THE UNIVERSITY OF KANSAS SPACE TECHNOLOGY LABORATORIES
2291 Irving Hill Dr, — Campus West Lawrence, Kansas 66044
Telephone:
NASA CR*
/_2£z^j££.
RADAR STUDIES RELATED TO THE EARTH RESOURCES PROGRAM
(NASA-CB- 1 41 643) EADAB STUDIES BELATED TO N75-18698'
THE EABTH KESOUECES PBOGBAM (Kansas Univ.)
171 p HC $6.25 CSCL 201
Unci as
G3/43 13363
CRES Technical Report 177-26
March, 1972
Supported by:
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
Johnson Spacecraft Center
Houston, Texas 77058
CONTRACT NAS 9-10261
REMOTE SENSING LABORATORY
INTERIM TECHNICAL PROGRESS REPORT
RADAR STUDIES RELATED TO THE EARTH RESOURCES PROGRAM
Table of Contents
I. INTRODUCTION 1
II. TASK PERFORMANCE 2
TASK 2.2 IMAGING RADAR SYSTEMS 2
PATTERN RECOGNITION ALGORITHMS 22
IMAGE DISCRIMINATION, ENHANCEMENT
AND COMBINATION SYSTEM (IDECS) 32
TASK 2.3 ALTIMETRY, SCATTEROMETRY AND
OCEANOGRAPHIC APPLICATIONS OF RADAR 40
TASK 2.4 RADAR GEOLOGY 57
TASK 2.5 RADAR APPLICATIONS IN
AGRICULTURE/FORESTRY 64
III. CONTRACT PUBLICATIONS
163
Interim Technical Progress Report
RADAR STUDIES RELATED TO THE EARTH RESOURCES PROGRAM
I. INTRODUCTION
Radar systems research at Kansas is directed toward achieving successful
application of radar to remote sensing problems in various disciplines, such as geology,
hydrology, agriculture, geography, forestry and oceanography. Such goals require
understanding (1) the performance of radar systems for the various applications in terms
of pertinent system and target parameters, and (2) the means for extracting information
from the radar image . This involves analysis of abilities of existing radars for the
different applications, and development of specifications for future radar systems and
their information extraction systems.
With these ideas as a guide, our radar system efforts have included:
1 . Support for modification of NASA imaging radar, related system analysis
and evaluation of resulting imagery.
2. Study of digital processing for synthetic aperture system.
3. Digital simulation of synthetic aperture system,
4. Averaging techniques studies.
5. Ultrasonic modeling of panchromatic system,
6. Panchromatic radar/radar spectrometer development,
7. Measuring octave “bandwidth response of selected targets.
8. Recommendations for future systems including a 1975 spacecraft system.
9. 13.3 GHz scatterometer system analysis including data processing
recommendations ,
10. Technical support of sea state, sea ice and agricultural missions and
analysis of the resulting scatterometry and imagery data.
IT. Construction and testing of a model Fresnel“zone processor for synthetic
aperture imagery.
12. Consultation with other disciplines conducting remote sensing and radar-
related studies.
1
The following sections summarize the progress accomplished. The subsections
have been numbered with reference to the Radar System Tasks of Technical and Cost
Proposal BG 921 -36-0- T5 P, dated August 1969. Reference should also be made to
Technical Report 1 77- 13, the Interim Progress Report, and the specific reports
indicated. Section 2. 2. 1.9 provides a good overall summary and is indicative of the
interrelationships between the various subtasks. Three of the subsections are more
substantive than the others since the detailed reports will be published at a later date.
II. TASK PERFORMANCE
IMAGING RADAR SYSTEMS
2.2. 1.1 Testing of SLAR Performance
2. 2. 1.2 SLAR Procedures Guide
2.2. 1.3 a, b Imaging Radar Modifications and Processing
The University of Kansas has provided technical support for NASA's Philco-Ford
DPD-2 SLAR, including recommendations for modification to its present form. As
presently configured this instrument is capable of producing useful imagery. The degree
of the usefulness of the imagery is limited by as-yefuncorrectcd system deficiencies
which include recording film transport speed errors, inoperative STC circuitry, gross
sensitivity difference between the like and cross polarization images and the occurrence
of intermittent system malfunctions. Successive missions have shown a general improve-
ment in the quality of the resulting imagery, with a failure in the synthetic aperture
channel on Mission 168, a major exception. Imagery from DPD-2 has been used when
possible, as has been reported on elsewhere. The original recommendation to acquire
the DPD 2 was based primarily upon its performance in flight tests for the Army and the
fact that it was the only synthetic-aperture radar unclassified at that time. Of the sev-
eral modifications to the basic instrument, the dual-polarized antenna and the abandon-
ment of quadrature detection have tended to degrade performance. One additional modi-
fication proposed by K.U. has now been incorporated; the Instrument can now be operated
in both the synthetic- and real-aperture modes. A review of the basic system operation,
including modifications, is contained in CRES Technical Memorandum 177-18.
2
A brief study has been performed as to the effect of quadrature versus non -
quadrature detection. The calculations and simulation on a digital processor program
demonstrate the loss of imagery quality with non-quadrature detection. It remains to
be seen whether the value of the additional polarizations added in lieu of the quad-
rature channel outweighs the loss incurred from the increased "speckle” in the radar return.
2,2,1.3 c Digital Processing Techniques
Digital processing for synthetic aperture imagers has been a major area of
study. This research has culminated in a dissertation by R. Gerchberg and a report
TR 177-10. The digital processor has been simulated on the University's computer with
various imager-processor configurations. A great deal of flexibility is built into the
program to permit simulation with varying antenna patterns, transfer functions, etc.
One of the most important aspects of this research was the inclusion of sub-
aperture processing. In lieu of full-focussed processing of the entire synthetic aperture,
subapertures could be processed more or less in parallel and then superimposed, thus
reducing the storage requirements and permitting averaging several Independent
samples to improve the image appearance. This improvement in gray scale is traded
for resolution. A digital processor system was postulated which appears feasible for
the 1975 time period if advances in the state-of-the-art in storage technology proceed
as expected .
The simulated processor has been used as a basic research tool and the program
has been repackaged for use on the PDP-15/20 computer at K.U. The simulated
processor was used in the non -quadrature studies cited above for evaluating imaging
radar performance with respect to specific target types.
2. 2. 1.5 Define Available Imagers
2.2. 1.6 Performance Specifications for Aircraft Polypanchromatic/Polypanchromatic
Radar
Due to the very nature of its efforts, the radar system group has kept abreast of
the imaging radar art. This effort includes a continuing literature survey, attendance at
appropriate meeting and symposia and informal contact with universities and industry.
Performance specifications have been written and described in several
technical memoranda, reports and NASA presentations. Technical Report 177-18
summarizes specifications for one candidate multispectral radar. Most of the effort
has been concentrated on satellite systems (see 2. 2. 1,9).
3
Modulation Bandwidth -400 MHz
IF Frequency - 37 KHz
Antenna - 3' diameter dish
Figure 1
BLOCK DIAGRAM OF RADAR SYSTEM
2. 2. 1. 7 Acoustic Modeling of Panchromatic/Polypanchromatic Radars (Parti a I ly
funded by Project THEMIS - Contract DAAK02-C-0089)
This task was completed during the first year of the contract period and is
reviewed in Interim Report TR 177-13, TR 177-11 is the definitive report on this
task. It describes the theory of panchromatic illumination and the acoustic (ultra-
sonic) simulation of panchromatic imaging. Variance reduction resulting from
panchromatic averaging and the improvement in image appearance was clearly
demonstrated ,
2,2, 1,8 Panchromatic Radar System (Partially funded by Project THEMIS -
Contract DAAK02-68-C-0089)
2,2.1. 8a System Description
The earlier theoretical and experimental work on polypanchromatic illumination
has clearly demonstrated the advantages of panchromatic averaging (c.f, TR 177-11).
Initial results with a pulsed polypanchromatic system verified acoustic tank simulations.
After verification of the value of panchromatic illumination was complete, a
decision was made to concentrate use of the system on cataloging microwave signatures
of targets of interest in environmental monitoring.
Difficulties in use of the pulse system at ranges under 20 m led to modifying
the broad-band spectrometer by use of frequency modulation instead of pulse modula-
tion. The FM system can operate at shorter ranges, and components are more suitable
for frequency-range expansion. Figure 1 shows a simplified block diagram of the radar.
During the 1971 growing season, initial measurements were made of the sig-
natures of crops. Problems with the absolute calibration and modulation techniques
have necessitated a refurbishment of the system.
Analysis of the 1971 data caused us to doubt the validity of the corrections
originally made on the basis of a single set of measurements of a calibration sphere.
Consistent variations that appeared in all the data led to the conclusion that the cor-
rections due to the sphere measurements were uniformly too large at some frequencies.
Consequently, the data have been reprocessed Into a relative format based on averages
of the field observations rather than on the calibration sphere.
A thorough analysis of the observations and many additional component cali-
brations have led to a modification of the original system using some new components
and incorporating additional internal calibration possibilities. Laboratory tests indicate
that the new configuration is more satisfactory, and more accurate results are anticipated
5
using it during the forthcoming 1972 growing season. Furthermore, a new boom truck
has been obtained that will permit greater ease of operation, more stability for the
antenna, and a larger minimum range.
The system originally known as the ground-based polypanchromatic radar has
now been rechristened a ground-based radar spectrometer, since the imaging capa-
bility of the original pulse system is no longer included. A passive measurement
capability is presently being added, so that the system will become a combined radar
spectrometer-microwave spectroradiometer. This system will be described in a
subsequent report.
Measurements with the new system will again be made first in the 4-8 GHz
spectrum, but the frequency range will be extended to 12.4 GHz shortly. The
4-8 GHz range was selected for initial use because it is the highest frequency range
for which full - octave components are generally available.
2,2,1.8b 1971 Data Measurement Program
Octave-bandwidth (4“8 GHz) radar spectrometer measurements were obtained
during a 2~week period in July 1971 in an attempt to associate signature definitions
to agricultural crops. The spectral response data was concentrated on agricultural
targets covering 71 fields with 4 crop types: corn, sorghum, soy beans, and
alfalfa; and plowed ground. The experiment consisted of measuring the spectral
responses of the agricultural crops across the 4-8 GHz frequency band at look angles
of 0°, 10°, 20°, 30°, 50°, and 70° for both horizontal and vertical polarizations.
Using a continuously tunable sweep oscillator, radar backscatter measurements
were recorded at 10 frequency points over the 4~8 GHz frequency range (4.2, 4.6,
5.0, 5.4, 5.8, 6.2, 6.6, 7.0, 7.4 and 7.8 GHz). Each measurement point was
an average over a 400 MHz bandwidth. For each agricultural field considered, this
procedure was followed for both polarizations (vertical and horizontal) and at each of
the 6 look angles. Though the collected data included spectral measurements at 0°
look angle, those measurements may be in error due to the proximity of the truck.
Hence, the 0° look angle set of data was deleted.
Measurements were conducted over 28 fields of corn, 7 fields of sorghum, 18
fields of soy beans, 14 fields of alfalfa and 4 fields of plowed ground. The data was
then averaged and plotted as shown in Figures 2 through 6 . The scattering coefficient
is expressed in relative and not absolute scale. The plowed ground category was deleted
since only four fields were covered. Variations in system performance over the frequency
range were indicated by the sphere measurements, but are not taken into account in
these illustrations. The presence of consistent variations is, however, apparent in the data.
6
Scattering Coelticient (dBl -- Relative Scale Scattering Coeiflclent (OB) - Relative Scale
HORIZONTAL POLARIZATION
VERTICAL POLARIZATION
I l a r , 1 -1 - I 1 -- -^---1 1 l l I I 1 1 1 J
4. 0 5. 0 6 . 0 7. 0 8. 0
frequency (CHzT
FIGURE 2 . RAW BROAD LAND SPECTRAL DATA AT 10° LOOK ANGLE.
Scattering Coefficient (dB> -- Relative Scale Scattering Coefficient IdB) - Relative Seal
HORIZONTAL POLARIZATION
FIGURE fe- RAW BROAD BAND SPECTRAL DATA AT 70’
11
LOOK ANCLE.
Because of the system variations with frequency, and lack of confidence in
the sphere measurement, a normalization technique was required. Two methods were
were used: the spectral data for the four crops may be plotted using one of the crops
as a standard; or for each frequency, look angle, and polarization, the four data
points corresponding to the four crops may be normalized with respect to their sum.
Both techniques tend to eliminate system variations with frequency.
Using corn as a standard, the spectral response curves at 10°, 30°, 50°
and 70 ° look angles are shown ?n Figures 7 through 10.
The concept of broad“band microwave imaging (polypanchromatic) radar
developed from the visual analog. To illustrate the value of this approach, the
observations were combined to produce visual displays of the “microwave color.”
For each crop, polarization and look angle, the 4-8 GHz data was divided into three
sub-bands; Red Band = 4-5.2 GHz, Green Band = 5. 2-6. 8 GHz, and Blue Band =
6.8 _ 8 GHz. The averaged return for each crop over each of the sub-bands was
normalized to the sum of the return from the four crops (to eliminate system variations
with frequency as discussed earlier) and then used to set the intensity level of the
corresponding light beam of a three-beam color combiner. The color signatures of the
four crop types were then grouped by look angle and polarization. Examples are shown
in Figure 11 corresponding to look angles of 10°, 30°, 50°, and 7U°.
Though the results Indicate that the four crop types can be easily discriminated using
any one of the four sets (both horizontal and vertical), the 50^ look angle set appears
optimum.
Possible effects due to plant or soil moisture were ignored above; the study
was concentrated on crop types rather than differences within a particular crop. In
conjunction with the radar spectrometer measurements, ground truth data were col-
lected including soil and plant samples. The moisture contents, by weight and by
volume, were determined for both the soil and the plant samples in the laboratory.
The next phase of the analysis will concentrate on variations of the radar return as
a function of look angle and moisture content.
12
Scattering Coefficient Relative To Corn MB) ^ Scattering Coefficient Relative to Corn fdB)
13
Scattering Coefficient Relative to Corn (dB) Scattering Coefficient Relative to Corn (dB)
HORIZONTAL POLARIZATION
CORN
Frequency (GHz)
VERTICAL POLARIZATION
FIGURE 8. RAW BROAD BAND SPECTRAL DATA AT 30 a LOOK ANGLE;
14
Scattering Coefficient Relative to Corn (dB) Scattering Coefficient Relative to Corn (dB)
■ *
HORIZONTAL POLARIZATION
VERTI CAL POLARIZATION
l — 1 1 1 j
4.0 5.0 6.0 7.0 8.0
Frequency (GHz)
FIGURE 9. RAW BROAD BAND SPECTRAL DATA AT 50° LOOK ANGLE.
15
Scattering Coefficient Relative to Corn (dB) Scattering Coefficient Relative to Corn MB)
HORIZONTAL POLARIZATION
VERTICAL POLARIZATION
“ CORN
1 1 i
5.0 6.0 7.0
Frequency fGIIZ)
Figure 10. Eroad Band Spectral Data at 70° Look Angle.
16
SORGHUM
SOYBEANS
ALFALFA
SORGHUM
SOYBEANS
ALFALFA
COLOR COMBINED NORMALIZED ONE -THIRD OCTAVE AVERAGES 30°
COLOR COMBINED NORMALIZED ONE -THIRD OCTAVE AVERAGES 70°
Figure 11
2. 2. 1.9 1975 Space System
Specification of remote sensing imaging radar systems is complicated by
economic, vehicle, and state-of-the-art considerations. The desirability of multi -
polarization polypanchromatic systems has been well documented at this point in
time. General recommendations for geoscience radar systems are contained in report
TR 177-18. The following two figures show representative systems which might be
flown on satellites. These figures are not to be considered as final specifications;
they could be modified in many ways depending upon the constraints associated with
the satellites.
Figure 12 shows the specifications for a radar for a small satellite capable of
supplying 450 watts of primary power. With this much power the system could have a
resolution os fine as 10 m with a swathwidth of 40 km. To achieve this large swath-
width, however, requires erecting an antenna on the small satellite with a length of
the order of 6 to 10 m. Note that, if the radar is only used 20% of the time, the
average power over an orbit is only 90 watts. The system selected was to operate out
to an incidence angle of 60° for geologic purposes, A 30° incident angle appears
quite adequate for agricultural purposes and many geologic purposes. With such an
incident angle, the power requirement could be reduced to 275 watts, which means
that the average over an orbit is only 55 watts.
Such a system would, we assume, transmit raw video data to the ground via a
wideband telemetry link. Processing would then take place on the ground. The power
for the telemetry link is not included in the figure , Of course, a system with a
poorer resolution could get by with significantly less power for both radar and telemetry.
Figure 13 shows a possible system for a large satellite such as a shuttle.
Here a polypanchromatic system Is postulated with a listing of first the power required
with no averaging, then that with 200 MHz, and then with 400 MHz averaging.
The 200 MHz averaging gives about 10 independent samples per cell due to pan-
chromatic illumination, and the 400 MHz gives about 20 per cell. Note that this
4-frequency system requires a total of 975 watts without averaging, but with 10
independent samples per cell the power requirement goes up by a factor of about 10,
The frequencies were chosen somewhat arbitrarily because we still do not have
adequate data over a wide frequency range . However, 16 GHz appears about the
highest reasonable frequency for a spacecraft imaging system both from the standpoint
of available components and of atmospheric effects. The 10 GHz band is quite common
and many data have been gathered at this frequency that indicate the value of radar.
18
SMALL SATELLITE SYSTEM
RESOLUT! ON: 10 m
SWATHVVIDTH: 40 km
HEIGHT: 600 km
o°: -25 dB
INCIDENCE ANGLE: 60°
SNR: 6 dB
AVERAGE TRANSMITTED POWER: 125 W
POWER (INCLUDING SYSTEM & LOSSES) * 450 W
AVERAGE POWER FOR A 20% ORBITAL DUTY CYCLE: 90 VV
30° INCIDENCE ANGLE GIVES A 8.4 dB SNR, ALLOWING
275 W AVERAGE POWER WITH ORBITAL AVERAGE OF 55 W.
Figure 12. Representative design for SLAR for small
satellite. Poorer range resolution would
allow lower power.
Our recent observations over the octave bandwidth 4“8 GHz indicate the value of
radar "color" in that band. All of these frequencies are near to frequencies that
may be available for allocation to spacecraft imaging radar systems.
As our analysis of the data gathered both with the imaging radars and the
radar spectrometer and scatterometers continues, we hope to be able to refine these
specifications. Nevertheless, we believe if a radar system were to be constructed
for spacecraft use immediately, these specifications should serve as a reasonable
guide .
System analysis efforts have also been concerned with specifying the relation -
ship between the radar signal and the film optical density in terms of the radar and
film system parameters. Both visual interpretation of radar imagery and analysis
19
LARGE SATELLITE SYSTEM
FREQUENCY AVERAGE RADI ATED
TRANSMITTER
WI DE-BAND AVERAGING (W)
SYSTEM
(GHz)
POWER (Wi
POWER (WI
200 MHz
400 MHz
POWER (W)
4
26
76
1020
2220
240
8
52
157
1870
3880
1
10
64
192
2500
5200
1
16
102
303
3980
8300
t
£ 735
9350
19400
POWER REQUIRED-
»- 975
9590
19640
Figure 13. Representative polychromatic/polypanchromatic
radar system for large space craft.
20
using densitometric data extracted from radar imagery are influenced by the
photographic parameters of the recording process.
Included in the study was the development of a relation for correcting
antenna pattern error effects in radar imagery. The success of such corrections is
dependent on sufficient knowledge of the system and an adequate dynamic range.
The pertinent relations are discussed in Technical Memorandum 177-2T.
2. 2. 2. 3 Test Synthetic Aperture Averaging
Delayed, to present Contract year. Preliminary report made in Technical
Report 177-7,
2. 2. 2. 4 Fresnel-Zone Plate Processor Test
In certain applications and under certain constraints imposed by spacecraft
or telemetary limitations, on“board and/or near real-time processing of synthetic
aperture radar data is required. One form of electronic processing was described in
Section 2.2.1 ,3c.
Another technique for such processing of synthetic aperture radar data is the
Fresnel zone plate processor. This processor is an analog/digital device which is
capable of processing synthetic aperture data to a resolution intermediate between
full-focussed synthetic and real apertures. It is less complex than is an electronic
version of the full "focussed processor that is now usually implemented optically.
During the period of this contract, the electronic model of the Fresnel zone-
plate processor was tested and found to provide a resolution close to that predicted
by theory. A description of the processor, test procedure and results are contained
in Technical Report 177-17.
It should be noted that the test model built at K ,U. is designed to process
the azimuth signal for one range bin only, and its implementation in a system is
somewhat dependent on the electronic state of the art. Merits of Fresnel zone-
plate processing were also investigated using the digital simulation.
21
PATTERN RECOGNITION ALGORITHMS
2.2,2. 1 Pattern Recognition Algorithms
The University of Kansas has taken a two-fold approach to the data processing
of remotely sensed imagery. Our approach has been based upon the need to have a
special purpose hardware facility for the near-real time processing of multi-image data
and the need to have a general purpose digital computer facility for the more sophisti-
cated non-real time processing. Cur near“real time facility is called IDECS (Image
discrimination Enhancement Combi nation System) and our non~real time facility is
called KANDIDATS (Kansas Digital Image Data System). These facilities have been
funded from both NASA and DOD sources. In this section we discuss KANDIDATS.
A Software System for Digital Image Processing: KANDIDATS
KANDIDATS (]<ansas DigitaIJmage Data System), currently being developed,
is a software package consisting of a monitor and a set of multi-image processing
programs designed to run on a GE-635 computer. The multi-image processing programs
are all written in FORTRAN IV and allow for image editing, registering, congruencing,
quantizing, clustering, feature extraction, image size and/or dimensionality reduction,
image texture analysis, and image pattern recognition. It has a variety of decision
rules, data display capability with scatterograms and histograms, grey-tone image
display with overprinting or digital image color map display. The KANDIDATS
monitor is a GMAP assembly ianguage program designed to integrate the multi-image
processing programs by handling all bookkeeping type and I/O operations and to
minimize the cost of processing image data by speeding up I/O time and overlapping
I/O time with execute time. Figure 14 illustrates a block diagram of the basic
KANDIDATS organization.
The KANDIDATS monitor inputs in free - format all instructions required by
the image processing program, supervises the execution of the programs, provides error
processing, and dynamic storage allocation and tape input and output for the programs.
The monitor has been written so that during a single activity of KANDIDATS many
processing programs may be sequentially executed using many different data sets.
The monitor does this by treating each program as a separate task and by allocating
and releasing data tapes as necessary.
22
BASIC KANDIDATS ORGANIZATION
Figure 14
Once remotely sensed data is converted to digital type format, it is necessary
to check the digitized tape to see if the conversion was made successfully. Preliminary
checking can be done by dumping the first few records on the tape; however, this is
by no means a complete check. The KANDIDATS image display program can make a
complste check by outputting the tape in picture format on the digital printer creating
the grey-tones by overprinting. If the image has so many resolution cells as to make
the digital picture printing awkward, a program may be utilized which reduces the
image size by averaging blocks of N x N resolutions or by selecting every N th column.
Examination of this picture output will indicate what kind of editing will have
to be done on the sides and top and bottom of the image, as well as indicate skewing
and A/D conversion distortion. (Skewing can occur because it may be impossible to
start digitizing each line of the image in exactly the same place. A/D conversion
distortion can occur when jitter or noise internal or external to the A/D conversion
makes the conversion go awry.) If necessary, a KANDIDATS deskewing program may
have to be used to remove skew and a special smoothing-replacement program may
have to be used for those resolution cells which were improperly converted.
When multi-image data is being processed, it is often necessary to align the
individual images to the same place. To do this KANDIDATS employs a registering
program. When different sensors or the same sensors with different look directions are
involved, it may be necessary to bring the images to the same geometry. In this case
a congruencing program must be used.
When the geometries on the images to be congruenced are quite different, the
congruencing job may be quite hard. However, where only minor geometric distortions
are involved, congruencing may be done by a KANDIDATS program which treats the
image as a rubber surface and expands or contracts it to best match up a set of given
corresponding points.
There are two formats by which multi-image data may be stored on tapes by
KANDIDATS. In the photo format all the grey-tones from the first image are stored on
a matrix followed by the grey-tones from the second image and so on. In the cor-
responding point format the grey-tone from image one resolution cell (1,1) is followed
by the grey-tone from image two resolution cell (1,1) and so on. Editing and con-
gruencing imagery from different sensors is usually done with data in photo form as
is image display and texture analysis. Most of the other programs work most easily
with the data in corresponding point format. KANDIDATS has programs which convert
multi-image data from one format to the other.
24
After initial editing and congruencing, it is convenient to obtain an intuitive
idea of what is happening in the data. To help with this, programs are available which
pick out specified regions on the image and display the data points in scatterogram or
histogram fashion. The scatterograms or histograms may be indexed by ground truth
categories when the ground truth is available. The axes of the scatterograms may be
combinations of pairs of the different sensor signals or the axes of a rotated coordinate
system. Rotation can be accomplished from principal component anejlysis or from
linear discriminant functions, and there are programs available for these operations.
Either of these operations will allow a significant reduction of dimensionality and,
therefore, allow a reduction in storage and display of data, especially in 12 or 24
channel multi - spectral scanner data.
Before pattern discrimination or clustering is done, a feature extraction is per”
formed which selects the relevant variables or which combines the original variables
in some optimum way. Sometimes as part of the feature extraction process quantizing
is done to normalize the data as well as to reduce the memory required for storage of
the data. KANDIDATS has available programs which do equal interval, equal
probability, minimum variance, and spatial quantizing.
When texture is an important feature for a category of interest, the dimen-
sionality of the images may be augmented by a texture analysis program which adds
dimensions providing texture type information.
Probably, the major workhorse of image data analysis consists of pattern dis”
crimination and clustering techniques. With pattern discrimination techniques, a
training set of data is gathered for which the correct category identification of each
distinct entity in the data is known. Then estimates are made of the required category
conditional probability distributions and a decision rule is determined from them. The
decision rule can then be employed to identify any other data set gathered under
similar conditions. With clustering techniques there is no training data set or decision
rule. Rather, the natural data structures are determined. Distinct structures are then
interpreted as corresponding to distinct objects or environmental processes.
The advantage of the discrimination techniques is that the scientist is able to
decide the types of environmental categories among which he wishes to distinguish.
The decision rule then determines as best as possible, to which environmental category
an arbitrary data entity belongs. The disadvantage of the discrimination techniques is
that they are sensitive to mis-calibrations . Any slight difference between the sensor
25
calibrations or state of environment for the training data and the new data will cause
error.
The advantage of the clustering techniques is that they are not sensitive to
calibration problems. Two smalharea patches of corn growing in the same field are
going to be detected as being similar because they have similar grey tone associated
with them. The disadvantage of the clustering techniques is that they are not able to
identify the distinct environmental structures they determine.
KANDIDATS has available iterative and chaining clustering programs and
pattern discrimination programs. The pattern discrimination programs use a variety
of decision rule types including a distribution-free Bayes rule which can only be used
on coarsely quantized data, a Bayes decision rule assuming the category conditional
probabilities are of some given type of multivariate distribution, a linear decision
rule, or a nearest neighbor decision rule.
Appendix A summarizes one of the things we are doing with a special purpose
KANDIDATS program for preprocessing of radar imagery.
26
2.2.2. 1 APPENDIX A: ENHANCEMENT AND NORMALIZATION OF RADAR
IMAGE TEXTURE
Texture has been of interest to engineers and geoscientists alike because of
its potential as a useful discriminant in image category identification. Hence, one
important preprocessing operation must be concerned with the enhancement and normali”
zation of image texture. Such an operation must bring out in normal form grey tone
variation due to texture and exclude grey tone variations due to look angles or flight
parameter fluctuations.
Antenna patterns and flight parameter fluctuations have been two factors most
responsible for degradation of radar imagery. If we regard the degradation as additive
noise, enhancement of the image would, in a sense, be appropriate if there were
means of removing the added noise. The 'streaks' parallel to the line of flight in an
image could be due to flight parameter fluctuations, scratches caused by handling of
the image before digitization, or due to antenna pattern. Perpendicular 'streaks'
could be due to scan lines.
Given below is a mathematical formulation, which in essence is the enhance-
ment technique .
Let Lx and Ly by the x and y spatial domains, G be the set of grey
tones and P:Lx x Ly — * G be some digital picture function of some more or less
'homogeneous' object 0,0 : Lx x Ly G ,
The relationship between P and O is assumed to be of the following form:
PO,j) = o{i,j) + 0(1) + e (j) (1)
where a(i) and 3 (j) can be thought of as additive row and column distortion
respectively. If we are interested in the texture of O, the average grey tone is not
important, and a function P(i,j) can be determined such that
P(U) = P(i,j) " a(i) - 60) (2)
where
I J
E E
i=i j=i
is minimized.
27
Hence, a(i)+SG) = "f" + V" ” IX (5)
A,
and the enhanced image P(i,j) is then obtained by substituting Equation (5) into
Equation (2) and
P0,j) = r<i,j) " ~r " t*- + tj- W
A,
The enhanced image P(i,j) is found to have a zero mean, and also each row
and column mean is zero.
Figure 15 shows a simulated 5x5 'homogeneous 1 image to which the enhancing
technique has been applied. The 5x5 image shown in (c) of Figure 15 is the model
with additive noise. The 5x5 enhanced image shown in (d) of Figure 15 clearly
shows a 'diffusion' of the additive noise. For simplicity of representation, image (d)
has been quantized, and therefore Joes not have a zero mean.
Figure 16 shows the enhancement technique applied to a radar image. Part (a)
shows a digitized radar image of a sorghum field. This field was isolated from the radar
image of a test site selected at Garden City, Kansas. The mission was conducted on
15 September 1965 by Westinghouse . Part (a) of Figure 16 shows a computer output of
the original . Streaks running vertically and horizontally show up very clearly on the
image. Part (b) shows the pictorial view of the 'noise' which was subtracted out of
(a). Part (c) shows the enhanced image. All three images are represented by 13 grey
tones and are quantized using an equal probability routine.
Figure 17 shows a larger area of the same test site and is made up of 14 fields.
