU estimates and Kc based CWU
estimates were largest during the mid-growing season. The CWU estimates for the drip
irrigated fields followed the same trend as noted in the case of corn. During the early part
of the growing season, CWU estimated by the Kc method was higher than the CWU
estimated by the Ksc method. During the late growing season, the CWU estimated by the
Ksc method was higher than the CWU estimates by the Kc method. During the mid-
season, the difference in estimates of CWU by these methods was small. For fields that
had the same planting dates (Field Nos. 1-1 and 1-2), the estimates of CWU by the Kc
method were the same for the entire growing season. As observed in the case of corn, the
Ksc-based CWU estimates were different for these fields for the same day, again showing
the ability of this method to distinguish between seemingly similar fields.
118
Fig. 4.4.27. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 1-1 (drip irrigated cotton). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
119
240
260
280
300
Fig. 4.4.28. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 1-2 (drip irrigated cotton). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
120
240
260
280
300
Fig. 4.4.29. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 2 (drip irrigated cotton). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
121
240
260
280
300
Fig. 4.4.30. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 3-1 (center-pivot irrigated cotton). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
122
240
260
280
300
Fig. 4.4.31. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 15-1 (furrow irrigated cotton). Also presented in this figure are the daily
estimates of CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
123
240
260
280
300
Fig. 4.4.32. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 13-1 (dryland cotton). Also presented in this figure are the daily estimates of
CWU by the regular crop coefficient (Kc ) method versus the day of the year.
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
124
240
260
280
300
Fig. 4.4.33 presents the seasonal CWU for the cotton fields in 2006. The seasonal
CWU estimated by the Kc method ranged from 501 to 544 mm, while the Ksc based
estimates showed a wider range in CWU from 156 to 495 mm. The average seasonal of
CWU values for the drip irrigated fields by Kc method was 510 mm. The seasonal CWU
of the three drip fields estimated by the Ksc method was approximately the same (average
483 mm), indicating that these fields indeed had comparable crop growth. For the centerpivot irrigated fields, the seasonal CWU by the Kc method was approximately the same
(average 508 mm) for all the fields. The Ksc-based CWU for the center-pivot irrigated
fields ranged from 296 mm (Field No. 3-2) to 385 mm (Field No. 6), and the average
seasonal CWU for center-pivot irrigated fields was 342 mm. Furrow irrigated fields had
the same Kc-based seasonal CWU (506 mm). The average seasonal CWU calculated by
the Ksc method for furrow irrigated fields was 262 mm. The seasonal CWU calculated by
the Ksc method was small for the dryland fields (156 and 183 mm), but the Kc based
estimates showed seasonal CWU comparable to other irrigated fields (average 483 mm).
Again the Ksc-based CWU estimates were site-specific and better able to handle nonoptimum growing conditions.
125
Fig. 4.4.33. Seasonal Crop Water Use (CWU) in mm estimated by the spectral crop coefficient (Ksc ) and regular crop coefficient
(Kc ) methods for cotton fields in 2006. Seasonal CWU is calculated by summing the daily values of CWU. Field Nos. 11, 1-2, and 2 are drip irrigated. Field Nos. 3-1, 3-2, and 6 are center-pivot irrigated. Field Nos. 15-1 and 15-3 are furrow
irrigated. Field Nos. 12-2 and 13-1 are dryland. The CWU by Kc method is approximately the same for all the fields,
while the Ksc-based CWU are different for each field.
600
Seasonal CWU (mm)
K sc
Ksc
Kc
Kc
400
200
0
1-1
1-2
2
3-1
3-2
6
Field Number
126
15-1
15-3
12-2
13-1
In 2006, eddy covariance measurements were made for two cotton fields (Field
Nos. 2 and 13-1). Fig. 4.4.34 presents the CWU estimated by the Ksc, Kc, and eddy
covariance methods for Field No. 13-1. The eddy covariance measurements were not
used for those days when the wind direction was not from the south. Results of the
Student’s t test comparing the CWU by the Ksc method and eddy covariance
measurements for Field No. 13-1 indicate that the Ksc-based estimates are not
significantly different from the actual measurements of CWU (t = -.289, 10 df, α = 0.05,
p = 0 .18). The CWU estimates by the regular crop coefficient method were also
compared against the actual field measurements of CWU. Student’s t test of the pairs of
observations of CWU by these two methods indicates that the CWU estimates are
significantly different (t = -3.85, 10 df, α = 0.05, p = 0.003). These results suggest that
the estimates of CWU by the Ksc method were the same as field measurements of CWU
using the eddy covariance method. As in previous analyses, the Kc method failed to
provide CWU estimates comparable to the actual field measurements.
Fig. 4.4.35 presents the CWU estimated by the Ksc, Kc, and eddy covariance
methods for Field No. 2. In this case, the CWU estimates by the Ksc method and actual
measurements were found to be significantly different for this field (t = 3.85, 16 df, α =
0.05, p = 0.001). Student’s t test of the pairs of observations of CWU estimates of Kc and
eddy covariance methods indicates that the CWU estimates by these methods are
significantly different (t = 3.65, 16 df, α = 0.05, p = 0.002). The results show that the Ksc
method tended to over-predict the daily estimates of CWU for this field. Although one
might expect this field to be non-stressed, the results suggest that this field might not
127
have been well-acclimated to the growing conditions, so that the stress factor Fs in
[Eq.1.2] was not 1.
