close

Вход

Забыли?

вход по аккаунту

?

A MULTISENSOR APPROACH TO CROP DISCRIMINATION: AN EXPERIMENTAL STUDY USING A MODULAR MULTIBAND RADIOMETER AND RADAR SCATTEROMETERS (REMOTE SENSING, MICROWAVE, OPTICAL SENSORS, GROUND TRUTH)

код для вставкиСкачать
INFORMATION TO USERS
This reproduction was made from a copy of a document sent to us for microfilming.
While the most advanced technology has been used to photograph and reproduce
this document, the quality of the reproduction is heavily dependent upon the
quality of the material submitted.
The following explanation of techniques is provided to help clarify markings or
notations which may appear on this reproduction.
1.The sign or "target" for pages apparently lacking from the document
photographed is "Missing Page(s)". If it was possible to obtain the missing
page(s) or section, they are spliced into the film along with adjacent pages. This
may have necessitated cutting through an image and duplicating adjacent pages
to assure complete continuity.
2. When an image on the film is obliterated with a round black mark, it is an
indication of either blurred copy because of movement during exposure,
duplicate copy, or copyrighted materials that should not have been filmed. For
blurred pages, a good image of the page can be found in the adjacent frame. If
copyrighted materials were deleted, a target note will appear listing the pages in
the adjacent frame.
3. When a map, drawing or chart, etc., is part of the material being photographed,
a definite method of "sectioning" the material has been followed. It is
customary to begin filming at the upper left hand corner of a large sheet and to
continue from left to right in equal sections with small overlaps. If necessary,
sectioning is continued again—beginning below the first row and continuing on
until complete.
4. For illustrations that cannot be satisfactorily reproduced by xerographic
means, photographic prints can be purchased at additional cost and inserted
into your xerographic copy. These prints are available upon request from the
Dissertations Customer Services Department.
5. Some pages in any document may have indistinct print. In all cases the best
available copy has been filmed.
University
Microfilms
International
300 N. Zeeb Road
Ann Arbor, Ml 48106
8529107
Jung, Bo
A MULTISENSOR APPROACH TO CROP DISCRIMINATION: AN
EXPERIMENTAL STUDY USING A MODULAR MULTIBAND RADIOMETER
AND RADAR SCATTEROMETERS
University of Kansas
University
Microfilms
International
PH.D.
300 N. Zeeb Road, Ann Arbor, Ml 48106
1985
PLEASE NOTE:
In all cases this material has been filmed in the best possible way from the available copy.
Problems encountered with this document have been identified here with a check mark V
1.
Glossy photographs or pages
2.
Colored illustrations, paper or print
3.
Photographs with dark background ^
4.
Illustrations are poor copy
5.
Pages with black marks, not original copy
6.
Print shows through as there is text on both sides of page
7.
Indistinct, broken or small print on several pages
8.
Print exceeds margin requirements
9.
Tightly bound copy with print lost in spine
.
^
^
10.
Computer printout pages with indistinct print
11.
Page(s)
author.
lacking when material received, and not available from school or
12.
Page(s)
seem to be missing in numbering only as text follows.
13.
Two pages numbered
14.
Curling and wrinkled pages
15.
Dissertation contains pages with print at a slant, filmed as received
16.
Other
. Text follows.
University
Microfilms
International
A MULTISENSOR APPROACH TO CROP DISCRIMINATION:
AN EXPERIMENTAL STUDY USING A MODULAR MULTIBAND
RADIOMETER AND RADAR SCATTEROMETERS
by
Bo Jung
B.S., Korea Military Academy, 1965
M.U.P., Seoul National University, 1980
Submitted to the Department of Geography and the
Faculty of the Graduate School of the University
of Kansas in partial fulfillment of the requirements
for the degree of Doctor of Philosophy.
Dissertation Committee:
Dissertation defended: December 1984
ii
ABSTRACT
To better understand the separate and combined capabilities of
optical and microwave remote sensing for crop discrimination as well
as the complementary nature of the two sensing techniques, an exten­
sive field experiment was planned and conducted for the two summers
of 1982 and 1983. The ground-based field research included ex­
perimental measurements of optical and microwave signatures from
agricultural crops such as wheat, corn, soybeans, potatoes, fallow,
grass, and hay over the large part of the growing season of each
crop. A Modular Multiband Radiometer as a Thematic Mapper simulator
and Mobile Agricultural Radar Scatterometers were mounted on the
boom of a truck for field operation in the vicinity of Lawrence,
Kansas.
The emphasis of the 1982 portion of the experiment was to
evaluate the capabilities of selected C- and X-band radar channels
to discriminate crops.
The separate and combined capabilities of
the Thematic Mapper bands and the radar channels were investigated
with the field data collected in 1983.
Contrary to the previous reports that recommended higher
microwave frequencies for crop discrimination, results from stepwise
discriminant analysis techniques indicated that 5.04 GHz crosspolarized data acquired at a 50 degree (CHV50) incidence angle were
the most sensitive to differences in the crop categories considered.
The Thematic Mapper channel 3 (TM3) was the best among optical chan­
nels considered for crop discrimination. When optical and microwave
channels were operated in common, a higher level of correct clas­
sification was achieved. When the two sensor systems were used
about a month apart, the accuracy level further increased, in­
dicating that the unique information provided by each sensor is
further enhanced by temporal separation which contributes to better
contrast among crop categories.
Similar studies are recommended
with airborne and spaceborne sensors to validate the optical and
microwave responses to vegetation and to determine the atmospheric
effects on the simultaneous operation of optical and microwave
sensing techniques.
ACKNOWLEDGEMENTS
This study would not have been realized without the aid and en­
couragement
of
many individuals.
A special note of gratitude goes
to my adviser, Professor T. H. Lee Williams, for his patient
tion
and
supervision for both this effort and my graduate program.
Professor F. T. Ulaby provided me the
topic
direc­
opportunity
to
pursue
this
within the framework of agricultural microwave remote sensing
in support of the Free Flying Imaging Radar
Experiment/Shuttle
tive
My former colleagues at
Microwave
Experiment (FIREX/SAMEX).
the University of Kansas Remote Sensing
Laboratory
provided
Ac­
tech­
nical assistance needed for the field operations. Finally, and most
importantly, to my family, goes a special debt of thanks
for
their
support and sacrifice without which I could not have succeeded.
iv
TABLE OF CONTENTS
Page
ABSTRACT
iii
ACKNOWLEDGEMENT
iv
LIST OF FIGURES
viii
LIST OF TABLES
xiii
CHAPTER 1. INTRODUCTION
1
1.1 General
1
1.2 Background
7
1.2.1 Developments in Optical Remote Sensing
7
1.2.2 Developments and Issues in Microwave Remote
Sensing Research
9
1.2.3 Joint Operation of Optical and Microwave Remote
Sensing
13
1.3 Objectives and Organization
13
CHAPTER 2. REMOTE SENSING APPLICATIONS TO CROP DISCRIMINATION
17
2.1 Electromagnetic Spectrum Important in Remote Sensing
17
2.1.1 Optical Range
17
2.1.2 Microwave Range
25
2.2 Optical Applications to Crop Discrimination
...
31
..
39
2.4 Multisensor Applications to Crop Discrimination . .
52
2.3 Microwave Applications to Crop Discrimination
v
CHAPTER 3. AN OPTICAL AND MICROWAVE APPROACH TO CROP
DISCRIMINATION
57
3.1 Experiment Description
57
3.2 Sensor Systems
61
3.3 Test Site and Test Fields
63
3.^ Optical Data Acquisition and Calibration
Procedures
64
3.4.1 MMR Field Operation
64
3.4.2 Calibration of MMR Data
72
3.5 Microwave Data Acquisition and Calibration
Procedures
77
3.5.1 Field Operation of Radar Scatterometers
....
3.5.2 Calibration of Radar Data
77
81
CHAPTER 4. MICROWAVE SIGNATURES TO DISCRIMINATE CROPS
....
91
4.1 Date Base and Selection of Date Variables
....
92
4.2 Techniques for Analysis
96
4.3 Timeframe 1: May 18 - July 9, 1982
104
4.4 Timeframe 2: June 16 - July 9, 1982
108
4.5 Timeframe 3: July 9, 1982
114
4.6 Timeframe 4: July 9 - August 16/18, 1982
118
4.7 Timeframe 5: August 16/18, 1982
126
4.8 Practical Number of Channels for Crop
Discrimination
132
vi
CHAPTER 5. OPTICAL AND MICROWAVE SIGNATURES: THEIR SEPARATE
AND COMBINED CAPABILITIES TO DISCRIMINATE CROPS . .
5.1 Data Evaluation and Selection of Date Variables
.
135
138
5.2 Date 1: July 6/8, 1983
140
5.3 Date 2. July 19/20, 1983
145
5.4 Date 3: July 27 - August 3, 1983
155
5.5 Importance of the Temporal Dimension and the
Complementary Nature of Radar for Crop
Discrimination
161
CHAPTER 6. SUMMARY AND CONCLUSIONS
166
6.1 Summary of Study
166
6.2 Summary of Major Results
168
6.3 Limitations and Recommendations
172
6.4 Conclusion
174
REFERENCES
176
APPENDICES
A. TIME HIL_-?Y OF SELECTED SCENE CHARACTERISTICS (1982)
B. TEMPORAL
186
JATIONS OF MEAN MMR AND RADAR
MEASUREMENTS (1983)
202
C. MMR MEASUREMENTS (1983)
215
D. RADAR MEASUREMENTS (1983)
226
E. MEASUREMENTS OF SELECTED SCENE CHARACTERISTICS (1983)
240
vii
LIST OF FIGURES
Figure
Page
2.1 The electromagnetic spectrum showing various blackbody
temperatures and windows important in remote sensing
. .
18
2.2 Typical spectral reflectance curve for healthy green
vegetation
22
2.3a Temporal variations of average photographic densities
for winterwheat, summer barley, natural hay, winter
rye, spring wheat, and sown hay
33
2.3b Temporal variations of average photographic densities
for winter barley, potatoes, rape, oats, and beets
. .
34
2.4 Variations of horizontal polarization returns with
soil moisture underlying four crop canopies
45
2.5 Sensitivity of radar return to changes in soil
moisture underlying four crop canopies plotted
versus angle of incidence
47
2.6 Diurnal variations of sorghum as measured at 7.25 GHz
and 2.75 GHz
48
2 7 Radar sensitivity to soil moisture change for different
frequencies and angles of incidence
49
3.1 Photograph of the MARS-X system observing a soybean
test field
59
3.2a Test site and field locations (1982)
65
3.2b Test site and field locations (1983)
66
viii
J
3.3 Definition of solar zenith and elevation angles
....
3.4 Reference panel measurements by hour on day 200, 1983
.
67
71
3.5 Plot of reflectance values of the reference panel
used in the experiment in relation to solar zenith
angles
76
3.6 Ninety percent confidence level for the Rayleigh
distribution
79
3.7 Field radar calibration using a luneberg lens
85
3.8 Area calculation of the radar footprint
89
4.1 Crops and study dates with corresponding radar
channels for 1982
95
4.2 Temporal variations of radar returns for selected
crops at selected times
97
4.3 Canonical transformation in two dimensions with four
crop categories
105
4.4 Performance of XW50 and XVH50 to discriminate wheat
and corn during the May 19 - July 9, 1982 timeframe
..
107
4.5 Performance of XW50 and XVH50 to discriminate wheat,
corn and soybeans during the June 16 - July 9, 1982
timeframe
113
4.6 Performance of XW50 and XVH50 to discriminate wheat,
corn, regular soybeans, and double-cropped soybeans
on July 9, 1982
116
4.7 Performance of XW50 and XVH50 channels to discriminate
ix
corn, regular soybeans, and double-cropped soybeans
during the July 9 - August 16, 1982 timeframe
119
4.8a Crop confusion status represented on the first two
canonical axes when two temporal measurements by
XVH50 are combined for classification during the
July 9 - August 16, 1982 timeframe
123
4.8b Crop confusion status represented on the first two
canonical axes when three temporal measurements by
XVH50 are combined for classification during the
July 9 - August 16, 1982 timeframe
124
4.8c Crop confusion status represented on the first two
canonical axes when three temporal measurements by
XVH50 are combined with one time measurement by XW50
during the July 9 - August 16, 1982 timeframe
4.9 Performance of
125
individual radar channels and
polarization combination to discriminate corn,
regular soybeans, and double-cropped soybeans
on August 16, 1982
128
4.10a Crop confusion status represented on the first two
canonical axes when CHV50 and XVH20 of August 16,1982
are combined for classification
129
4.10b Crop confusion status represented on the first two
canonical axes when three radar channel measurements
made on August 16, 1982 are combined for classification
x
130
4.10c Crop confusion status represented on the first two
canonical axes when measurements made by four radar
channels on August 16, 1982 are combined
for classification
131
4.11 Change of-percent correct classification level as
additional channels are combined according to
F-to-enter values
5.1
133
Reference panel measurements by date (11:00-12:00) . . 139
5.2a Performance of individual channels to discriminate
winter wheat, cut grass, spring wheat, hay, corn, and
soybeans on July 6/8, 1983
141
5.2b Combined classification accuracy change as channels
are added according to F-to-enter values on July 6/8,
1983
5.3
141
Performance of channels as a group to classify winter
wheat, cut grass, spring wheat, hay, corn, and
soybeans on July 6/8, 1983
144
5.4a Performance of individual channels to classify
corn, fallow, potatoes, soybeans, swamp, and
winter wheat on July 19/20, 1983
146
5.4b Change of combined classification accuracy as channels
are added according to F-to-enter values on July 19/20,
1983
5.5
146
Performance of channels as a group to classify corn,
xi
fallow, potatoes, soybeans, swamp, and winter wheat
on July 19/20, 1983
151
5.6a Confusion status of crops observed by M3 and CHV50 on
July 19/20, 1983 as represented on the first two
canonical axes
152
5.6b Confusion status of crops observed by M3, CHV50, and
Ml on July 19/20, 1983 as represented on the first two
canonical axes
153
5.6c Confusion status of crops observed by M3, CHV50, Ml,
and M7 on July 19/20, 1983 as represented on the first
two canonical axes
154
5.7a Performance of individual channels to classify alfalfa,
milo, corn, cut grass, regular soybeans, double-cropped
soybeans, spring wheat, and bare soil during the
July 27 - August 3, 1983 period
157
5.7b Change of combined claccification accuracy as channels
are added according to F-to-enter values during the
July 27 - August 3, 1983 period
157
5.8 Performance of channels as a group to classify alfalfa,
milo, corn, cut grass, regular soybeans, double-cropped
soybeans, spring wheat, and bare soil during the
July 27 - August 3, 1983 period
158
5.9 Performance of selected channels to classify crops
during the July 27 - August 3, 1983 period
xii
160
LIST OF TABLES
Table
Page
1.1
Space SAR Missions 1978-1991
2.1
Common Designations for Optical Wavelengths
19
2.2
Operating Wavelengths for Landsat MSS
19
2.3
TM and MMR Spectral Bands and Their Major Applications
24
2.4 Microwave Band Designations and Wavelengths
27
3.1
62
MARS-C/X Systems Specifications
3.2 Reflectance Values (%) for
4.1
Reference Panel
Radar Measurement Log for 1982
4.2 Summary of 1982
4.3
Correlation Matrix of XW50 and XVH50 for Wheat, Corn,
109
Importance of Channel Variables to Discriminate Wheat,
Corn, and Soybeans
4.5
Ill
Performance Comparison of Depolarization Ratio and
Combination of XW50 and XVH50
4.6
5.1
115
Confusion Matrix Among Wheat, Corn, Regular Soybeans,
and Double-Cropped Soybeans
4.7
93
98
and Soybeans
4.4
75
Radar Measurements for the Five
Timeframes
117
Biomass Comparison of Soybeans and Weeds in
Soybean Field 2
121
MMR and Radar Measurement Log for 1983
136
xiii
jJ
5
5.2a Performance of M3 on July 19/20, 1982
148
5.2b Performance of CHV50 on July 19/20, 1982
148
5.2c Performance of Ml on July 19/20, 1982
149
5.2d Performance of M7 on July 19/20, 1982
149
5.3
Performance Comparison of MMR and Radar Channels
to Classify Crops on Date 1 and Date 3
xiv
163
CHAPTER 1
INTRODUCTION
1.1
General
Like any other professionals, geographers
deavoring
to
solve
problems
by
are
constantly
en­
employing new techniques such as
remote sensing, multivariate analysis, modeling and simulation, and
computer
applications
to
geographical
questions. Discriminating
crop types on a global basis using remote sensing is
proach
to
phenomena.
the
Conventional methods such as field surveys,
farm-report
technology,
of
aerial photography are costly, time-consuming,
and often incomplete.
sensing
mapping
Aerial photography, a major
is
area
of
remote
a valuable tool for measuring spectra from
various field-cover features; however, it is restricted in areal
well
as
ap­
distributed
even
and
such
spatially
analyses, and
inventory
one
wavelength
coverage
as
and additional steps are required to
produce interpretable data.
Although remote sensing began to be recognized as a science
the
1960's,
it is one of the most dynamic fields in the scientific
community today.
presents
a
in
It encompasses a
variety
of
broad
research
span
of
opportunities
general and for geographers in particular.
technology
and
for scientists in
Satellite remote sensing
has been used to study spatially distributed phenomena such as iden­
tifying the areas of the world where crop
identifying
the
crop
types,
production
is
adequate,
and estimating area and yields.
1
The
Large Area Crop Inventory Experiment (LACIE) demonstrated the poten­
tial
of visual and infrared remote sensing to provide timely infor­
mation for food throughout the world
(Erb
and
Moore,
1978;
Mac-
Donald, 1983).
Remote sensing technology traditionally has
and
infrared
portions
of
the spectrum.
used
the
visible
The signals from the in­
frared range behave very much like those from the visible range
are
routinely
processed like visible data.
Hence, visible and in­
frared "windows" in the spectrum are often called the optical
and
the
terms
such
At
the
microwave
present time, spaceborne sensors using the optical
range, i.e., the
operational.
range
as optical sensors and optical remote sensing
have been developed especially to distinguish it from the
range.
and
Landsat
Landsat
Multispectral
MSS
data
Scanner
(MSS),
are
were extensively used in the Crop
Identification Technology Assessment
for
Remote
Sensing (CITARS)
project for corn and soybeans in Indiana and Illinois in 1973 - 1975
(MacDonald, 1983).
publicized
LACIE
This was followed by the highly
project
successful
and
for global wheat inventory during 1973 -
1978.
Satellite remote sensing in the 1970's
conventional
multispectral
analysis
about
largely
based
on
aerial photography, the space laboratory, and aircraft
observations
techniques
the
was
earth
were
from
made
in
the
1960's.
Machine
data
developed to rapidly extract information
the
observations
2
made
by
multispectral
scanners.
User
confidence and interest have developed.
been better scientific understanding of the complex
There has
electromagnetic
energy interactions among the sensor systems, atmosphere, and ground
scenes.
In addition to the Landsat MSS, the
optical
spectrum
was
further divided into seven Thematic Mapper (TM) bands to provide im­
proved spectral, spatial, and radiometric information over the
MSS,
mainly for vegetation sensing.
One of the limitations of landsat operations, however, is cloud
interruption.
In
many parts of the world, the weather often makes
it impossible for sensors working in the optical
ground
scenes.
range
1983).
It
is
to note that areas covered with cloud less than 30 percent
of the time are mostly desert areas in Africa, Asia, and
whereas
observe
Screening for cloud cover was one of the important
tasks of the multitemporal LACIE study (MacDonald,
relevant
to
cropland
areas
lie
Australia,
mostly in the zone covered with cloud
more than 30 percent of the time.
Also, in the Northern Hemisphere,
cloud conditions generally are worse in July when vegetation is lush
than in January (Scrimshaw, 1965). An analysis
the
Illinois
for
summer months shows that after three looks (36 days assuming an
18-day satellite pass), the percentage of
can
over
one-section
fields
that
be expected to be seen on all three passes is only 2.45 percent
because of cloud cover (Bush and Ulaby, 1977).
which
has
significant
Even
with
the
TM
spectral, radiometric, and spatial improve­
ments over the MSS, the cloudcover constraints are not resolved.
3
Microwave sensors are relatively immune to
and
are independent of solar illumination.
dency, optical
synchronous
sensors
orbit.
such
as
Landsat
weather
conditions
Because of solar depen­
are
placed
in
a
sun-
The orbit of space borne microwave sensors can
be chosen with much more freedom, because the time of day at which a
ground
scene is viewed does not matter.
be flown on any space mission.
flown
the
Space
For example, the space
series
planned
future.
the
near
of
shuttle
microwave
Europe
has
experiments
It is noteworthy that the European
Space Agency has selected microwave systems as
for
shuttle
Shuttle Imaging Radar - A (SIR-A) and -B (SIR-B).
In addition, there is a
for
Space radars can therefore
the
future
sensors
and that there is a strong emphasis in U.S. space borne
remote sensing research on the microwave region (Table 1.1).
Within the microwave range,
operated:
active and passive.
known as radar.
whereas
two
different
sensor
types
Active microwave sensors are better
Radars supply their
own
energy
or
illumination,
passive microwave systems sense the naturally available low
levels of microwave energy emitted by the ground within their
of
are
view.
field
Because of the variety of possible energy sources and the
weak magnitude of energy that
reaches
sensors, passive
microwave
returns yield a "noisy" signal compared to those of active microwave
sensor systems.
Due
to
their
low
spatial
resolution,
passive
microwave systems are not considered for crop discrimination studies
at the present time.
Table 1.1
Space SAR Missions 1978-1991
Missions
Launch
Date
Agency
Mission
Duration
Altitude
Band
Polarization
Seasat-A
1978
NASA
3 months
794km
L
HH
20°
SIR-A
1981
NASA
5 days
252 km
L
HH
47°
MRSE
1983
NASA/ESA
7 days
350 km
X
HH.VV
SIR-B
1984
NASA
7 days
250-325 km
L
ERS-1
1987
ESA
3 days
675 km
SAMEX-A
-B
-C
-0
1988
1989
1989
1990
NASA
NASA
NASA
NASA
7
7
7
7
250-325
250-325
250-325
250-325
RADARSAT
1990
Canada
3-5 years
1001 km
C
N/S
30-45°
25 m
150 km
Land and sea
FIREX
1991
NASA
3 years
N/S
L.X
HH(L)
HH.HV,VV(L)
15°, 30°
45°,60°
25 m
150 km
Land and Sea
SPIREX
1991
NASA
5 years
N/S
C,X
HH.HV,VV
15°, 30°
45°, 60°
25 m
200 km
Global
monitoring
days
days
days
days
km
km
km
km
Incidence
Angle
Nominal
Resolution
Swath
Width
Primary
Applications
25 km
ino km
Ocean dynamics
35 m
54.6 km
Geology/land
31-54°
?5 m
8.5 km
Land.and sea
HH
15-60°
35 m
50 km
Land and sea
r.
N/S
N/S
N/S
N/S
Land and sea
L,C
L,C
L.C.X
L.C.X
HH
HH(L),HH,HV
HH.HV,VV
HH.HV.VV
15-60°
35
35
35
35
50
SO
50
50
m
m
m
m
km
km
km
km
Land and sea
(After Ulaby, 1982a)
Notes:
1.
2.
L-band = 1.275 GHz; C-band = 5.3 GHz; X-band = 9.5 GHz
Abbreviations:
SIR - Shuttle Imaging Radar
MRSE - Shuttle/Space Lab Microwave Remote Sensing Experiment
ESA - European Space Agency
ERS - ESA Remote Sensing Satellite
SAMEX - Shuttle Active Microwave Experiment
FIREX - Free-Flying Imaging Radar Experiment
SPIREX - Space Station Imaging Radar
N/S - not specified
A multisensor approach to crop discrimination involving optical
and
microwave
sensors
has been recommended since the early 1960's
(Steiner, 1970; MacDonald, 1983).
sensing
and
the
The advantages
of
radar
remote
reported improvement in information extraction by
merging optical and microwave data have been
recognized (Eyton
et
al., 1979; Li et al., 1982; Ulaby et al., 1982c; Wu, 1980).
Unfor­
tunately, there is no spaceborne microwave sensor system
in
opera­
microwave
sensor
tion that is comparable to Landsat sensors.
It seems
systems
that
the
operation
of
orbiting
has not been deeply explored because of the relatively high
power levels required to operate the systems and their
resolution in comparison with optical sensors.
of spatial resolution by the use
of
the
low
spatial
With the improvement
synthetic-aperture
radar
(SAR) and the emergence of the space shuttle, it has become possible
to explore the true potential of radar
sensing
tool
for
as
an
operational
many applications including agriculture.
remoteSeasat
was the first spacecraft that carried a synthetic-aperture radar, to
monitor
the
status of sea waves, although it failed after 105 days
of operation in space (Ford et al., 1980; Elachi and Granger, 1982).
SIR-A
was flown on the Space Shuttle Columbia in 1981 and SIR-B was
flown in October 1984.
The early 1990's will see orbiting microwave sensors mounted on
a
space
station, and they will have more flexible polarization and
incidence angle options than they have now (Table 1.1).
6
The
radar
missions
are
planned to last about five years, which is comparable
to the Landsat mission.
tribute to that goal.
types, estimate
yields,
map
This paper represents
acreage,
detect
soil-moisture
successful
however,
is
effort
to
con­
Remote sensing could be used to identify crop
disease
application
accurate
and
insects,
determine
conditions, and project irrigation and
ground water requirements and supplies.
the
an
crop
of
The first
remote
sensing
identification.
Two
requirement
to
for
agriculture,
sensor
systems
operating for a common purpose but in different regions of the spec­
trum will be invaluable to realize the potential of
remote
sensing
for agricultural applications because of the complementary nature of
the systems.
1.2
Background
1.2.1 Developments in Optical Remote Sensing
In the area of optical remote sensing, five Landsat satellites
have
been
put
into
orbit since 1972.
Because the Landsat's four
band MSS turned out to be successful for a variety of geoscience ap­
plications,
the experimental MSS was declared an operational sensor
system in 1979.
changed
from
The use of Landsat MSS data in some areas
basic
has
now
research to conditionally operational applica­
tions (Tom and Miller, 1984).
In
addition
to
the
cloud-cover
problem
discussed
in
the
previous section, however, the MSS bands sample the vegetation spec­
trum coarsely and only over a limited range. This is partly due
the
to
fact that the band designations are rather conveniently divided
into four
spectral
delineation
of
regions
visible
However, the bands are
according
to
range (except
not
the
the
specifically
traditional
MSS
designed
color
band
7 (MSS7)).
for
vegetation
studies (Table 2.1).
The seven TM bands provide
vegetation
rather
detailed
coverage
geological
applications.
added
spatial
resolution
primarily
The important design goal of the TM
was to achieve better radiometric sensitivity in all bands
proved
the
spectrum, because they were chosen primarily for vegeta­
tion monitoring, although the TM band 7 (TM7) was
for
of
over
and
im­
the Landsat MSS (EDC, 1982).
The
ground resolution is 30 meters in all
but
pixel size of 120 meters on the ground.
TM6, which
achieves
a
It has been recognized that
the availability of additional spectral bands as
well
as
the
im­
proved spatial resolution in those bands provide a greater potential
for deriving information
research
and
from
development
these
period
data (EDC, 1984).
that
The
TM
terminates in January 1985
calls for extensive studies to utilize the TM's capability for prac­
tical applications.
Because of inconveniences involved in research
or
satellite
using
sensor systems and their data, remote sensing reseach
efforts are frequently conducted with ground-operated field
ments.
airborne
experi­
Airborne operations are generally more difficult regardless
8
of the sensor systems employed because of
missions
on
an
on-call
basis, the
calibration difficulties, and large
costs.
Ground-operated
sensors
the
lack
problems
of
operation
of
flying
a stable platform,
and
data-processing
are better suited for experiments
aimed at understanding the interactions between the various
signatures
and
the
scene
at
different
temporal stages.
systems used for ground-based studies in the optical
the
channel
Sensor
range
include
Exotech-100, Exotech-20C, and Barnes 12-1000 radiometers.
Exotech systems are often used as
Barnes
12-1000
is
a
Landsat
a
TM
simulator
called
Radiometer (MMR) with
an
additional
band
MSS
The
simulator.
The
the Modular Multiband
in
the
1.15
-
1.30
micrometer range.
1.2.2 Developments and Issues in Microwave Remote Sensing Research
Active microwave research is frequently conducted
scatterometers.
using
radar
A radar scatterometer is a device that measures the
energy return from the scene illuminated.
In fact, any
radar
that
makes an accurate measurement of the strength of the observed signal
is a scatterometer.
imaging
Most radar
scatterometers,
systems.
As
ground-based operations, Bradley
and
there
are
not
sensor systems. For geoscience applications, their primary
purpose has been to collect information that can
design
however,
imaging
were
radar
correlations
ranging
be
used
to
an example of the validity of
Ulaby (1980) reported
that
from 0.8 to 0.92 between ground-
operated and airborne scatterometer measurements.
9
then
In
the
wake
of
developments
in the optical remote sensing, scatterometers have un­
dergone continual modifications and
improvements
since
the
early
been
con­
1970's.
Generally, studies involving microwave sensing have
ducted
independently from those of the optical range.
The research
emphasis has been to define sensor system specifications, and to un­
derstand
and model the microwave signatures from the various compo­
nents of vegetation and soil.
Since the late 1950's, Ohio
radar
scatterometer
data
State
University
investigations
collected
from a wide variety of both agricultural
and cultural scenes, using a truck-mounted radar.
prehensive
has
The two most com­
of radar response to vegetation began in
the Netherlands and at the University of Kansas in the early
1970's
(de Loor and Jurriens, 1974; de Loor et al., 1984; Ulaby, 1982a).
Early efforts using the radar scatterometer
measuring
selected
very
much
at
microwave
systems
the same way as the spectral response is measured in
the visible region.
moisture
directed
crops over an entire growing season and showing
that added information can be obtained by active
in
were
underlying
Further studies found that the amount
the
of
soil
various crops had a significant effect on
radar backscattering; in general, radar return was found
to
be
an
selected
so
increasing function of soil moisture (Ulaby, 1975).
For soil moisture detection, radar parameters are
that
returns
are strongly sensitive to soil moisture and minimally
10
sensitive to soil surface roughness variations and vegetation
types.
cover
C-band (Table 2.4) at an incidence angle of about 15 degrees
has been found to meet
that
requirement (Ulaby
et
al.,
1982b).
Likewise, for crop-classification purposes, radar sensor parameters
are selected so as to
canopy
while
maximize
the
influence
of
the
vegetation
minimizing the effects of variations in the soil sur­
face roughness and soil moisture content.
(X-band) and
incidence
angles
Frequencies above
higher
GHz
than 40 degrees have been
recommended for crop-discrimination purposes (Ulaby
Ulaby, 1982a).
8
et
al.,
1979;
Stated simply, radar energy bounces off vegetation
cover and does not penetrate it at higher frequencies and angles
of
incidence.
Currently, investigations are
sources
of
backscattering
specifications.
row
effects
on
the
understanding
above
Special investigations into areas such as a
crop's
radar return, the diurnal variations of radar
defoliation
experiments,
individual
plant
parts
and
different
others, continue to be made (Batlivala and
al.,
bare-ground
under different surface conditions, the effects of dew
on radar backscattering, analysis of the dielectric
of
the
system
the
using
on
mentioned
signatures, step-by-step
measurements
centered
characteristics
soil conditions, among
Ulaby, 1976;
Ulaby
et
1978; Ulaby, 1982b; Aslam et al., 1983; Jung et al., 1983; Al­
len, 1984).
11
There is
no
doubt
that
microscale
investigations
and
the
modeling of the findings for applications elsewhere have contributed
tremendously to the improved scientific understanding of the complex
electromagnetic
components
findings
of
interactions between the sensor systems and various
the
target
characteristics.
Nevertheless,
need to be tested from the geographer's perspective, whose
interests include differentiating crop types for mapping on a
scale.
the
Energy
ficient for crop
returned
from
large
a crop canopy alone may not be suf­
discrimination.
It
may
be
more
effective
to
analyze sensor returns from a crop type in association with analysis
of the underlying soil and its condition, because these are
associated
with
the
agricultural crop.
closely
As an extreme case, paddy
rice on which more than half of the world's population relies
staple
food
requires
except at harvest time.
very
similar
as
a
standing water throughout its growth season,
On the other hand, dry field rice, although
to paddy rice, grows much like wheat and does not re­
quire standing water at any stage of its growth
cycle.
Therefore,
it is very likely that an X-band radar with an incidence angle of 50
degrees or higher would classify the two differently cultivated rice
crops
as
one, when the purpose of the sensing effort is to map the
two practices for agricultural planning.
was
a
low-frequency, L-band
sensing; nevertheless, it
geoscience
applications
has
Recall that the Seasat SAR
sensor, designed
been
reported
for
to
be
ocean remote
useful
for
also (Wu, 1980; Clark, 1982). The sensor
12
channel designed for soil moisture may serve better as a crop
clas­
sifier than the one recommended for crop studies (see Chapter 2).
1.2.3 Joint Operation of Optical and Microwave Remote Sensing
Although it has been recognized that two sensors
common
in
the
in
different wavelength regions are far more effective
than either one alone, only a
separate
operating
few
studies
have
investigated
the
and combined capabilities of optical and microwave sensing
to discriminate crops.
Since several spaceborne microwave sensors are planned for
late
1980's
the
and the research and development period of TM is until
January 1985, an investigation involving simultaneous ground
opera­
tions of the MMR and radar scatterometer will provide a valuable op­
portunity to enhance the knowledge of spectral
characteristics
and
perhaps provide an input to improving the accuracy of identification
of crops and eventually to the assessment
of
crop
conditions
and
yields for practical purposes.
1.3
Objectives and Organization
Making a remote sensing system operational is
problem.
There
are
in
multi-faceted
many technical aspects which are often beyond
the interest or scope of the geographer.
areas
a
Some of the major
problem
a multisensor approach that concern geographers are: (1)
defining the sensor channels to be used for
(2) understanding
the
specific
applications,
interaction among different sensor channels
13
and with the scene characteristics, (3) determining the best time to
mobilize
remote-sensing
resources for particular applications, and
(4) improving the methods of data interpretation.
Rather than attempting to encompass all the various aspects
these
of
problems, the approach taken in this investigation is to (1)
outline several specific areas, and (2) attempt to
through
provide
answers
research based on field data collected for this study using
the Barnes 12-1000 Modular Multiband Radiometer (MMR) and the Mobile
Agricultural Radar Scatterometers (MARS).
as mentioned
developed
in
at
the
the
previous
University
section.
of
Kansas
The MMR is a TM simulator
The
MARS
systems
were
and
operate at 5.04 GHz
(C-band) and 10.2 GHz (X-band), with incidence angles set at 20 and
50
degrees
Chapter 2.
for
this research.
The band designations are found in
The specifications and operations of the sensor
systems
are described in Chapter 3.
Specific objectives of this investigation are as follows:
(1) To compare the relative ability of optical and microwave
sensors
to
discriminate
various
crop
types
within
restricted
geographic areas and crop ranges.
(2) To examine the complementary nature of microwave and op­
tical
remote
sensing techniques in an effort to define operational
multisensor channels for crop discrimination.
(3) To investigate the tradeoff between the number
of
mul­
tisensor channels employed and the number of temporal data sets used
14
for crop discrimination analysis.
In addition, it is hoped that
used
in
this
the
multivariate
techniques
report will demonstrate a methodology capable of in­
vestigating the complex structure of multiresponse
data
sets
con­
sisting of different measurement scales.
The discussion
Chapter
1
in
this
paper
was a general overview.
is
carried
tisensor
studies
approaches.
presented
here
in
a
A
3
procedures.
tion
discusses
follows:
remote-sensing
brief
evaluation
general
time
of
previous
work
and
design
future
and
developments.
data-acquisition
This chapter is an important aspect of this
investiga­
in terms of the quality of the data used in the analysis.
and
is
sequence to indicate how this
studies
experimental
practical aspects of optical
and
involving optical, microwave, and mul-
research is related to previous
Chapter
as
Chapter 2 gives a review of the
electromagnetic spectrum, which is important to
crop-discrimination
out
radar
field
research
are
The
also
discussed here.
Chapters 4 and 5 present the analysis
and
evaluation
of
the
field measurements.
Chapter 4 discusses the experiment conducted in
the summer of 1982.
After a brief discussion of the techniques
and
methodology used in this study, it presents extensive discussions on
signature returns from each test field, individual channel
tions, and
the
transformation of different signatures.
evalua­
Chapter 5
deals with the 1983 investigation into the two sensor operations for
15
crop discrimination:
their advantages, tradeoffs, and complementary
nature.
The conclusions of this study, the limitations of
ment, and
recommendations
for
future
Chapter 6.
16
research
the
experi­
are contained in
CHAPTER 2
REMOTE SENSING APPLICATIONS TO CROP DISCRIMINATION
2.1
Electromagnetic Spectrum Important in Remote Sensing
Electromagnetic energy is considered to span
wavelengths
from
10
the
spectrum
micrometers (jam), the cosmic rays, to 10*®
Jim, the broadcast wavelengths. Figure 2.1 presents the
electromagnetic
spectrum
curves are for the various
curve
important
black
in
body
temperature
of
The upper
sources.
The
in the middle section shows the spectral transmittance of the
shows
the
location
of
spectrum
is
The lowest sec­
the various regions of importance in
remote sensing. For convenience, the continuum of
netic
portion
remote sensing.
atmosphere as a function of wavelength (frequency).
tion
of
the
electromag­
often divided into bands or regions such as ul­
traviolet, visible, infrared, and
microwave.
The
boundaries
of
these bands are not precise but are zones of transition (Table 2.1).
2.1.1
Optical Range
The optical wavelength region extends from 0.30 to 15 |Jmthese
wavelengths, electromagnetic
energy
retracted with solid materials like a mirror
being
can
or
be
reflected and
lens,
capable
of
manufactured to precision tolerance (Silva, 1978). The human
eye responds to radiation between wavelengths of approximately
to
At
0.72
Jim, referred to as the visible range.
uses this range (Figure 2.1).
0.38
Photography largely
The region between 0.72 and 3.0 pm is
17
BLACK BODY RADIATION CURVES AND SUN'S RADIATION
Black body at 5.800° K
•Sun's energy
>
u
fit
•Black body at 1,200°K
UJ
Z
"Black body at 600°K
-Black body at 300° K
i i • •••'
4.0 6.0
6 .8 1.0
10
40
60
100
J
200
,5mm 1mm 1cm
1m
10m 100m
100
z
Blocking effect
<n
CO
of earth's
5
w
atmosphere
o
z
<
tc
SPECTRAL RANGE OF OPERATION FOR COMMON REMOTE SENSING INSTRUMENTS
Radar
Human eye
Photography
r*
m
I
Thermal scanners _ .
2.0
4.0 6.0
f
Passive Microwave
^j
-L.
.6 .8 1.0
I »
X O CL
H
Mulli-Spcctral scanners
.3
A3 CD CD
10
20
40
60
100
-\V
200 .5mm 1mm
J
1cm
1m
10m 100m
WAVELENGTH. MICRONS
Figure 2.1
The electromagnetic spectrum showing various black body temperatures and
windows important in remote sensing (from Robinson et al., 1978).
Table 2.1
Common Designations for Optical Wavelengths
Designation
Spectral Region (ym)
Visible
0.38 - 0.72
Blue
0.40 - 0.50
Green
0.50 - 0.60
Red
0.60 - 0.70
Near Infrared
0.72 - 1.30
Middle Infrared
1.30 - 3.00
Far Infrared
.
7.00 - 15.0
Table 2.2
Operating Wavelengths for Landsat MSS
MSS Band
Wavelength (um)
Spectral Region
1
0.5 - 0.6
Green
2
0.6 - 0.7
Red
3
0.7 - 0.8
Near Infrared
4
0.9 - 1.1
Near Infrared
Note:
The Landsat MSS bands 1-4 are bands 4-7 on the Landsat 1-3 MSS
sensors. It is a change in designation only.
19
subdivided
into
the
near infrared (0.72 - 1.30 pm) and the middle
infrared (1.30 - 3.0 pm) wavelength regions (Table 2.1
2.2).
and
Figure
Energy sensed in the 0.38 - 3.0 pm range is primarily radia­
tion originating from the sun and reflected
by objects on the earth
(Silva, 1978).
The region from 3.0 to 7.0 pm is so greatly affected by the at­
mosphere
that
these
wavelengths
are
satellite remote sensing applications.
wavelengths
spectrum.
radiated
from
normally not considered for
Electromagnetic
this
range
and
indicates,
the
energy
is largely from 300 °K blackbody sources,
i.e., the earth, rather than from the sun.
sive"
in
7.0 to 15 pm is in the far infrared region of the
As the upper curve in Figure 2.1
at
energy
Therefore, terms
"emis­
"thermal" are also used to designate this portion of the
spectrum.
The numerous remote sensing devices that sense
divided
into
active
and
energy
may
be
passive sensor systems. Passive sensors
detect natural energy, either reflected or emitted.
Active
sensors
generate the energy that is transmitted to and subsequently received
from a target scene.
systems is radar.
Probably
the
most
is
of
the
active
An example of passive sensing is photography when
it is used without any artificial source of
flashlight
popular
illumination.
When
used, photography becomes active remote sensing.
the windows in Figure 2.1 show, it is obvious that the optical
As
por­
tion of the spectrum has been the most popular in remote sensing.
20
a
Typical reflectance for healthy green vegetation
Figure 2.2.
is
shown
Note that the curve is characterized by "peaks and val­
leys". The valleys in the visible portion of the spectrum are
tated
by
the
pigments
in
plant
tered at 0.45 and 0.65 pm.
the
blue
regions
cen­
If a plant is in some form of stress and
produces less chlorophyll, then there will be less
in
dic­
leaves, especially chlorophyll.
Chlorophyll strongly absorbs energy in the wavelength
sorption
in
and red bands.
chlorophyll
ab­
This information is used to
detect vegetation stress and also to differentiate crops.