The images shown in the figure are positioned the same, relative to one another, as
they were on the ground. Each image in the figure is a representation of the noise sub"
tracted out from it. The streaks occurring in one field carry on into the neighboring
fields. The vertical streaks (perpendicular to the line of flight) are almost periodic and,
as stated earlier in this section, could be due to scan lines* The horizontal streaks may
have been caused due to scratches on the negative or due to antenna pattern, but to
pinpoint their cause at this stage, without further research, would be difficult.
28
////// r/f/f/Y///s/Y//s/f
///*///////
//////
777777 * y / / / /
/ / S / f f t S f/S s
r/s//S f/////
//✓/// t///s/
// // /S f/s/f/
////// '/////
/////* /s/ // /
////////////
///Z//Z/Z//Z
////// //////^
//////////// s
//ft//"/'".
SS////////V'
/ff/ ///////*
/s//////////
/✓//////////
/S////////S'
////// /////>
////// f // s//
/ s * /*/ t //*//
<///// ff////
y*//// r//s/f
k///// ^/////
VfYS'f '/*///
///// *////*
zz/// v///z
zzzzz v/zz/
Z/ZZZ ?/////
/z/zz vzz/z
b) Additive Noise
(Row and Column)
c) Mode! Obtained after
Adding Noise
//zzzz
Vzzzzz
y ZZZZZ/
^z/zzzz
z/z/zz
zzzz/^z/z/z/zzzzz/ ZZ/Z/^ZZZZZZ
/zzzz/izzzzzyzzzz/zzzzz/jzzzzzz
zzzzzzi/zzzz^zzzzzz zzzzzzjz/Zzz/
ZZZZZitZZZZZziZZZZZZZZZZZyJZZZZZZ
z/z/z/zzzzzZ
zzzzzzzzzzzk
zzzzzzz/zzzk
ZZ/ZZZZZ/ZZ^
zzzzz/z/zz/zi
d) Image Obtained after
Enhancing Model
Shown in (c)
LEGEND
Figure is. Shows a simulated 5x5 homogeneous image with additive noise (row and column) and
its removal by the enhancement technique.
ORIGINAU PAGE 7E
OF POOR QUALITY
«s
!pc
8i
£>
cj rO
Direction of flight.
*
Figure 17. Part of the Garden Ci tv test site showing the
subtracted out of the 14 fields.
noise
sSljjji
IN
te'iiiiSiiife j
IMAGE DISCRIMINATION, ENHANCEMENT
AND COMBINATION SYSTEM (IDECS)
2. 2,2.2 Image Discrimination, Enhancement and Combination System (IDECS)
During the contract period the IDECS hardware was virtually redesigned and
rebuilt. Partial support for this development came additionally from a continuation of
the project THEMIS Contract DAAK 02 - 68"C“0089 and ETL Contract DAAK 02-71-C"
0482 with the Army. The IDECS was converted from an essentially analog device to
a hybrid system capable of full computer control by a PDP-15/20.
Early in the design phase it was suspected that the two major problems of system
noise (too much) and experiment repeatibility (too little), were related to the packaging
and grounding techniques previously employed. Therefore, after a close study, it
was decided that a major redesign of all IDECS subsystems be undertaken. As a result,
all of the subsystem elements are now independent, removable "black boxes" with
appropriate inputs and outputs capable of being selected and configured through a
central switching matrix. All of the new circuits which were added and virtually all of
those which were carried over with minor changes from the previous model, were made
on printed circuit boards. Finally, all of the subsystems were repackaged in a new
vertical rack assembly which insures better air flow for greater temperature stability.
The heart of the new digital circuitry is an IDECS Central Processing Unit (CPU),
which is interfaced to a PDP - 15/20 computer. The computer has 12 K of 1 8“ bit memory
plus four DEC Tape transports and the usual complement of small computer peripherals.
In the computer mode, the PDP~15/20 will gather image data through the IDECS,
calculate the associated statistics and generate an appropriate decision rule. The
IDECS will then implement the rule through the IDECS“CPU. Thus the computer" *
controlled CPU is able to direct data flow and processing anywhere in the IDECS system.
A detailed description of the digital circuitry and specifications can be found in "System
Hardware Specification Manual/IDECS," P. N. Anderson, CRES Technical Report 133 -
26, September 1971, and in "IDECS Hardware Reference Manual, Part I," P. N.
Anderson, CRES Technical Memorandum 133~34, December 1971.
Some twenty new analog circuits were added. The major changes included a
20 x 20 analog/digital switching matrix which allows a high degree of flexibility in
system configuration, a new signature selector which has greater bandwidth than the
earlier version and allows for automatic level selection via the IDECS - CPU, and a new
linear combiner circuit was designed for manual or computer manipulation of video
32
signals to fit an arbitrary linear combination. Additional new circuits include:
automatic discrimination and framer / composite operator's panel, video preamplifiers,
A-scan unit, and a color generating panel,
A detailed description of all the analog circuitry will be found in "IDECS
Hardware Reference Manual, Part II, " T. E. Polcyn and P.N. Anderson, CRES
Technical Memorandum 133-35, January 1972,
The third major element which was redesigned during this period was the flying
spot scanner package. Three synchronous flying spot scanners were developed which
are comparable with a wide range of film formats. The scanner assemblies were con-
structed to allow for rapid changing of images, and the associated circuitry enables
the operator easily to congruence and register images of different scales. The basic
scan mode is a raster, but a dot scan is available, and the scanner circuitry is inter-
faced to the computer so that an arbitrary computer generated scan is possible. Details
of the scanner circuits are found in the previously mentioned "IDECS Hardware Reference
Manual, Part I and Part 11." A description of the scanner hardware is found in "IDECS
Users Manual," J. C. Barr and P.N. Anderson, CRES Technical Report 177-27,
January 1972.
2.2,2.2a TEST IDECS APPLICATION
During most of the contract period the IDECS was unuseable while new circuitry
was being added and while the overall package was being rebuilt. Therefore, com-
prehensive experiments could not be performed until the end of the period when the
entire system was operable. With the new design it was possible for the first time to
repeat thematic mapping experiments at widely spaced intervals and receive identical
results. Two of the earlier problems of noise, which manifested itself in pictures which
seemed fuzzy or out of focus, and jitter, which was similar to a misthreaded movie
projector, are now gone, thereby simplifying the process of category definition by grey
tone selection on multi-imagery.
With the new signature selector, up to four video channels may be manipulated
in real time. The upper and lower levels of grey tone for each of the channels may be
independently set and the resultant selected signals combined with a logical AND.
The ANDed output signal is then available to the switching matrix for further manipula-
tion or display in any color. An experiment was conducted in which several people
33
independently constructed thematic maps from multi-image data, by using the signature
selector and the digitizing capabilities of the IDECS for category storage. These maps
were compared and the only significant differences were in the aesthetic preference
towards different colors for display.
In experiments with radar imagery it was found that the inconsistent grey level
for a single category from near to far range could be compensated by inputting the
video signal from one image into two of the signature selector channels. In this way,
one channel would be keyed to the category in the near range and the other channel
to the same category in the far range. By switching between the two the category was
adequately selected.
Multi-image category selection has been semi-automated through the use of
a scattergram program which will be described in Section 2.2.2.2c. A detailed
description of the techniques employed in category selection is to be found in "Image
Processing Applications — IDECS," P. N. Anderson, et al ., CRES Technical Report
177-28, January 1972. Other applications of the IDECS will be found in geoscience
subtasks associated with this contract and others.
2.2.2.2b IDECS INTERACTION HARDWARE
As mentioned in the introduction, the IDECS was interfaced to a PDP-15/20
computer, and an extensive Central Processing Unit (CPU) was designed and built to
direct the configuration of the IDECS subsystems. A twenty-four channel digital disc
is the central element of the current IDECS, providing the synchronization for the
sweep circuitry in the scanners, the monitors and the vidicon, as well as providing the
timing necessary for data transfers throughout the system. The most commonly used
algorithms between disc channels, e.g., logic AND, OR, EXCLUSIVE-OR, have
been built into the CPU as hardware elements.
The CPU is capable of interpreting 256 different instructions from the PDP.
The outputs of a set of storage registers in the CPU can be used as control signals for
some of the analog circuits, such as providing reference levels for the signature selector
A buffer memory associated with the disc has been incorporated to allow for data trans-
fers between the disc and the PDP . In this way the computer could configure the IDECS
monitor the results, calculate appropriate statistics and then reconfigure the IDECS to
display the "second level" results.
34
Another hardware task associated with this contract had to do with the flying
spot scanners. New circuits were designed and built which greatly simplified con -
gruencing and registering multi~images . The operation of this circuit is easily under -
stood by a description of the image handling procedure. A set of images is placed in
film holders and slipped into the scanners where they are individually focused and
enlarged to approximately the same size as viewed on a black and white television
monitor. One image is then used as a reference for the rest. One at a time the other
images are registered to the reference. One method of performing the operations is
to have the I DECS alternate rapidly between the reference and the other on the monitor
so they appear superimposed. Through the new hardware each image may be translated
horizontally or vertically, elongated or compressed horizontally or vertically and
rotated about an axis perpendicular to the center of the face of the screen. In this way,
exact superposition is possible, even if the film transparencies are of different sizes
or are not spatially true in one direction. An experienced operator can perform the
operation described above in five minutes or less.
Another hardware task covered by this contract had to do with construction of
new color generating circuitry. The IDECS will not normally be used for direct film -
den$ity _ to“hue conversion, but rather discrete colors will be arbitrarily assigned to
categories. Hence, an intrinsic problem occurred of determining the number of colors
that could be used without confusion. The human eye has a remarkable capability for
arranging subtle changes in hue into a continuous spectrum; however, very few people
are able to correctly identify the same color if it is separated by other colors or even
a solid background of another color. For example, small spots of yellow and biege on
a green background may be indistinguishable if random spots of other colors are also
present. Because of this problem considerable experimentation and research was done
before settling upon a set of ten colors (including white) for automatic encoding. The
circuit was designed for three modes of simultaneous operation: any of the twenty”
four disc channels may be displayed in any of the colors; any signal input to the
20 x 20 switching matrix may be displayed in any of the ten colors; and any signal
input to the matrix may be displayed in any color in a continuous spectrum from violet
to red.
Another circuit which was redesigned and built is a combined, framer/automatic
discriminator. This circuit is operated in the following manner. The framer is
switched through the matrix to the monochrome monitor. By adjusting the framer controls
35
the operator is able to define a white rectangle (or "frame") of arbitrary dimensions
upon a black background, and position it anywhere on the screen. The IDECS is then
configured to alternate rapidly between the frame and one of the video signals, so
that the frame appears superimposed and may be positioned over some category in the
image. The video signal is also fed via the matrix to the automatic discriminator.
This circuit examines that portion of the image which lies "under" the frame and
determines the maximum and minimum values of Intensity, i.e., grey level, present.
These values are applied as references for a circuit which gives an output only when
the video signal of the image lies within this min“max range. The output is available
to the matrix for display in any of the previously mentioned colors. As an example,
consider that the framer was positioned over, say, a wheat field in an air photo, and
the output of the automatic discriminator was displayed in yellow. Then on the color
monitor all fields which had the same range of grey levels as that under the framer
would be displayed in yellow. The outputs of both the signature selector and the
automatic discriminator are capable of being digitized, and stored on an arbitrary disc
channel by pressing a button on the composite operator's panel or by a command from
the computer.
Some typical configurations which are possible using the switching matrix are
given below to illustrate how the IDECS is used in practice.
Level Selection, Storage and Display.
Three video signals (S) are switched through the matrix (M) to the level selector (LS)
where grey level slicing is independently done for each signal with the outputs ANDed
together. The ANDed output from the level selector is then routed to the disc operating
36
panel (DO), where it may be directly displayed on the color monitor. After disc
storage (DC) the digitized image may also be displayed from the disc operating panel.
The output of the level selector is also available on the matrix where it can be routed
to either the monochrome or color monitors.
Automatic Discrimination and Display.
A video signal (S) and the framer (F) are routed via the matrix (M) to the composite
operating panel (OP) where they are alternately displayed or flickered (FL) on the
monochrome display (MD). The rapid alternation causes the frame to appear superim -
posed on the video image. The frame size is adjusted and the frame positioned over a
category of interest in the image.
The video signal is also routed through the matrix to the automatric discriminator
(AD) and then back through the matrix to the color display (CD). The framer circuitry
and the automatic discriminator circuitry work together so that the output of automatic
discriminator is only that part of the image which is within the range of grey levels
beneath the frame.
Automatic Discrimination/ Storage and Display.
37
This ts the same as the previous example except the output of the automatic discriminator
(AD) goes to the disc operating panel (CO) where it is stored on the disc (DC) and
the stored information is viewed on the color monitor (CM).
A complete description of the analog and digital circuitry is found in the
previously mentioned "IDECS Hardware Reference Manuals, Part I and II."
2.2.2.2c IDECS INTERACTION SOFTWARE
Two types of software have been written for the IDECS: diagnostic and
processing/control. The diagnostic programs are designed to test all of the subsystems
and locate any malfunctions. Within the CPU the matrix registers, the parameter
registers, the buffer memory, the circuitry for prograrrrcontrol led scans and the computer -
driven displays are configured and monitored to test for faults. It is also possible to
monitor the various supply voltages within the system and compare them to a reference
generated by the computer. A complete description of the diagnostic programs is found
in "IDECS Maintenance Manual," L. Haas and P, N. Anderson, CRES Technical
Memorandum 133-33, October 1971.
Processing/control programs relate to image data processing and control of the
IDECS. As mentioned above, through the 1DECS - CPU the PDP is able to control almost
all of the subsystems in the IDECS. For example, if several sets of images were to be
processed on the IDECS according to the same rules, it is possible to have the IDECS
repeat all the steps normally made by an operator in creation of a thematic map. That
is, the operator would manually manipulate the first set, and then the computer would
duplicate his actions for the remaining sets. The only steps the computer cannot
duplicate are those which obviously need human actions like positioning the framer
around a specific field or registering a set of images.
Processing programs have been written which analyze image data, calculate
associated statistics and use these statistics for IDECS control or display.
As an example, a two-dimensional scattergram program has the operator
position the framer over a field of interest, say, wheat, and then it determines the
range of grey levels within the frame for both images. This grey level data is stored
on a disc channel in a format which when displayed corresponds to having as axes the
grey levels of one image versus the grey levels of the other. The program then asks
the operator for another category and the process is repeated, with the scattergram
38
stored on a different disc channel. After all the categories have been selected, all
the calculated scattergrams are simultaneously displayed, each in a different color,
on the color monitor. Typically, the scattergrams will overlap somewhat. The program
then names each of the categories and directs the operator to position the framer over
that portion of the composite scattergram which he desires to define the category. In
this way the operator is able to make a decision as to which part of the overlapped
areas are to be called which category. The computer will read these values and then
direct the IDECS to level select each category and store the results on different disc
channels for thematic map display.
Other examples of programs which have been written include: generation of
histograms of the grey levels in single images; calculating the area of a selected
category in any type units, e.g., the total area in acres of a selected category is
calculated and printed on the teletype; and displaying any information which can be
stored in matrix form.
All of the programs have been written in a form which either directs the operator
via the teletype to do some task, e.g., "position the framer over such and such," or
else asks the operator simple questions, e.g., "what is the name of the next category?"
\ detailed description of these programs is found in the previously mentioned "Image
Processing Applications."
39
ALTIMETRY, SCATTEROMETRY AND
OCEANOGRAPHIC APPLICATIONS OF RADAR
2.3 Altimetry, Scatterometry and Oceanographic Applications of Radar
Various subsections of this task relating to scatterometry system analysis and
oceanography have been combined in a single section 2.3. 1,0. They represent a
completed study and will be the subject of a definitive report in the near future.
The substantial summary presented here is felt to be necessary in view of the importance
of the results and timeliness of the conclusions.
It should also be noted that all tasks relating to altimetry were deleted at the
the request of NASA/MSC. Additionally, no missions relating to the Gulf Coast
and Hurricane tasks were flown.
2.3. 1. 0 Scatterometry
The radar scatterometer is an instrument designed to measure the radar scat-
tering coefficient cr° (radar cross-section normalized to the illuminated area) as a
function of the illuminated incidence angle © (angle measured from vertical). The
purpose of these measurements is to determine the scattering signature data for various
types of terrain as well as the ocean surface and to determine the ability of radar to
discriminate between various terrain types. In addition, the <x° vs © data may be
used to aid in understanding radar imagery. The primary radar scatterometer used
in the NASA Earth Resources program operates at 73.3 GHz.
Eight missions were conducted over the North Atlantic ocean to investigate
the applicability of the 13.3 GHz radar scatterometer to measuring the instantaneous
wind field over the surface of ocean. The purpose of these missions was to determine
the relationship between radar scattering coefficients and ocean surface wind speed
and direction. Analysis of 13.3 GHz scatterometer data from the first few missions
indicated operational problems with the scatterometer. Thus a program of theoretical
analysis was begun to thoroughly study the operation of the system and thereby
recommend changes in order to improve the performance of the scatterometer. Of
particular importance was the study of the system calibration method, the signal
analysis of the receiver, the effect of non-linearities on the measurement data and
the interaction of navigational parameters on the system operation.
40
A complete signal analysis and computations of system sensitivity and data
measurement accuracy were made for the 13.3 GHz scatterometer. The effect of
phase errors on the data measurement was also considered. The effect on the backscatter
measurement of non-linearities and aircraft-flight parameters data were also taken
into account. An analysis of the calibration system showed that the absolute level
of the returns was not accurately known.
As a result of these analyses, the scatterometer performance was improved
and the data measured subsequently with the system could be more readily interpreted
and adjusted to compensate for the system operation.
The 13.3 GHz scatterometer system contains two vertically polarized phased”
array antennas, a CW transmitter and a two - channel homodyne receiver. The
receiver and transmitter antennas have a combined gain pattern that is narrow in the
direction normal to the flight direction (cross-track) and is wide in the direction
parallel to the flight direction (along-track). The transmitter continuously illuminates
an area corresponding to nominal angular dimensions of 3° in the cross~track and
+60° in the along - track directions. The energy incident at the receiving antenna
corresponds to the backscatter response of the surface illuminated. Because of air-
craft motion in the along-track plane, the radar returns at the illuminated incidence
angles are "coded JI by Doppler shift to frequencies displaced from the radar center
frequency. The return signal is immediately translated down in frequency with the
carrier shifting to zero frequency. The negative Doppler frequencies, corresponding
to aft information are folded onto the positive Doppler frequencies corresponding to
fore information. Separation of the information is achieved by using a two”channel
receiver with one channel in quadrature phase with respect to the other channel.
It is then possible to separate fore and aft information by appropriately summing and
differencing the outputs of the two channels. This operation is performed in the
digital data processing operation. During data processing, returns at the angles at
the angles between +60° are selected by selecting the appropriate frequencies.
The radar differential scattering coefficient c r° is calculated by solving the
classical radar equation for radar cross-section, then normalizing the cross-section
to the resolved surface area. The scattering signature for a specific area on the
surface is constructed by plotting <r° as a function of the incidence angle ©.
Since the cr° is measured at different times for different incidence angles, the time
must be "compressed 11 to yield a <r° vs © plot for a particular patch on the surface.
Figure 18 shows the system block diagram, including the data processing
technique used at MSC.
4?
SCATTEROMETER
REDOP 2
SLOTTED ARRAYS
DATA PROCESSING
Figure 18.
THE UNIVERSITY OF KANSAS
REMOTE SENSING LABORATORY
Scatterometer Sensitivity and Data Measurement Accuracy
It was shown that the scattering coefficient must be greater than ”24.4 dB
for a signal“fo”noise ratio of 10 dB at the receiver output, with aircraft altitude
3,000 ft, velocity of aircraft 180 kts, and incidence angle 60°, If the minimum
signal“to~noise ratio required for acceptable data quality is OdB, then the
minimum scattering coefficient that can be measured accurately is “34.4 dB.
In order to determine the signal”to~noise ratio required for an acceptable
measurement, the distribution of the mean power measured was considered for non”
homogeneous terrain (agriculture, sea ice). The number of independent scatterers for
the 0.3275 integration time of a single point is about 200, which gives the standard
deviation of the measurement as approximately 10% of the mean signal for signal”
to-noise ratio of 10 dB. The standard deviation increases to 16% and 22% of mean
signal received if the signa|-to“noise ratio is decreased to 3 dB and 0 dB, respectively.
Thus, scattering coefficients of less than ”34.4 dB can be measured with the scat”
terometer, but with a substantial increase in the standard deviation of the measure-
ment.
In the case of homogeneous terrain (ocean surface) measurements, the minimum
number of independent scatterers increases to about 800 due to longer time averaging.
The scatterometer can then measure scattering coefficients down to “40 dB with a
standard deviation of about 16% of the measured mean signal. Scattering coefficients
lower than this can be measured with a corresponding increase in the standard deviation
of the measured signal. The calibration signal recorded on tape is proportional to the
product of transmitter output power and total gain of the receiver. Fluctuations in the
operating parameters should be reflected by a corresponding fluctuation in the call"
bration signal .
Signal Analysis
The purpose of this exercise was to better understand the operation of the
several stages in the scatterometer. It was shown that the side bands appearing from
the interaction of the calibration signal and data signal are below the noise level and
therefore do not interfere with the data signal .
43
Phase Error
Analytical treatment of the passage of the signal through the dual quadrature
channels of the receiver and digital processor shows the insensitivity of the calculated
radar scattering coefficient to phase error introduced prior to the sign“sensing
operation in the data reduction procedure. It was shown that the total phase error
between the two Redop channels can be as high as 20° before seriously affecting the
calculation of scattering coefficient. The scattering coefficient error was found to
be less than 0.2 dB for a phase error as high as 20°.
Calibration System
The purpose of the calibration system in the 13.3 GHz radar scatterometer is
to provide a reference signal to permit compensating for variations in the system
parameters. This reference signal, when properly calibrated, provides an absolute
calibration of the value calculated for <x° . In the present 13. 3 GHz scatterometer
a sample of the transmitted power is amplitude modulated by a ferrite modulator and
the sidebands are used as reference signals. The ferrite modulator is driven by an
audio oscillator and operates as a variable attenuator utilizing controlled Farraday
rotation to attenuate the rf energy passing through the modulator. It was shown that
the voltage transfer response of the ferrite modulator as a function of the solenoid
voltage is linear in most of the region. The bias, or quiescent, point is established
by the remanent magnetism of the ferrite material. The amplitude of the modulation
sideband which becomes the calibration signal is critically dependent upon the
location of the quiescent point and thereby on the linearity of the transfer response.
Since the quiescent point is determined by the remanent magnetism of the ferrite
material, it can change as a result of changes in temperature, nearby magnetic fields,
vibration, or accidental direct currents. Thus, it was found that the present technique
does not provide a stable calibration of the scattering measurements.
A number of different calibration techniques were investigated and proposed.
The one recommended utilized a PIN diode modulator as a switching device to inject
the reference signal into the RF receiver circuitry. It was shown that the calculation
of a° depends only on the insertion gain of the PIN diode which is a stable constant.
This, then, would certainly offer a better calibration of the calculation of cr° .
44
Effect of Receiver Nonlinearity on Measurement Data
Agricultural missions conducted in summer of 1968 over Garden City, Kansas,
produced saturated data* An effort was then made to determine the effect of non -
linearity of the receiver and the recorder on calculation of <r° . Bradley showed
that the effect of nonlinear amplification of a multi-frequency input signal resulted
in intermodulation distortion within the Doppler frequency signal spectrum. A
third order power series with variable coefficients was used to approximate the
amplifier transfer function and the percentqge distortion at two points in the Doppler
signal spectrum was calculated. The greatest distortion, i.e., the greatest error
in the calculation of backscatter coefficients, occurred at the highest Doppler
frequency. A typical value of distortion at 6 KHz, the higher end of the Doppler
spectrum, is 5.14% which corresponds to an error of 0.44 dB in the calculation of
o
o- .
Aircraft Flight Parameters and Measurement Data
It was shown that the aircraft flight parameters have significant effect in the
measurement of backscatter coefficient from inhomogeneous terrain, notably sea ice
and agricultural terrain. In the measurement and analysis of backscatter coefficient
from sea ice and agricultural terrain, it is important, therefore, to assure that flight
parameters are maintained. The effect of drift, pitch, velocity and altitude variations
on the computation of scattering cross-section was considered.
2,3,2. 1 Ocean Wind Measurements
Measurements of the response of the radar scattering coefficient at 13.3 GHz
have been made using MSC aircraft since 1966. Four major missions were mounted
over the North Atlantic in 1968 (Mission 70), 1969 (Mission 88), 1970 (Mission 119,
JOSS I), and 1971 (Mission 156, JOSS II). In each case, the radar cross-section <r°
was observed to increase with increasing wind speed up to the highest winds observed.
Missions 70 and 88 were flown in the vicinity of weather ships permanently stationed
by various nations in the North Atlantic and North Sea; Mission 119 was flown near
the Argus Island "Texas Tower" near Bermuda, and Mission 156 was flown off the
U.S. east coast near the unmanned weather buoy XERB“1 . Because of the nature of
observations made routinely on the weather ships, the surface wind conditions during
Missions 70 and 88 are less accurately known than those during the later missions,
and the observations scatter more widely for the earlier missions as a consequence.
45
Improvements were made In the programs used for processing the scatterometer
data at MSC after the initial processing of Mission 70 and 88 data, and only the
Mission 88 data were reprocessed. Consequently the subsequent study of the data
has concentrated on Missions 88, 119, and 156. Because of the improved wind moni-
toring during Missions 119 and 156, particular emphasis was given to analyzing data
from these missions, but Mission 88 encountered 49 kt winds as contrasted with a
maximum of 33 kts for the later missions, so these data have also been included in
the analysis.
Absolute calibration of the returned radar signals is in question because of the
problem with the ferrite modulator calibration scheme used, as described above.
For this reason the analysis has concentrated on use of the ratio of the scattering
coefficient at the angle being studied to that at a reference angle, which has been
selected as 10°. This ratio is insensitive to changes in system gain and transmitter
output power, the two quantities the calibration system is intended to cover. The
kinds of parameter changes that might affect this ratio are (1) changes in the frequency
response of the receiver audio system, (2) changes in the antenna pattern, (3) errors
in measured aircraft speed that cause a given return frequency to be assigned to the
wrong angle, and (4) errors in aircraft pitch information that cause the wrong antenna
gain to be assigned to a particular angle. The first two quantities are not likely to
change over long periods of time, although a change in antenna pattern occurred
from Mission 1 19 to Mission 156 because the antenna was moved from the P3 to the
CI30 aircraft. All measurements made with the Doppler navigator for speed deter-
mination should be as accurate as the Doppler navigator itself; for a few runs this
instrument was inoperative, and the speed measurement may have larger error, which
was considered in the analysis. The amount of error introduced by erroneous readings
of aircraft pitch is not known .
Results of the observations with the three missions for incident angles of 25°
and 35° are summarized in Figures 19 and 20. Here the scattering coefficient ratios
for Mission 156 have been adjusted to the same scale used for the previous missions,
since they were different because of the different antenna pattern on the C130.
Clearly the wind speed dependence has been established as being significant, but
somewhat lower for crosswind than for upwind “downwind observations; hence,
knowledge of the direction of the wind is essential if a scatterometer is to be used
for anemometry .
Analysis of the variability of the Argus Island anemometer record as contrasted
46
UPWIND
with that of the radar return from nearby indicated that much of the fluctuation of
the radar return could be accounted for in terms of wind speed fluctuations, but that
an additional component of variability remained after these corrections. This
analysis is continuing.
The entire sea backscatter program has been a joint effort involving, in
addition to many NASA, Navy, and NOAA personnel, very close cooperation between
Professor Pierson's group at New York University and the Kansas group. Numerous
papers have been published separately and jointly by these groups. A major part of
the analysis of the Mission 1 19 and 156 data is contained in a Ph. D. dissertation by
G. A. Bradley, soon to be published as a technical report.
2. 3. 2. 4 Experiments (Agricultural Terrain)
Five scatterometer missions were flown in the period from September 1969 to
September 1972 over the Garden City, Kansas, test site. The purpose of the scat
terometer agricultural missions is to help determine the ability of a radar to identify
agricultural features such as crop type, crop vigor, soil moisture content, and crop
maturity throughout the agricultural growing season. The optimum radar parameters
(such as incidence angle, frequency, bandwidth, etc.) for discriminating agricultural
features will be determined using results from the scatterometer, spectrometer and
imaging radars.
Missions 130, 133 and 153 were conducted in May, June, and October of
1970 respectively, and Missions 165 and 168 in May and June of 1971 . Missions 130
and 133 contained data from both the 13.3 GHz and 400 MHz scatterometers, whereas
Missions 153, 165 and 168 had only the 13.3 GHz data. With the exception of
Mission 153, all scattering data have recently been received and analysis has just
begun. A summary of the missions and data is presented in Table I.
2. 3. 2. 5 Analysis of Sea Ice Observations
Ice Scatterometry
The ability of radar to discriminate different types of sea ice was demonstrated
by J. W. Rouse, Jr. in Technical Report 121-1 . It was shown that it is possible to
identify different categories of sea ice by the radar backscatter. A number of analyses
were carried out on the data obtained from Mission 47. The main effort was concentrated
on finding the “roughness factor," for each ice type based on the Kirchhoff“Huygens
principle. The results were encouraging but not definitive because of the limited
amount of data available.
49
Table I,
Site 76 (Garden City, Kansas) Scatterometer Inventory
Date
Received
Mx
#
Date of
Mx
Frequency
Data
Copy
Line
Run
1 I/I 8/71
130
5/70
400 MHz
<j° time history
2
1-6
1
Reflect plots
2
1-6
1
PSD plots
2
2-5
1
3/9/72
Output tape
1
1-6
1
13.3 GHz
Dump of first & last ten
records of each file
1
1-6
1
Reflect plots
1
1-5
1
PSD plots, filter tab
1
1-5
1
c t time history plot
I
1-5
1
Output tape (6 files)
1
1-6
1
11/18/71
133
6/70
400 MHz
<j° time history plot
2
1-6
1
Reflect plot
2
1-6
1
PSD plots
2
1-6
1
3/9/72
Output tape (6 files)
1
1-6
1
13,3 GHz
cr° time history plot
1
1-6
1
PSD plots, filter plots
1
1-6
1
Reflect plots
1
1-6
1
Output tape
1
1-6
1
153
10/70
13,3 GHz
Not received
2/16/72
165
5/71
13.3 GHz
Output tape
1
1-6
1
Filter plots & tab,
PSD plots
1
1-6
1
cr time history plot
1
1-6
I
Reflect plots
2
1-6
1
2/16/72
168
6/71
13.3 GHz
Output tape & format
1
1-6
1
<t time history plot
I
1-6
1
Reflect plots
1
1-6
1
PSD plots, filter plots
I
1-6
1
50
Interpretation for backscatter radar returns were presented in TM 185"! in
terms of surface roughness, volume scatter, effective conductivity, and relative
dielectric constant of the scattering media. These interpretations were discussed with
respect to statistical analysis of 13.3 GHz backscatter of various categories of sea ice
at different angles of incidence. Results obtained from the statistical analysis carried
out on the data obtained from Mission 47 indicated that it is possible to identify both
water or thin ice and multi-year ice from first-year ice without any misidentifications
using angles of incidence greater than 15°. Further, it seemed possible to identify
a few additional ice categories within the first-year ice group.