128
Fig. 4.4.34. Daily Crop Water Use (CWU) estimated by the spectral crop coefficient (Ksc), regular crop coefficient (Kc), and eddy
covariance (EC) methods for Field No. 13-1 (dryland cotton) in 2006.
10
K sc
Ksc
Kc
Kc
Daily CWU (mm)
8
EC
EC
6
4
2
0
188
189
190
191
192
193
228
Day of the Year
129
229
230
235
236
Fig. 4.4.35. Daily Crop Water Use (CWU) estimated by the spectral crop coefficient (Ksc), regular crop coefficient (Kc), and eddy
covariance (EC) methods for Field No. 2 (drip irrigated cotton) in 2006.
10
Ksc
K sc
Kc
Kc
Daily CWU (mm)
8
EC
EC
6
4
2
0
180 181 182
183 184
207 209 210
211 212 249
Day of the Year
130
250 252
256 257 258
Pearl Millet
As in 2007, there was only one pearl millet field (Field No. 19-3) in 2006, which was
center-pivot irrigated. Values of the daily CWU estimates by the Ksc method are
presented in Fig. 4. 4.36. This field was planted on 1 May 2006 and attained a maximum
of 65% GC. This resulted in mid-season Ksc values that were low compared to the pearl
millet field in 2007. The seasonal CWU estimated for this field by the Ksc method was
443 mm.
Forage Sorghum
In 2006, CWU was estimated for two fields, Field No. 20-1 and Field No. 4-2,
which were center-pivot irrigated. Values of the CWU estimated by the Ksc method for
Field No. 20-1 are presented in Fig. 4.4.37. CWU was not estimated by the Kc method
since there were no published regular crop coefficients specific to forage sorghum in this
region. The seasonal CWU estimated for this field by the Ksc method was 252 mm.
Values of daily CWU estimates for Field No. 4-2 by the Ksc method are presented in Fig.
4.4.38. This field was harvested twice. The seasonal CWU estimated up to the first
harvest was 385 mm. The seasonal CWU estimated for the second harvest was 242 mm.
Grain Sorghum
In 2006, CWU estimates were made for only one grain sorghum field (Field No.
15-4). This field was furrow irrigated. Fig. 4.4.39 shows the CWU estimates for this
field by the Ksc and Kc methods. Except for few days in the growing season, the Kc-based
estimates of CWU were larger than the Ksc-based CWU estimates. The seasonal CWU
calculated by the Kc method was 456 mm, and by the Ksc method was 220 mm.
131
Fig. 4.4.36. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 19-3 (center-pivot irrigated pearl millet).
Crop Water Use (mm/day)
12
10
Ksc
8
6
4
2
0
120
140
160
180
200
220
Day of the Year
132
240
260
280
300
Fig. 4.4.37. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 20-1 (center-pivot irrigated forage sorghum).
Crop Water Use (mm/day)
14
12
Ksc
10
8
6
4
2
0
160
170
180
190
200
Day of the Year
133
210
220
230
240
Fig. 4.4.38. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 4-2 (center-pivot irrigated forage sorghum). Field No. 4-2 was harvested twice in
2006.
Crop Water Use (mm/day)
16
Ksc First crop
Ksc Second crop
14
12
10
8
6
Harvest
4
2
0
120
140
160
180
200
220
Day of the Year
134
240
260
280
300
Fig. 4.4.39. Daily estimates of Crop Water Use (CWU) in 2006 determined by the spectral crop coefficient (Ksc ) method plotted
versus the day of the year for Field No. 15-4 (furrow irrigated grain sorghum).
Crop Water Use (mm/day)
12
Ksc
Kc
10
8
6
4
2
0
120
140
160
180
200
Day of the Year
135
220
240
260
280
4.5. Comparison of CWU
Comparison of the CWU for cotton fields that were planted to cotton in both 2006
and 2007 shows year to year differences (Fig. 4.5.1). The rainfall in 2006 was much
lower during most of the growing season compared to the rainfall in 2007 (Fig. 4.5.2).
Reduced rainfall is associated with drier air and dry air enhances evapotranspiration
under well-watered conditions (Allen, 1999). This was observed in the current study
also. Since 2006 was a warm, dry year, the CWU values estimated for drip irrigated
cotton fields (Field Nos. 1-1 and 1-2) were larger compared to the seasonal CWU
estimates in 2007. For Field No. 2 (drip cotton), the seasonal CWU in 2006 and 2007
were approximately the same. For the center-pivot (Field Nos. 3-1 and 6) and furrow
(Field No. 15-1) cotton fields, the CWU estimated in 2007 was higher than in 2006.
Similar kinds of results were obtained for Field No. 24-1 that was planted to corn in 2006
and 2007. The seasonal CWU of corn was 719 mm in 2006. In 2007, the seasonal CWU
of corn in this field was less than in 2006 (549 mm). This indicates that the irrigation on
these fields was significantly supplemented by rainfall in 2007, and alone was not
sufficient to support the evapotranspiration demand in 2006. In 2007, these fields
received ample amounts of rainfall in each month of the growing season. This increased
availability of water caused an increase in CWU for these fields in 2007. In general, the
drier conditions in 2006 associated with reduced rainfall resulted in greater CWU by
fields in the study. Greater humid-ity, which has a strong effect on evaporation through
vapor pressure deficit, generally reduced CWU in 2007.