At the near infrared range, the internal cell structure of
leaves controls the level of reflectance.
the
At this range, about half
of the incident energy is reflected, nearly half is transmitted, and
very little is absorbed by the leaf.
Because the internal structure
of the leaves is highly variable between plant species, reflectance
measurement
in
this
range
often
is used to discriminate between
species, even if they look the same in visible wavelengths (Hoffer,
1978; Lillesand and Kiefer, 1979).
In the middle infrared portion of the
spectrum, the
spectral
response of green vegetation is dominated by strong water-absorption
bands which occur near 1.4, 1.9, and 2.7 pm (Figure 2.2).
region,
In
this
the incident energy absorbed by vegetation varies according
to the moisture content of the leaf and the thickness of the leaf.
Plant leaves reflect, absorb, and transmit
in
incident
radiation
a manner that is uniquely characteristic of pigmented cells con-
21
Leaf
.
pigments;
Cell
structure
Dominant factor
> controlling leaf
reflectance
Water content
J
Water
absorption
Chlorophyll
absorption
Primary
[>- absorption
J bands
a
u
c
o
5=
o
0£
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
Wavelength (//m)
Visible :
-
= 5 u - ^ ear infrared I
5 Si os
a
Spectral
region
Reflective infrared
/
Middle infrared
Figure 2.2
Typical spectral reflectance
green vegetation (Hoffer, 1978).
22
curve
for
healthy
taining water solutions (Hoffer, 1978). These relationships vary as
a function of wavelength and can be shown generally as:
R = I -( A + T )
where R denotes energy reflected,
denotes
energy
I
denotes
incidence
absorbed, and T denotes energy transmitted.
0.3 to 3.0 |Jm portion of the spectrum, most sensor
this
reflected
systems
In the
measure
energy which is equal to the incidence energy minus
the energy which is either absorbed or transmitted.
tral
energy, A
Since the spec­
response patterns are unique for each plant species, the spec­
tral variability at a given wavelength
provides
or
a
band
of
wavelengths
the basis for crop discrimination, as will be discussed in
the following section.
Combining the information summarized in Figures
the
functional
for
sensor
design
in
consideration
characteristics of green vegetation.
windows
with
The TM band
TM bands are primarily for vegetation sensing except
Table
2.2,
2.2 shows, Landsat
can
be
the spectral
designations
their potential applications are listed in Table 2.3.
As
and
relationship between atmospheric transmissivity and
absorption is established, and specific wavelength
selected
2.1
and
Note that the
TM6
and
TM7.
MSS bands generally are rather con­
veniently divided in accordance with the common
for the visible region.
23
color
designations
Table 2.3
TM and MMR Spectral Bands and Their Major Applications
TM
Band
MMR
Band
Wavelength
Range (wn)
TM1
Ml
0.45 - 0.52
Coastal water mapping;
Deciduous/coniferous differentiation;
Soi1/vegetation di fferenti ation
TM2
M2
0.52 - 0.62
Vegetative vigor; (Green
reflectance by healthy vegetation)
TM3
M3
0.63 - 0.69
Plant species differentiation
(Chlorophyll absorption band)
TM4
M4
0.76 - 0.90
Biomass surveys;
Water body delineation
TM5
M5
1.55 - 1.75
Vegetation moisture measurement;
Snow/cloud differentiation
TM6
M6
10.4 - 12.5
Plant heat stress management;
Thermal mapping
TM7
M7
2.08 - 2.35
Hydrothermal mapping;
Rock type discrimination
M8
1.15 - 1.30
Vegetation health
Potential Applications
Zk
2.1.2
Microwave Range
The spectrum between 15 pm and about
remote
is
not
used
for
Also detectors sensitive enough to be operational
region
windows
in
mosphere.
in
have not been developed (Slater, 1980). The curves in
Figure 2.1 are for clear weather conditions.
CO2,
mm
sensing mainly because of water absorption in the atmosphere
(Figure 2.1).
this
1
the
optical
range
is
gaseous
The problem
CH^ gases.
the
absorption in the at­
They are obstructed due to absorptions by
N2O, and
with
C^, 0^, ^0,
Water in the form of clouds is the most
troublesome cause of attenuation to sensors operating in the optical
windows.
The radio
window
extends
from
1
mm
to
10
m.
At
these
wavelengths, obstruction is mainly by the oxygen and hydrogen gases.
Although
the
wavelengths,
atmosphere
operation
very
little
relatively
above
ionospheric reflections.
tributes
is
3
"transparent"
these
m is nearly impossible because of
Below 3 m, however,
to
at
the
ionosphere
con­
the total atmospheric propagation losses
(Bush and Ulaby, 1977).
Within the microwave range approximately from
serious
1
to
1
m,
atmospheric effects are confined to the shorter wavelengths
of less than 3 cm (Figure 2.1).
wavelength
increases.
Heavy
Generally, attenuation decreases as
rain has a strong influence on radar
operating at 1 cm or less, but, the effects of rain
with
mm
wavelengths
greater
than
3
25
cm.
Cloud
are
negligible
attenuation in the
microwave region, particularly in the longer
orders
wavelength
range,
of magnitude smaller than the attenuation in the visible and
infrared region (Ulaby, 1982a). Short of the conditions of
heavy
rain, radar
night
a
very
can be used through clouds, smoke, fog, or vir­
tually any atmospheric conditions.
and
is
sensing
This feature, combined with
day
capability, makes radar a particularly valuable
tool to form a multisensor system in geoscience applications.
Because the microwave range is so wide compared
tical
the
op­
spectrum, different regions of the microwave range are often
referred to by letter codes.
in
with
Table 2.4.
The common wavelength bands are listed
The letter codes for the various bands such as X, C,
and L were originally used by the military during World War
security
II
for
purposes in the development and employment of radar. They
continued to be used to describe the microwave spectrum
even
after
the release from military security classification in 1970 (Jensen et
al., 1977).
As Figure 2.1
available
in
the
shows,
there
is
very
little
emitted
energy
microwave range, and very sensitive equipment is
necessary to sense the naturally radiated energy in this part of the
spectrum.
Passive sensors operating in the microwave range are not
considered for crop discrimination purposes
ferior resolution.
study
of
their
in­
The most commonly used active microwave bands in
geoscience applications are L, X, and K.
under
because
because
of
its
Recently, C-band has
been
relative capability to penetrate the
26
Table 2.4
Microwave Rand Designations and Wavelengths
Band
Wavelengths (cm)
Frequency (GHz)
P
100-30
0.3-1
L
30-15
1 - 2
S
15-7.5
2 - 4
C
7.5
X
3.75 - 2.4
Ku
2.4
- 3.75
- 1.67
4-8
8
- 12.5
12.5 - 18
K
1.67 - 1.18
18.0 - 26.5
Ka
1.18 - 0.75
26.5 - 40
Millimeter
1.00 - 0.1
30
- 300
(Long, 1975)
Notes:
1.
Hertz (Hz) = cycles/second
2.
Giga Hertz (GHz) = 10^ Hertz
3.
Wavelength (cm) = 30/frequency (GHz)
27
vegetation canopy.
While optical responses are sensitive to
structure, water
contents
of
plants,
and
leaf
pigments,
cell
background soil color
(Rosenthal and Blanchard, 1984), active
microwave
radars,
roughness) and dielectric
respond
to
the
geometry (or
properties of vegetation and background
radar
return
soil.
sensors,
More
i.e.,
specifically,
is a function of operating wavelength (microwave fre­
quency), polarization configuration of the transmit and receive
tennas,
the angle of illumination, and the geometric and dielectric
properties of a given scene.
interact
to
produce
The radar system and scene
one radar return.
canopy
and
the underlying soil.
vegetation increases as the wavelengths
lowing
the
vegeta­
The attenuation through the
become
shorter,
thus
al­
less penetration at short wavelengths. The cause of the at­
tenuation is reported to
(Ulaby,
parameters
The geometry may depend on
the angle of incidence and the depth of penetration in
tion
an­
1975;
Long,
be
1975).
largely
If
moisture
the
content
in
plants
microwave energy penetrates
through the vegetation canopy, the underlying soil
may
also
con­
tribute to the radar energy return.
Whether a particular scene appears "rough" or "smooth" on radar
return largely depends on the wavelengths of radar.
A paved airport
runway may stand out clearly against
ground
the
runways.
grass-covered
between
At L-band the grass appears rough on imagery and the
runways appear quite smooth, resulting in a strong contrast.
28
As the
wavelengths
decrease, however, the
contrast
becomes less (Long,
1975).
Radar imageries of an agricultural scene contain variations
tone
and
texture and these variations define boundaries of various
types of crops.
smooth
in
with
In general, the radar return of a
respect
scene
that
is
to the wavelength is weaker than a rough sur­
face, except when a flat surface is viewed at near-normal incidence.
When
looking straight down with radar, a smooth scene can be a much
stronger target than a rough scene because more energy is
If
the
returned.
angle of illumination is increased to 40 to 50 degrees, the
radar energy is known to bounce off the vegetation
reflected
energy
received
is
vegetation (Ulaby, 1982a). A
mainly
study
from
using
canopy
and
the
the upper part of the
side
looking
airborne
radar (SLAR) reported that there was no significant penetration of
vegetation observed at X-band.
penetrated
low
At L-band, however, the radar energy
grasses and reeds standing 0.9 to 1.5 meters out of
the water (Drake et al., 1974).
Radars originally operated
equipment
decimeter
wavelengths
because
had not been developed for short wavelengths, but by 1945
operating wavelengths
shorter
at
wavelengths
had
been
extended
were being used.
into
K-band,
and
even
The advantage of the shorter
wavelengths is that narrower beamwidths (see Chapter 3 for
defini­
tion) and thus better resolution, is obtainable with a given antenna
size. Disadvantages include a greater loss of signal
29
strength
as­
sociated
with atmospheric absorption and scattering, greater inter­
nal system losses, and less transmitter power (Long, 1975; Ulaby
et
al., 1982a).
Recently,
there
has
been
renewed
interest
in
decimeter
wavelengths (Table 2.2). The Seasat program and the Shuttle Imaging
Radar (SIR) missions have flown L-band systems, for example.
wavelengths
permit
penetrate vegetation
target
canopy
detection
and
the
at
These
very great distances and
soil
surface.
The
latter
characteristic highlights terrain features on radar return, which is
suiatable for geologic applications.
Depending on the degree of penetration, the underlying soil can
contribute significantly to the overall radar return.
At low angles
of incidence with C-band, soil contribution dominates the
tion
from
the
vegetation.
If
soil
contribu­
moisture is discernible for
mapping, it is potentially important to irrigated agriculture.
Although tone and texture are primary features for scene
iden-
tifiction on radar imagery, other elements such as shape, size, pat­
tern, association, and site are also
very
important
in
the
crop
discrimination process. For a given size of antenna aperture, beamwidths obtainable are directly proportional
For
radar
wavelengths.
example, at 4.3 mm (70 GHz) a 1° beamwidth can be obtained with
an antenna approximately 30 cm (1 foot) in
(0.7
to
GHz),
an
antenna
diameter.
At
42.9
cm
diameter of about 30 m (100 feet) would be
necessary to achieve the same beamwidth (Long,
30
1975).
Therefore,
there has been continuing interest for using the shorter wavelengths
for providing improved target recognition.
Because
of
information
of tone and texture at long wavelengths and detailed information ob­
tainable at short wavelengths, Drake et al. (1974) suggested that Land X-band radars be used for vegetation growth stage measurements.
2.2
Optical Applications to Crop Discrimination
Efforts to classify agricultural crops through
began
in
the 1950's.
remote
Before 1950, it was ascertained that certain
features, mostly physical, could be mapped with speed
from
aerial
photographs,
but
and
the
corresponding
and
aerial
Goodman compared ground
photographic
appearance
selected fields in Northern Illinois, at nine intervals
from
May to October 1950.
of
their
of
of
growth
She showed that farm crops could be dif­
ferentiated on aerial photographs by the unique tonal
qualities
accuracy
no definitive method of identifying
agricultural crops existed (Goodman, 1959).
truth
sensing
and
textural
photographic images, and by objects which were
commonly found in association with them.
A number of investigations to identify crop types on black-andwhite
and color aerial photographs were conducted in the 1950's and
early 1960's.
It was recognized that temporal
black-and-white
and
color coverage during the growing season might permit identification
of individual crops with certainty (Colwell, I960;
These
conventional
techniques, however, were
costly, and limited to small areas.
31
Steiner, 1970).
slow, subjective,
In the
automatic
late
methods
was believed that
1960's, efforts
of
the
to
develop
semi-automatic
or
crop classification began.
At that time, it
combination
number
of
a
larger
of
crop
parameters such as photographic densities, canopy heights, and spec­
tral returns, was more likely to produce a correct identification of
crops than the use of the single parameter of image tone.
In Switzerland, Steiner (1970) was concerned with the
and
the
classification
analysis
of densitometric and stereometric measure­
ments from black-and-white, infrared, and natural and color infrared
photographs
using
a
digital
computer.
He combined densitometric
measurements of June and July true color data and was able to
the
crop
raise
discrimination accuracy to 75 percent as compared with 40
percent (June) and 29 percent accuracy (July) on single
date
photo
sets.
In another instance, Steiner (1970) and his collaborators
lected
densitometric
measurement
categories in Switzerland from
thirteen
different
times
photographic
of
panchromatic
during
the
years. Figures 2.3a and 2.3b present
average
data
major
photographs
growing
the
eleven
col­
crop
taken
at
seasons of several
temporal
variations
of
densities for the eleven crop types under in­
vestigation.
For automated classification
analysis,
average
den­
sities
were
read off the graphs at ten day intervals from April 10
to August 8.
Using an estimated average value of 0.04 for the stan­
dard deviation, a normal distribution was simulated to draw nine ob-
32
Winter Wheat
Summer Barley
Winter Rye
Spring Wheat
— Sown Hay
0
20
40
April 10 April 30 May 20
60
80
100
June 9 June 29 July 19
Time - Days
120 130
Aug 8
Figure 2.3a
Temporal variations of average photographic den­
sities for winter wheat, summer barley, natural hay, winter rye,
spring wheat, and sown hay (Steiner, 1970; from Bush and Ulaby,
1977).
33
Winter Barley
Potatoes
Rape
— Oats
— Beets
\
\
\
/
\
•
0
20
40
April 10 April 30 May 20
•
'
60
80
100
June 9 June 29 July 19
Time - Days
120 130
Aug 8
Figure 2.3b
Temporal variations of average photographic den­
sities for winter barley, potatoes, rape, oats, and beets
(Steiner, 1970; from Bush and Ulaby, 1977).
3^
servations around the mean for each crop and each
F-ratio
date.
power.
discriminatory
Based upon classification accuracy, he concluded that June 9
was the best among the 13 dates to
types (Figures
2.3a and 2.3b).
differentiate
the
eleven
crop
With the single day data, he could
correctly classify 55 percent of the samples.
Further, he
selected
dates that discriminated the crop types with a 90 percent ac­
curacy.
could
an
of among-categories variance to within-categories variance,
Steiner selected the dates that provided the largest
three
Using
The results suggested that a reliable
be
crop
classification
achieved if a few but relevant temporal observations were
available.
Although the Steiner data inevitably
exposure
involved
differences
and processing from film to film and also of variations in
crop phenology from location to location and from year to year,
study
the
demonstrated the importance of the temporal dimension, one of
the two basic dimensions which provide multivariate
remote
in
sensing.
information
in
Note that the Steiner study used only one spectral
dimension, i.e., panchromatic aerial photography.
The temporal dimension provides valuable information
for
crop
discrimination, because the dynamic nature of continuously changing
crop conditions influences the variations in the reflectance
(see Figures
2.3a
and 2.3b for example).
Another major dimension
which probably has been more emphasized is the
The
spectral
curves
spectral
dimension.
response patterns from healthy, green vegetation tend
35
to be generally the same as shown in Figure 2.2. However, there are
small
or
large
variations
both
within
and
among
vegetative
categories, depending upon the leaf pigments, cell structure,
water
plant
content, stress conditions, background soil effects, cultural
practices, and so on.
Because of the variations, it is possible
to
identify different plant species.
If two wavelengths or bands of wavelengths are considered,
reflectance
values
observed
for
each crop form a cluster of data
points on a plane formed by the two
points
can
be
bands.
The
cluster
of
data
be described by a statistical distribution and are sub­
ject to mathematical manipulation.
can
the
used.
If
data
from
In addition, more than two bands
p-bands are used to classify h crop
categories, for example, each crop can be conceived as
a
swarm
of
points in p-dimensional space centered at a point characterized by a
mean vector.
ellipsoidal
The data points may be clustered about this mean in an
pattern characterized by the variance-covariance matrix
(Seal, 1964).
lesser
degree
The ellipsoids of crops may overlap to a
and
the
mean
vectors
separated (see Chapter 4). In the
the
analysis
greater
are more or less distinctly
spectral
task involves two aspects:
dimension,
therefore,
one being how to select a
few important spectral dimensions and another being deciding how
partition
the
multidimensional
or
to
space into regions associated with
each crop for correct classification.
36
Important investigations were conducted with the development of
multispectral
scanners.
The mid-1960's saw the introduction of the
first airborne multispectral scanner (MSS) by
the
University
of
Michigan (MacDonald, 1983). Design simulations for the first earth
resources satellites were conducted in 1970.
airborne
At the same time,
the
MSS was used in the Corn Blight Watch, which was the first
large-scale application of remote sensing in agriculture in 1970.
Research into data classification techniques
targets
by
Purdue.
The purpose was to process airborne
data
to
machine
produce
a
The
agricultural
also began at the Universities of Michigan and
line
printer
received, thus skipping the
1966).
for
map
photography
multispectral
directly
or
from the signals
imagery
stage (Lowe,
first successful computer recognition of wheat was ac­
complished using aircraft multispectral measurements in
research
than
three
bands,
it
is
very
visually the spectral information.
example, takes
up
When dealing
For
recognition
with
difficult to fully evaluate
The additive color
viewer, for
to only four band data for color rendition.
overcome this limitation, computers were used in the
tern
1966.
purposes, multispectral scanners with as many as 24 chan­
nels were tested at the same time in the 1960's.
more
scanner
procedures.
spectral
To
pat­
By dealing with image data quantita­
tively, the spectral information in virtually any number of channels
could be evaluated (Lillesand and Kiefer, 1979).
37
Probably the most extensive agricultural remote sensing program
would
be
the Large Area Crop Inventory Experiment (LACIE). It was
initiated in 1974 as a
global
wheat
monitoring
system (Erb
and
Moore, 1978). Its organized research activities were established in
1965 by the U.S.
of
several
Department of Agriculture and NASA.
The
feasibility investigations conducted with Landsat-1 led
to the design and initiation of LACIE in 1973 - 1974.
computer-aided
statistical
of
The
digital,
pattern recognition techniqufes employed
in LACIE were designed to take advantage of
response
success
crop types over time.
the
changing
spectral
Thus Landsat data were acquired
throughout the crop season, screened for cloud cover, registered
previous
acquisitions,
and
the 5x6 nautical mile (9 x 11 km)
sample segments were extracted in
was
done
digital
format.
Classification
by selecting training samples of less than one percent of
each sample segment as either wheat or non-wheat.
final
to
production
estimate
The
1977
Soviet
released in January 1977 was 92 million
metric tons and the LACIE final estimate
tons, a difference within one percent.
was
91.4
million
metric
The project demonstrated the
potential to provide timely information on
the
global
extent
and
condition of various crops needed for food throughout the world.
Landsat 4 was launched in 1982 with the TM and
and
Landsat
5
was
launched
in
specifically selected for vegetation
1984.
MSS
on
board,
Most of the TM bands were
sensing.
However, not
many
crop discrimination studies involving TM data are available, perhaps
38
partly due to system problems that developed with
1984).
At
landsat 4 (EDC,
this writing, visual interpretation of TM band combina­
tions is being studied. Preliminary results indicate that a combina­
tion
of bands 2, 3, and 4 is preferred for color infrared composite
for general use, and that a composite generated from bands 3, 4, and
5 is preferred to locate geologic land features (EDC, 1984).
2.3
Microwave Applications to Crop Discrimination
The potential of active microwave remote sensing has
subject
of
studies
only
the
since the mid-1960's (McDonald, 1983), but the
place of microwave sensing in geoscience
recognized
been
recently.
In
remote
sensing
has
been
his analysis of photographic data,
Steiner (1970) realized the difficulties produced by the presence of
frequent
cloud cover over certain areas, and stated that the use of
another spectral band which was less affected by clouds might be the
solution.
The first major radar survey using side looking airborne
(SLAR) was
a
complete
mapping
in 1968 of the Darien Province of
Panama connecting Central and South America.
this
area
has
radar
Prior
to
that
time,
never been mapped in its entirety because of nearly
perpetual cloud cover.
The rainforests of
the
Amazon
Basin
also
were among the first areas to be mapped by radar.
Perhaps the first serious study of the use of radar to classify
crop
types was performed by Simonett et al. in 1967.
film density of beets, wheat, and corn on Ka-band
39
They measured
imagery
acquired
by
the
AN/APQ-97
radar
over a test site near Garden City, Kansas
from October to November 1964 and from
They
related
to
September
1965 -
each ground truth variable to the measured density on
radar imagery by crop type on several
plant
August
canopy
dates.
They
reported
that
height, percent ground cover, and percent plant mois­
ture are the significant factors that account for
the
the radar return, in descending order of importance.
that acquiring imagery on a monthly
basis
variance
in
They indicated
throughout
the
growing
season would significantly add to the discriminatory capabilities of
radar.
Schwarz and Caspall (1968) digitized dual
acquired
in
1965
1966
imageries
over the Garden City test site.
As a
result of analyses using clustering techniques, they indicated
that
multi-polarization
and
polarized
coverage
was quite helpful to discriminate crop
types but that temporal coverage was more effective for
tion
purposes.
Also
discrimina­
it was felt that multiple frequencies should
have many of the same advantages for crop
identification
found
by
similar studies using multispectral data in the visible and infrared
regions.
helpful
It was also noted that
a
satellite
to provide repetitive coverage.
platform
should
be
Similarly, Haralick et al.
(1971) employed a Bayesian decision rule to attempt crop discrimina­
tion, and
Morain
and Coiner (1970) used data generated from radar
imagery to define unique crop signatures.
40
In another study of the Garden
City
Ka-Band
imagery, Berger
(1970) attempted to make use of the texture information available in
the imagery for target classification.
significant
analysis.
cess
in
new
information
could
It was
be
determined
that
extracted from the texture
On the other hand, Brisco and Protz (1982) reported
identifying
no
suc­
corn fields using image tone and texture from
multidate, multichannel radar imagery.
Another example of radar's capabilities as a remote crop
sifier
was
the
work
clas­
of Batlivala and Ulaby (1975). For this in­
vestigation, L-band HH and HV imageries were acquired with
the
En­
vironmental Research Institute of Michigan (ERIM) synthetic aperture
radar (SAR) system in 1973 over a Huntington
site.
The
They reported that the highest
classification
tion data.
test
vegetation categories studied were corn, soybeans, pas­
ture, and woodland.
correct
county, Indiana
probability
of
resulted from using the HH and HV polariza­
The percentage of correct classification increased
from
64.6 percent using HH data only to 71.2 percent using both HH and HV
data.
By now it becomes
mid-1960's
to
clear
that,
studies
the
period
have
crop
discrimination
the
of
studies.
not been entirely successful in discriminating
crops due to the fact that the densitometric data are
representation
from
the mid-1970's, densitometric data derived from SLAR
imagery had been used in a number of
These
during
complex
radar
kl
a
simplified
data and that the radar channels
used were designed mainly for terrain mapping and geologic
applica­
tions.
The Skylab satellite
remote sensing.
a
significant
event
in
microwave
It was the platform of the S193 microwave system, a
triple purpose active
terometer,
was
and
and
passive
instrument (radiometer, scat-
altimeter), operated at 13.9 GHz (21 mm).
Although
the spatial resolution of 11 km at nadir was rather coarse, the S193
viewed
an
unparalleled
number of targets at varying angles of in­
cidence (Sobti, 1975; Moore et al., 1980).
Prom
microwave
the
mid-1970's,
sensor
efforts
intensified
the
sensor
channels
define
and
target
The successful launch of ERTS-1 (later changed to
Landsat-1) excited the scientific community and started
optical
to
channels for vegetation sensing and to better un­
derstand the interaction between
characteristics.
have
remote sensing.
a
boom
in
The Skylab experiment proved that orbiting
microwave sensors can be effectively used to provide an overview
of
an area of subcontinental magnitude with maximum economy of time and
money.
One of the major problems in microwave remote sensing, however,
has
been
the fact that specifications for radar sensor systems for
agricultural
applications
specifications, often
have
referred
not
been
defined.
illumination.
system
to as system "parameters," include
the frequency (wavelength), angle of illumination, and
of
The
polarization
Microwave sensors operate in a wide range of the
kz
spectrum from one millimeter (mm) to one meter (m), and
radar
sig­
nals can be transmitted and received in different modes of polariza­
tion.
That is, the signal can be filtered in such a
electrical
that
its
wave vibrations are restricted to a single plane perpen­
dicular to the direction of wave propagation.
possibility
of
dealing
Thus,
there
is
the
with four different combinations of signal
transmission and reception: Horizontal (H) transmit
transmit
way
H
receive, H
Vertical (V) receive, V transmit V receive, and V transmit
H receive. Like polarized data result from the HH
or
W
combina­
tions, and cross polarized data are obtained from HV or VH combina­
tions.
Since various objects modify the polarization of the
energy
they reflect to varying degrees, the mode of signal polarization in­
fluences how the objects will be represented in the remotely
sensed
data.
In addition, the angle of incidence can be anywhere from
75
degrees
for
practical
experiments
to
applications, with 0 being the nadir.
Because of this dynamic nature
radar
0
of
microwave
combination
options,
in agriculture have been mostly to determine op­
timum sensor parameters as function of varying target scene
charac­
teristics. For example, SIR-B was flown to measure a given scene on
six successive days at six different angles of incidence to test the
sensor-target
interactions over a range of viewing
and atmospheric
conditions, although frequency and polarization were fixed at L-band
and like polarization.
^3
Investigations of microwave sensor parameters for crop studies
were
possible with the development of radar scatterometers suitable
for ground-based field research
for
without
important
interruption.
terometer
is
not
backscatter", the
an
It
is
imaging
strength
a
prolonged
radar.
of
the
Rather
in
the
late
it
time
measures
power
Initiated by
1950's, researchers
of
realize that a scat-
transmitted
returned to the antenna by a given scene.
University
to
period
built
the
that
is
Ohio
State
radar
scat­
the
radar
terometers and conducted comprehensive investigations of
response to vegetation.
In a series
specifications,
of
efforts
perhaps
the
to
define
microwave
University
of
Kansas
Sensing Laboratory (RSL). One of the goals of the first RSL
scatterometer design was to show that added information can
tained
system
most complete data base of scattering
data on vegetation has been developed at the
Remote
sensor
by
be
ob­
measuring radar backscattering over the microwave range,
much as the spectral response of a target is measured in the visible
and infrared portion of the spectrum (Bush and Ulaby, 1977).
After the first system design experiment in
program
was
1971, a
research
initiated at RSL in 1972 (Ulaby and Burns, 1979).
The
major finding from the data obtained near Eudora, Kansas in 1972 was
that the content of soil moisture underlying the various crops had a
significant effect on radar backscatter.
of
this
effect.
Figure 2.4 is
an
example
This figure shows that radar backscatter and soil
44
20r Crop Type: Milo
Crop Type: Corn
Crop Height 2.4 Meters
Frequency: 4.7 GHz
Polarization: HH
•
e • 0°(Normal)
»
e • 30°
Crop Height- 1.0 Meter
el5h Frequency: 4.7 GHz
Polarization: HH
•
e * 0° (Normal)
e-30°
%
_10
c
OJ
°<J
£ 5
8
0
01
o>
x:
-5-10.
5
10
15
20
25
30
Percent Moisture Content by Weight
20
15
ho
C
o>
u
s
'15
ho
g . 30°
£ 5
5
8
o
C_J
•?o
>«
f0
O)
is
ro
r
o -5
IS)
'
r
o -5
in
-10
0
5
10
15
20
25
30
Percent Moisture Content by Weight
Figure 2.4
35
20
Crop Type: Soybeans
Crop Height 1.0 Meter
Frequency: 4.7 GHz
Polarization: HH
•
e • 0° (Normal)
.
5
10
15
20
25
30
Percent Moisture Content by Weight
35
35
Crop Type: Alfalfa
Crop Height 0.5 Meter
Frequency: 4.7 GHz
Polarization: HH
•
e • 0° (Normal)
•
e • 30°
-10,
0
5
10
15
20
25
30
Percent Moisture Content by Weight
35
Variations of polarization returns with soil moisture underlying four crop
canopies (Ulaby, 1975).
moisture content has a linear relationship.
As soil
moisture
con­
tent increases, generally the radar return also increases.
From the slopes of the best fit lines shown in
sensitivity
factor
the
2.5.
It
20
is
clear
curves in Figure 2.5 that the crop canopy has minimal ef­
fect on radar return at the incidence angles ranging from
to
2.4, a
to describe the radar sensitivity to soil mois­
ture change was defined and is shown in Figure
from
Figure
degrees.
Note
also
that
the
about
10
effect of soil moisture is
minimized at 7.1 GHz.
Further experiments made over a 24 hour period showed that
radar
the
return measured at 2.75 GHz increases gradually and reaches a
maximum around dawn (Figure
markedly
2.6).
However,
the
variations
reduced with the measurements made at 7.25 GHz.
are
It is ob­
vious from the curves that higher frequency is better if radar is to
be a day and night vegetation monitor.
Using the data collected during 1972 beans, alfalfa, and
1974 from
corn,
soy­
milo fields, researchers at the University of
Kansas constructed a set of curves that showed
the
sensitivity
of
the radar return to soil moisture change at various incidence angles
as a function of frequency.
radar
to
monitor
crops
It is apparent from Figure 2.7 that the
should operate at frequencies of 8 GHz or
higher and incidence angles of 40 degrees or greater to minimize the
effects from underlying soil.
l±6
Polarization ; HH
Frequency : 4.7 GHz
o
Corn
•
Mi lo
a
Alfalfa
a
Soybeans
10 20 30 40 50 60
Incidence Angle in Degrees
Polarization: VV
Frequency: 4.7 GHz
o
Corn
Milo
- Alfalfa
- Soybeans
70
0
10 20 30 40 50 60
Incidence Angle in Degrees
VV Polarization, 4.7 GHz
HH Polarization, 4.7 GHz
Polarization : HH
Frequency: 7.1 GHz
o
Corn
Milo
a
Alfalfa
a
Soybeans
Polarization : VV
Frequency; 7.1 GHz
o
Corn
Milo
a
Alfalfa
a
Soybeans
0.4 L~~
V.\°
10 20 30 40 50 60
Incidence Angle in Degrees
70
70
•o
CO
HH Polarization, 7.1 GHz
,
10 20 30 40 50 60
Incidence Angle in Degrees
VV Polarization, 7.1 GHz
Figure 2.5
Sensitivity of radar return to changes in soil
moisture underlying four crop canopies plotted versus angle of
incidence (Ulaby, 1975).
47
70
Crop Type: Hybrid of Sorghum (Sand M)
Polarization: VV
Frequency (CHz): 7.25
10° a
-a
30°
50°
.£ -5
%
c
8
o
p-io
M7 S3
S6
-15.
SI
-U-
J4-
18
21
24
Midnight
M6 S4
M5 52
M2
-J_L
S5
X
12
Noon
Time of Day
15
M3
Ml M4
18
Crop Type: Hybrid of Sorghum (S andM)
- Polarization: VV
Frequency (GHz): 2.75
-a
o
£-io
-15
M5 S4
M2
M4
-20.
Midnight
Time of Day
Noon
Figure 2.6
Diurnal variations of sorghum as measured at 7.25
GHz and 2.75 GHz. Measurements were made with vertical likepolarization for both frequencies (Ulaby and Batlivala, 1976).
<+8
Angle of Incidence (Degrees):
o
0°
9
10°
a
20°
30°
40°
50°
8
10
12
Frequency (GHz)
14
Figure 2.7
Radar sensitivity to soil moisture change for dif­
ferent frequencies and angles of incidence (Ulaby et al., 1979).
49
Further research conducted at many different frequencies in the
microwave
region
has
led to the conclusion that the dependence on
soil surface roughness and vegetation
operating
a
radar
at
8
be
minimized
by
degree
range.
On
the
other
minimize the effects of variations in soil surface rough­
ness and moisture content on the
above
can
a frequency in the 4 to 5 GHz region and at
angles of incidence in the 10 to 20
hand, to
cover
GHz
and
angles
backscattered
power, frequencies
of incidence higher than 40 degrees were
recommended for crop discrimination (Ulaby, 1982a).
Using the results, extensive experiments
with
scatterometers
of
5.04
have
been
conducted
GHz at 20 degree incidence angle for
soil moisture detection, and 10.2 GHz (X-band) at
cidence angle for crop studies.
50
degree
in­
It seems, however, an important is­
sue concerning the use of radar for remote sensing of vegetation has
been raised recently.
The reflected microwave energy from the upper
part of the vegetation canopy
discriminate
crops.
alone
might
not
be
sufficient
to
That is, L- or C-band could be better for crop
discrimination than X-band
because
the
energy
return
at
longer
wavelengths includes contributions from the lower part of the canopy
and the underlying soil.
Shanmugam et al. (1981) conducted a crop
classification
study
using airborne scatterometer data of HH and HV polarization measure­
ments at C-band (4.75 GHz) and L-band (1.6 GHz) acquired over a test
site
near Colby, Kansas during the summer of 1978.
50
The look angles
used in the analysis were 5 to 50 degrees in steps
The
accuracy
level
to
of
5
degrees.
classify corn, wheat stubble, pasture, and
fallow increased as the look angle increased.
Under wet field
con­
ditions, namely, after a rain, the use of HV polarized C-band at a
look angle of either 40 or 45 degrees produced the
classify
the
four
cover categories.
best
result
to
Since the Ku-band (13.3 GHz)
data acquired at the time of the NASA's C-130 overflight was not in­
cluded
in
the analysis, it is difficult to conclude that C-band is
the most suitable wavelength for crop discrimination.
frequencies
The microwave
flown at the time included 13.3 GHz (Ku-band), 4.75 GHz
(C-band), 1.6 GHz (L-band), and 0.4 GHz (P-band) with
angles
from
varying
5 to 50 degrees in steps of 5 degrees each.
quency had HH and
HV
polarization
except
Ku-band
look
Each fre­
which
had
W
polarization.
Besides the Colby test, radar scatterometers with the same con­
figuration
were
flown
Analyses of the airborne
over
several
test sites in 1978 and 1980.
scatterometer
data
obtained
in
Webster
County, Iowa in 1980 showed that the C-band HV polarized measurement
at an incidence angle of 50 degrees is the best of 50 channels
sidered
to discriminate corn and soybeans (Paris, 1983). Rosenthal
and Blanchard (1984) investigated the
obtained
con­
airborne
scatterometer
data
over Guymon, Oklahoma in August 1978 and Dalhart, Texas in
August 1980.
Fields included in the study
were
corn, bare
soil,
sorghum, pasture, wheat stubble, weed, and bare soil at Dalhart, and
sorghum, alfalfa, and
polarization
most
polarized data.
bare
soil
at
Guymon.
The
frequency
and
sensitive to crop type differences was C-band HV
Incidence angles of 40 and 45 degrees were
efficient in discriminating crop type variations.
were only able to distinguish vegetated
P-band (0.4 GHz) data
and
equally
Ku-band responses
non-vegetated
fields.
differentiated between corn and sorghum —
crops with high phytomass — compared to the high microwave frequen­
cies.
These
results
conflict
with results found by Ulaby et al.
(1975, 1982b) which indicated that
greater
than
high
frequency
microwave
data
8 GHz (X-band) can discriminate crops more accurately
than low frequency microwave data.
In summary, the ability of radar to produce vegetation and crop
maps
has
been
demonstrated by SLAR operations, ground-based scat-
terometer experiments, image simulations, and Seasat SAR data.
the
use
But
of radar in crop discrimination and crop growth monitoring
has been slow to come of age (Moore, 1983). As yet there is no con­
sensus
as
to the specifications of radar channels for agricultural
applications. Perhaps, the realization of orbiting microwave system
will
initiate
a
boom in the development and applications of radar
data, as the launch of Landsat did in the application of
MSS
tech­
nology.
2.4 Multisensor Applications to Crop Discrimination
The multisensor approach involves the examination and
correla­
tion of data of the same agricultural scene collected simultaneously
52
by sensors operating in pre-selected portions of
different
regions
of the electromagnetic spectrum (Lancaster and Feder, 1966). At the
present time, sensors operating in the optical and microwave
ranges
are being considered for a multisensor combination.
In addition to near all-weather capability of microwave
systems, sensors
operating in the microwave region provide new in­
formation, because the electrical properties of most
vary with the electromagnetic wavelengths.
target
a
given
have little relationship with each other, thus providing dif­
ferent information about the same scene.
in
scenes
Of particular importance
to crop studies is that optical and microwave signals from
scene
sensor
the
optical
range
may
Scenes that appear "rough"
be "smooth" seen by microwave sensors.
Wheat grain heads, for example,.have strong effects on radar returns
(Allen,
1984), while the optical returns are relatively low as they
are largely influenced by the greenness
of
the
vegetation.
Corn
plant leaves reflect strong infrared radiation in the optical range,
but corn cobs and stalks were found to be a
strong
contributor
to
Calla et al. (1979) reported after a temporal investigation
of
radar backscatter (Ulaby, 1982b).
paddy
rice
using
X-band
radar that a significant change in radar
return had been observed just before and after harvest
during
the
appearance
of the spike.
as
wheat
and
rice.
as
These findings could provide
valuable information to help predict the heading and harvest
of
well
stages
The Heydt and McNair (1973) investigations on
53
rice in the Philippines using temporal Landsat data, revealed a rise
in
reflectance
7.
The rise occurs from planting to a few weeks before harvest,
the
rice
response in all MSS bands, particularly bands 6 and
plants
reflect
increasing
radiation
Therefore, if Landsat data and radar measurements
would
be
possible
to
as
are
they
as
grew.
combined
it
monitor the rice crop more effectively, in­
cluding the heading and harvest stages.
To examine the advantages in combining
data
microwave
and
optical
for crop discrimination, Ahern et al. (1978) experimented with
a 13.3 GHz scatterometer and a nadir-viewing radiometer with landsat
MSS
bands,
both
mounted on an aircraft in Canada.
supervised maximum likelihood classification for
they
As a result of
single-date
data,
reported that microwave and optical sensors provide complemen­
tary information.
When combined, the two sensors
permit
the
most
accurate classifications.
They also found that the most significant
features
the
derivable
from
scatterometer
data
were
the
dual
polarized data at nadir.
Li et al. (1982) evaluated crop classification
combined
Landsat/radar
multitemporal
data.
The
accuracy
using
radar data were
simulated from scatterometer data measured at 10.6 and 14.2 GHz with
an
incidence
angle of 50 degrees near Eudora, Kansas in 1976.
channels were used including 10.6 and 14.2 GHz
polarizations (HH, HV,
radar
data
Ten
with
3
and W), and Landsat bands 4, 5, 6, and 7.
The time gaps in the data set were filled by
5k
linear
interpolation.
Test
site cells were extracted from Landsat data, and were compared
with simulated radar data.
They reported that the best minimal
set
of Landsat MSS and radar scatterometer was bands 5 and 7, plus radar
14.2 GHz HV and W polarization.
This combination of
sensor
chan­
nels produced 96.7 percent accuracy for two dates.
A similar examination was
performed
again
by
Ulaby
et
al.
(1982c) using W polarized 13.3 GHz airborne scatterometer data col­
lected over an agricultural test site near Colby, Kansas
The
major
conclusion
in
1978.
derived was that each sensor system could be
complementary and that the combination of two sensors is superior to
either one alone.
Although it was a land cover classification study rather than a
crop
study, Wu (1980) found an improvement of classification ac­
curacy by merging Seasat SAR digital data
both
with
Landsat
acquired in 1978 over a site near New Orleans.
MSS
data,
In this study,
high contrast Seasat digital data were smoothed using a 7x7
fil­
tering window to make them statistically compatible with Landsat MSS
data.
merged
The smoothed Seasat data were treated as a Landsat
with
Bands
5
and
7.
Based
on
the
method
likelihood classification, the results indicated that
band
and
of maximum
the
addition
of microwave data allowed for further subdivision in the classifica­
tion of forested wetlands which were
data alone.
55
not
identified
with
Landsat
In spite of various techniques employed in crop
analysis,
discrimination
many investigations noted that the accuracy level of crop
classification with radar is dependent on the
acquisition
date
of
data and that the use of multisensor and multidate information could
significantly improve the percentage of correct classification.
studies
Few
involving TM or TM-related data and microwave data for crop
discrimination from a temporal perspective have been available up to
this writing.
56
CHAPTER 3
AN OPTICAL AND MICROWAVE APPROACH TO CROP DISCRIMINATION
The basis for remote sensing technology for crop discrimination
is
in
the
variations
(Chapter 2).
conditions
1979).
the
of
spectral
characteristics
Much of the early work was performed under
generally
to
examine
the
variations of crop spectra.
conditions
would
conduct
to
laboratory
measurements
in
complexity of the spectral and spatial
The understanding obtained under
field
serve better to develop the full potential of the
remotely sensed data acquired from space-borne
contribute
crops
using freshly-picked plants (Robinson et al.,
It is obviously more realistic to
field
among
the
development
of
remote
sensor
systems
and
sensing technology for
agricultural applications.
3.1 Experiment Description
To provide insights
into
the
optical
and
microwave
sensor
operations for crop discrimination, a field experiment was performed
over a number of commercial agricultural fields north
of
Lawrence,
Kansas, in the two summers of 1982 and 1983 over the greater part of
the growth cycles of the various crops selected.
used
for
the
field
measurements
were:
(a) The Barnes 12-1000
Modular Multiband Radiometer (MMR) and (b) Two
Radar Scatterometers (MARS).