The area coverage requirement of a given sea ice category was described in
terms of aircraft operating altitude, speed, antenna patterns, and signal frequency.
This description also covered the minimum terrain area necessary for discrimination
against other sea ice categories.
The presence of volume scattering was considered to be a factor, but further
studies were recommended to study its effect.
The statistical analysis of Mission 47 data was carried out using pattern
recognition techniques. The purpose of this was to determine whether it was at all
reasonable to expect that certain basic ice types could be differentiated on the basis
of their 2.5 cm radar~backscattered return profiles. Each return profile (cr° vs ©
curve) contained values of <r° at incidence angles of 2.5° / 3°, 7 , 15°, 25 ,
35°, 45°, 50°, and 65°. From the Arctic mission, there were 363 backscattered
profiles, each taken of a small area (30 m x 30 m) ground patch of an ice type
reliably observed by ground or air. Initially, the ice identifications were divided
into three major ice categories: water or thin ice, multi "year ice, and first-year ice.
If ice types could be distinguished on the basis of their radar backscatter profiles,
certainly these major groups would be distinguished.
One approach (and an optimum one at that) to the discrimination problem
involves the use of a simple Bayes decision rule. Such a rule assigns the cross-section
to the most likely ice categories. This approach was used here to discriminate
ice types.
From the data sample of 363 measurements, 195 measurement profiles were
selected at random for the training set. Assignments of ice categories were made on
the basis of photo interpretation and comments during flight by a U.S. Navy ice
observer. The prediction set consisted of the remaining 168 measurements. Each
measurement was assigned an ice category by the constructed decision rule. Table
51
Table II.
Contingency table for prediction set using all 9
angles of scatterometer data. Rows are true category identi-
fication, and columns are category identifications assigned
by a Bayes decision rule based upon multivariate normal
conditional distribution.
Water, Thin Ice
IsfYear Ice
Multi-Year Ice
Water, Thin Ice
5
3
1
lst“Year Ice
4
71
6
Multi-Year Ice
0
0
78
II illustrates the resulting contingency table of true ice identifications versus
assigned ice identifications.
Other analyses were conducted using different training sets. A detailed
list of the results obtained is given in TM T85“l .
The current research effort is concerned with analysis of Mission 126
(see 2. 3. 2. 4). A detailed discussion of the objectives and location of the mission
is given in TM I77“14. The experimental data included multi“polarization 400 MHz
scatterometry, vertical polarization 13.3 GHz scatterometry, dual polarization
16.5 GHz imagery and aerial photography.
The preliminary analysis of the data has only permitted qualitative conclusions.
The real value of the data can only be demonstrated when a detailed analyses of the
data is carried out. Figures 21, 22 and 23 illustrate some of the preliminary results.
Figures 21 and 22show the spread of the average scattering coefficient for
nine different ice types as a function of angle of incidence. Figure 23 shows this con-
verted into dynamic range requirements for the ice imager. Clearly an imager
operating at steep angles of incidence would need dynamic range from about “12 to
+ 15 dB; whereas an imager whose closest angle of incidence is only about 35° can
get by with a dynamic range of considerably less than 20 dB as shown, but must have a
sensitivity of at least “14 dB, Analysis of the data along these lines is continuing.
52
Incidence Angle (Degrees)
Figure 21. Means for 9 Ice Types Observed on Mission 126.
53
1. OPEN WATER .
2. NEW ICE . A GENERAL TERM FOR RECENTLY FORMED ICE WHICH
INCLUDES FRAZIL ICE, GREASE ICE, SLUSH, & SHUGA.
3. NILAS. A THIN ELASTIC CRUST OF ICE HAVING A MATTE
SURFACE AND A THICKNESS OF UP TO 10 CMS.
4. YOUNG ICE (COMPACT PACK) . ICE IN THE TRANSITION STAGE
BETWEEN NILAS AND FIRST YEAR ICE, 10-30 CMS IN THICKNESS,
AND HAVING A CONCENTRATION OF 8 / g . NO WATER IS VISIBLE.
5. YOUNG ICE (VERY CLOSE PACK) . SAME AS ABOVE BUT WITH A
CONCENTRATION OF 7/ g TO LESS THAN 8 / g .
6. FIRST YEAR ICE (COMPACT PACK). ICE OF NOT MORE THAN
ONE WINTER'S GROWTH HAVING A THICKNESS OF 30 CM TO 2M.
CONCENTRATION 8 /g •
7. SECOND YEAR ICE (COMPACT PACK) . ICE WHICH HAS SURVIVED
ONLY ONE SUMMER'S MELT. CONCENTRATION 8 / g .
8. MULTI-YEAR ICE (COMPACT PACK). ICE WHICH HAS SURVIVED
MORE THAN ONE SUMMER’S MELT. THICKNESS UP TO 3 M.
CONCENTRATION 8 / g .
9. MULTI-YE AR ICE (VERY CLOSE PACK). SAME AS ABOVE.
CONCENTRATION 7/ g TO LESS THAN~5/g .
Figure 22, Classification of Sea Ice for Figure 21.
54
2. 3. 2, 6 Mission 126
Technical/scientific support was provided for Mission 126 of the NASA/MSC
aircraft to Pt. Barrow, Alaska, for a sea ice experiment. This included assistance in
arranging the flight plan and ground support both in advance of and during the mission.
Mission 126 was conducted in April 1970, with Dr. A. W. Biggs, Dr. G. A. Bradley
and Mr. R. L. Walters acting as scientific observers on the experiment team.
56
RADAR GEOLOGY
2.4 Radar Geology
The utilization of imaging radars in geologic research must be preceeded by
the development of an understanding of both system and surface parameters. This has
been the stated goal of the geologic research group since its organization and partic-
ipation in the program for radar studies related to earth resources. Briefly outlined,
our objectives have been:
1. System Parameter Evaluation
Look Direction
Frequency
Depression Angle
Polarization
Resolution
2. Target Parameter Evaluation
Moisture
Vegetation
Slope and Relief
3. Applicability
Geologic Structure
Geologic Materials
Geomorphology
Our research objectives under contract NAS 9-10261 have to some degree
been modified because of modifications in the flight and data acquisition programs.
However, whereas this has resulted in de-emphasis in some areas of investigation,
it has also resulted in more intensive investigation in other areas. With a sparsity
of new imagery, one becomes increasingly aware of the vast amount of existing
imagery in which still may be locked the solution to many problems of interpretation.
57
RESULTS OF INVESTIGATIONS
Geologic Cross~Polarized Anomalies
A prime example of the utilization of existing imagery is seen in the
investigation of cross-polarized anomalies in selected areas of volcanics and
sandstones in the western United States (McCauley, 1972). Utilizing imagery
over essentially arid areas In which the time lapse between imaging and field
study would be least significant, it was determined that surface roughness, not
rock type, was the key factor in the development of tonal reversals in the cross-
polarized imagery evaluated.
Outcrops of these rock types share certain features; planar rock surfaces that
are targe In comparison with the wavelength of the incident radar are abundant and
detrital material and vegetation are of secondary importance; the planar surfaces
appear to significantly contribute to the returning radar energy with this energy
maintaining a constant polarization; and the outcrop areas are of sufficient size and
sufficiently uniform character to be delineated on small scale K-band imagery.
These rock types, for differing reasons, produce terrains in which radar return
is dominated by specular reflection from planar surfaces. For a specular reflector to
be recorded on the radar image, its orientation should be orthogonal to the path of
the impinging radar; and for such an orientation, the depolarized component of the
reflected radar energy is at a minimum. The result would be a higher return on the
like-polarized image and a lower return on the cross-image. This is in compliance
with the observed behavior of the volcanics and sandstones evaluated.
Look Direction Significance
The need was early recognized for a study of the effect of look direction. In
fact,as a result of a study of imagery in the Ouachita Mountains of Arkansas (Dellwig,
et al . 1966) multiple looks from diverse diverse directions were requested and obtained
during the NASA imaging program utilizing the AN/APQ-97 radar. Conclusions
reached regarding the relationship between the look-direction and detectability of
geologic features in this environment conflicted with the conclusions reached by
MacDonald in his evaluation of Darien Province, Panama imagery (MacDonald, 1969).
58
Evaluation of NASA flight program imagery from additional areas in which multiple
coverage was available provided sufficient data to reconcile the differences and
provide a better understanding of the effect of look-direction in diverse environments
(MacDonald, et al., 1970).
It was apparent that look-direction did indeed influence the detectability
of geologic features expressed in the terrain configuration. Depending on the relative
topographic relief, effective incident angle of radar energy, and look-direction,
geologic features would be advantageously enhanced or completely suppressed. Thus,
to obtain the maximum benefit from geological reconnaissance studies utilizing radar
in poorly mapped areas, it would be desirable to image the specific region from four
orthogonal look-directions. For more detailed studies where a known terrain con-
figuration would be imaged, a minimum of two opposing look-directions would suffice
for optimum geologic interpretation.
Depression Angle
Shadow enhancement of terrain configuration has proved to be a reasonably
adequate compensatory mechanism for the lack of stereoscopic coverage. As useful
as shadowing is in low relief terrain, it is equally undesirable in areas of extreme
high relief. Recognizing the need for consideration of depression angle variation
relative to relief, MacDonald and Waite (1971) evaluated terrain relief on a world-
wide basis and calculated the optimum depression angles for various terrain categories.
In low relief areas, the oblique illumination and resultant shadowing by
imaging radars can generally provide enhancement of topographically expressed geo-
logical features, but in mountainous terrain , radar shadowing can deter geological
interpretation. Especially in rugged terrain, two inherent disadvantages of a radar
imagery format which can hinder geologic interpretation are extensive shadowing
and layover. Radar depression angle and terrain slope define the range over which
shadow and layover will occur, but the extent of either parameter is defined by relative
relief. For most operational side- looking radar systems, the interpretative data loss
increases as terrain slopes exceed 35 degrees and local relief surpasses 1000 meters.
Trade-offs between loss of geologic data due to radar shadow and layover, versus
swath-coverage, have been evaluated. Similarly, the advantage of slight radar
shadowing in low relief areas is considered. Near and far range depression angles
have been recommended according to five global terrain categories, and imaging
altitudes are considered for both airborne and spaceborne platforms.
59
Resolution
The comparative evaluation of coarse and fine resolution radar imagery was
first realized with the acquisition of the University of Michigaris high resolution
imagery in the Ouachita Mountains, an area previously imaged by several coarser
resolution systems (Dellwig and McCauley, 1971).
In general the value of increased detail is offset by the distraction of minor
topographic and vegetal features and the loss of the distinctive pattern of major
features. Although judged to be real, this undesirable feature may be in part due to
the loss of the synoptic presentation with the decrease in swath width. As has been
demonstrated on numerous occasions in the past, the synoptic presentation and the all
weather capability of radar are its major advantages. In a relatively heavily vegetated
area, ihe increased resolution provides distracting detail and the decrease in swath
width compounds this loss; thus only the all weather capability remains as the capability
not offered by the aerial photography. The conclusions reached in this study, although
possibly adaptable to other terrain environments are only preliminary in nature and await
verification through evaluation of terrain in other environments.
Frequency
The determination of the value of imagery of other than K-and X-bands for
geologic investigations continues to be a significant problem in need of further
investigation. A preliminary study in the Pisgah Crater, California area (Dellwig,
1969) pointed out the need of controlled comparisons and a nearly completed similar
study of imagery of diverse sources and acquisition dates in the Florida Panhandle
emphasizes the need. Based on a bare minimum of data, it appears that simultaneous
imaging with high and low frequency radars will be of greater value to geomorphologists,
hydrologists, and geographers, than to geologists.
TARGET PARAMETER EVALUATION
Moisture
Unfortunately a great deal of the present imagery was produced without the
benefit of simultaneous acquisition of ground truth data. Whereas studies in areas
such as Pisgah Crater or Mono Craters, California are not so dependent on simultaneously
acquired surface data (because of lack of vegetation and aridity), investigation of
60
the effect of vegetation and soil moisture on the return signal are hampered in
more humid environments when ground truth acquisition and imaging are widely
spaced in time. Fortunately some more recent data were obtained in the Gulf
Coast area which demonstrated a correlation between results obtained from a
controlled measuring device and ground conditions (MacDonald and Waite, 1971).
The data presented in that study suggest that presently available dual
polarized, K~band, side-looking imaging radars provide a capability for revealing a
qualitative estimate of soil moisture content. When used as a supplement to aerial
photography in temperate climates, radar imagery analysis will decrease the ambiguity
of soil type reconnaissance. In the Arctic, an imaging radar may provide data for
mapping regions of permafrost, and this process could be accomplished in a sequential
manner regardless of weather or time of day. The use of additional multifrequency
multipolarization imaging radars and the relative foliage penetration of each should
be investigated as a possible means of gathering quantitative soil moisture information.
Vegetation Penetration
Utilizing the same Gulf Coast imagery, a closely allied study concerning
vegetation penetration provided valuable data for the user community (MacDonald and
Waite, 1971). Although limited in scope, some progress was indicated in the under-
standing and separation of effects of soil moisture and vegetation. In this study it
was suggested that within specific terrain environments, the penetrative effects of Ka-
band radar energy can be recognized on the imagery format. Boundaries were defined
on the radar imagery that are directly related to differences in soil moisture content,
irrespective of either the vegetation type or density. Use was made of the depolarized
return signal to distinguish between differences due to soil moisture and gross vegetation
differences, thus providing a practical advantage for a multipolarization mode.
Appl i cat ions
Studies aimed primarly at evaluation of system and terrain parameters necessarily
pointed out the potential of radar as a device for geological investigation. In addition
to these studies, other reports dealt with the utilization of radar in a variety of investi-
gations in diverse terrains (Kirk, 1970; MacDonald and Waite, 1972; Wing and Dellwig,
1970; Wing, et al., 1970). The best demonstrated utilization appears to be in the area
61
of fracture detection and analysis, regardless of the degree of development of masking
vegetative cover, relief, or overall structural complexity.
With the completion of these studies, documentation of radar as an operational
tool for geologic investigation has been realized. This to more than a limited degree
has been substantiated by the development and widespread utilization of three
commercial radar systems.
This does not imply that the evaluation of SLAR as a geoscience tool is complete.
We visualize that with further investigation as outlined below, the full capability and
limitations of radar will be more precisely defined.
1 . Multifrequency imaging for geologic and geomorphic studies.
2. Varying combinations of dual polarization imaging in geologic,
hydrologic, and geomorphic studies.
3. Radar as a detector of soil moisture, swamps, ice and permafrost.
4. Resolution ranges for geologic and geomorphic studies for a wide
range of map scales.
5. Radar imagery data content as compared with high altitude and orbital
photography.
62
Task 2 .4 APPENDIX A: References
Dellwig, L. F., J. N. Kirk and R. L. Walters, 1966, "The Potential of Low
Resolution Radar Imagery in Regional Geologic Studies," Jour. Geophys.
Res., vol. 71, pp. 4995-4998. ~ “ ~ ~
Dellwig, L. F., 1969, "An Evaluation of Multifrequency Radar Imagery of the
Pisgah Crater Area, California," Modern Geology , vol. 1, pp. 65-73.
Dellwig, L. F. and J. McCauley, 1971, "Evaluation of High Resolution X-Band
Radar in the Ouachita Mountains," CRES Technical Report 177-21, University
of Kansas Center for Research, Inc., Lawrence, Kansas.
Kirk, J. N., 1970, "A Regional Study of Radar Lineaments Patterns in the Ouachita
Mountains, McAlester Basin- Arkansas Valley, and Ozark Regions of Oklahoma
and Arkansas," CRES Technical Report 177-4, University of Kansas Center for
Research, Inc., Lawrence, Kansas.
MacDonald, H.C., 1969, "Geologic Evaluation of Radar Imagery for Darien
Province," CRES Technical Report 133-6, University of Kansas Center
for Research, Inc., Lawrence, Kansas.
MacDonald, H . C . , J . N . Kirk, L. F . Dellwig and A. J . Lewis, 1970, 'The
Influence of Radar Look-Direction on the Detection of Selected Geological
Features , " Proc. Sixth Symposium on Remote Sensing of Environment,
University of Michigan, Ann Arbor, pp. 637-650.
MacDonald, H. C. and W, P. Waite, 1970, "Optimum Radar Depression Angles
for Geological Analysis," CRES Technical Report 177-9, University of
Kansas Center for Research, Inc., Lawrence, Kansas.
MacDonald, H. C. andW. P. Waite, 1971, "Soil Moisture Detection with Imaging
Radars," Water Resources Research, vol, 7, no. 1, pp. 100-110.
MacDonald, H. C. and W. P. Waite, 1972, "Terrain Roughness and Surface Materials
Discrimination with SLAR in Arid Environments," CRES Technical Report
177-25, University of Kansas Center for Research, Inc,, Lawrence, Kansas.
McCauley, J.,1972, "Surface Configuration as an Explanation for Lithology-Related
Cross-Polarized Radar Image Anomalies," Earth Resource Program 4th Annual
Program Review, January 1972, NASA Manned Spacecraft Center, Houston,
Texas.
Waite, W. P. and H. C. MacDonald, 1971, "Vegetation Penetration with K-3and
Imaging Radars," IEEE Transactions on Geoscience Electronics, vol. GE-9,
no. 3, July 1971, pp. 147-155. “ ~ ~
Wing, R. S. and L. F. Dellwig, 1970, "Radar Expression of Virginia Dale Precambrian
Ring-Dike Complex, Wyoming/Colorado," Geol , Soc ♦ America Bui I . , vol . 81,
pp. 293-298.
63
RADAR APPLICATIONS IN AGRICULTURE / FORESTRY
Table of Contenfs
General Introduction 65
Conclusions and Recommendations for Tasks 2.5.2 and 2.5.3 65
System Preferences for an Airborne Agricultural Imaging Radar 68
An Information Dissemination System for Agriculture 69
2.5.2 Radar Sensing In Agriculture: An Overview 82
2.5.2. 1 Local Level Agricultural Practices and Individual
Farmer Needs as Influences on SLAR Imagery Data
Collection 89
2. 5. 2. 2 Image Interpretation Keys to Support Analysis of
SLAR Imagery 96
2. 5. 2. 3 Basic Parametric Studies: The "Standard Farm"
Design Philosophy and Initial Results 105
2. 5. 2. 4 Remote Determination of Soil Texture and Moisture
Using Active Microwave Sensors 113
2.5.3 Radar for Vegetation Studies: An Overview 11/
2.5.3. 1 Vegetation Mapping with Side“Looking Airborne
Radar: Yellowstone Park 123
2. 5. 3. 2 SLAR Imagery for Evaluating Wildland Vegetation
Resources 134
2. 5. 3. 3 The Potential of Radar for Small Scale Land Use
Mapping 139
Appendices
Appendix A: Example of a Dichotomous Key Algorithm 142
Appendix B: Differential Scattering Cross~Section (cv^ © in
dB) for a Silty Clay Loam Soil under Varying
Roughness and Moisture Conditions 152
Appendix C: Crop Signatures fora "Typical Standard Farm" 158
Appendix D: References 159
64
General Introduction
The following report summarizes Geoscience research efforts under the
Agriculture /Forestry tasks of Contract NAS 9~1 0261 (tasks 2,5.2 and 2.5.3). Sub”
tasks under each of these are keyed specifically to work proposed under Technical and
Cost Proposal BB321“55”0”60”P dated June 1970. The report has been subdivided into
three parts: 1. General Introduction, System Design and Recommendations; II. Sum”
mary of Research on Tasks 2,5.2 and 2.5.3, and III. Appendices and References,
In Part 1. we outline our current estimate of requirements for an operational
agricultural radar and for the utilization of its data in an information system. A list
of general conclusions and recommendations is presented as derived from the subtasks.
Part II opens with an overview of the use of radar imaging In Agriculture
(2.5.2) and is followed by a series of brief subtask reports (2.5.2.1-4). Each of
these is a condensed version of technical reports and/or published documents. The
interested reader is referred to them for more details. Following the subtask reports
for Agriculture, a summary for Forestry (task 2.5.3) is given. An overview for resource
mapping in general is presented for the specific case of Arid Zones and this is followed
by subtask reports 2.5.3. l - 3.
Finally, in Part III we present sample data and results from studies reported in
the subtasks. The amount of data and computer printouts is so vast that only a sample
Is feasible .
Conclusions and Recommendations for Tasks 2.5.2 and 2.5.3
Task 2.5.2
(a) Multi-frequency radar imagery will be required for accurate crop identi-
fication. The wider use of imagery is presently hindered by (1) the need for parametric
information on cr°, incidence , polarization and the effects of terrain variations;
and (2) the development of interpretation techniques.
(b) Using automated interpretation keys, radar data hold great promise in
providing early crop statistical estimates. Timely and efficient interpretation through-
out the growing season could provide statistics such as acres in production, acres
harvested, progress of harvest, etc.
(c) We recommend systematic cr° vs © investigations on crops in various
stages of growth, under varying kinds of stress, etc. , to begin unraveling within-
category changes in image appearance.
65
Subtask 2.5.2. 1
(a) It is axiomatic that every time a farmer performs an operation in his field
the radar ground echo will change. It is small wonder that interpretation anomalies
and ambiguities result considering the mutual interactions of farmer operations and
physical variables. Nevertheless, experienced interpreters can utilize information
contained in the radar image to make accurate determinations.
(b) The complexities of land use changes and the unique kinds of agricultural
information requested in interviews with farmers strongly suggest that initial inter-
pretation of radar data should take place at the county level. Interpretation tech-
niques, such as keys, seem the most advantageous for satisfying information needs,
especially when combined with the county agents' in-depth knowledge of local
conditions and trends.
Subtask 2. 5. 2. 2
(a) Visual image attributes can be effectively used to prepare dichotomous
keys to aid interpretation of radar imagery. A variety of single purpose keys can
be produced by concentrating on different attributes. When these are automated and
combined with ADP techniques, rapid and consistent information extraction is possible.
(b) Interpreter tests of a series of preliminary keys yielded crop identification
accuracies of slightly less than 50%. Though these are unacceptable in practice,the
tests showed where the keys could be improved.
(c) Results from preliminary tests of automated keys are encouraging but
inconclusive. Correct identification is on the order of 50%.
Subtask 2. 5. 2. 3
(a) Row direction appears to be a major factor in the reflectance of crops
such as corn, grain sorghum, and sugar beets. By extension, row direction may be
an important parameter in identifying all such row crops. The data hint that there
is more angular dependence in the HH signal return from crops with rows orthogonal to
the look direction than in those with rows parallel to the antenna.
(b) Return is influenced by polarization in that like polarizations tend to be
brighter than cross polarizations under similar gain conditions. Bare ground is an
exception to this statement.
(c) In the case of available systems (X- and K - bands) range variations for
non -row crops does not appear to effect radar return significantly. Similar means and
standard deviations are characteristic of near, mid and far ranges for wheat and pas-
ture. However, the radar returns for corn, sorghum and sugar beets appear to be
angularly dependent.
66
Subtask 2. 5. 2. 4
(a) For soils, variation in backscatter appears to increase at steeper depression
cngles and this variation is believed to be due largely to effects of moisture.
(b) Shallow depression angles appear to be most useful in making distinctions
in soil roughness, although vegetation effects may cause confusion.
(c) It is very likely that cross-over points in radar backscatter, arising from
combined effects of moisture and roughness,ex!st in the frequency and angular domains
such that a moist fine textured soil "looks" identical to a dry, coarse textured soil —
all other parameters equal.
(d) Our studies suggest that the generally quoted value of VlO f° r a smooth
surface needs re-evaluation for angles greater than 30°. We recommend modeling
studies to simulate the effects of soil texture under uniform conditions of moisture,
slope and vegetation.
(e) The unique potential for sub-surface soil information from long wavelength
systems should be evaluated along with panchromatic imagers.
Subtasks 2.5.3. 1 and 2. 5. 3. 2
(a) Interpretation of Ka-band imagery from Yellowstone Park Wyoming,
Horsefly Mountain, Oregon, and Buck's Lake, California all reveal that regional,
physiognomic vegetation mapping is possible with SLAR — especially in regions with
markedly contrasting plant community structures.
(b) For vegetation analysis from SLAR, best results are obtained in regions
where little information is lost due to shadowing. In mountainous areas better interpre-
tation is possible when the flight line is orthogonal to the "grain" of the mountains.
(c) Recommendations for future vegetation data collection include: 1)
multiple look directions; 2) 40% sidelap between passes.
(d) Frequency analysis of 50' resolution radar imagery is expected to yield in-
formation on plant community structure enabling identification of broad regional types.
(e) Estimates of timber volume seem impractical with coarse resolution imagery
except insofar as gross values may be useful in highly remote areas. The possibility
for calculating usable timber volume estimates from fine resolution, poly frequency
imagery should be explored for use in monitoring the growth progress of large planta-
tions, etc. The strategy may be very similar to that proposed for crop statistical esti-
mates, i.e., average volume/acre X number of acres as calculated from imagery.
67
System Preferences for an Airborne Agricultural Imaging Radar
The studies which follow this section address various aspects of the type
of imaging presently considered necessary to obtain agricultural data. Any active
microwave sensor designed for an image output should approach the parameters given
in Table 1 .
Table 1
Parameters of a SLAR
for Agricultural Data Collection
frequency
resolution
Ka, X, L* (ultimately we suggest a broadband
system in the 4-16 GH region)
1 - 5m
polarization HH/HV/VV
geometry of output image ground range display
dynamic range expanded in low medium signal return, sliced
in very low-high signal return.
depression angle range limited to 10° at approximately 30 to
40° depression.
*L, although only limited work has been done, this may be important
in collection of soil information.
The final image is a result not only of the SLAR system but the aircraft
planning constraints. Our estimate, based on the following research, has shown that
flight planning considerations may be as critical as systems parameters. These flight
planning parameters include (1) the temporal sequencing of flights; (2) the time of
day of acquisition; (3) the number and direction of flight lines; and (4) the con~
sistency of equipment operation. Table 2 summarizes some preliminary findings
related to flight planning.
68
Table 2
Time interval of missions:
Diurnal frame:
Mission coverage:
Side lap:
Set operation:
The above considerations are derived from analysis of available imagery,
all of which is SLAR. We have no spaceborne data on which to approximate radar
design parameters for Agriculture. Nevertheless, we feel that on a cost/benefit
basis, a satellite system would be most appropriate. It is increasingly apparent that
in order to chart future research efforts, a mechanism for information extraction and
dissemination from microwave imagery is needed. The types of information obtainable
for various levels in the decision-making hierarchy, the timeliness of the information
extraction and the interpretation approach are all uniquely dependent on the system
used to disseminate knowledge. For example, a federal or state agency monitoring
the agricultural status of any given county seems hardly worthwhile (and may not even
be relevant) if the results are not made immediately available to that county. At the
other extreme, monitoring by local county agents may be inefficient, if their immediate
needs differ from information required at the state or national levels. The purpose in
this next section is to propose and discuss one strategy for an information system.
1 .2 An Information System for Agriculture
So much has been written about "information systems" for remote sensing
that clarification of its use in the present context is essential. Holmes (1968), Fu,
et al.(1969). Holmes and MacDonald (1969) and Langley,et al.(1970) have all des-
cribed segments of an "information system" for agriculture based on multispectral
scanner and photographic data, its manipulation and interpretation. These reports
20 to 30 days
nearly identical time of day from mission to
mission. Pre-noon to minimize effects
of turbulence.
multiple look directions
multiple coverage; 40% for extensive area
analysis.
as nearly consistent as possible from mission to
mission; data should include flight
information on roll, pitch, yaw, etc.
69
all concentrate on the means for converting data to information, but they fail to
suggest where in the management hierarchy such conversion should take place or how
the resulting information should be used. As Holmes and MacDonald state (1969,
p. 629), "Recorded or telemetered data are put through preliminary data handling
and reformating for delivery to a data processing section where data analysts cor-
relate photographic and scanner or electronic camera data." When considered in
terms of a system complete with feedback loops, these reports describe data acquisi-
tion and conversion systems rather than information systems. In short, there seems to
be a preoccupation with automatic data processing (ADP) for discrimination and
categorization in advance of clearly defined user needs. Those people who really
need remote sensor data are "primary" users. Once they have extracted this infor-
mation, data can be annotated and held for retrieval for use throughout the hierarchy.
Rather than regard the "user" as a nebulous entity who periodically makes
predetermined requests of information from a computerized data bank, a more viable
system would beonecapable of catering to expected as well as "unusual" information
needs from specific user groups. The system described in this report extends the
strategies described elsewhere by focusing attention on the location for data conversion
and the flow of information throughout the agribusiness community. In other words.
It begins where most other systems terminate. . .at the user interface. In diagrammatic
form the difference between this and previous approaches is shown in Figure 24.
Design and Function of the Information System
Heany (1968) outlines eight steps in the development of information systems
for use in business. These are: (1) establish or refine an information requirement;
(2) develop gross system concepts; (3) obtain approval; (4) detail the design;
(5) test; (6) implement; (7) document; and (8) evaluate. Although it is true that
the acquisition and processing aspects of remote sensor data in agriculture (that is,
the ADP part) are now in stages four and five, the flow of results as information
are, clearly,only in stage 2. It is essential that these two aspects be re-alligned,
lest the former (by virtue of strength) leads us along uncertain paths, or the latter
(by contrasting weakness) delays implementation.