136
Fig. 4.5.1. Comparison of Crop Water Use (CWU) determined by the spectral crop coefficient (Ksc) method for fields that were
planted to cotton in both 2006 and 2007. Field Nos. 1-1, 1-2 and 3 were drip irrigated, Field Nos. 3-1 and 6 were centerpivot irrigated, and Field No. 15-1 was furrow irrigated.
Seasonal CWU (inches)
600
495
468
400
2006
2007
485 490
401
393
358
385 397
344
327
224
200
0
1-1
1-2
2
3-1
Field No
137
6
15-1
Fig. 4.5.2. Monthly average rainfall data for 2006 and 2007 recorded at the mesonet
weather station in Plainview, TX.
120
2006
2007
Rainfall (mm)
100
80
60
40
20
0
Jan Feb mar
Apr May Jun
Jul
Month
138
Aug Sep Oct
Nov Dec
The average seasonal CWU values estimated by the Ksc and Kc methods for cotton
fields under different irrigation management systems for 2006 are presented in Fig. 4.5.3.
The average seasonal CWU estimated by the Kc method is approximately the same for all
the fields irrespective of the irrigation conditions, since the value of the regular crop
coefficients is fixed for a given crop (Allen et al., 2004). As hypothesized, the spectral
crop coefficient was able to capture the variability in crop growth and its effect on water
use for individual fields. The Ksc-based estimates of seasonal CWU followed the trend:
drip > center-pivot > furrow > dryland. For the irrigated fields, this trend reflects the
relative efficiencies of the irrigation systems.
The average seasonal CWU values estimated by the Ksc and Kc methods for cotton
fields for 2007 are presented in Fig. 4.5.4. As observed in 2006, the average seasonal
CWU estimated by the Kc method is approximately the same for all the fields irrespective
of the irrigation conditions. The Ksc method showed comparable seasonal estimates for
the drip and center-pivot irrigated fields. The additional rainfall received in 2007
compared to 2006 caused the center-pivot irrigated cotton to grow as well as the drip
irrigated fields. An interesting observation in 2006 that the Ksc method was able to show
was that the dryland field was growing as well as the furrow irrigated fields. This is also
observed in the field during frequent field visits. This is strong evidence that the spectral
crop coefficients are superior to the regular crop coefficients in capturing the variability
in growing conditions in the filed, thus providing more accurate estimates of CWU. It
also shows the relative inefficiency of furrow irrigation.
139
Fig. 4.5.3 Comparison of seasonal Crop Water Use (CWU) determined by the spectral
crop coefficient (Ksc) and regular crop coefficient (Kc) methods averaged for
all cotton fields in the study by irrigation type in 2006. Average values
represent results from 3 drip irrigated, 3 center-pivot irrigated, 2 furrow
irrigated and 2 dryland cotton fields.
600
Seasonal CWU (mm)
K sc
Ksc
Kc
Kc
400
200
0
Drip
Center-pivot
Furrow
Irrigation type
140
Dryland
Fig. 4.5.4
Comparison of seasonal Crop Water Use (CWU) determined by the spectral
crop coefficient (Ksc) and regular crop coefficient (Kc) methods averaged for
all cotton fields in the study by irrigation type in 2007. Average values
represent results from 3 drip irrigated, 2 center-pivot irrigated, 3 furrow
irrigated and 1 dryland cotton fields.
600
Seasonal CWU (mm)
K sc
Ksc
Kc
Kc
400
200
0
Drip
Center-pivot
Furrow
Irrigation type
141
Dryland
Fig. 4.5.5 presents the average seasonal CWU calculated for all the fields in
the study by the Ksc method for 2006 and 2007. In 2007, the seasonal CWU of corn was
the highest. In 2006, the seasonal CWU followed the trend: corn > pearl millet > forage
sorghum > grain sorghum > cotton. In 2007, the seasonal CWU followed the trend: corn
> pearl millet ≈ forage sorghum > grain sorghum ≈ cotton. Averaged over the two years,
the lowest CWU in the study was for dryland grain sorghum. This confirms the general
observation that grain sorghum is a well-adapted crop for this region.
142
Fig. 4.5.5
Comparison of seasonal Crop Water Use (CWU) determined by the spectral
crop coefficient (Ksc) method averaged for all fields in the study by crop in
2006 and 2007. In 2006, average values represent ten cotton fields, four corn
fields, one grain sorghum field, three forage sorghum fields, and one pearl
millet field. In 2007, average values represent nine cotton fields, four corn
fields, three grain sorghum fields, one forage sorghum field, and one pearl
millet field.