57
The sensor systems
Mobile
Agricultural
The sensor systems were mounted on the
boom
of
a
operated by controls in the driver seat of the vehicle.
truck
and
The MMR has
eight optical bands including the Thematic Mapper (TM) bands (Table
2.3) and
both
(X-band) each.
MARS
operated
at 5.04
A photograph of the
MARS
GHz (C-band) and 10.2 GHz
system
in
operation
is
shown in Figure 3.1.
The emphasis of the 1982 portion of the experiment was the
amination
of the effects of the inter-field variation of radar sig­
nals on crop discrimination.
Therefore, ten fields
soybeans, and winter wheat were observed.
each
of
angles
set
corn,
In the early summer, only
MARS-X was available with W and VH polarization configuration
incidence
ex­
with
at 50 degrees. The C-band was added in the
latter part of the summer with HH and HV polarization.
Measurements
were made with incidence angles of 20 and 50 degrees for both bands.
Originally, the higher angle was for
vegetation
modeling
and
the
lower angle was for soil moisture studies (Chapter 2).
Weather permitting, the fields were
week
during
the
months
from
revisited
three
April to November 1982.
weather problems and equipment failures, however, some
was maintained in scheduling the observation activities.
times
a
Because of
flexibility
A measure­
ment log was maintained by each ground truth team and for each
sen­
sor system in order to identify and tabulate the fields measured and
problems encountered, and to
record
any
other
observations
might be helpful in the interpretation of the data.
58
that
Figure 3.1
Photograph of the MARS-X system observing a soybean
test field.
The modified version used in 1983 had both C- and
X-band antennas mounted on the boom. For optical measurements,
the X-band antenna was replaced by the MMR instrument.
59
Selected crop and field
data
were
taken
within
characteristics,
one
i.e.,
ground
truth
hour of the radar observations.
ground truth measurements were designed to obtain sufficient
mation
to
describe
agricultural
scene
weather measurements were obtained from
Weather
Service,
were
University
Daily
of
Kansas
and included wind speed and direction, precipita­
tion amounts, and temperature.
ments
infor­
characteristics.
the
The
recorded
on
a
Supplemental precipitation
daily
basis
established in the vicinity of the test
from
four
site (Figure
measure­
rain
gauges
3.2a).
The
details on radar and ground truth data requirements, and acquisition
procedures are discussed in Aslam et al. (1983) and
(1983) which
also
contain
the
Jung
et
al.
complete listings of the acquired
data.
The 1983 experiment was designed to address specific
pertaining
to
agricultural
crop discrimination from spectral data
obtained from optical and active microwave systems.
ments
were
made
questions
Field
with the MMR and two MARS systems.
channels were employed at this time:
eight
MMR
measure­
A total of 12
bands,
C-band
HH
polarization at 50 degree incidence angle (CHH50), CHV50, XW50, and
XVH50.
The
crops
observed
included
corn,
fallow,
grass,
hay,
potatoes, soybeans, and winter and spring wheat.
The field revisit
interval was approximately ten days.
were
observed
for
While there
ten
fields
each crop in the summer of 1982, one or two fields in
the test site were selected for observation for each crop during the
60
summer
of
1983.
Free Flying
The experiments were conducted in support of the
Imaging
Radar
Experiment
(FIREX)/Shuttle
Active
Microwave Experiment (SAMEX) program (Table 1.1).
3.2
Sensor Systems
MMR is a portable, battery-operated
eight independent radiometers.
lel to each other.
taneously
by
sensor
system
containing
All eight radiometer axes are paral­
Thus, a single target scene can be viewed simul­
all eight channels.
The bands represent the seven TM
bands, plus an additional band in the 1.15 - 1.30 pm range.
strument
was designed as a landsat TM simulator suitable for opera­
tion from a helicopter, small plane, truck,
forms.
The in­
or
from
tripod
The sensor system measures about 23 x 25 x 28 cm.
put signals have an analog scale factor
ranging
from
0
plat­
The out­
to
5
DC
volts, and are recorded automatically onto a polycorder data logger
during field measurements.
The MARS systems
University
of
Kansas
ten years or so.
have
been
developed
and
by
the
Remote Sensing Laboratory (RSL) for the last
MARS-C was operated at 5.04
MARS-X was operated at 10.2 GHz (2.94 cm).
GHz (5.95
cm) and
Both systems have a dual
polarization capability and variable look angles.
(91
modified
The
three
foot
cm) dish antenna was used for C-band, and the smaller one foot
(30 cm) dish antenna was used for the
Table
3.1
summarizes
X-band
radar (Figure
3.1).
the major characteristics of the radar scat-
terometers.
61
Table 3.1
MARS-C/X System Specifications
C-Band
X-Band
Center Frequency:
5.04 GHz
10.2 GHz
Modulation Bandwidth Af:
270 MHz
420 MHz
IF Frequency:
22 kHz
22 kHz
Polarization:
HH, HV
VH, VV
VV and VH
Transmitted Power:
10 dBm
17 dBm
Modulation Rate (FM Rate):
50-900 Hz
20-660 Hz
Modulation:
Triangular
Triangular
IF Bandwidth
Wide Filter
Narrow Filter
6600 Hz
100 Hz
6600 Hz
100 Hz
Antennas:
Parabolic Dish for
HH, HV, VH, VV
Parabolic Dish for
VV
Standard-Gain Horn
for VH
Diameter:
36" (91.5 cm)
12" (30.5 cm)
Feed:
Dual-Pol. Ouad-ridged
Horn - 4-18 GHz
Double Dipole
Product Beamwidth:
Elevation
Azimuth
Elevation
Azimuth
HH
HH
VV
VV
VV
VH
Dynamic Range
70 dB
70 dB
Look Angle in Mobile Mode
10° - 80°
10° - 80°
Height Above Ground
10 m
10 m
Calibration
Internal
External:
Delay Line (16 m)
Luneberg Lens
Delay Line (10 m)
Luneberg Lens
3.28°
3.75°
3.59°
3.30°
62
4.14°
5.86°
Measurements can be made in the stationary sweep-mode, with the
outriggers
down and the boom extended, or in the mobile drive mode,
with the boom raised up and driving the truck very slowly.
3.3. Test Site and Test Fields
The test site is located on the Newman and Holiday terraces
the Kansas River flood-plain north of Lawrence, Kansas.
intensively cultivated and represents a typical
of
non-irrigated
winter
of
The area is
agricultural
wheat, soybeans, and corn.
scene
The criteria
used to select the target fields were as follows:
(1) Fields must be representative of the general
population
of fields at the test site.
(2) Fields must have a relatively uniform surface
condition
in terms of soil type, surface slope, drainage, and planting charac­
teristics, in order to minimize within-field variability.
(3) Fields for each crop should have a range
of
soil
tex­
tures in order to maximize between-field variability. For soil textural classifications of each field, see Jung et al. (1983).
(4) Fields
must
be
readily
accessible,
and
should
be
amenable to the data collection activities of the truck-mounted sen­
sor systems as well as of the ground truth teams.
Test fields were carefully selected
crop
height, density, and
weed
to
minimize
variance
infestation. Fields with a wide
roadside ditch, roadside power lines, and high fences were
No
in
avoided.
field exceeded a gradient of two percent (Dickey, et al., 1977).
63
Figures 3.2a and 3.2b show the locations of the test fields.
3.4 Optical Data Acquisition and Calibration Procedures
3.4.1
MMR Field Operation
The objective of the optical data acquisition procedure
is
to
obtain a property of the agricultural scene which is nearly indepen­
dent of the incident irradiation and atmospheric conditions
time
of
the
measurement.
the
experiment
The field procedures used
closely follow those used since 1974 by the
Purdue University Laboratory for Applications of Remote Sensing
the Field
Research
the
Data should be comparable from time to
time, site to site, and sensor to sensor.
during
at
Team,
and
NASA Johnson Space Center (Robinson and
Biehl, 1979).
Essentially, the field procedure consists of the comparison
the
response
of
the sensor observing a test field to the response
from a level reference
known.
Reference
surface
surfaces
whose
sulfate
reflectance
properties
are
include pressed barium sulfate powder,
painted barium sulfate, and canvas.
barium
of
For this experiment, a
panel (4x4PNL-ll) was used.
painted
Painted barium sulfate
is highly reflective from 0.4 to 2.4 jam range.
If applied properly,
the panel differs no more than five percent from Lambertian for 0 to
55 degree zenith angles. Painted barium sulfate panels are
usually
used as the reference surface for truck-mounted multiband radiometer
systems.
The solar zenith angles (Figure 3.3) normally
encountered
during
allowed
or
data collection by truck-mounted systems are 15
64
ED
^t-Y
if;
•W^ (
'y'
Hi, ,/ /
<afc
ScKMI"
£5*£*ssj
/
J
OS
V
Q Rain Gauges
O Wheat
QCorn
/\Soybean!
^Bar« Soil
/
NOITH UWHNCE QJ
OJ
(Tj
t35
Figure 3.2a
Test site and field locations (1982).
Note that
Soybean Fields 1, 2, 7, 8, and 9 are double-cropped with wheat
and thus correspond to Wheat Fields 2, 3, 5, 6, and 7, respec­
tively.
65
©
M'dlond
©
Municjpol
j
Komot Tumelli#
NORTH LAWRENCE
Mile
Figure 3.2b
Test site and field locations (1983).
Field 14
(alfalfa) is off the map near Perry, Kansas, The numbers on the
map are:
1 - Corn 1
2 - Soybean 1
3 - Hay
4
Wheat 3
5 - Corn 2
6 - Wheat 1
7 - Cut Grass
Soybean
8
Fallow
9 - Wheat 2
10 - Swamp
11 - Potatoes
12
Milo
13 - Soybean 3
14 - Alfalfa
15 - Bare soil
16
66
SUN'S RAYS
SOLAR
ZENITH
ANGLE
SOLAR ELEVATION
OR ALTITUDE
ANGLE
SUBSATELLITE
POINT
Figure 3.3
Definition of solar
(Slater, 1980).
TANGENT
PLANE
EARTH ELLIPSOID
zenith
6?
and
elevation
angles
to 45 degrees (Robinson and Biehl, 1979).
When viewing the reference panel,
filling
the
field of view on the panel.
at a height of about two meters above
reference
care
panel.
To
is
to
ensure
the 4x4 foot (1.2 x 1.2 m)
At this height, it was relatively easy to locate
observed
ensure
taken
The sensor was positioned
the 0.53 meter field of view on the panel.
panel
was
Generally, the reference
every 10 to 15 minutes during the measurements.
measurement
accuracy,
however,
the
white
panel
was
measured every time each field was observed.
To acquire meaningful field
data,
measurements at the appropriate scale.
it
is
necessary
to
make
This entails positioning the
sensor at a proper distance above the target scene, and giving care­
ful consideration to the field of view.
Measurements were made with
the MMR mounted on the MARS truck boom (Figure 3.1).
the
crop,
the
boom
was
raised
meters above the top of the crop
When observing
up to a height of at least three
canopy.
At
a
height
of
three
meters, the diameter of the 15 degree field of view is approximately
0.8 meters.
The field calibration procedure for each field was:
(1) Place a lens cover over the front of the MMR to obtain a
dark
measurement for each band.
All dark level measurements should
be zero or very close to zero because all incoming flux to
the
should
The dark
have
been
blocked by the thick, dark lens cover.
readings are subtracted from the reference panel
68
and
MMR
corresponding
target scene measurements during data calibration.
(2) Remove the dark lens cover and take the readings of
white
reference
panel.
At this time, the reference surface should
be levelled by adjusting bubbles in the corners of the
instrument .should
the
panel.
The
be placed in the middle of the white panel about
two meters above it.
(3) Raise the boom up to a height of at least
above
three
meters
the top of the crop canopy and level the MMR instrument using
the control in the driver seat of the truck.
Start
to
sweep
the
boom over the target area, and record signal returns for each band.
The above procedure was repeated for each
observation,
all
the
data
logger
During
Omnidata
International, Inc.
was programmed to stop when 20 individual signals
were entered from 20 individual spots scanned. Therefore, 20
tially
the
measurements were recorded by a data logger,
Model 516 Polycorder manufactured by
The
field.
distributed
samples were taken per observation.
spa­
Later, the
field measurements including dark readings, white readings, and tar­
get
scene
readings
were transferred to the mainframe computer for
processing.
It is necessary to make sure every time that no shade
either
is
cast
on the reference panel or on the target scene by MMR itself,
boom, or truck.
All field measurements were made
during
conditions or no cloud in the general vicinity of the sun.
and Biehl (1979) reported that, as a large cumulus cloud
69
cloudless
Robinson
approaches
the
solar
disc,
a
sensor
of global irradiance will indicate in­
creased intensity (as much as 30 percent).
Then as the cloud begins
to cover the solar disc, the intensity decreases markedly (typically
to 20 percent of the original
made
during
intensity).
Thus,
any
measurement
these events will be subject to error due to intensity
changes with time.
Clouds that appear on the horizon on
have
otherwise
clear
day
less effect in that they irradiate from a large angle from the
solar disc.
tion
an
of
Bands of clouds covering half the horizon to an
eleva­
20 degrees (zenith of 70 degrees) produce little effect on
the measurement of reflectance
(Robinson
solar zenith
3.3) normally allowed or encountered
angles (Figure
during data collection by
degrees.
truck-mounted
and
Biehl,
operation
was
Most
during 9:30 - 12:00 a.m.
of
the
MMR
measurements
to
45
MMR
channels
(July 19), 1983.
were
made
under cloudless condition.
Figure 3.4 is a plot of the white reference panel
9:30
15
The
This corresponds to 9:30 a.m. - 3:00 p.m. timeframe local
Daylight Savings Time.
by seven
1979).
measurements
from 9:30 to 12:25 a.m. on Julian date 200
It indicates that the measurements obtained before
a.m. may not be adequate due to their time rate-of-change.
the time of day progresses, the panel measurement
increase, but at a slow rate.
70
values
As
gradually
5.0
4-.0
Ch
3.0
Ch
2.0
Ch
Ch
Ch
Ch
Ch
0.0
900
1000
1100
1200
Time of Day
Figure 3.4
Reference panel measurements by hour on day 200, 1983.
1300
3.A.2
Calibration of MMR Data
The discussions contained in this section could comprise an ap­
pendix. However, the data calibration procedures not only are vital
to the quality of data used for analysis but also represent
important aspect of optical remote sensing.
a
very
Therefore, the decision
was made to include the data calibration procedures in the main body
of this report.
The spectral data obtained by the MMR were processed into
parable
units, namely
a
com­
reflectance factor expressed in percent.
The reflectance of a surface depends on both the
direction
of
the
irradiating flux and the direction along which the reflected flux is
detected.
radiant
A reflectance factor here is defined as the ratio of
flux
actually
the
reflected by a sample surface scene to that
which would be reflected into the same reflected beam geometry by an
ideal (lossless) perfectly diffuse (Lambertian) standard surface ir­
radiated in exactly the same way as the sample scene (Robinson
and
Biehl, 1979).
Since the field of view is small (less
than
20
degrees
full
angle), the term bidirectional reflectance factor (BRF) is also used
to describe the measurement.
associated
with
the
It is so called as
one
direction
is
observing angle (usually nadir) and the other
direction is the angle associated with the position of
the
sun
at
the time of the measurement.
If S is the analog output voltage of
s
scene,
and
Sr
observing
the
is for the output voltage of the MMR observing
the
72
the
MMR
level reference surface, then the essential calibration equation
to
produce a reflectance factor (R ) is:
s
R =
s
ss
. R
r
s
r
where Rf is the reflectance factor of the
calibration
of
the
reference
surface.
MMR field data, the following reformatting al­
gorithm was used for each sample of each MMR channel.
was
For
The algorithm
developed at Purdue University and accounts for the differences
in dark level measurements, gain settings, and the sun angle
varia­
tions between the scene and the reference measurements.
R
=
Ss " ds
G cos 0
r. R
—? !. _I
s
S
r
- d
r
G cose
s
s
r
where R : Spectral bidirectional reflectance factor of scene (%).
s
S : Spectral response of MMR to scene (volts).
s
Sr: Spectral response of MMR to reference panel (volts).
dg: Dark level response of MMR nearest in time to the scene
(volts).
dr: Dark level response of MMR nearest in time to the reference
panel (volts).
G : Gain setting of MMR during observation of target scene.
s
Gr: Gain setting of MMR during observation of the reference
73
panel.
0g: Solar zenith angle at time of observation of the target
scene.
0r: Solar zenith angle at time of observation of reference
panel.
R : Bidirectional reflectance factor of reference panel (%).
The bidirectional reflectance factor
panel (Rr) is
a perfect diffuser.
reference
the
white
reference
required to correct for the non-ideal properties of
the reference panel.
April 1982.
of
The
panel
The reference surface is almost, but not quite
The reference panel (4x4PNL-ll) was prepared in
bidirectional
used
for
reflectance
the
measurements
for
experiment were made by the Purdue
University Laboratory for Applications of Remote Sensing on
1982.
Table
3.2
degrees.
and
55
degrees
with
angles
8,
of
10,
a view angle set at zero
The measurements are plotted in Figure 3.5.
The reference panel values (Rr) needed to calibrate a
field
May
includes the reflectance values of the reference
panel for the seven MMR bands for irradiance zenith
20, 30, 40, 50,
the
measurement
were
determined
specific
by first calculating the solar
zenith angle at the time and location of the field measurement, and
then
interpolating
between
the
angles from Table 3.2.
zenith angle at the test site and at the time of each
ment, was
(1978).
calculated
using
the
procedure
reported
MMR
measure­
by Walraven
This procedure is an approximation to an equation
7^
The solar
used
to
Table 3.2
Reflectance Values (%) for Reference Panel (4x4PNL-ll)
Date panel prepared:
April 19R2
Date Panel calibrated:
May 8, 1982
Instrument usee:
Exotech 20C
Location:
Purdue University/LARS Optics Lab
View zenith angle:
0°
Number of measurements:
9
MMR Band
Designation
Wavelength
(ym)
Irradiance Zenith Angle (degrees)
10
20
30
40
50
55
1
0 .45 - 0 .52
96 .5
95 .0
92 .0
91 .0
87 .5
86 .5
2
0,.52 - 0 .60
96 .0
94 .7
91,.8
90 .5
87 .1
86 .2
3
0,.63 - 0 .69
95 .6
93 .8
91 .5
90 .2
87 .0
86,.0
4
0,.76 - 0 .90
94,.7
92 .4
90,.3
89 .5
86 .2
85,.2
8
1,.15 - 1 .30
92,.5
90
88,.1
87 .6
84 .6
84,.0
5
1..55 - 1 .75
88,.2
85,.9
83,.9
83 .5
81.0
80,.5
7
2,.08 - 2 .35
77..8
75,.6
73,.8
73 .7
71 .3
71,.0
Notes:
1.
The MMR band designations correspond with those of the
Thematic Mapper (TM) except Rand 8, which is unique to MMR.
2. Band 6 is the thermal band and is calibrated using different
criteria. See Barnes (1979).
75
10 DEG:
50 DEG:
20 DEG:
55 DEG:
30 DEG:
1.2
40 DEG:
1
Wavelength (um)
Figure 3.5
Plot of reflectance values of the reference
used in the experiment in relation to solar zenith angle.
76
panel
generate
the
Nautical Almanac and computes the position of the sun
rapidly to an accuracy of one hundredth of a degree.
the
test
site
location,
the
For
input
of
grid for Lawrence Municipal Airport
(39°00'30"N 95°13'06"W) was used, because
it
is
located
approx­
imately in the middle of the test site.
Further details on the MMR system
found
in
Barnes (1979), Daughtry
(1979), Robinson and
(1980),
and
and
its
operation
can
et al. (1981), Robinson et al.
Biehl (1979), Biehl (1982), Bauer
Robinson
et
be
al. (1981).
et
al.
The calibrated MMR data are
reported in Appendix C.
3.5
Microwave Data Acquisition and Calibration Procedures
3.5.1
Field Operation of Radar Scatterometers
Because radar scatterometers permit more
of
detailed
observation
radar scattering behavior than radar imagers, most field experi­
ments
use
radar
scatterometers.
calibrated, scatterometers
While
imagers
are
seldom
are almost routinely calibrated.
since scatterometer measurements are
recorded
in
digital
Also,
format,
they are much easier to process and analyze.
In coherent radar systems, either constructive
interference
may
occur
or
between the signal components reflected by
the individual scatterers within the resolution cell of
This
destructive
the
radar.
effect is manifested as a rapid fading (or fluctuation) of the
signal received.
Thus, the
phenomenon
7?
of
fading
is
a
function
directly
1975).
dependent
on
the
target
being
sensed (Bush and Ulaby,
In imaging radars, fading appears in the image as speckle.
This fading observed in the radar return complicates the
of
the
scattering
properties of a particular scene.
stantaneous measurement of scattered power
much fading is actually taking place.
a number of measurements, which can
fading
effects.
The
fading
will
A single in­
indicate
how
Thus it is necessary to make
then
be
averaged
characteristics
Rayleigh statistics, and the number of
not
study
samples
to
reduce
can be described by
needed
to
achieve
various confidence levels in the measurement of the mean backscatter
is shown in Figure 3.6.
Bush and Ulaby (1975) provide a
review
of
fading characteristics and the effects of averaging the radar return
from agricultural scenes.
The radar scatterometers used in the experiment can take
tral
data
spec­
in either of two modes: the stationary sweep mode or the
mobile drive mode.
Since the test fields were distributed over
the
test site, the mobile drive mode was used for the measurements.
The
combinations of frequency (wavelength), polarization, and
incidence
angle were as follows:
C-band
HH, HV
20, 50 degrees
X-band
W, HV
20, 50 degrees
In field measurements,
passed
over
the
truck
with
the
sensors
mounted
the field for each look angle at a speed of about 5 km
78
UJ
-b
co
-8
oc
-n3
VO
-10
UJ
o
2
UJ
Q
o
CJ
-12
-11
-16
-18
J
3
4
I
L
5 6 7 8 910
20
30
10 50
Number of Independent Samples
Figure 3.6
Ninety percent confidence levels for the Rayleigh distribu­
tion (Aslam et al., 1983).
or less per hour while the
angle of 90 degrees.
radar
antenna
continuous
In this
integration
was
mobile
used,
drive
mode
of
samples
were
ob­
at the approximate rate of one sample every time the antenna
moved a distance of d/2, where d is the
The
azimuth
meaning that as the
truck drove alongside the edge of the field, new
tained
the
In other words, the conical axis of the anten­
nas was perpendicular to the road.
operation,
maintained
integration
seconds.
feet.
time
could
be
diameter
selected
from
A typical distance travelled during 19.6
The
diameter
of
the
antenna.
9.8 seconds to 36
seconds
was 60
of C-band antenna is 3 feet and so d/2 is 1.5
feet. The number of spatial samples was therefore: 60/1.5 = 40.
The actual number of independent samples are computed from spa­
tial
averaging
and frequency averaging.
The number of independent
samples provided by frequency averaging increases with angle of
cidence.
At 50 degrees incidence angle, the number of independent
samples per measurement is calculated to be around
responds
in­
to
12,
which
cor­
a 90 percent confidence interval of -2.4 dB to +1.8 dB
relative to the mean (Stiles et al., 1979).
To account for the fading
added
reliability, 30
or
35
characteristics
of
minutes
to
take
and
for
independent samples were taken.
terms of measurement time, it was also adequate; it
five
radar
took
three
In
to
30 or 35 samples per field, while the truck
moved slowly on the road.
The areal samples were averaged
to
com­
pute a single signal value, a0, for each field according to the data
80
processing procedure to be discussed shortly in this section.
3.5.2
Calibration of Radar Data
Procedures to calibrate radar data acquired in
well
the
field
are
established (Stiles et al., 1979 for example). The techniques
employed to process the data acquired for this report are
here.
The
discussion represents an essential part of radar remote
sensing and would be beneficial for geographers as
including
discussed
well
as
others
microwave engineers wishing to understand the calibration
techniques
used
at
the
University
of
Kansas
Remote
Sensing
laboratory and to better understand radar data in remote sensing.
The,performance of a radar is described by the radar
It
relates
the
power
received
by
characteristics and to the parameters
1983).
equation.
the radar to the target scene
of
radar (Fung
and
Ulaby,
The equation is:
Pr =
r
P G G k2o
3 4
(4n) R
(D
where Pr: the average power received by radar (watts)
Pt: the power transmitted by radar (watts)
Gt: gain of transmit antenna
6 : gain of receive antenna
A.: wavelength (m)
o
a: radar backscattering cross-section (m )
81
R: the range from antenna to the scattering target scene (m)
Since
most
distributed
geoscience
scene
rather
applications
involve
a
spatially-
than a point target for ranging, and the
radar cross-section of a spatial target varies with the geometry
illumination,
is introduced.
as
the
of
average differential cross-section per unit area
The average differential cross-section is also known
the radar backscattering coefficient, and is commonly denoted by
o°:
a
CT°
=
(2)
where A is the illuminated area or foot-print of the
calculated
radar
and
is
from the geometry on the basis of measured values of the
radar beamwidths for each frequency-polarization
configuration
and
the range from antenna to the target.
From equations (1) and (2),
P r (4n) 3 R 4
a0 =
-
(3)
PtAAGtGr
As the equation shows, a0 is generally a function of: (1) the
angle
of
incidence relative to the target, (2) microwave frequency
of the incident energy, (3) the polarization configurations
transmit
and
receive
antennas,
of
the
and (4) the geometrical and elec­
82
trical properties of the target scene as discussed in Chapter 2.
When a
wavelength,
radar
and
is
built,
transmitted
polarization are fixed.
radar system and range parameters,
power, antenna
gains,
Aside from the effects of
therefore,
the
average
return
strength varies only with CT°. Hence, variations in tone on a
power
radar image for example are mainly in response to variations in
magnitude
of
a° (Ulaby et al., 1981).
the
The a0 values are computed
through a two-step calibration procedure to ensure that the measured
power return can be accurately related to the radar cross-section of
the target scene.
The first step is to eliminate the relative short
tions
term
varia­
internally in the radar system by replacing the antennas with
a delay line with constant loss.
The second step is to
relate
the
measured power to the actual radar cross-section of the target scene
by referencing all power returns from the test fields to
return
from
a
the
power
target of known cross-section and distance from the
antenna (Moore et al., 1980; Aslam et al., 1983).
In case of the optical sensor MMR, the painted
panel
barium
was used as a reference to calibrate field data.
periment, a Luneberg
because
lens
was
the
calibration
sulfate
In this ex­
target
primarily
of its high cross-section and its relative insensitivity to
orientation.
Other calibration targets
such
as
flat
rectangular
plates, flat circular plates, and corner reflectors may be used in a
laboratory environment, but it is difficult to achieve
83
high
direc­
tional accuracies with them (Ulaby et al., 1982a).
A spherical Luneberg lens is a dielectric sphere whose index of
refraction
varies
of the lens.
as a function of r, the distance from the center
It consists of many concentric dielectric shells, with
the relative permittivity increasing from slightly larger than 1 for
the outermost shell to 2 at the center (Figure 3.7).
permittivity
variation,
a
plane
refracted towards the focal point.
is
present
on
the
wave
incident
lens
and
upon the lens is
If a metallic reflecting surface
back side of the sphere, the wave is reflected
back in the direction from which it is oriented.
Luneberg
Because of the
field
calibration
Further details on
procedures are discussed in
Ulaby et al. (1982a) and Aslam et al. (1983).
For measurement of signal return
from
a
Luneberg
lens,
the
radar equation can be rewritten as:
p1 =
rl
PtGtGr^°l
T~!
(4)
(ta) 3 ^
where P^: the power received from the lens
a^: the true radar cross-section of the lens
Rr: the range from antenna to lens
On the other hand, the delay line power (P^) is defined:
8^
Radar Antenna
Calibration Target
Nonrellecting Pole
Plane Wave
Focal Point
Figure 3.7
Field radar calibration using a Luneberg lens. The
bottom figure shows that a plane wave is focused to a point and
is reflected back (Ulaby et al., 1983a).
85
Pdl =
j2
^
where "L" is the one-way loss factor of the delay line.
If the equation (4) is divided by the equation (5):
Prl
_
pdi
Ft
V/"2
p
t
m3
"1
4
If "k" is introduced to represent the fixed values determined by the
radar system:
P,
a,
=
~
k
--T
'»
R1
dl
G.G L 2 \ 2
where: k =
(4n) 3
By taking the log of the above equation:
l°g Prl - log Pdl = log k + log
- 4 log
Because 0° is a normalized quantity, i.e., a ratio
scattering
cross-section
to
of
the
antenna
the physical cross-section of the il­
luminated area and because its angular response usually varies
several
orders
of
magnitude, it
decibels (dB):
86
is
customary
over
to express it in
a°(dB) = 10 log a0
Hence,
Pdl(dB) = 10 log Prl - 10 log k - 10 log o1 + 40 log ^
(7)
Therefore, the delay line power of a radar is determined from
power
returns
from the lens, radar system parameters, radar cross-section
of the lens (known), and range from antenna to
the
lens.
By
the
same token, for a spatially-distributed actual scene:
Pr
<J°A
=
k
pdi
.
r4
Solving the above equation for 0°:
<,°= JL. A
Pdl
^
By taking the log on both sides:
log a0 = log Pr - log Pj j + 4 log R - log A
Since log a°(dB) = 10 log cr°,
a°(dB) = 10 log P - 10 log P,, + 40 R - 10 log k - 10 log A
r
*dl
87
(8)
The quantities except "A" in
recorded
by
the
equation
are
radar beamwidths and the range.
Figure 3.8.
the
basis
7t
ab
4
where "a" is the major axis (m) and "b" is the minor axis (m).
From Figure 3.8, the range to target (R) is:
h
R =
COS0
To determine "a" from Figure 3.8:
d
tan (0 + —)
2
dj =
h
h [ tan (e + -|) ]
2
h
p
= h [ tan (e —£~) ]
1
a
1
2
=
d1 - d2
88
of
It appears as an ellipse as in
The area of the ellipse is given by:
A =
d9
and
the radar from the target scene and the Luneberg lens.
The footprint (A) is calculated from the geometry on
the
measured
Figure 3.8
Area calculation of the radar footprint (Aslam et al., 1983).
=
h [ tan (0 +
- tan (6
1) ]
where 0: angle of incidence (known)
h: the height of the antenna above ground (known)
pg: the elevation beamwidth (known)
The minor axis (b) can be determined from
(p ) of
a
the
antenna
which
is
the
determined
azimuthal
by
the
beamwidth
radar system
parameters (Table 3.1).
b
=
K
2R tan—5L
2
cos G
2
With the calculation of the radar
from
the
equation (8).
Further
footprint, cr0
details
on
the
is
computed
scatterometer
systems, field operation, and data calibration can be found in Asalm
et
al. (1983) and Gabel et al. (1981). Radar data for the 1982 ex­
periment are documented in Aslam et al. (1983) and
radar
data
the 1983 experiment are listed in Appendix D of this report.
90
for
CHAPTER 4
MICROWAVE SIGNATURES TO DISCRIMINATE CROPS
In this Chapter, an analysis will be made
to
investigate
the
capabilities of selected microwave channels to discriminate selected
crops.
there
Although Landsat MSS is operational in the
are
no
spaceborne
microwave
better understanding of the
needed
behavior
optical
region,
sensors comparable to MSS.
of
microwave
signatures
A
is
before the operation of orbiting microwave sensor systems in
the 1990's.
The analysis in this Chapter uses the data
1982.
The
in
Analysis of the 1982 data
is mainly to examine the capability of active microwave signa­
tures to discriminate crops.
of
developed
1982 data are radar scatterometer measurements from ten
fields each of wheat, corn, and soybeans.
base
base
optical
and
The separate and
combined
capability
microwave signatures to discriminate crops will be
investigated using the 1983 data base in Chapter 5.
The
1983
data
are multiple measurements from one or two fields for each crop using
MMR and radar scatterometers.
An effort to compare the tradeoffs between the number
poral
data
tem­
sets and the number of sensor channels employed will be
made for both data bases.
used
of
Canonical
analysis
techniques
to help look into the structure of the data set.
will
be
The criteria
to evaluate the performance of each channel or date variable will be
the ability to correctly classify different crop types observed at a
91
particular time.
4.1
Data Base and Selection of Date Variables
Because the fields were not visited on a daily basis, there are
time
gaps
between
the
measurements.
between the measurements was 3.6 days.
The average number of dates
The number of fields visited
a day was 11.9 fields on the average.
There have been instances where the gaps between
ments
are
linearly
the
measure­
interpolated to make available a complete time
history of spectral measurements for each crop for its growth
(Steiner,
1?70; Bush
and Ulaby, 1977; Li et al., 1982). For this
study, however, no data values have been interpolated or
All
cycle
generated.
values were obtained from field measurements. For the 1982 ob­
servations, one 0° value was obtained per field per measurement day.
There were 10 fields each for wheat, corn, and soybeans.
Therefore,
10 data values for a given cover type would be available for a given
day.
The
approach is equivalent to extracting one training sample
value for a given cover type from ten different
fields
in
Landsat
imagery data.
Table 4.1 shows the time history of measurements for
data base.
1982
The MARS-X was available with W and VH polarization for
the first half of the summer. From July 26, 1982, a
sion
the
modified
ver­
of MARS was employed and data were collected using eight chan­
nels: CHH20, CHV20, CHH50, CHV50, XW20, XVH20, XW50, and XVH50.
92
Table 4.1
Radar Measurement Log for 1982
April
V
s 16
W1 0
W2 00
W3 0 0
W4 0
W5 0
W6
W7 0
W8 0
W9 0
wic 2
CI
C2 _L
C3
C4
C5
C6
C7
C8
C9
CIO
SI
S2
S3
S4
S5
S6
S7
1
0
0
0
0
0
0
•
S8
S9
S101 ~
0
0
u
W - Wheat
C - Corn
S- Soybeans
/ CHH20 & CHV20
\ XVV20 & XVH20
— CHH50 & CHV50
O XVVSO & XVH50
June
May
July
AuRUSt
September |1 October
1
7 10 14 17 19 21 24 26 28 1 4 7 9 14 18 23
9 12 13 20 23 26 28 30 2 3 4 6 9 16 18 2 9 16 23 28 30 5 7 12 14 19 21 25
0 Q0 0 0
00Q Q Q
00P 0
0 0 0 0pp 0
3
00O0 Q
0 0 o 0u
3 3
0
00 0
0
0 0 £)
3
03
7)
000
0 0 00
3
03
D
0
00 00
3
03
Z)
o
Q
Q
Q
0
3
0
3
b
3
3
0
9
r\
3 Q Q0
00
0
y
Q Q0
0 0 0 03
3
00 Q
0
0 0
ft ft ft ft ft ft
3 o 59 ft ft ft
3
Q0 o
0
3 0 9BBBI
0 & 9 9R 0
1 1 DDDDI OOlODDDDDDDDDI m mnlRORlRl 1 iRlftlselRiftlAI^ I! II II
D
0 3 oQ 0
ft
0 3
Q 3 33 &
o 3 0o
0
00 3
0 3 o0 0
0 D 5 3 o o 05 3 0 3 0 0
o 3 O o ft ft $
ft
$ ft
o 0 3 ft ft ft $
3 d00 0 3 0 0 0
ft f?
Q
o
o 3 ft ft& 3
0o 0 0 0 0 3
0
o0
0p 0
0 3 ft ft ft & ft
y
03 3
_j
00 £)
03
J
$ ft ft& ft ft H
! ft
R 0 ft # ft
ft ft ft s
ft ft a
o R R ft ft ft
&
ft <?slos
3 0 0 p 0 33
i
0 3 r>3 3o R R
ft S) ft
i
3 0 3 0 o3 0 3 R R cakti ft
ft S ft
ft
0 3 0 3 3 0 o3 3 3 0 ft © ft
ft
ftN>R
s100^1
3
ft $
ft
f? foferl
ft
ft
,k
&
jj)
0.•i3 Q
—
2 QL _ J
I I I
_L
—
'
__ _
&
&
"
•
In order to examine the most significant dates to
discriminate
crop types, seven "Dates" have been selected based upon the interval
between measurement dates, the number of sample values available
the
in
timeframe, cover types observed, and the number of sensor chan­
nels used.
three
Note that a Date in Figure 4.1 actually involves two
measurement
each field.
cover
days
to
include at least one sample value from
It was assumed that
the
characteristics
of
a
type would not change significantly over a few days.
terval between the Dates
venience's
sake,
is
each Date
measurement date.
or
approximately
is
usually
two
weeks.
given
The in­
For
con­
represented by the first
The sensor channels used in
the
experiment
are
analysis,
the
also indicated in Figure 4.1.
Based largely upon the crop types available for
data
set
of
analysis.
cludes
the
seven Dates was divided into five timeframes for
Timeframe 1 has five Dates from May 19 to July 9 and
wheat
and
corn.
This is the entire period when wheat and
corn have been observed simultaneously.
from June
16
to
July
corn, and soybeans.
observed
because
tually bare.
9.
Timeframe 2 has three Dates
Crops observed at this time are wheat,
Before this timeframe, soybean fields were
plants
in­
not
were less than 10 cm and fields were vir­
Wheat was harvested and
not
available
for
analysis
after this timeframe.
Timeframe 3 is one Date with four crops — wheat, corn, regular
soybeans,
soybeans
double-cropped
94
after
the wheat harvest.
This
Timeframe
Crops:
Wheat
v.-..•••>••'•iW-.i
Soybeans
Dates:
5/19
5/21
Channels:
6/1
6/16
6/25
7/9
7/26
8/16
6/2
6/1
6/28
7/13
7/28
8/18
XVV50
XVV50 XVV50 XVV50 XVV50
XVV50 XVV50
XVH50
XVH50 XVH50 XVH50 XVH50 XVH50 XVH50
XVV20
XVH20
CHH50
CHV50
CHH20
CHV20
Figure 4.1
Crops and
channels for 1982.
study
dates
95
with
corresponding
radar
timeframe provides an opportunity to evaluate the four
The
double-cropped
soybeans
types.
were treated as a separate cover type
from regular soybeans because they
For
crop
have
different
growth
cycles.
instance, regular soybeans were already about 40 to 50 cm tall
resulting in about 80 percent ground cover, while the double-cropped
soybeans
were
about 10 cm tall in the wheat stubble and bare back­
ground at this time (Appendix E; Jung et al., 1983).
Timeframe 4 covers three Dates and three crops of corn, regular
soybeans,
and double-cropped soybeans.
This timeframe allows us to
analyze the three crop types with three temporal perspectives.
last
timeframe
is
one
Date
of August 16/18 for three crops with
eight scatterometer channels for analysis.
is
the
last Date
of
The
This timeframe
actually
Timeframe 4. However, Timeframe 4 involves
XW50 and XVH50 and Timeframe 5 has eight radar channels for evalua­
tion.
The temporal variation of the
backscattering
coefficients
of
the crops considered at different stages of growth is represented in
Figure 4.2.
deviation
The vertical bars in the figure
indicate
±1
standard
each. Descriptive statistics for the selected timeframes
are shown in Table 4.2.
4.2 Techniques for Analysis
The performance of an individual or a
evaluated
group
of
channels
was
according to their capability to correctly classify given
crop categories.
The classification accuracy
96
of
a
channel
or
a
-5
XW50
-10
®-15
"D
XVH50
MD
-O
S-20
- wheat
— soybeans
-25
. soybeans
double-cropped
-30
May 19
Figure 4.2
June 1
June 16
June 25
July 9
July 26
August 16
Temporal variations of radar returns (mean values ±1 standard deviation)
from selected crops at selected times (1982).
Table 4.2
Summary of Radar Measurements for the Five Timeframes
Wheat
vO
OD
Date
Channel
Hay
19/21
XVV50
XVH50
June
1/2
June
16/18
s.d.
COVA
-15.8
-25.0
2.4
1.7
-1.5
-0.7
XVV50
XVH50
-16.2
-25.8
2.7
2.5
XVV50
XVH50
-13.3
-22.3
-U.7
-22.2
COVA
-12.1
-25.1
2.5
2.4
2.1
-1.0
-1.7
-1.0
-12.8
-24.2
1.1
1.9
-0.9
-0.8
1.2
1.6
-0.9
-0.7
-10.1
-21.5
0.9
0.8
-0.9
-0.4
1.5
1.8
-1.3
-0.8
-7.7
-19.2
1.5
1.8
XVV50
XVH50
-8.7
-16.0
XVV50
XVH50
x
Soybeans, n .
s.d.
COVA
-14.1
-26.9
2.5
2.3
-1.8
-0.8
-1.9
-0.9
-B.3
-17.3
1.3
2.7
-1.6
-1.5
1.6
1.7
-1.8
-1.1
-10.1
-17.6
1.5
1.4
-9.4
-17.3
1.3
1.2
-1.4
-0.7
-8.7
-16.1
August XVV20
16/18 XVH20
-4.9
-13.5
3.2
1.9
-6.7
1.4
CHH50
CHV50
-9.1
-17.2
2.4
1.7
CHH20
CHV20
-5.0
-14.2
3.0
1.9
July
26/28
x
Soybeans
s.d.
July
XVV50
9 / 1 2 / 1 3 XVH50
x
Corn
s.d.
COVA
-11.2
-19.9
2.4
1.3
-2.1
-0.6
-1.5
-0.8
-11.5
-20.0
3.0
1.6
-2.1
-0.8
1.8
1.9
-2.0
-1.2
-lfl.O
-lfl.l
1.7
1.3
-1.7
-0.7
-4.2
-13.7
0.7
1.4
-1.7
-1.0
-5.3
-15 .R
1.61.1
-3.0
-0.7
-2.6
-1.0
-11.1
-14.8
2.5
3.0
-2.3
-2.0
-12.(1
-18.0
2.8
3.3
-2.3
-1.8
-6.1
-1.3
-4.1
-13.8
3.2
2.9
-7.7
-2.1
-6.8
-16.3
2.5
2.9
-3.6
-1.8
Notes: 1 . x - sample mean
2.
s . d . - standard deviation
3.
COVA - c o e f f i c i e n t o f v a r i a t i o n .
4.