70
ORIGINAL PAGE IB
OP POOR gUALIM
1
WRI S FLOW CHART (FROM LANGLEY ET AL. I970J
Figure 25 outlines a concept of an information system for agricul-
ture. In the terminology of systems design (Heany 1968) it can properly be regarded
as a "manual information system." This implies that its utility lies fundamentally in
human operations assisted by necessary computer and office hardware. However, it
is more than a manual system in that to function successfully it will require ADP
hardware and its associated software. These latter components are described else-
where (Lorsch, 1968; Bernstein and Cetron, 1969); and usable versions are now being
developed at The Kansas University Center for Research (Anderson, 1971).
According to stage 1 of the system, broadband microwave data are collected
by a satellite designed specifically for agricultural monitoring (here referred to as
AGRISAT). These data are then transmitted to a Data Return Facility, perhaps like
the one planned for Sioux Falls, South Dakota to handle ERTS and EROS data
( Geotimes , September 1971, p.26).
Stage 2 of the design takes advantage of information and expertise avail-
able at the local level (see subtask 2,5.2, 1). The strategy thus enables agricultural
county agents (representative of ASCS, SCS and AES) to "call" data as required from
the Return Facility to extract both their routine and specific information needs.
As presently conceived, information may be extracted most easily through
a combination of human interpretation and ADP techniques. For routine determina-
tions of agricultural statistics .(e .g . wheat acreage planted and harvested) automated
dichotomous keys provide one approach (Coiner and Morain, 1971), By this strategy
the in-depth knowledge of events and trends in agriculture provided by local agents
is retained in the system. Furthermore, this knowledge can be put to effective use
implicitly and explicitly in the process of creating interpretation keys for their own
county (see subtask 2. 5. 2. 2).
Once created, the keys should be able to rapidly process data coming in on
a regular basis. The variety of keys and related ADP algorithms would depend upon
the kinds of information desired by local users and by the needs of higher levels in
the agricultural chain-of-command. The obvious advantage of beginning information
extraction at the "grass roots" level is that non-aggregate data can be employed.
Timely maps of agricultural production, of diseases and other crop damage, trends in
cropping practices, introduction and diffusion of new crops, and a host of other local
phenomena could be more readily produced and monitored by beginning data inter-
pretation at the county level. Figure 26 is an artist's conception of a future county
72
Figure 25.
STAGE 1
STAGE 2
I?.<
a
jawijtjuoihda
agricultural office with automatic image and digital processing equipment.
Following the initial data review and information extraction by county
agents, stage 3 of Figure 25 indicates a vertical flow to state agricultural boards and
U5DA regional centers. What was considered "information" at the county level
now becomes essentially "non-aggregate data" requiring further synthesis and re-
evaluation to meet the information needs of regional analysts and state statisticians.
It is at this level that policy decisions begin to play a significant role in national
agribusiness. Thus, these intermediate levels in the hierarchy are as important as
the primary level, but for quite different reasons.
Remote sensing centers for agricultural information might be located in
each of the presently recognized 20 farm production regions of the United States
(Figure 27). Counties within each of these regions would pass their processed non-
aggregate data and the Information derived from them to similar ADP units at the center
(Figure 26B). By following this path, and remembering that each county has based its
interpretation on locally prepared algorithms, much of the distance-decay problem
associated with identification and categorization over large areas is effectively
avoided. In the parlance of probability theory, each county in the region becomes
a "training area," thus diminishing greatly the "prediction area."
Taking the Central Great Plains Winter Wheat and Range Region as an
example (H), a total of ±250 counties, one can more easily picture the kinds of
information extractable at stage 3. A sensing program aimed specifically at winter
wheat with data arriving bi-monthly from April through July would provide the fol-
lowing near "real-time" information: (1) acres planted and harvested; (2) northward
progress of ripening and harvest; (3) direction and rate of spread of infestations;
(4) estimates of soil moisture status and drought prediction; (5) allocation of railroad
stock; (6) yield prediction (probably based on per/acre yield estimates coupled with
acreage data). Together with the archival function of regional centers, one might
add to the above list: (7) spread of innovation (cultivation practices and new crops
or crop varieties); (8) historical summaries.
In the scheme presented here, regional centers are regarded as the seat
of archival data; namely, digital tapes, imagery, supporting ground data, etc. From
these inputs information can be assembled for use at the national and international
levels. Stage 4 of the design represents the top of the agricultural hierarchy in
75
LAND RESOURCE REGIONS AND MAJOR LAND RESOURCE AREAS
OF THE UNITED STATES (Exclusive of Alaska and Hawaii)
A northwestern forest, forage, and
SPECIALTY CROP REGION
0 NORTHWESTERN WHEAT AND RANGE REGION
c CALIFORNIA SUBTROPICAL FRUIT, TRUCK, AND
SPECIALTY CROP REGION
Q WESTERN RANGE AND IRRIGATED REGION
E ROCKY MOUNTAIN RANGE AND FOREST REGION
p NORTHERN GREAT PLAINS SPRING WHEAT REGION
0 WESTERN GREAT PLAINS RANGE AND
IRRIGATED REGION
H CENTRAL GREAT PLAINS WINTER WHEAT AND
RANGE REGION
| SOUTHWESTERN PLATEAUS AND PLAINS RANGE
AND COTTON REGION
J SOUTHWESTERN PRAIRIES COTTON AND
FORAGE REGION
K NORTHERN LAKE STATES FOREST AND fORAGE REGION
L LAKE STATES FRUIT, TRUCK, AND DAIRY REGION
M CENTRAL FEED GRAINS AND LIVESTOCK REGION
N »ST AND CENTRAL GENERAL FARMING AND FOREST
REGION
O MISSISSIPPI DELTA COTTON AND FEED GRAINS
REGION
p SOUTH ATLANTIC AND GULF SLOPE CASH CROP,
FOREST, AND LIVESTOCK REGION
R NORTHEASTERN FOREST AND FORAGE REGION
S NORTHERN ATLANTIC SLOPE TRUCK, FRUIT, ANO
POULTRY REGION
T ATLANTIC AND GULF COAST LOWLAND FORESTAND
TRUCK CROP REGION
U FLORIDA SUBTROPICAL FRUIT, TRUCK CROP, ANO
RANGE REGION
Figure 27.
76
EXAMPLE 2 EXAMPLE 1
this country. It consists of a host of Services, Sections and Branches, the information
needs for which are so complex that they cannot be considered in detail here. The
most immediate needs are discussed by Houseman (1969), ERS (1967), Park (1969)
Mayer and Heady (1969) and others. As examples of the information derivable from
the system described here, we cite particularly the work of Mayer and Heady.
Among the most sweeping needs facing policy makers in American agri-
culture are those describing rates of change. Especially important are the mutually
related causes influencing farm size and number, farm employment, capital invest-
ment, effectiveness of production control programs, yield trends, and the supply
and demand of crop land. As will be shown, information pertaining to some of these
changes can be obtained by remote sensors.
In Mayer and Heady (1969), for example, the potential land available in
the United States by 1980 for the seven major crops is given at 252 million acres
(p.381). The distribution of these acres, including land idled by government pro-
grams in 1965, is given in Figure 28. Among other parameters, it is almost certain
that the distribution of idled land will change in the intervening decade before 1980
according to supply and demand functions created by changes in government pro-
grams. Retired land in one region may come under renewed cultivation forcing land
in another region to pass out of production. In Figure 29(a - d)the geography of idle
land in 1980 is predicted on the basis of 4 distinct production and trade programs.
Even a quick glance reveals the difference between an acreage quota program (29a) and
crop production under a free market economy(29c) . The quota system apparently dis-
tributes idle land fairly equitably throughout the 144 recognized producing regions. The
free market situation, however, eliminates from production those regions that cannot
compete .
Production on land not idled can be estimated by appropriate strategies.
Average feed grain yields for 1980, for example, are predicted by Mayer and Heady
by the equation (1):
Y fgk= t P ik Y !k <’>
1=1
where Y f i = weighted average feed grain yield in the k ^ production region,
tgk k = 144
P.^ = proportion of feed grain acreage devoted to feed grain crop
in 1964 (calculated as P.^ = A;[/y^ A.^ where
A = acreage) '
= actual average yield of the crop in the k^ region
77
Figure 28a.
Cropland
for major field crops
by regions projected
Region
1980
Uni ted
States
251 ,171
(Thousands of
A.
Northeast
5,711
B.
Lake States
24,708
C.
Corn Belt
70,306
D.
Northern Plains
60,613
E.
Appelachian
11,654
F.
Southeast
11,483
G •
Delta States
11 ,265
H.
Southern Plains
30,712
J.
Mountain
16,434
K.
Paci f i c
8,285
* Land
base is for wheat, corn
, oats.
barley, grain
soybeans, and cotton. Other cropland used for fruits,
vegetables, and minor crops has been subtracted from the
total. The figures do not include land devoted to tame
hay in rotation with other crops or grown alone. But
the figures do include cropland idled under government
programs in 1965.
Figure 28b. The 10 farm production regions of the
United States.
Note: The Farm production regions shown in 5b. are not
the same as Crop production regions as delineated
by Mayer and Heady. The acreages in 5a. are
derived from aggregating data for all 144 CPR's
and partitioning according to FPR'S.
78
Figure 29.
c. PROPORTION OF TOTAL CROPLAND UNUSED IN EACH OF THE 144 CROP
PRODUCING REGIONS FOR MAJOR CROPS UNDER A FREE MARKET MODEL
WITH 1965 LEVEL EXPORTS IN 1980.
d. PROPORTION OF TOTAL CROPLAND UNUSED FOR MAJOR CROPS IN EACH OF
THE 144 CROP PRODUCING REGIONS UNDER A FEED GRAIN PROGRAM WITH
TREND LEVEL EXPORTS IN 1980.
80
In both of the above examples (distribution of idle land and crop yield),
it is obvious that acreage data for various agricultural land uses is of central impor-
tance, It has always been so in agricultural monitoring, and, at least in American
agriculture, this information has traditionally been supplied from lower to upper
levels in the community. Consequently, the case for utilizing remote sensors for
data acquisition and for dissemination of information by the scheme suggested here
needs further clarification and justification. On-going work in this area will focus
on a cost/benefit analysis of the proposed system.
81
RADAR SENSING IN AGRICULTURE: AN OVERVIEW*
Stanley A . Morain
Task 2,5,2 Radar Sensing in Agriculture: An Overview*
For any sensor system time is a key discriminant. Research on the spectral
reflectivities of plants has shown that instantaneous unique signatures are unlikely
to exist and that time-sequential imaging may be required to identify crops (Haralick,
et al., 1970; Park, 1969; Wiegand, et al., 1969). Basically there are two temporal
frameworks in which to work: seasonal and year-to-year. Under both there exist
within- and between-crop radar variations, but the economic and social implications
attendant upon each are vastly different. Radar work to date has been largely
under the heading of seasonal change between crops (cf. Schwarz and Caspall, 1968).
Results from these efforts have shown that numerous variables must be considered to
make even the simplest determinations.
The intent of this summary is to outline current capabilities for radar in agri-
culture and to sketch a few economic benefits attending their use.
Seasonal Changes Between Crops
Schwarz and Caspall (1968), Morain, et al. (1970), and Morain and Coiner
(1970), working with imagery from Ka-, Ku~, and X-band frequencies respectively,
have shown that major agricultural crops can be segregated, though not unambiguously
identified, using simple two-dimensional plots of HH and HV film densities. In
Table 3 percentages are given to indicate the degree of isolation by crop type for
each of the systems.
Although the data** represent only a few crops and at present fall below
acceptable levels of accuracy for most crop reporting services in this country,
*Condensed from: CRES Technical Report 177-14, December 1970.
**These values should not be used to judge system capabilities. They were derived
from imagery taken at different times in the growth cycle and from images of
highly varying quality.
82
TABLE 3
CROP SEGREGATION ON SCATTERGRAMS
AS A FUNCTION OF RADAR FREQUENCY
AND DATE IN THE GROWING SEASON
Crop 7-66 9-4-69 9-15-69 10-69
Type Ka-Band Ku-Band Ka-Band X-Band
% % % %
Wheat Not
Present
Grain
Sorghum — 69 77
Corn 82 28 92
(cropped)
Alfalfa 50
Sugar 92 — 97 64
Beets
Bare 91 90 83 91
Ground
83
they suggest a capability that might benefit agriculture in several ways. It is well
known that reliable crop statistics for many developing countries do not exist, or,
become available too late for any but historical use. Regional and world-wide
figures for total cultivated acreage, crop diversity, or acres in particular crops would
be welcome input for agricultural planners both here and abroad. Tracing trends
in global cultivated acreage would aid significantly in formulating population policies
and in making production or carrying capacity predictions (Ehrlich and Ehrlich, 1970).
Bachman (1965), Kellogg (1963), and others have stressed that most developing
countries still have scope for increasing agricultural acreage, though this potential
should decrease with increasing population. As illustrated in Table 3, the bulk
of all cultivated acreage would be revealed at some time during the cropping cycle
as bare ground. The advantage of using either airborne or spaceborne radars as
monitors for this is their ability to operate in clouded environments, where low sun
altitudes prevent photographic sensing, or where small scale synoptic coverage is
demanded.
It is not, in fact, the ability of radar to collect useful information that stymies
its wider application, but our ability to interpret the data. As Haralick, et al. (1970)
point out, it will take a Herculean effort to create automatic data processing routines
for crops whose spectral properties vary continuously in time and space. Wheat, the
world's most important crop, is produced in scores of varieties under as many culti-
vation practices. Clearly, it is unrealistic to expect radar or any other sensor to
provide interpretable agricultural data without knowing the nature or magnitude
of within-crop geographic and phenologic variations. Basically, the problem reduces
to knowing which radar and terrain parameters are critical in making accurate land
use identifications. In searching for these we have overlooked one of the simplest,
most useful aids to identification yet devised — the dichotomous key (see subtask
o r ^ n\
L * J .Z * z} .
Using keys, and provided imagery can be disseminated quickly enough for
primary agencies to interpret the data in at least a sampling framework, improvement
in statistics from the crop reporting services could be realized. At present, data are
gathered by an army of volunteer observers in coordination with regularly mailed
questionnaires. By the end of the year it is often true that acreages, and yield
predictions for wheat in the Great Plains are accurate to within 3 per cent. However,
predictions prior to harvest are often gross estimations. It is not in improving accuracy
84
for which radar holds great promise but in providing early estimates and in decreasing
uncertainties inherent in the predictions.
Prompt and efficient interpretation of radar imagery generated at regular
intervals throughout the growing season could dramatically improve such statistics
as number of acres in particular crops, progress of the harvest, number of acres
harvested and others. All of these would result in better planning at the county
and state level, even if the margin of improvement in accuracy were small.
Seasonal Changes Within Crops
Detecting within-crop changes assumes that the crop has been identified.
The most dramatic benefits to accrue from agricultural sensing reside in our ability
to unravel these within-category changes. The following paragraphs outline
the magnitude of these benefits and the evidence behind our belief that they can
eventually be achieved.
Figure 30 illustrates actual and anticipated radar images for crops as they
might appear at Ka-band four times in the growing season. Wheat is especially
interesting. It suggests that in the Winter Wheat Belt economy, statistics such as
acres planted, harvested, and lost could be calculated. In June the drop in return
from these fields would signal that harvest had occurred. Such changes through time
could almost certainly be tallied. One requirement, however, since wheat is har-
vested somewhere in the world all the time (Thompson, 1969) is the ability to obtain
synoptic coverage every 2-3 weeks regardless of weather conditions.
During Spring in the Winter Wheat Beit natural pasture, perennial alfalfa,
and wheat represent the only growing crops and should be mutually distinguishable
on the basis of terrain context and image attributes. By March or April an estimate
of winter wheat acreage could be made and compared to estimates from the previous
season. Under the present system the first forecast in Kansas is made in December,
followed in the Spring by monthly up-dates beginning in April. Final tabulations
appear six months after the harvest (Pallesen, 1970).
What value would radar derived crop statistics have for society? One answer
is that by improving the precision of weekly and monthly crop reports, better yield
predictions could be made. Errors of a few tenths per cent in acreages for a crop
as important as wheat may send shock waves throughout the network of domestic
85
Sugar
Beets
Figure 30. Typical radar appearance at Ka-band for economically important crops
at Garden City, Kansas (derived from July and September imagery)
arrayed against their expected appearances at the beginning and end
of the growing season (May and November).
and foreign trade. World-wide surveillance of wheat by radar might help alleviate
problems associated with planning acreage allotments, periodically reassessing estimated
production. Recommendations might be feasible early in the season to either graze
or plow under surplus acreage, preventing excessive production; or to increase
southern hemisphere acreage (Australia, Argentina) to compensate for poor harvests
in the northern hemisphere.
The most important aspects of within-crop seasonal change are those subtle^,
differences associated with crop quality or variety. Both are part of a complex yield
function the discerning of which lies at the foundation of agricultural reconnaissance.
With the "Green Revolution" in progress in Asia (Wharton, 1969) it is imperative that
both acreage and variety be considered in production estimates. Willett (1969)
reports that farmers in southwest Asia have already shown their enthusiasm for adopting
new varieties, but that at the national level marked inequalities exist in the rates
86
of adoption. By the 1967-68 reporting period India had converted 20 per cent of
its wheat acreage to new varieties whereas Pakistan had converted only 12 per cent.
At present there is no evidence that radar, or any other sensor, can detect
(at acceptable levels of confidence) differences between varieties, let alone identify
them. There may be hope for some estimates, however, by using surrogates related
to time (if some varieties ripen earlier or permit double or triple cropping) or space
(if they occur in restricted localities) (Brown, 1968). Unlikely to be detected are
small differences in a° arising from increased head size, longer stalks, or other
geometric parameters. For the foreseeable future it seems that radar's most valuable
contribution to yield prediction rests largely with acreage calculations inserted
into a more comprehensive yield function.
Two aspects of crop quality lend themselves to radar monitoring. One is
moisture status; the other is physical damage due to lodging, hail damage, or extreme
defoliation. Present evidence of how moisture status influences a° is inconclusive.
This evidence resides in mottling patterns within fields. We have observed, for
instance, that tonal irregularities exist for both corn and sorghum in October. At
this time of year these crops are harvestable but differences in planting date, irrigation
history, and local variations in ripening are manifested as tone mottling. In most
instances crop geometry within corn or sorghum fields can be assumed to be uniform
and that differences within fields are due to moisture patterns. To confirm this
assumption, experiments are needed to derive crop dielectric as a function of
changing moisture and to establish the backscattering cross-sections of crops at
various moistures and stages of growth.
Physical damage (lodging or extreme defoliation) arising from heavy rain
or hail can be detected on fine resolution data as a change in tone or texture. In
mature corn, when row orientation is orthogonal to look direction, differences are
observable which pinpoint areas of damage or poor quality (see subtask 2. 5. 2. 3).
As a parameter of the yield function, cause of poor quality is not important. It is
sufficient to know simply that low quality fields are also low yielders and should
be weighted accordingly in making production predictions. However, if causes and
effects can be related by using radar, tremendous economic benefits accrue. By
following the rapid expansion of diseased areas or by tracing damage along storm
tracks (both of which demand near all-weather capability) assessment of economic
loss or preventive measures could be quickly made. The cost of monitoring would
87
be small compared to the savings. In Kansas alone loss of sorghum from aphid infes-
tations, which spread across the state in two weeks, amounted to $14,733,000 in
1968-69 (Kansas Board of Agriculture, 1970). Over $13 billion was lost to American
agriculture in 1969.
Year to Year Changes Between Crops
The benefits of long term agricultural sensing have not been seriously con-
sidered. Land use histories compiled from data collected over the years and stored
for rapid retrieval could be useful in developing production control measures. In
higher echelons there is a tradition of juggling the amount of land planted to control
surplus and deficit (Doll, et al., 1968). However, problems of cross compliance*
and input substitution have hindered any startling successes. By using automatic
data processing, accurate regional histories and local land use trends could be
mapped. Such projects are not possible today because historical records of field
size, crop type, or production are not available. It would be highly desirable to
trace diffusion rates and directions of new crops (varieties?) or of the development
of farm-to-market road nets (Brown, 1968); to follow the geographic spread of in-
novation such as the use of polyethylene protectors on sugar beet seedlings or the use
of sub-surface asphalt layers; and to monitor the progress of land reform policies.
Efficiency of production has been the keynote of 20th Century agriculture; yet, the
policies instituted to achieve that efficiency are mostly based on inadequate aggre-
gate** statistics rather than detailed local data. Disaggregation may hold promise
for substantially improving agricultural economic theory, and for this reason alone
radars may prove their value by supplying synoptic coverage of any desired region.
Year to Year Changes Within Crops
The least studied aspect of agricultural surveillance, the gradual changes in
crop reflectivities over the years, may ultimately be the most important in aiding
plans for world food supplies and production since these indicate the spread of the
"Green Revolution" or improvements in crop vigor. Most of the increase relates to
gradually improving yields which leads us to agree with Pendleton (1970) that the
Cross-compliance refers to participation in several government programs sometimes
with conflicting requirements.
**Aggregate statistics refers to averages rather than individual values. See, for
example, Grunfeld and Griliches, 1960.
88
concept of a "yield plateau" is a myth. Increasingly man must rely on greater
production from each acre to feed expanding populations. The best mechanism for
doing this is by creating better varieties and engaging in other forms of input sub -
stitution; namely, fertilizer application, irrigation, etc. We are quite uncertain
how radar will prove economically beneficial, but certainly a most worthy pursuit
would be to explore the numerous possibilities.
Subtask 2.5.2. 1
LOCAL LEVEL AGRICULTURAL PRACTICES
AND INDIVIDUAL FARMER NEEDS AS INFLUENCES ON
SLAR IMAGERY DATA COLLECTION*
Floyd M. Henderson
Introduction
Crop identification has long been one of the objectives of radar imaging
systems. Yet, there are many phenomena that can be studied apart from this one
simple aspect. It is the purpose of this paper (1) to discuss the complexities and
variables in land use practices that affect crop variation and lead to observed dif-
ferences in landscape patterns from region to region, (2) to illustrate that everything
in the environment is so closely interrelated that an attempt to isolate one factor is
extremely complicated, and (3) to describe and list other information obtainable from
radar apart from crop identification.
Remote Sensing at the Local Level
What a farmer thinks and how he perceives his land are important variables
determining the patterns ultimately imposed on the landscape. His perceptions of
what crop and variety to plant; when to plant it; how much to plant; where to plant
it and in how large a field; whether to irrigate; how to irrigate and when; how to
plow and plant a field; his Idea of the future market and government programs; and
*Condensed from CRES Technical Report 177-15, July 1971.
89
how and when to control weeds will affect each and every field. The ways and extent
to which such decisions affect the imagery will vary among sensors, but it is a
variable confronted by the interpreter in analyzing the inter-connected and related
aspects of the environment. It is small wonder that imagery anomalies and incon-
sistencies result when all the physical variables possible are crossed with all the
cultural perceptions of how to vary an environment. It is apparent that land use prac-
tices are as variable as the mechanical parts of the sensor. Data that are needed to
improve farm management as perceived by the farmer and county agent are assessed
with regard to radar's potential to supply answers.
Until recently, the information needs of users at primary levels (farmers and
county agents) have been largely neglected (Lorsch, 1969). Yet, it is at this level
that many of our broadest claims for uses of remote sensor data are made. In July,
1970, data were collected in interviews with 112 farmers and agricultural agents.
By working at the local level, it was possible to determine some of the needs regarding
land use and farming practices as perceived by these people. Three counties (Fin-
ney, Wichita, and Grant) in the High Plains of Western Kansas were selected to
serve as a study area.
This is admittedly a small sample considering the total number of farm types
and operations in the United States. Problems paramount in other environments have
not been determined but will surely have an impact on the potential usefulness of
radar programs. In compiling responses to the interview, a decision was made to
include only those answers most often given to avoid minor or singular requests.
Those designated with an asterisk (*) indicate possible radar applications. Clearly,
many of the problems listed are not amenable to radar analysis or to radar analysis
alone. It should be noted also that asterisks represent present as well as potential
future capabilities. A complete defense of each present or future application is
beyond the scope of this report; consequently, a brief summary foiiows relevant
responses.
In reply to the first question "Which aspects of your farm and its operation
would you like to know better but cannot now determine or predict?" farmers
answered:
(1) Proper fertilizer application — optimum time and amount.
*(2) Land production capability — This might prove possible by measuring
(1) crop yield by field; (2) the slopes and drainage systems of the land;
90
(3) soil condition, i.e. existing nutrients in soil, what fertilizers are
needed, degree of soil salinity; and (4) determining better crop rotation
strategies. The ability of radar to monitor and aid the calculation of
(1) and (4) might enable algorithms to be applied to indicate a field
or area's production capability. To date, however, only radar's ability
to determine slope and drainage system (2) has been proven (McCoy,
1969). The detection of soil salinity and other soil condition changes
might prove feasible in the future if the dielectric properties of these
soils can be more accurately defined. Use of dichotomous keys indi-
cates that consistent crop identification may be possible in the near
future with fine resolution radar. With these data it should be possible
to relate ground truth information and expand the key to detect other
crop parameters.
(3) Knowledge of expected market prices early in planting season.
(4) Long range accurate rainfall prediction before planting.
(5) What the next government price supports are going to be.
(6) Which crops to plant and how many acres per crop.
*(7) Accurate irrigation guidelines (e.g., optimum time and duration of
application). Periodic analysis of fields could be made by a sensor
system capable of determining soil moisture and/or crop vigor. At a
specified soil moisture content or degree of plant stress, a signal from
the sensor could be sent to a computer center which would in turn send
a signal to the automatic irrigation system and water would be applied to
the field. The actual implementation of such a system is obviously
somewhere in the future. Due to present system limitations it is impossible
to know if such a radar - satellite - computer - automatic field water
applicator chain will function on a large scale. However, studies by
MacDonald and Waite (1971) and Hardy, et al., (1971) have shown that
microwave imaging systems carry moisture content information in their
near-range presentation. It seems possible that with further study, the
moisture relationship viewed in the radar image may be correlated with
moisture requirements of crops.
*(8) Soil moisture and content — Differences in soil moisture might be detect-
able by radars capable of penetrating the surface. The differences in
91
return (e.g. grey level or texture) of certain fields would indicate the
amount of soil moisture present. See also number 7 above.
(9) Water Table level .
(10) Prediction of hazards (e.g. hail, tornadoes).
*(11) Periodic soil analysis to determine soil fertility — This might be possible
in very general terms. It must also be remembered that a soil's fertility
is relative to the crop being grown. A change in crop yield or changes
in a bare field's appearance, as seen from time sequential imagery over
a period of years, might indicate a change in soil nutrients. Extreme
cases such as a rise in salt content, which will in turn influence the
dielectric properties of the soil, might be more easily "spotted." A
dichotomous key might be used to develop a detection capability.
(12) Income stability.
(13) How to make a profit on a farm.
(14) How to increase yields.
*(15) Early plant disease and insect infestation detection — Variation or
anomalies in a field's texture and/or tone pattern might indicate a
disease or infestation in a crop. If a sensor could relay this information
fast enough and early enough to the farmer, proper preventive measures
could be completed to minimize crop loss. Other farmers in the area
could be warned, so precautions could be taken by them. Again a
dichotomous key procedure might be developed to indicate if field
anomalies were a result of disease, or of crop and soil variations.
Although some pre-visual detection may be possible, caused by shifts
in the gross surface roughness not otherwise observable from single
point locations on the ground, the probability of active microwave
detection of insect or plant disease infestation in a single field seems
unlikely. The more realistic role is In the determination of areas in
the path of migrating infestation, as a system to provide data to develop
infestation control strategies and direction of spread.
(16) Insect and disease elimination prior to and after field infestation.
(17) How to cut operation costs — By reviewing historical trends on broad
scale imagery, it might be possible to develop better crop rotation
systems, field arrangement patterns and shapes, irrigation applications.
92
and soi I conservation practices. Such information and imagery could be
made available to local county agents (provided with some training
and/or interpretation keys). The farmer should realize a cut in operation
costs when these recommended farm operation practices are employed.
After answering question (1), farm operators were asked “What kind of information
might come from remote sensing experiments that would be of use to you?" Their
most frequent replies were:
*(1) Prediction of pests and disease in crops — See 15 above.
*(2) Changes in soil fertility — See 11 above.
*(3) Optimized water application — Limited capability in soil moisture
detection has already been demonstrated by MacDonald and Waite (1971).
More accurate delimitation of soil moisture might be possible if the com-
plex inter-relationship of the dielectric constant to moisture, salt
content, and nutrients in the soil can be defined and consistently identi-
fied. It would then be necessary to calibrate and detect variations
in each of these at small calibrated intervals.
*(4) Current field and crop conditions — Such information could be made
available to the farmer, if automatic data processing of current field
conditions (e.g. soil fertility, crop stress) could be incorporated into
image analysis. This information would have to be available to the
county agent or directly to the farmer by phone. An integral part of
such a system is the development of dichotomous keys that can analyze
field variables quickly as automatic data inputs.
(5) Drought prediction.
(6) Location of ground water.
(7) What nutrients soils need.
*(8) Accurate acreage measurements — The degree of accuracy needed to
improve present methods varies according to the level of economic
development existing in parts of the world. Using a system of equations
developed by Sabol (1968) it Is possible to determine field acreages on
certain parts of a radar image. Such information would permit the compu-
tation of the amount of land under cultivation; expansion, contraction,
or changes in areas regarding land use alterations; and better field size
93
and arrangement to improve farm management. Houseman (1970) states
that remote sensing data could provide highly valuable supplementary
and collateral information regarding crop statistics and forecasts.
*(9) When and how much to irrigate — See 7, question 1 .
Answers to the third question, "If such information as periodic analysis of predicted
crop yields, soil moisture content, or plant vigor were available, how would you use
them on the farm?" were:
(1) To increase profits.
*(2) More efficient farm management — See 2, 7, 8, 15, and 18 of question 1 .
*(3) Optimize water application — See 7, Question 1.
*(4) To increase yields — See 2, 7, 11, and 17 of Question 1; 4 and 8 of
Question 2.
(5) Optimize planting time .
*(6) Detect and control disease and insects — See 15, Question 1 .
(7) Be informed of problems and correct them.
(8) Make me a better farmer.
(9) Income prediction.
(10) What crops to plant and knowledge of their yields.
*(11) For farm planning — See 2, Question 3.