800
Seasonal CWU (mm)
2007
2006
600
400
200
0
Cotton
Corn
Grain
Forage
Sorghum Sorghum
143
Pearl
Millet
4.6. Evaluation of the stress factor
In chapter 1, it was hypothesized that for a crop acclimatized to its environment,
the stress factor Fs in [Eq.1.2] should be approximately 1. This hypothesis can be
evaluated from results of this study. Values of CWU estimated by the spectral crop
coefficient versus the corresponding field-based actual observations of CWU by the eddy
covariance method are plotted in Fig. 4.6.1. If Fs = 1, the points should cluster along the
1:1 line. The linear regression line through these points shows a small deviation from the
1:1 line (slope = 0.8, intercept = 0.1, with an R2 of 0.8). This deviation was primarily due
to the data from the drip irrigated Field No. 2 (cotton) mentioned earlier in this
discussion. The data from the other fields cluster along the 1:1 line.
[Eq.1.2] states that CWU = GC x PET x Fs. By re-arranging this equation, Fs can
be evaluated as:
Fs = CWU / (PET x GC)
Fig. 4.6.2 shows the distribution of the stress factor calculated from the data in Fig. 4.6.1,
where the actual CWU comes from the eddy covariance measurements and PET x GC
represents the spectral crop coefficient estimates of CWU. In this figure, it is observed
that the data from Field No. 2 lies within the general scatter of points from the various
fields. All the points tend to cluster along the Fs = 1 line, with a mean value of Fs of
approximately 1. While more data is needed to make a conclusive determination, this
analysis suggests that the hypothesis that the stress factor is close to 1 is supported.
144
Fig. 4.6.1.
Daily Crop Water Use (CWU) estimated by the spectral crop coefficient
(Ksc) method plotted versus corresponding values of daily CWU measured
using eddy covariance. Diagonal line represents the 1:1 line. Dotted line
represents the regression between estimated and measured CWU values.
Results are presented for 2006 and 2007.
145
Fig. 4.6.2. Calculated values of the stress factor Fs plotted versus corresponding values
of measured daily CWU using eddy covariance. Horizontal solid line
represents Fs = 1.
146
Chapter V
Conclusions
The spectral crop coefficient (Ksc) is a novel method for estimating the water use
of field crops. The most common method for estimating crop water use involves the use
of reference evapotranspiration and an empirical crop coefficient developed for the
regions of interest. The spectral crop coefficient is evaluated from actual remote sensing
observations (satellite or aircraft imagery) of the field and thus is specific to the crop
growth characteristics in the field. This method assumes that the crop is acclimated to its
environment and determines CWU based on the product of potential evapotranspiration
and remotely sensed crop ground cover.
Specific conclusions drawn from this study are as follows:
(1.) The use of the Perpendicular Vegetation Index (PVI) was effective as a
means of evaluating crop ground cover (GC) from medium-resolution multispectral
satellite imagery and high resolution aerial imagery. By plotting the scatterplot of the red
and NIR digital counts for pixels from a satellite or aircraft image, it was routinely
possible to determine the equation of the bare soil line used in calculating PVI. From that
same scatterplot, it also was possible to identify the 100% GC point used to convert PVI
to GC. Removing pixels corresponding to non-agricultural targets (such as buildings,
paved surfaces, water bodies, clouds, and cloud shadows) from the remote sensing image
data simplified the application of this procedure. Statistical analysis of estimated and
field-measured GC from a large number of fields indicated that the procedure for
estimating crop GC from remote sensing imagery was accurate so that, on average,
147
estimates of GC determined using this procedure should be within 6 percent of their true
values.
(2.) The use of the spectral crop coefficient method was effective in estimating
daily values of CWU for fields in the study. The values of GC determined for days with
remote sensing data could be used in a crop model to produce a simulation of GC for
each day of the growing season. These daily values of GC represented values of the
spectral crop coefficient Ksc used to estimate daily CWU from daily PET values
calculated using the Penman-Monteith Equation and observed weather data. Comparison
of the CWU calculated by the Ksc method with field-based measurements of CWU
measured using the eddy covariance method showed that, with the exception of a single
field, the estimated and measured CWU values were statistically the same.
(3.) Summing the daily values of CWU calculated using the Ksc method resulted
in a seasonal CWU estimate for each field in the study. The seasonal CWU estimated by
this method showed the differences in water utilization by individual fields. Comparison
of these seasonal CWU values among the fields in the study was effective in showing
differences related to year, crop, and irrigation type.
(4.) Comparing daily values of CWU estimated using the spectral crop
coefficient method and the regular crop coefficient method recommended for crops in the
Texas High Plains with actual measurements made using the eddy covariance method
showed that the spectral crop coefficient method was consistently more accurate than the
regular crop coefficient method. The regular crop coefficient method produced
approximately the same results for each field of a given crop type. However, the spectral
crop coefficient values developed from remotely sensed GC reflected the actual growth
148
of the crop in each field, so CWU estimates were unique to each field in the study. The
regular crop coefficient method was not able to show the difference in CWU among
neighboring fields with the same crop but different irrigation types. The Ksc method
could show these differences, and could show the spatial variation in CWU within
individual fields.
(5.) It was hypothesized that for crops that were acclimated to their environment,
the value of the stress factor Fs in the equation for calculating CWU using the Ksc method
should be approximately 1. In this study, the stress factor could be evaluated from the
ratio of actual CWU measured using eddy covariance to Ksc-based CWU estimates.
Analysis of these data suggests that this hypothesis is correct.
The spectral crop coefficient method developed in this study is an accurate,
effective way to estimate the daily and seasonal CWU of agricultural crops. It is superior
to the regular crop coefficient method recommended for crops in the Texas High Plains.