X
Sample size - wheat ( 1 2 ) , corn ( 1 4 ) , soybeans ( 1 0 ) , double-cropped soybeans (10)
COVA =
x 10
x
group of channels was measured by linear discriminant analysis.
Linear discriminant analysis defines
based
on
a
canonical
a
discriminant
transformation of the original variables to
maximize the separability of observations representing two
categories.
Then
function
or
more
the classification functions are superimposed on
the hyperspace formed by the canonical axes to classify unknown
servations
provides
Linear
into
an
one
of
introductory
discriminant
the
known
categories.
explanation
analysis
is
of
ob­
Johnston (1978)
discriminant
analysis.
superior to the Bayesian maximum
likelihood classifier in accuracy, time, cost, and nonsensitivity to
the
statistical
variance
and number of variables (Tom and Miller,
1984).
In remote sensing research, linear
especially
useful
because
discriminant
The discriminant func­
derived from the original variables are uncorrelated and each
function is defined to produce the 'best' analysis of variance.
is
is
highly correlated multichannel data are
reduced to a more manageable linear problem.
tions
analysis
It
a method of producing hybrid variables so as to produce the best
possible
separation,
or
discrimination,
among
the
various
categories.
Discriminant analysis assumes that the
multivariate
observations
follow
a
normal distribution and that they have equal variances
for each category.
However, it is not seriously affected by limited
departures from the assumptions (Davis, 1973). In practice, the as­
99
sumptions have been proven to be quite robust over a wide
range
of
applications (Duda and Hart, 1973; Merembeck and Borden, 1978).
For
input
to
the
discriminant
analysis,
the
multichannel
responses obtained in the field were coded using the format shown in
Appendices C and D.
BMDP7M (Jennrich
each case.
A
jackknife
validation
option
available
in
and Sampson, 1981) was used to reduce the bias in
This procedure determines the category to which each ob­
servation belongs by using a classification function for each obser­
vation which does not include the particular observation about which
a
decision is being made.
to reduce the bias
that
This technique is particularly important
results
from
classifying
the
same
in­
dividuals used to determine the classification function.
The BMDP7M stepwise linear discriminant analysis operates in
stepwise
fashion
and
chooses variables to enter or to remove from
the discriminant function at each step based on the
ratio
of
among
a
F-statistic,
a
categories variance to within categories variance.
The variable with the largest F at any
time
is
entered
into
the
function, provided the F is above the cut-off level.
A very useful way to look into the structure
data
is
of
multiresponse
to transform the variables into canonical variates and in­
vestigate the responses on a space formed by the first two canonical
axes.
The
multivariate
technique
of
statistics
canonical analysis (Seal, 1964) uses the
associated
with
categories
of
interest
within the data to find an orthogonal linear transformation C of the
100
form:
(1)
Y = CX
where Y is the transformed q-vector, C
that
defines
data vector.
statistically
majority
of
is
the
coefficient
matrix
the canonical discriminant axes, and X is a p-channel
The equation defines
independent
the
a
relatively
discriminant
discriminant
axes
small
that
number
of
explain
the
the
variance
associated
with
by
finding
the
categories.
The solution for C is found
first
A,
q x q
diagonal matrix of eigenvalues of the equation (1) and is the trans­
formed among categories covariance matrix.
tial
equations, the
among
categories
By calculus of differen­
matrix
A
and
the within
categories matrix W for the original variables are changed into
the
canonical form defined by the characteristic equation:
|A - AW|= 0
where \ works as the
Lagrange
(2)
multiplier. To
yield the spherical
unit variance, the above equation can be rewritten:
101
|W1/2AW"1/2 -
XI 1=
0
(3)
To enforce the condition that the among categories covariance matrix
is maximized in the transformation,
CAC1 = A
is imposed.
(4)
To minimize the within categories variance, the
within
categories covariance matrix for the transformed variables is set to
the q x q identity matrix I to scale the variances into unity:
CWC' = 1
(5)
Under the above conditions of (A) and (5), C is found
plete the transformation of equation (1).
variance-covariance matrices.
are
equal
to
related.
Because all the off-diagonal elements
are
are
variables.
orthogonal,
and
This means
hence
uncor-
The diagonal elements of A are chosen such that:
X 11 > A 22 >
These
com­
Both A and I are diagonal
zero, all covariances are equal to zero.
that the transformed variables
to
the
among
k
categories
Therefore,
the
qq
variances
first
102
for
canonical
the
axis
transformed
explains
the
greatest variance and the second
variance and so on.
axis
explains
the
largest
It is not uncommon to have in excess of 95 per­
cent of the discriminatory variance explained on the
three
next
first
two
or
canonical axes regardless of the original number of variables
(Gnanadesikan, 1977).
Therefore, the patterns
represented
by
the
first two canonical axes will generally show the separations of crop
categories from the best two-dimensional view.
Canonical variate analysis can be applied in two ways in remote
sensing
data
analysis.
One is to transform the original variables
into a few significant canonical axes and then define classification
functions
to
classify
observations into various categories.
Many
"canned" packages do not scale the observations of each category
to
unity, however (Merembeck and Borden, 1978).
The other approach is
original
or
selected
to
determine
the
correlation
between
scene variables (height, soil moisture, leaf
area index, etc.) and statistically significant canonical axes.
The
interpretation of the correlation between the variables and the axes
is that variables highly correlated with the high-variance axes
are
important (Horton et al., 1968). For excample, if an axis is highly
correlated with variables characterizing corn, then the axis can
termed
"corn
axis".
In combination with other important axes, the
corn axis can be used for classification or for examination
sources
be
of
the
of correlation between sensor channels and the target scene
characteristics.
Although
ground
103
truth
measurements
were
sophisticated (Jung et al., 1983), the latter attempt to examine the
relationship between scene variables and
1982 data base was not successful.
sensor
channels
for
the
Canonical analysis in this study
therefore is used to graphically represent the confusion
status
of
various scene categories when a number of channel variables are com­
bined in the analysis.
Applications of the canonical techniques
to
remote sensing are found in Maxwell (1976), Merembeck et al. (1976),
Merembeck and Borden (1978), Turner et al. (1978), Jenson and
(1979),
and
Merembeck
Waltz
and Turner (1980). The geometric aspect of
the canonical transformation is explained in Figure 4.3.
4.3 Timeframe 1: May 18 - July 9, 1982
(Wheat and Corn)
The first five Dates indicated along
Figure
the
horizontal
axis
4.2 constitute Timeframe 1. Since two channels of XW50 and
XVH50 are used, the dimension or the number of channel variables
Timeframe 1 data is:
timeframe
which
F-ratio:
variance.
has
that
variance
from
each
to
within-categories
XW50 values are higher than XVH50 returns
for all crops observed (Figure 4.2).
are
variable
the largest discriminatory power was based on
among-categories
Notice
of
5 dates x 2 channels = 10.
The technique employed to select the best
the
in
It is obvious that XW50
data
better than XVH50 signatures for discriminating wheat and corn.
It seems that XW50 data obtained either on June 16 or
clearly discriminate between the two crop categories.
104
July 9
can
Indeed, XW50
Figure 4.3
Canonical transformation in two dimensions with
four crop categories. Arbitrary crop categories A, B, C, and D
are plotted against original axes in solid lines and transformed
axes in dashed lines.
The ellipsoidal shapes representing
within-categories dispersions have been scaled to unit variance
as shown in circles. The original axes have been shifted and
rotated in the transformation to provide better separation among
categories.
For example, categories A and B are confused to
some degree on X. axis but are separable on the transformed Y1
axis (after Jenson and Waltz, 1977).
105
returns obtained on June 16 has an overall
classification
accuracy
level of 96.2 percent for wheat and corn (Figure A.4).
Classification accuracies for each
channel
first timeframe are represented in Figure 4.4.
variable
for
the
It is interesting to
note that like-polarization (XW50) performs far better than
polarization (XVH50) for wheat and corn discrimination.
cross-
The XW50
returns achieved such a high accuracy throughout the period that ad­
dition
of
another
variable
does
overall accuracy of classification.
the
not
significantly
Rather it could
improve the
add
noise
to
discriminant information resulting in decrease in accuracies as
indicated by June 1 and July 9 results:
XW50
and
XVH50 is
lower
the
combined
accuracy
of
than that of XW50 alone for these two
dates (Figure 4.4).
As Figure 4.4 shows,
using
the
therefore,
any
single
day
XW50 channel during the two month timeframe from May 19
to July 9
can
discriminating
achieve
between
better
wheat
than
and
80
corn.
percent
The
accuracies
Overall, the
performance
of
distinctly
to
the
average.
XVH50 increases as wheat matures and
corn grows healthier (Figure 4.4).
more
for
use of XVH50 alone
achieves an accuracy level of less than 60 percent on
responds
measurement
This
may
suggest
that
XVH50
lush corn and mature wheat than does
XW50.
Ground truth that depicts wheat and corn growth cycles is shown
in
Appendix
A.
It
was noted that the canopy height of the wheat
106
vO
vO
CT\
00
00
CO
CT\
00
oo
o
oo
oo
oo
<r
00
vO
00•
o
00
00
—
CO
m
m
vO
i—H
vO
CO
C*4
May
19
June
1
June
June
16
25
July
9
Figure 4.4 Performance of XW50 and
XVH50
channels
to
discriminate wheat and corn during the May 19 - July 9, 1982
timeframe. For each Date, the first bar represents the correct
classification level for XW50 and the second bar for XVH50.
The third bar of each Date indicates the combined classification
accuracy.
107
plant saturates around May 25 and that the
starts
to
saturate
around
July 15.
canopy
height
end
of
May
while plant moisture content of corn rapidly increases at this
time.
due
corn
However, top leaf height and
plant moisture content of wheat start to drop from the
1982
of
The relatively high level of classification accuracy
to
may
be
this contrast between the two crops — one becoming dry and
another lush.
4.4
Timeframe 2: June 16 - July 9, 1982
(Wheat, Corn, and Soybeans)
The
next
Analysis
timeframe
involves
and July 9.
timeframe
The
is:
data
number
contains
wheat,
corn,
and
soybeans.
acquired on three Dates: June 16, June 25,
of
channel
variables
2 channels x 3 dates =6.
As
involved
in
this
the soybean crop is
added, visual evaluation to select the best single variable to
dif­
ferentiate the three categories of crops is not an easy task (Figure
4.2 and Table 4.2).
suggest
what
Furthermore, it
becomes
difficult
even
to
variable would be the next best to use in combination
with the first variable, because there is a certain degree
relation among the variable (Table 4.3).
of
cor­
This type of analysis will
be handled using a stepwise F-statistics procedure available in
the
BMDP package (Jennrich and Sampson, 1981).
The importance of each of the
dicated
in
Table
4.4 according
descending order of importance.
six
channel
variables
is
in­
to their discriminatory power in
The initial F-values
108
indicate
the
Table 4.3
Correlation Matrix of XVV50 and XVH50 for
Wheat, Corn and Soybeans
JUNE 25
JUNE 16
XVV50
XVH50
XVV50
XVH50
JUNE
XVV50
16
XVH50
0.83
1
JUNE
XVV50
0.36
0.31
25
XVH50
0.34
0.39
0.89
JULY
XVV50
0.07
0.18
0.09
0.25
9
XVH50
0.48
0.41
0.12
0.10
(Timeframe 2:
JULY 12
XVV50
XVH50
1
1
June 16 - July 9, 1982)
109
1
1
0.39
1
relative
amount of araong-categories variance which is accounted for
by each channel.
variable
The
reduction
entered
after
each
channel
Note that XVH50 and XW50 variables of June 25 are not
in the procedure because their F-values to enter fall below
the cut-off F-value of 1.
the
F-values
is entered is due to the correlation between the variables
(Table A.4).
to
in
F
statistics
The F-to-enter for a variable corresponds
computed from a one-way analysis of variance
(ANOVA) on the variables for the categories
used
in
the
analysis
(Jennrich and Sampson, 1981).
Within timeframe 2, therefore, XVH50 data set taken
16
has
from
jackknifed
classification
the
stepwise
It
should
be
noted
procedure in selecting channel variables is op­
timal in terms of the F-ratio criterion
mentioned
necessarily in terms of classification accuracy.
not test each subset of data for optimality.
wise
show
XW50 accuracy of the same Date is higher than XVH50 accuracy:
75.0 percent over 66.7 percent (Figure 4.5).
that
June
the single largest discriminatory power based upon F-ratio.
However, the accuracy levels
that
on
above,
but
not
The procedure does
Therefore, the
step­
procedure should be considered not as an maximal procedure but
as an efficient technique to locate the
important
variables (Tat-
suoka, 1971).
Within the timeframe from June 16 to July 9, the dual polariza­
tion
data
taken
88.9 percent.
on
June 16 achieved a classification accuracy of
This is comparable to 91.7 percent achieved using all
110
Table 4.4
Importance of Channel Variables to Discriminate
Wheat, Corn, and Soybeans
Step
Variable
Initial
F-Value
Cumulative
Entered
F-Value
to Enter
Classification
Accuracy (%)
1
XVH50 (6/16)
37.59
37.59
66.7
2
XVV50 (6/16)
21.97
30.71
88.9
3
XVV50 (7/9)
26.53
12.33
91.7
4
XVH50 (7/9)
15.24
5.89
91.7
5
XVH50 (6/25)
8.49
—
—
6
XVV50 (6/25)
3.78
—
—
(Timeframe 2:
June 16 - July 9, 1982)
111
six
variables
although the last three variables are not entered in
the procedure (Table 4.4).
data
were
used,
the
reached 83.3 percent.
a
combination
of
only 69.4 percent.
When all three
classification
temporal
like-polarized
accuracy for the three crops
This accuracy is higher than the result
from
three temporal cross-polarized returns which was
As Figure 4.5 shows, like-polarization
performs
better than XVH50 on June 16 and July 9.
In an effort to understand and utilize the complementary nature
of
dual
polarization,
a
depolarization
ratio
depolarization ratio represents the ability of
soil
to
a
was
crop
transform incident radiation in one plane
tested.
The
canopy
and
of polarization
into the orthogonal plane of polarization compared to
the
original
plane (Paris, 1983). A depolarization ratio, D is:
D = XW50 / XVH50
D(dB) = XW50 - XVH50
The overall accuracies of classification using the
tion
ratio
depolariza­
were 83.3, 80.6, and 61.1 percent for June 16, June 25,
and July 9 respectively. The depolarization ratio does not
classification
accuracy.
To
compute
the
ratio, however,
polarization is used. The classification level from
of
dual
improve
a
dual
combination
polarization was 88.9, 77.8, and 72.2 percent for the same
Dates (Figure 4.5). In case of the observations made
112
on
June
25,
00
00
in
r^
<N
r-»
\D
<D
cn co
vO
June 16
June 25
•
oo
m
July 9
Figure 4.5
Performance of XW50 and
XVH50
channels
to
discriminate wheat, corn, and soybeans during the.June 16 - July
9, 1982 timeframe. For each Date, the first bar represents the
correct classification level for XW50 and the second bar for
XVH50. The third bar of each Date indicates the combined clas­
sification accuracy.
113
notice
the fact that an accuracy level of 80.6 percent was achieved
from the depolarization ratio of XW50/XVH50.
because
The
improvement
is
the classification accuracy for soybeans changed from 60 to
70 percent (Table 4.5). It seems that the depolarization ratio
may
be used to separate a specific crop category from others.
Overall, employment of X-band
dual
polarization
on
June
16
turns out to be the most practical approach to classify wheat, corn,
and soybeans during Timeframe 2.
4.5
Timeframe 3: July 9/13, 1982
(Wheat, Corn, Regular Soybeans, and Double-Cropped Soybeans)
During Timeframe 3, the X-band
radar
acquired
backscattering
data from ten fields each of wheat and corn, and five fields each of
soybeans and double-cropped soybeans.
harvested
leaving
40
By this time, wheat has
to 50 cm tall stubble in the field.
soybeans had an average height of 45.3 cm
beans were 10.7 cm tall on the average.
and
been
Regular
double-cropped
soy­
They had been planted among
the wheat stubble by the "no till" method which
minimally
disturbs
the soil during planting.
Overall classification accuracies are indicated in Figure
4.6.
Both polarizations have similar performance levels.
The highest ac­
curacy was 65.2 percent obtained by combining XW50
and
the
four
crop
categories.
4.6
for
The low level of classification may be
attributed to the growth stages of
Table
XVH50
crops
observed
shows the classification status.
114
at
this
time.
It seems quite natural
Table 4.5
Performance Comparison of Depolarization Ratio and
Combination of XVV50 and XVH50
Classified As:
Actual
Category
Wheat
Corn
wheat
83.3 / 83/3
16.7 / 16/7
corn
0.0 / 0.0
85.7 / 85.7
soybeans
0.0 / 0.0
40/0 / 30.0
Soybeans
0.0
/ 0.0
14.3 / 14.3
60.0 / 70.0
(Date:
Notes:
1.
2.
Sample
Size
12
14
10
June 25, 1983)
The upper figures indicate correct classification in
percent achieved from a combination of XVV50 and XVH50.
The lower figures represent the classification accuracy
obtained from the depolarization ratio.
Notice that the accuracy level for soybeans changed from
60 to 70 percent when the depolarization ratio is used.
115
65.2
50.0
45.7
43.5
XW50
XVH50
XVV50
+
XW50
XVH50
XVH50
Figure 4.6
Performance of XW50 and
XVH50
channels
to
discriminate wheat, corn, regular soybeans, and double-cropped
soybeans on July 9, 1982.
116
Table 4.6
Confusion Matrix among Wheat, Corn, Regular Soybeans,
and Double-Cropped Soybeans
Classified As:
Sample
Actual
R. Soybeans
D. Soybeans
Size
Wheat
Corn
Wheat
66.7
8.3
8.3
16.7
12
Corn
14.3
71.4
14.3
0.0
14
R. Soybeans
0.0
30.0
60.0
10.0
10
D. Soybeans
20.0
10.0
10.0
60.0
10
Category
(Date:
Notes:
1.
2.
July 9, 1982)
The figures represent correct classification in
percent achieved by combining the XVV50 and XVH50
channels used on July 9.
Overall correct classification: 65.2%.
117
that wheat stubble fields are confused with
fields, because
the
double-cropped
soybean
latter were emerging among wheat stubble.
the same time, the wheat stubble field had
some
green
At
vegetation,
namely, emerging wheat plants from stray wheat grains which had fal­
There
len during the harvest operation.
between
corn
and
is
also
some
confusion
soybeans due to lush biomass and the plant water
content of the crops (Allen, 1984).
4.6
Timeframe 4: July 9 - August 16/18, 1982
(Corn, Regular Soybeans, and Double-Cropped Soybeans)
The timeframe from July 9 to August 16 is
soybeans, and
double-cropped
soybeans
when
corn,
regular
were observed. Figure 4.7
shows the performance of the XW50 and XVH50 channels separately and
jointly
during this timeframe.
Corn is confused with soybeans, and
soybeans are
consistently
confused
during
period.
this
this
saturate
in
canopy
At
with
time,
height, while
double-cropped
soybeans
regular soybeans mature and
double-cropped
soybeans
grow
rapidly to finish their growth cycle within a limited time (Appendix
A). Double-cropped soybean fields have such dense broad leafs
that
they have a biomass comparable to that of regular soybean fields.
In fact, the double-cropped
weeds
that
soybeans
so
infested
with
a weed defoliation operation was conducted to determine
the extent of weed infestation and its
radar
were
backscatter.
possible
influence
on
the
When weeds and soybeans were separated from the
radar swath of Soybean Field 2 (Figure 3.2a), wet biomass
118
of
weeds
00
tn
00
00
v£>
VO
VO
vO
<r
<N
00
CO
July 9
CM
ci
July 26
m
cn
August 16
Figure 4.7
Performance of XW50 and
XVH50
channels
to
discriminate corn, regular soybeans, and double-cropped soybeans
during the July 9 - August 16, 1982 timeframe. For each Date,
the first bar represents the correct classification level for
XW50 and the second bar for XVH50. The third bar of each Date
indicates the combined classification accuracy.
119
in
2
2
the field was 0.35 kg/m , while that of soybeans was 1.16 kg/m ,
a ratio of 1 : 3.3. However, the
based
on
moisture
content
of
the
weeds
wet weight was 61.22 percent, while soybean moisture con­
tent was only 24.30 percent (Table 4.7). It has been reported
free
that
water hanging in a vegetation canopy affects radar backscatter
(Attema and Ulaby, 1978; Allen, 1984).
conducted
late
in
the
As the weed defoliation
summer, it is certain that the weeds had a
degree of influence on the radar return.
radar
return
from
was
This might have caused the
the double-cropped soybeans to be confused with
regular soybeans.
It is interesting to note that the cross-polarized channel per­
forms
better
than
the
like-polarized
Rosenthal and Blanchard (1984) also
channel.
reported
Paris (1983) and
that
C-band
cross-
polarized
radar seems to be the best among Ku-, C-, L-, and P-bands
for
classification.
crop
timeframe
are
used
in
When
all
channel
variables
in
this
the stepwise procedure, it turned out that
XVH50 of July 26 had the largest discriminatory power,
followed
by
XVH50 (July 9), XVH50 (August 16), XW50 (July 9), XW50 (July 26),
and XW50 (August 16) in descending order of importance.
XVH50 of July 26 alone had an overall
of 61.8
percent.
classification
accuracy
When XVH50 of July 9, the next best channel for
discrimination, was added to XVH50 (July 26) in the jackknifed clas­
sification
procedure, the overall accuracy level increased to 73.5
percent.
120
Table 4.7
Biomass Comparison of Soybeans and Weeds in Soybean Field 2
Soybeans
Weeds
24.30
61.22
0
Wet biomass density (kg/m )
1.16
0.35
Dry biomass density (kg/m^)
0.88
0.14
Plant water density (kg/nr)
0.28
0.22
Plant water content (%)
Notes:
1.
2.
The measurements were made on October 21, 1983.
Plant water content is based on wet weight.
121
Since six channel variables are
data
involved
for
three
temporal
sets in this timeframe, it is not easy to follow the crop con­
fusion status as the next best channel
sification
procedure.
As
variance.
combined
in
the
clas­
mentioned in Section 4.2, the first two
axes from canonical transformation
discriminatory
is
provides
Therefore,
the
majority
of
the
the patterns of observations
represented by the first two canonical axes will
show
the
separa­
tions of crop categories from the best two dimensional view.
Figures 4.8a, 4.8b, and 4.8c show
first
two
canonical
axes
the
for Timeframe
radar
4.
returns
on
the
The circles in the
figures indicate the upper limit of the 90 percent confidence inter­
val.
The
ellipsoidal
shape
of the data clusters in the original
population are supposed to be represented by a circle
two
dimensional
canonical space.
canonical
of
unity
in
space and a sphere in three dimensional
The data points are scattered out to
some
degree
in the plot because samples are not a true representation but an ap­
proximation to a population.
A few outlying values for soybeans may
reflect irregularities of the target scene, or internal fluctuations
in the radar system, or possible inconsistency in the field calibra­
tion.
The canonical plots show graphically the
channel
combinations.
Figure
4.8a
shows
performances
the
of
the
confusion overlap
between the three categories using XVH50 on July 26th and July
9th.
Figure 4.8b shows that addition of XVH50 on August 16th improves the
122
Channels used:
XVH50 (7/26)
XVH50 (7/9)
®- CORN
SOYBEANS
+- SOYBEANS a
Double-cropped soybeans
(80%)
corn
2-i
Regular soybeans
(70%)
Overall correct
classification: 73.5%
-31
-4
-3
-2
canonical
axis
1
Figure 4.8a
Crop confusion status represented on the first
two
canonical axes when two temporal measurements by XVH50 are combined for
classification during the July 9 - August 16, 1982 timeframe.
Channels used:
XVH50 (7/26)
XVH50 (7/9
XVH50 (8/16)
®- CORN
a- SOYBEANS
+- SOYBEANS a
Double-cropped
soybeans
yt
(90%)
/
corn
0-
Regular soybeans
(70%)
-2
-a
Overall correct classification:
79,4%
-A
-3
-2
-1
0
canonical
1
axis
1
Figure 4.8b
Crop confusion status represented on the first
two
canonical axes when three temporal measurements by XVH50 are combined
for classification during the July 9 - August 16, 1982 timeframe.
Channels used:
XVH50 (7/26)
XVH50 (7/9)
XVH50 (8/16)
"XVV50 (7/9)
o- CORN
SOYBEANS
+- SOYBEANS 0.
Doublecropped
soybeans
(90%) /
corn
(92.9%)
0-1
Regular soybeans
(60%)
-2
Overall correct
classification: 82.4%
-31
-4
-3
-2
canonical
axis
1
Figure 4.8c
Crop confusion status represented on the first two canonical
axes when three temporal measurements by XVH50 are combined with one time
measurement by XW50 during the July 9 - August 16, 1982 timeframe.
separation between corn and double-cropped
4.8c
shows
soybeans, while
that further addition of XW50 on July 9th improves the
separation of each of these from regular soybeans.
August
16
When
XVH50
of
was added to XVH50 (July 26) and XVH50 (July S), the ac­
curacy increased to 79.4 percent from 73.5 (Figure
cropped
Figure
soybeans
4.8b).
Double-
increased from 80 to 90 percent and corn improved
from 71.4 to 85.7 percent. However, regular soybeans
remained
at
70 percent mainly due to confusion with corn.
It is interesting to note that when
(XW50
the
fourth
best
channel
on July 9th) in terms of F-ratio was added in the procedure,
the overall accuracy level increased to 82.4 percent due to the
crease
in
corn from 85.7 to 92.9 percent, but the accuracy of soy­
beans actually dropped to 60 per cent. These results
the
XW50
channel
contains
consistently
superior
suggest
that
information peculiar to corn if it is
combined with the other three channels.
was
in­
to
XVH50
Recall that
XW50
channel
for discriminating corn from
wheat and soybeans in the previous dates.
4.7
Timeframe 5: August 16/18, 1982
(Corn, Regular Soybeans, and Double-Cropped Soybeans)
The same crop types as Timeframe 4 were observed on this
Radar
returns
obtained on August 16/18 involve eight channels:
and X-bands with dual polarization at 20 and
angles
each.
Date.
50
degrees
C-
incidence
The incidence angle of 20 degrees was used to detect
soil moisture as discussed in Chapter 2.
126
The
performance
of
in­
dividual channels is shown in Figure 4.9. The relatively low levels
of classification accuracy may be due to the
fact
that
plants
at
this time were drying up (see Appendix A and Jung et al., 1983).
The stepwise procedure indicated that the importance
nels
of
chan­
was XVH20, CHV50, CHH50, XW20, CHV20, XVH50, CHH20, and XW50
in descending order of importance.
channel
that
has
the
highest
polarization performs better
teresting
Note
in
accuracy
than
4.9
is CHV50.
that
It
is
that XVH20 turned out to have the highest F-value.
of
the
Again, cross-
like-polarization.
radar energy penetrates better at low angles
lying
Figure
incidence,
in­
Since
under­
soil conditions have more influence on the energy return than
crop canopy at this time of the year.
Plots using the two canonical axes for different
binations
are
shown
in Figures
circles
of
the
XVH20
and
of
CHH50,
crop
CHV50.
As
90 percent confidence limit indicate,
there is confusion between corn and regular soybeans.
dition
com­
4.10a, 4.10b, and 4.10c. Figure
4.10a shows the result of transformation of
overlapping
channel
classification
With the
ad­
accuracy increases rapidly
(Figure 4.10b). Regular soybeans are 90 percent separable from corn
and double-cropped soybeans.
The latter two have a degree of confu­
sion but the circles are more separated than those shown
4.10a.
If
a
study
Figure
is to discriminate only regular soybeans from
corn and double-cropped soybeans,
XVH20,
in
then
a
channel
combination
of
CHV50, and CHH50 would produce the highest accuracy at this
127
100
80
<r
60 -
m
m
CNJ
^r
<r
r--
<r
<r
•st
CM
m
<r
«<r
O
CM
33
33
O
o
CM X
> 4J
33 o
CJ PQ
LO
40
CM
cn
CO
20 -
X.
4J
o
CQ
O
in
sc
»
u
o
m 43
> •U
33 o
U CQ
Figure 4.9
Performance of individual radar
channels
and
polarization combination to discriminate corn, regular soybeans
and double-cropped soybeans on August 16, 1982.
Notice that
CHV50 achieves the highest accuracy level among the eight radar
channels.
128
Channels used:
CHV50
XVH20
® — CORN
*- SOYBEANS
+- SOYBEANS a
Corn (57.1%)
Double^
cropped
soybeans (70%)
Regular soybeans
(60%)
Overall correct classification: 61.8%
-31
-4
-3
-2
1
0
canonical
1
axis
1
Figure 4.10a
Crop confusion status represented on the first two
canonical axes when CHV50 and XVH20 of August 16, 1982 are combined for
classfication.
Channels used:
CHV50
XVH20
CHH50
®- CORN
•»- SOYBEANS
+- SQY3EANS D.
Corn •
(78.6%)
r••I
0-1
a Regular soybeans
Double-cropped
soybeans (90%)
(80%)
-2
Overall correct
classification:
-31
-4
-3
76.5%
-2
canonical
axis
1
Figure 4.10b
Crop confusion status represented on the first two
canonical axes when three radar channel measurements made on August 16,
1982 are combined for classification.
Channels used:
CHV50
XVH20
>
CHH50
/
"XW20
/
A
o- CORN
SOYBEANS
+- SOYBEANS 0.
Regular
Soybeans (80%)
Corn
(85.7%)
•
9-i
Double-cropped
soybeans (80%)
Overall correct
classification: 82.4%
-2
-31
-4
-3
-2
canonical axis
1
Figure 4.10c
Crop confusion status represented on the first two
canonical axes when measurements made by four radar channels on August
16, 1982 are combined for classification.
time.
When XW20, the fourth important channel is added in the
knifed
procedure,
jack-
an overall accuracy of 82.4 percent is achieved.
The three crop categories also separate from each other as indicated
by minimal circle overlapping
and accuracies of individual crops in
Figure 4.10c.
4.8 Practical Number of Channels for Crop Discrimination
It was mentioned in the previous section that Timeframe
part of Timeframe 4, but involved eight radar channels.
Of
turned
out
to
be
XVH50
measurements.
Although
comparison can be made with the eight radar channels of
Timeframe 5 and the three temporal variables of Timeframe 4.
4.11
shows
bined.
chan­
the six channel variables used in Timeframe 4, the three
most important
limited, a
was
Timeframe 4
involved three temporal measurements by dual-polarized X-band
nels.
5
changes
in
Figure
accuracies as additional channels are com­
The dotted curve indicates aggregated classification
levels
with the gradual addition of the XVH50 channels employed during July
9 to August 16.
The solid curve represents percent of correct clas­
sification as additional channels used on August 16 are combined.
When compared with the temporal dimension,
the
channel
dimensions
are
lower
for
the
accuracies
the first three channels.
After the 82.4 percent accuracy, the performance curve of the
nel
dimension
drops
or stays at the saturation level.
performance curves of this type
saturate
132
of
after
a
chan­
Generally,
certain
point.
100
79.4
80 -
82.4
79.4
82.4
85.3
82.4
76.5
61.8 - *
60 -
61.8
47.1
20 -
•
XVH50
CHV50
+XVH20
+ July 26 "••July 9
+CHH50
+Aug 16
+XVV20
+CHV20
+XVH50
+CHH20
+XW50
Figure 4.11
Change of correct classification level as additional channels are combined
according to F-to-enter values.
The horizontal axis indicates the order of channels
added. The three temporal measurements of XVH50 are also plotted in a dotted line.
The
crops observed are corn, regular soybeans and double-cropped soybeans. Notice that when
the fouth channel is added, the accuracy curve starts to level off.
When
sample
size
is
rather
small, the multivariate discriminant
levels represented by curves such as these sometimes even drop after
a
certain
point (Gnanadesikan, 1977). The sample size here is 34:
14 for corn and 10 each for regular soybeans and double-cropped soy­
beans.
Therefore,
the
accuracy of 82.4 percent appears to be the
saturation point. From then on, information added by the
of
additional
channel
variables
does
inclusion
not influence the accuracy
level significantly. •
For this particular situation, it could be concluded that
dimensions
of
channel
variables are needed in order to achieve an
accuracy level of about 80 percent to classify
beans,
and
timeframe.
crop
double-cropped
soybeans
during
corn,
the
July
to
that
soy­
August
same
by XVH50, the comparison of channel variables with
temporal variables is limited. However, the curves in
suggest
regular
Since there are only three Dates that observed the
categories
four
Figure
4.11
the temporal measurements are more important than em­
ploying multiple channels in one day.
Also, cross-polarized
chan­
nels perform better than like-polarized radar channels for this par­
ticular case.
nel
variables
For practical applications, it seems that four chan­
of either temporal or channel dimension will reach a
saturation point in the overall classification level.
13^
CHAPTER 5
OPTICAL AND MICROWAVE SIGNATURES: THEIR SEPARATE
AND COMBINED CAPABILITIES TO DISCRIMINATE CROPS
The use of the Modular Multiband Radiometer (MMR) in the summer
of
1982
was not systematic and measurements were sporadic, because
the emphasis of the experiment was
directed
towards
understanding
microwave signatures from typical agricultural crops, as dicussed in
Chapter 4.
In 1983, however, both MMR and radar
simultaneouly
to
acquire
signatures
from
were
selected
used
almost
agricultural
scenes.
One or two fields for each crop were observed on
a
weekly
basis.
The measurement log indicating crop categories measured and
data availability by date is shown in Table 5.1.
Although eight optical channels and four
used
in
radar
channels
the experiment (Tables 2.3 and 3.1), MMR channel 6 was ex­
cluded because it is the
discrimination studies.
thermal
channel
not
suitable
for
crop
One time measurement made by each MMR chan­
nel consists of 20 samples. In other words, as the truck boom
the
were
with
MMR system mounted on top of it scans the target scene, each of
eight MMR channels viewed and recorded
areal spots.
signals
from
20
different
These individual samples were calibrated as discussed
in Chapter 3 and are listed in Appendix C.
On the other hand, each radar
channel
scanned
and
recorded
return signals from 30 or 35 different spots as the truck boom swept
the target scene in much the same way as the MMR channels, but these
135
Table 5.1
MHR and Radar Measurement Log for 1983
/
C-Band
\ X-Band
O MMR
August
July
June
Date
6 8 14 19 20 27 29 2 3 25
6 15 17 20 30 1 5
157 166 168 171 181 182 186 187 189 195 200 201 208 210 214 215 237
Crop
Corn 1
Corn 2
Fallow
Grass
Hay
®
®
®
®
®
®
®
X
O
®
®
®
®
®
®
®
X
O O
O
O
O
O
O
O
O X
O
O
X
O
X
O O
X
X
O
®
X
®
Milo
Potato
Soybean 1
Soybean 2
® X
® ®
® X
O
O
®
®
®
X
O
O
O O
X O
O
X
X
X
X
X
O
O X
X
O
O O X
X
O O
X
O
O
X
O
Soybean D.
Swamp
Wheat 1
Wheat 2
S. Wheat
O
® ®
® ®
® X
® X
®
®
®
®
X
O
O
O
O X
O
O
O
O
X
X
X
O
X
O
X
®
Alfalfa
®
Bare Soil
136
O
radar signals were integrated during data collection in the field to
account for the fading statistics of radar discussed in
Chapter
3.
Therefore, one sweep of the boom produces 20 independent reflectance
values for each MMR channel, but only one averaged backscatter coef­
ficient (a0) for a radar channel.
Radar measurements had
sample
size
comparable
to
to
be
that
repeated
to
of the MMR.
backscatter coefficient values were obtained for
This
acquire
adequate
Generally, 10 to 30
each
test
field.
is different from the field measurement procedure of 1982 when
only one backscatter coefficient was acquired
measurement
for
each
field
per
day by driving the radar truck slowly along the edge of
the field until 30 independent spots were
scanned (Aslam
et
al.,
1983).
In the summer of 1983, the radar truck initially moved back and
forth
along
the
edge
of
the
field until a sufficient number of
backscatter values were obtained. It took approximately
to
make
general
locations
five
to
in a given field with a minimum movement of
the boom truck for each sweep within a location.
was
hours
30 sweeps, i.e., 30 o° values using the mobile drive mode.
To reduce the measurement time, radar data were taken from
six
two
When one
location
completed, the radar truck moved to the next location until the
desired number of sweeps were made for the test field.
137
5.1 Data Evaluation and Selection of Date Variables
The field data thus gathered, both optical and microwave,
calibrated
according
to
the procedures discussed in Chapter 3 and
are listed in Appendices C and D.
field
by
were
Their
average
values
for
each
each channel are plotted as a function of day of year and
are shown in Appendix B.
The field data were cleaned during
but
an
the
calibration
processes
evaluation was also made of the calibrated data before they
were used for analysis. Figure 5.1 is a plot of the white reference
panel
measurements
taken
function of day of year.
made
during 11:00 - 12:00 in the morning as a
Since the plotted values are
measurements
during a one hour period each day from day 157 (June 6) to day
237 (August 25) under similar cloudless conditions (see Chapter 3),
the returns from the barium sulfate reference surface should be very
similar throughout
taken
the
period.
Unfortunately,
the
measurements
before Julian date 181 (June 30) indicate irregularities.
particular, the "valley"
on
day
171 (June
20) suggests
In
either
weakened battery condition or a malfuntion of the MMR instrument it­
self, thus making the quality of the reflectance data
The
MMR
measurements
plotted
questionable.
in Appendix B confirm these doubts.
The trend in the plots shows that the reflectance measurements
during
day 170 (June 19) to day 181 (June 30) are irregular, having
generally a marked depression in the curves.
tance
made
values
acquired
Therefore, the reflec­
throughout this period were not used in the
analysis.
138
0)
•+-*
VjJ
o
>
VO
Ch 4
Ch ft
Ch 7
0.0
150
160
170
Figure 5.1
180
190
200
210
Juffcjn Date
220
230
Reference panel measurements by date (11:00-12:00).
24-0
250
Based on the quality of data acquired, the availability of both
MMR and radar data, crop types observed, sample sizes available, and
the number of dates involved to observe a given field, three "Dates"
of July 6-8, July 19 - 20, and July 27 - August 3, were selected
for analysis.
tural
cycle
The July time is an important period in
an
agricul­
in the midlatitude of the Northern Hemisphere since it
has been reported to be the best period to discriminate crops (Good­
man, 1959; Steiner, 1970).
5.2 Date 1: July 6/8, 1983
The first Date consists of MMR measurements made on July 6 and
radar measurements made on July 8.
Multiple radar measurements took
so much time that simultaneous employment of MMR and radar
cases
was
not
practical
in
terras
boom.
Crop
Since
each
mounted
on
the
same
categories observed at this time included corn,
cut grass, hay, soybeans, winter wheat, and
5.1).
some
of personnel requirements and
logistics, although both sensor systems were
truck
in
spring
wheat (Table
crop type had one or two test fields, data were
extracted from both or either one of the two fields in much the same
way
as
training
scene category.
samples
are taken from landsat data for the same
The characteristics of the target scenes
are
sum­
marized in Appendix E.
Figure 5.2a shows the performance levels of individual channels
for
the
first
Date
as measured by the percent of overall correct
classification of each channel for the six scene categories.
140
Of the
100
82.5
80
-
75.0
64.2
«
60.8
60
53.3
52.5
52.5
52.5
44.2
37.5
40
25.8
20
Ml
M2
M3
M4
M8
M5
M7
CHH
50
CHV
50
XW
50
XVH
50
Figure 5.2a
Performance of individual channels to discriminate
winter wheat, cut grass, spring wheat, hay, corn, and soybeans
on July 6/8, 1983. The TM channels are labeled Ml - M7 (see
Table 2.3).
D
O
•—
(0
0)
o
o
K
87.5
99.2
99.2
99.2
—A
c 100r
o
90 +>
' 1A
—~A
99.2
98.3
+M8
+XVH +M5
50
99.2
99.2
80
82.5
70
60
50
-+-
•+-
M3
+M2
+M4
+M7
+M1
+XVV +CHV
50
50
-h
4CHH
50
Figure 5.2b
Combined classification accuracy change as chan­
nels are added according to F-to-enter values on July 6/8, 1983.
1^1
11
channels
involved,
M3
has the highest classification accuracy
(82.5 percent) followed by M2, M4, and
ordered
based
on
the
M7
F-ratio (Figure
when
5.2b).
the
channels
It is as expected
because channel 3 (M3 for MMR and TM3 for TM) is in the
absorption
region
of
tiating crop species.
superior
to
the
are
chlorophyll
the spectrum and is designated for differen­
It is obvious that the optical
channels
are
radar channels as a crop classifier in this case.
The M3 channel alone achieves an overall classification accuracy
82.5
of
percent which is higher than the 81.7 percent achieved by com­
bining all four radar channels.
It is interesting to note that the
XVH50
have
the
radar
CHV50
and
same accuracy level of 52.5 percent, which is much
higher than those for CHH50 and XW50 (Figure
the
channels
5.2a).
Recall
that
results from the 1982 experiment showed that cross-polarization
was better than like-polarization for multiple crop discrimination.
When both the optical and radar channels
stepwise
discriminant
are
entered
in
the
procedure, XVH50 ranked as sixth most impor­
tant after the five optical channels and CHV50 ranked tenth (Figure
5.2b).
The
low
ranking
for
CHV50
despite discriminatory power
similar to that of XVH50 is due to the appreciable degree of
dancy
in
redun­
information content between the two cross-polarized chan­
nels.
It was noted in Chapter 4 that
a
practical
number
of
radar
channels needed to reach a saturation point in a correct classifica­
142
tion curve is four.
first
The same holds for optical channels.
Date, M3 achieves an 82.5 percent accuracy level.
For
the
When M3 is
combined with M2 and M4, the level increases to 95.8 percent (Figure
5.2b).
The
combined
level
of
correct
classification reaches a
saturation point at 99.2 percent for a four channel
M3,
combination
of
M2, M4, and M7. From then on, the accuracy level changes very
little even with the addition of the
remaining
optical
and
radar
channels.