*(12) Knowledge of soil fertility — See 8 and 11, Question 1 .
*(13) To make government reports more accurate — A sensor capable of monitoring
crop acreages periodically throughout the growing season would vastly
improve world harvest and even U.S. crop harvest estimates. Present
methods rely on volunteer reports by farmers and/or government agents.
A radar data collection and interpretation system having all-weather
capability and providing synoptic coverage could detect and rapidly
report crop acreages, predict losses due to environmental hazards, and
forecast market prices. Although not operable with present systems, this
objective should be studied in developing future generation radars.
See also 8, Question 2.
(14) What to plant and when.
Five local agricultural officials were asked, "What information would improve your
ability to aid farm operations and farm planning?" Although this represents a small
sample, it represents not only their needs but the needs of hundreds of farmers as
94
viewed by persons immediately involved with them. Their replies were:
*(1) The prediction of yields by soil moisture depth in fall seeding time for
dryland crops (see above) — See 8 and 11, Question 1.
*(2) The effect of irrigation water on the soil with specific information on
soil salinity — See 7, Question 1.
*(3) Insect and disease detection — See 15, Question 1 .
*(4) Soil permeability by field.
(5) Compaction of soil.
(6) Soil classification by texture and structure.
*(7) A better overall picture of a farm than could be obtained by walking.
This included: (a) drainage and erosion - topography; (b) optimum land
use versus actual use in relation to slope and conservation practices;
and (c) better field and building arrangement — See 2 and 17, Question
1; 8, Question 2.
(8) More accurate survey of livestock numbers and feedlot arrangements.
(9) Pollution control measures.
*(10) Flood control measures — Past studies have proven the capability of
radar to delimit drainage basins, stream networks and topography (McCoy,
1969; Lewis, 1971; and MacDonald, 1969). The average amount and
time of precipitation for an area could be obtained from weather bureau
stations and state water resources (USGS) personnel. Knowing the
general soil types and degree of hill slope the potential areal runoff could
be computed using a combined formula. These areas could then be checked
by local county agents to see if remedial or preventive erosion control
measures existed or needed to be constructed. Increased land productivity
and more efficient farm management should accrue from such efforts.
(11) Moisture stress on crops on a weekly basis.
*(12) Degree of water weeds in irrigation ditches and larger water bodies —
Present systems cannot adequately detect such changes in small irrigation
ditches. Changes in larger water bodies might be possible with fine
resolution imagery in a time sequential framework to monitor changes in
the size of the water body and its reflective properties. Such changes
could be analyzed and the result related to loss of water area due to
drought, silting, or weed and plant growth in the water. It should be
95
noted that, at present, such data seem to be obtainable only with
classified systems of high resolution. For this kind of monitoring, low
incidence angle may be preferred.
The description of problems and variables presented here may leave the
impression that the interpreter is faced with an insurmountable task. The diverse
land use practices appear so complex, and the local needs so detailed that there may
appear little hope for radar. This is certainly not the case. These variables are not
only problems, but also significant clues in analyzing and interpreting agricultural
land uses from radar imagery. With this information the reason for the non-homogeneous
appearance of a single field or the different appearance of two identical corn fields
may be resolved. These relatively minute changes, differences, and perceptions are
meaningful inputs to the development and refinement of: (1) radar systems and
(2) such interpretive aids as the dichotomous key and tone-texture analyses.
In essence, these data bits may provide some of the information necessary to
explain anomalies and inconsistencies in a radar image. When this information is
incorporated into interpretation keys, identification and monitoring accuracy should
rise several levels. The more ambiguities that can be explained and eliminated, the
better present and future radar systems will function as a viable research tool.
Subtask 2. 5. 2. 2
IMAGE INTERPRETATION KEYS TO SUPPORT
ANALYSIS OF SLAR IMAGERY*
Jerry C. Coiner and S. A. Morain
Introduction
Before SLAR can be used generally as a method of agricultural investigation,
a number of technical difficulties involving both the sensor and the interpretation
of the imagery must be overcome. One of the most pressing of these difficulties is
the lack of an interpretation technique for SLAR imagery that is simple to develop
and accurate to use. Ideally, such a technique should require a minimum of image
interpretation expertise while providing accurate and repeatable interpretations. As
a contribution toward filling this need, the present study is an investigation into the
*Condensed from ASP Fall Meeting Proceedings, San Francisco, Sept. 1971, paper ^71.
96
adaptation of image interpretation keys for analysis of SLAR imagery.
The subject matter of the keys developed and studied in this paper are drawn
from two areas: agricultural crop discrimination and natural vegetation mapping.
Three agricultural keys were developed utilizing imagery from Garden City,, Kansas
(NASA test site 76) and two natural vegetation keys were developed for imagery of
Horsefly Mountain, Oregon and Yellowstone National Park, Wyoming. Examples
of some of them are included solely to illustrate the methodology; they in no way
exhaust the variety of foreseeable approaches, nor do they represent limits on the
applicability of the technique.
Examples of Keys
The use of keys to interpret radar imagery for agricultural information was
initially conceived by Morain and Coiner (1970). The first attempt was developed
for X-band radar imagery obtained in October 1969. The ground truth data were
compared with the imagery on a field by field basis to determine the status of the crops
in the fields. Then, for each crop in each of several growth stages, imagery work
sheets were prepared. These work sheets constituted the basis for preparing the
dichotomous key shown in Table 4. This table is a revised and simplified version
of the previous key. Table 5 is another example, but from Ka-band imagery acquired
in September 1965. This key relates specifically to crop type discrimination and is
not the limit of information contained in the image. It is based on data from the
mid radar range (depression angles betweenl8 and 23°) and the HH/HV image pair.
There is little quantitative information other than gray scale (tone) on which to base
the key.
The construction of keys for natural vegetation is more involved than that
for agriculture because the interpretation requires integration of both tone and texture
information. One way to handle this added level of complexity is to alter the key
format, as shown in Table 6. This key was necessarily constructed subsequent to the
interpretation, thus it represents a method by which the generalization of the inter-
pretation can be validated. Such "matrix keys" may also provide an excellent training
aid for other interpreters.
The matrix key offers several advantages to the more experienced interpreter.
(1) It is not necessary to retrace the entire logic of the interpretation, as is necessary
97
TABLE 4
DIRECT DICHOTOMOUS KEY FOR CROP TYPES AT GARDEN CITY, KANSAS
(For use with fine resolution X-band imagery for October)
Field is light gray to white on HH
Field is not light gray to white on HH
Go to B
Go to D
Field gray tone shifts from light gray/ white HH to medium gray HV- Go to C
Field gray tone shifts HV lighter than HH cut alfalfa
Field gray tone on HV homogeneous —
Field gray tone on HV not homogeneous
Field has medium to dark gray tone on HH
Field has very dark gray tone on HH — — *
Go to E
sugar beets; or wheat >3°
fallow
recently tilled
Field gray tone is homogeneous — — Go to F
Field gray tone is not homogeneous — • Go to I
Field has lineations parallel to long axis
Field does not have lineations
Go to G
maturing alfalfa
Field has medium coarse texture —
Field does not have medium coarse texture
Go to H
grain sorghum (rows J- flight line)
Field has same gray tone on HH and HV
Field has moderate gray tone shift HH to HV
wheat 3"
- alfalfa >12"
Field has a cultivation pattern observable
Field does not have cultivation pattern observable but
displays pronounced boundary shadowing
emergent wheat
mature corn
TABLE 5
DIRECT DICHOTOMOUS KEY FOR CROP TYPES AT GARDEN CITY, KANSAS
(For use with AN(/APQ-97 Ka-band imagery for September)
'O
S3
A HH and HV are white — —
A* HH Is not white Go to B
B HH Is light gray Go to C
B* HH is not light gray —————— Go to D
C HH and HV are light gray
C* HH is light gray, HV almost white
D HH is gray Go to E
D 1 HH is not gray — Go to G
E HH has even gray tone — Go to F
E' HH has uneven gray tone (also HV)
F HV has similar gray scale to HH — — — — «
F' HV has lighter gray scale than HH
G HH is dark gray to black,
even gray scale Go to H
G' HH is dark gray to black,
uneven gray scale
(possibly more noticeable in HV)~ Go to I
H Area has regular boundaries
H' Area has irregular boundaries —
I HV shows major shift toward gray
to light gray
I' Field shows onfy minor shift
toward gray
sugar beets
corn
alfalfa
fallow
wheat
sorghum
recently tilled
standing water
pasture
fallow
absent faint medium coarse
(TEXTURE) (TEXTURE) (TEXTURE) (TEXTURE)
Black Dark Medium Light Dark Medium Light Dark Mediu m Light Dark Medium Light
TABLE 6
MATRIX KEY FOR AN/APQ -97 RADAR IMAGERY
YELLOWSTONE NATIONAL PARK *
HH
COARSE
(TEXTURE)
MEDIUM
(TEXTURE)
Tone Dark Medium Light
1 ;
• i
* i 1
i i i
, L _ T h*rrrv*l t
\tA"S El*v i Ap*«* ‘
1 fortll * | fit
+ - -j- - ■*- -H
i »
Dark Medium Light
i
FAINT
(TEXTURE)
Dark Medium Light
ABSENT
(TEXTURE)
Black Dark Medium Light
low
O'*
I Art*
\edgtpofe
1 Pk*
*Med
Ht
• o T
- i Mi ted
• Coral
I fie*
» On
1 Arti
1
1
1
1
\ 1
l
1
1
1
1
1 Mirth
t
t (Nik
1
1
!
1
I
1
l-,.. .
■
i
*
*
—
1
1
1
1
1
-f
1
1
1
1
1
i
1
-r —
i
1
1
1
1
1
1
1
—
1
1
1
l
i
i
-4
1
1
1
(
\ "
1
I
M**7! '
1
1
~H*n _
1
1
_ L
1
1
1
l
R«d*r|
t*. \
t
( !
5h* 1
Or i
1
1
JS±
Art* ]
— J
1
1
NEH -1371
SSSSfSSSS
100
in the case of the dichotomous key. (2) The key provides an excellent method of
information transfer from one interpreter to another and may substantially increase
the consistency of a given interpretation. (3) The matrix key surfaces those points
within the interpretation where the image fails to clearly separate categories of
information, for example the overlap between Lodgepole Pine, Mixed Conifer and
Medium Elevation Dry Area. Identification of ambiguity allows the interpreter
to concentrate on associate information (in the form of associate keys) for those
categories where clear-cut identification is not possible.
Testing Keys
Two of the direct dichotomous keys were subjected to interpreter testing,
primarily to determine whether, in practice, results were sufficiently rewarding to
warrant additional research. Secondarily, we needed to learn which of the human
and system factors most influenced the successful use of keys. Lastly, we attempted
to better appreciate the nature and type of collateral inputs required to make high
validity interpretations.
The tests were relatively simple in nature. Each interpreter was given a
packet consisting of materials providing a brief background to the subject area;
images and keys for Garden City for July 1970 (Ku-band) and September 1965 (Ka-
band); and appropriate instructions. Individually, ten experienced radar interpreters
from around the U.S. were asked a series of questions relating to their interpretation
experience. They were also requested to describe the method they used to interpret
the images. No attempt was made to control interpretation technique, as it was
felt that this would not provide a measure of the key's effectiveness and would place
undue emphasis on a given method of interpretation. Each interpreter was asked
to make an interpretation using each of the two keys provided. The interpreters
used the direct dichotomous key to identify crops in 55 fields selected by the investi-
gators. Slant ranges for the fields were the same as those used in developing the key.
The test of a Ku-band was based solely on the use of a textual key with no
supporting graphic. Range of correct interpretations varied from 25 per cent to 68
per cent, with the average at 49 per cent (Table 7). Higher percentages were
achieved by interpreters who had the least overall experience in interpretation.
The test for a Ka-band key was based on a textual key accompanied by a
visual graphic showing typical crop responses. It required a finer degree of
TABLE 7
PERCENTAGE CORRECT CROP IDENTIFICATION BY TEN IMAGE INTERPRETERS
USING KEYS DERIVED FROM Ka-BAND AND Ku-BAND IMAGERY
Interpreters
(Ranked according to
general interpretation
experience; 1 = high)
1
2
3
4
5
6
7
8
9
10
Average
% Correct Crop Identification
AIS/APQ-97 DPD-2
Ka-band
9/65
Ku-band
7/70
44 1
34
31
35
57
i 43%
55
48
52
35
50
43 i
68
33
68
44
39%
25
52 '
41
26
57
41
49
45%
51%
1. Recently Tilled
Fallow
Pasture
|!« Sur'yhum
Wheat
Corn
Ml. Alfalfa
Corn
Sorghum
IV. Sugor beets
Alfalfa
TABLE 8
IDENTIFICATION OF CROP GROUPS*
Cron Grouo . JbJumbfiiL.filJield^ d . ej3l L fied by interpr e t e r s
U .P. sugar beets corn offal fa wheat sorghum posture fallow tilled
52
29
1 20 12
2 35
36
10
8 48 59
7 27
43
95
13 2
.Per csnL
Corrsct-
73
86
87
98
^Derived by combining the responses of oil interpreters.
**Bold type = fields correctly identified; regular type = incorrect identification.
102
differentiation of crops than "hat needed by the other key. This resulted in a
narrower range of correct interpretations (26 per cent to 57 per cent) with somewhat
lower average correct interpretations (41 per cent). When individual crops were
grouped into clusters simulating agricultural scenes typical of different times in the
growing season, however, it was clear that accuracies exceeding 75 per cent could
be achieved by using keys. The figures in Table 8 compare very well with similar
crop groupings derived through cluster analysis by Haralick,et al . (1970).
The accuracy of the initial keys was below that acceptable as a minimum
by the authors. However, the tests showed areas where the direct dichotomous key
could be improved. Therefore, results of these tests should be assessed from a
research point of view as a basis on which to build an operational interpretation key.
Several factors entered into the limited effectiveness of this "first approx-
imation:"
1) the lack of extensive coverage of the test site by any given system,
2) the lack of time sequential coverage, and
3) the lack of non-image information for use with the direct dichotomous key.
Even in the light of the above problems, the key appears to provide a capability
for increasing interpretation accuracy. When the key is coupled with a test sequence
for research purposes, a feedback loop can be established to identify areas of low
accuracy.
The two tests conducted on the preliminary keys have given some estimate
of the initial accuracy and the problems inherent in the construction of interpre-
tation keys. The tests have pointed out the need for feedback and collateral infor-
mation in achieving high validity interpretations. Prior knowledge of an area under
interpretation and expertise in the type of information being identified (in this case
agriculture) may be advantageous for users of keys. Although the tests resulted in
only a 50 per cent accuracy, for individual crops, improved visual support graphics
and the expansion of the key to categorize more cases on the image, should increase
the level of accuracy. However, to reach extremely high (over 90%) correct inter-
pretations, the use of collateral information and time sequential imagery (introduced
into the interpretation in the form of associate keys) will be essential.
Key Automation
The procedure for automating keys is essentially one of substituting actual
values for qualitative word descriptions of tone and texture. For the key shown in
103
Table 5, only image tone was used. The imagery for both the horizontal and vertical
polarizations was digitized using 256 channels and a 50p cell size. By producing
frequency distributions for each of the images, it was possible to compress the 256
channels into 5 equal probability classes (arbitrarily designated as 0,1, 2, 3, 4).
These classes were assumed to be roughly equivalent to the five levels of gray
visually detectable on each image. Following this, the tape coordinates for each
test field were tabulated so that the newly defined 5 levels could be summed and a
simple average "gray tone" calculated. On the basis of these gray tone designations
for both the HH and HV images, each field could be classified according to crop
type. Appendix Ais a discussion of the algorithm used in this approach and provides
documentation for programs as they are presently developed.
Results from preliminary tests of the computer algorithm are encouraging but
inconclusive. The number of correct crop identifications so far has been on the
order of 50 per cent; no better nor worse than that achievable by human interpre-
tation. The testing program is continuing, however, and we are hopeful of better
results as we refine the strategy for calculating gray levels. We suspect for example
that the use of equal probability classes does not adequately portray the scene as
it was originally imaged, and that better identifications can be obtained by pre-
selecting regional concentrations within the frequency distribution.
With additional research into methods and problems of preparation and auto-
mation, there is a high expectation that accurate and repeatable interpretations
from radar data can be achieved — even perhaps in the absence of fully calibrated
systems. This approach to radar interpretation could provide a method of analysis
adaptable to government agency (mission oriented) needs, as well as the more
specific and individual needs of academic research.
104
Subtask 2. 5.2 .3
BASIC PARAMETRIC STUDIES:
THE STANDARD FARM DESIGN PHILOSOPHY
AND INITIAL RESULTS*
William Lockman & Phillip Jackson
Introduction
It is quite impossible to evaluate the full matrix of radar frequencies, polar-
ization, times, incidence angles, and resolutions for the Garden City Test site if
one keeps in mind the myriad terrain parameters affecting backscattering cross-
sections (crop types, height, cover, moisture, row direction, stage of growth, etc.).
Consequently, we are attempting to model the data in such a way as to hold constant
as many of the system/terrain variables as possible. For each imaging date over
Garden City we are producing a set of "standard farms" with each farm consisting of
pure, weed-free, uniform fields representing the major crop types. Such models
will be created for each frequency, polarization, incidence angle, and resolution
presently available. These models are to be prepared according to a standard
format for human interpretation as well as pattern recognition and IDECS interpre-
tation. In addition, the model can be digitized on a scanning densitometer so
that image density may be studied in a quantitative manner. The Standard Farm
model will provide us with a method to "fingerprint" average film tones for high quality
fields on a crop by crop basis in order that we may eventually be able to assess
departures from these ideal levels.
On radar imagery, vegetation and crop information depends not only on
vegetation type and growth, but also on a number of other variables. Phase I of
the Standard Farm Project is a controlled parametric experiment where crop quality,
a terrain parameter, has been held constant. Key radar system parameters to be
considered as variables are incident angle («£), polarization (P), and frequency (f).
Pert charts as exemplified by Figure 31 demonstrate the design. Studies include
experiments to determine the influences of individual radar parameters on crop
identification. If successful results are obtained from the isolation of these parameters,
*Condensed from CRES Technical Report 177- (in preparation).
105
LICt CUT MUI
4>
EASI-
EST
t«l-
rtsi
“ CONTINUED AS ABOVE
EAST-
WEST
NQRTK-
SOUTH
GRAtH SORGiWfl
Zhc®.
~ j j — m
~ j ( — pj"1
] p-W
a
a
■H
a
{F)
a
j_^ j— al
- j j — TmT)
— j p— [hh]
Q— CI®_
y-dE.
a
a
a
a
a
a
*MT SI UEBLL | Q
BARE CRgjND
if — T hTI
p * rm H T-L ^
H=®-®
| HAIURt ftUALfA~j j
ZhrfU
a
j^HEAf S T Ut-E L£
BAftt GROUMD ~[ — f~a ^
— h d^ B
CRAIH SC'RGHuT] — | ^
SU6AA 3E£TS ~~1 1 jjjj
MATURE ALFALFA | ~~- | j^-j
CLU ALfAIfA |— tZj=L- Qjjj
WHEAT SUABLE. 1 1 ^ |^j
EflFEGMMU 1
Z^ZKi^a
I g
CRA1H SOKHUrt"] 1 ^ |^-|
| SUGAR BEETS IH ^- B
| TIAUlfc ALFALFA ] r ^TTl
WHEAT SUJBBLi | f ^
BARE C ROUHp "| f ® p^-j
a
a
a
a
a
a
a
a
a
"All/fif A If ALFA
~j__ i— a
Zha®.
l~ l | — a
u-c®.
]_ |— B
- p-a
3-dEL
-^-a
Figure 31 • A flow chart for analysis of polar-
ization. In this case, row direction, inci-
dence angle, time (growth stage) and radar
frequency are held constant for each of several crop types so that the effects of
E olarization can be studied. For each of the polarization boxes a film density
istogram along with statistical measures will be compiled into a catalog.
106
more complicated studies will be undertaken using less than optimum field conditions.
"Standard farm" provides the framework by which the above experimental
design is being pursued. To assemble it, a single high quality field from each image
is selected from each of eight crop categories (corn, grain sorghum, sugar beets,
mature alfalfa, cut alfalfa, wheat, bare ground and pasture) along each of three
roads for each of two crop row directions (N-S and E-W). Ideally, roads are par-
allel to the line of flight; therefore, all of the fields along a given road are imaged
at the same depression angle . An example of one standard farm is given in Figure 32,
In all we have created TOOstandard farms for parametric analysis, using the data
listed in Table 9.
Methodology
The first step after collection of imagery is digitization. The scanning
parameters are as follows: raster, 50 microns; aperture, 50 microns; optical density
range, 0 to 2D (we are now adopting 0-3D); gray levels, 256.
Depending on the particular radar system, the 50 micron aperture is on the
order of 0.5 to 1 .0 resolution cell on the film transparency (i.e. the raster and
aperture size is compatible with the system resolution). The 50 micron raster (samples
taken every 50 microns) corresponds to a minimum of approximately 400 samples
per field for the smaller fields and over 10,000 samples per field for larger fields.
The optical density range of 0 to 2D has been divided into 256 gray levels. The
smallest measurable level corresponds to a density interval of 2/256= 0.0078.
Thus, a op for film of 0.0064 is approximately equal to the minimum measurable
interval.
Due to film characteristics and the 0-2D range, mean values and histograms
that approach or extend beyond the OD or 2D (O or 256 for 256 levels in the 0 2D
range) are of questionable value. A number of fields in the models under study
approached these limits. Small values for standard deviation (approximately zero)
were obtained when the mean density was greater than 2D as would be expected.
It should be noted that the statistics of imagery from different radar systems are not
directly comparable due to differences in system parameters, film type, and processing.
In order to locate standard farms on the map print-outs so that field statistics
could be obtained, it was necessary to collect coordinates. For each field involved.
107
FIGURE 32.
STANDARD FARM FORMAT
DATE July 27 , 1966 RADAR SYSTEM Westinghouse AN/APQ-97
ROAD INf.lDFNtF A, 30-39°
RON DIRECTION
POLARIZATION _ 5 !L
ENLARGEMENT 5X
CORN
M 11, F 4
GRAIN SORGHUM
M 10, F 3
!
SUGAR BEETS
M 7, F 7
I
MATURE ALFALFA
M 1, F 3
r
1 farad
HP
CUT ALFALFA
M 1, F 11
BARE GROUND
M 9, F 5
WHEAT STUBBLE
M 3, F 3
PASTURE
M 11, F 1
108
Table 9
GARDEN CITY DATA BASE
Test Site 76
Date
Radar System
Polarization
Field Data
September 1965
Westinghouse
AN/APQ-97
4
'
Crop
Parameters
July 1966
Westinghouse
AN/APQ-97
4
Crop
Parameters
October 1969
Michigan High
Resolution
2
Crop
Parameters
September 1971
(expected)
Michigan High
Resolution
2
Crop Parameters
Crop and Soil
June 1970
16.5 GHz
13.3 GHz
400 MHz
Scatterometers
2
Crop
Parameters
May 1971
NASA DPD-2
16.5 GHz
Scatterometers
2
Crop Parameters
Crop and Soil
Moisture
June 1971
NASA DPD-2
16.5 GHz
Scatterometers
2
Crop Parameters
Crop and Soil
Moisture
July 1971
Michigan High
Resolution
KU Scatterometer
2
Crop Parameters
Crop and Soil
Moisture
109
it was necessary to obtain four figures - a row and a column coordinate for the upper
left corner of the field, and then a'down"and an 'bcross"value to provide field dimen-
sions. From these coordinates and dimensions it was possible to retrieve the digitized
standard farm fields from the tapes containing the entire digitized test site and to put
them on a single tape in order that they may be rapidly obtained for statistical work.
For analysis, three descriptive statistics were obtained by computer programs
means, standard deviations, and histograms. Other statistics can be obtained at a
later date if desired. The analysis presented in the following section is a human com-
parison of histograms, means, and standard deviations.
Distinctive differences or similarities between attributes of a parameter are
the main concern for the first part of the analysis. For example: the parameter Row
Direction has two attributes: (1) east-west (parallel to line of flight) and (2) north-
south (orthogonal to line of flight). To assess the effect of row direction on radar
return, the attributes are compared under similar conditions of radar frequency, polar-
ization, incidence angle, crop type and growth stage. The statistical variation of
the two histograms is assessed to determine if the attributes of row direction affect
radar return differently. If no difference is found for example between the parallel
and orthogonal attributes then the suggestion Is that row direction is not a significant
factor influencing radar return for that specific crop situation. Conversely, if a
discernable major difference is observed for row direction in either the means or
standard deviation (or both), there is some evidence that this attribute is a factor
influencing radar return. Typical output from the standard farm is given in Appendix C
Preliminary Results
Table 10 presents means and standard deviations characteristic of the radar
returns for "developing" crops holding frequency, row direction, polarization and
incident angle constant. The higher the value for mean density, the brighter (whiter)
the crop appears on imagery. To take an extreme example, compare sugar beets (row 3,
column 2a) with bare ground (row 6 column 2a). Sugar beets appear almost white
with a mean density of 219 while bare ground with a mean density of 93 appears
almost black.
110
Croe
1 Corn
2 Grain sorghum
3 Sugar Beets
4 Cut Alfalfa
5 Wheat
6 Pasture
7 Bare Ground
Table 10
Comparative Radar Return from Crops in their "Developing"* Stage
Radar Variables
_L
2 _
_3_
Row direction
Polarization
Incidence
(orthogona
1 to LOF)
(HH)
(Mid range)
• a**
b
a
b
a
b
155(29)
249(13)
209(26)
237(26)
193(14)
240(16)
221(22)
221(12)
213(27)
206(47
182(28)
204(52)
201(63)
222(18
219(39)
252(49)
—
171(72)
—
—
184(22)
212(50)
184(22)
225(52)
—
253(5)
—
253(5)
—
107(5)
—
—
185(10)
254(4)
185(10)
117(8)
80(13)
191(25)
93(28)
175(15)
123(10)
214(13)
By selecting a stage in the growth cycle, time becomes a variable. Wheat, for
example, is "developing" in May whereas corn develops in July and August.
** Column a under each of the system variables represents Ka-band imagery; column
b represents Ku**band data. They cannot be compared except internally.
Return from row crops was found to have consistently high standard deviations.
These were interpreted as being a function of scatter due to rows. Although alfalfa
is not a row crop, it too was often observed to have a high standard deviation. This
may be explained by the irrigation ridges which are highly visible when alfalfa is at
a low growth stage.
It appears that (especially in the HH polarization Ku band) bright but
textured return is characteristic of developing row crops, whereas high but purer
tone return is distinctive of wheat or pasture. Low return (less bright) is characteristic
of bare ground in all cases. Textural effects may be due to soil surface conditions or
stubble in the field, but low mean brightness is a good indication of bare ground.
Other findings include the following:
1 . For row crops having rows parallel to the flight line, crop growth stage appears
to be reflected in the standard deviation. Larger standard deviations are characteristic
of emerging crops and this statistic becomes smaller with maturation.
2. Return is influenced by polarization in that like polarizations tend to be
brighter than cross polarizations under similar conditions except on bare ground
where the opposite is the case.
3. Range variation for non-row crops does not effect radar return. Similar means
and standard deviations are characteristic of near, mid and far ranges for wheat and
pasture. Corn, sorghum and sugar beets show distinct variations in return with change
in range.
The standard farm experiment seems then, on the basis of initial results, to be
yielding useful data concerning the relative importance of radar parameters to crop
identification. An expansion of the original design to include a larger number of
sample fields and more quantitative analysis is presently forthcoming.
112
Subtask 2. 5. 2. 4
REMOTE DETERMINATION OF SOIL TEXTURE AND MOISTURE
USING ACTIVE MICROWAVE SENSORS*
S. A. Morain & J. B. Campbell
Soil texture and moisture, under certain circumstances, are susceptible to
measurement by active microwave sensors. The most useful sensor parameters relating
to radar backscatter (<j°) from soils are: (1) frequency, (2) polarization, and (3) angle
of illumination. The primary soil characteristics contributing to backscatter are:
(1) soil texture, which, as it ranges from clay through boulder categories, varies in
surface roughness and thus influences the amount of backscatter; and (2) soil moisture,
which alters electrical conductivity of the soil, thus influencing the depth of signal
penetration and the amount of re-radiation. In theory, low frequency radar (L-band)
should best be able to detect boulder surfaces (assuming a flat, dry surface free of
vegetation and of uniform roughness) and higher frequencies (V-band) should be
sensitive to soils comprised of medium sand. Analysis of side-looking airborne radar
(SLAR) imagery of an area near Tuscon, Arizona, produced soil texture patterns
corresponding to soil survey patterns.
As soil moisture increases, soil reflectance in the microwave also increases.
The effects are most pronounced at steep depression angles (small angles of incidence).
SLAR imaging of soil moisture is illustrated in the report by imagery of Hutchinson, Minn
In this region topographic depressions containing peat and muck soils image stronger
than surrounding soils, especially at steep angles.
The combined effects of moisture and roughness should be considered in order
to properly characterize radar backscatter from soils. It is likely, for example, that
at low frequencies and steep depression angles, ambiguities will arise between high
returns from dry boulder surfaces and those from moist loamy surfaces. Additional
complications may result at shallower depression angles because the roughness com-
ponent resides more in vegetation patterns than in soil patterns.
*Excerpfed from 177-23 (in preparation).
113
The past decade of research 01 . the microwave properties of soils has focused
on two approaches. The first is illustrated by research conducted by Lundien (1966,
1971 ) in which the reflectance of soils was measured at selected frequencies under
artificial conditions of texture and/or moisture. The results of these works lead
toward an understanding of basic parametric interactions. They provide a link between
radar theory and application in a controlled environment. More recently, since
1965, a second approach has appeared. Sheridan (1966) and Barr (1970), among
others, have shown that a correspondence exists between known soil patterns and
patterns observed on radar imagery. They have described these trends and specu-
lated on their utility for engineering soil studies and for other "user needs." Both
approaches are essential if we are to extend our knowledge and capitalize on micro-
wave reflectance from soil. We see, however, that in order to increase interpretive
skill, we need yet a third approach; one which strives to explain image patterns on
the basis of parametric interactions. The thrust of this paper is toward that approach.