It can easily be evaluated from medium-resolution multispectral satellite imagery and
high resolution aerial imagery, and does not rely on empirical relationships. Additional
research should be conducted to further explore the capabilities of this method.
149
REFERENCE CITED
Allen, R. G., M. E. Jensen, J. L.Wright, and R. D. Burman. 1989. Operational estimates
of evapotranspiration. Agron. J. 81, 650–662.
Allen, R. G., M. Smith, and L. S. Pereira. 1994. An update for the definition of Reference
Evapotranspiration. ICID Bulletin, 43 (2): p. 29.
Allen, R.G., L.S. Pereira, D. Raes, and M. Smith. 1998. Crop evapotranspiration –
Guidelines for computing crop water requirements. FAO Irrigation and
Drainage Paper 56, p. 330.
Allen, R.G. 1999. Concept paper - accuracy of predictions of project-wide
evapotranspiration using crop coefficients and reference evapotranspiration in a
large irrigation project. In Proc. United States Committee on Irrigation and
Drainage Conference on “Benchmarking Irrigation System Performance Using
Water Measurement and Water Balances”, San Luis Obispo, CA, 10–13 March
1999.
Allen, R. G. 2000. Using the FAO-56 dual crop coefficient method over an irrigated
region as part of an evapotranspiration intercomparison study. J. Hydro. 229, 27–
41.
Allen, R G., M. Smith, L. S. Pereira, D. Raes, and J. L.Wright. 2000. Revised FAO
Procedures for Calculating Evapotranspiration: Irrigation and Drainage Paper No.
56 for testing in Idaho. In Proc. Watershed Management and Operations
Management. June 20-24, Fort Collins, CO.
150
Allen, R. G. 2003. Crop Coefficients. In Stewart, B. A and T. A Howell (eds.),
Encyclopedia of Water Science. Marcel Dekker Publishers, New York. p. 87-90.
Allen, R.G., I. A. Walter, R. Elliot, T. Howell, D. Itenfisu, and M. Jensen (eds). 2005.
The ASCE Standardized Reference Evapotranspiration Equation. American
Society of Civil Engineers Environmental and Water Resource Institute
(ASCE-EWRI). P. 59.
Barnes E. M., M. S. Moran, P. J. Pinter Jr, and T. R. Clarke. 1996. Multispectral Remote
Sensing and Site-Specific Agriculture: Examples of Current Technology and
Future Possibilities. In Proc. of the 3rd International Conference on Precision
Agriculture. June 23-26, Minneapolis, Minnesota. ASA, 677 S. Segoe Rd.,
Madison, WI 53771, USA. p. 843-854.
Bashir, M. A., T. Hata, A. W. Abdelhadi, H. Tanakamaru and A. Tada. 2006. Satellitebased evapotranspiration and crop coefficient for irrigated sorghum in the Gezira
scheme, Sudan. Hydrology and Earth System Sciences Discussions, 3 (2), 793817.
Bausch, W. C., and C. M. U. Neale, 1987. Crop coefficients derived from reflected
canopy radiation: A concept. Trans. ASAE., 30(3):703–709.
Bausch, W. C., and C. M. U. Neale, 1989. Spectral inputs improve corn crop coefficients
and irrigation scheduling. Trans. ASAE. 32(6):1901–1908.
Bausch, W.C. 1993. Soil background effects on reflectance-based crop coefficients for
corn. Remote Sen. Environ. 46(2):213–222.
Bausch, W.C. 1995. Remote-sensing of crop coefficients for improving the irrigation
scheduling of corn. Agric. Water Manage. 27(1):55–68.
151
Benli. B., S. Kodal, A. Iibeyi, and H. Ustun. 2006. Determination of evapotranspiration
and basal crop coefficient of alfalfa with a weighing lysimeter. Agric. Water
Manage. 81 (3), 358-370.
Bouman, B. A. M., D. Unek, and A. J. Haverkort. 2006. The estimation of ground cover
of potato by reflectance measurements. Potato Research, 35 (2), 111-125.
Brown, P.W., C.F. Mancino, M.H. Young, T.L. Thompson, P.J. Wierenga and D.M.
Kopec. 2001. Penman Monteith crop coefficients for use with desert turf systems.
Crop Science, 41:1197-1206.
Carlson, T. N., E. M. Perry and T. J. Schmugge. 1990. Remote estimation of soil
moisture availability and fractional vegetation cover for agricultural fields. Agric.
For. Meteorol. 52, p. 45–69.
Carlson, T. N., and D. A. Ripley. 1997. On the relation between NDVI, fractional
vegetation cover, and leaf area index. Remote Sen. Environ. 62(3):241-252.
Chander, G., and B. Markham. 2003. Revised Landsat-5 TM radiometric calibration
procedures and postcalibration dynamic ranges. IEEE Tans. Geoscience and
Remote Sens. 41(11):2674-2677.
Choudhury, B. L., and S.B. Idso. 1985. An empirical model for stomatal resistance of
field-grown wheat. Agric. For. Meteorol. 36, 65–82.
Choudhury, B. J., N. U. Ahmed, S. B. Idso, R. J. Reginato, and C. S. T. Daughtry. 1994.