Being passive in nature, optical sensors can have a
channels
on
scanners.
a
space-borne
system,
as
is
the
ground
scene,
of
that
could
energy
Hence, the
be made available for crop discrimination may
well be limited to one, even though a multichannel microwave
will
be
available in the future.
of
using
three
optical
to
obtain
a
(Figures 5.2b and 5.3).
be
reasonable
to
channels and one radar channel for
practical applications, because four is
channels
sensor
In a multisensor situation using
optical and microwave sensors, therefore, it may
think
to
the power requirement is one of the
major considerations in active microwave remote sensing.
channels
of
seen by multispectral
Since radar has to provide its own source
illuminate
number
satisfactory
the
practical
number
of
classification accuracy level
Two optical channels and one radar
channel
may be used in a color composite for enhanced visual interpretation.
For the July 6/8, 1983 applications
crops,
the
recommended
to
discriminate
the
optical channels are M2, M3, and M4.
143
six
This
96.7
100
96.7
95.0
89.2
89.2
81.7
84.2
80
£
60
55.0
53.3
40 -
33.3
30.8
20 -
M3
XVH
M3
CHV
M3
M2
XVH
M3
M2
CHV
M3
M2
MA
XVH
M3
M2
M4
CHV
CHV
CHH
CHH
CHV
XW
XVH
XVV
XVH
XVV
CHV
XVH
CHH
Figure 5.3
Performance of channels as a group to classify
winter wheat, cut grass, spring wheat, hay, corn, and soybeans
on July 6/8, 1983. An incidence angle of 50 degrees was used
for all radar channels.
3M
coincides with the TM channels (TM2, TM3, and TM4) recommended
produce
to
a false color composite for vegetation studies (EDO, 1984).
Rosenthal and Blanchard (1984) also recommended these three
channels
for
crop
classification.
XVH50 or CHV50 is recommended.
optical
As for radar channels, either
Note in Figure 5.3 that
a
correct
classification level of 96.7 percent is obtained by combining either
CHV50 or XVH50 with the three optical channels.
Also, note in Figure 5.3 that a depolarization
C-band
and X-band performs poorer.
ratio
of
both
The accuracy level of 81.7 per­
cent achieved by combining the four radar channels is comparable
that
of
the
1982
to
results (see Figure 4.11 for example), although
there are differences in the number of
cover
types
date of measurements, crop conditions, and so on.
involved,
the
For a three chan­
nel combination including a radar channel for color composite rendi­
tion, a
combination
for this Date.
of M3, M2, and XVH50 would be the best choice
Using CHV50
instead
of
XVH50
is
also
effective
(Figure 5.3).
5.3 Date 2: July 19/20, 1983
The second Date for analysis
19 and July 20.
is the data base obtained on July
The radar measurements were made on July 19 and op­
tical measurements were made on July 20 (Table 5.1).
observed
by
types
both systems include corn, fallow, potatoes, soybeans,
swamp, and winter wheat.
nels
Cover
The performance level of individual
chan­
are represented by bars in Figure 5.4a. Compared with Date 1,
145
100
c
o
80
'•P
o
65.8 63.5 62.5
o
62.5
•—
'm
tfl
o
o
57.5
60
53.3
50.8
44.2
40.0
g
40
40.0
36.7
t_
o
o
N
20
Ml
M2
M3
M4
M8
M5
M7
CHH
CHV
XW
XVH
50
SO
50
50
Figure 5.4a
Performance of individual channels to classify
corn, fallow, potatoes, soybeans, swamp, and winter wheat on
July 19/20, 1983.
D
O
*—
(0
99.2
99.2
—A-
99.2
-A
98.3
80
V)
a
70
o
60
K
99.2
—&—
99.2
100 r
c
o
tp 90
62.5
50
-t-
M3
+CHV
50
+M1
+M7
+M5
-+•
+M4
-+-
+M8
+M2
+CHH +XVV
50
50
+XVH
50
Figure 5.4b
Change of combined classification accuracy as
channels are added according to F-to-enter values on July 19/20,
1983.
146
the performance of the radar channels has improved slightly and that
of
the
optical channels has deteriorated, due to a change in cover
categories.
the
best
The Ml channel can discriminate the six crop categories
with
an
overall correct classification accuracy of 65.3
percent, followed by M2, M3, and CHV50.
When the channels are ordered according to their discriminatory
power based upon the F-ratio criterion, M3 turns out to be the first
to be selected to discriminate the six agricultural scene categories
at
this time followed by CHV50, Ml, and M7.
The horizontal axis in
Figure 5.4b indicates the importance of the 12 channels entered from
left
to
right according to the F-ratio criterion.
other channels and CHV50 is the next best
the six cover categories.
channel
M3 out-performs
to
discriminate
Also, it is interesting to see that CHV50
is found to be the second best over a number
of
optical
channels.
This could be due to the redundancy of information among the optical
channels, while
the
radar
channels
provides
unique
information
suitable for crop discrimination.
Since each region of the electromagnetic spectrum can represent
unique
information
on
the scene it observes, it would be a worth­
while effort to look into
discriminate
the
how
each
of
the
major
four
ground scenes for this Date individually. Tables
5.2a through 5.2d show the abilities of M3, CHV50, Ml, and
discriminate
the
channels
six
crop
M7
to
categories on July 19/20, 1983. Table
5.2a shows that M3 can identify winter wheat and soybeans perfectly.
1^7
Table 5.2a
Performance of M3 on July 19/20, 1983
Classified As:
Actual
Category
Corn
%
Correct
Corn
Soybeans
Wheat
45
45
Potatoes
Swamp
5
Fallow
50
100
Soybeans
100
W. Wheat
100
Potatoes
40
Swamp
60
Fallow
30
TOTAL
62.5
100
40
35
25
5
20
60
15
50
20
30
Table 5.2b
Performance of CHV50 on July 19/20, 1983
Classified As:
Actual
Category
%
Correct
Corn
Soybeans
Wheat
Potatoes
Swamp
Fallow
Corn
95
5
Soybeans
35
W. Wheat
100
Potatoes
45
Swamp
50
15
50
35
Fallow
60
10
30
60
TOTAL
62.5
95
35
50
15
45
25
100
10
20
148
Table 5.2c
Performance of Ml on July 19/20, 1983
Classified As:
Actual
Category
%
Correct
65
Corn
Soybeans
100
W. Wheat
95
Corn
Soybeans
Wheat
65
Potatoes
Swamp
5
Fallow
30
100
95
Potatoes
65
30
Swamp
10
30
Fallow
60
5
5
65
10
5
10
10
40
10
25
60
Swamp
Fallow
20
40
50
5
Table 5.2d
Performance of M7 on July 19/20, 1983
Classified As:
Actual
Category
%
Correct
Corn
Soybeans
Corn
20
20
20
Soybeans
20
10
20
W. Wheat
95
Potatoes
40
45
5
Swamp
75
15
10
Fallow
70
25
5
Wheat
Potatoes
15
95
1^9
5
40
10
75
70
Confused
with
and corn.
percent
each other at this time are fallow, swamp, potatoes,
The CHV50 channel alone discriminates
and
corn 95 percent.
winter
wheat
Others are confused with each other
(Table 5.2b).
Ml performs similarly to M3 (Table 5.3) with
provement
classification
in
100
an
im­
of fallow, potatoes, and corn (Tables
5.2a and 5.3).
M7 correctly classifies winter wheat
95 percent, swamp 75 per­
cent, and fallow 70 percent. Considering the fact that the TM chan­
nel 7 was designed for geologic remote sensing, the result here does
seem to be by chance alone.
not
Winter wheat at this time was ripe
and dry, swamp and fallow fields were also dry because of
sustained
drought and hot weather in the summer of 1983 (see Appendix E). The
categories with considerable amount
corn, soybeans, and
of
potatoes, are
vegetation
poorly
biomass, i.e.,
classified
with this
geologic channel (Table 5.2d).
Again, it is illustrative to see the change of confusion status
on
the
first two canonical axes as channels are added for combined
classification.
bined, an
When the best two channels, M3 and CHV50, are
accuracy of 87.5 percent is reached.
com­
The improved level
of correct classification for each category and confusion status are
shown
in Figure 5.6a
on a two dimensional plane formed by trans­
forming and scaling the original M3 and CHV50 data.
the circles show percent correct classification.
tion is indicated by arrows.
in
The misclassifica-
Of the six categories,
150
The numbers
winter
wheat,
100
87.5
86.7
88.3
85.8
85.0
87.5
80.8
80
76.7
c
o
IP
O
0
1
(0
o
o
60
o
0)
40
L.
L.
67.5
36.1
33.3
o
o
20
M3
CHV
M3
CHH
M3
Ml
CHV
M3
Ml
CHH
M3
Ml
M7
M3
Ml
M7
M5
CHV
CHH
CHH
CHV
XVH
XVV
XW
XVH
CHV
XVV
XVH
CHH
Figure 5.5
Performance of channels as a group to classify
corn, fallow, potatoes, soybeans, swamp, and winter wheat on
July 19/20, 1983. An incidence angle of 50 degrees was used for
all radar channels.
151
Date: July 19/20. 1983
Channels: M3. CHV50
ru
cn
corn
•r-i
soybeans
X
ro
w. wheat
to
•iH
potatoes
1
c
o
c
swamp
o 31
-5
Overall accuracy: 87.5%
-7
-13
-11
-g
-7
-5
-a
—j
canonical axis 1
Figure 5.6a
confusion status of crops observed by M3 and CHV50 on July 19/20 1983
of
?n
*•5lrSt ^W° canonical axes- The circles represent the upper'limit
percent confidence interval. The confusion among crops is shown by the arrows
Date: July 19/20, 1983
Channels: M3, CHV50, Ml
soybeans
^100%
95%
J
corn
1
" 60%
w. wheat
potatoes
100%
•M-i
c
10
llow
05%
swamp
80%
r3
Overall accuracy: 88.3%
-7
- 1 5
- 1 3
- 1 1
- 9
- 7
- 5
-
3
canonical axis 1
-
1
1
3
Figure 5.6b
Confusion status of crops observed by M3, CHV50, and Ml on July 19/20, 1983
as represented on the first two canonical axes. The circles represent the upper limit of
the 90 percent confidence interval. The confusion among crops is shown by the arrows.
5
7
Date: July 19/20. 1983
Channels: M3. CHV50, Ml. M7
soybean
100%
ai
en
-rH
X
(D
w. wheat
£
•p
-r-t
swamp
100%
(O
1
c
o
c
O
5
15
fallow
95%
d
20
potatoes
80%
-5
Overall accuracy: 92.5%
-7
-15
-13
-11
-7
-5
_L
_L
-3
-1
canonical axis 1
Figure 5.6c
Confusion status of crops observed by M3, CHV50, Ml, and M7 on July 19/20, 1983
rL "Present^d °° the first t"0 canonical axes. The circles represent the upper'limit of
the 90 percent confidence interval. The confusion among crops is shown by the arrows.
soybeans,
more.
corn,
and
fallow are correctly classified 95 percent or
Potatoes and swamp are largely confused with each other.
When Ml is added to M3
level
and
CHV50,
the
combined
increases slightly to 88.3 percent, due to classification im­
provement of swamp from 75 to 80 percent (Figure
the
performance
geologic
channel
5.6b).
When
is added to form a four channel combination,
the two most difficult categories to discriminate, namely
and
swamp, are
M7,
potatoes
better classified to boost the overall accuracy to
92.5 percent (Figure 5.6c).
Again, the
cumulative
accuracy
curve
starts
to saturate after the fourth best channel is entered (Figure
5.6b).
Therefore, it may be better not to include the fifth or
following
channels
for
crop
the
classification studies. The perfor­
mances of other channel combinations
on
this
Date
are
shown
in
Figure 5.5.
5.4 Date 3: July 27 - August 3, 1983
The third Date involves eight categories of
grass,
milo, regular
soybeans, double-cropped
wheat, alfalfa, and bare soil.
are
shown
in Table 5.1.
3,
1983.
crop categories
agricultural
This
such
scene
as
corn, cut
soybeans,
spring
The dates of measurements by
fields
The third Date actually is a "window" in
the growth stages of crops that spans eight days
August
crops:
from
July
27
to
timeframe was selected in order to include
alfalfa
categories
and
milo (Table
5.1).
Other
included at this time are corn, cut
grass, regular soybeans, double-cropped soybeans, spring wheat,
155
and
bare soil.
The performance levels of
Figure
5.7a.
M7
is
individual
channels
are
shown
in
the best optical channel to discriminate the
eight scene categories and XW is the best among the radar channels.
When
the
channels
are
ordered
according to their discriminatory
power based on F-ratio, M7 is followed by M3, M8, CHV50, and
(Figure
5.7b).
so
on
The performances of different channel combinations
are shown in Figure 5.8.
The fact that the geologic channel (M7) turns
best
out
to
be
the
classifier of the eight categories may be due to the crop con­
ditions at this period.
Spring wheat at this time
was
crisp
dry.
It was planted early in the spring not as a grain crop but as a setaside cover for soil conservation.
never
reached
53 cm in height.
The
average
soybeans
were
were
left.
been
cut
soil
It was sparsely vegetated
lawn
The cut grass field category actually is
of
Grant
School
in the area.
grass on the lawn remained almost constant at 9 E).
It
Alfalfa
on
a
with an average plant density of 23.5 stems per 30 x 30
cm sample plot.
maintained
E).
recently and the mean canopy height from five sample
locations was 11.6 cm on August 3.
sandy
Double-
characterized by uneven growth, and several
bare spots were noticable in the test field (Appendix
had
canopy
The leaves started to dry out from
June 28, and by July 19 few green grain heads
cropped
vegetation
provides
a
the
well
The height of the
12
cm (Appendix
good reference for both optical and microwave
156
100
c
o
80
73.8
53
o
o
k
CO
o
o
•fj
u
£
66.3
60.0
60 -
55.6
61.9
54.4
52.5
43.8
40
38.1
35.0
31.3
L.
o
0
20
Ml
M2
M3
CHH
CHV
XVV
XVH
50
SO
50
50
Figure 5.7a
Performance of individual channels to classify al­
falfa, milo, corn, cut grass, regular soybeans, double-cropped
soybeans, spring wheat, and bare soil during the July 27 August 3, 1983 period.
M5
A—
M7
98.8
95.6
99.4
99.4
99.4
A
—A
99.4
—A
99.4
80
a
to
o
70
o
60
N
M8
97.5
c 100 r
o
tp 90
D
O
£
M4
50
73.8
M7
•+-
-+-
•+-
+M3
+M8
+CHV
+M2
+XVH +M4
50
-tXW
50
+CHH
50
•+-
-t-
+M1
+M5
Figure 5.7b
Change of combined classification accuracy as
channels are added according to F-to-enter values during the
July 27 - August 3, 1983 period.
157
100r
95.0
86.9
80.6
94.4
88.1
83.1
81.9
80
70.6
*=
58.8
60
46.9
©
40
31.3
20
M7
CHV
M7
M3
M7
M3
CHV
M7
M3
M2
M7
M3
M2
CHV
M7
M3
M2
XW
CHV
CHH
CHH
CHV
XW
XVH
XW
XVH
XW
CHV
XVH
CHH
Figure 5.8
Performance of channels as a group to classify al­
falfa, milo, corn, cut grass, regular soybeans, double-cropped
soybeans, spring wheat, and bare soil during the July 27 August 3, 1983 period. All radar measurements were made at 50
degrees incidence angle.
158
signals as shown by the relatively flat, unchanging curves in Appen­
dix B.
Bare soil was rough, recently tilled and disked.
The M7 channel alone correctly classifies
soybeans,
spring
cut
grass, regular
wheat, double-cropped soybeans, alfalfa, and bare
soil 80 percent or better (Figure 5.9). On the other hand, corn and
milo
are poorly classified at 35 and 30 percent respectively.
shows that the geologic channel is
tall,
leafy
vegetation.
tall and lush with
several
starting
yellow.
to
turn
not
effective
for
This
classifying
At this time corn was about three meters
leaves
at
the
bottom
of
the
plant
The milo canopy was 126 cm tall on the
average on August 3, 1983 (Appendix E).
When M7 is combined with the next best channels, namely M3
M8, the
5.9).
ferent
accuracy
levels of corn and milo increase rapidly (Figure
M3 is in the chlorophyll absorption band and responds to dif­
species
of vegetation because the dominant factor that con­
trols the reflectance from vegetation in this part of the
is
and
leaf
pigments.
spectrum
Pigmentation differences cause a distinct dif­
ference in reflectance in this wavelength region (see Chapter 2).
M8 is an additional channel added to the
Purdue University agronomists.
TM
plant
leaves
largely
by
the
It is in the near infrared region of
the spectrum (Figure 2.2). In this region, the
of
channels
internal
structure
controls the reflectance (Hoffer, 1978).
Therefore, the addition of M3 and M8 upgraded the classification ac­
curacy
significantly
by improving classifications of leafy vegeta-
159
100
90
grass
bare soil
soybeans
alfalfa
80
c
o
'•£
70
«
60
•s
50
o
40
o
o
o
<D
L.
_ soybeans, d,
s. wheat
corn
30
milo
20
10
0
M3
M7
M3
M8
M7
M3
M8
CHV50
Figure 5.9
Performance of selected channels to classify
during the July 27 - August 3, 1983 period.
160
crops
tion (Figure 5.9). The addition of CHV50 at this time only slightly
increases
the overall classification accuracy, from 95 to 95.6 per­
cent (Figure 5.7b).
The accuracy level achieved by the four best channels
third
for
the
Date is somewhat lower than those for the previous two Dates,
because the number of scene categories has increased to
eight
from
six and the scene categories involved are different from each other.
It seems from this segment of the analysis that
with
other
channels
discriminates
M7
in
combination
an agricultural scene which is
less dominated by vegetation, such as dryland crops in semi-arid re­
gions.
The
results from Date 2 also showed that M7 contributed to
increase the accuracy level of the almost
dry
vegetated
M7
potato
categories.
Although
swamp
is
and
designated in the
Thematic Mapper for geologic applications, it also can
for
agricultural
applications
sparsely
be
valuable
depending upon the area under study
and the scene categories in question.
5.5
Importance of the Temporal Dimension and the Complementary
Nature of Radar for Crop Discrimination
Since measurements obtained early in the summer up to
were
30
avoided because of questionable data quality, temporal evalua­
tion of the performance of the channels is limited.
Dates
June
considered
in
Of
the
three
this Chapter, Date 1 and Date 3 were selected
for a multidate study, based on the separation of the two Dates
mutually
viewed
scene categories.
and
The two Dates are about a month
161
apart.
The categories observed by both Dates are corn,
cut
grass,
soybeans, and spring wheat.
Although the seven MMR channels and
available
four
radar
channels
for analysis, only the channels found to be of importance
in the previous discussions have been selected for comparison.
are
M2,
M3, M4, and CHV50.
geologic
applications
vegetative types.
formed
well
They
M7 was the best classifier for Date 3,
but it was excluded for evaluation because the
for
are
and
channel
is
largely
the categories considered here are
The selection of CHV50 was made because
it
per­
among the radar channels under evaluation for both the
1982 and 1983 experiments.
Table 5.3 summarizes the performance
of
the channels, either individually or in groups, to classify the four
categories.
the
last
tion.
The rows represent the channels used and the figures in
column
The
indicate the level of overall correct classifica­
groups
were
selected
to
represent
practical
mul-
tidate/sensor/band combinations.
The table shows that M3 has the highest classification accuracy
and
CHV50 has the lowest for both Dates.
When the four channels of
both Dates are combined to discriminate the
the
level
overall
classification
reaches
crop
categories,
97.5 percent.
The
dimension involved here is 4 channels x 2 dates = 8. When only
the
three
of
four
MMR channels for the two Dates are used, an accuracy level of
91.3 percent is achieved.
dimension
used
is
This is rather unsatisfactory because the
3 MMR channels x 2 dates = 6. Also recall that
162
Table 5.3
Performance Comparison of MMR and Radar Channels
To Classify Crops on Date 1 and Date 3
Date 1
(July 6/8, 1983)
Date 3
(July 27-Au.gust 3, 1983)
M2
71.3
M3
71.3
M4
62.5
CHV50
53.B
67.5
M2
-
87.5
M3
75.0
M4
CHV50
M2
M3
M4
M2
M3
M3
M4
M3
M2
Note:
M3
M3
CHV50
M4
92.5
M2
M3
M4
M2
M3
M4
M4
H4
90.0
M4
CHV50
M2
51.3
87.5
M2
H2
Overall
Correct
Classification
(*)
CHV50
M2
M3
M4
CHV50
M2
M3
M4
CHV50
93.8
91.3
CHV50
62.5
CHV50
95.3
93.8
CHV50
97.5
The crop categories observed on both dates are corn, cut grass,
soybeans, and spring wheat.
163
the
curves
"plateau"
indicating
when four
the
correct
dimensions
classification
were
used
as
reached
the
discussed in this
Chapter and Chapter 4. The level of 91.3 percent obtained using the
two
temporal
data sets of MMR measurements is also not a great im­
provement over the one temporal data set levels
of
87.5
and
90.0
approach
for
this
percent.
Table 5.3 also shows that the
case
is
better
multisensor
than the one sensor system temporal approach. For
Date 1, for example, an overall correct classification of 92.5
cent
is
together.
reached
when
three
MMR
per­
channels and CHV50 are combined
The accuracy level increases to 93.8 percent for the same
combination for Date 3.
Note that the level is higher than when all
MMR channels of two temporal data sets were
combined,
a
value
of
91.3 percent.
The fact that a single day multichannel of four is better
than
using two temporal data sets of dimension six may appear to conflict
with the discussion in Chapter 4, where it was suggested
that
tem­
poral measurements might be better than single day measurements with
multiple radar channels (Figure 4.11).
channels
of
one
However, both
cases
use
sensor system operating in the same region of the
spectrum, one being in the microwave region and the other in the op­
tical
region.
It
is noteworthy that the increased accuracy level
here, i.e., 92.5 percent for Date 1 and 93.8 for Date 3 are achieved
using one radar channel and three MMR channels respectively.
164
To illustrate further the complementary
crop
discrimination
studies,
a
nature
of
in
the
radar
systems
for
reasonable simulation can be made
that only TM data are available for one date because of
tion
radar
a
malfunc­
and only radar data are available for
another date because of cloud interruptions.
Table 5.3
shows
that
the three MMR channels of Date 1 and CHV50 of Date 3 jointly achieve
an accuracy level of 95.3 percent.
When CHV50 of Date 1
and
three
MMR channels of Date 3 are available, an accuracy level of 93.8 per­
cent is achieved.
This is significant in that it involves only four
channels and that it excels any channel combination of the same sen­
sor systems and is comparable or better than single day measurements
using
two
sensor
systems.
Note that when CHV50 of both Dates are
combined, a poor accuracy level of only
62.5
percent
is
achieved
(Table 5.3).
It can be concluded that when
two
sensor
systems
are
used,
separate use of the systems with a time gap in between could provide
better crop classification than simultaneous use of the two systems.
Change
in
scene characteristics over time provides new information
that contributes to better contrast among scene categories.
165
CHAPTER 6
SUMMARY AND CONCLUSIONS
6.1
Summary of Study
After an overview of remote sensing in general and its applica­
tions
to agriculture, problem areas were identified and a course of
investigation was undertaken to further understand the
of
optical
and
capabilities
microwave channels, individually or as a group, to
differentiate various crop categories.
The objectives of this study were:
(1) Compare the relative ability of selected channels of op­
tical and microwave sensors to discriminate various crop types.
(2) Examine the complementary nature of
tical
channels
in
an
effort
to
define
microwave
and
op­
operational multisensor
specifications for crop discrimination.
(3) Investigate the tradeoffs between the number of channels
employed and the temporal variable used for crop discrimination.
As a first step to fulfill the above objectives,
literature
was
reviewed
microwave
remote
ducted over a number of
Lawrence, Kansas
extensive
to consolidate the existing body of know­
ledge and understanding relevant to combined
and
an
sensing.
operation
of
optical
Then, a field experiment was con­
commercial
agricultural
fields
north
of
during a large part of their growth cycles in the
summers of 1982 and 1983.
166
The sensor systems used were
(MMR) with
eight
(TM) channels, and
a
Modular
Multiband
Radiometer
channels that included the seven Thematic Mapper
two
Mobile
Agricultural
Radar
Scatterometers
(MARS) operating at C-band (5.04 GHz) and X-band (10.2 GHz) with in­
cidence angles set at 20 and 50 degrees each.
Chapter
3
contained
the descriptions of the sensor systems, operations and data acquisi­
tion procedures, and field data calibration procedures.
The 1982 experiment consisted of radar
ments
scatterometer
measure­
and supporting ground truth observations from ten fields each
of winter wheat, corn,
capability
of
and
soybeans.
In
order
to
examine
active microwave signatures to discriminate crops at
different stages of growth, five time frames were selected
performance
of
each
channel
or
group
evaluated. The criteria used to measure
channel
or
of
the
radar
and
the
channels
was
performance
of
each
group of channels were the classification accuracies of
agricultural scene categories sampled in the
tion,
the
canonical
techniques
timeframe.
In
addi­
were employed to look into the complex
structure of the multiresponse data set.
In the summer of 1983, MMR and MARS
taneously
to
categories.
range
and
acquire
signatures
were
used
almost
simul­
from selected agricultural scene
The MMR had the seven channels of the TM in the optical
the
MARS systems involved four radar channels of CHH50,
CHV50, XW50, and XVH50
for
this
experiment.
Three
Dates
were
selected and analyzed based on the results of field data evaluation,
16?
availability of both MMR and
MARS
observations,
scene
categories
mutually viewed, and sample sizes for each category observed,
three
Dates were selected and analyzed.
6.2 Summary of Major Results
Major findings are summarized as follows:
(1) XW50 performs better than XVH50
winter wheat and corn.
to
categories.
between
During the timeframe from May 19 to July 13,
1982, XW50 employed on June 16 was the best
crop
discriminate
to
classify
the
two
As the crops mature, the performance of XVH50 in­
creases gradually, but it does not exceed that of XW50.
(2) XW50 is slightly better than XVH50 to
winter
wheat, corn, and soybeans.
discriminate
among
Again, June 16 was the best time
to correctly classify the three crop categories.
(3) When the timeframe shifts to July - August, XVH50 generally
outperforms
XW50.
Wheat was harvested and double-cropped soybeans
emerged by the end of July, 1982. XVH50 on July
best
to
correctly
classify
corn,
regular
26
the
soybeans, and double-
cropped soybeans, with an overall accuracy of 61.8
four
performed
percent.
When
radar channels were used, a classification accuracy of over 80
percent was achieved generally.
(4) The performance levels of the radar channels gradually drop
as
crops
typical
become
plant
backscatter.
mature and drier.
leaves
The
are
July
a
This supports the reports that
dominant
contributor
to
radar
timeframe seems to be the best to classify
168
wheat, corn, and soybeans, although wheat was
either
harvested
or
ready for harvest by late July.
(5) When eight radar channels of dual polarized C- and
at
incidence
angles
of 20 and 50 degrees were employed, CHV50 had
the highest correct classification level (55.9
regular
X-bands
percent) for
corn,
soybeans, and double-cropped soybeans in August 1982.
The
next best was XVH20 with 47.1 percent. This result contradicts
the
findings
that
X-band is better than C-band for crop discrimination
based on microscale modeling and
source
of
radar
return.
investigations
to
determine
However, this result coincides with the
results of studies using airborne radar scatterometers.
polarized
the
Also, like-
channels performed more poorly than cross-polarized chan­
nels.
(6) A radar depolarization ratio does not
perform
any
better
than the combination of dual polarization used to compute the ratio.
However, it seems to have the
crops.
capability
to
which
is
equivalent
discriminate crop types.
TM3
of
certain
Therefore, the ratio can be used selectively.
(7) Optical channels outperform any radar
M3
discriminate
the
Thematic
to
TM3
was
found
channels.
to
be
the
Overall,
best to
This coincides with the design purpose
Mapper.
The
M3
of
channel alone obtained an
overall classification accuracy of 82.5 percent for corn, cut grass,
hay, soybeans, winter wheat, and spring wheat on July 6, 1983.
was almost the same
as
the
81.7
percent
169
correct
This
classification
achieved by combining all four radar channels.
(8) Although M7 (equivalent to TM7) is
designed
for
geologic
applications, the channel was found to be valuable in crop studies.
For measurements made on July 19, 1983, M3, Ml, and
CHV50
achieved
an overall accuracy of 82.5 percent to correctly classify corn, fal­
low, potatoes, soybeans, swamp, and winter wheat.
to
When M7 was added
the three channels, the accuracy level increased to 92.5 percent
due mainly to classification improvement in
categories.
swamp
and
potato
The swamp was dry and the potato field was rather bare
because of planting practice and moisture
fields
the
became
drier
stress.
When
the
and leafy categories were not included in the
classification, M7 turned out to be the best
classifier
among
seven
Therefore,
M7
optical
valuable for
and
test
four
agricultural
radar
channels.
applications
depending
categories in question and the area under study.
upon
the
can be
the
crop
It is particularly
valuable for semi-arid areas such as dryland farming situation.
(9) The practical number of channels needed to achieve a satis­
factory
classification level is four whether they are from one sen­
sor system or combined from two
channels
reach
sensor
systems.
a "plateau" of the accuracy curve.
Generally, four
The addition of
extra channels does not influence significantly the combined perfor­
mance of channels.
(10) When two sensors using different regions of
are
operated
in
the
spectrum
common, the classification accuracy improves sig­
170
nificantly.
(a
Two temporal measurements using three optical
dimension
of six) achieved an accuracy level of 91.3 percent to
classify four crop categories, but when the three
were
channels
operated
in
common
with
optical
channels
CHV50 the same day (a dimension of
four), the classification accuracy was 92.5 percent with four
categories
in early July, and 95.3 percent with the same crop types
a month later.
common,
it
When two different sensor systems
are
employed
in
is certain that the performance level increases impres­
sively because each sensor provides unique
ferent
cover
information
while
dif­
channels of the same sensor inevitably include redundant in­
formation.
(11) When two different sensor systems are used separately with
a
time
gap
in between, the accuracy level is comparable or better
than that obtained by two sensor systems used in common on one
day.
In a simulated situation where only the MMR was available because of
radar failure one day and only CHV50 was available
one
on
another
day
month later because of cloud cover, an overall accuracy of 96.3
percent was achieved when the two temporal data sets were
combined.
In a vice versa situation where radar was used on the first Date and
the MMR one month later, the accuracy level was 93.8 percent.
is
impressive
because CHV50 alone achieved an accuracy of a little
more than 50 percent under the same condition
involve
only
This
four
channels.
It
seems
and
both
situations
that the new information
provided by the two sensor systems is further enhanced
by
temporal
separation
that
contributes
to
better
contrast
among
crop
categories.
6.3 Limitations and Recommendations
The limitations of this investigation can also be considered as
recommendations for further research.
(1) The
Because
it
experiment
is
and
possible
its
They are:
results
are
location-specific.
to discriminate certain crop types in one
given physiographic region, this may not demonstrate
the
universal
application for all regions and for all crops.
(2) It should be emphasized that the
GHz),
cross-polarization,
and
50
degree
represents a result from analyzing the
categories
on
of
C-band (5.04
incidence angle (CHV50)
effect
of
different
scene
a few selected active microwave channels, and conse­
quently is not intended to imply optimum
crop
choice
discrimination.
sensor
specification
Furthermore, since the microwave range is so
dynamic and there are such a large number of sensor
binations, there
is
for
parameter
com­
a need for repetitive case studies using dif­
ferent system parameters in many different agricultural environments
with different crop types, varieties, and diseases.
nels in this study were limited to Cwere
preset
studies.
the
for
soil
moisture
and
X-bands.
The radar chan­
The
channels
detection and crop discrimination
The incidence angle of 50 degrees was available throughout
field
operations,
but
the
20
degree
incidence
available only for one timeframe in this report.
1?2
angle
was
(3) The general choice of the July time
to
best
discriminate
crops should be interpreted as the time when most vegetation is lush
thus providing the best representation of crop characteristics.
best
The
time should be understood as flexible because the time history
of measurements is not continuous in terms of measurement dates
data
quality
in
one
categories over time.
part, and
observation
of
and
different
Also recall that winter wheat
was
crop
harvested
in July.
(4) The results are from ground-based investigations only.
In
conduction with ground-based operations, there is a need to fly air­
borne missions and satellite overpasses to analyze the relationships
among
the
sensor,
the
teristics.
Similar
studies
representation
and
atmosphere,
image
to
and agricultural crop charac­
produce
results
in
cartographic
format should be conducted for practical
applications.
(5) The results are derived from the evaluation of channels
classify
given crop categories.
to
It is necessary to examine optical
and microwave sensor returns in relation with quantified target crop
charateristics
such
as
leaf area index, canopy height, plant den­
sity, percent ground cover, soil moisture, and so on.
applications, an
attempt should be made to combine collateral data
such as soil surveys, precipitation, elevation, and so
precisely
identify
vigor and health.
For practical
not
only
crop
on
to
more
categories but also vegetation
Overall evaluation of channels and microscale in­
173
vestigations
to
determine sources of signal returns should be con­
ducted in close coordination.
(6) In terms of field measurement efforts, every possible
was
exercised to acquire reliable observations.
care
However, there are
several factors that could have affected sensor operations including
speed of the truck, angle adjustment of the truck boom, and defining
the sampling locations.
from
For example, data should
have
been
taken
a large number of fields instead of from several general loca­
tions with minimal movement of the truck.
(7) There was a problem of absolute radar calibration from
system
to
another
and
from year to year.
one
In case of radar scat-
terometers, field data were taken over the two summers with modified
systems each time.
(8) MMR and radar
because
operations
were
not
exactly
simultaneous
of personnel and logistic requirements involved.
Sometimes
windows were defined in the crop growth cycle to obtain large enough
sample sizes for analysis.
6.4 Conclusion
The first requirement for the successful application of
sensing
to
agriculture
is
accurate crop identification.
sensors, especially the Thematic Mapper, provide
tion
on
vegetation
discriminating crops.
and
are
superior
to
detailed
microwave
Mainly because of atmospheric
remote
Optical
informa­
sensors
in
interruptions,
however, a complete time history of agricultural observations using
17^
optical sensors can not be realized.
tively
immune
to
sensing capability.
common
is
far
Microwave
sensors
complementary
rela­
weather conditions and possess the day and night
Operation of optical and microwave
more
effective
than
either
nature
sensors
sensor
discriminating crops because it provides an opportunity
the
are
alone
to
in
in
utilize
of the two different sensing techniques.
If the two sensor systems are used with
a
time
gap
in
the
crop
growth cycle, further improvement in crop discrimination is achieved
due to the fact that the unique information provided by each
system
sensor
is further enhanced by temporal separation which contributes
to better contrast among crop categories.
175
REFERENCES
Ahern, F. J., D. G. Goodenough, A. L. Grey, R. A.. Ryerson, R. J.
Vilbikaitis, and M. Goldberg, 1978. Simultaneous Microwave and
Optical Wavelength Observations of
Agricultural
Targets,
Canadian Journal of Remote Sensing, 4, No. 2, August, pp.
127-142.
Allen C. T., 1984. Modeling the Temporal Behavior of the Microwave
Backscattering Coefficient of Agriculttural Crops, Ph. D. Dis­
sertation, University of Kansas.
Aslam, A., F. Ulaby, B. Jung, M. Hemmat, M. Dobson, D. Brunfeldt,
and J. Paris, 1983. Multi-Channel MARS C/X-Band Scatterometer
System, Vol. 1: Design and Operation, Vol. 2: System Specifica­
tions and Experimental Data, RSL Technical Report 580-1,
University of Kansas Center for Research, Inc., Lawrence, Kan­
sas.
Attema, E. P. W., and F. T. Ulaby, 1978. Vegetation Modeled
Water Cloud, Radio Science, Vol. 13, No. 2, pp. 357-364.
as
a
Barnes Engineering (Barnes), 1979. Operation and Service Instruc­
tions: Modular Multispectral 8-Channel Radiometer, Stamford,
Connecticut.
Batlivala, P. P., and F. T. Ulaby, 1975. Crop Identification from
Radar Imagery of the Huntington County, Indiana Test Site, RSL
Technical Report 177-58, University of Kansas Center for
Research, Inc., Lawrence, Kansas, November.
Batlivala, P. P., and F. T.Ulaby, 1976. Radar Look Direction and
Row Crops, Photogrammetric Engineering and Remote Sensing, Vol.
42, No. 2, pp. 233-238.
Bauer, M. E., V. C. Vanderbilt, B. F. Robinson, and C. S. Daughtry,
1980.
Spectral Properties of Agricultural Crops and Soils
Measured from Space, Aerial, Field and Laboratory Sensors,
Proceedings of the International Archives of Photogrammetry,
XXIII (B7j: 5S-73, International Society for Photogrammetry,
XIV Congress, Hamburg, West Germany, July 13-15.
Berger, D. H., 1970. Texture as a Discriminant of Crops on Radar
Imagery, IEEE Transactions on Geoscience Electronics, Vol.
GE-8, No. 4, October, pp. 334-3?8.
Biehl, L. L., 1982. Data Processing, Management, and Base, Paper
presented at the 1982 Machine Processing of Remotely Sensed
1?6
Data Symposium, Purdue University.
Bradley, G. A., and F. T. Ulaby, 1980. Aircraft Radar Response to
Soil Moisture, RSL Technical Report 460-2, University of Kansas
Center for Research, Inc., Lawrence, Kansas.
Brisco, B., and R. Protz, 1982. Manual and Automatic Crop Iden­
tification
with
Airborne
Radar Imagery, Photogrammetric
Engineering and Remote Sensing, Vol. 48, No. 1, pp. 101-99.
Bush, T. F., and F. T., Ulaby, 1975. Fading Characteristics of Pan­
chromatic Radar Backscatter from Selected Agricultural Targets,
IEEE Transactions of Geoscience Electronics, Vol. GE-13, No. 4,
October, pp. 149-157.
Bush, T. F., and F. T. Ulaby, 1975.
On the Feasibility of
Monitoring Croplands with Radar, Proceedings of the 10th Inter­
national Symposium on Remote Sensing of Environment, Environmental Research Institute of Michigan, Ann Arbor, October 6-10,
pp. 1111-1121.
Bush, T. F., and F. T. Ulaby, 1976. Radar Return from
Vegetation
Canopy,
IEEE
Transactions
on
Propagation, Vol. AP-24, No. 3, pp. 269-276.
a Continuous
Antennas and
Bush, T. F., and F. T. Ulaby, 1977. Cropland Inventories Using an
Orbital Imaging Radar. RSL Technical Report 330-4, University
of Kansas Center for Research, Inc., Lawrence, Kansas, January.
Bush, T. F., and F. T. Ulaby, 1978. An Evaluation of Radar as a
Crop Classifier, Remote Sensing of Environment, 7, pp. 15-36.
Calla, 0. P. N., N. S. Pillai, 0. P. Kaushik, and Sivaprasad, 1979.
Temporal Study on Paddy (Rice) Using X-Band Scatterometer,
Paper presented at the 13th International Symposium on Remote
Sensing of Environment, Ann Arbor, University of Michigan, pp.
23-27.
Clark, J., 1982. Improved Land Use Classification from Landsat and
Seasat Satellite Imagery Registered to a Common Map Base, Tech­
nical Papers Presented at the 47th Annual Meeting of the
American
Society
of
Photogrammetry, Washington, D. C.,
February, pp. 22-27.
Colwell, R. N., 1960. Some Uses and Limitations of Aerial Color
Photography in Agriculture, Photogrammetric Engineering, Vol.
26, No. 2, pp. 220-222.
Colwell, R. N., 1973. Remote Sensing as an Aid to the Management of
Earth Resources, American Scientist, Vol. 61, pp. 175-183.
1??
Cooley, W. W., and P. R. Lohnes, 1971. Multivariate Data
John Wiley and Sons, New York.
Analysis,
Daughtry, C. S. T., V. C. Vanderbilt, and V. J. Pollara, 1981.
Variability of Reflectance Measurements with Sensor Altitude
and Canopy Type, Technical Report SR-Pl-04191, Laboratory for
Applications
of
Remote Sensing, Purdue University, West
Lafayette, Indiana.
Davis, J. C., 1973. Statistics and Data Analysis in
Wiley and Sons, New York.
Geology, John
de Loor, G. P., and A. A. Jurriens, 1974.
The Radar Backscatter
from
Selected
Agricultural
Crops, IEEE Transactions on
Geoscience Electronics, Vol. GE-17, No.2, April
de Loor, G. P., P. Hoogeboom, and E. P. W. Attema, 1982. The Dutch
Rove Program, IEEE Transactions on Geoscience and Remote Sen­
sing, Vol. GE-20, No. 1, January, pp. 3-11.
Dickey, H. P., J. L. Zimmerman, R. 0. Plinsky, and R. D. Davis,
1977.
Soil Survey of Douglas County, Kansas, U. S. Department
of Agriculture Soil Conservation Service, July.
Drake, B., R. A. Shuchman, R. F. Rawson, F. L. Smith, and R. W. Lar­
son, 1974.
Feasibility of Using Multiplex SLAR Imagery for
Water Resource Management and Mapping Vegetation Communities,
9th International Symposium on Remote Sensing of Environment,
Environmental Research Institute of Michigan, April, pp. 15-19.
Duda, R. 0., and P. E. Hart, 1973. Pattern Classification and Scene
Analysis, John Wiley and Sons, New York.
Elachi, C., and J. Granger, 1982. Spaceborne Imaging Radars
'in Depth', IEEE Spectrum, Vol. 19, No. 11, pp.24-29.
Probe
Erb, R. B., and B. H. Moore, 1978. The Large Area Crop Inventory
Experiment (LACIE): A Summary of Three Years' Experiment,
Proceedings of the 5th Canadian Symposium on Remote Sensing,
Victoria, B.C.
Ernst, H., and G. Fischbeck, 1984. X- and L-Band Radar Backscatter
Analysis of Agricultural Crops in West Germany, Proceedings of
IGARSS'84 Symposium, Strasbourg, August 27-30.
EROS Data Center (EDC), 1982. Landsat Data Users Notes, No.