The discussion in Technical Report 177-23 is divided into two segments. The
first is largely a consideration of important radar system and soil variables; their
mutual interactions; and their theoretical appearance on radar imagery. In the second
part, examples are used to illustrate the degree to which theory carries over to
practice. For the present report we summarize only our research efforts with the
polypanchromatic radar system developed at KU.
Backscatter from Soils in the Field
In an effort to further investigate radar scattering from soil surfaces the authors
have conducted preliminary experiments at the Center for Research, Inc. at Kansas
University. Field conditions were altered to study the interaction of soil moisture and
roughness with frequency, polarization, and incidence angle. Three roughness
categories and two moisture conditions were investigated. The soil was a Grundy
Silty Clay Loam developed from shale interbedded with limestone. Normal field
capacity for the soil is reported to be between 30 and 40% by weight. Our measure-
ments were taken at about 15% (considerably below field capacity) and at about
30% moisture content. Microrelief was altered by roto-tilling and raking the surface
to desired roughness.
A frequency modulated, continuous wave (FM-CW) scatterometer was used
to take data at 10 frequency bands, each .4 GHz wide, between 4 GHz and 8 GHz;
114
this frequency range corresponds to portions of C and X bands. Backscatter data
were collected for two polarizations (HH and VV) at 6 look angles between 0° and
65° - Sample data are presented in Appendix B *
The curves in Figure 33 a,b,c represent selected examples from the complete
data set and should be regarded as preliminary. At present we are unprepared to
discuss cause and effect between soil and radar parameters. We include the infor-
mation only to demonstrate the range of variation observed in the frequency and
angular domains as roughness and moisture are altered. Curves A and B illustrate
the response over the frequency range (4-8 GHz) at 30° and 65° incidence angles,
respectively. Only the HH polarization is shown here. Curve C shows the angular
response for the .4 GHz band centered at 7.8 GHz.
In comparing the backscattering characteristics for the curves in A and B the
following points emerge. First, an immediate observation is that greater fluctuations
occur at steeper depression angles (shallow incidence angles). The response at 65°
is considerably flatter than the one for 30°. Of particular interest is the apparent
effect of moisture on the curves for smooth soils at 30° (curve A).
Second, the effect of roughness (comparing the two dry curves in A and B)
is to influence the frequency response more at near range than at far range. Also
there are fewer crossover points at 65° than at 30°. These crossover points could
aid in interpretation if seen in the context of a broad bandwidth over a period of
time — however, when seen at isolated frequencies these points can only confuse
any attempt to identify soil characteristics, since the effect of moisture and rough-
ness appear to have similar responses at certain frequencies. This effect suggests that
there may be optimum frequencies (not yet ascertained) for distinguishing certain
soil characteristics.
Analysis of the angular data (c) yields the following observations. First, at
higher frequencies (7.8 GHz) there is a considerable flattening of the curves at
shallow depression angles for all soil conditions studied. Although it would be dif-
ficult to distinguish dry from wet smooth surfaces, there appears to be a significant
difference between a rough and a smooth surface, at least at the frequency illus-
trated. At present we have little knowledge of the capability of existing imaging
radar systems for making the kinds of distinctions important in soil studies. From
both the geo-science point of view and the engineering point of view, this topic
merits further consideration.
115
FREQUENCY (GHz)
FREQUENCY (GHz)
0 10 20 30 50 65
INCIDENCE ANGLE IN DEGREES
ROUGH = MEDIUM TO FINE CLODS
MICRORELIEF 25-30 CM
OR GREATER
SMOOTH
= FINE CLODS
DRY/SMOOTH
WET/SMOOTH
DRY *
(2-8 CM DIA)
LITTLE MICRORELIEF
ABOUT 15X
DRY/ROUGH
WET =
ABOUT FIELD CAPACITY
RADAR BACKSCATTER FOR SOIL SURFACES AT SELECTED
FREQUENCIES AND INCIDENCE ANGLES FOR GRUNDY SILTY
CLAY LOAM.
Figure 33.
116
RADAR USES FOR VEGETATION:
AN OVERVIEW WITH EMPHASIS ON ARID ZONES*
Stanley A. Morain
Task 2.5,3 Radar for Vegetation Studies; An Overview
It has been argued that imaging radars will serve an ancillary role to photo -
graphy in sensing dry environments (Simonett, et al., 1969a). Normally the atmos-
phere over arid lands is cloud free and contains little moisture; hence the number of
hours available annually for conventional and infra-red photography is much higher
than would normally be experienced in either the humid low latitudes or in high
latitudes. Spectacular photographs obtained by Gemini and Apollo spacecraft over
the Middle East, Australia and the southwestern United States testify to the potential
of photography for acquiring resource data from these regions (OSSA, 1970),
Data losses can occur, however, through the combined effects of high albedos
with dust and aerosol concentrations in the atmosphere. These phenomena tend to
reduce image contrast and scatter short wavelength signals. In addition there are
film and terrain related color ambiguities which confound the job of terrain identifi-
cation (Artsybashev, 1962; Simonett, et al., 1969b). Considering that the arid
lands of the world constitute almost 36% of the land area, are poorly known, remote,
sparsely mapped, and largely unphotographed from aircraft altitudes, it is important
to investigate sensors which are independent of solar illumination and most weather
conditions. This leads naturally to the consideration of an active microwave remote
sensor as a potential data collector for these areas.
Arid Zone Applications
Initially, space oriented resource inventories will focus on such broadscale
topics as soil, vegetation, geology (geomorphology), drainage and hydrology (including
irrigation) and agriculture. As thematic maps and other land use information become
*Condensed from Morain, S.A. (1970), Radar Uses for Natural Resources Inventories
in Arid Zones, presented at First World Symposium on Arid Zones, Mexico City,
November 1970. In press by McGraw-Hill of Mexico.
117
available for these Interests, more comprehensive management and development
schemes can be implemented. It is therefore important to appreciate the role radar
data might play in some of these initial surveys. The following section deals speci-
fically with mapping natural vegetation.
Vegetation patterns have been observed on radar imagery from a wide variety
of environments. In only a few cases, however, have desert communities been iden-
tified directly rather than as a surrogate related to soils or topography (Morain and
Simonett, 1966). The reason, primarily, is that desert communities are generally low
and sparse. Consequently, their backscattering cross-sections are complicated by
"noise" from lithology, soil surfaces, moisture contents, and possibly by salinity
patterns. This is precisely the kind of problem confronting interpretation of coarse
resolution color space photography. For this and other reasons it Is widely agreed
that unambiguous identification of are a- extensive terrain categories can seldom be
achieved by a single sensor without the aid of local surrogates, perhaps not even
then. Given a basic pattern, local interpreters could make sound judgments regarding
the distribution of plant communities.
Figure 34a, b shows two aspects of the desert vegetation in Escalante Valley,
southwestern Utah. The region is largely characterized by great basin sage (Artemisia
tridentata) but in lower lying and more saline localities shadscale scrub (Atrip lex
confertifolia ) becomes the dominant type. In moister, less saline situations a variety
of grasses form the association and on the upper slopes of the alluvial fans juniper
woodlands occur. The sagebrush type is uniformly low but has a variable density.
In more open areas a sandy (pebbly) surface is often exposed to radar signals, but in
denser stands there may be relatively little return contributed by soil — especially
at higher incidence angles where objects protruding from a flat surface intercept
most of the impinging signal .
From the radar imagery it is possible to delimit the shadscale scrub community,
the larger localities of grass, and, by inference, the area dominated by sagebrush
(Figure 34c). Juniper woodlands cannot be confidently located due to their associa-
tion with rough topography. Preliminary field investigations have confirmed that plant
cover more than soil differences influence the radar return from this area, though a
more detailed examination might show equal dependence. Unquestionably, the soil
and vegetation patterns are highly correlated. Figure 34c suggests that the most
important influences on backscatter from desert terrain are height and density of veg-
etation and soil texture. If so, significant differences in image texture, tone and
118
ORIGINAL PAGE IS
OE POOR QUALITY/
b
VEGETATION TYPES NEAR MILFORD, UTAH
(After Field Investigation)
□ Medium high return
(Sagebrush shrub)
Medium low return
(Shadscale shrub)
Low return
(Grass meadow)
Variable return
(Bare or near bare land)
Agricultural land
Radar return obscured
c
polarization should be observable in places such as central Australia where huge
areas are dominated by mulga scrub (Acacia aneura ), spinifex plains (Triodia spp.)
or mixtures of the two. The bluebush deserts of South Australia consisting mainly of
Kochia should be analogous to the great basin sage of North America.
As a second example of mapping potential for vegetation. Figure 35 shows
a high altitude area north of Flagstaff, Arizona. Within the region three vegetational
zones are known to exist; pine forest (mainly Pinos, j pndero w); grassland; and juniper
woodlands. From the preceding discussion one would expect to find clear boundaries
attending such grossly different height and density phenomena, at least at higher inci-
dence angles. Inspection shows that this is indeed the case. The variations are so
great in fact that image textures as well as tones can be delineated and categorized.
For areas of pine forest (see Figure 36) a definite "popcorn" texture is observed on
the original imagery. Since both polarizations display nearly the same tone and
texture, it appears that ponderosa forests have little depolarizing potential. "Texture"
is a complex combination of phase relationships between trees of varying height and
resonances associated with needle spacing. Phase may be particularly important
mature stands where scattering facets facing the receiver are added in time phase
(giving a bright spot on the imagery) and those facing away from the receiver are
subtracted (giving a dark spot)*
On enlarged versions of the cross and like polarized components, intricate
patterns of tone and texture can be recognized, then categorized according to known
ecological relationships and image appearance. Figure 35 demonstrates the capacity
for mapping not only the broad structural groups, but several subgroups as well .
Although the cross polarized image is not shown in Figure 36, the addition of this
information permits a few boundaries only marginally distinguishable on the HH to
become clear. The technique is analogous in photography to adding a different film-
filter combination. Presumably, if the full complement of polarizations and viewing
directions were available, highly reliable delineations could be made. The reasons
why some plant communities depolarize signals more than others are not fully known.
Some preliminary statements can be found in Morain (1967) and Morain and Coiner (1970)
In the subtask reports that follow, we summarize our research efforts and those
under subcontract to the Forestry Remote Sensing Laboratory at Berkeley in more
temperate regions of the U.S.
120
Cover Types for the
Kendrick - Humphreys
Peak Area, Arizona
0 5 10 Km
1 H H
0 3 6 Mi
Scale
PINE FOREST
PINE PARKLAND
JUNIPER/GRASS
DRY GRASSLAND
SAM/CRES/70
□
H
SHOW
NO DATA
SOURCES:
Boundaries -
K-band SLAR
Categories -
SLAR & ecolooic
Inference
MOIST
MEADOW
Figure 35
Figure 36. K-band like polarized imagery of the Kendrick-Humphreys Peak
area , Arizona . Tones and textures represent broad plant commu
nitles .
122
Subtask 2.5.3. 1
VEGETATION MAPPING WITH SIDE-LOOKING
AIRBORNE RADAR: YELLOWSTONE NATIONAL PARK*
Norman E . Hardy
The purpose of this study is to delimit the vegetation communities of Yellow-
stone Park, from SLAR imagery and to explore the feasibility of identifying vegetation
types by power spectrum analysis.
The park was chosen for three reasons: (1) the availability of AN/APQ-97
SLAR imagery for the greater part of the park, (2) the willingness of the National
Park Service to provide ground support data including a vegetation map, and (3)plant
communities in the park are structurally simple and conform to those typical of the
eastern Rocky Mountains.
The interpretation approach was to define the boundaries of the vegetation
communities. After boundary definition, a classification of vegetation types was
established based upon elevations, moisture and slope features evident in the SLAR
image. Image boundaries corresponded well with those on the ground truth map, but
the classification derived from the imagery was not in total agreement with that shown
on the ground truth map. To assist in the classification of the vegetation communities,
a matrix interpretation key was constructed (see subtask 2. 5. 2. 2).
Comparison of the radar produced map (Figure 37) with the pre-existing
National Park Service map (Figure 38) suggests that most of the physiognomic changes
detectable on the ground, are also clearly defined on the radar imagery. However,
the radar map does contain varying degrees of reliability (Figure 39), since much of
the park has only received single coverage. Consequently, most of the park is
covered by only a fraction of the incidence angle range available from the system.
Only an area at the north end of the Park has received multiple coverage, and it is
this area where plant communities are most accurately defined (compare Figure 37
wi th 38 and 39).
Although problems of interpretation have been encountered, this study has
*Condensed from an article of the same name in NATO Advisory Group on Aerospace
Research and Development (AGARD) Conference Proceedings No. 90 on Propagation
Limitations in Remote Sensing, 1971, pp. T 1/1 — 1 1/19.
123
PAGE
QUALITY FIGURE 37
RADAR VEGETATION
YELLOWSTONE NATIONAL PARK
ORIGINAL
POOR
Lodgepole Pine - Mi*ed Coniferous Forest
Douglas Fir
Forest
mm Dry Alpine Grassland
uli m i m ilu (Above 3000 m.; Short Grasses, Some Sedges, Flowering Plants)
Dry Sub- Alpine Grassland / Conifer Complex
k&Z& a (Lodgepole Pine, Drought Tolerant Grasses)
Marsh (Sedges. Wetland Grasses)
] Thermal Areas (Varying Vegetation Types)
fEggpta Moist Sedge grass /Shrub Complex EE5553 water Mass
(2 200 -2 700m.; Willows, Wet land Sage , Grasses) ^ ■
1 Dry Lowland Grass/ Sagebrush Complex
J (2200 - 2700 m.; Dryland Sage, Drought Tolerant Grasses)
Dry Sub -Alpine Sagebrush Meadow (2700- 3000 m.)
No Radar Coverage
124
FIGURE 38
FIELD DERIVED VEGETATION MAP
YELLOWSTONE NATIONAL PARK
Mmiau W»Mfcnr< AH Ele*#l WU.Msnh Lb"*)
MUM T»1ul *0d Lfrwtactalad Mi 9 M«rtdl
Jdric 3oflflbru<t.-Q»«»l»a(ftiOO‘ Llmrtl
M*pdc- (fiMtf-WOy. fill wd will
$f%r
f f Tldrmil A(*» U»*<V«l* 1 P d
Source WiltlamHHiniiricktM trd K.
YlUowalom Nation*! P»rX. Nal tonal Park Sarvica,
U.S. Dtp*rlr»et>1 pi th* Inlarigr. Flfe'ua'v. IOTO
125
UNCONTROLLED RADAR MOSAIC
we DELINEATION OF PARK BOUNDARIES, 0*10, AND SCALE ON THIS MOSAIC IS APPROXIMATE.
COMPILED BY THE U. 3. GEOLOGICAL SURVEY
'?u\ 0 V«7 11 *«> »• '*« ">■ tHI IAHTH MIOUKtt
THIOUOl? t«? NA,,OHAL At«OMAUtlC» AND SAACf ADMINISTRATION
THIOOOH THE COOPERATION OP THE U. 3. ARMY ELECTRONICS COMMAND.
” t°€ STmQ H omi f 7/»r».r J r * MAK MVIl6MD »* THE AEROSPACE DIVISION,
WE STIMO HOUSE ELECTRIC CORPORATION fOI THE U. S ARMY ELECTRONICS COMMAND.
been successful. The major recommendation relates to mission planning. From this
study, it is concluded that analysis of natural vegetation communities is dependent
upon at least a 40% sidelap. This ensures that all areas covered will be found at
least once in near, mid and far range, thus allowing comparison between vegetation
reflectivity at different look angles. Unlike geology, where it has been suggested
that flights parallel to topographic features are most advantageous, we recommend that
vegetation analysis requires flights which are perpendicular to, as well as parallel to,
major topographic features.
In an effort to glean vegetation information from the imagery beyond routine
boundary delineation and categorization, we are extending the investigation to
include the use of lasers. For the purposes of this report it is not necessary to go into
detail regarding the internal workings of lasers. Nevertheless, a brief discussion
is worthwhile to understand the strategy and evaluate the results.
The term "laser" is an acronym meaning Jjght amplification byjtimulated
emission of radiation. It is a device for producing a powerful monochromatic light
beam in which the waves are coherent. Once produced, the light must be filtered
to eliminate extraneous noise; then refocused. The refocused light is then passed
through a preselected "scan" area on an image. The pattern within the scan area is trans-
formed and a Fraunhofer Diffraction Pattern is produced in the transform plane of the lens.
Once the transform has been created, wedge and annular ring filtering techniques, which
are the basis of this study, are applied to the FDP. Measurements of transmitted
power representing various terrain directions and periodicities can then be noted.
Wedge Filtering
When a coherent light beam is passed through an image, any angular char-
acteristics present on the image will be manifested as axes within the FDP. These
axes are symmetrical in opposite quadrants of the transform and pass directly through
the origin of the pattern. The object of wedge filtering is to filter all but a very
small amount of the total light in the transform over a sequence of small discrete
angles from 0° to 180°. As the filter is rotated about the origin of the transform,
light passes only if some degree of directionality is present. If the surface possesses
strong directionality, it will be a fairly intense axis in some direction; by integrating
the light energy of the FDP along a radial line, the total contribution in one direction
is obtained independent of frequency (and independent of location in the scan area).
127
This will be represented as a strong peak on the INTENSITY vs. 0 curve (Figure 40).
Annular Ring Filtering
When coherent light is passed through an image it is transformed and linear
features are displayed as axes in the transform, while high frequency features and
periodic features are displayed as points of high intensity at various distances from
the center of the transform. It is this characteristic of the transformed image which
is the basis of the annular ring filtering technique, and the resultant intensity
vs. radius curves. For this study, the annular rings are a series of 27 rings which
range from approximately 0.5 mm to 22 mm in diameter, with a ring width of approx-
imately 0.5 mm. To generate a data set, the rings are moved across the transform and
power readings taken from each. When values have been derived for each of the 27
rings, they are plotted as intensity vs. radius curves. For the area scanned, any
peaks on the curve will show if the surface has a distinct periodic characteristic;
if it has, the location along the X-axis of the curve will indicate whether the phenom-
enon responsible is of high or low frequency (Figure 41).
Optical Processing As a Potential Aid to Image Interpretation
At this point we need to discuss the kinds of imagery used in the Yellowstone
experiment, how each type lends itself to optical processing, and the kinds of data
to be expected from each type. Remember that as image scale and resolution relative
to the ground surface increase, different characteristics of the surface will influence
the appearance of the image. This simply means that as the interpreter examines
several types of imagery, different characteristics of the surface will influence his
interpretation.
In the Yellowstone example we employed: (I) 35 mm photographs taken from
an altitude of about 2000' and having a ground resolution of one foot; (2) USGS
black and white aerial photography at a scale of 1:37,400 and ground resolution of
+ 5'; and (3) Ka-band imagery at a scale of 1:155,000 and a ground resolution of about
50'.
128
35mm Aeriol Photography
For photography with T resolution the interpreter must consider individual
plants and their geometries when attempting to obtain optical data from the image.
A coherent light beam of 2 mm passing through an image of large scale and high
resolution will be transformed largely according to the characteristics of individual
trees, and the local environment in which the trees are located. Therefore, the
i ntensity vs. G curves (wedge filtering) for a forest may show a distinct linearity
which is related to the branches and branchlets of a tree. Meanwhile, the intensity
vs. rad'us curves (annular ring filtering) may show frequencies related to the number
of branches per unit area of the tree, and if the tree is sparsely branched they may
show features relating to the understory and even the ground. Thus, from imagery
of this type, it is reasonable to expect to distinguish the tall, straight, sparsely branched
Alpine Fir from the shorter, much bushier Douglas Fir. However, it is essential
that the interpretation be made using both the curves and the original photography
or image in order to determine which feature is responsible for any periodicity.
Vertic al Aerial Ph ot ography
The imagery employed in this stage of the experiment was standard aerial
photography of 5‘ resolution. In terms of image interpretation this implies that grasses,
flowering plants and many shrubs (such as small willows or sagebrush) should appear
as a relatively smooth, homogeneous surface. For the application of the optical
processor, this suggests that meadowland would not likely show any frequency or
periodicity characteristics, but that forest would have a distinctive pattern dominated
by relatively low frequency components (e.g. trees spaced rather far apart).
From the standpoint of vegetation structure the resolution of the system demands
that the interpretation be focused upon frequency relationships of the trees and their
environments; upon the biotope and localized structure rather than on physiological
characteristics of individual plants, as was the case in the previous example using
35 mm photography. It is at this scale that the study begins to assume a spatial
characteristic, and as a result becomes more of a geographic study than a purely
botanical project. From this point, it becomes possible to examine the characteristics
of infraspecific tree spacing, as well as interspecific distribution of trees, or trees
and shrubs.
130
Ideally, If frees of a given spec’es tended to exhibit a fairly even distribution, as
do many of the plants of arid zones, it would be possible to derive Intensity vs.
Radius curves which would reflect this periodic nature.
S LAR Imagery
The radar imagery for Yellowstone Park is at a scale of 1:155,000, with a
ground range resolution of approximately 50 feet. At this scale and this resolution,
it is obvious that individual branches, branchlets, or even individual trees will not
be detectable on the image. Thus, to make optimum use of the information con-
tained within the radar image, the interpretation approach will probably have to be
based upon the broader community study of individual plants. This means that the
research must be directed toward the analysis of ecosystems, with emphasis placed
upon complexes such as ecclines and ecotones and the problems of specific ecologic
range.
With these concepts as a basis, it is assumed that the optically derived radar
data will not show characteristic signatures for the individual plants; however, it is
likely that the curves produced will reflect broad community structure for each
community studied. Thus, it is quite likely that the biomes of grassland, desert,
savanna, and forest will be readily distinguishable, and inter-community composi-
tions within each biomes will be possible through analysis of gross frequency char-
acteristics of communities which are larger than the resolution of the system. Therefore
if the optical system can derive data of this nature from the imagery. It will probably
be possible to break the communities down to the species level through inferences based
upon knowledge of regional and local environments.
Analysis of Optically Processed Radar Imagery
In the analysis, five areas which appear totally different to the naked eye
were selected from the radar imagery. These areas were optically scanned for frequency
and directionality data, as were the corresponding areas on the photographs. For this
report only the results from Lodgepole Pine Forest are presented.
Lodgepole Pine is the dominant conifer in Yellowstone. It forms the main
portion of the upper canopy, and homogeneous stands are relatively easy to locate on
the radar image.
131
The examination of the frequency analysis for radar must take cognizance of .
the 50' resolution (Figure 42a, b). High frequency indicators are related to the film
grain, system noise, flaws in the film, or scan lines but definitely not to surface related
phenomena. The inspection must be restricted to the low frequency portion of the
Intensity vs. Radius curve. Within this portion of the curve two peaks appear. The
first is at 97 feet (132 lines per inch) and the second at 65 feet (200 lines per inch).
Comparing this with the curve derived from aerial photograph (suitably reduced to
the radar scale) a comparable pair of peaks can be observed (94 feet and 62 feet).
Since the images ere of comparable scale, it is possible that these frequences genuinely
represent unique surface phenomena. We must explore further the question of whether
the phenomena are vegetation related.
Turning briefly to the directional data derived from the wedge filtering
operation, it appears that for radar a strong directional characteristic is present.
However, detailed comparison of the curve and the imagery reveals that the high
intensity peaks at 0° and 180° represent mechanically caused scratches. The peak
at 90° represents the cross range dominance of the radar scan lines. These are discrete
lines imparted to the film by the radar pulse as it is returned and transferred through
the CcR.T.
132
INTENSITY
LOG INTENSITY
Figure 42
Subtask 2. 5. 3. 2
SLAR IMAGERY FOR EVALUATING WILDLAND
VEGETATION RESOURCES*
Steven J.Daus & Donald T. Lauer **
Introduction
The objective of the research reported here was to determine the
utility of SLAR imagery for evaluatingwlldland vegetation resources. Specifically,
comparisons were made, with the help of a group of skilled photo interpreters, between
certain ground features such as aspect, slope and major vegetation/ terrain type and
corresponding tonal/texture image characteristics for each feature or groups of fea-
tures as seen on the SLAR imagery. In addition, qualitative evaluations were made
regarding the overall usefulness of SLAR imagery.
Description of Study Area and SLAR Imagery
The Bucks Lake-Meadow Valley site, in Plumas County, California, was
chosen as the study area for four reasons: (1) the site is located in the north-central
Sierra Nevada Mountains in the heart of a mixed conifer forest type which encom-
passes a variety of types, compositions, species, densities and age classes of vege-
tation; (2) the site possesses a variety of topographic conditions ranging from 80%
slopes adjacent to the middle fork of the Feather River to flat pasture lands in Meadow
Valley; (3) a large portion of the site previously had been flown with a SLAR system
and the resulting imagery was considered to be of excellent quality; and (4) an abun-
dance of ground data was available about the site.
Analysis Procedure
Conventional photo interpretation performed on vertical aerial photographs
involves a complex process of evaluating a number of image characteristics such as
* Condensed from ASP-paper 71-333 presented at ASP-ACSM Fall Convention, San
Francisco California, September 1971.
**The authors are located at the Forestry Remote Sensing Lab, University of Califor-
nia. This study was performed under subcontract 1775-9. A final report is in
preparation .
134
size, shape, texture, tone, shado ' and stereo parallax. The analysis of these factors
combined with the additional powers of the human brain (5,e., subjective reasoning,
intuition, convergence of evidence, past experience, etc.) allows the interpreter to
recognize, identify and deduce the significance of objects or conditions seen on the
photographs. However, this task is further complicated when working with SLAR
imagery. Rarely are size, shape, shadow and stereo parallax useful inputs to the
interpretation process because either they do not exist in the imagery or cannot be
recognized. Consequently, the interpreter generally relies on his ability to discrim-
inate levels of image tone and texture. Thus, when attempting to map wildland
vegetation types on SLAR imagery, a reference key illustrating discrete tonal/texture
categories for each type of interest would be useful to the interpreter. First, however,
it is necessary to determine if there is a consistent tone or texture value that can be
assigned to any one of several vegetation/terrain types.
Photo interpretation tests were performed whereby numerous systematically
selected plots on the SLAR imagery were classified into one of nine tonal/texture
categories. The interpreters were not asked to identify objects and conditions on the
SLAR imagery; they were instructed only to categorize the plots in terms of tone and
texture .
Three skilled interpreters working independently with the same SLAR imagery
classified each plot. A reference key, showing examples of each tonal/texture
category, was used by the interpreters as they evaluated the image characteristics
of each plot. The interpretation key was constructed in such a manner that each
point could be matched with one of nine chips representing a particular tonal-texture
category. Once the interpreters classified all of the plots as to image tone and
texture, it was possible to relate their results to ground truth data collected for each
plot (i.e., aspect or orientation of terrain with respect to sensor, steepness of slope
and major vegetation/terrain type). Thus, with the tabulated data, correlations
could be made between the tonal/texture properties of an image and the corresponding
ground features (see Tables 11 and 12). Note that the first row in Table 11 should be
read as follows: 78 image plots were classified as smooth-white; 86% of those plots
were on slopes normal to the beam while 14% were on slopes oblique to the beam;
30% of the plots were on 20-40% slopes; and 20% on 40-60% slopes and 50% on
60-80% slopes; and 9% were in dense conifer, 33% in sparse conifer and 58% in dry
site hardwoods. The remaining rows in Table 11, as well as those in Table 12, should
135
TABLE 11
RESULTS GROUPED BY TEXTURALAONAL CATEGORY
TexhjralA onc| l
TextureAo na l Points (Percent of Total For
Each Category Falling Within Each Type)
Category
(and number of
Aspect
Slope
VegetationA errain
points in each
category)
Normal to beam
Facing away from
beam
Oblique to beam
0-20%
20-40%
40-60%
60-80%
Dense conifer
Sparse conifer
Dry site hardwood
Brushfield
Bare ground
Riparian and
meadows
Water
Smooth
White (78)
86
14
0
30
20
50
9
33
58
0
0
—
0
Smooth
Grey (104)
46
19
35
42
10
8
16
25
43
4
12
-
0
Smooth
Black (108)
2
72
26
20
18
36
4
30
52
0
3
-
11
Medium
White (127)
76
9
15
17
36
22
15
35
44
8
6
-
0
Medium
Grey (206)
40
35
40
9
14
48
31
10
6
-
0
Medium
Black (123)
8
34
26
19
21
20
46
28
4
4
0
Rough
White (29)
52
14
34
24
38
31
7
4
49
43
0
4
-
0
Rough
Grey (81)
5
75
20
46
30
4
10
20
54
21
1
3
-
0
Rough
Black (44)
6
71
23
30
41
10
19
18
55
27
2
2
“
0
136
TABLE 12
RESULTS GROUPED BY VEGETATION/TERRAIN TYPE
Vegetation/Terrain Points (Percent of Total Within Each
Type Falling Within Each Category)
Vegetation/
Aspect
Slope
Textural/Tonal Catego
ry
Terrain
Type
(and number
of points in
each type)
£
o
0)
_Q
O
15
E
0
Z
“E —
a
0)
-Q
E
0
i-
>-
a
£
o
CD
C
o
a
LL.
E
a
a)
o
a>
3
a“
3
O
O'*
O
CN
1
O
£
Tf
l
O
CM
•sp
£
O
1
o
£
00
1
o
o
o
+-
IS
f
_c
-k-
o
o
E
m
>s
<D
s—
CD
\
o
o
E
CO
u
o
_Q
-C
o
0
E
m
<u
IE
£
i
E
p
m ~o
a>
2
<D
s-
CD
E
3
0)
5
o
0
1
E
D
15
o
<u
Z
$
i
-C
CD
3
O
O'
>V
(D
u
CD
1
_C
CD
3
O
O'
U
a
_Q
1
-C
CD
3
O
Dense
Conifer
(134)
34
25
41
27
38
5
30
6
13
3
16
24
19
1
12
7
Sparse
Conifer
(369)
35
26
39
24
38
22
16
7
8
9
12
23
15
4
13
7
1
Dry Site
Hardwood
(338)
37
31
32
24
31
23
22
13
14
16
16
18
10
3
5
4
Brushfield
(15)
80
20
0
100
0
0
0
0
33
0
7
14
33
0
7
7
Bare
Ground
(46)
26
0
74
67
20
7
6
0
26
7
15
30
11
4
4
2
Riparian-
meadow
(0)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Water
03)
0
100
0
100
i 0
0
0
0
0
100
0
0
0
0
0
0
137
be read in the same manner.