Relations between evaporation coefficients and vegetation indices studied by
model simulations. Remote Sen. Environ. Vol. 50, no. 1, 1-17.
152
Curran P. J., 1983. Multispectral remote sensing for the estimation of green leaf area
index. Philosophical Transactions of the Royal Society of London, Series A:
Mathematical, Physical, and Engineering Sciences, 309: 257–270.
Doorenbos, J., and W. O. Pruitt. 1977. Crop water requirements. Irrigation and Drainage
Paper No. 24, (rev.) FAO, Rome, Italy, p. 144.
Duchemin . R., R. Hadria, S. Erraki, G. Boulet , P. Maisongrande, A.Chehbouni, R.
Escadafal, J. Ezzahar, J. C. B. Hoedjes, M. H. Kharrou, S. Khabba, B.
Mougenot, A. Olioso, J.-C odriguez, and V. Simonneaux. 2005. Monitoring
wheat phenology and irrigation in Central Morocco: On the use of relationships
between evapotranspiration, crops coefficients, leaf area index and remotelysensed vegetation indices. Agric. Water Manage. Vol. 79-1 p. 1-27.
Er-Raki, S., A. Chehbouni, N. Guemouria, B. Duchemin, J. Ezzahar, and
R. Hadria. 2007. Combining FaO-56 and ground-based remote sensing to estimate
water consumptions of wheat crops in a semi-arid region. Agric. Water Manage.
87: 41–54.
Foody, G.M., R. M. Lucas, P. J. Curran, and M. Honzak. 1997. Mapping tropical forest
fractional cover from coarse spatial resolution remote sensing imagery. Plant
Ecol. 131, pp. 143–154.
Fox G. A., G. J. Sabbagh, S. W. Searcy and C. Yang. 2004. An Automated Soil Line
Identification Routine for Remotely Sensed Image. Soil Sci. Soc. Am. J. 68:13261331.
Fox G. A., and R. Metla. 2005. Soil property analysis using principal components
analysis, soil line and regression models. Soil Sci. Soc. Am. J. 69:1782-1788.
153
Fox, G.A., G. J. Sabbagh, S.W. Searcy, and C. Yang. 2004. Evaluation of an automated
soil line identification routine. Soil Sci. Soc. Am. J. 68:1326–1331
Fuchs, M. 2003. Evapotranspiration, Reference and potential. In Stewart, B. A., and T.
A Howell (eds.), Encyclopedia of Water Science. Marcel Dekker Publishers, New
York. p. 264-266.
Goel. N. S., and T. Grier. 1986. Estimation of canopy parameters for inhomogeneous
vegetation canopies from reflectance data. II. Estimation of leaf area index and
percentage of ground cover for row canopies. Int. J. Remote Sens. Vol 7, No. 10,
p.1263 – 1286.
Gitelson, A. A., Y. J. Kaufman, J. R. Stark, and D. Rundquist. Novel algorithms for
remote estimation of vegetation fraction. Remote Sens. Environ. 80(1):76-87.
Glenn, E. P., Huete, A. R., Nagler, P. L., K. K. Hirschboeck, and P. Brown. 2007.
Integrating remote sensing and ground methods to estimate evapotranspiration.
Critical Reviews in Plant Sciences, 26:3, 139 – 168.
Gutman, G., and A. Ignatov, 1998. The derivation of the green vegetation fraction from
NOAA/AVHRR data for use in numerical weather prediction models. Int. J.
Remote Sens. 19, 1533–1543.
Heilman, J.L., W. E. Heilman, and D. G. Moore. 1982. Evaluating the crop coefficient
using spectral reflectance. Agron. J. 74:967-971.
Hirano, Y, Y. Yasuoka , and T. Ichinose. 2004. Urban climate simulation by
incorporating satellite-derived vegetation cover distribution into a mesoscale
meteorological model. Theoretical and Applied Climatology. Vol. 79, Nos. 3-4,
pp. 175-184.
154
Howell, T. A., S. R. Evett, J. A. Tolk, and A. D. Schneider. 2004. Evapotranspiration of
Full-, Deficit-Irrigated, and Dryland Cotton on the Northern Texas High Plains. J.
Irrig. and Drain. Engg. 130 (4), 277-285.
Howell, T. A., S. R. Evett, J. A. Tolk, K. S. Copeland, D. A. Dusek, D, and P. D.
Colaizzi. 2006. Crop coefficients developed at Bushland, Texas for corn, wheat,
sorghum, soybean, cotton, and alfalfa. In Proc. World Water and Environmental
Resources Congress. Examining the Confluence of Environmental and Water
Concerns, May 21-25, 2006, Omaha, Nebraska. 2006 CDROM.
Huete, A.R., R. D. Jackson and D.F. Post. 1985. Spectral response of a plant canopy with
different soil backgrounds. Remote Sens. Environ.17, 37–53.
Huete, A. R. 1988. A soil-adjusted vegetation index (SAVI). Remote Sens. Environ.
25:295-309.
Hunsaker, D.J. 1999. Basal crop coefficients and water use for early maturity cotton.
Trans. ASAE. 42(4):927-936.
Hunsaker, D.J., P.J. Pinter Jr., E.M. Barnes, and B.A. Kimball. 2003. Estimating cotton
evapotranspiration crop coefficients with a multispectral vegetation index.