U.S. Geological Survey, Sioux Falls, S. Dakota, July.
23,
EROS Data Center (EDC), 1984. Landsat Data Users Notes, No.
U.S. Geological Survey, Sioux Falls, S. Dakoda, March.
30,
178
Eyton, J. R., R. Li, and F. T. Ulaby, 1979.
Combined Radar and
Landsat Multitemporal Crop Classification, RSL Technical Report
360-10, University of Kansas Center for Research,
Inc.,
Lawrence, Kansas, February.
Ford, J. P., R. G. Blom, M. L. Bryan, M. I. Daily, T. H. Dixon, C.
Elachi, and E. C. Xenos, 1980. Seasat Views North America, the
Caribbean, and Western Europe with Imaging Radar, JPL Publica­
tion 80-67, Jet Propulsin Laboratory, California Institute of
Technology, Pasadena, California.
Fung, A. K., and F. T. Ulaby, 1983.
Matter-Energy Interaction in
the Microwave Region, Manual of Remote Sensing, Chapter 4, Vol.
I, R. N. Colwell (ed.), American Society of Photograimnetry.
Gabel, P. F., Jr., F. T. Ulaby, and D. R. Brunfeldt, 1981. MARS
X-Band, Scatterometer, RSL Technical Report 360-14, University
of Kansas Center for Research. Inc.,
Lawrence,
Kansas,
November.
Gnanadesikan, R., 1977. Methods for Statistical Data Analysis
Multivariate Observations, John Wiley and Sons, New York.
of
Goodman, M. S., 1959. A Technique for the Identification of Farm
Crops on Aerial Photographs, Photogrammetric Engineering, Vol.
25, No. 1, March, pp. 131-136.
Haralick, R. M., F. R. Caspall, and D. S. Simonett, 1971.
Using
Radar Imagery for Crop Discrimination: a Statistical and Con­
ditional Probability Study, Remote Sensing of Environment, 1,
No. 2, pp. 131-142.
~
Henderson, F. M., 1979. Land Use Analysis of Radar Imagery,
Photogrammetric Engineering and Remote Sensing, Vol. 45. vv.
6
295-307.
Heydt, H. L., and A. J. McNair, 1973. Engineering Analysis of
Data for Rice in the Philippines, Cornell University.
ERTS
Hoffer, R. M., 1978.
Biological and Physical Considerations in
Analyzing Computer-Aided Analysis Techniques to Remote Sensor
Data, Remote Sensing: The Quantitative Approach, Chapter 5, P.
H. Swain and S. M. Davis (eds.), McGraw Hill, New York.
Horton, I. F., J. S. Russel, and A. W. Moore, 1968.
MultivariateCovariance and Canonical Analysis: A Method for Selecting the
Most Effective Discriminators in a Multivariate Situation,
Biometrics, Vol. 24, No. 4, Journal of the Biometric Society,
December, pp. 845-858.
179
Jennrich, R., and P. Sampson, 1981. Stepwise Discriminant Analysis,
BMDP Statistical Software, W. J. Dixon ed., University of
California Press, Berkeley, California, 1981, pp. 519-537.
Jensen, H., L. C. Graham, L. J. Porcello, and E. N. Leith, 1977.
Side Looking Airborne Radar, Scientific American, Vol. 237, No.
4, pp. 84-95.
Jenson, S. K., and F. A. Waltz, 1979. Principal component analysis
and Canonical Analysis in Remote Sensing, Proceedings of the
American Society of Photogrammetry, Vol. "TJ 45th
Snnual
Meeting, March 18-21, 1979, Washington, D.C., pp. 337-348.
Johnston, R. J., 1980.
Multivariate
Geography, Longman Inc., New York.
Statistical
Analysis
in
Jung, B., K. Gillespie, M. Hemmat, A. Aslam, D. Brunfeldt, M. Dobson, and F. Ulaby, 1983. Radar Remote Sensing for Crop Clas­
sification and Canopy Condition Assessment:
Ground
Data
Documentation, RSL Technical Report 580-2, University of Kansas
Center for Research, Inc., Lawrence, Kansas, April.
Kumar, R., and L. F. Silva, 1977.
Separability of Agricultural
Cover Types by Remote Sensing in the Visible and Infrared
Wavelength Regions, IEEE Transactions on Geoscience Elec­
tronics, Vol. GE-15, No. 1, January, pp. 42-79.
Lancaster, C. W., and A. M. Feder, 1966. The Multisensor
Photogrammetric Engineering, Vol. 35, pp. 484-494.
Mission,
Langrebe, D. A., 1978. Useful Information from Multispectral Image
Data: Another Look, Remote Sensing: The Quantitative Approach,
Chapter 7, P. H. Swain and S. M. DaviFTEds.), McGraw-Hill, New
York, pp. 336-374.
Larson, R. W., P. L. Jackson, R. J. Dallaire, R. Shuchman, and R. R.
Wsoon, 1975.
Interpretation and Measurement of Multichannel
Microwave SAR Imagery, 10th International Symposium on Remote
Sensing of Environment, Vol. 1, pp. 53-69.
Li, R. Y., K. S. Shanmugam, and F. T. Ulaby, 1982. Crop Classifica­
tion Using Radar and Landsat Data, IEEE Transactions on
Geoscience and Remote Sensing, Vol. GE-20, NO. 1, January, ppT
Lillesand, T. M., and R. W. Kiefer, 1979. Remote Sensing and
Interpretation, John Wiley and Sons, New York.
Image
Long, M. W., 1975. Radar Reflectivity of Land and Sea, D. C. Heath
and Co., Lexington, Massachusetts.
180
Lowe, D. S., and J. G. N.. Braithwaite, 1966. A Spectrum Matching
Technique for Enhancing Image Contrast, Applied Optics, Vol. 5,
No. 6, pp. 893-898.
MacDonald, 1983. A Summary of the History of the Development of
Automated Remote Sensing for Agricultural Applications, NASA
Johnson Space Center, Houston, Texas.
Maxwell, E. L., 1976. Multivariate System Analysis of Multispectral
Imagery, Photogrammetric Engineering and Remote Sensing, Vol.
42, No. 9, pp. 1173-1186.
Merembeck, B. F., and F. Y. Borden, 1978. Principal Components and
Canonical Analysis for Reduction of the Dimensionality of Large
Data Sets, ORSER Technical Report 5-78, Office for Remote
Sensing of Earth Resources, Pennsylvania State University,
University Park, Pennsylvania.
Merembeck, B. F., and B. J. Turner, 1980.
Directed Canonical
Analysis and the Performance of Classifiers Under its As­
sociated Linear Transformation, IEEE Transactions on Geoscience
and Remote Sensing, Vol. GE-18, No. 2, pp. 190-196.
Merembeck, B. F., F. Y. Borden M. H. Podwysocki, and D. N. Applegate, 1976. Application of Canonical Analysis to Multispec­
tral Scanner Data. Proceedings of the 14th Annual Symposium on
the Application of Computer Methods in the Mineral Industries,
October 4-8, 1976, University Park, Pennsylvania, pp. 867-879.
Moore, R. K., 1983. Radar Fundamentals and Scatterometer, Manual of
Remote Sensing, Chapter 9, 2d ed., Vol. I, R. N. Colwell ed.,
American Society of Photogrammetry, Falls Church, Virginia.
Moore, R. K., K. A. Soofi, and S. M. Purduski, 1980.
A Radar
Clutter Model: Average Scattering Coefficients of Land, Snow,
and Ice, IEEE Transactions on Aerospace and Electronic Systems,
Vol. AES-16, No. 6, November.
Morain, S. A., and J. C. Coiner, 1970.
An evaluation of Fine
Resolution Radar Imagery for Making Agricultural Determina­
tions, CRES Report 177-7, University of Kansas Center for
Research, Inc., Lawrence, Kansas.
Morain, S. A., and D. S. Simonett, 1967. K-Band Radar in Vegetation
Mapping, Photogrammetric Engineering, Vol. 33, No. 7, pp.
730-740.
National Aeronautics and Space Administration (NASA), 1981, 1981
Ground Data Survey, AgRISTARS Enumerators Manual, Lyndon B.
Johnson Space Center, Houston, Texas 77058.
181
Paris, J. F., 1983. Radar Backscattering Properties of Corn and
Soybeans at Frequencies of 16, 4.75, and 13.3 GHz, IEEE Tran­
sactions on Geoscience and Remote Sensing, GE-21, No. 3, July.
Podwysocki, M. H., F. J. Gunther, and H. W. Blodget, 1977.
Discrimination of Rock and Soil Types by Digital Analysis of
Landsat Data. NASA Preprint X-923-77-17, Goddard Space Flight
Center, Greenbelt, Maryland.
Robinson, A., R. Sale, and J. Morrison, 1978.
Elements
tography, 4th ed., John Wiley & Son, New York.
of
Car­
Robinson, B. F., M. E. Bauer, D. P. Dewitt, L. F. Silva, and V. C.
Vanderbilt, 1979.
Multiband Radiometer for Field Research,
SPIE Vol. 196-Measurements of Optical Radiations, Society of
Photo-optical Instrumentation Engineers, Bellingham, Washing­
ton, pp. 8-15.
Robinson, B. F., and L. L. Biehl, 1979. Calibration Procedures for
Measurement of Reflectance Factor in Remote Sensing Field
Research, SPIE Vol. 196-Measurements of Optical Radiations,
Society of Photo-optical Instrumentation Engineers, Bellingham,
Washington, pp. 16-26.
Robinson, B. F., R. E.Buckley, and J. A.Burgess, 1981.
Performance
Evaluation and Calibration of Modular Multiband Radiometer for
Remote Sensing Field Research, SPIE Vol. 308-Contemporary In­
frared Standards and Calibration, Society of Photo-optical In­
strumentation Engineers, Bellingham, Washington, pp. 146-157.
Rosenthal, W. D., and B. J. Blanchard, 1984.
Active Microwave
Responses: An Aid in Improved Crop Classification, Photogrammetric Engineering and Remote Sensing, Vol. 50, No. ZTj April,
pp.461-468.
Schwarz, D. E., and F. R. Caspall, 1968. The Use of Radar in the
Discrimination of Agricultural Land Use, Proceedings of the 5th
Symposium on Remote Sensing of Environment, Universitry of
Michigan, Ann Arbor, April, pp. 233-247.
Scrimshaw, N. S., 1965. Food, Readings in Agriculture from Scien­
tific American, W. H. Freeman & Co.. San Francisco, pp.
50^214:
Seal, H. L., 1964.
Multivariate
Statistical
Biologists, John Wiley & Sons, Inc., New York.
Analysis
for
Shanmugam, K. S., F. T. Ulaby, V. Narayanan, and C. Dobson, 1981.
Crop Classification Using Multidate/Multifrequency Radar Data,
RSL Technical Report 360-17, University of Kansas Center for
182
Research, Inc., Lawrence, Kansas, December.
Silva, L. F., 1978.
Radiation and Instrumentation in Remote
Sensing, Remote Sensing: The Quantitative Approach, Chapter 2,
P. H. Swain and S. M. Davis ed., McGraw Hill, New York.
Simmonet, D. S., J. E. Eagleman, A. Erhard, D. C. Rhodes, and D. E.
Schwarz, 1967.
The Potential of Radar as a Remote Sensor in
Agriculture I: a Study with K-band Imagery in Western Kansas.
CRES Report 61-21, University of Kansas, Lawrence, Kansas.
Slater, P. N., 1980. Remote Sensing: Optics and Optical Systems,
Addison-Wesley Publishing Co., Reading, Massachusetts.
Sobti, A., 1975.
Terrain Response to an Orbiting
Microwave
Radiometer/Scatterometer, Ph.D. Dissertation, University of
Kansas, Lawrence, Kansas.
Steiner, D., 1970. Time Dimension for Crop Surveys from
Photogrammetric Engineering, Vol. 36, pp. 187-194.
Space,
Stiles, W. H., D. Brunfeldt, and F. T. Ulaby, 1979.
Performance
Analysis of the MAS (Microwave Active Spectrometer) Systems:
Calibration, Precision, and Accuracy, RSL Technical Report
360-4,
University
of Kansas Center for Research, Inc.,
Lawrence, Kansas, April.
Tatsuoka, M. M., 1971.
New York.
Multivariate Analysis, John Wiley & Sons,
Tom, C. H., and L. D. Miller, 1984. An Automated Land-Use Mapping
Comparison of the Bayesian, Maximum Likelihood and Linear
Discriminant Analysis Algorithms, Photogrammetric Engineering
and Remote Sensing, Vol. 50, No. 2, February, pp. 193-207.
Turner, B. J., D. N. Applegate, and B. F. Merembeck, 1978.
Satel­
lite and Aircraft Multispectral Digital Data User Manual,
ORSER Technical Report 9-78, Office for Remote Sensing of Earth
Resources, Pennsylvania State University, University Park,
Pennsylvania.
Ulaby, F. T., 1975. Radar Response to Vegetation.
IEEE Transac­
tions on Antennas and Propagation, AP-23, No. 1, pp. 36-45.
Ulaby, F. T., T. F. Bush, and P. P. Batlivala, 1975. Radar Response
to Vegetation, Part II: 8-18 GHz Band, IEEE Transactions on An­
tennas and Propagation, Vol. AP-23, No. 5^ September, pp.
608-618.
183
Ulaby, F. T., 1982a. Radar Signatures of Terrain, Proceedings of the
IEEE, Vol. 70, No. 12, December, pp. 1410-142IT!
Ulaby F. T., 1982b. Review of Approaches to the Investigation of
the Scattering Properties of Material Media, International
Geoscience and Remote Sensing Sympositun (IGARSS) Digest, Vol.
Ulaby, F. T., and P. P. Batlivala, 1976.
Diurnal Variations of
Radar Backscatter from a Vegetation Canopy, IEEE Transactions
on Antennas and Propagation, Vol- AP-24, No. 1, pp. 11-17.
Ulaby, F. T., P. P. Batlivala,
Backscatter Dependence on
Soil Texture, Part I:
Geoscience
Electronics,
286-295.
and M. C. Dobson, 1978.
Microwave
Surface Roughness, Soil Moisture, and
Bare Soil, IEEE Transactions
on
Vol. GE-16, No. TT, October. PP7
Ulaby, F. T., P. P. Batlivala, and M. C. Dobson, 1979.
Microwave
Backscatter Dependence on Surface Roughness, Soil Moisture, and
Soil Texture, Part II:Vegetation-Covered Soil, IEEE Transac­
tions on Geoscience Electronics, Vol. GE-17, No. 2, April, pp.
3FS0. *
Ulaby, F. T., and G. Burns, 1979.
Statistical Properties of the
Radar Scattering Coefficient of Agricultural Crops, RSL Tech­
nical Report 360-2, University of Kansas Center for Research,
Inc., Lawrencxe, Kansas, July.
Ulaby, F. T., R. K. Moore, and A. K. Fung, 1981.
Microwave Remote
Sensing:
Active
and
Passive, Vol. I (Fundamentals and
Radiometry), Addison-Wesley publishing Co., Reading, Massachu­
setts.
Ulaby, F. T., R. K. Moore, and A. K. Fung, 1982a. Microwave Remote
Sensing: Active and Passive, Vol. II (Radar, Remote Sensing and
Surface Scattering and
Emission
Theory),
Addison-Wesley
Publishing Co., Reading, Massachussetts.
Ulaby, F. T., A. Aslam, and M. C. Dobson, 1982b. Effects of Vegeta­
tion Cover on the Radar Sensitivity to Soil Moisture, IEEE
Transactions on Geoscience and Remote Sensing, Vol. GE-20, No.
4, October,pp. 476-481.
Ulaby, F. T., R. Y. Li, and K. S. Shanmugam, 1982c.
Crop Clas­
sification Using Airborne Radar and Landsat Data, IEEE Transac­
tions
on Geoscience and Remote Sensing, Vol. GE-20, No. 1. pp.
5__
184
Walraven, R., 1978. Calculating the Position of the Sun, Solar
Energy, Vol. 20, Pergamon Press, Great Britain, pp. 393-397T
Wu, S. T., 1980.
An Improvement in Land Cover Classification
Achieved by Merging Microwave Data,with Landsat Multispectral
Scanner Data, ASCM-ASP Convention, St. Louis, Missouri, March
1980, pp. 293-309.
185
APPENDIX A
TIME HISTORY OF SELECTED SCENE CHARACTERISTICS (1982)
1. The plots included
in
this
Appendix
represent
the
time
history of selected scene characteristics for each field measured in
1982.
2. Soybean Fields 1, 2, 7, 8, and 9 were also Wheat
Fields
2,
3, 5, 6, and 7 double-cropped after the wheat harvest.
3. The complete listings of
pertinent
crop-canopy
measurements are reported in Jung et al. (1983).
186
and
soil
CORN FIELD 1
130
155
180
205
JULIAN DATE
230
255
280
Q-CANOPY
A-P1ANT MOISTURE
+-SOIL MOISTURE
CORN FIELD 2
330
100
220'
I-
50
z
LLI
a
tr
LD
OL
LLI
1 1O X
25
130
155
180
205
JULIAN DATE
A-PLANT MOISTURE
+-501L UOISTUHE
O-CANOPY HKICHT
18?
230
255
280
CORN FIELD 3
330
100
220'
O
2
50
z
UJ
1101
o
a:
w
25
£L "
130
155
180
205
JUUAN DATE
230
255
280
A-PLANT MOISTURE
-t-SOlt. MOISTURE
D-CAM0PY HEICIIT
CORN FIELD 4
100
K
£
330
75
220'
tn
o
2
iz
Ul
o
q:
uj
a.
50
1101
25|-
130
155
180
205
JUUAN DATE
a-plant moisture
••-soil moisture
o-canopy heicht
188
230
255
280
PERCENT MOISTURE
PERCENT MOISTURE
HEIGHT (CM)
HEIGHT (CM)
PERCENT MOISTURE
PERCENT MOISTURE
Ul
ui
a>
o
O
O
33
2
> ro
3
zo
rn
_ Ul
JO
01
O
DJ
1
oo
1
»
I
I
oi
-»
K>
M
01
o
o
HEIGHT (C»$
°
HEIGHT (CM)
o
ro
PERCENT MOISTURE
HEIGHT (CM)
PERCENT MOISTURE
HEIGHT (CM)
SOYBEAN FIELD 1
100
120
230
JUUAN DATE
*-PUW MOISTURE
•«-SOa. MOISTURE
O-CANOPr HEIGHT
SOYBEAN FIELD 2
120
100
u
C
1
O
2
-7K
75r-
i!
J80
Q
- 40
£
o
I
50 j-
z
U1
o
ir
ui
a.
i
25-
01—
155
180
205
230
JUUAN DATE
*-PUW MOISTURE
-t-SOIL MOISTURE
O-CANOPY HEICHT
192
255
280
305
UI
PERCENT MOISTURE
PERCENT MOISTURE
10
r*o
cn
r
tn
cn
cn
o
"T"
~T"
cn
O
O
a>
o
ro
cn
cn
M
• f
CD
o
u
o
o
HEIGHT (CM)
ro
o
mo
*
o
..-J..
CD
O
HEIGHT (CM)
ro
o
SOYBEAN FIELD 5
120
100
u
a:
cn
5o
o:
155
180
230
JUUAN DATE
A-PLANT MOISTURE
4-SOIL. MOISTURE
M-CANOPY HEIGHT
SOYBEAN FIELD 6
120
100
80
!iE
UJ
<J
4-0 I
a:
K
25
155
180
205
230
JUUAN DATE
A-PLANT MOTSTURE
+-SAL MOISTURE
M-CAWOPY HEJCHT
19^
255
280
305
PERCENT MOISTURE
cn
in
PERCENT MOISTURE
M
Ifl
O
in
O
-.j
cn
o
o
oo
o
fO
u
o
to
in
cn
to
CD
O
_ 1.
•p-
o
o
HEIGHT (CM)
fO
4-1+ _L
•P-
o
j_
i
00
o
HEIGHT (CM)
L
SOYBEAN FIELD 9
120
100
UJ
tz
V)
o
2
5
UJ
o
tr
w
40 x
a.
230
JUUAN DATE
MOtSIURE
•4—SOIL MOISTURE
Q-CANOPY HEIGHT
SOYBEAN FIELD 10
120
100
40 I
230
JUUAN DATE
*-PtANT MOISTURE
-t-SOIL MOISTURE
O-CANOPT HEIGHT
196
WHEAT FIELD 1
120
80
o
V-/
r * S r
I
E
o
H40 I
o1—
90
110
130
150
JUUAN DATE
170
190
210
••-PLANT MOISTURE
t-SOIL MOISTURE
D-TOTAL HEIGHT
»-TOP LEAF HQGHT(AFTER HEADING)
®-STUBBLE HEIGHT
WHEAT FIELD 2
120
100
ui
o:
cn
o
2
H
Z
Ui
o
w
40 x
a.
a.
150
JUUAN DATE
PLANT MOISTURE
••-SOIL MOISTURE
o-TOTAL HEIGHT
•-TOP LEAF HEIGHT(AFTER HEADING)
®-STUBBLE HEIGHT
197
WHEAT FIELD 3
100
150
JUUAN DATE
-•-PLANT MOISTURE
••—SOIL MOISTURE
m-TOTAL HEIGHT
•-TOP LEAF HEIGHT(AFTER HEADING)
®-STUBBLE HEIGHT
WHEAT FIELD 4
120
100
til
q:
1
o
2
O
40 x
a.
Ui
a.
150
JUUAN DATE
--PLANT MOISTURE
••—SOIL MOISTURE
o-TOTAL HEIGHT
"••TOP LEAF HEIGHT(AFTER HEADING)
<o-STUBBLE HEIGHT
198
WHEAT FIELD 5
120
100
40 I
150
JUUAN DATE
PLANT MOISTURE
••—SOIL MOISTURE
m-TOTAL HEIGHT
"-TOP UEAF HEIGHT(AFTER HEADING)
®-STUBBLE HEIGHT "
WHEAT FIELD 6
120
100
40 X
150
JUUAN DATE
—PLANT MOISTURE
••—SOIL MOISTURE
o-TOTAL HEIGHT
•-TOP UEAF HE1GHT(AFTER HEADING)
o-STUBBLE HEIGHT
199
WHEAT FIELD 7
120
100
ui
a:
2
o
o
2
o
UJ
Ui
o
a.
ui
X
£L
150
JUUAN DATE
PLANT MOISTURE
+—SOIL MOISTURE
n»-TOTAL HEIGHT
""-TOP LEAF H0GHT(AFTER HEADING)
®-STUBBLE HEIGHT
WHEAT FIELD 8
120
100
ui
a:
o
2
ui
o
40 x
a.
hi
a.
150
JUUAN DATE
—PLANT MOISTURE
••—SOIL MOISTURE
tD-TOTAL HEIGHT
"-TOP LEAF HE!GHT(AFTER HEADING)
®-STUBBLE HEIGHT
200
WHEAT FIELD 9
1120
100
40 X.
150
JUUAN DATE
A-PLANT MOISTURE
•(—SOIL MOISTURE
D-TOTAL HEIGHT
W-TOP LEAF HE1GHT(AFTER HEADING)
©-STUBBLE HEIGHT
WHEAT FIELD 10
120
100
/
150
JULIAN DATE
A-PLANT MOISTURE
H—SOIL MOISTURE
0-TOTAL HEIGHT
He—TOP LEAF HDGHTfAFTER HEADING)
©-STUBBLE HEIGHT
201
APPENDIX B
TEMPORAL VARIATIONS OF MEAN MMR AND RADAR MEASUREMENTS (1983)
1. The MMR measurements are calibrated into percent reflectance
and
the
mean
values
for each channel are plotted by day for each
field.
2. The radar power returns
have
been
calibrated
into
radar
backscattering coefficients and their mean values are plotted by day
for each field.
3. Wheat 1 and 2 are winter wheat.
202
Wheat 3 is spring wheat.
MMR Measurementa
Corn 1
60
50
Ch *
ch a
30
o 20
ch a
10
Ch 7
Ch 2
Ch 1
150
I
J.
160
170
_h ,
3
x
180
190
Julian Date
200
210
220
230
210
220
230
Radar Measurements
Corn 1
.
»
CHHSO
/
'*•/\/
XVH50
CHV90
150
160
170
180
180
Julian Date
203
200
MMR Meaaurements
Corn 2
60
50
•"N
K
40
©
Ch 4
§ 30
Ch a
U
o
«
?:
o
o: 20
10
Ch
Ch 7
Ch 3
Ch 1
150
160
170
180
190
Julian Date
200
210
220
230
Radar Measurements
Corn 2
-5.0xwso
-10.0
a
•o
O-15.0
CHH90
E
o
G5
XVHSO
-20.0
CHV30
-25.0 •
-30.0
150
160
170
180
100
Julian Oato
zob
200
210
220
230
MMR Moaauromonto
Fallow
60
50
K 40
Ch
Ch
30
20
Ch S
10
Ch 7
Ch 2
Ch 1
«-
150
160
170
x
x
J.
180
190
Julian .Date
200
210
220
230
Radar Measurements
Fallow
xwso
chhso
XVHSO
CHVSO
190
160
170
180
190
Julian Data
205
200
210
220
230
MMR Meaauramonta
Grass
180
190
Julfan Date
200
Radar Measurements
Crass
xwso
XVHSO
CHHSO
CHVSO
150
160
170
180
190
Julfan Data
206
200
210
220
230
MMR Meaauremonta
Hay
Ch o
ch a
190
230
Julian Date
Radar Measurements
Hay
-10.0
a
•u
o-IS.O
CHHSO
£
w
(/)
-20.0
-25.0
-30.0
150
160
170
180
190
Julian Date
207
200
210
220
230
MMR MaaBuromenta
Potato
60
50
K 40
Ch
s
30
Ch 4
o 20
Ch 3
10
Ch 7
Ch 1
x
150
160
170
180
x
x
190
Julian Date
200
210
220
230
Radar Measurements
Potato Reld
xwso
CHHSO
CHVSO
XVHSO
150
160
170
180
190
Julian Data
208
200
210
220
230
MMR Meoauremonta
Soybean 1
60
50
Ch 4
Ch
30
Ch B
3 20
10
o1—
150
Ch
Ch 3
Ch 2
Ch 1
160
170
160
190
Julian Date
200
210
220
230
Radar Measurements
Soybean 1
XW90
CHHSO
XVHSO
CHVSO
-30.0
190
160
170
180
190
Jullon Date
209
200
210
220
230
MMR Moasuromonto
Soybean 2
60
50 Ch *
40
•'
-
ch
e
©
o
c 30 o
o
®
C
o
Ch 5
a: 20
10
,«i-rr- • .
0
150
I
160
.
I
170
.
•
.....
180
Ch 7
Ch a
Ch l01 3
i
.
i
„
i
.
100
200
210
.
i
220
230
Julian Date
Radar Measurements
Soybean 2
-5.0
xwso
^-10.0
m
T>
w
CHHSO
O-15.0
e
XVHSO
a
CHVSO
a
-20.0 •
-25.0 -
-30.0
190
160
170
180
1S0
Julian Date
210
200
210
220
230
MMR Measurements
Swamp
60
50
K 40
Ch
30
Ch
® 20
Ch
10
150
160
170
X
I
180
190
200
210
220
230
220
230
Julian Date
Radar Measurements
Swamp
-
CHHSO
XVH30
CHVSO
150
160
170
180
190
Julian Data
211
200
210
MMR Measurements
Wheat 1
60
50
K 40
ch e
30
Ch B
Ch 4
® 20
Ch 7
Ch 3
Ch 2
Ch 1
X
150
160
170
180
•L.
X
190
Juiran Date
200
210
220
230
Radar Measurements
Wheat 1
XWSO
CHHSO
150
170
180
100
Julian Date
212
200
210
220
230
MMR Meaourerr its
Wheat 2
K 40
Ch
a
Ch 9
8 20
Ch 7
Ch 3
Ch 2
Ch 1
190
Julian Date
200
210
220
230
Rodar Measurements
Wheat 2
0.0
-5.0.
xwso
-10.0 -
m
•o
O-15.0
y
-
CHHSO
£
to
-20.0
XVH3CJ
~
—
CHVSO
-25.0
-30.0
150
160
170
180
190
Julian Oato
213
200
210
220
230
MMR Mflosurements
Wheat 3
K 40
Ch o
Ch 3
Ch +
180
230
Jultan Date
Radar Measurements
Wheat 3
0.0
-5.0
-10.0 -
m"
•o
XWSO
O-15.0
CHHSO
XVHSO
CHVSO
-25.0
-30.01—
150
100
ISO
Julian Data
214
200
210
220
230
APPENDIX C
MMR MEASUREMENTS (1983)
1. This Appendix tabulates calibrated MMR measurements.
2. Each observation by MMR is described in
first
three
lines.
The
line identifies crop type and conditions under which measure­
ments were made.
The mean and one standard deviation of 20
reflec­
tance
values
for each channel are listed in lines 2 and 3 for each
case.
The first column represents channel 1 (Ml), the second column
M2, and so on.
3.
harvest.
Soybean
3
is
double-cropped
Wheat 3 is spring wheat.
215
soybeans
after
the
wheat
Alfalfa
8.556
0.799
Code:16-801 Date: 8/ 3/83(215) Time 10:35 S.Z.Angle: 42.48
11.663 14.322 31.894 40.856 34.436 21.560 306.978 Mean
1.242
1.846
2.545
2.494
1.213
0.929
0.517 S.D.
Alfalfa
8.899
0.775
Code:16-801 Date: 8/ 3/83(215) Time 10:46 S.Z.Angle: 40.46
11.975 14.644 32.387 41.761 35.020 21.771 308.034 Mean
1.114
1.677
2.409
2.266
0.984
0.750
0.986 S.D.
Bare Soil Code:15-901 Date: 7/27/83(208) Time 11:58 S.Z.Angle: 27.18
10.536 13.537 18.014 24.9/8 37.038 39.013 33.544 300.528 Mean
0.764
0.831
0.599
0.538
1.171
1.083
3.058
0.548 S.D.
Bare Soil Code:15-901 Date: 7/27/83(208) Time 12: 3 S.Z.Angle: 26.45
10.629 13.643 18.194 25.212 37.157 39.092 33.818 302.488 Mean
0.705
0.755
0.607
0.748
1.025
0.957
2.716
0.685 S.D.
Bare Soil Code:15-901 Date: 8/25/83(237) Time 13: 5 S.Z.Angle: 28.37
8.146 10.196 13.411 18.294 30.239 34.730 32.208 294.210 Mean
0.412
0.616
0.806
1.070
1.112
1.445
0.903
1.164 S.D.
Corn 1
3.672
0.654
Code: 1-201 Date: 6/ 6/83(157) Time 11:36 S.Z.Angle: 26.96
4.668
5.678
9.930 18.439 19.073 13.362 308.598 Mean
0.885
1.376
1.764
2.392
2.304
1.510 15.956 S.D.
Corn 1
6.888
0.958
Code: 1-201 Date: 6/15/83(166) Time 13:44 S.Z.Angle: 16.64
8.714 10.291 18.371 28.187 29.773 21.819 299.723 Mean
1.184
1.934
1.101
1.461
2.673
2.701
1.512 S.D.
Corn 1
2.891
0.602
Code: 1-201 Date: 6/20/83(171) Time 10:46 S.Z.Angle: 36.01
4.338
3.928 25.288 30.103 16.815
8.768 305.346 Mean
1.058
1.120 11.325
9.827
4.533
2.274
1.854 S.D.
Corn 1
3.390
0.366
Code: 1-201 Date: 6/30/83(181) Time 12:48 S.Z.Angle: 17.35
5.368
4.214 37.078 35.619 21.293
8.422 308.391 Mean
0.557
1.069 10.050
5.445
2.244
2.691 0.848 S.D.
Corn 1
3.214
0.468
Code: 1-201 Date: 6/30/83(181) Time 12:54 S.Z.Angle: 16.86
4.883
4.070 32.192 31.539 19.843
8.698 308.121 Mean
0.577
1.042
8.713
4.379
1.798
2.580
0.931 S.D.
Corn 1
3.188
0.554
Code: 1-201 Date: 7/ 1/83(182) Time 9:46 S.Z.Angle: 48.17
4.667
3.861 30.579 31.773 21.206 10.548 304.625 Mean
0.821 1.156
8.362
4.612
3.152
3.617
0.583 S.D.
Corn 1
Code: 1-201 Date: 7/ 1/83(182) Time 9:52 S.Z.Angle: 47.00
3.041
4.390
3.561 30.857 31.079 19.804
9.240 304.524 Mean
0.577
0.894
1.155
7.678
4.649
3.506
3.120
0.548 S.D.
216
Corn 1
3.374
0.598
Code: 1-201 Date: 7/ 1/83(182) Time 10:10 S.Z.Angle: 43.51
4.965
3.944 35.784 35.940 22.942 10.650 304.677 Mean
0.861 1.174
7.239
4.535
3.923
3.542
0.592 S.D.
Corn 1
3.299
0.696
Code: 1-201 Date: 7/ 1/83(182) Time 10:15 S.Z.Angle: 42.54
4.839
3.981 34.246 34.477 22.689 10.987 304.591 Mean
1.023
1.311
7.235
4.430
3.830
3.791
0.637 S.D.
Corn 1
3.258
0.440
Code: 1-201 Date: 7/ 1/83(182) Time 10:44 S.Z.Angle: 36.99
4.744
3.931 31.480 32.285 21.581 11.041 304.870 Mean
0.533
1.171
7.333
3.515
2.336
3.276
0.653 S.D.
Corn 1
Code: 1-201 Date: 7/ 1/83(182) Time 10:49 S.Z.Angle: 36.05
3.371
4.946
4.305 31.339 32.605 22.261 11.614 304.853 Mean
0.535
0.643
1.110
6.773
3.035
2.608
3.481
0.801 S.D.
Corn 1
4.343
0.454
Code: 1-201 Date: 7/ 6/83(187) Time 9:34 S.Z.Angle: 50.90
6.190
5.466 35.186 36.986 26.163 14.364 300.673 Mean
0.726
0.794
8.927
5.834
3.320
2.677
0.587 S.D.
Corn 1
4.157
0.319
Code: 1-201 Date: 7/20/83(201) Time 12:22 S.Z.Angle: 22.51
5.919
4.963 36.989 34.205 20.224
8.852 310.270 Mean
0.495
0.474
3.227
1.668
1.247
1.334
0.963 S.D.
Corn 2
4.446
0.248
Code: 5-202 Date: 6/ 6/83(157) Time 12:12 S.Z.Angle: 21.39
5.404
6.238 11.147 18.937 20.065 15.230 315.800 Mean
0.358
0.338
1.182
1.495
1.495
1.153 18.276 S.D.
Corn 2
5.457
1.670
Code: 5-202 Date: 6/15/81(166) Time 14:41 S.Z.Angle: 23.57
7.508
7.811 20.988 26.569 22.405 16.917 302.345 Mean
2.247
2.739
5.143
5.485
5.987
6.724
1.742 S.D.
Corn 2
2.433
0.682
Code: 5-202 Date: 6/20/83(171) Time 11:30 S.Z.Angle: 27.95
3.494
2.812 20.660 25.692 12.352
6.907 306.246 Mean
1.162
0.755 10.090
8.527
3.639
2.026
1.051 S.D.
Corn 2
2.145
0.979
Code: 5-202 Date: 6/30/83(181) Time 9:58 S.Z.Angle: 45.76
2.990
2.031 29.348 24.178 12.292
4.180 306.349 Mean
1.401 0.767 11.050
9.186
4.697
1.644
0.370 S.D.
Corn 2
1.817
0.728
Code: 5-202 Date: 6/30/83(181) Time 10: 3 S.Z.Angle: 44.79
2.390
1.742 25.941 21.004 10.550
3.933 305.957 Mean
1.214
0.982 11.641
8.588
4.263
1.690
0.462 S.D.
Corn 2
2.820
0.298
Code: 5-202 Date: 7/ 6/83(187) Time 10:22 S.Z.Angle: 41.60
3.989
3.063 35.444 31.005 16.479
6.330 306.355 Mean
0.374
0.280
2.004
1.331
1.123
0.706
0.391 S.D.
21?
Corn 2
4.645
0.843
Code: 5-202 Date: 7/20/83(201) Time 12:44 S.Z.Angle: 20.06
6.040
5.542 32.098 30.954 18.428
8.493 311.393 Mean
1.206
1.094
4.721 4.421
3.204
1.697
0.442 S.D.
Corn 2
3.866
0.299
Code: 5-202 Date: 7/27/83(208) Time 11:16 S.Z.Angle: 34.03
5.480
4.588 33.752 31.624 18.100
7.850 312.128 Mean
0.538
0.578
4.119
2.575
1.674
1.519
0.425 S.D.
Corn 2
3.716
0.366
Code: 5-202 Date: 7/27/83(208) Time 11:22 S.Z.Angle: 32.99
5.176
4.256 31.978 29.806 16.810
7.113 312.409 Mean
0.688
0.739
5.886
3.604
1.488
1.024
0.514 S.D.
Corn 2
Code: 5-202 Date: 8/25/83(237) Time 13:13 S.Z.Angle: 28.21
5.641
8.382
8.603 27.994 32.383 24.575 14.785 297.666 Mean
0.454
0.761 1.336
2.703
2.116
1.835
1.604
0.368 S.D.
Fallow
Code:12-601 Date: 6/ 6/83(157) Time 15:29 S.Z.Angle: 32.63
3.491 5.049
4.894 16.763 18.687 14.161
7.762 299.505 Mean
1.810
2.466
3.429
6.793
8.992
9.121
6.634
6.614 S.D.
Fallow
4.327
0.590
Code:12-601 Date: 6/17/83(168) Time 11:11 S.Z.Angle: 31.27
7.130
5.071 30.783 30.228 19.274
8.830 302.966 Mean
0.938
1.179
4.473
3.108
2.085
2.050
1.008 S.D.
Fallow
2.268
0.449
Code:12-601 Date: 6/20/83(171) Time 13: 4 S.Z.Angle: 15.91
3.836
1.974 32.591 22.533
9.649
2.616 308.004 Mean
0.771 0.611
6.846
4.980
2.498
0.954
0.684 S.D.
Fallow
2.955
0.826
Code:12-601 Date: 6/30/83(181) Time 12:24 S.Z.Angle: 19.99
5.121
2.601 35.558 27.888 14.181 5.396 309.282 Mean
1.604
1.179
4.914
5.636
3.970
1.619
0.836 S.D.
Fallow
3.417
0.870
Code:12-601 Date: 6/30/83(181) Time 12:30 S.Z.Angle: 19.24
6.011
3.253 39.018 31.689 16.658
5.757 308.797 Mean
1.811
1.486
6.227
5.814
3.979
1.749
0.915 S.D.
Fallow
6.625
0.292
Code:12-601 Date: 7/ 6/83(187) Time 11:42 S.Z.Angle: 26.91
10.139 10.044 33.136 39.442 26.747 13.694 303.011 Mean
0.407
0.967
2.553
3.113
1.740
0.981
0.757 S.D.
Fallow
4.580
0.177
Code:12-601 Date: 7/20/83(201) Time 9:31 S.Z.Angle: 52.79
6.624
5.067 32.213 32.066 19.037
8.128 308.149 Mean
0.094
0.362
1.628
0.951
0.881
0.627
0.353 S.D.
Fallow
4.864
0.160
Code:12-601 Date: 7/20/83(201) Time 9:37 S.Z.Angle: 51.62
6.702
5.253 31.117 31.033 18.555
8.160 308.039 Mean
0.192
0.396
1.709
1.394
0.861
0.487
0.473 S.D.
218
Grass
4.656
0.346
Code: 7-502 Date: 6/ 6/83(157) Time 12:58 S.Z.Angle: 16.85
8.351
6.667 36.287 38.170 23.715 10.215 291.012 Mean
0.382
0.520
1.042
0.849
1.206
0.898
2.942 S.D.
Grass
5.457
0.259
Code: 7-502 Date: 6/15/81(166) Time 14:54 S.Z.Angle: 25.69
9.163
7.802 36.877 40.254 26.043 11.593 305.974 Mean
0.437
0.643
1.233
0.771
1.102
0.819
0.689 S.D.
Grass
3.817
1.241
Code: 7-502 Date: 6/20/83(171) Time 11:45 S.Z.Angle: 25.38
6.412
5.986 31.890 32.658 19.727
8.056 305.426 Mean
2.103
1.917
6.979
8.003
5.506
2.464
0.769 S.D.
Grass
3.017
0.999
Code: 7-502 Date: 6/30/83(181) Time 10:14 S.Z.Angle: 42.66
4.837
3.844 29.064 31.676 18.835
6.132 306.481 Mean
1.687
1.625 13.158 16.684 12.048
4.184
0.410 S.D.
Grass
2.897
1.204
Code: 7-502 Date: 6/30/83(181) Time 10:20 S.Z.Angle: 41.51
4.597
3.674 27.011 30.325 18.126
7.090 307.027 Mean
1.996
1.873 14.111 17.800 12.735
4.305
0.588 S.D.
Grass
4.279
0.345
Code: 7-502 Date: 7/ 6/83(187) Time 10:38 S.Z.Angle: 38.54
7.176
5.678 35.901 37.453 21.691
8.761 303.948 Mean
0.502
0.706
1.449
0.650
0.604
0.601
0.423 S.D.
Grass
4.312
0.255
Code: 7-502 Date: 7/20/83(201) Time 9:55 S.Z.Angle: 48.13
6.966
6.288 31.364 34.23S 20.307
8.495 307.483 Mean
0.371
0.391
0.862
0.591
0.630
0.415
0.391 S.D.
Grass
4.996
0.190
Code: 7-502 Date: 7/27/83(208) Time 11:36 S.Z.Angle: 30.63
8.172
7.168 35.052 38.179 23.148 10.011 312.745 Mean
0.353
0.312
1.435
1.115
0.812
0.441
0.283 S.D.
Grass
5.074
0.225
Code: 7-502 Date: 7/27/83(208) Time 11:41 S.Z.Angle: 29.82
8.214
7.229 35.004 38.336 23.260
9.911 312.217 Mean
0.400
0.330
1.238
0.947
0.673
0.396
0.405 S.D.
Hay
5.027
0.512
Code: 3-501 Date: 6/ 6/83(157) Time 11:20 S.Z.Angle: 29.73
9.024
8.702 25.945 30.640 22.953 16.673 285.269 Mean
0.485
0.881
2.226
1.272
1.773
2.004 26.187 S.D.