Discussion of Results
The data in Table 1 1 show a consistent relationship between certain tonal/
texture categories and slope aspects. In fact, aspect of the terrain, in relation-
ship to the positioning of the sensor system, seems to have a profound effect on the
image characteristics. There does not appear, however, to be any consistent
relationship between image tone and texture, and vegetation/terrain types (except
for large bodies of water) as indicated in Table 12.
The data suggest that vegetation typing will be inaccurate in areas of rugged
terrain. However, this does not mean that radar imagery is useless. It was
discovered that an interpreter could effectively delineate a variety of tonal and
textural anomalies on a SLAR image, and he could also consistently identify
(1) bodies of water, (2) drainage networks, (3) aspect and relative steepness of
slope, and (4) watershed boundaries. In addition, in relatively flat areas, delin-
eated boundaries often relate to changes in vegetation type. The types on each
side of the boundary can rarely be identified on the SLAR imagery alone, but
stratifications indicating differences in vegetation type and condition can be made.
Basic map information showing homogeneous terrain features can be coupled with
supplemental data derived from other sources to produce preliminary maps and
statistical data about the vegetation resources.
138
Subtask 2. 5. 3.3
THE POTENTIAL OF RADAR FOR SMALL SCALE
LAND USE MAPPING*
F.M. Henderson
Although many studies have employed SLAR imagery for one purpose or another,
no one has analyzed its ability to create small scale general land use maps over an
extensive area containing a variety of geographic settings. This study attempts to
create such a map. Specifically, the practicality of Side-Looking Airborne Radar
(SLAR) to study and delimit general land use regions for small scale maps will be
investigated. This study will analyze an imaged area from eastern Minnesota to
northern Utah to test the hypothesis that SLAR imagery can be used to create small
scale land use maps.
The study area consists of an area approximately fifteen miles wide and 1500
miles long stretching from eastern Minnesota, across South Dakota andwyoming to
Northern Utah. The area was flown by Westinghouse on October 10, 1965.
This imagery is the longest traverse available and transects a number of agricultural
land uses, soil types, vegetation communities, physical land forms and topography,
and climatic regions.
A map delimiting "SLAR land use regions" has been completed using the SLAR
imagery without a priori knowledge of the study area. That is, no ground work had
been done nor was the interpreter familiar with the area. Land use regions were
delimited solely on their appearance on the radar imagery. As land use borders were
drawn, criteria for their selection were recorded. Figure 43 illustrates two SLAR
land use regions and describes the decision making process involved in identifying them.
To be considered a region, the combination and inter-action of the five factors
had to produce a homogeneous area that was different from adjacent areas. Initially,
twenty-one different land use regions were delimited. The characteristics of each
region in terms of the five factor criteria were recorded. The next step will be to
compare the characteristics of the twenty-one areas and to consolidate regions with
nearly identical criteria.
*This report represents a new effort for which only the formative stages are complete.
139
Topography: Region A is one of gentle relief and appears almost level while Region B
appears much more dissected and eroded. Region B also contains plateaus and more
obvious relief features.
Natural Vegetation: Almost none is visible in Region A as the entire area is culti-
vated except for the stream banks. Some texture coarseness in Region B might imply
a low vegetation pattern on many of the non-cultivated slopes.
Settlement : Although scattered there are more farmsteads in Region A and they are
more regularly spaced than in Region B.
Field Geometry : Region A obviously contains a rectangular field pattern with some
borders conforming to terrain limitations. In addition the pattern is virtually ubiqui-
tous. Region B contains fewer fields, and the general pattern is one of larger fields
and natural pastures with some ponds confined to plateau areas.
Transportation : A rectangular system is visible in Region A with major and minor
arteries. Power lines are also visible across the image. In contrast, there are fewer
roads in Region B, and they are less visible except on the plateau areas. In addition
only parts of one power line are visible.
Figure 43
140
In the follow-on phase we will devise an interpretation key using the criteria
originally employed to determine a "SLAR land use region." This key will then be
tested using two groups: (1) "non-remote sensers" — geographers who have a back-
ground in mapping and land use terminology but not in remote sensing or photo inter-
■N.
pretation, and (2) "remote sensers" — geographers who have experience in working
with land use regions, remote sensing and radar interpretation. The degree of accuracy
of each group in detecting land use regions using the key will then be computed.
Those areas identified correctly and incorrectly will be examined to determine what
possible factors aided or hindered identification. Last, an evaluation of SLAR for
mapping land use regions will be made based on (1) the success of the interpretation
key, (2) terrain phenomena visible, (3) training required to analyze SLAR imagery,
and (4) possible uses for maps of "SLAR land use regions."
If land use regions comparable to those compiled by Anderson (1970), Marschner
(1959), or Austin (1965) can be consistently delineated from SLAR (using the same or
different category criteria) such maps could be updated much faster and kept current.
A change in land use such as the introduction of irrigation to a grazing area, or the
progress of certain land reforms could be easily monitored. The need for small scale
general land use maps exists, but no one knows if SLAR can be used to fill this need.
It is hoped this study will at least mitigate some of the voids between theory
and proven fact. First, by analyzing a large, extensive area, it is possible to study
radar's consistency between environments. For example, is the same level of informa-
tion and detail identifiable from region to region? Second, by comparing the "radar
land use regions" with regions already delimited on existing maps, the degree of
compatibility can be derived. Third, the development of a radar interpretation key
will test the validity and consistency of radar land use regions created by various
interpreters.
141
Task 2.5 APPENDIX A: An Automatic Interpretation Program Derived from
a Radar Interpretation Key
r
Percy P. Batlivala and J. C. Coiner
Purpose: To use radar data that has been digitized to interpret crops by converting
a human based interpretation key into a form that can be employed as a
computer algorithm.
Method: The radar image is digitized using 50 micron spot size. It is stored on tape,
and then printed out as a computer map using the KANDIDATS Picture Program.
The fields to be interpreted are then selected and called out onto a separate
tape with each field having two files (one for the HH polarization image and
one for the HV polarization image). The field is called from the tape with each
file being reduced by mean and mean assignment to a single integer value
between 1 and 5. The HH file integer value then is used as value I while the
HV integer value is used as value J. These integers (1 and J) are used to
define allocation within the matrix A, which consists of crop labels to be
applied to the field (i.e., if the HH value is reduced to 3 and the HV value
to 3, the location in matrix A is then defined as Ay and contains the label
'£jrain sorghum". This crop label with field identification is then printed out.
Figure 44 is a simplified flow diagram of the program, and Figure 45 is a
pictorial representation of a preliminary label matrix derived from an image
interpretation key.
Subroutine Name: DIKEY
Calling Statement : CALL DIKEY (I LABEL, N LEVEL, IARRAY, NROWM, NCOLM,
II, 1MIN, IMAX, IDUM1, 1DUM2, NIMAGE, NFILE, IFILE)
Argjments:
1LABEL is the input array of labels used for classification.
NLEVEL are the number of levels used for classification. ILABEL is
dimensional (NLEVEL, NCEVEL)
IARRAY is the field to be classified.
142
NROWM
NCOLM
II
1MIN
IMAX
IDUM1
IDUM2
NIMAGE
NFILE
IFILE
is the largest number of rows on any field,
is the largest number of columns on any field.
is an integer value from 1 to 10 depending on the type of processing
to be done.
is the minimum brightness level on any field,
is the maximum brightness level on any field.
are scratch vectors of size NIMAGE.
is the number of fields to be classified.
is the file code used to designate the file on which the images
are placed. The images are read a line at a time. An "end of file"
mark must designate the end of an image,
is the file array of dimension NIMAGE.
143
FIGURE 44. DIGITIZED IMAGE FLOW DIAGRAM FOR SUBROUTINE Dl KEY
OF POOR
' page js
Quality?
144
FI GURE 45 PICTORIAL REPRESENTATION OF A PRELIMINARY LABEL MATRIX
DERIVED FROM AN IMAGE INTERPRETATION KEY.
HH
DARK 1
2
3
4
LIGHT 5
HV DARK 1 2 3 4 5 LIGHT
•NO CROP ASSIGNED
RECENTLY
Tl LLED
FALLOW
NCA*
NCA*
NCA*
FALLOW
FALLOW
WHEAT
WHEAT
NCA
NCA
NCA
SORGHUM
ALFALFA
NCA
NCA
NCA
NCA
ALFALFA
SUGAR
BEETS
NCA
NCA
NCA
NCA
SUGAR
BEETS
02-22-72 15,564
DIMENSION ILABEL(5»5), IARR AYi i33j1A9 ) » 1 D"M { 3 A , 2 ) , I F 1 L E £ 3 ^ )
C Y a I N LINE
VlSVElsS
VR0WMn33
VJC0LM = 1A ?
VI MAGEs 34
HlMfO
I M AX=255
V F 1 L £ = 1
! I s 2
CALL DIK = Y( ILABLLi ‘ iLivEL. I A ri? AY,NROwM,NCOLM, u , I MJN,
2 1 yak, int'H, v image ,nf up, i r i l e '
iiTOP
= VD
PB1GHAB b
jOF POOR pJALTHi
146
ZM
^ So
•c'G“5I j 3ni A \i I SS^TD IjNV H13BV1I * XI H i V ^ AfiODgTVO^ri^OSV'-S
•— * — — — • — * ■• " 1HM if,c: I
hv3wI*{T »k vt>i> >^rai on
f-JV3wl s ^ 2 JWPCi 1
. . , ••• - - - 3DVKrK-V«y W = j>i>>
OH Oi 19 « 39vwlN’3“*kOO<>i).it
«o»® T>BO^»Be'#'#!rsTr5'tt » # v »r» 005 »t« MWTfrsTfv# « * 5 °*
•ATdAtiJaeSab S3‘JVHI 0 3Z I faV'-Oc
A h Q NTT M"H 3 H 1 i"0 ■ SR V 3TT"3'H I WHor AvIidV "AKWrCl M I a * C I S
■— TT{V3Wl • 39 VK n '*>>>’ T
N VsNS* XV hj* NJJ2 lMI*rt103N«WM0aN* hOdR , 10 0N:i 3A = >; V ViMd V n N J|hP _ 1 ‘ ^
3HM1NC- III
rd3H"3 JTSog? 1TVD ( O ’lTUT-I.'Vl AWVI * i* otP u
- — nr oaNsMCt'N
(£1 * »«noaN. 'xffil « .■MClfcN. *XT)iVMfcCj 50T
• -TC073 *-(5i (TgN "(9D‘T *9 J 3 ITfc r
n^lb'isICDN
... _ - • - rrHyVMdsN
dUiS ( u 'IT nBVTT'H>X>>"3laVlB) JL
3GVwIKn = kX>i>' T c:g
> * fl
— TiTtr'i'rr'SN lnvsa 'it' vis
39VW'lN*S = = 3Vhlh
3 f i N 1 1 N C 3 I*
— 1 1 j I * c J 1 1 > 5T I 3 I
< El •X9>ivkdCj: 201
- - Tt3+ ’ fJOT*S >ev2fc
3‘jVwlN * * = a I 1 ** Oil
• •• - < C9Vt*Xe)S ) IVWfCd t» U T
M3a31n*t*i , <i3AaiN , t*r* ( r* i )“3 bvii ) > no in >31 iar
■C//»XIM1VW ABO021V3 3Hl*‘//XT>lVHfCi fiUX
(SflT*9>31Ifa. v
-trrrrrrvKecj uu
3
3
3
3
J
3
3
nqA3iN‘T»i , naA3iN , T«r , (rM>i3ev , iin txox*s»av3fc
*^**************o^^***^******«**********»»*«******* ft * w * <, **“ tt * 4l0< *
'aaiawvavd irdKi v_sv cnsn si om
a3NIwHl'IWa'd Sl.l39Tirrx-IaTVR- m53i*3 ^i'AferSVwI bJCVfe
a = ii3«dd3iNl AllVnSIA JO s3ll01NH33i S JSfl 3* 1 if Ot tins SIM.
*4^i^J^.*^***4***********^*««
IrSVlfc UrEalM
( TTMtH' (33VWINJ3H jl ’ Oti
«=»0V H lN»Hnai«(13A3lN , 13AaiN)l3BVllMW103N , Hf'0BN)AvBfcyi (S Q i S N 3 S I Ci
1 I E N I N I^NtdiX;
( dll JI * j*UJN '5SVH IN 1 HfQ I 'xvhPMhll
- r T prns^N 'Td^rrrr-TBS Trmsirmrrrrc-i? et s
3
3
3
3
3
S9 ( j* 6 1
ZfZZ-ZJ
g<\:
N^riEt
31-MiM.: L
( «, [ 1 i s N ? 5 i 't! 1 i *C-N G 1 3 I J rt h « * X T l
*/Sl**sN»3k i 'il'i *GN 111313 Mh I'lCUjvhtC;
< 2 • u> * > i-j'nn r r < w y'x > 3*i I j I M l * ?,J h> ) t-na I ' ( u>i> ) 31 i 3 1 (io’i'viam*
( 2 • h>» ) wnc i • < ) 31 1 jl * ( I \,yir ) «rui I v ( ) m u i ( tu x • ) -i 1 1- ••
a t JV*IN , I = rt>'>' i Lu
'ii'insHfc 3Hi hONHc INV iMtc -)
qriMiNC: i.
(ui
1 itj . '<**', v «iv ll3iilSSVlO SVM i 1 £ I ' * M SEhHi, 3Dvi„l 3Ni t ' CKC > i VKfcL.: £l'T
(rr* i mafcV'ii • < 1 1 i ) j*u ji (eot ' v)5ut:/
( 2 • n I jGffl I = r f
(T‘ in ) n n i = 1 1
• ■ ' ■■ h 3 vwtn'*t='i i r ? cc
*
CL “22“2G
6 n
wms&WteW
si aovd tvnjqiso
o
N0li»TIdW03 SIHi Ab Uasr A M 0h3h jC S 1 J d 0 « iSLUi
N«ni3fc
inv 3 mI ' aavHiN^iXxx'Nvswa* iNins nv3 cz'ca'ii) ji
t N V 3 t-i a * X ^ W I *r,lhi HsAalwUMfiS TWO U'DH'IDjI
■j i) v « 1 ia S ,*i b v) 0 d Cii U 53 I 1 riOIJ. d^‘ 5)1 L P4 1 i3 b G *» v *3
< l ' 9 J • i = N V dh 3 < ‘XTUVK’kC: 1 0 l
iv V 3 W □ HOI *9) 3lli5*
< < ’1<J3N*M(;ai< ) 1 VQlj > / C C WH\) i VO“ i ) = N
( r M ) Avabv i + wnr-,s^n,\ i
io:iM‘t = rtOj
f-OfcN*T*IXCC
Osh ll N
JUVl-,1 Shi JO NVEW 3hi 0 ^ I - 0
ft ft ft ft ft « ft J
<HlOON ' WhOtiN ) A vat v 1 NO I SN = fc 1 C
ft ft ft ft* ft*« ****** *** ft *o «*« **««*•**•♦'**** »•*»»««**«.*.«•».♦*#« «« * *#« J
’qMirobfPt o
3i^5t)oCJeder V Oi 11v3 3 hi Sti3jSKfai * l I * NOIljO 3hU Ol ENlflbCC-v D
N3HJ. CivV 3‘jVwI H0V3 JO NV3.J dHi SQnU 3Nl IT Qt 0hS SIH J
5 N I 'aaVwlN'^XVX'NVaWt'l
aVa 1 ‘Nidi ' l 1 *W1C3lv'WMOdN*1o3N *VUdN * *1 JAjlh ' i*fc H v I ) N V 3hT 3N I J.rCfef Pi
5 6 a * £ T
tL“£ d’20
OSl
NUllVlidWOa S I Hi ah Cl = sr AHOhak JC SUbuR H 99 !lZ
- CK =
IVfcfUSi:
TCTWn « n * 3 h a « = n » * x t u *irt Cj tut
N V 3 W 0 ( T Of * V > ai I ty f
aiiMiNCs i
£ Ci CC
l = M = hj; z
t ci c*:
Z~qI 01 (13 ATTv •• I V k ¥3 h r? J i
l«iMs 1 = Ar"lv
iaA3^N*tsi x cn
>***iH*#*»***o»«c*tttt«*«a**««»»e*«##«
'nV 3 k 9 ' jo aTiiv'/i "dMl 3 ivmvA 3 -at> yci\
Al i N / TN T wr-xnn> = AM
>**«*#*»»*#»#***#*tt#*#«»»tt«#*««**«ff
" 37 I s ’ 1?Aa = ii\ i 3hi ClNlj
•noiiTZiiKivno ivAfcgiNl ivpCs
NV3k a>u s 3 z I i n v ncj aimrutans sii-i
66 ‘i’iT
t L'ZZ-l'i
N0li*HdK03 SIHj. AS ci= sr AMUk“A jc S 3 all, I tG9k:i
■ ■ - - "• "' " Ck =
Ndn=d ^ u t
(> - fvV^WI {‘OS<: , lVN^3w9 , CJNr ,, 06T , ia*NV = hE)JI
p s !•; V 3 W 1 < 1 li6T ’ i'1 1 N V 3 t-i i) 1 Qr ,- V ' ‘ 0 t>T * ’is ' M VrfcE ) j I
S' s I'-V^Wl ( f GfrT ’ i*T nV 3*S ‘ Qi‘. V i *§UT * 19 * NVakS ) • l
Z * N»3wl t *£ut ’.i'l’.MVjWC ■ £1 N V • * JZ*i*J*NV='V.£) J I
t = f'l V 3 >; 1 ( ' Oi* il'NVik^'Qf.jV' *d'l9'NV = ^2) Jl
1VAj*3lNI a 3 m hjadiSQHttd Mi i M bBSaiNl Ol HH SifcaANC: J
2cT Gi C'G
c, = r, V gWI < , 062 , n , N»3wS , QNV , *04T*ig , N»swEJJi
> = K V 3 WI r 0 £T *iVfcV3*3*ilNV ' 'sTT 'io'NVshC )-'i
£ * Nv3uI t J\V3 WC*UNv‘ ■i»C*x£i , N» = wEJ J I
Z « ,\V3wI" \ ‘ uS'l^i^HS ' UNV * ‘ 5$ ' IE)’ N* =aS > J I
T = N V 3 H I ( *59 * i VNVikS* GNV 1 ' U ’ i 9 ' N V SkS ) £\
•~3NlWd3i3 r J3bd ONI SI! {j3t)3iM Qi NV3|* AH Si d J A A C 3
OrvIcODMa lVAd = iM tzcCtc 2
Gi SiC^alti Gf-.v 3‘llj AH dG HH NV Si 311 j - 1 S;nIl^d = ii 2 G G
Tot Gi C 3 ( 3 0 T h I N * 3“*v»>>>) - 1
S3*llJ Ah UMv hh 3hi JO SNV3 h 3 Hi tOj ST^AddlM G
SbVlG QdN lHa3l9U3«d AfdM 01 LdSH SI 3Mil*0tflnS SlHi 3
(ri?3hl '3yvUi-;^>f^i^3h5V liiiVS 3Mirc«H*£
(. l - iZ-Z'* y
i a • c t
APPENDIX B-l
Differential Scattering Cross Sections (a v $ 0 in dB)
for Silty Clay Loam Soil
Surface conditions: rough , dry (15% h^O by vol)
r=?Eo
LOOK ANGLE
IN DEGREES'
“ ‘ — - —
P 01
IN QWZ
0
10
20
30
50
65
4.2
•4.10025
-8.53393
-4 . 33096
-4 . 97766
-6,14679
-10.37384
H
4.6
-4.99192
-2 . 42565
0.27752
-5 . 36902
-7.93545
-7.58954
H
5 . 0
-4.90254
-4.53625
-3,63303
-4.27942
-7.00813
-8.29152
N
5,4
-1 . 3 3 R 2 7
-0.77191
-7.56866
-3.21495
-9.17605
-9. $9099
H
5.8
-I . 30450
-2, 73513
-3.03483
-3.18102
-3.34843
-6.69740
M
6.2
0. 18875
-2.74435
-9.54153
-4 . 18765
-6.03851
-8.42396
H
6.6
1 . 64749
-4 . 28613
-6.08275
-7.72880
-5.24253
-9.16726
H
7.0
1 .56054
-1.37307
-3.66969
" -4.81570
-8 . 47367
-8.44064
H
7.4
5.41731
-5,51633
-1.31290
-3.45883
-12.24393
-7.75644
H
7,8
5.74957
0 .31593
-0 . 98059
3.87351
-6.02262
-7168440
H
4.2
- 4 . 1 0 n 2 5~
-4703393“
-5,83096
-5.97766
-9.64679
-11.87384
V
4.6
99i 92
-0.92565
-5 . 72248
-6 . 36902
-6 . 33545
-3.58954
V
5.0
" 5, 4 i) 7 5 4
-5 , 3 3 6? 5
3.96697
-4.27942
-4 . 00 8 1 5
-7.79 t 52
V
5.4
-7. R 38 2 7
-7. 77i 9{
-6.06966
-7.21495
-10 . 17605
-8.99099
V
5,8
-3. 30450
-10.73813
’-6.03483 “
-5.18102
-8 . 34848
-7 ♦ 1 974 0
V
6.2
-0.3ll25
-7 , 74436
-10.54i53
-2.18765
*l2 . o385l
”8 . 92396
V
6.6
-0 .35751' "
-3. 786i5
-7.58275
-7 . 22880
-6 . 24253
-11.16726
V
7.0
-0 .43946
-5.87J07
-2.66969
-9.31570
-13.47367
-14.44064
V
7.4
? . 4 ! 7 3 i
-4 , ?i630
- 4 . 9 1 2 9 o
-1 . 45883
*12.74393
-£2.75644
V
7.8
5 . 74957
0.31593
-0.48059
-4.12649
-2.0226 2
-8 . 08440
V
page is
APPENDIX B-2
Differential Scattering Cross Sections (crv s © in dB)
for Silty Clay Loam Soil
Surface conditions; Uneven, dry (13% f^O by Vol)
FREQ
IN GHZ
0
10
LOOK ANGLE
20
IN DEGREES
30
50
65
POL
4.2
-4.60025
-3,03393
-7 . 33096
-0.97766
-4.64679
-7.87384
■H
4.6
-0.491 92
-8.42563
-0.72248
4 . 63098
-2.33545
-7.08954
H
5.0
0 . 59746
-4.33623
-4.63303
-4.77942
-2.00815
-6-. 79152
H
5.4
3.661 73
-2.271 9 {
-7,56966
-3.21495
-5.17605
-7.99099
H
5.8
2.1 9550
1 . 76197
-4.03483
-3.68102
-3.34843
-4.69740
H
6.2
4 .68875
3,25514
-1.04153
-5.68765
-2,53851
-6.92396
H
6.6
0 . 64749
-3. 2B613
-12.58275
-5.22880
-4.74253
-8.16726
H
7.0
2.06054
-0 . 37307
-5.16969
-6.81570
» 6 * 9736 7
-8.44064
H
7.4
5.41731
2 . 98370
-0 . R1290
-8.45883
3.87351
0.25607
-3.25644
H _
7.8
3.24957
1 . 31593
3.01941
2.47733
-2.58440
H
4.2
2.39975
-1733393 ’
0,66904
-1.47766
-2.64679
-9.37384
V
4.6
6.50808
4.07440
4.27752
-1.36902
-2 . 83545
-5.58954
V
5.6
5 . ^97 4 6
1 . l638fj
— 1 , 6 3 3 q 3
-2.7794?
-2. OflQl 5
-4 .2*1 5 2
V
5.4
2 . 1 6173
- 3 . 2 7 1 9 {
-0 . 56366
-8.21495
-8 . 17605
-6.99099
V
578
-0.30450
-3 .23813'
-6.53483
- 1 . 68102
-6 .84848
-4.69740
V ’
6.2
3. {8875
0.?55'<4
-2. 54t53
-0-18765
-6 . o385i
-3.92396
V
676
1 .‘1 4749
-4 .2961 3
-3.58275
-6.72880
-3.24253
-12.16726
V
7.0
-0 . 93946
-9.37307
-14.16969
-11.81570
-14.97367
-9.44064
V
774
r - 4l73 —
-5‘, 5{ 63 5 '
"-10 . 312^0
-0.45883
-4.74393
-5 , ?5644
V
CO
N
3.74957
-0. 1040?
-0 , 98059
-0.12649
-1.52262
-1. 0844 0
V
APPENDIX B-3
Differential Scattering Cross Sections (crv s G in dB)
for Silty Clay Loam Soil
Surface conditions: smooth, dry (15% H 2 O by Vol)
FREQ
LOOK ANGLE
IN DEGREES
IN GHZ
0
10
2 0
30
50
65
4.2
2.39975
0 .96602
0 . 16904
-3 . 47766
-4.64679
-7.87384
4.6
9 . 0 0 80 8
" -1.42560
-2.22248
-3.86902
-3.33545
-7.58954
5.0
2.09746
-6.33620
-9.13303
-2.27942
-3.50915
-9.29152
5.4
5.66173
-6. 271.91
-2 . 06966
1.2B505
-0 , 67605
-5.49099
5.0
1 . 69550
-8.73813
-2 . 53483
0 .31898
-2.34843
-6.-69740
6.2
7.1 8875
0.75514
1,45847
5.31235
-7 . 53851
-4 .92396
6.6
1 . 1 4 749
3.71390
0.41725
-6.22080
-2.24253
-7.66726
7.0
0 . 06054
3,62693
3 . 66^69
-4.81570
•*5,9 7 367
-5.94065
7.4
0 • 91731
-0 . 01633
-2.91290
-7,45883
-4.74393
-2.75644
7 . 8
2 . 24957
-0 . 18402
3.51941
0.87351
-0 . 52262
-0.08440
4.2
4 . 39975
3:96602
2.66904 "
-6.47766
2.35321
-1.87384
4.6
3 .00808
-0.92560
-1 . 72248
-2.36902
-2.33545
-4.58954
5.0
"l. 09746
0 . 16305
-2 , 633q3
-1.27942
“1. 5 O 0 1 5
-4.79 t 52
5.4
-10 . 03827
- 7 , 7 7 1 9 {
-4,06866
-0.21495
-3.67605
-3.99099
“ 578
-0 . 80450
-4 .73813
1.96517
3.31898
0.15152
-6.69740
6.2
-D. 3n"25
-5.54466
6 . 45847
8.31233
-2.53851
-1.92396
6.6
27 . 64749
3.21390
- 4,58275
-2.22880
-3.24253
-5.66726
7.0
-7 . 93946
2.62693:
-7.16969
-8.31570
-7.47367
r 4. 44064
774
— . 5¥2'6-9
-8.5X633
5.l87 10
-4 , 95883
— 5,74393
-1,75644
7.0
0.74957
9.31595
5,01941
2.37351
3.9773S
-2.58440.
APPENDIX B-4
Differential Scattering Cross Sections (ctv $ G in dB)
for Silty Clay Loam Soil
Surface conditions: rough, wet (31% by Vol)
FREQ "LOOK ANGIE IN DEGREES’" " ” POi
IN GHZ
0
10
20
30
50
65
-
4.2
-10.60025
-2.03393
-4.33096
-6.97766
-1 . 14679
-3.37384
H
4.6
-7 . 49i 92
-0 . 92560
-2.22248
-3 . 86902
2.16455
-5.08954
H
5.0
-1 . 90254
-2.83620
^ 2 . 1330 3
-0.77942
0 . 49185
0.70948
H
5.4
-7.83827
-4 . 27 1 9 {
-7.06966
-10.21495
-2.67605
0.50901
H
5.8
-3.80450
4.26187
-2.53483
-4.18102
-0 . 94848
2.30260
M
6.2
-5.31125
-2.24485
-5 . 04153
-1 . 18765
2 . 96149
6.57604
H
6.6
-8.85251
-6,78610
-3.08275
-5.22880
-2,74253
-8.16726
W
7.0
-9.43946
-1,37307
-2.16969
-3.31570
-5.47367
-9,94064
H
7.4
-li .58269
-1 . 01630
-0.81290
-9.45883
-9.74393
"12.75644
_w __
'sj
CD
-0.75043
1 . 31593
1.51941
-1.12649
-1.02262
-4.08440
H
4.2
-0.10025
" 07 9 6 6 P 2
-3 . 33096
-1.47766
-1 . 64679
-2.87384
V
4.6
4 .00808
3.57440
-3.72248
-3.86902
2.16455
-0.08954
V
5.0
4f,'Qo?54
-3 . 33620
0 . 36697
1.22058
1.99185
- 2 . ?9i 52
V
5.4
-9.83827
-3.77l'9\
-8 .06866
-9.21495
-7.17605
-2.49099
V
5.8
-6 . 80450
-2.23313
-3.03483
-4.18102
-1.34843
1.80260
V
5.2
-3. 8ii 2 5
.4.74436
1.95947
1.31235
1 . 96149
7 . o76o^
V
6.6
-4 . 35251
’ -9.79610
-5.08275
-5.22880
-1.24253
2.33274
V
7.0
-4 . 43946
-11 . 37307
-9.66969
-8.81570
-8.97367
-4.44064
V
7; 4
-X. 5 8 ? 6 9
-8.01635 -
“10 .3i29o
-9.95803
-9,74393
-13.75644
V
7.8
6 . 74957
4 . 31593
1 . 01941
1.37351
2.97738
-6.0844 0.
V
APPENDIX B-5
Differential Scattering Cross Sections (av s G in dB)
for Silty Clay Loam Soil
Surface conditions: Uneven, wet (29% h^O by Vol.)
FREQ
LOOK ANGLE
IN GHZ
0
_ 10
20
4.2
1 . 99975
-0 . 53399
6.36904
4.6
i . 50808
-0 . 92560
2 .27752
5.0
-0 .40254
-3.33620
-1,13303
5.4
-5.33827
-3 . 77191
0.43134
5.8
0 . 19550
-2.23^13
0 .46517
6.2
5 * 6 8 R 7 5
12.25514
0 , 5415 3
6.6
3.64749
5.71390
-2.58275
7.0
3.56054
3.12693
-2.16969
7.4
5 . 91731
1 . 9837 0
-3,01290
7.8
12.74957
-0 . 6840?