Irrigation Sci. 22(2):95-104.
Hunsaker, D. J., P. J. Pinter Jr., and B.A. Kimball. 2005. Wheat basal crop coefficients
determined by normalized difference vegetation index. Irrigation Sci. 24(1):114.
Jackson, R. D., S. B. Idso, R. J. Reginato, and P. J. Pinter. 1980. Remotely sensed crop
temperatures and reflectances as inputs to irrigation scheduling. Irrigation and
155
Drainage Special Conference Proc., 23-25 July, Boise, Idaho, ASCE Newyork, p.
390-397.
Jensen, M. E., R. D. Burman, and R. G. Allen. 1990. Evapotranspiration and irrigation
water requirements, ASCE Manual vol. 70, American Society of Civil Engineers,
New York. p. 332.
Kauth, R. J., and G. S. Thomas. 1976. The tasseled cap -- a graphic description of the
spectral-temporal development of agricultural crops as seen in Landsat. In Proc.
Symposium on Machine Processing of Remotely Sensed Data, West Lafayette,
Indiana, June 29 -- July 1, p. 41-51.
Ko, J., S. J. Maas, R. J. Lascano, and D. Wanjura. 2005. Modification of the GRAMI
model for cotton. Agron. J. 97:1374-1379.
Ko, J., S. J. Maas, S. Mauget, G. Piccinni and D. Wanjura. 2006. Modeling waterstressed cotton using within-season remote sensing data. Agron. J. 98:1600-1609.
Krieg D. R. Cotton water relations. Special Report 198: In Proc. 2000 Cotton Research
Meeting and Summaries of Cotton Research in Progress. p. 7-15.
Maas, S. J. 1993a. Parameterized model of gramineous crop growth: I. Leaf area and
dry mass simulation. Agron. J. 85:348-353.
Maas, S. J. 1993b. Parameterized model of gramineous crop growth: II. Within-season
simulation calibration. Agron. J. 85:354-358.
Maas, S. J. 1998. Estimating cotton canopy ground cover from remotely sensed scene
reflectance. Agron. J. 90, pp. 384–388.
Maas, S. J. 2000. Linear mixture modeling method for estimating cotton canopy ground
cover using satellite multispectral imagery. Remote Sens. Environ. 72, 304–308.
156
Maas, S. J. 2001. Use of yield prediction models in the Yield Tracker project. Abstracts,
Annual Meetings, Amer. Soc. Agronomy. Charlotte, NC. (CD-ROM)
Maas, S. J., R. J. Lascano, and D. E. Cooke. 2002. YieldTracker: A Yield Mapping and
Prediction Information Delivery System. In Proc., IFAFS Workshop. (CD-ROM)
Maas, S. J., R. J. Lascano, and D. E. Cooke. 2003. Web-based Yield Prediction
Information Delivery System. In Proc., Integrated Biological Systems Conf., San
Antonio, TX. (http://beaumont.tamu.edu/conference/presentation.asp)
Maas, S. J., R. J. Lascano, D. E. Cooke, C. Richardson, D. R. Upchurch, D. Wanjura, D.
R. Krieg, S. Mengel, J. Ko, W. A. Payne, C. M. Rush, J. Brightbill, K. F.
Bronson, W. Guo, and S. Rajapakse 2004. Within-season estimation of
evapotrasnspiration and soil moisture in the High Plains using YieldTracker. In
Proc., 2004 High Plains Groundwater Resources Conference. Lubbock, TX. pp.
219-226.
Maas, S. J., N. Rajan, J. Duesterhaus, R. J. Lascano, and J. Ko. 2005. Remote sensing
method for estimating daily crop water use. In Proc., 20th Biennial Workshop on
Aerial Photography, Videography, and High Resolution Digital Imagery for
Resource Assessment. ASPRS, Weslaco, TX. (CD-ROM)
Marek, G. W., B. W. Auvermann, T. H. Marek, and K. Heflin. 2006. Evaluating the use
of reference evapotranspiration data as an estimator of feedyard evaporation.
ASAE Paper No. 064025. In Proc. 2006 International ASABE Annual
Conference, Portland, Oregon: ASABE.
157
Monteith, J.L. 1965. Evaporation and environment. p. 205–234. In G.E. Fogg (ed.) The
state and movement of water in living organisms. In Proc. Symp. Soc. Exp. Biol.
Vol. 19. Academic Press, New York.
Morsdorf, F., B. Kötz, E. Meier, K. I. Itten, and B. Allgower, Estimation of LAI and
fractional cover from small footprint airborne laser scanning data based on gap
fraction. Remote Sens. Environ. 104 (2006), p. 50.
Neale C. M. U., H. Jayanthi, and J. L. Wright. 2005. Irrigation water management using
high resolution airborne remote sensing. Irrig. and Drain. Systems, 19
(3-4), 321-336.
Neale C. M. U., W. C Bausch and D. F. Heerman. 1989. Development of reflectance
based crop coefficients for corn. Trans. ASAE, 28: 773-780.
Neale, C. M. U., R. H. Ahmed, M. S. Moran, P. J. Pinter Jr, J. Qi, and T. R. Clarke.