Hay
4.600
0.754
Code: 3-501 Date: 6/15/81(166) Time 14:16 S.Z.Angle: 19.92
12.143
8.651 31.498 32.248 19.795 10.100 303.167 Mean
2.125
1.995
7.625
6.958
3.984
2.308
1.480 S.D.
Hay
3.110
0.226
Code: 3-501 Date: 6/20/83(171) Time 11: 0 S.Z.Angle: 33.38
5.894
4.429 27.936 31.643 20.013
8.291 302.856 Mean
0.498
0.272
3.566
2.720
1.644
0.832
0.627 S.D.
219
Code: 3-501 Date: 6/30/83(181) Time 9:22 S.Z.Angle: 52.77
Hay
5.816
3.877 27.819 31.788 19.846
7.892 301.568 Mean
3.115
0.597
0.372
5.950
4.320
1.470
0.718
0.383 S.D.
0.231
Hay
3.177
0.338
Code: 3-501 Date: 6/30/83(181) Time 9:27 S.Z.Angle: 51.79
5.888
3.856 28.430 32.312 19.167
8.160 301.616 Mean
0.795
0.434
6.059
4.935
3.415
1.037
0.417 S.D.
Hay
3.822
0.289
Code: 3-501 Date: 7/ 6/83(187) Time 9:54 S.Z.Angle: 47.01
6.722
4.566 31.727 35.436 21.710
8.728 302.760 Mean
0.869
0.436
5.115
4.096
2.101
0.969
0.419 S.D.
Milo
5.129
1.236
Code:16-909 Date: 8/ 3/83(215) Time 13:36 S.Z.Angle: 21.60
7.334
8.091 26.947 30.015 20.712 11.429 313.400 Mean
1.168
1.496
3.956
3.446
3.437
2.883
0.305 S.D.
Milo
4.896
1.050
Code:16-909 Date: 8/ 3/83(215) Time 13:44 S.Z.Angle: 21.89
7.048
7.749 26.136 28.623 19.914 11.029 310.196 Mean
1.298
1.261
4.660
4.287
2.903
2.622
7.678 S.D.
Potato
6.700
1.373
Code:ll-401 Date: 6/ 6/83(157) Time 15: 1 S.Z.Angle: 27.67
9.178
9.401 25.056 28.607 25.105 18.917 293.192 Mean
1.611
2.931
6.216
5.581
6.637
6.846
3.734 S.D.
Potato
6.764
2.126
Code:ll-401 Date: 6/ 6/83(157) Time 15:13 S.Z.Angle: 29.75
9.454
9.727 28.383 31.108 25.690 18.510 291.493 Mean
2.574
3.612
6.460
6.268
7.375
7.291
7.288 S.D.
Potato
6.056
2.226
Code:11-401 Date: 6/17/83(168) Time 10:56 S.Z.Angle: 34.04
8.867
8.528 33.819 34.428 25.825 15.833 302.072 Mean
2.870
4.233
8.875
6.518
6.612
8.809
1.431 S.D.
Potato
Code:11-401 Date: 6/20/83(171) Time 12:48 S.Z.Angle: 16.94
2.537
3.337
2.338 22.140 19.359 11.113
3.606 306.976 Mean
1.061
1.329
1.485
8.430
9.238
7.709
3.805
0.973 S.D.
Potato
3.567
0.632
Code:11-401 Date: 6/30/83(181) Time 12: 5 S.Z.Angle: 22.69
6.835
3.384 40.061 33.357 17.734
5.836 307.453 Mean
1.389
0.838
7.934
7.137
3.980
1.295
0.729 S.D.
Potato
3.600
0.577
Code:11-401 Date: 6/30/83(181) Time 12: 9 S.Z.Angle: 22.08
6.870
3.426 39.490 33.321 17.939
5.995 307.771 Mean
1.326
0.752
7.059
6.417
3.755
1.350
0.671 S.D.
Potato
5.172
0.572
Code:11-401 Date: 7/ 6/83(187) Time 11:28 S.Z.Angle: 29.33
8.646
6.549 40.437 37.742 23.868 12.388 304.455 Mean
0.707
1.172
3.391
1.462
1.690
2.545
0.705 S.D.
220
Potato
4.593
0.564
Code:11-401 Date: 7/20/83(201) Time 9: 4 S.Z.Angle: 55.00
7.050
6.304 29.965 31.775 20.595 11.370 305.040 Mean
0.421
0.815
2.913
2.071
1.421
1.379
0.917 S.D.
Potato
4.412
0.312
Code:11-401 Date: 7/20/83(201) Time 9:10 S.Z.Angle: 55.00
6.974 6.182 28.966 30.618 20.147 11.170 305.843 Mean
0.387
0.616
2.216
1.590
1.359
1.407
1.276 S.D.
Soybean 1 Code: 2-301 Date: 6/ 6/83(157) Time 11:32 S.Z.Angle: 27.64
3.658
4.795
5.663 11.426 18.850 18.290 12.191 304.292 Mean
0.302 0.387
0.668
1.374
1.140
1.004 0.786 12.164 S.D.
Soybean 1 Code: 2-301 Date: 6/15/83(166) Time 13:49 S.Z.Angle: 17.02
3.677
4.943
5.779 12.828 20.013 18.743 15.755 309.462 Mean
0.728
1.060
1.511
3.415
4.943
5.014
4.271
0.840 S.D.
Soybean 1 Code: 2-301 Date: 6/20/83(171) Time 10:39 S.Z.Angle: 37.33
3.415
5.169
4.998 21.599 26.693 21.397 12.435 307.205 Mean
0.753
1.413
1.188
9.635
7.759
4.640
2.166
0.265 S.D.
Soybean 1 Code: 2-301 Date: 6/30/83(181) Time 13: 0 S.Z.Angle: 16.46
2.663
4.365
3.662 30.254 29.606 17.597
9.044 308.610 Mean
0.699
1.384
1.525
7.984
7.940
5.226
3.297
0.438 S.D.
Soybean 1 Code: 2-301 Date: 6/30/83(181) Time 13: 5 S.Z.Angle: 16.20
2.586
4.208
3.619 37.755 41.238 17.382
9.084 308.820 Mean
0.792
1.458
1.699
9.013 11.028
5.679
3.753
0.330 S.D.
Soybean 1 Code: 2-301 Date: 7/ 1/83(182) Time 9:28 S.Z.Angle: 51.67
2.612
4.584
2.552 40.179 38.165 22.849
9.233 305.319 Mean
0.111
0.305
0.135
2.767
1.968
1.243
0.576
0.237 S.D.
Soybean 1 Code: 2-301 Date: 7/ 1/83(182) Time 9:40 S.Z.Angle: 49.33
2.627
4.556
2.554 38.253 36.796 22.067
8.930 306.315 Mean
0.112
0.188
0.260
1.972
1.284
0.671
0.347
0.361 S.D.
Soybean 1 Code: 2-301 Date: 7/ 1/83(182) Time 10:20 S.Z.Angle: 41.58
2.669
4.658
2.691 36.866 35.726 22.405
9.389 306.088 Mean
0.127
0.245
0.177
2.175
1.788
0.933
0.416
0.281 S.D.
Soybean 1 Code: 2-301 Date: 7/ 1/83(182) Time 10:54 S.Z.Angle: 35.11
3.035
5.254
3.317 39.574 38.767 25.056 11.625 306.117 Mean
0.156
0.191
0.248
1.951
1.200
0.637
0.560
0.312 S.D.
Soybean 1 Code: 2-301 Date: 7/ 6/83(187) Time 9:41 S.Z.Angle: 49.54
2.448
4.133
2.327 40.518 36.328 21.212
8.398 303.053 Mean
0.202
0.444 0.191
3.323
3.207
2.036
0.796
0.368 S.D.
221
Soybean 1 Code: 2-301 Date: 7/20/83(201) Time 12:16 S.Z.Angle: 23.29
2.382
3.829
2.314 45.039 36.243 24.278
9.402 312.783 Mean
0.485
0.816
0.436
8.177
6.684
5.039
2.169
0.372 S.D.
Soybean 1 Code: 2-301 Date: 7/27/83(208) Time 10:20 S.Z.Angle: 44.31
2.214
3.408
2.102 44.699 38.768 20.505
7.334 309.183 Mean
0.087
0.151
0.105
1.916
1.139
0.421
0.256
0.390 S.D.
Soybean 1 Code: 2-301 Date: 7/27/83(208) Time 10:25 S.Z.Angle: 43.37
2.190
3.418
2.070 44.407 38.593 20.411
7.256 308.886 Mean
0.050
0.109
0.120
2.159
1.687
0.824
0.335
0.262 S.D.
Soybean 1 Code: 2-301 Date: 8/25/83(237) Time 12:46 S.Z.Angle: 29.17
2.853
4.715
3.324 36.445 33.320 19.228
8.184 316.164 Mean
0.396
0.747
0.479
2.387
2.050
1.645
0.841
0.313 S.D.
Soybean 2 Code: 8-302 Date: 6/ 6/83(157) Time 13:15 S.Z.Angle: 16.38
3.990
5.082
6.127 10.377 17.140 17.385 13.480 260.799 Mean
0.305
0.496
0.779
1.278
1.588
1.766
1.869 37.413 S.D.
Soybean 2 Code: 8-302 Date: 6/17/83(168) Time 10:12 S.Z.Angle: 42.42
5.146
6.578
7.734 15.053 23.246 24.393 21.421 305.375 Mean
0.374
0.539
1.060
3.129
2.944
2.867
3.188
0.520 S.D.
Soybean 2 Code: 8-302 Date: 6/20/83(171) Time 11:56 S.Z.Angle: 23.58
4.679
6.084
7.397 13.180 26.320 27.948 14.622 307.163 Mean
0.837
1.139
1.574
3.593
7.740 10.157
6.050
1.438 S.D.
Soybean 2 Code: 8-302 Date: 6/30/83(181) Time 10:51 S.Z.Angle: 35.60
2.631
3.376
3.059 15.895 21.462 19.032 10.407 310.092 Mean
1.101
1.672
1.781
9.189
9.412
7.520
4.398
0.449 S.D.
Soybean 2 Code: 8-302 Date: 6/30/83(181) Time 10:56 S.Z.Angle: 34.66
2.479
3.063
2.774 12.762 19.446 17.987
9.883 309.386 Mean
0.992
1.548
1.708
7.369
9.091
8.125
4.963
0.633 S.D.
Soybean 2 Code: 8-302 Date: 7/ 6/83(187) Time 10:49 S.Z.Angle: 36.46
4.186
6.088
5.605 30.762 34.133 25.631 15.965 304.877 Mean
0.431
0.522
0.712
1.447
1.472
1.746
1.902
0.753 S.D.
Soybean 2 Code: 8-302 Date: 7/20/83(201) Time 11: 6 S.Z.Angle: 34.71
3.117
4.539
3.166 40.940 37.133 21.341
8.607 309.650 Mean
0.141
0.258
0.132
1.753
1.014
0.625
0.302
0.394 S.D.
Soybean 3 Code:13-303 Date: 7/27/83(208) Time 10:39 S.Z.Angle: 40.74
7.528 13.431 18.636 28.727 37.105 27.793 17.138 304.490 Mean
0.772
1.520
2.145
3.089
3.500
2.643
1.747
1.010 S.D.
222
Soybean 3 Code:13-303 Date: 7/27/83(208) Time 10:45 S.Z.Angle: 39.62
7.600 13.478 18.701 28.911 37.180 27.843 17.144 305.132 Mean
0.668
1.355
1.851
2.550
3.072
2.366
1.604
0.885 S.D.
Soybean 3 Code:13-303 Date: 8/25/83(237) Time 12:57 S.Z.Angle: 28.63
9.776 14.007 16.329 27.890 31.833 26.459 17.480 313.752 Mean
2.576
3.396
4.894
2.872
1.867
4.362
4.238
0.590 S.D.
Swamp
3.183
0.452
Code:10-401 Date: 6/ 6/83(157) Time 13:41 S.Z.Angle: 17.20
4.812
3.626 26.699 19.944 10.897
4.229 284.286 Mean
0.778
0.617
6.606
5.486
3.030
1.041 20.171 S.D.
Swamp
3.228
1.173
Code:10-701 Date: 6/15/83(166) Time 14:30 S.Z.Angle: 21.91
4.879
4.610 30.175 27.736 16.166
8.481 307.461 Mean
1.800
2.809
4.677
6.347
5.183
3.732
1.169 S.D.
Swamp
3.740
0.431
Code:10-701 Date: 6/17/83(168) Time 10:33 S.Z.Angle: 38.39
5.430
4.763 22.223 20.281 11.717
5.012 305.010 Mean
0.790
0.745
8.218
7.877
4.467
1.820
1.263 S.D.
Swamp
2.532
0.340
Code:10-701 Date: 6/20/83(171) Time 12:25 S.Z.Angle: 19.39
3.967
3.724 24.080 17.996
6.933
1.851 308.146 Mean
0.731
0.783
8.745
7.841
3.310
1.008
1.337 S.D.
Swamp
2.444
0.262
Code:10-701 Date: 6/30/83(181) Time 11:40 S.Z.Angle: 26.74
3.887
2.995 30.823 31.543 12.323
4.540 307.892 Mean
0.472
0.361
5.908
5.883
2.014
0.557
0.594 S.D.
Swamp
2.513
0.358
Code:10-701 Date: 6/30/83(181) Time 11:46 S.Z.Angle: 25.73
3.967
3.070 30.572 31.937 12.689
4.736 307.285 Mean
0.703
0.737
7.664
7.683
3.043
1.223
0.615 S.D.
Swamp
3.499
0.376
Code:10-701 Date: 7/ 6/83(187) Time 11:11 S.Z.Angle: 32.38
5.477
5.164 31.292 29.446 15.907
6.436 308.578 Mean
0.524
1.469
3.222
1.710
1.060
0.903
0.385 S.D.
Swamp
4.768
0.621
Code:10-701 Date: 7/20/83(201) Time 11:17 S.Z.Angle: 32.74
6.618
5.916 32.774 27.991 15.087
6.107 311.443 Mean
1.066
0.513 19.886 17.878
9.371
3.551
0.352 S.D.
Wheat 1
3.059
0.159
Code: 6-101 Date: 6/ 6/83(157) Time 12:52 S.Z.Angle: 17.19
5.521
4.153 29.076 25.564 14.316
6.168 290.252 Mean
0.259
0.392
2.529
1.327
0.678
0.720
3.506 S.D.
Wheat 1
2.797
1.517
Code: 6-101 Date: 6/15/81(166) Time 14:59 S.Z.Angle: 26.54
4.332
4.983 25.230 30.090 22.531
8.916 307.146 Mean
2.019
2.228 11.018 15.084 14.873
6.338
1.113 S.D.
223
Wheat 1
3.899
1.256
Code: 6-101 Date: 6/20/83(171) Time 11:41 S.Z.Angle: 26.05
6.149
6.702 21.591 24.150 15.528
7.386 307.322 Mean
2.199
1.539
5.461
4.482
2.465
1.158
1.364 S.D.
Wheat 1
3.603
0.756
Code: 6-101 Date: 6/30/83(181) Time 10:35 S.Z.Angle: 38.63
8.565 15.127 23.343 31.055 26.733 21.457 309.780 Mean
2.111
3.266
4.920
5.470
5.172
4.274
5.325 S.D.
Wheat 1
6.414
1.667
Code: 6-101 Date: 6/30/83(181) Time 10:38 S.Z.Angle: 38.06
7.673 13.056 20.060 36.324 36.790 19.152 307.327 Mean
1.858
3.311
5.009
6.940
5.770
3.463
0.481 S.D.
Wheat 1
6.791
0.976
Code: 6-101 Date: 7/ 6/83(187) Time 10:33 S.Z.Angle: 39.49
11.587 16.068 23.536 32.874 26.023 15.534 303.735 Mean
1.871
2.438
3.117
3.649
3.071
2.153
3.417 S.D.
Wheat 1
6.829
0.727
Code: 6-101 Date: 7/20/83(201) Time 10: 1 S.Z.Angle: 46.97
11.256 15.457 24.654 33.804 26.700 15.725 304.069 Mean
1.356
1.946
2.189
2.559
2.258
1.673
0.522 S.D.
Wheat 2
2.973
0.573
Code: 9-102 Date: 6/ 6/83(157) Time 13:30 S.Z.Angle: 16.63
5.179
4.187 26.782 24.150 13.698
6.097 292.230 Mean
0.897
1.180
3.165
1.561
2.948
2.542
3.796 S.D.
Wheat 2
3.030
0.514
Code: 9-102 Date: 6/17/83(168) Time 10:22 S.Z.Angle: 40.49
4.790
6.641 18.342 23.277 15.551
7.678 301.452 Mean
0.891
1.321
3.783
4.378
3.129
1.431
0.350 S.D.
Wheat 2
2.087
0.518
Code: 9-102 Date: 6/20/83(171) Time 12:16 S.Z.Angle: 20.58
2.715
6.206 19.045 23.729 17.718
6.026 308.358 Mean
0.653
1.811
5.967
8.659
6.851
4.351
1.651 S.D.
Wheat 2
3.654
1.503
Code: 9-102 Date: 6/30/83(181) Time 11:10 S.Z.Angle: 32.07
3.515
6.856 11.900 28.677 18.232
6.703 309.766 Mean
1.542
3.180
4.969 10.946
8.323
3.550
0.062 S.D.
Wheat 2
4.454
1.186
Code: 9-102 Date: 6/30/83(181) Time 11:29 S.Z.Angle: 28.65
4.671
9.288 15.222 36.923 22.288 10.003 310.003 Mean
1.478
3.043
4.515 10.469
7.088
3.621
0.040 S.D.
Wheat 2
6.916
1.174
Code: 9-102 Date: 7/ 6/83(187) Time 11: 3 S.Z.Angle: 33.85
11.094 14.722 20.979 29.912 26.517 16.531 308.128 Mean
2.155
2.513
3.073
3.711
3.230
2.037
3.297 S.D.
Wheat 3
3.870
3.141
Code: 4-103 Date: 6/ 6/83(157) Time 12:18 S.Z.Angle: 20.59
5.434
3.368 30.826 27.342 14.477
6.047 293.685 Mean
1.269
0.421
2.129
2.243
1.602
0.996
7.209 S.D.
224
Wheat 3
1.956
0.550
Code: 4-103 Date: 6/20/83(171) Time 11:13 S.Z.Angle: 30.99
3.194
2.013 28.158 21.549 10.480
3.584 305.789 Mean
1.020
0.722 10.412
8.128
4.091
1.270
0.450 S.D.
Wheat 3
3.382
0.426
Code: 4-103 Date: 6/30/83(181) Time 9:38 S.Z.Angle: 49.65
4.771
5.471 24.246 29.328 20.986 10.998 304.212 Mean
0.731
0.904
3.920
2.890
2.273
1.733
0.353 S.D.
Wheat 3
3.160
0.515
Code: 4-103 Date: 6/30/83(181) Time 9:42 S.Z.Angle: 48.87
4.459
5.127 24.172 30.518 22.157 11.598 304.102 Mean
0.881
0.847
3.052
2.781
2.527
2.024 0.317 S.D.
Wheat 3
3.461
0.140
Code: 4-103 Date: 7/ 6/83(187) Time 10: 7 S.Z.Angle: 44.49
5.460 6.691 21.128 26.603 19.640 10.392 303.989 Mean
0.262
0.556
0.951
1.071
0.933
0.650
0.461 S.D.
Wheat 3
6.229
0.224
Code: 4-103 Date: 7/20/83(201) Time 12:35 S.Z.Angle: 20.97
9.678 14.193 22.790 32.186 27.651 16.477 306.989 Mean
0.464
1.390
0.997
1.307
0.649
0.674
0.726 S.D.
Wheat 3
5.132
0.277
Code: 4-103 Date: 7/27/83(208) Time 11: 1 S.Z.Angle: 36.70
8.454 12.232 21.634 29.701 24.331 13.967 308.362 Mean
0.586
1.699
1.211
2.057
1.915
0.987
0.682 S.D.
Wheat 3
5.213
0.318
Code: 4-103 Date: 7/27/83(208) Time 11: i S.Z.Angle: 35.62
8.554 12.362 22.027 30.338 24.782 14.449 307.495 Mean
0.714
2.036
1.477
2.432
2.152
1.148
0.591 S.D.
225
APPENDIX D
RADAR MEASUREMENTS (1983)
1. Radar
backscattering
power
returns
have
been
calibrated
into
radar
coefficients and are listed in the last four columns
of this tabulation in the order of CHH50, CHV50, XW50, and XVH50.
2. Soybean 1 and 2 are regular soybeans.
double-cropped
wheat.
after
the
wheat harvest.
Wheat 3 is spring wheat.
226
Soybean 3 is soybeans
Wheat 1 and 2 are winter
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
alfalfa
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
bare soil
14 801 08/03/83 (215) 09:32 -18.23 - 2 1 . 3 6
14 801 08/03/83 (215) 09:37 -17.10 -22.14
14 801 08/03/83 (215) 09:40 -16.52 -21.57
14 801 08/03/83 (215) 09:44 -17.45 - 2 2 . 7 0
14 801 08/03/83 (215) 09:47 -16.39 -21.92
14 801 08/03/83 (215) 09:50 -16.54 -22.49
14 801 08/03/83 (215) 09:54 "•16.44 -21.94
14 801 08/03/83 (215) 09:56 -16.55 -22.90
14 801 08/03/83 (215) 10:00 -16.43 -22.50
14 801 08/03/83 (215) 10:02 -16.46 -21.20
14 801 08/03/83 (215) 10:04 -16.55 -20.83
14 801 08/03/83 (215) 10:06 -16.48 -21.44
14 801 08/03/83 (215) 10:08 -16.65 -21.07
14 801 08/03/83 (215) 10:10 -17.29 -20.72
14 801 08/03/83 (215) 10:12 -17.90 -21.98
14 801 08/03/83 (215) 10:16 -15.63 - 2 2 . 2 2
14 801 08/03/83 (215) 10:20 -15.51 -22.45
14 801 08/03/83 (215) 10:21 -15.85 - 2 1 . 8 8
14 801 08/03/83 (215) 1 0 : 2 2 -15.55 -23.01
14 801 08/03/83 (215) 10:24 -16.36 -23.91
-12.23
-12.70
-12.39
-12.60
-12.30
-12.42
-18.83
-19.40
-19.19
-19.30
-18.70
-20.12
-12.08 -18.88
-12.14
-12.71
-13.00
-11.75
-19.14
-17.81
-19.90
-19.85
-12.18 -19.38
- 1 1 . 8 2 -19.02
-13.32 -20.02
-11.13 -19.83
-11.17 -18.97
-10.75 -20.05
-12.22 -20.02
-11.52 -19.82
-11.61 -19.81
15 901 07/27/83 (208) 13:00 -22.41 97.50 -13.27 -23.07
15 901 07/27/83 (208) 13:01 -22.24 -31.04 -11.50 -25.20
15 901 07/27/83 (208) 13:03 -19.98 -30.50 - 1 2 . 8 6 -26.56
15 901 07/27/83 (208) 13:04 -20.95 -31.34 -11.08 -25.58
15 901 07/27/83 (208) 13:06 -21.14 -33.26 -11.19 -25.69
15 901 07/27/83 (208) 13:09 -21.53 -30.24 -11.62 - 2 6 . 2 2
15 901 07/27/83 (208) 13:13 -21.38 -30.96 -11.85 -26.25
15 901 07/27/83 (208) 13:14 -21.54 -31.13 -12.84 -27.44
15 901 07/27/83 (208) 13:15 -22.20 -30.25 -11.70 -24.50
15 901 07/27/83 (208) 13:18 -19.83 -30.15 -10.46 -24.46
15 901 07/27/83 (208) 13:22 -17.60 -27.79 -10.34 -22.64
15 901 07/27/83 (208) 13:24 -19.35 -28.14 -10.91 -23.51
15 901 07/27/83 (208) 13:26 -18.58 -27.73 -11.30 -23.60
15 901 07/27/83 (208) 13:29 -18.97 97.50 - 1 1 . 0 0 -23.10
15 901 07/27/83 (208) 13:31 -20.04 -29.23 - 1 1 . 0 8 -23.68
15 901 07/27/83 (208) 13:34 -20.09 -28.20 - 1 1 . 0 0 -23.80
15 901 07/27/83 (208) 13:36 -19.28 -29.17 -10.87 -24.67
15 901 07/27/83 (208) 13:39 -17.59 -27.71 -10.50 -23.30
15 901 07/27/83 (208) 13:41 -18.09 -27.89 -11.23 -23.63
15 901 07/27/83 (208) 13:43 -18.64 -27.35 -10.70 -23.60
15 901 07/27/83 (208) 13:45 -24.16 -29.86 -13.34 -24.14
15 901 07/27/83 (208) 13:47 -22.90 -29.93 -12.92 -23.62
15 901 07/27/83 (208) 13:50 -23.02 -30.34 -12.64 -23.74
15 901 07/27/83 (208) 13:52 -24.56 -30.83 -12.93 -24.03
15 901 07/27/83 (208) 13:54 - 2 2 . 6 1 -30.91 -11.53 -23.63
15 901 07/27/83 (208) 13:56 -22.91 -31.04 -12.89 -24.49
15 901 07/27/83 (208) 13:58 -22.96 -30.84 -12.23 -24.13
15 901 07/27/83 (208) 14:00 - 2 2 . 8 0 -30.19 -12.38 -23.98
15 901 07/27/83 (208) 14:02 -22.53 -29.93 -12.10 -23.00
22?
bare soil
15 901 07/27/83 (208) 14:05 -22.94 -29.67 -13.08 -24.28
corn 1
01 201 06/06/83 (157) 11:05 -13.93 -23.64
-5.83 -15.63
corn 1
01 201 06/15/83 (166) 12:05 -18.03 -22.68
-9.09 -18.89
corn 1
01 201 06/20/83 (171) 14:10 -10.11 -18.29
-6.62 -15.62
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
01
01
01
01
01
01
01
01
01
01
1
1
1
1
1
1
1
1
1
1
201
201
201
201
201
201
201
201
201
201
07/19/83 (200) 09:02
07/19/83 (200) 09:04
07/19/83 (200) 09:06
07/19/83 (200) 09:08
07/19/83 (200) 09:10
07/19/83 (200) 09:12
07/19/83 (200) 09:14
07/19/83 (200) 09:16
07/19/83 (200) 09:18
07/19/83 (200) 09:20
-13.16
-12.94
-13.70
-15.68
-15.50
-15.05
-15.24
-15.11
-16.16
-15.28
-20.71
-21.58
-21.61
-21.48
-21.92
-21.89
-22.02
-22.00
-22.11
-22.31
-10.28
-10.27
-8.65
-7.81
-7 14
-7.38
-9.04
-8 67
-9 23
-8.10
-17.68
-17.87
-19.35
-18.21
-17.14
-16.98
-18.84
-19.37
-20.03
-18.50
corn 2
05 202 06/06/83 (157) 11:45 -12.76 -23.49
-8 01 -18.21
corn 2
05 202 06/15/83 (166) 10:50 -12.41 -20.34
-8 48 -18.48
corn 2
05 202 06/20/83 (171) 14:40
-8 57 -17.77
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
2
2
2
2
2
2
2
2
2
2
05
05
05
05
05
05
05
05
05
05
202
202
202
202
202
202
202
202
202
202
07/08/83 (189) 14:50
07/08/83 (189) 14:52
07/08/83 (189) 14:54
07/08/83 (189) 14:56
07/08/83 (189) 15:00
07/08/83 (189) 15:03
07/08/83 (189) 15:04
07/08/83 (189) 15:06
07/08/83 (189) 15:08
07/08/83 (189) 15:10
-14.55
-17.07
-15.76
-15.50
-15.27
-15.39
-14.88
-14.61
-15.42
-15.30
-21.28
-21.69
-20.71
-20.37
-20.29
-21.84
-21.72
-22.87
-20.81
-21.76
-8
-9
-8
-8
-7
-7
-7
-7
-7
-7
37
04
55
58
55
72
87
71
93
70
-18.17
-17.94
-17.85
-17.48
-17.75
-17.62
-17.67
-17.11
-17.73
-17.50
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
2
2
2
2
2
2
2
2
2
2
2
2
2
2
05
05
05
05
05
05
05
05
05
05
05
05
05
05
202
202
202
202
202
202
202
202
202
202
202
202
202
202
08/02/83 (214) 11:00
08/02/83 (214) 11:02
08/02/83 (214) 11:04
08/02/83 (214) 11:07
08/02/83 (214) 11:10
08/02/83 (214) 11:13
08/02/83 (214) 11:15
08/02/83 (214) 11:18
08/02/83 (214) 11:21
08/02/83 (214) 11:24
08/02/83 (214) 11:28
08/02/83 (214) 11:30
08/02/83 (214) 11:32
08/02/83 (214) 11:35
-15.09
-15.13
-14.74
-14.91
-13.80
-16.06
-14.95
-15.08
-15.86
-15.46
-15.37
-14.46
-17.29
-14.58
-22.58
-22.78
-23.15
-22.21
-21.57
-22.48
-22.13
-21.81
-22.93
-22.50
-24.12
-24.47
-22.52
-22.99
-8.06
-7.60
-7 83
-8.03
-7 50
-8.53
-8 76
-9.13
-9.10
-7.63
-7.87
-8.26
-8.10
-8.60
-16.56
-16.90
-17.13
-17.83
-17.60
-16.83
-17.56
-18.03
-17.30
-17.53
-17.87
-17.66
-18.10
-17.80
228
-9.87 -17.46
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
corn
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
05
05
05
05
05
05
05
05
05
05
05
05
05
05
05
05
202
202
202
202
202
202
202
202
202
202
202
202
202
202
202
202
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
11:38
11:40
11:42
11:45
11:48
11:50
11:54
11:56
11:58
12:00
12:03
12:07
12:09
12:11
12:13
12:15
-15.89
-14.77
-15.00
-14.80
-15.56
-16.12
-14.48
-13.68
-14.37
-13.65
-14.58
-14.80
-13.99
-14.27
-15.21
-14.19
-23.14
-22.08
-22.60
-22.98
-22.45
-23.12
-21.04
-20.80
-21.22
-23.84
-22.92
-21.01
-21.31
-21.72
-22.54
-21.37
-8.39
-8.69
-8.36
-9.09
-8.33
-7.83
-8.16
-8.15
-7.99
-8.52
-8.32
-7.65
-8.45
-7.42
-7.16
-8.69
-18.09
-18.09
-18.06
-18.19
-17.73
-17.73
-17.76
-18.35
-17.29
-18.32
-17.62
-17.95
-18.25
-18.02
-18.06
-17.89
fallow
12 601 06/06/83 (157
15:00 -14.48 -20.14 -9.28 -17.88
fallow
12 601 06/15/83 (166
12:25 -14.85 -21.19
-8.89 -17.09
fallow
12 601 06/20/83 (171
15:50 -12.72 -21.81
-9.72 -19.52
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
601
601
601
601
601
601
601
601
601
601
601
601
601
601
601
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
07/14/83 (195
13:30
13:32
13:34
13:36
13:38
13:40
13:42
13:44
13:46
13:48
13:50
13:53
13:55
13:58
14:00
-10.33
-10.79
-10.35
-10.45
-10.03
-13.89
-13.08
-13.50
-13.87
-13.14
-12.25
-12.80
-12.50
-12.58
-13.94
-17.00
-17.10
-17.37
-16.95
-16.60
-15.07
-15.57
-15.17
-15.65
-15.37
-17.08
-18.02
-16.93
-16.88
-15.98
-7.79
-8.24
-7.49
-7.82
-7.12
-9.02
-9.11
-8.48
-9.28
-9.32
-8.88
-7.18
-7.92
-7.51
-8.46
-16.39
-16.94
-16.09
-17.52
-16.72
-16.52
-16.11
-16.58
-16.58
-16.72
-15.78
-15.78
-16.52
-15.71
-17.36
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
fallow
12
12
12
12
12
12
12
12
12
12
12
601
601
601
601
601
601
601
601
601
601
601
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
10:45
10:48
10:51
10:54
10:57
11:00
11:05
11:08
11:10
11:13
11:15
-14.41
-13.10
-12.43
-12.61
-13.24
-12.04
-12.44
-11.97
-12.61
-12.50
-10.75
-17.61
-16.60
-16.32
-16.25
-16.05
-16.15
-16.49
-16.62
-16.56
-16.47
-15.75
-9.55
-10.03
-8.90
-8.71
-8.98
-7.20
-6.85
-7.33
-7.17
-7.08
-7.89
-15.85
-15.03
-16.60
-15.11
-15.38
-16.80
-16.25
-16.93
-16.97
-16.48
-15.99
229
601
601
601
601
07/19/83 (200
07/19/83 (200
07/19/83 (200
07/19/83 (200
11:17
11:20
11:23
11:25
i
-15.70 -8.05
-16.08 -7.69
-16.88 -7.59
-16.74 -7.32
-15.55
-15.79
-15.89
-15.72
12
12
12
12
grass
07 502 06/06/83 (157
grass
07 502 06/15/83 (166] 11:02 -17.06 -21.68 -11.18 -19.58
grass
07 502 06/20/83 (171
14:50 -14.28 -22.18
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
hay
hay
hay
hay
hay
hay
hay
hay
hay
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
07 502 07/08/83 (189
03 501 07/08/83 (189
03 501 07/08/83 (189
03 501 07/08/83 (189
03 501 07/08/83 (189
03 501 07/08/83 (189
03 501 07/08/83 (189
03 501 07/08/83 (189
03 501 07/08/83 (189
03 501 07/08/83 (189
17:09. -16.33
17:11 -15.18
17:12 -17.59
17:14 -17.76
17:16 -17.72
17:17 -15.83
17:19 -16.14
17:21 -15.80
17:22 -15.69
17:24 -15.13
17:26 -16.00
17:28 -15.48
17:30 -17.24
17:32 -18.12
17:34 -17.37
17:36 -16.15
17:38 -15.82
17:40 -16.12
17:42 -15.38
17:44 -15.43
12:30 -13.59
12:32 -11.25
12:34 -16.73
12:36 -14.37
12:38 -15.39
12:40 -18.06
12:42 -18.71
12:43 -17.74
12:45 -17.70
-21.92
-22.57
-22.69
-22.24
-23.23
-21.89
-21.85
-22.47
-23.63
-21.79
-21.48
-23.02
-22.24
-22.68
-22.77
-22.33
-21.41
-22.92
-23.16
-22.23
-21.86
-23.07
-21.81
-21.70
-23.21
-23.83
-23.70
-25.24
-23.65
-8.72 -16.72
-9.02 -17.32
-8.25 -16.55
-8.86 -16.96
-8.16 -16.46
-8.91 -18.01
-8.19 -17.89
-8.52 -16.32
-8 72 -17.72
-7 95 -17.15
-8 55 -16.39
-9 20 -17.67
-8 08 -16.22
-9 04 -17.30
-8 00 -16.13
-9 09 -18.37
-8 03 -17.53
-8 69 -16.65
-8 55 -17.37
-8 11 -17.49
-11 63 -21.03
-11.44 -20.64
-13 25 -20.15
-11 52 -20.02
-11 93 -21.33
-13.62 -20.92
-14 50 -21.30
-11 23 -20.63
-13.27 -20.17
grass
grass
grass
grass
grass
grass
grass
grass
grass
07
07
07
07
07
07
07
07
07
10:15
10:17
10:19
10:21
10:23
10:25
10:28
10:30
10:32
-20.22
-20.20
-20.39
-20.60
-20.47
-22.24
-21.81
-22.24
-22.92
-9.20
-9.15
-8.65
-10.07
-8.43
-8.42
-8.73
-7.99
-7.62
502
502
502
502
502
502
502
502
502
07/27/83 (208
07/27/83 (208
07/27/83 (208
07/27/83 (208
07/27/83 (208
07/27/83 (208
07/27/83 (208
07/27/83 (208
07/27/83 (208
12:20 -16.29 -22.81 -11.85 -18.05
230
I
-10.09
-10.35
-10.35
-10.41
fallow
fallow
fallow
fallow
-17.83
-17.19
-16.92
-16.54
-17.00
-18.86
-18.70
-15.73
-15.62
-7.97 -16.67
-17.50
-17.15
-17.15
-17.07
-17.13
-17.82
-17.83
-16.59
-15.82
502
502
502
502
502
502
502
502
502
502
502
07/27/83 (208) 10:35
07/27/83 (208) 10:37
07/27/83 (208) 10:39
07/27/83 (208) 10:41
07/27/83 (208) 10:44
07/27/83 (208) 10:46
07/27/83 (208) 10:50
07/27/83 (208) 10:52
07/27/83 (208) 10:54
07/27/83 (208) 10:56
07/27/83 (208) 10:58
-14.64
-18.94
-19.21
-18.40
-18.25
-20.51
-18.29
-19.20
-22.69
-18.50
-19.10
-22.11 -7.35 -16.65
-22.11 -9.29 -17.59
-22.59 -9.05 -17.75
-22.88 -8.35 -18.35
-22.37 -9.39 -17.09
-23.55 -8.24 -16.54
-21.54 -9.29 -17.99
-21.29 -9.10 -17.10
-22.72 -10.25 -18.15
-21.62 -9.33 -17.63
-22.85 -9.74 -19.24
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
grass
07
07
07
07
07
07
07
07
07
07
07
hay
03 501 06/06/83 (157) 11:30 -13.92 -19.39 -13.83 -19.43
hay
03 501 06/15/83 (166) 11:40 -14.01 -22.51 -12.72 -19.42
hay
03 501 06/20/83 (171) 14:20 -13.60 -21.96 -11.86 -17.66
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
hay
03
03
03
03
03
03
03
03
03
03
03
03
03
03
03
03
03
03
03
03
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
07/08/83 (189) 12:30 -13.59
07/08/83 (189) 12:32 -11.25
07/08/83 (189) 12:34 -16.73
07/08/83 (189) 12:36 -14.37
07/08/83 (189) 12:38 -15.39
07/08/83 (189) 12:40 -18.06
07/08/83 (189) 12:42 -18.71
07/08/83 (189) 12:43 -17.74
07/08/83 (189) 12:45 -17.70
07/08/83 (189) 12:47 -16.30
07/08/83 (189) 12:49 -13.32
07/08/83 (189) 12:51 -11.47
07/08/83 (189) 12:53 -16.40
07/08/83 (189) 12:56 -14.66
07/08/83 (189) 13:02 -15.08
07/08/83 (189) 13:05 -18.42
07/08/83 (189) 13:07 -18.34
07/08/83 (189) 13:10 -18.09
07/08/83 (189) 13:13 -17.35
07/08/83 (189) 13:15 -16.63
-21.86 -11.63 -21.03
-23.07 -11.44 -20.64
-21.81 -13.25 -20.15
-21.70 -11.52 -20.02
-23.21 -11.93 -21.33
-23.83 -13.62 -20.92
-23.70 -14.50 -21.30
-25.24 -11.23 -20.63
-23.65 -13.27 -20.17
-22.77 -14.53 -20.73
-21.42 -11.40 -20.61
-23.53 -11.67 -21.05
-21.37 -12.99 -19.75
-22.13 -11.75 -20.42
-22.75 -11.69 -20.90
-24.31 -13.89 -21.34
-23.23 -14.21 -20.87
-25.74 -11.45 -21.04
-23.18 -13.00 -19.77
-23.23 -14.82 -21.14
milo
milo
milo
milo
milo
milo
milo
milo
milo
milo
milo
16
16
16
16
16
16
16
16
16
16
16
909
909
909
909
909
909
909
909
909
909
909
08/03/83 (215) 12:20
08/03/83 (215) 12:22
08/03/83 (215) 12:25
08/03/83 (215) 12:30
08/03/83 (215) 12:31
08/03/83 (215) 12:33
08/03/83 (215) 12:35
08/03/83 (215) 12:36
08/03/83 (215) 12:37
08/03/83 (215) 12:40
08/03/83 (215) 12:52
-19.03
-20.85
-19.36
-19.68
-19.58
-19.59
-20.84
-19.19
-19.37
-20.56
-19.99
231
-14.67
-13.40
-13.21
-14.32
-14.06
-15.87
-13.58
-12.38
-15.29
-13.91
-14.78
-11.13 -21.63
-11.22 -21.52
-8.73-18.73
-11.03 -19.43
-10.42 -20.12
-9.94 -22.04
-9.20 -19.00
-11.00 -17.50
-7.58 -18.18
-9.23 -17.53
-9.74 -21.60
909
909
909
909
909
909
909
909
909
08
08
08
08
08
08
08
08
08
03/83 215
03/83 215
03/83 215
03/83 215
03/83 215
03/83 215
03/83 215
03/83 215
03/83 215
12:55
12:57
13:00
13:03
12:42
12:44
12:46
12:48
12:50
-13.83
-12.89
-13.86
-12.70
-15.08
-13.56
-13.15
-13.59
-12.96
-20.39
-19.92
-21.10
-20.32
-20.40
-19.99
-20.33
-20.69
-20.73
-9.38
-10.78
-7.73
-9.05
-9.94
-9.20
-11.00
-7.58
-9.23
-19.38
-17.15
-18.54
-17.18
-22.04
-19.00
-17.50
-18.18
-17.53
milo
milo
milo
milo
milo
milo
milo
milo
milo
16
16
16
16
16
16
16
16
16
potatoes
potatoes
J1 401 06 06/83
11 401 06 06/83
157
157
14:10 -13.69 -20.21
14:20 -12.43 -18.35
-8.89 -17.79
-7.32 -15.92
potatoes
11 401 06 15/83
166
12:50 -12.51 -17.27
-7.69 -16.89
potatoes
11 401 06 20/83
171
15:30 -10.04 -16.03
-6.57 -21.87
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
14/83
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
195
11:55 -10.47 -13.91
12:00 -10.29 -15.14
12:04 -10.25 -14.83
12:07 -10.30 -14.77
12:08 -10.70 -14.95
12:10 -10.62 -14.94
12:12 -11.16 -15.71
12:14 -9.66 -14.53
12:17 -9.56 -16.18
12:20 -9.28 -16.97
12:24 -11.09 -16.82
12:26 -10.72 -15.20
12:30 -10.75 -15.24
12:34 -11.59 -15.37
12:37 -10.76 -15.29
12:40 -12.60 -16.70
12:43 -12.36 -17.15
12:46 -11.51 -17.62
12:48 -12.32 -17.12
12:51 -11.76 -18.90
12:55 -12.81 -16.83
13:00 -10.67 -14.76
13:02 -10.43 -15.16
13:02 -10.60 -14.96
13:06 -10.18 -15.19
13:08 -10.23 -14.88
13:10 -10.52 -14.99
13:13 -11.20 -14.23
13:15 -11.27 -14.11
13:17 -11.02 -15.15
-10.24 -18.04
-8.67 -17.87
-7.92 -16.62
-7.06 -16.36
-6.79 -15.59
-7.61 -16.81
-7.73 -16.53
-7.85 -16.15
-7.49 -16.59
-7.63 -17.13
-8.07 -16.97
-7.56 -16.96
-8.09 -16.59
-7.41 -17.51
-8.82 -16.92
-7.52 -15.72
-8.15 -17.25
-6.67 -16.17
-6.69 -15.79
-7.62 -15.32
-6.89 -16.09
-6.89 -16.39
-7.04 -16.54
-6.23 -16.03
-7.04 -16.84
-8.66 -16.66
-7.94 -15.84
-7.93 -16.93
-7.60 -15.70
-7.03 -15.23
potatoes
potatoes
11 401 07 19/83
11 401 07 19/83
200
200
11:40 -10.39 -14.37
11:42 -9.77 -14.85
-7.15 -14.95
-7.52 -16.92
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
401
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
232
401
401
401
401
401
401
401
401
401
401
401
401
401
07/19/83 (200 11:45 -10.25
07/19/83 (200 11:48 -10.82
07/19/83 (200 11:50 -10.16
07/19/83 (200 11:55 -10.51
07/19/83 (200 11:57 -10.75
07/19/83 (200 12:00 -8.68
07/19/83 (200 12:02 -10.39
07/19/83 (200 12:05 -11.17
07/19/83 (200 12:08 -9.90
07/19/83 (200 12:10 -10.28
07/19/83 (200 12:15 -10.11
07/19/83 (200 12:18 -10.12
07/19/83 (200 12:20 -10.37
-14.67
-14.85
-14.89
-16.70
-16.29
-16.66
-18.17
-17.34
-18.20
-18.24
-16.93
-16.79
-16.37
-8.59
-8.04
-7.27
-8.94
-9.05
-7.03
-8.90
-8.23
-7.65
-8.63
-7.59
-7.65
-8.23
-16.79
-16.74
-15.97
-17.34
-16.65
-17.23
-18.00
-16.83
-16.95
-17.43
-16.09
-16.95
-16.83
potatoes
potacoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
potatoes
11
11
11
11
11
11
11
11
11
11
11
11
11
soybean 1
02 301 06/06/83 (157
11:14 -14 10 -21.23
-6.72 -18.22
soybean 1
02 301 06/15/83 (166
12:00 -11 86 -18.99
-9.89 -19.99
soybean 1
02 301 06/20/83 (171
14:05 -11 26 -18.26
-7.81 -17.71
soybean
soybean
soybean
soybean
soybean
soybean
1
1
1
1
1
1
02
02
02
02
02
02
301
301
301
301
301
301
06/30/83 (181
06/30/83 (181
06/30/83 (181
06/30/83 (181
06/30/83 (181
06/30/83 (181
14:30
14:32
14:34
14:36
14:38
14:40
-9
-9
-10
-10
-10
-11
97
99
04
13
20
43
-17.31
-16.36
-17.54
-17.61
-17.27
-16.97
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
11:15
11:20
11:25
11:28
11:30
11:32
11:35
11:38
11:42
11:46
11:48
11:54
11:56
12:00
12:04
12:07
12:10
12:14
12:16
12:20
12:25
12:30
12:33
-12 81
-12 78
-13 18
-13 11
-13.68
-14.23
-13.65
-13.08
-12.77
-11.94
-13.90
-12.26
-13.02
-12.12
-13.04
-11.89
-12.19
-13.17
-14.47
-13.83
-14.14
-14.05
-12.81
-18.03
-18.10
-18.62
-18.52
-19.36
-18.44
-19.07
-19.08
-19.48
-18.79
-19.55
-19.73
-19.02
-19.44
-18.79
-18.13
-18.32
-18.70
-18.71
-18.18
-18.77
-18.90
-19.27
233
-8
-7
-9
-8
-8
-8
46
58
40
22
68
19
-19.16
-18.88
-19.40
-20.12
-19.48
-18.49
-10.50
-8.43
-7 98
-8 25
-9 19
-9 62
-7 96
-8 70
-8 18
-8 14
-8.67
-7 81
-8 44
-8 05
-8.46
-7 58
-9.40
-8.22
-8.68
-8.19
-8.51
-9.15
-9.87
-20.80
-18.73
-19.08
-19.05
-19.59
-19.82
-18.96
-19.00
-18.88
-18.94
-19.47
-18.91
-18.84
-18.85
-19.16
-18.88
-19.40
-20.12
-19.48
-18.49
-19.21
-20.45
-20.57
soybean
soybean
soybean
soybean
soybean
soybean
soybean
1
1
1
1
1
1
1
02
02
02
02
02
02
02
301
301
301
301
301
301
301
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
07/05/83 (186
12 35
12 37
12•39
12 43
12•46
12:49
12 52
-12.86
-12.65
-12.70
-12.5?