1.51941
4 . 2
-10 . 1 0025
~- 875 3 39 3"
“-1 . 33096
4.6
-1 . 491 92
-0 . 92560
-1.72248
— _ * / / A H
5". O'
1 . n9746
3.16385
0 , 86697
5.4
-10.33827
-8.771 9{
-9.56966
5.8
-5.30450
-4.73R13
-5.03483
6.2
n . 68875
-4 .74485
-0 . 0^153
6.6
-VTR5251
-4.78610
-5.08275
7.0
^-9 . 93946
-9.87307
-9.16969
7 - 4
-10 .002^9
- 8 . 5 i 6 3 0
~ -10. 31290
QO
•
N
1 .24957
0,01595
1 . 01941
2.97766
3.13098
■1 , 77942 _
■1.21495
•3.1B102
2.31233
•1.22880
3.18430
■5.95883
■0.62649
0 . 02234
-0 . 86902
1.72058
>7.71495
■4 . 18102
0.81233
-5.22880
-9.31570
-9.45883
0.373 5
4 . 64679
2.16455
3.49185
8.67605"
0 . 34848
0 . 53851
3.74253
3.47367
3.24393
3.97733
•2.64679
■ 6 . 33545
3.99185
•7.17605
■2. 84648
2 . 461 49
-0.24253
-8.47367
-9.24393
1,97738
-7.87384
-2. 08954
-0.29152
"il. 49099
1.30260
-5.42396
-6 ; 66726
-5.94065
-5.75644
-4 . 08440
APPENDIX B-6
Differential Scattering Cross Section (av s 0 in dB)
for Silty Clay Loam Soil
Surface conditions: smooth / wet (33% H 2 O by Vo I .)
FREQ
LOOK A MGtE
IN DEGREES
IN GHZ
0
_ 10
_ . ?0
30
50
65
4.2
-l . 60025
1 . 4660?
-5. B3096
1 . 02234
-2.14679
-6.37384
4.6
7 . 50 S 0 8
'6.57443
4.7775 2
7 . 130 98
1.16455
-1.00954
5.0
3 . 59746
3 . 16383
3.36697
-2 . 27942
; -0.5Q815
-1.29152
5.4
1 .1617 3
2.7250?
2.43134
-5.71495
-3 * 17 6 0 ^
-3.49099
5.8
3.69550
5. 261 87
3,96517
1.31898
-0 . 84848
-5.19740
6.2
-3.81125
-Y.' 74486
5.95947
-0.68765
-1.53851
-3.92396
6.6
3.64749
6.21390
-2.08275
-5.22880
-2.74253
-4.-16726
7.0
5 . 06054
9 . 52693
2.83031
0.66430
-2.97367
-2.44064
7.4
5.41731
11 . 98370
-0.31290
-1.45883
1.75607
-3.25644
7.8
5.24957
12.31593
3 .01941
1.37351
-0 . 52262
-3.08440
4.2
5.89975
3796602 “
2 . 66904
0.02234
-2.64679
-5.373B4
4.6
10. 00808
6.07440
0.77752
7 . 63098
0.66455
-1.08954
5.0
1 . 59746
'2‘. 1638 5“
3.36697
-1 . 27942
1 . 99185
-l.?9i52
5.4
-0.33827
-4.77191
-0 . 06366
-0.71495
-8.17605
-4.49099
5,8
5.19550
-5.73813
0 .46517
0.31898
-3.34848
: 3. 19740
6.2
- fi • 5 1 1 2 5
7 , ? 5 5 1 4
-2. 54i53
-0.18765
-0.53851
-1.42396
6.6
-3.85251
2.21390
3.41725
-5. 728B0
-3.24253
-7.66726
7.0
-2.43946
-0 ,37307;
0.83031
-1.81570
-14.47367
-4,44064
7V 4
5 . 4i73i
5 , 9 8375
1 , 687io
1 .54ii7
~2 , 24393
-3 .25644
7.8
11 . 24957
9.81599
2.51941
2.87351
3,97733.
-1.08440
Figure 46
TYPICAL CROP HISTOGRAM FOR 7/66 (Ka-BAND)
CORN GRAIN SORGHUM
SUGAR BEETS
PASTURE
Task 2.5 APPENDIX C: Crop Signatures for a ’'Typical Standard Farm
Task 2.5 APPENDIX D; References
Anderson, J.R., 1970, "Major Land Uses," National Atlas, Washington, D.C.,
pp. 157-159.
Anderson, P.N., 1971, "System Hardware Specification Manual/IDECS, " CRES
Technical Report 133-26 , University of Kansas Center for Research, Inc.,
Lawrence, Kansas.
Artsybashev, Y.S., 1962, "Study of the Spectral Brightness of Some Landscape
Elements for Interpretation of Ground Water on Aerial Photographs,"
U.S. Army Foreign Science and Technology Center Technical Translation
FSTC-HT-23-353-68, USSR.
Austin, M.E., 1965, "Land Resource Regions and Major Land Resource Areas of the
U.S.," USDA Agriculture Handbook, ^296, pp. 82.
Bachman, K.L., 1965, "Can We Produce Enough Food? 11 in World Population and
Food Supplies, 1980, American Society of Agronomy, Special Publication
No. 6, 50 pp. (see pages 42-48).
Barr, D.J. and R.D. Miles, 1970, "SLAR imagery and Site Selection," Photogrammetric
Engineering , Vol . XXXVI, pp. 115-1170.
Brown, L.R., 1968, "The Agricultural Revolution in Asia," Foreign Affairs, 46(4),
pp. 688-698.
Coiner, J.C. and S.A. Morain, 1971, "Image Interpretation Keys to Support Analysis
of SLAR Imagery," Proceedings, Amer. Soc . Photog . (Fail Meeting, San
Francisco), paper no. 71-334, pp. 393-412.
Doll, J.P., V.J. Rhodes, and J.G. West, 1968, Economics of Agricultural Produc -
tion. Markets, and Policy. Hoewood, Illinois, Richard D. Irwin, Inc.,
557 pp;
Ehrlich, P.R. and A.H. Ehrlich, 1970, Population Resources Environment: Issues
in Human Ecology , San Francisco, W.H. Freeman, 383 pp.
Fu, K.S., D.A. Landgrebe and T.L. Phillips, 1969, "Information Processing of
Remotely Sensed Agricultural Data," Proceedings IEEE (4): 639-653.
Geotimes, September 1971, p.26.
Grunfeld, Y. and Z. Griliches, 1960, "Is Aggregation Necessarily Bad?" Review
of Economics and Statistics (42(1), pp. 1-13.
Haralick, R.M., F. Caspall, and D.S. Simonett, 1970, "Using Radar Imagery for
Crop Discrimination: A Statistical and Conditional Probability Study,"
Remote Sensing of Environment 1(1), pp. 131-142.
159
Hardy, N.E., J.C. Coiner and W. O. Lockman, 1971, " Vegetation Mapping with
SLAR: Yellowstone National Park," Proceedings of XVII Symposium of
AAGARD Electromagnetic Wave Propagation Panel on Propagation Limita-
tions in Remote Sensing, June 21, 1971 to June 25, 1971, pp. 11-1 to 11-19.
Heany, D.F., 1968, Development of Information Systems, New York: Ronald
Press, 415 pp.
Henderson, F.M., 1971, "Radar Monitoring of Agricultural Land Use: Some Problems
and Potentials at the Local Leve," Proceedings Amer. Soc. Photog . (Fall
Meeting, San Francisco), Paper no. 71-332, pp. 368-385.
Holmes, R.A., 1968, "An Agricultural Remote Sensing Information System," EASCQN
1968 record, pp. 142-149. .
Holmes, R.A. and R.B. MacDonald, 1969, "The Physical Basis of System Design for
Remote Sensing in Agriculture,: Proceedings IEEE .
Houseman, E.E., 1970, "Remote Sensors — A New Data Source for Agricultural Sta-
tistics,: AIAA, Earth Resources Observations and Information Systems Meeting,
Annapolis, Maryland, Paper ^70-312, 4 pp.
Kansas Board of Agriculture, 1968-69, 52nd Annual Report of Kansas Agriculture.
Kellogg, C.E., 1963, "Potential for Food Production," Farmer's World, 1964 USDA
Yearbook of Agriculture.
Langley, P.G., et al., 1970, "The Development of an Earth Resources Information
System Using Aerial Photographs and Digital Computers," Annual Progress
Report of the Forestry Remote Sensing Laboratory, University of California,
Berkeley.
Lewis, A.J., 1971, "Geomorphic Evaluation of Radar Imagery of Southeastern
Panama and Northwestern Colombia, CRES Technical Report 133-18,
Center for Research, Inc., University of Kansas, Lawrence, Kansas.
Lorsch, H.G., 1969, "Agricultural Resources Information System — The User's Point
of View," Joint meeting of the American Astronautical Society and Opera-
tions Research Society, paper ^Y110.5.
Lundien, J.R., 1966, "Terrain Analysis by Electromagnetic Means," U.S. Army
Engineer Waterways Experiment Station, Technical Report No. 3-693.
Lundien, J.R., 1971, "Laboratory Measurement of Electromagnetic Propagation
Constants in the 1.0-1. 5 GHz Microwave Spectral Region," Terrain Analysis
by Electromagnetic Means, Report 5, Technical Report no. 3-693, U.S.
Army Waterways Experiment Station, Vicksburg, Mississippi.
MacDonald, H.C., 1969, "Geologic Evaluation of Radar Imagery for Darien Province,
CRES Technical Report 133-6.
160
MacDonald, H.C. and W.P. Waite, 1971, "Vegetation Penetration with K-bond
Imaging Radars," Transactions of the IEEE Geoscience Electronics .
Marschner, F.J., 1959, "Land Use and its Patterns in the United States," Agri~
cultural Handbook No. 153, USDA, Washington, D.C.
Mayer, L.V. and E.O. Heady, 1969, "Projected State and Regional Resource
Requirements for Agriculture in the United States in 1980, 11 Iowa State
University; Research Bulletin no. 568 , pp. 372-413.
McCoy, R.M. , 1967, "An Evaluation of Radar Imagery as a Tool for Drainage
Basin Analysis," CRES Technical Report 61-31 for NASA Contract NSR
17-004-003 and Grant NSG 298, 102 pp.
Morain, S.A. and D.S. Simonett, 1966, "Vegetation Analysis With Radar Imagery, "
4th Symposium on Remote Sensing of Environment, Ann Arbor, Michigan,
April 11-14,
Morain, S.A., 1967, "Field Studies on Vegetation at Horsefly Mounting, Oregion
and its Relation to Radar Imagery," CRES Report 61-23, 19 pp.
Morain, S.A. and J.C. Coiner, 1970, "An Evaluation of Fine Resolution Imagery
for Making Agricultural Determination," CRES Technical Report 177-7,
Center for Research, Inc., University of Kansas, Lawrence, Kansas.
Morain, S.A., C. Wood, and D. Conte, 1970, "NASA Earth Observations Survey
Program 90-day Mission Report, NASA/MSC mission 102, site 76,"
CRES Technical Memo 169-4, Center for Research, Inc., University of
Kansas, Lawrence, Kansas, 16 pp.
OSSA (Office of Space Science and Applications), 1970, "Ecological Surveys from
Space," NASA SP-230, 75 pp.
Pallesen, J.E., 1970, "Statistics Inform the Wheat Industry," in 52nd Annual Report
of Kansas Agriculture, 1968-69, Kansas State Board of Agriculture (see
expecially pp. 131-133.
Park, A.B., 1969, "Remote Sensing of Time Dependent Phenomena," Proceedings of
the Sixth Symposium on Remote Snesing of Environment , Institute of Science
and Technology, University of Michigan, Ann Arbor, pp. 1227-1236.
Pendleton, J.W., 1970, "Advances in Crop Cultural Practices," Agronomy Abstracts ,
American Society of Agronomy, Abstract of a Paper PresentecTaM-he Annual
Meetings, Tuscon, August 23-27.
Sabol, J., 1968, "The Relationship Between Population and Radar Derived Area of
Urban Places," The Utility of Radar and Other Remote Sensors in Thematic
Land Use Mapping from Spacecraft, Annual Report, USGS Contract 14-08-
000T-T0848, pp. 46-74.
Schwarz, D.E. and F. Caspall, 1968, "The Use of Radar in the Discrimination and
Identification of Agricultural Land Use,: in Proceedings of the Fifth Symposium
on Remote Sensing of Environment , Institute of Science and Technology,
University of Michigan, Ann Arbor, pp. 233-247.
161
Sheridan, M.F., 1966, "Preliminary Studies of Soil Patterns Observed in Radar Images,
Bishop Area, California, 11 USGS Technical Letter NASA-63.
Simonett, D.S., J.R. Eagleman, J.R. Marshall, and S.A. Morain, 1969a, "The
Complementary Roles of Aerial Photography and Radar Imaging Related to
Weather Conditions," The Utility of Radar and Other Remote Sensors in
Thematic Land Use Mapping from Spacecraft , 2nd Annual Report USGS
Contract No. 14-08-0001-10484.
Simonett, D.S., G.R. Cochrane, S.A. Morain and D.D. Egbert, 1969b, "Environ-
ment Mapping with Spacecraft Photography: A Central Australian
Example, " The Utility of Radar and Other Remote Sensors in Thematic Land
Use Mapping from Spacecraft, 2nd Annual Report USGS Contract No.
T4-'0g-0O0PT0835.
Thompson, L.M., 1969, "Weather and Technology in the Production of Wheat in the
U.S.," Journal of Soil and Water Conservation 24(6), pp. 219-224.
Wharton, C.R., Jr., 1969, "The Green Revolution: Cornucopia or Pandora's Box?"
Foreign Affairs 47, pp. 464-476.
Wiegand, C.L., R.W. Learner, D.W. Weber and A. H. Gerbermann, 1969, "Compari-
son of Multi-base and Multi-emulsion Photography for Identifying Crop and
Soil Conditions from Space, " Southern Plains Branch Soil and Water Conserva-
tion Research Division, Agricultural Research Service, USDA, Weslaco,
Texas.
Willett, J.W., 1969, "The Impact of New Grain Varieties in Asia," USDA Foreign
Regional Analysis Division, ERS-Foreign 275 , 26pp.
162
III.
CONTRACT PUBLICATIONS (NAS 9- 1 0261)
TECHNICAL REPORTS
Technical Report 177-1, "An Analysis of Methods for Calibrating the 13.3 GHz
Scatterometer," G. A. Bradley.
Technical Report 177-2, "Signal Analysis of the Single "Polarized 13.3 GHz
Scatterometer G. A. Bradley, May 1970,
Technical Report 177-3, "Imaged Textural Analysis by a Circular Scanning Technique,
G. O. Nossaman, June 1970.
Technical Report 177-4, "A Regional Study of Radar Lineaments Patterns In the
Ouachita Mountains, McAlester Basrn-Arkansas Valley, and Ozark Regions
of Oklahoma and Arkansas," J. N. Kirk, June 1970.
Technical Report 177-5, "An Analysis of the Effects of Aricraft Drift Angle on
Remote Radar Sensors," G. A. Bradley and J . D. Young, August 1970.
Technical Report 177-6, "Radar Lineament Analysis, Burning Springs ; Area, West
Virginia — An Aid in the Definition of Appalachian Plateau Thrusts,
R. S. Wing, W. K. Overbey, Jr., and L. F. Dellwig, July 1970.
Technical Report 177-7, "An Evaluation of Fine Resolution Radar Imagery to Making
Agricultural Determinations," S. A. Morain, J. Coiner, August 1970.
Technical Report 177-9, "Optimum Radar Depression Angles for Geological
Analysis," H . C . MacDonald and W. P. Waite, August 1970.
Technical Report 177-10, "Synthetic Aperture Radar and Digital Processing," Ph.D.
Dissertation, R. Gerchberg, September 1970.
Technical Report 177-11, "Panchromatic Illumination for Radar; Acoustic Simulation
of Panchromatic Radar," Ph.D. Dissertation, G. Thomann, September 1970.
Technical Report 177-12, "Discrete Pattern Discrimination Using Neighborly
Dependence Information," R. M. Haralick, October 1970.
Technical Report 177-13, "Interim Technical Progress Report Radar Studies Related
to the Earth Resources Program," March 1971.
Technical Report 177-14, "Radar Sensing in Agriculture, A Socio-Economic View-
point," S. A. Morain, J. Holtzman and F. M. Henderson, December 1V/U.
Technical Report 177-15, "Local Level Agricultural Practices and Individual Farmer
Needs as Influences on SLAR Imagery Data Collection," F. M. Henderson,
April 1971.
Technical Report 177-16, "Detectability of Water Bodies by Side-Looking Radar,"
C. Roswell, October 1969.
163
Technical Report 177-17, "A Fresnel Zone“Plate Processor for Processing Synthetic
Aperture Data," G. Thomann, R. Angle and F. Dickey, May 1971,
Technical Report 177-18, "Geoscience Radar Systems," G, Thomann and F. Dickey,
May 1971.
Technical Report 1 77“ 19, "SLAR Image Interpretation Keys for Geographic Analysis,"
M.S. Thesis, J, C. Coiner, April 1972,
Technical Report 177-20, "Multi-Year Program in Radar Remote Sensing," R. K,
Moore and J. C, Holtzman, August 1971.
Technical Report 177-21, "Evaluation of High Resolution X-Band Radar in the
Ouachita Mountains," L, F, Dellwig and J. McCauley, August 1971.
Technical Report 177“22, Being Completed, Author: G. A. Bradley.
Technical Report 177-23, "Soil Mapping from Radar Imagery: Theory and Preliminary
Applications," S . A . Morain and J . Campbell , March 1972.
Technical Report 177“ 24, Being Completed, Author: N. Hardy.
Technical Report 177-25, "Terrain Roughness and Surface Materials Discrimination
with SLAR in Arid Environments," H. C, MacDonald and W. P. Waite,
January 1972.
Technical Report 177-26, Annual Report.
164
TECHNICAL MEMORANDA
Technical Memorandum 177-1, "An Analysis of RF Phase Error in the T3.3 GHz
Scatterometer," G. A. Bradley, November T 969,
Technical Memorandum 177-2, "Mathematical Theory of Filtering Program,
R. M. Haralick, December 1969.
Technical Memorandum 177-4, "Informal Log, 13.3 GHz Single-Polarized Scat -
terometer, 400 MHz Dual-Polarized Scatterometer, Mission 119, Argus
Island, Bermuda, 19 January 1970^“ *27 January 1970," G. A. Bradley,
February 1970.
Technical Memorandum 177”5, "Principal Component Analysis," R. M. Haralick,
April 1970.
Technical Memorandum 177-6, "Informal Log, Mission 126, Pt. Barrow, Alaska,
G. A. Bradley, June 1970.
Technical Memorandum 1 77-7, "Frequency Averaging for Imaging Radars,"
G. C. Thomann, June 1970.
Technical Memorandum 177-8, "Informal Log, Mission 130, Garden City, Kansas,"
J. D. Young, May 1970.
Technical Memorandum 177-9, "Informal Log Mission 133, Garden City, Kansas,
Site 76," G. A. Bradley, August 1970.
Technical Memorandum 177-10, "Ninety“Day Mission Analysis Report, Mx 108,
DPD-2, Side-Look Radar, Pisgah Crater, California," L. F. Dellwig,
July 1970.
Technical Memorandum 177-11, "Correlated Averaging to Enhance Radar Imagery,"
R. W. Gerchberg, September 1970.
Technical Memorandum 177-12, "Analysis of Sea State Missions 20-60,"
J. Young, September 1970,
Technical Memorandum 177-13, "A Note on the Antenna Beamwidth Term Used in
the Scatterometer Data Reduction Program," J. D, Young and G. A.
Bradley, October 1970.
Technical Memorandum 177-14, "Mission 126, 90-Day Report, Pt. Barrow, Alaska,"
G. A. Bradley, April 1970.
Technical Memorandum 177-15, "NASA Earth Observation Survey Program 90-Day
Mission Report, Garden City, Kansas, Site 76, Mission 133, 9 July 1970,"
W. O. Lockman, L. T. James, J. C. Coiner, December 1970.
Technical Memorandum 177-16, "Satellite Radar Power Calculations," G. C.
Thomann, December 1970.
165
Technical Memorandum 177“ 17, "Informal Log, Mission 156, JOSS II, North
Atlantic Ocean, Site 166, 8— 19 February 1971," G. A. Bradley, Marcy 1971.
Technical Memorandum 177-18, "DPD-2 System Analysis Review," F. Dickey and
J. Holtzman, May 1971 .
Technical Memorandum 177-19, "Informal Log, Mission 165, Garden City, Kansas,"
J. D. Young, June 1971.
Technical Memorandum 177-20, "NASA Earth Observation Survey Program, 90-Day
Mission Report, Garden City, Kansas, Site 76, Mission 165, 20 May 1971, 1
F. M. Henderson and F. M. Dickey, August 1971.
Technical Memorandum 177-21, "Basic Considerations for Extracting Quantitative
Data from Photographically Stored Radar Imagery," F. Dickey and J.
Holtzman, August 1971.
Technical Memorandum 177-22, "90“Day Mission Report," F. M. Henderson and
F. Dickey, November 1971.
Technical Memorandum 177-23, "densitometric Data as a Basis for Agricultural
Analysis from Fine Resolution Radar Imagery," J. C. Coiner and J. B.
Campbell, November 1971.
Technical Memorandum 177-24, "Preliminary Agricultural Data Analysis," J. Young,
December 1971.
Technical Memorandum 177~25, Being Completed, Author: S. A. Morain.
Technical Memorandum 177-26, "IDECS-System Development, September 1970 -
September 1971," P . N . Anderson, et a I .
Technical Memorandum 177-27, "90-Day Mission Report, Pt. Barrow, Alaska,"
F. Dickey and 5. Parashar, February 1972.
166 '
RELATED PUBLICATIONS:
Berger, D. H “Texture as a Discriminant of Crops on Radar Imagery," IEEE
TRANSACTIONS, vol. GE“8, no. 4, October 1970, pp. 344-348.
Coiner, J. C. and S. A, Morain, "Image Interpretation Keys to Support Analysis
of SLAR Imagery," PROCEEDINGS, Amer. Soc. Photog., Paper No. 71-334,
1977, pp. 393-412.
Dellwig, L. F.,H. C. MacDonald and J. N. Kirk, "Technique for Producing a Pseudo-
Three” Dimensional Effect with Monoscopic Radar Imagery , " PHOTOGRAM-
METRIC ENGINEERING, September 1970, pp. 987-988.
Gillerman, E., "Roselle Lineament of Southeast Missouri," GEOLOGICAL SOCIETY
OF AMERICA BULLETIN, vol. 81, March 1970, pp. 975-982. Also CRES
Technical Reprint 118“5.
Haralick, R. M., F. Caspall, and D. S. Simonett, "Using Radar Imagery for Crop
Discrimination: A Statistical and Conditional Probability Study," REMOTE
SENSING OF ENVIRONMENT, vol. 1, no. 1, 1970, pp. 131-142.
Haralick, R. M., "Data Processing at The University of Kansas," PROCEEDINGS,
4th Annual Earth Resources Program Review, NASA Manned Spacecraft Center,
Houston, Texas, January 1972.
Hardy, N. E., "Vegetation Mapping with Side-Looking Airborne Radar; Yellowstone
National Park," AGARD IXVII EPP Technical Meeting, Colorado Springs,
Colorado, June 1971.
Henderson, F. M., "Radar Monitoring of Agricultural Land Use: Some Problems and
Potentials at the Local Level," PROCEEDINGS, AMER. SOC. PHOTOG.,
Paper No. 71-322, 1971, pp. 368-385.
Henderson, F. M., "Space Photography as a Tool in Delimiting Transportation Net-
work," PROCEEDINGS, AMER. SOC. PHOTOG., vol. II, August 1970,
pp. 71-73.
Holtz man, J., G. A. Bradley and L. F. Dellwig, "Radar Remote Sensing Technology,"
PROCEEDINGS, IEEE International Conference on Systems, Networks, and
Computers, Mexico, January 1971, pp. 832“836.
Lewis, A. J., "Geomorphic Evaluation of Radar Imagery of Southeastern Panama and
Northwestern Colombia," CRES Technical Report 133-18, University of Kansas
Center for Research, Inc,, Lawrence, Kansas, 1971.
Lewis, A. J., H. C. MacDonald, and D. S. Simonett, "Detection of High Return Linear
Cultural Features on Multiple Polarized Radar Imagery," Technical Reprint 188“
4, PROCEEDINGS, 6th Symposium on Remote Sensing of Environment, University
of Michigan, Ann Arbor, October 1969.
MacDonald, H. C., "Effective Radar Look-Direction for Geological Interpretation,"
PROCEEDINGS, 2nd Annual Earth Resources Aircraft Program Status Review,
NASA Manned Spacecraft Center, Houston, Texas, September 1969.
167
MacDonald, H . C . , J . N . Kirk, L. F . Del I wig and A , J . Lewis, "The Influence of
Radar Look- Direction on the Detection of Selected Geological Features,"
PROCEEDINGS, 6th Symposium on Remote Sensing of Environment, 11 University
of Michigan, Ann Arbor, 1970, pp. 637-650.
MacDonald, H. C., "Geologic Evaluation of Radar Imagery for Darien Province,"
CRES Technical Report 133-6, University of Kansas Center for Research, Inc.,
Lawrence, Kansas, 1969.
MacDonald, H. C. and W. P. Waite, "Soil Moisture Detection with Imaging Radars,"
WATER RESOURCES RESEARCH, vol. 7, no. 1, 1971, PP . 100-110.
MacDonald, H. C., A. J. Lewis and W. P. Waite, "Radar Geomorphology in Louisiana
Coastal Marsh and Swamp," ABSTRACTS, Annual Meeting Geogria Academy
of Science, April 1971.
MacDonald, H. C. , A. J. Lewis and R. S. Wing, "Mapping and Landform Analysis of
Coastal Regions with Radar," GEOL. SOC . OF AMER. BULLETIN, vol. 82,
no. 2, February 1971, pp. 345-358.
McCauley, J., "Surface Configuration as an Explanation for Lithology-Related Cross-
Polarized Radar Image Anomalies," PROCEEDINGS, 4th Annual Earth Resources
Program Review, NASA Manned Spacecraft Center, Houston, Texas, January
1972. 7
Mora in, S. A., "Radar Uses for Natural Resources Inventories in Arid Zones,"
presented at Mexico Symposium, November 1970, McGraw-Hill of Mexico.
Morain, S. A., “Active Microwave Systems In Agricultural Remote Sensing: A Look
Ahead," PROCEEDINGS, 3rd Annual Earth Resources Aircraft Program Status
Review, NASA Manned Spacecraft Center, Houston, Texas, December 1970.
Morain, S. A., J. Holtzman and F. M. Henderson, "Radar Sensing in Agriculture, A
Socio-Economic Viewpoint," CONVENTION RECORD, Electronic and Aero-
space System (EASCON '70). Published IEEE, pp. 280-287.
Morain, S. A., "The Status of Parametric Studies in Radar Agriculture," PROCEEDINGS,
• 4th Annual Earth Resources Program Review, NASA Manned Spacecraft Center,
Houston, Texas, January 1972.
Moore, R. K., Radar and Data Processing," PROCEEDINGS, 2nd Annual Earth
Resources Aircraft Program Status Review, NASA Manned Spacecraft Center,
Houston, Texas, September 1969.
Moore, R. K, and G. A. Bradley, "Radar and Oceanography," PROCEEDINGS, 2nd
Annual Earth Resources Aircraft Program Status Review, NASA Manned Space-
craft Center, Houston, Texas, September 1969.
Moore, R. K., "Remote Sensing at The University of Kansas in Radar Systems,"
PROCEEDINGS, 3rd Annual Earth Resources Aircraft Program Status Review,
NASA Manned Spacecraft Center, Houston, Texas, December 1970.
168
Moore, R. K. and W. J. Pierson, Jr,, "Worldwide Oceanic Wind and Wave Pre~
dictions Using a Satellite Radar Radiometer," AIAA TECHNICAL PAPER 70-
310, March 1970; also JOURNAL OF HYDRONAUTICS, vol. 5, no. 2,
April 1971, pp. 52-60.
Moore, R. K., "Radar and Microwave Radiometry, " presented at NASA International
Workshop on Earth Resources, Ann Arbor, Michigan, May 1971.
Moore, R. K., "Radar Imaging Applications: Past, Present and Future," PROPAGATION
LIMITATIONS IN REMOTE SENSING, AGARD Conference Proceedings No. 90,
NATO AGARD, Neuilly - sur“Seine, France, October 1971, pp. 9/1 “9/1 9.
Moore, R. K. and G. C. Thomann, "Imaging Radars for Geoscience Use," IEEE
TRANSACTIONS, vol. GE“9, no. 3, July 1971, pp. 155-164.
Moore, R. K., "Radar Signature and Systems Studies at Kansas University," PRO-
CEEDINGS, 4th Annual Earth Resources Program Reivew, NASA Manned Space-
craft Center, Houston, Texas, January 1972.
Moore, R. K. and W. J. Pierson, Jr., "The Extrapolation of Laboratory and Aircraft
Radar Sea Return Data to Spacecraft Altitudes," PROCEEDINGS, 4th Annual
Earth Resources Program Review, NASA Manned Spacecraft Center, Houston,
Texas, January 1972.
Waite, W, P. and H. C. MacDonald, "Snowfield Mapping with K“Band Radar,"
REMOTE SENSING OF ENVIRONMENT, vol. 1, no. 2, 1970. CRES Technical
Reprint 133“7.
Waite, W. P. and H. C. MacDonald, "Vegetation Penetration with K-Band Imaging
Radars," IEEE TRANSACTIONS, vol. GE-9, no. 3, July 1971, pp. 147-155.
Wing, R. S. and L. F. Dellwig, "Radar Expression of Virginia Dale Precambrian
Ring-Dike Complex, Wyoming/Colorado, " GEOL. SOC. AMER. BULLETIN,
vol. 81, 1970, pp. 293-298.
Wing, R. S., L. F. Dellwig and H. C. MacDonald, "Tectonic Analysis from Radar-
Central and Eastern Panama," PROGRAM ABSTRACTS, Geological Soc. of
America Southcentral Meeting, April 1970, pp. 305-306.
169
* U.S. GOVERNMENT PRINTING OFFICE: 1974— €71 192/ 165
Live Music Archive
Librivox Free Audio
Metropolitan Museum
Cleveland Museum of Art
Internet Arcade
Console Living Room
Books to Borrow
Open Library
TV News
Understanding 9/11