1996. Estimating cotton seasonal evapotranspiration using canopy reflectance. In
Proc. International Conference on Evapotranspiration and Irrigation
Scheduling, 3–6 November, San Antonio, Texas
North, P. R. J. (2002) Estimation of f(APAR), LAI, and vegetation fractional cover from
ATSR-2 imagery. Remote Sens. Environ. 80: 114-121
Ormsby, J. P., B. J. Choudhury, and M. Owe. 1987, Vegetation spatial variability and its
effect on vegetation indices. Int. J. Remote Sens. 8, 1301-1306.
Ostle, B., and R. W. Mensing. 1975. Statistics in research. Iowa State Univ. Press,
Ames, IA.
Penman, H. L. 1948. Natural evaporation from open water, bare soil, and grass. Proc.
Roy. Soc. London A193:120-146.
158
Pinter, P. J., J. L. Hatfield, J. S. Schepers, E. M. Barnes, M. S. Moran, C. S. T. Daughtry,
and D. R. Upchurch. 2003. Remote Sensing for Crop Management. Photogramm.
Eng. Remote Sens. 69 (6):647-664
Pickup, G., V. H. Chewings, and D. J. Nelson. 1993. Estimating changes in vegetation
cover over time in arid rangelands using Landsat MSS data. Remote Sen.
Environ. 43, 243-263
Purevdorj, T. S., R. Tateishi, T. Ishiyama and Y. Honda. 1998. Relationships between
percent vegetation cover and vegetation indices. Int. J. Remote Sensing. 19(18):
3519-3535.
Qi, J., R. C. Marsett, M. S. Moran, D. C. Goodrich, P. Heilman, Y. H. Kerr, G. Dedieu,
and A. Chehbouni. 2000. Spatial and temporal dynamics of vegetation in the San
Pedro river basin area. Agric. For. Meteorol. 105, 55–68.
Richardson, A. J., and C. L Wiegand. 1977. Distinguishing vegetation from soil
background information. Photogramm. Eng. Remote Sens. 43:1541-1552.
Rajan. N., and S. J. Maas. 2006.Estimating Daily and Seasonal Crop Water Use of High
Plains Cropping Systems Using Remote Sensing and Crop Modeling. In Proc.
Southern Conservation Systems Conference, Amarillo TX, June 26-28 p.25-29.
Ray, S. S., and V. K. Dadhwal. 2001. Estimation of crop evapotranspiration of irrigation
command area using remote sensing and GIS. Agric. Water Manage. 49, pp. 239–
249.
Raupach, M., and J. Finnigan. 1988. Single-layer models of evaporation from plant
canopies are incorrect but useful, whereas multi-layer models are correct but
useless. Aust. J. Plant Physiol. 15:705–716.
159
Ringersma, J., and A. F. S. Sikking. 2001. Determining transpiration coefficients of
Sahelian vegetation barriers. Agroforestry Systems. 51(1): 1-9.
Sakuratani, T. 1981. A heat balance method for measuring water flux in the stem of intact
plants. J. Agric. Meteorol. 37:9–17.
Small, C. 2001. Estimation of urban vegetation abundance by spectral mixture analysis.
Int. J. Remote sensing. 22: 1305-1334.
Suleiman, A. A., C. M. Tojo Soler, and, G. Hoogenboom. 2007. Evaluation of FAO-56
crop coefficient procedures for deficit irrigation management of cotton in a
humid- climate. Agric. Water Manage. 91 (1-3), 33-42.
Thornthwaite. C. W. 1948. An Method towards a Rational Classification of Climate.
Geogr. Rev. Vol. 38, No. 1. pp. 55-94.
Turner, D. P., W. B. Cohen, R. E. Kennedy, K. S. Fassnacht and J. M. Briggs. 1999.
Relationships between Leaf Area Index and Landsat TM Spectral Vegetation
Indices across Three Temperate Zone Sites. Remote Sen. Environ. 70:52–68
Wanjura, D. F., D. R. Upchurch, S. J. Maas, and J.C. Winslow. 2003. Spectral detection
of emergence in corn and cotton. Precision Agriculture. 4(4):385-399.
White, M. A., G. P. Asner, R. R. Nemani, J. L. Privette, and S. W. Running. 2000.
Measuring fractional cover and leaf area index in arid ecosystems. Digital camera,
radiation transmittance, and laser altimetry methods. Remote Sen. Environ. 74,
45–57
Wittich, K. P., and O. Hansing. 1995. Area-averaged vegetative cover fraction estimated
from satellite data. Int. J. Biomet. Vol 38, No 4, pp. 209 -215.
160
Wright J. L. 1982. New evapotranspiration crop coefficients. J. Irrig. Drain. Div. ASCE,
108: 57 – 74.
Wright J. R., and C. L. Hanson. 1990. Crop coefficients for rangeland. J. Range Manage.
43, pp. 482–485.
Xiao, J., and A. Moody. 2005. A comparison of methods for estimating fractional green
vegetation cover within a desert-to-upland transition zone in central New Mexico,
USA. Remote Sen. Environ. 98(2-3): 237-250.
Zeng. X., R. E. Dickinson, A. Walker, M. Shaikh, R. S. DeFries, and J. Qi, 2000.
Derivation and evaluation of global 1-km fractional vegetation cover data for land
modeling. J. Appl. Meteor. 39, 826–839.
161