-13.44
-13.93
-12.07
-17.94
-18.71
-18.52
-18.79
-17.76
-19.30
-18.76
-8.37
-8.15
-9.16
-8.42
-8.27
-7.57
-7.93
-18.87
-18.95
-18.66
-19.42
-18.37
-17.77
-17.93
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
07/08/83 (189
10
10
10
10
10
11
11
11
11
11
11
11
11
11
11
37
40
42
56
58
00
02
04
06
08
10
22
24
26
27
-15.61
-14.24
-13.28
-13.07
-13.07
-12.75
-12.86
-13.58
-12.44
-13.70
-13.36
-13.20
-11.94
-12.76
-12.29
-17.48
-17.29
-17.89
-17.09
-18.67
-18.56
-19.05
-18.47
-18.12
-16.95
-17.90
-17.08
-17.68
-17.04
-17.38
-7.21
-6.59
-9.50
-9.93
-8.83
-8.72
-8.65
-7.60
-7.96
-7.33
-7.97
-8 87
-8.34
-8 86
-7.21
-17.01
-17.89
-18.90
-20.93
-18.43
-17.72
-19.75
-16.80
-18.36
-17.63
-16.67
-17.07
-17.54
-19.16
-17.01
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
02
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
301
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
07/29/83 (210
09:30
09.35
09:38
09.41
09:43
09:45
09.47
09.50
09:52
09-55
09:59
10:02
10:04
10:06
10:08
10:11
10:13
10:16
10:18
10:20
-14.63
-11.80
-11.13
-11.48
-12.50
-12.68
-13.16
-12.91
-14.24
-13.26
-13.00
-11.39
-12.56
-13.16
-14.52
-11.40
-13.85
-14.13
-13.00
-13.32
-18.75
-19.04
-18.86
-18.87
-19.54
-19.05
-19.28
-19.02
-19.14
-19.18
-21.30
-19.50
-20.07
-20.97
-18.19
-18.81
-19.32
-19.04
-18.97
-19.40
-8.23
-6 55
-6 33
-7 38
-6 33
-5 80
-5 67
-6 13
-6 72
-6 60
-7 53
-7 55
-6.63
-7 16
-7 53
-6.68
-6 19
-5 91
-6 33
-6 23
-17.13
-17.55
-17.43
-18.58
-17.93
-17.60
-17.27
-17.63
-17.62
-17.90
-17.03
-17.35
-17.63
-18.36
-17.43
-17.88
-18.09
-18.71
-18.63
-18.33
soybean
soybean
soybean
soybean
soybean
1
1
1
1
1
02
02
02
02
02
301
301
301
301
301
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
08/02/83 (214
09:25
09:28
09:30
09:33
09:35
-14.16
-14.01
-15.03
-14.63
-14.54
-17.06
-16.57
-16.26
-16.59
-15.91
-6
-6
-5
-6
-5
-18.27
-16.18
-16.15
-15.85
-15.45
23^+
77
08
95
95
65
soybean
soybean
soybean
soybean
soybean
1
1
1
1
1
02
02
02
02
02
301
301
301
301
301
08
08
08
08
08
02/83 (214) 09:38
02/83 (214) 09:41
02/83 (214) 09:44
02/S3 (214) 09:46
02/R5 (214) 09:49
-13.97
-13.58
-13.14
-13.37
-13.71
-17.33 -5.65
-18.21 -5.07
-18.77 -5.54
-18.59 -6.92
-18.49 -7.42
-16.55
-16.57
-17.14
-16.62
-15.72
soybean 2
08 302 06 06/83 (157) 12:45 -13.65 -22.25
soybean 2
08 302 06 15/83 (166) 11:10 -11.76 -21.39 -10.00 -20.20
soybean 2
08 302 06 20/83 (171) 15:00 -13.37 -21.77
soybean
soybean
soybean
soybean
soybean
soybean
soybean
2
2
2
2
2
2
2
08
08
08
08
08
08
08
302
302
302
302
302
302
302
07
07
07
07
07
07
07
05/83 (186) 13:41
05/83 (186) 13:44
05/83 (186) 13:47
05/83 (186) 13:52
05/83 (186) 13:56
05/83 (186) 13:58
05/83 (186) 14:00
-13.12
-13.44
-13.35
-14.34
-15.17
-15.30
-12.74
-20.96 -9.18 -18.48
-20.05 -8.84 -19.54
-19.82 -7.66 -18.36
-19.56 -8.54 -18.34
-20.30 -8.51 -19.11
-19.66 -8.03 -18.23
-19.06 -8.20 -17.00
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
08
08
08
08
08
08
08
08
08
08
08
08
08
08
08
302
302
302
302
302
302
302
302
302
302
302
302
302
302
302
07
07
07
07
07
07
07
07
07
07
07
07
07
07
07
14/83 (195) 09:15
14/83 (195) 09:23
14/83 (195) 09:25
14/83 (195) 09:27
14/83 (195) 09:30
14/83 (195) 09:36
14/83 (195) 09:42
14/83 (195) 09:44
14/83 (195) 09:47
14/83 (195) 09:50
14/83 (195) 09:53
14/83 (195) 09:57
14/83 (195) 10:01
14/83 (195) 10:03
14/83 (195) 10:05
-13.56
-13.23
-11.70
-11.70
-12.87
-13.70
-12.81
-12.17
-10.59
-12.34
-12.95
-12.75
-13.18
-11.34
-10.98
-18.24
-18.54
-19.02
-19.46
-17.49
-17.64
-17.77
-18.01
-17.55
-17.01
-18.07
-19.08
-19.15
-18.47
-19.05
-7
-6
-8
-8
-8
-7
-7
-6
-6
-7
-7
-7
-7
-7
-7
30
69
22
93
23
98
02
73
86
00
43
33
08
67
77
-19.20
-17.49
-19.32
-19.13
-17.83
-19.38
-16.32
-17.93
-18.46
-17.80
-17.13
-18.03
-17.88
-18.37
-17.87
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
2
2
2
2
2
2
2
2
2
2
2
2
2
2
08
08
08
08
08
08
08
08
08
08
08
08
08
08
302
302
302
302
302
302
302
302
302
302
302
302
302
302
07
07
07
07
07
07
07
07
07
07
07
07
07
07
19/83 (200) 14:05
19/83 (200) 14:10
19/83 (200) 14:14
19/83 (200) 14:16
19/83 (200) 14:20
19/83 (200) 14:24
19/83 (200) 14:25
19/83 (200) 14:28
19/83 (200) 14:30
19/83 (200) 14:33
19/83 (200) 14:37
19/83 (200) 14:39
19/83 (2.00) 14:41
19/83 (200) 14:43
-14.31
-12.90
-14.56
-13.66
-11.89
-12.71
-13.40
-12.27
-12.04
-12.15
-14.52
-12.80
-13.02
-12.91
-18.45
-19.73
-19.12
-19.13
-17.26
-17.53
-16.80
-18.47
-17.73
-16.69
-19.75
-18.72
-18.88
-18.47
-7 83
-8 55
-8 42
-7 51
-7 41
-6.98
-6.45
-6 63
-6.62
-6.05
-5.45
-5.95
-6.16
-6.18
-16.73
-17.25
-17.22
-16.21
-15.71
-16.68
-16.65
-16.73
-16.52
-16.95
-16.65
-16.75
-16.76
-16.28
235
-8.04 -18.34
-9.86 -20.06
soybean 2
08 302 07/19/83 (200) 14:46 -14.43 -19.20
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
soybean 3
13 303 07/29/83 (210) 10:30 -18.00 -24.45 -14.91 -21.21
13 303 07/29/83 (210) 10:33 -18.27 -26.52 -15.32 -24.22
13 303 07/29/83 (210) 10:35 -18.61 -26.33 -14.61 -24.41
13 303 07/29/83 (210) 10:37 -18.64 -26.88 -12.97 -24.17
13 303 07/29/83 (210) 10:39 -18.80 -23.68 -14.78 -24.68
13 303 07/29/83 (210) 10:41 -13.95 -26.61 -15.80 -25.30
13 303 07/29/83 (210) 10:43 -19.39 -24.70 -13 '9 -24.29
13 303 07/29/83 (210) 10:45 -18.77 -27.67 -15.43 -24.63
13 303 07/29/83 (210) 10:47 -19.80 -27.52 -15.58 -24.88
13 303 07/29/83 (210) 10:49 -19.56 -27.67 -13.68 -24.58
13 303 07/29/83 (210) 10:51 -22.28 -25.59 -15.87 -22.77
13 303 07/29/83 (210) 11:00 -18.17 -25.47 -15.72 -23.82
13 303 07/29/83 (210) 11:03 -18.42 -25.85 -14.21 -23.41
13 303 07/29/83 (210) 11:06 -20.69 -27.70 -15.84 -24.44
13 303 07/29/83 (210) 11:08 -20.79 -28.97 -14.75 -24.45
13 303 07/29/83 (210) 11:10 -20.86 -29.26 -15.99 -24.99
13 303 07/29/83 (210) 11:12 -21.58 -29.96 -14.28 -24.38
13 303 07/29/83 (210) 11:14 -21.32 -29.38 -15.75 -24.75
13 303 07/29/83 (210) 11:16 -21.43 -29.67 -14.60 -24.70
13 303 07/29/83 (210) 11:18 -21.73 -29.57 -14.82 -24.92
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
soybean
13 303 08/02/83 (214) 10:00
13 303 08/02/83 (214) 10:02
13 303 08/02/83 (214) 10:04
13 303 08/02/83 (214) 10:07
13 303 08/02/83 (214) 10:10
13 303 08/02/83 (214) 10:14
13 303 08/02/83 (214) 10:16
13 303 08/02/83 (214) 10:18
13 303 08/02/83 (214) 10:20
13 303 08/02/83 (214) 10:23
3
3
3
3
3
3
3
3
3
3
-21.65
-21.81
-21.49
-22.01
-20.52
-18.78
-18.61
-18.86
-17.82
-19.78
-25.39
-26.05
-26.47
-26.90
-26.79
-23.71
-23.23
-23.00
-23.04
-21.95
-6.40 -17.30
-16.02
-15.22
-15.42
-16.19
-14.94
-14.08
-13.60
-14.88
-13.74
-14.78
-23.42
-24.02
-24.02
-25.39
-24.64
-22.58
-23.00
-22.98
-22.74
-23.58
swamp
10 701 06/06/83 (157) 13:15 -15.29 -22.93
swamp
10 701 06/15/83 (166) 11:25 -14.26 -19.81 -11.56 -17.86
swamp
10 701 06/20/83 (171) 15:10-15.20 -20.28 -13.16 -20.86
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
10
10
10
10
10
10
10
10
10
10
701 07/14/83
701 07/14/83
701 07/14/83
701 07/14/83
701 07/14/83
701 07/14/83
701 07/14/83
701 07/14/83
701 07/14/83
701 07/14/83
(195) 10:55 -16.27
(195) 10:59 -15.84
(195) 11:01 -14.87
(195) 11:03 -15.95
(195) 11:07 -19.10
(195) 11:12 -15.43
(195) 11:15 -16.86
(195) 11:18 -14.03
(195) 11:21 -15.05
(195) 11:25 -16.06
236
-19.44
-20.36
-19.29
-20.51
-23.23
-20.00
-19.32
-19.57
-20.45
-20.33
-9.45 -17.55
-9.35
-7.98
-9.97
-10.22
-12.01
-17.26
-13.29
-8.88
-9.64
-9.12
-17.45
-16.58
-18.87
-18.42
-20.01
-20.06
-17.69
-17.48
-15.34
-15.92
swamp
swamp
swamp
swamp
swamp
10
10
10
10
10
701
701
701
701
701
07/14/83 (195)
07/14/83 (195)
07/14/83 (195)
07/14/83 (195)
07/14/83 (195)
11:30
11:32
11:35
11:38
11:42
-13.41
-13.40
-12.73
-12.28
-14.49
-17.72 -9.48 -17.28
-16.53 -8.70 -16.90
-17.07 -8.65 -17.45
-16.68 -10.00 -16.70
-19.37 -10.67 -18.37
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
swamp
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
701
701
701
701
701
701
701
701
701
701
701
701
701
701
701
07/19/83 (200) 12:40
07/19/83 (200) 12:45
07/19/83 (200) 12:48
07/19/83 (200) 12:51
07/19/83 (200) 12:53
07/19/83 (200) 12:56
07/19/83 (200) 12:58
07/19/83 (200) 13:00
07/19/83 (200) 13:06
07/19/83 (200) 13:08
07/19/83 (200) 13:10
07/19/83 (200) 13:13
07/19/83 (200) 13:15
07/19/83 (200) 13:20
07/19/83 (200) 13:22
-15.13
-14.66
-14.38
-16.99
-15.71
-13.01
-14.14
-14.09
-13.49
-13.54
-13.79
-12.95
-13.37
-15.44
-12.69
-18.22
-16.50
-18.07
-17.31
-18.39
-21.32
-21.64
-19.89
-19.40
-19.26
-18.80
-19.93
-19.98
-19.86
-19.89
wheat 1
06 101 06/06/83 (157) 12:05 -13.98 -22.18 -15.85 -21.85
wheat 1
06 101 06/15/83 (166) 11:00 -14.16 -25.86 -14.44 -22.64
wheat 1
06 101 06/20/83 (171) 14:45 -12.86 -22.03 -17.54 -25.04
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
1
1
1
1
1
1
1
1
1
1
06
06
06
06
06
06
06
06
06
06
101
101
101
101
101
101
101
101
101
101
07/08/83 (189) 16:05
07/08/83 (189) 16:07
07/08/83 (189) 16:08
07/08/83 (189) 16:\
07/08/83 (189) 16:12
07/08/83 (189) 16:14
07/08/83 (189) 16:16
07/08/83 (189) 16:19
07/08/83 (189) 16:21
07/08/83 (189) 16:23
-16.18
-16.34
-16.33
-16.27
16.08
-15.97
-15.49
-15.65
-15.65
-16.12
-24.36
-24.60
-24.20
-24.16
-23.91
-24.75
-26.14
-26.22
-24.41
-25.02
-11.45
-11.96
-10.86
-10.42
-11.99
-10.72
-10.71
-9.42
-9.76
-10.04
-20.75
-21.86
-20.76
-20.42
-21.69
-20.22
-20.51
-21.42
-20.46
-20.74
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
1
1
1
1
1
1
1
1
1
1
1
06
06
06
06
06
06
06
06
06
06
06
101
101
101
101
101
101
101
101
101
101
101
07/19/83 (200) 14:55
07/19/83 (200) 14:59
07/19/83 (200) 15:01
07/19/83 (200) 15:04
07/19/83 (200) 15:06
07/19/83 (200) 15:10
07/19/83 (200) 15:12
07/19/83 (200) 15:14
07/19/83 (200) 15:16
07/19/83 (200) 15:18
07/19/83 (200) 15:22
-14.33
-13.96
-14.32
-14.09
-14.71
-17.14
-18.21
-17.96
-18.18
-17.72
-16.96
-27.80
-28.09
-31.46
-26.89
-27.27
-25.17
-25.44
-25.74
-25.14
-25.75
-25.64
-13.50
-10.96
-11.84
-11.43
-12.04
-10.70
-10.65
-12.14
-11.27
-11.75
-9.78
-23.90
-22.66
-22.54
-23.23
-23.54
-23.00
-23.35
-24.44
-22.97
-23.65
-23.08
237
-9.53
-8.66
-9.25
-8.54
-10.16
-10.32
-9.57
-9.80
-10.38
-9.73
-12.92
-9.81
-10.17
-10.29
-11.15
-16.43
-16.26
-17.25
-18.64
-18.56
-19.52
-20.37
-18.00
-17.98
-17.03
-20.42
-17.81
-18.57
-18.39
-17.05
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
1
1
1
1
1
1
1
1
1
06
06
06
06
06
06
06
06
06
101
101
101
101
101
101
101
101
101
07/19/83 (200) 15:25
07/19/83 (200) 15:27
07/19/83 (200) 15:29
07/19/83 (200) 15:31
07/19/83 (200) 15:35
07/19/83 (200) 15:38
07/19/83 (200) 15:40
07/19/83 (200) 15:43
07/19/83 (200) 15:45
-16.77
-16.27
-17.24
-16.28
-16.87
-17.01
-16.96
-16.15
-16.20
-26.26
-28.17
-25.72
-25.89
-26.03
-27.01
-26.15
-25.86
-27.37
-9.98
-9.13
-9.67
-9.99
-10.75
-9.82
-11.80
-11.20
-6.25
-23.28
-22.83
-22.97
-23.09
-22.65
-22.42
-22.90
-22.70
-18.25
wheat 2
09 102 06/06/83 (157) 13:00 -14.49 -23.91 -16.80 -21.00
wheat 2
09 102 06/15/83 (166) 11:20 -12.37 -19.20 -15.81 -22.81
wheat 2
09 102 06/20/83 (171) 15:05 -13.20 -20.63 -16.96 -23.46
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
09
09
09
09
09
09
09
09
09
09
09
09
09
09
09
102
102
102
102
102
102
102
102
102
102
102
102
102
102
102
07/14/83 (195) 10 15
07/14/83 (195) 10•17
07/14/83 (195) 10 19
07/14/83 (195) 10 21
07/14/83 (195) 10 25
07/14/83 (195) 10 27
07/14/83 (195) 10.30
07/14/83 (195) 10 33
07/14/83 (195) 10 35
07/14/83 (195) 10 37
07/14/83 (195) 10 40
07/14/83 (195) 10 43
07/14/83 (195) 10:45
07/14/83 (195) 10 47
07/14/83 (195) 10 50
-23.11
-21.45
-21.04
-20.76
-20.67
-14.49
-14.50
-13.94
-12.90
-13.39
-16.08
-15.92
-16.13
-15.78
-15.94
-26.11
-25.85
-25.69
-26.01
-26.08
-23.63
-23.74
-23.32
-23.10
-23.65
-21.78
-21.51
-21.60
-21.64
-21.64
-11.62
-12.24
-11.42
-11 88
-11.99
-11 18
-9 40
-10 18
-9 27
-9 51
-9 55
-9 09
-9 77
-9 20
-9 08
-21.62
-21.94
-21.92
-21.98
-22.39
-22.28
-21.30
-21.48
-21.17
-21.11
-16.95
-16.79
-16.97
-17.60
-17.68
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
09
09
09
09
09
09
09
09
09
09
09
09
09
09
09
102
102
102
102
102
102
102
102
102
102
102
102
102
102
102
07/19/83 (200) 13.25
07/19/83 (200) 13 28
07/19/83 (200) 13.30
07/19/83 (200) 13 33
07/19/83 (200) 13:35
07/19/83 (200) 13- 38
07/19/83 (200) 13:40
07/19/83 (200) 13:43
07/19/83 (200) 13:45
07/19/83 (200) 13:48
07/19/83 (200) 13:50
07/19/83 (200) 13:52
07/19/83 (200) 13:55
07/19/83 (200) 13:57
07/19/83 (200) 14:00
-14.25
-14.75
-14.06
-13.87
-14.22
-17.08
-16.10
-16.13
-16.75
-16.43
-15.91
-15.00
-14.86
-14.80
-14.84
-23.35
-23.06
-22.46
-23.09
-22.60
-23.56
-22.87
-23.23
-22.57
-23.08
-23.86
-23.92
-24.61
-24.55
-24.63
-8.48
-8 38
-8.10
-8.86
-8.29
-9.85
-9.97
-8.86
-8.84
-8.78
-10.84
-10.88
-10.29
-10.26
-11.59
-20.08
-20.38
-20.10
-20.36
-20.49
-20.75
-21.47
-20.06
-20.74
-20.48
-20.94
-20.88
-20.99
-20.86
-20.59
wheat 3
04 151 06/06/83 (157) 11:54 -11.28 -18.97 -10.23 -17.83
238
wheat 3
04 151 06/15/83 (166) 10:45 -13.63 -23.01 -12.05 -21.05
wheat 3
04 151 06/20/83 (171) 14:30 -12.55 -21.22 -10.48 -19.38
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
04
04
04
04
04
04
04
04
04
04
3
3
3
3
3
3
3
3
3
3
151 07/08/83 (189) 13:45
151 07/08/83 (189) 13:47
151 07/08/83 (189) 13:49
151 07/08/83 (189) 13:51
151 07/08/83 (189) 13:53
151 07/08/83 (189) 13:56
151 07/08/83 (189) 13:58
151 07/08/83 (189) 14:00
151 07/08/83 (189) 14:02
151 07/08/83 (189) 14:05
-13.18 -19.06
-13.50 -18.67
-12.20 -19.33
-12.26 -19.03
-12.89 -19.84
-15.14 -22.03
-16.28 -22.54
-16.42 -22.32
-16.19 -21.92
-15.70 -23.04
-12.75
-14.18
-13.09
-12.61
-13.05
-13.89
-14.23
-13.51
-13.54
-13.19
-18.95
-19.98
-18.79
-18.41
-19.85
-19.99
-20.93
-20.01
-19.84
-20.09
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
wheat 3
04 151 07/29/83 (210) 11:15 -17.81 -20.81 -11.64 -19.84
04 151 07/29/83 (210) 13:18 -13.38 -15.92 -11.62 -20.02
04 151 07/29/83 (210) 13:22 -17.87 -21.10 -11.75 -19.05
04 151 07/29/83 (210) 13:24 -17.44 -21.58 -11.74 -20.44
04 151 07/29/83 (210) 13:26 -17.61 -21.86 -11.67 -20.47
04 151 07/29/83 (210) 13:29 -18.23 -22.29 -12.33 -20.53
04 151 07/29/83 (210) 13:31 -19.24 -23.51 -11.75 -20.25
04 151 07/29/83 (210) 13:34 -17.94 -21.28 -11.98 -19.98
04 151 07/29/83 (210) 13:36 -17.59 -22.08 -11.99 -20.49
04 151 07/29/83 (210) 13:38 -17.92 -22.00 -12.38 -20.68
04 151 07/29/83 (210) 13:43 -14.94 -18.86 -10.26 -16.76
04 151 07/29/83 (210) 13:45 -14.83 -18.64 -10.61 -16.91
04 151 07/29/83 (210) 13:47 -14.74 -18.90 -10.28 -17.08
04 151 07/29/83 (210) 13:49 -15.05 -19.05 -10.48 -17.08
04 151 07/29/83 (210) 13:51 -15.11 -18.76 -9.95 -17.15
04 151 07/29/83 (210) 13:54 -14.94 -19.84 -10.22 -17.12
04 151 07/29/83 (210) 13:55 -15.07 -18.89 -10.29 -16.69
04 151 07/29/83 (210) 13:58 -14.68 -18.86 -10.17 -17.17
04 151 07/29/83 (210) 14:00 -14.85 -18.70 -11.99 -17.79
04 151 07/29/83 (210) 14:04 -14.65 -19.42 -11.49 -17.19
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
wheat
04
04
04
04
04
04
04
04
04
04
3
3
3
3
3
3
3
3
3
3
151 08/02/83 (214) 10:30 -17.64 -23.36
151 08/02/83 (214) 10:32 -17.90 -25.82
151 08/02/83 (214) 10:35 -21.64 -23.21
151 08/02/83 (214) 10:37 -21.54 -24.07
151 08/02/83 (214) 10:39 -18.52 -22.75
151 08/02/83 (214) 10:42 -16.82 -22.17
151 08/02/83 (214) 10:44 -17.27 -22.93
151 08/02/83 (214) 10:47 -17.12 -23.06
151 08/02/83 (214) 10:49 -17.41 -22.78
151 08/02/83 (214) 10:51 -16.97 -23.24
239
-12.35
-12.59
-12.40
-12.40
-12.64
-12.25
-12.63
-13.15
-12.95
-12.24
-20.15
-20.79
-20.70
-20.80
-20.14
-19.75
-19.53
-19.75
-19.45
-19.54
APPENDIX E
MEASUREMENTS OF SELECTED SCENE CHARACTERISTICS (.1983)
1. This Appendix contains tabulated ground truth data for
field.
The
measurements
were
obtained
in
accordance
each
with the
procedures discussed in Jung et al. (1983).
2. For growth
"AgRISTARS
stage
Enumerator's
codes
and
percent
ground
cover, see
Manual," 1981 Ground Data Survey, National
Aeronautics and Space Agency/Johnson Space Center, Houston, Texas.
3. Row separation was measured between leaves and between rows.
Percent
soil
moisture (S.M.) and percent plant moisture were cal­
culated based on wet weight.
ZkO
Corn 1 (Field 1)
Growth
Stage
Canopy
Height
Row Sep
Stems
32
60
77
20
10
BW prints; Ridge and
furrow not visible
22
59
25
76
20
26
BW prints and color
slides
10:15
22
72
14
75
25
24
(179)
09:20
23
135
0
76
40
6/30
(181)
13:40
7/05
(186)
13:45
7/08
(189)
11:45
7/12
(193)
14:20
31
254
0
74
80
38
7/19
(200)
10:30
41
300
0
75
90
37
Time
6/06
(157)
14:00
22
6/17
(168)
10:00
6/20
(171)
6/28
24
182
0
74
% Ground
Cover
80
Stem Dia
(mm)
t S.M.
Row Sep
Leaves
Date (Julian)
0-2 cm
* S.M. % Plant
2-5 cm
Moist
27.3
26.4
22.2
23.3
91.8
Notes
BW prints and color
si ides
14 plants per 5 m row
27
4.7
13.5
85.9
Color print
Corn 2 (Field 5)
Row Sep
Leaves
Row Sep
Stems
22
33
61
75
10
9
No ridge and furrow
visible; BW prints
10:35
22
75
9
76
25
22
Recently cultivated;
BH prints and color
slide
(171)
10:45
22
80
7
77
35
31
6/28
(179)
09:45
23
142
0
77
45
6/30
(181)
10:20
7/05
(186
15:15
7/08
(189)
13:20
7/12
(193)
13:40
7/14
(195)
7/19
.. (200)
09:35
41
272
0
73
100
7/29
(210)
10:35
42
261
0
75
95
8/03
(215)
11:00
43
273
0
72
95
28
Color slide
8/17
(229)
10:15
44
256
0
74
95
29
Lower 1/3 leaves are
brown; Color slide
Time
6/6
(157)
11:15
6/17
(168)
6/20
Growth
Stage
% Ground
Cover
Stem Dia
(mm)
% S.M.
0-2 cm
% S.M. % Plant
2-5 cm
Hoist
Canopy
Height
Date (Julian)
24.3
206
0
77
90
257
0
72
95
20.7
96.0
31
BW prints
18 plants per 5 m rov
3.9
33
22.1
Color slide
20.0
24
Notes
10.0
85.7
34
Color print
27
Top cob length:24 cm
Bottom leaves drying
Fallow (Field 12)
Growth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
% Ground
Cover
Date (Julian)
Time
6/06
(157)
12:45
6/17
(168)
6/20
(171)
12:40
33
90
6/30
(181)
12:30
44
95
7/05
(186)
11:00
15
7/14
(195)
10:30
7/19
(200)
11:00
13
Stem 01a
(mm)
% S.M.
0-2 cm
% S.M. % Plant
2-5 cm
Moist
70
Notes
Majority is clover.
BW prints
BH prints and color
slide
35.1
83.1
BW prints; color sUdi
Cut recently
9.5
30
33.8
15.1
72.6
New growth;
color prints
Cut grass (Field 7)
Date (Julian)
Time
Grovrth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
% Ground
Cover
Stem Dia
(mm)
% S.H.
0-2 cm
% S.H. % Plant
2-5 cm
Hoist
Notes
6/17
(168)
9
BW print
Color slides
6/20
(171)
9
BH print
6/28
(179)
09:50
9
Color slide
6/30
(181)
11:20
7/05
(186)
7/12
(193)
7/14
(195)
7/19
(200)
09:45
8
7/29
(210)
10:45
8
8/03
(215)
11:00
9
8/17
(229)
10:20
12
35.1
29.6
59.5
BW print
12
14:00
9
Color print
Color slides
Hay
Date (Julian)
Time
Growth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
(Field 3)
% Ground
Cover
Stem Dia
(tin)
% S.M.
0-2 cm
* S.M. % Plant
Hoist
2-5 cm
Notes
6/17
(168)
26
BW prints
Color slides
6/20
(171)
34
BW prints
6/30
(181)
09:45
42
7/05
(186)
14:30
40
7/08
(189)
12:30
7/12
(193)
14:30
7/14
(195)
7/19
(200)
36.1
31.8
74.5
13.8
17.0
60.6
BW prints
Color slides
37
Color prints
10:45
16
Hay cut recently
Mllo (Field 16)
Growth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
t Ground
Cover
Stem Oia
(mm)
Date (Julian)
Time
8/03
(215)
12:30
126
0
73
85
14
8/17
(229)
11:00
125
0
75
85
18
t S.M.
0-2 cm
3.9
% S.M. % Plant
2-5 cm
Hoist
5.4
Notes
Color slides; 11 head!
per meter; 19 plants
per meter;
Tall heads brown,
short green
Color slides
Potatoes (Field 11)
Growth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
13:30
Bud
stage
34
31
75
45
10
Peak to peak: 80 cm
Peak height: 15 cm
(168)
11:20
Flower­
ing
40
16
75
50
15
10 plants per meter
BVJ prints and color
slides
6/20
(171)
12:25
46
0
74
55
6/30
(181)
12:10
48
23
72
70
7/05
(186)
10:30
49
3
0
70
Date (Julian)
Time
6/06
(157)
6/17
% Ground
Cover
Stem Dia
(mm)
% S.M.
0-2 cm
20.7
% S.M. % Plant
2-5 cm
Hoist
18.2
86.5
(195)
11:15
46
0
0
65
7/19
(200)
11:45
39
0
0
75
7/29
(210)
10:00
13
Plant moisture under
ground: 78.4%
Plants are wilting;
Light insect damage
15
-
7/14
Notes
4.9
5.7
86.5
Plant moisture under
ground: 75.01
Some weeds;
Color prints
2.0
2.3
Harvested
Soybean 1 (Field 2)
Growth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
% Ground
Cover
Stem Dia
(mm)
22
11
65
78
20
2
Surface smooth;
BW prints
19
50
73
20
3
Cultivated 6/15; Light
insect damage; Color
slide.
23
26
55
73
30
4
23
48
25
72
40
Date (Julian)
Time
6/06
(157)
10:45
6/17
(168)
09:50
6/20
(171)
10:05
6/28
(179)
09:30
6/30
(181)
13:20
7/05
(186)
13:30
7/08
(189)
13:00
7/12
(193)
14:15
7/14
(195)
7/19
(200
10:25
40
80
0
73
85
7/29
(210)
10:30
42
85
6
76
85
8/03
(215)
11:00
42
97
0
73
95
10
8/17
(229)
09:45
43
92
0
75
95
9
22
32
32
58
66
20
12
74
75
70
80
% S.M.
0-2 cm
32.4
% S.M. % Plant
2-5 cm
Moist
Notes
Rained over weekend
27.7
Light insect damage
6
6
21.0
21.2
81.9
BW prints; color
slides.
4.2
14.1
81.4
21 plants per meter;
Moderate insect damage
4.9
14.1
81.7
Moderate insect damage
Color prints
8
Soybean 2 (Field 8)
% S.M,
2-5 cm
Row Sep
Leaves
Row Sep
Stems
12:30
12
9
64
74
10
3
RW prints
(168)
10:45
21
14
56
74
15
4
Recently cultivated;
RW print and color
slide
6/20
(171)
11:00
22
18
56
75
20
3
Smooth surface
6/28
(179)
10:00
23
36
43
74
30
6/30
(181)
11:35
7/05
(186)
14:45
32
42
25
74
50
6
7/12
(193)
13:25
32
54
18
76
55
5
7/14
(195)
10:10
7/19
(200)
09:50
6/06
(157)
6/17
Stem tlia
(mm)
% S.M.
Canopy
Height
Time
% Ground
Cover
0-2 cm
Growth
Stage
Date (Julian)
65
4
73
95
7
Notes
32.4
27.7
20.7
19.9
81.6
BW print and color
slide
3.4
10.9
82.2
19 plants per meter
Light insect damage
2.6
41
t Plant
Moist
4.1
80.3
Soybean 3 (Double-cropped soybeans) (Field 13)
Date (Julian)
Time
Growth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
% Ground
Cover
Stem 01a
(tmi)
% S.H.
0-2 cm
% S.H. % Plant
2-5 cm
Hoist
Notes
Wheat stubble: 48 cm
7/12
(193)
13:15
12
10
70
75
2
7/19
(200)
09:45
21
14
70
75
3
7/29
(210)
10:15
21
20
69
76
3
Color slides
8/03
(215)
10:45
22
13
61
74
3
Uneven growth
color slides
8/17
(229)
09:45
23
34
40
74
3
Flowering, but growing
leave; color slides
rv>
-fr
vo
Swamp (Field in)
Date (Julian)
Time
Growth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
% Ground
Cover
Stem nia
(mm)
% S.H.
0-2 cm
t S.M. * Plant
2-5 cm
HolSt
Notes
6/06
(157)
RW prints
6/17
(168)
RW prints
color slides
6/30
(181)
Color slide
7/14
(195)
10:30
7/19
(200)
10:00
42.8
34.6
Color prints
Water has receded
~ 30 ft. from road
Wheat 1 (Winter Wheat) (Field 6)
Growth
Stage
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
11:30
33
97
0
19
95
3
Some bare areas;
BW prints
(168)
11:00
42
109
0
19
95
3
31 stems per 15 cm row
BW print and color
slides
6/20
(171)
12:00
42
102
0
14
80
3
Heads bending;
Leaves dry
6/28
(179)
09:50
43
100
0
15
80
6/30
(181)
10:35
7/05
(186)
15:00
61
100
7/08
(189)
14:10
61
100
Date (Julian)
Time
6/06
(157)
6/17
% Ground
Cover
Stem Oia
(tun)
% S..1.
0-2 cm
% S.H. * Plant
2-5 cm
Hoist
Notes
Host heads bent
22.2
22.3
39.6
BW print and color
slide
129 stems per meter
21.9
21.2
9.7
skssmatd
Wheat 2 (Winter Wheat) (Field 9)
Growth
Stage
Canopy
Hei ght
Row Sep
Leaves
Row Sep
Stems
11:30
41
106
0
17
90
3
RW prints
(168)
11:10
42
112
0
17
90
3
BW prints and color
slides
6/20
(171)
12:10
42
101
0
13
80
3
6/28
(179)
10:05
43
109
0
19
80
6/30
(181)
11:50
7/05
(186)
11:15
61
42
7/12
(193)
13:30
61
45
7/14
(195)
10:15
61
7/19
(200)
09:55
61
Date (Julian)
Time
6/06
(157)
6/17
% Ground
Cover
Stem Dia
(mm)
% S.M.
0-2 cm
21.7
« S.H. 5 Plant
2-5 cm
Hoist
25.4
Notes
Heads bending;
leaves are dry
Some heads turning
dark
27.4
25.7
35.1
BW prints
139 stems per meter;
stubble
10.2
18.7
8.9
Color prints
Stubble tilled under
very large clods
Wheat 3 (Spring Wheat) (Field 4)
Canopy
Height
Row Sep
Leaves
Row Sep
Stems
11:00
23
23
0
18
60
2
BW prints
(168)
10:25
25
57
0
16
80
4
Many weeds; BW and
color slides
6/20
(171)
10:40
32
70
0
13
90
3
6/28
(179)
09:40
34
73
0
20
80
6/30
(181)
10:10
7/05
(186)
15:30
7/08
(189)
13:50
7/12
(193)
13:50
7/14
(195)
7/19
(200)
09:25
52
70
0
13
80
3
Grains shrlvled, a few
green heads left
7/29
(210)
10:25
53
68
0
18
85
3
Color slides
8/03
(215)
10:00
53
62
0
14
85
3
Heads bent to the
north; difficult to
see rows
8/17
(229)
10:00
Time
6/06
(157)
6/17
42
43
68
68
0
0
11
15
% Ground
Cover
75
75
Stem Dia
(mm)
% S.M. X Plant
2-5 cm
Moist
Growth
Stage
Date (Julian)
% S.M.
0-2 cm
25.0
Notes
24.0
Leaves yellowing;
Color slide
21.1
18.1
70.0
BW prints
2.6
6.5
54.8
63 plants per meter;
leaves drying out
3
3
Color print
Crop tilled under;
color slides
Документ
Категория
Без категории
Просмотров
0
Размер файла
10 905 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа