EPA-600/4-80-051
November 1980
SUMMARY OF THE WESTERN ENERGY
OVERHEAD MONITORING PROJECT
by
P. Ishikawa, Jr.
Lockheed Engineering and Management
Services Company, Inc.
Remote Sensing Laboratory
Las Vegas, Nevada 89114
Contract No. EPA 68-03-2636
Project Officer
G. A. She!ton
Advanced Monitoring Systems Division
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada 89114
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA 89114
-------
EPA-600/4-80-051
November 1980
SUMMARY OF THE WESTERN ENERGY
OVERHEAD MONITORING PROJECT
by
P. Ishikawa, Jr.
Lockheed Engineering and Management
Services Company, Inc.
Remote Sensing Laboratory
Las Vegas, Nevada 89114
Contract No. EPA 68-03-2636
Project Officer
G. A. Shelton
Advanced Monitoring Systems Division
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada 89114
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA 89114
-------
DISCLAIMER
This report has been reviewed by the Environmental Monitoring Systems
Laboratory, U.S. Environmental Protection Agency, and approved for
publication. Approval does not signify that the contents necessarily reflect
the views and policies of the U.S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorsement or
recommendation for use.
-------
SUMMARY
In June of 1975 the Environmental Protection Agency (EPA) and the National
Aeronautics and Space Administration (NASA) entered into a five-year, EPA-
funded project which later was to be called the Western Energy Overhead
Monitoring Project. The purpose of this joint undertaking was to transfer
hardware and software techniques for processing remotely sensed digital data
from NASA to EPA in order that EPA could develop the capability of estab-
lishing and maintaining a fully operational remote sensing monitoring system."
Such an operational system was perceived to be a valuable tool for monitoring
surface reclamation activities at coal strip mines in the western United
States. This being a major consideration, the overall objective of the
project was to define and develop operational remote sensing techniques and
demonstrate that these techniques could be used in monitoring, in a rapid and
cost-effective manner, the success with which an energy-related extraction
site has been, or is being, rehabilitated.
The project was divided into three phases. The main emphasis in Phase I
was the transfer of a Data Analysis System (DAS) and its software technology
and mai ntenance requi rements to the Envi ronmental Moni tori ng Systems
Laboratory in Las Vegas, Nevada (EMSL-LV). In Phase II, EMSL-LV used the
system on a shakedown basis while the NASA National Space Technology "
Laboratory (NSTL) in Slidell, Louisiana, used a similar system to investigate
and research specific problems defined by EMSL-LV. In Phase III, EMSL-LV
used and tested the system in an operational mode and upgrades in hardware and
softwa re were made. "
The bulk of the technology transfer was successfully accomplished during
Phase I but periodic upgrades continued into Phase III. Throughout the
project valuable experience was acquired in developing and maintaining an
operational remote sensing monitoring system, in obtaining remotely sensed.
digital data, and in processing the digital data on the EMSL-LV DAS.
The acquisition and processing of the scanner data during the Hestern
Energy Project demonstrates that the EPA remote sensing monitoring system can
be used to successfully detect and map, in a cost-effective manner, land
use/land cover features associated with strip mining activities in the western
United States. It is a versatile system with various processing options
available. With the system, it is possible to geographically correct scanner
data so that classified scanner images can be directly correlated to
topographic maps for easy mapping of land use category distributions. In
addition, accurate area and percentage figures can be obtained for each
classified land use category.
Hi
-------
Airborne multispectral scanner data acquired over selected western strip
mines can be effectively processed and analyzed at the EMSL-LV. However,
since the EMSL-LV OAS is not a fully automatic system free of analyst
interaction, a considerable amount of interactive computer processing is
required to produce a computer-generated product that is correctly classified.
Because of this, the EMSL-LV DAS is more suitable for mapping strip mining
reclamation activities on a case-by-case basis but is less suitable for
production-type monitoring that requir~s the quick turnaround processing of
data for hundreds of energy-related extraction sites.
iv
-------
CONTENTS.
Summa ry . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fi gures . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tab les. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgement. . . . . . . . . . . . . . . . . . . . . . .
1.
2.
3.
4.
I ntroducti on . . . . . . . . . . . . . . . . ... . . . .
Cone 1 us ions. . . . . . . . . . . . . . . . . . . . . . .
Recommenda ti ons . . . . . . . . . . . . . . . . . . . . .
Ph as e I . . . . . . . . . . . . . . . . . . . . . . . . .
Data acquisition. . . . . . . . . . . . . . . . . .
Data reduction and processing procedures. . . . . .
Con venti ona 1 ana 1ysi s pro~edures . . . . . . . . . .
Phase I res u 1 ts .. . . . . . . . . . . . . . . . .
Phase I I ......... . . . . . . . . . . . . . . . .
5.
6.
Data acquisition. . . . . . . . . . . . . . . . . .
Data processing. . . . . . . . . . . . . . . . . .
Data analysis. . . . . . . . . . . . . . . . .
Phase II results. . . . . . . . . . . . . . . . . .
Phase I I I . . . . . . . . . . . . . . . . . . . . . . . .
Software upgrade. . . . . . . . . . . . . . . . . ..
Hardware upgrade. . . . . . . . . . . . . . . . . .
Phase II res u 1 ts . . . . . . . . . . . . . . . . . .
Cost estimate. . . . . . . . . . . . . . . . . . .
Annotated Bibliography
. '. . . . . . . . . . . . . . . . . .
v
Page
iii
vi
vii
viii
1
3
4
5
5
7
10
12
13
13
13
14
16
18
18
20
20
27
28
-------
,
Number
1
2
3
4
5
6
7
8
9
10
11
FIGURES
locations of Mine Sites Surveyed in Phase
Multispectral Scanner Data Reduction and
Processing Flow. . . . . . . . . . . .
I . . . . .
. . . .
.....
. . . . .
Graphical Outline of the Comtal Varian Interactive
Processing System (CVIPS) Program. . . . . . . .
Daedalus 1260 Multispectral Scanner System and
RC-8 Metric Mapping Camera. . . . . . . . . .
. . . . .
. . . . . . .
Contact Print of the Classified Aircraft
Digital Scanner Data Produced by the
Clustering Algorithm UNSUP. . . . . . .
. . . . . . . . . .
Data Analysis System as Defined and Assembled
at NASA/NSTl. . . . . . . . . . . . . . . .
. . . . . . . .
Updated Processing Hardware Configuration of the
EPA Data Analysis System. . . . . . . . . . . . . . . . . .
Decker Strip Mine Classified Using MAXl4. .
. . . . .
. . . .
Aircraft Photograph of Decker Strip Mine. . . . .
. . . . . .
Unregi stered Scanner Data Versus Gee-referenced
Scanner Data. . . . . . . . . . . . . . . . . . . . . . . .
Center Strip Mine Classified Using ENCClS . . . . .
.....
vi
Page
6
8
9
14
16
21
22
23
24
25
26
-------
Number
1
2
3
TABLES
Land Cover Classification Hierarchy for
Coal Strip Mi ne Monitoring. . . . . .
. . . . . . . . . . .
MSS Wavelength Bands. . . . .
. . . . . .
Cost Estimate for the Overhead Monitoring
of One Hundred Western Surface Mines. .
vii
. . . .
...:'.r~. .
......
. . . . .
Page
10
15
27
-------
ACKNOWLEDGEMENT
The Western Energy Overhead Monitoring Project and this project summary
report were made possible through the cooperation of the u.s. Envi ronmental
Protection Agency and the National Aeronautics and Space Administration's
National Space Technology Laboratory in Slidell, Louisiana.
Sincere thanks are extended to Mr. Charles E. Tanner, formerly of Lockheed
Engineering and Management Services Company, Incorporated, for documenting
Phases I, II, and III of the Western Energy Overhead Monitoring Project and
for assisting in the preparation of this summary report.
viii
-------
SECTION 1
INTRODUCTION
Since its establishment in 1970, the U.S. Environmental Protection Agency
(EPA) has been involved in energy-related environmental research, including
the development of pollution control technology, as necessary to meet its
statutory responsibilities. EPAls primary mission is environmental
protection, and its objective in the energy area is to enable energy
development to progress as rapidly as possible while assuring that national
environmental goals are met.
As the nation strives to reduce its dependence on foreign energy,
increased attention must be given to the environmental problems arising from
domestic energy resource development. On the federal level, this need has
been recognized by the expansion of energy-related environmental research and
development programs undertaken by a number of federal departments and
agenci es.
Centralized coordination of a number of these expanded efforts is provided
via the Federal Interagency Energy/Environmental Research and Development
Program formerly administered by the Environmental Protection Agency's Office
of Energy, Minerals, and Industry (OEMI). The primary purpose of the inter-
agency program is to assure that our national energy goals are matched with an
effective research and development program in the critical areas where energy
needs and environmental protection goals overlap.
-With-this in mind the Environmental Protection Agency and the National
Aeronautics and Space'Administration (NASA) entered into a 5-year, EPA-funded
project in June of 1975. The purpose of this interagency project was twofold:
to transfer hardware and software techniques for processing remotely sensed
digital data from NASA to EPA and to assist EPA in developing the capability
of establishing and maintaining a fully operational remote sensing monitoring
system .
Laboratories through which the 5-year plan would be implemented were
designated by both agencies. The NASA National Space Technology Laboratory
(NSTL, formerly the Earth Resources Laboratory) located in Slidell, Louisiana,
and the EPA Environmental Monitoring Systems Laboratory in Las Vegas, Nevada,
(EMSL-LV) were the two agency laboratories involved.
The overall objective of this 5-year project was to define, develop, and
demonstrate operational remote sensing techniques to rapidly monitor, in a
1
-------
cost-effective manner, the success with which an energy-related extraction
site has been, or is being, rehabilitated to a state suitable for its intended
or previous land usage.
The two participating agencies agreed to divide the 5-year project into
three distinct phases based on operational requirements. Phase I was begun in
July of 1975 and continued for 18 months. During this time technology trans-
fer of software and hardware capabilities, and data reduction of aircraft-
acquired and Landsat multispectral scanner data were performed and evaluated.
Phase II of the project was begun in January of 1977 and was 18 months in
duration. During this phase, EMSL-LV used the system on a shakedown basis for
monitoring purposes. Also, during this time, NSTL used a similar system to
investigate and research specific problems defined by EMSL-LV.
During Phase III, EMSL-LV used and tested the system in an operational
mode. In addition, software and hardware upgrades were made and the system
was refined to make processing faster and more economical.
This report summarizes the events that occurred in the various phases of
the Western Energy Overhead Monitoring Project. Technical details of. the data
acquisition, data processing, and analytical procedures will not be addressed
in this report. These details have been discussed in previous reports
documenting each phase of the Western Energy Overhead Monitoring Project.
These earlier reports are listed in the annotated bibliography.
2
-------
SECTION 2
CONCLUS IONS
The Western Energy Overhead Monitoring Project demonstrates that airborne
multispectral scanner data and computer processing techniques can be used to
detect and map land "use/land cover features associated with coal strip mining
activities in the western United States. Through the joint efforts of NSTL
and EMSL-LV, a Data Analysis System for Monitoring strip mining activities has
been established at EMSL-LV.
As a result, airborne multispectral scanner data acquired over selected
western strip mines can be effectively processed and analyzed at EMSL-LV.
However, since the EMSL-LV DAS is not a fully automa~ic system free of analyst
interaction, a considerable amount of interactive computer processing is
required to produce a computer generated product that is correctly classified.
Because of this the EMSL-LV DAS is more suitable for mapping strip mining
reclamation activities on a case-by-case basis rather than monitoring a large
number of coal strip mines on a production basis.
3
-------
SECTION 3
RECOMMENDATIONS
Based on the results and conclusions from the three phases of the Western
Energy Overhead Monitoring Project, the following recommendations are proposed
for consideration:
.
For optimum processing, obtain scanner data at altitudes that
will enable the scanner to resolve features of interest and at sun
angles that will minimize shadow interference; eliminate excessive
computer processing by obtaining scanner data at scales that will
allow the entire target area to be displayed on a single computer
scene; and edit excessive data.
. Augment the use of automatic data processing with photographic
interpretation as the means for developing baseline information.
Computer classification techniques can be relied upon to quickly
and correctly classify scanner data containing features having
unique spectral characteristics, but in situations where features
of interest cannot be resolved spectrally, resulting in the
misclassification of data, photographic interpretation proves to
be more productive.
.
Investigate the feasibility of digitizing photo-interpreted
aerial photography for use as baseline information if production
work is proposed.
Investigate the use of the EPA data base as a means to store and
update changes that occur at strip mining sites.
.
4
-------
1--
SECTION 4
PHASE I
Phase I of the Western Energy Overhead Monitoring Project began in July of
1975 and was 18 months in duration. The overall objective of this phase was
to develop and demonstrate the techniques and procedures for deriving
vegetation/land cover classifications of coal strip mine sites and their
environment for the purpose of subsequent monitoring of the rehabilitation
efforts on mined areas. The specific objectives were to:
. Transfer the Data Analysis System and its associated software
technology and maintenance requirements to EMSL-LV.
. Perform automated analysis of aircraft MSS data and manual
classification of photographic data over selected coal strip mine
sites.
. Determine the utility of Landsat MSS data for performing a
regional land cover classification of a portion of the Powder
River Basin.
Compare the results obtained from the aircraft MSS classifica-
tion with the manually photointerpreted results.
The Western Energy Project was appropriately named because the study area
was composed of seven states located west of the Mississippi River. These
states,moving westwardly, were North Dakota, Montana, Wyoming, Colorado, New
Mexico, Arizona, and Utah. All states had active surface mining operations in
many areas. Because of aircraft flight limitations and other criteria, eight
mines were selected from five of the seven states for the study.
.
DATA ACQUISITION
Data acquisition occurred during the months when the chlorophyll content
of vegetation was at its maximum at the selected coal strip mine sites.
(Figure 1).
Flight lines were drawn on partial or full-frame color infrared film
enlargements (1.02 meter format) and also black-and-white full-frame
enlargements produced by EMSL-LV personnel from NASA high-altitude aerial
photography acquired prior to the low-altitude data collection.
5
-------
I
!
COLORADO
TEXAS
MONTANA
ROSEBUD )(
x BIG SKY
DECKER X
BELLE AYR. X
IDAHO
DAVE X
JOHNSTON
NEVADA
UTAH
WYOMING
HENRY X
MOUNT AIN
BLACK X
MESA
X NAVAJO
Figure 1.
Locations of mine sites surveyed in Phase I.
6
-------
These photographic enlargements, along with corresponding 7~-minute or
IS-minute quadrangle maps, were annotated with required flight lines and start
and stop coordinates for each mine. All aircraft data were collected by the
data acquisition system on board NASA/Johnson Space Center's NP-3A aircraft.
During the missions multispectral scanner data were collected at altitudes
of 914, 1,828, and 3,657 meters (3,000, 6,000, and 12,000 feet) above mean
terrain elevation. Also photographic coverage was obtained at altitudes of
305, 914, 1,828, and 3,657 meters (1,000, 3,000, 6,000, and 12,000 feet) using
both color and color-infrared film with 60 percent forward overlap.
After the aircraft data collection was completed the film was processed by
EMSL-LV and duplicate positive transparencies and prints were supplied to NSTL
for use in their analysis. The multispectral data were reduced and processed
at NSTL.
DATA REDUCTION AND PROCESSING PROCEDURES
Reduction and processing of the aircraft MSS data were accomplished
through task-designed software modules on the Data Analysis System that are
linked in such a manner that output from one module serves as input to
another. Each module can be composed of several software programs that
accomplish one major task. Figure 2 illustrates a simplified flow diagram
, that outl ines the reduction and processing of data' from an aircraft
multispectral scanner.
Immediately following data acquisition, theMSS data were decommutated and
reformatted. The decommutation/reformatting process converts the PCM (Pul se
Code Modulated) format of the original aircraft data into a digital format
compatible with the DAS. The digital tape that is produced by the
decommutation/reformatting procedure is checked for data anomalies such as
data drop-outs, sun angle manifestations, recording problems, etc. Most
anomalies can be corrected by the application of software programs in the
preprocessing and transformation block shown in Figure 2. A data quality
check can be performed by analyzing a hard copy product that is the result of
intermediate processing, or by viewing the data on the interactive display
system.
After the data decommutation operation was completed, a supervised
approach to pattern recognition was employed by the analyst and was initiated
in the Pattern Recognition Modules. This approach requires that the analyst
know something about the scene because he is responsible for "training" the
computer to recogni ze the va ri ous 1 and cover categori es that a re of interest.
, This was accomplished by using the Interactive Processing System (Figure 3) to
outline areas of known physical characteristics and then computing statistics
(means, variance, correlation matrices, etc.) for each training sample and
grouping the samples into land cover classes.
The Pattern Recognition Modules were written to accommodate a maximum of 4
channels of digital data. It was therefore necessary to run the Channel
7
-------
HARD COpy
OUTPUT
DEVICE
PREPROCESS ING
DATA
TRANSFORMA TION
I
I
I
I
I
I
I
I
I
I
I
I
I
1--
INTERACTIVE
DISPLA Y
SYSTEM
Figure 2.
- --
PA TTERN RECOGNITION
.TRAINING SAMPLE SELECTION
.STATISTICS COMPUTATIONS
.CHANNEL SELECTION
.CLASSIFICATION
- --
ACREAGE
COMPILA TIONS
THEME
INVENTORY
ETC.
Multispectral scanner data reduction and processing
flow (simplified).
8
-------
MODE 2
cs
COMPUTE
STATS
CV
CHANGE
VALUE
MS
MAIN
STREAM
"
LF CL
LOAD - CLASSIFI.
FUNCTION CATION
PL
PLOT
FUNCTION
IE
IMAGE
ENHANCE-
MENT
LF
LOAD
FUNCTION
LC
LOAD
COLOR
LC
LOAD
COLOR
,/
PT
PICK
TRAINING
FIELD
LF PT
LOAD PICK
FUNCTION - TRAINING
FIELD
RT
, READ
TARGET
LC
LOAD
COLOR
cs
COMPUTE
STATS
PL
PLOT
FUNCTION
PT
PICK
TRAINING
FIELD
CL
CLASSIFI-
CATION
cs
COMPUTE
STATS
I
\
LC
LOAD
COLOR
LF
LOAO
FUNCTION
Figure 3.
CL
CLASSIFI-
CATION
cs
COMPUTE
STATS
LC
LOAD
COLOR
LF
LOAD
FUNCTION
Graphical outline of the Comtal Varian Interactive
Processing System (CVIPS) program.
9
-------
Selection Program to determine the'4 channels for use in classifying each data
set. Once the 4 channels had been determined the analyst proceeded as
outlined in Figure 2.
CONVENTIONAL ANALYSIS PROCEDURES
Included also in Phase I was the analysis of coal strip mines using aerial
photography. The photo analysis of the mines was performed using a zoom-
~ stereoscope, a light table, and the classifkation hierarchy presented in
Table 1. First, appropriate stereo-pairs of each coal strip mine were located
within the roll of duplicate positive transparencies. Next, transparent
overlay material was placed over one frame of imagery and the features of
interest were outlined and annotated on this overlay. When the frame of,
imagery was completely interpreted, the overlay was forwarded to the EPA
cartographic laboratory for scribing (scribing produced a "clean" product,
e.g., lines are of equal width, numbers are the same size, etc.). These
interpreted sets of photography were later used to evaluate the accuracy of
the multispectral analysis.
TABLE 1.
LAND COVER CLASSIFICATION HIERARCHY FOR COAL
STRIP MINE MONITORING
00 - Active Mine Features
I .
I
01 - Advance cut
02 - Topsoil stockpile including embankment
03 - Stripping bench, highwall, open cut, pit, and related features
04 - Exposed coal seam
05 - Raw spoil bank and/or sideslope
06 - Coal storage pile
07 - Recontoured spoil
08 - Haul road including cuts, fills, turnouts, etc.
09 - Misc. disturbed areas within active mine complex
10 - Barren Land
14 - Shorelines, river banks
15 - Badlands (barren silts and clays, related metamorphic rocks)
18 - Man-made barrens
20 - Water Resources
21 - Settling ponds
22 - Pit ponds
23 - Undifferentiated ponds, lakes, and
24 - Water courses, including canals
reservoi rs
(continued)
10
-------
TABLE 1.
(Continued)
30 - Natural Vegetation
31 - Herbaceous types
32 - Shrub/scrub types
33 - Savanna-like types
34 - Forest and woodland types
40 - Cultural Vegetation (plantations and seedings)
Numerators
41 - Grass/forb seedings
42 - Tree/grass or tree/scrub plantations
45 - Seeding trails and test plots
Denominators
A - Vegetative cover 80 - 100
B - Vegetative cover 60 - 80
C - Vegetative cover 30 - 60
D - Vegetative cover 5 - 30
E - New seeding, vegetative cover less than 5
1 - Production level better than native level
2 - Production level about equal with native level
3 - Production level less than native level
50 - Agricultural Production
51 - Field crops
56 - Fallow land
60 - Urban/Industrial
61 - Residential
62 - Commercial and services
63 - lndustri al
64 - Industri al
65 - Transportation, communications
66 - Resource extraction
and utilities
11
-------
PHASE I RESULTS
The results of Phase I analysis indicated the feasibility of using
aircraft-acquired multispectral scanner data for monitoring and evaluating
coal strip mine rehabilitation activities. It was determined that multi-
spectral scanner data acquired by an aircraft at an altitude of approximately
3,660 meters (12,000 feet) appeared to possess sufficient spatial resolution
for vegetation/land cover classifications of western coal strip mine sites and
their environs. Also, adequate details of land cover features could be
obtained at 3,660 meters. It was further determined that the additional
details gained from low-altitude coverage would not compensate for the
necessary increase in data processing.
Finally, it was concluded from a Chi-square analysis that there were no
significant differences in the results of the land cover mapping achieved
through computer-implemented techniques and the land cover mapping obtained
from conventional photointerpretation methods for the same land-cover classes.
12
-------
SECTION 5
PHASE II
Phase II was begun in January of 1977 at the laboratories involved.
EMSL-LV was responsible for performing a shakedown of the system for moni-
toring purposes. During the same time frame, NSTL personnel investigated
problems specifically defined by EMSL-LV which required additional research.-
During Phase II, EMSL-LV addressed the development of a procedure that would
maximize the data acquisition system defined and assembled at NASA/NSTL.
IITurn-around-timeli was of critical importance in the foundation of this
procedure.
DATA ACQUISITION
The aircraft multispectral scanner and aerial camera shown in Figure 4
were used to collect data over the selected coal strip mines in late June of-
1977. The airborne multispectral system is ~apable of collecting data at
-altitudes ranging from 152 meters (500 feet) to 6,096 meters (20,000 feet)
above ground level. The system was designed to collect and record radiant
energy data in the near ultraviolet through the thermal infrared portions of
the electromagnetic spectrum (Table 2).
The aerial camera that is part of the onboard instrumentation is a 15.240
centimeter (6-inch) focal length metric mapping camera. Its primary use is as
a ground-coverage documenting camera. The film/filter combinations and
exposure techniques are chosen to obtain photographic imagery in the desired
spectra of interest.
DATA PROCESSING
Data processing began after the scanner data had been recorded on a high
density digital tape and returned to the EPA laboratory. The aircraft-
acquired multispectral scanner data was converted into another format because
conventional computers, such as the Varian operating system for the DAS,
cannot read the data format produced by an aircraft multispectral scanner.
Therefore, a special device had to be designed to convert aircraft multi-
spectral scanner data into the format expected by conventional computers.
This reformatting procedure is termed decommutation and it occurs in the high
density digital tape frontend hardware. Following the decommutation
operation, standard data processing techniques were applied to all other
functions (Figure 2) in order to produce a classified image and appropriate
statistical data. -
13
-------
Figure 4.
Daedalus 1260 multispectral scanner system and RC-8
metric mapping camera.
DATA ANALYSIS
Processing of the data sets was initiated after a photointerpretation
hierarchy and a MSS data processing hierarchy were developed. In this phase
it was felt that a Levell/II hierarchy would prove to be more economical and
a faster means for producing results that were needed to implement monitoring-
type operations over coal strip mines.
In Phase II a combination approach to pattern recognition was used.
First, the raw data were classified by a sequential clustering algorithm
(UNSUP) and, second, the image generated by the sequential clustering
algorithm was then used to facilitate the selection of homogeneous training
samples for use in other classification approaches to pattern recognition.
14
-------
TABLE 2. MSS WAVELENGTH BANDS
Wavelength
Channel Band
1 0.38
2 0.42
3 "0.45
4 0.50
5 0.55
6 0.60
7 0.65
8 0.70
9 0.80
10 0.92
11 8.00
- 0.42 ym
- 0.45 ym
- 0.50 ym
- 0.55 ym
- 0.60 ym
- 0.65 ym
- 0.70 ym
- 0.79 ym
- 0.89 ym
- 1.10 ym
- 14.00 ym
Color/Spectrum
Near ultraviolet
Blue
Blue
Green
Green
Red
Red
Near infrared
Near infrared
Near infrared
Thermal infrared
Classification Program - UNSUP
The clustering program used to produce the hard copy product shown in
Figure 5 uses an algorithm that groups pixels with similar statistical
characteristics. The major assumption in the program is that the data exhibit
some homogeneous features. Success at implementing this program was dependent
upon the analyst's knowledge of the data set. An understanding of the
spectral range, approximate number of classes that can be generated, the
categories that may possibly cause confusion, the best channels to use for
discrimination, and the number of channels with which to work were essential.
The data flow for the clustering program is straightforward. First, an
initial population is established; then statistical profile tables for each
class are developed. Third, the program compares the picture element with
each of the established populations to determine where it best fits. Finally,
additional populations are established whenever needed.
15
-------
Figure 5. Contact print of the classified aircraft digital
scanner data produced by the clustering algorithm UNSUP.
As stated earlier, each pixel of the digital data is compared to the
existing statistical profiles to determine which population it best fits.
When all the pixels had been assigned to one of the active populations or to
zero population, the run terminated and the results were saved on a 9-track
(800 BPI) magnetic tape. Statistics that were generated during the run and
stored in the computer are dumped to the line printer device for printing.
These hard copies are used in color-coding the images and grouping similar
land-cover areas.
After color-coding the images, a special software program was used to
convert the classified digital imagery to a format a~ceptable to the film
recorder. The images were filmed in color and prints such as that shown in
Figure 5 were then produced for use in presentation and/or reports similar to
this.
PHASE II RESULTS
The aircraft-acquired MSS data assessment of the areas affected by strip
mining and reclamation activities was performed in a single phase analysis
through the use of the DAS. The use of the MSS Computer-compatible tapes with
the DAS and other available software provided a reliable means for detecting
affected areas at coal strip mines.
16
-------
The clustering program, which was the primary software classification
algorithm, was capable, in most instances, of detecting and distinguishing
various earth surface conditions and land cover situations (mine-related and
nonmine-related) in the study area.
The results, six classification maps, generated by the clustering
algorithm provided needed information for future planning and data for incor-
poration into the EPA data base. The data base will be an important tool in
the temporal analysis of MSS data. The data reduction and processing
techniques developed at EMSL-LV produced additional capabilities and provided
a means of quicker assessment of coal strip mining. Also, the acreage
compilation algorithm generated reliable statistical data which are essential
in determining the progress of reclamation work conducted by coal mining
companies.
17
-------
SECTION 6
PHASE I II
In the Phase III operations, new software was acquired from NSTL and
additional hardware was purchased and added to the old DAS configuration.
These upgrades afforded the analyst an opportunity to develop a procedure for
processing scanner data acquired over coal strip mines. .
The clustering approach to pattern recognition, UNSUP, used in Phase II
was not an ideal procedure. There were several problems associated with UNSUP
but the two more critical problems that justified the search for a better
procedure were: .
. UNSUP has no memory in the form of statistical data that
can be retained from one run to the next. Each set of data
to be processed is treated independently. Consequently, the
only manner in which the program can form classes is to
start at the beginning of each tape or tape segment. Since
the data are statistically different in each tape, the
classes will be formed differently on each UNSUP run making
it almost impossible to butt-join the results of two tapes.
. New. classes are formed by the UNSUP program as itencoun-
ters statistically different materials. This results in mate-
rials being sorted into one class for a portion of the scene
and then being inserted into a new class when it is formed
by the program. When colors are assigned, this anomaly appears
as a discontinuity in the classified data that occurs at the
scan line at which the new class was formed and continues
throughout the data set. This phenomenon may occur several
times during a run and is beyond control of any input parameter.
SOFTWARE UPGRADE
Phase III of the Western Energy Overhead Monitoring Project was begun in
June of 1978. The overall objective of this phase was to develop an update to
the procedure formulated in Phase II. This updated procedure would utilize
aircraft-acquired MSS data, the new software developed at NSTL, and the
expanded hardware capability of the DAS. Among the software received from
NSTL were three programs that served as the nucleus for the updated procedure.
The first program, SEARCH, was developed as a trainer for a maximum likelihood
classifier. SEARCH evaluates a six-sean-line by six-element block of
4-channel input data by the use of its covariance matrix. The statistics
18
-------
developed from a SEARCH run are stored in the Comtal Varian Interactive
Processing System (CVIPS) statistics file. These statistics can be updated,
merged, and used to classify a data set using the second NSTL program, the
maximum likelihood classifier (MAXL4).
The NSTL version of a maximum likelihood classifier is limited to
processing four channels 'of data; hence the acronymn MAXL4. MAXL4 classifies
each vector!, where xt = (X , X2' X3' Xq), as belonging to one of 63
classes or under certain conditions as "other". The mean vector, M, and the
covariance matrix, C, are estimated from training information supplied by the
statistics file from the program SEARCH.
The third program developed at NSTL provided the long-needed aircraft
registration capability that would allow classified data to be input to a data
base for comparative analysis. The aircraft geographical reference module was
designed to be used on multispectral scanner data collected by aircraft that
lacked the sophisticated instrumentation for recording altitude parameters.
This module features an option to correct the multi-channel multispectral
scanner data using bilinear interpolation and an option to correct classified
data using a nearest neighbor resampling method. There is also an option to
choose the resampling size (not necessarily square). An important advantage
to this module lies in the fact that the results are adaptable to fit a
georeferenced data base system.
Such a data base for storing LANDSAT information is available and a
similar one for storing aircraft-acquired data is being developed and tested
by NSTL. When delivered it will greatly enhance the low and medium altitude
monitoring capabilities at EMSL-LV, making temporal analysis and continual
data updating possible. This will facilitate the progressive monitoring of
land cover changes related to coal strip mining reclamation activities.
. In addition to the above programs delivered by NSTL, an unsupervised
clustering program called the Iterative Self-Organizing Clustering System
(ISOCLS) was recently modified at EMSL-LV and put into operation. Developed
at the NASA Johnson Space Center, ISOCLS was rewritten and adapted to the
EMSL-LV DAS. This EPA adaptation of ISOCLS has subsequently undergone further
changes to speed up the procedure and is now called the Environmental
Clustering Classification System (ENCCLS). The significance of ENCCLS is that
it is an extremely rapid and accurate clustering procedure. This promising
program arranges all pixels into distinct clusters whose centers are
represented by their respective cluster-means. The program routinely computes
the distances between each pixel and all the cluster centers and assigns each
pixel to its nearest cluster. New cluster means are then computed and the
cluste.rs are examined to see if they should be split or combined. This entire
process continues repeatedly until the cluster mean values cease to change
significantly from one iteration to the next. The resulting clusters
represent the categories derived from the MSS data.
19
-------
HARDWARE UPGRADE
The DAS, as defined and originally assembled by NSTL, became obsolete
because of increased requirements for automatic data processing. As config-
ured by NSTL (Figure 6) the DAS had 65K of memory, 2 tape drives, a 47 x 10
word disk file, a line printer, a STATOS 33 printer/plotter, 3 interactive
terminals and a PCM (pulse-code-modulated) subsystem.
In its updated configuration (Figure 7) the DAS has 192,000 words (16 bit)
core memory, 235 megaword disk storage, 5 magnetic tape drives and is capable
of supporting background runs and foreground activities simultaneously. The
increase in memory improves processing efficiency by 80 percent and the
additional tape drives support background and foreground processing. The
additional disk space supports data base applications and multiple interactive
users. .
PHASE III RESULTS
The Phase III updated procedure provides a more flexible and accurate
means of processing aircraft-acquired multispectral scanner data than the
earlier multi-channel clustering program, UNSUP. This clustering program
lacked the capability to input statistics, either initially or following a run
that had less than 90 percent of the data classified. However, the analyst
had the option to change input parameters to the program to improve the
classification. This action did not always solve the problem and in some
cases produced less meaningful results than the initial run. In essence,
UNSUP required too manY iterations to produce comparable results with first or
second run results from the maximum likelihood classifier.
The program SEARCH that was intended to replace UNSUP was a faster method
of clustering but it did not produce a classification tape that could be used
to produce hard copies of the image for use in reports and presentations.
Because of its speed in clustering and storing the data in the 8100 STAT (sta-
tistics) file this disadvantage became only an inconvenience to the analyst
who had to punch the cards for the maximum likelihood classification program.
By far the most difficult problem encountered while running SEARCH was knowing
what values to enter for the four critical input parameters. These param-
eters, entered through the interactive mode, consisted of the upper and lower
limits for the standard deviation, the maximum coefficient of variation, and
the desired divergence for fields and classes. Regardless of the values
entered for these parameters there was no way to evaluate the statistics
developed by SEARCH other than processing the data through the classifier, and
color-coding the image. Figure 8 is an example of such MAXL4 run and Figure 9
is the aircraft photography that was 'simultaneously acquired with the scanner
data. .
Color-coding is a little difficult because in developing statistics the
program SEARCH does not have any knowledge of what the materials are other
than that they are statistically similar. Individual classes produced by
SEARCH are given names based on numerical order, e.g., CLS-Ol, CLS-02, CLS-03
... CLS-Nn. These data, when classified, will require the analyst to assign,
20-
-------
OPERATOR'S TERMINAL
AND CARD READER
9-TRACK MAGNETIC
TAPE DRIVES
PLAYBACK SYSTEM AND
CENTRAL COMPUTER
Figure 6.
Data analysis system as defined and assembled at NASA/NSTL.
21
-------
M4 M3 M2 M1 MO
-
TAPE DRIVES
LINE PRINTER
ELECTROSTATIC PLOTTER
7530 DISK MEMORY
7530 DISK MEMORY
OPERATOR CARD
CONSOLE READER
7510 DISK
CENTRAL PROCESSING UNIT
CORE MEMORY
INPUT/OUTPUT
PULSE CODE MODULATED
INTERFACE
MEMORY AND
INPUT/OUTPUT EXPANSION
CROMEMCO
BIT SYNCS
INTERACTIVE
DISPLAY
SYSTEM
COLOR FILM
RECORDER
PCM SUBSYSTEM
Figure 7.. Updated processing hardware configuration of the
EPA data analysis system.
22
-------
--
I
I
1
Figure 8.
Decker strip mine classified using MAXL4.
on an individual class basis, an identification/location (ICLOC) color that
will, when used in conjunction with ancillary data, locate and identify all
the particular classes of material.
Various versions of the maximum likelihood classifier used in these
updated procedures have been evaluated many times over the past ten years by
research groups and the remote sensing community and each has proven to be a
very reliable algorithm. We concur with these findings and add that it is
superior to the UNSUP clustering program.
23
-------
.r~~ ~ --- -
I
I
-..-.- - ..",..-~
Figure 9.
Aircraft photograph of Decker strip mine.
The aircraft geographic referencing program performed beyond expectations
(Figure 10). With this capability we will be able to enter detailed
analytical results into the EPA data base for future reference. The medium-
altitude aircraft-acquired multispectal scanner data, LANDSAT data, and other
digital input to the data base will certainly enhance and extend the
capabilities of this laboratory.
A newly developed program that, at the time of this study, performed well
is ENCCLS. The obvious advantage in using ENCCLS is the elimination of the
time-consuming tasks of selecting training samples and generating training
sample statistics required in the MAXL4 procedure. ENCCLS uses no training
24
-------
Unregistered Data
,!".
j
...
Geo-referenced Data
Figure 10.
Unregistered scanner data versus geo-referenced scanner data.
25
-------
. - --
-""'~~ ~ -
~--
fields, requires very little analyst interactive time, and produces
classification results comparable to, or better than, the MAXL4 results. The
ENCCLS program requires further investigation to evaluate its capability for
reducing the analyst interactive processing time on future projects. Figure
11 shows the results of an ENCCLS run.
Fi gure 11.
Center strip mine classified using ENCCLS.
26
-------
The recent hardware expansion extended the analysis capabilities of the
DAS. The extra memory, tape drives and disk space made it possible to perform
interactive analysis and run background jobs simultaneously. This capability
allows the analyst the opportunity to use his interactive time more effi-
ciently and discriminately. Where once the analyst had to decide whether a
batch job or interactive analysis was more important, he may now implement
both jobs.
COST ESTIMATE
A key factor to be considered in evaluating the feasibility of an overhead
monitoring project, such as monitoring coal strip mining reclamation, is the
cost involved. This cost is related to certain activities such as acquiring
overhead aircraft data, processing and analyzing the data~ and compiling the
results of the analysis in a report. In Table 2 a projected cost estimate for
the initial baseline monitoring of one hundred surface mines in the western
United States has been put together to provide an appraisal of the cost for
undertaking an extensive surface mine monit~ring project.
The total man hours required for such a project is estimated to be in
excess of 3,000 hours with approximately 1,800 hours allotted for analysis and
documentation and the remaining 1,200 hours going to support tasks. It should
be noted that the data acquisition cost will vary depending on the regional
location of the mines and weather conditions at the time of overflight. For
example, adverse weather conditions interfere with aerial data collection
which increases collection costs. Also, mines located in the eastern United
States, would add to the acquisition costs because of the travel distances
involved for the EMSL-LV leased aircraft based in Las Vegas.
"
. ".":i
. .
~.: ,.- .~~.
'.. .1
. ~.~
.;~~. ~~
:~-~~; .
The cost figures in Table 3 are based on the current system configuration
and capabilities at EMSL-LV.
~':'.-. ~:'
l' ~
. . .~ :;.
l'
,
..
TABLE 3.
COST ESTIMATE FOR THE OVERHEAD MONITORING OF
ONE HUNDRED WESTERN SURFACE MINES
...
Item
Cost (0011 ars)
Labor
Data acquisition
(aircraft platform)
Computer processing
Materials
Miscellaneous
39,000
25,,000
230,000
7,400
1,200
Total
302,600
27
-------
ANNOTATED BIBLIOGRAPHY
Anderson, J. E., and C. E. Tanner. Remote Monitoring of Coal Strip Mine
Rehabilitation. EPA-600/7-7B-140, July 1978. EPA, Environmental
Monitoring Systems Laboratory, P. O. Box 15027, Las Vegas, Nevada.
Discusses the accomplishments of the Western Energy Phase I
operations. The results of manual photo-interpretation and automated data
analysis are compared. Included are the results of a feasibility study to
use Landsat data as a regional planning tool for pre-mining environmental
impact evaluation.
Tanner, C. E. Computer Processing of Multispectral Scanner Data Over Coal
Strip Mines. EPA-600/7-79-080, March 1979. EPA, Environmental Monitoring
Systems Laboratory, P. O. Box 15027, Las Vegas, Nevada 89114.
Discusses the results of the Western Energy Phase II operations
conducted by the Environmental Monitoring Systems Laboratory, Las Vegas.
The report outlines the techniques developed and employed to obtain the
classification results on a number of .western coal strip mines.. .
Overhead Multispectral Scanner Monitoring of Western Energy
Coal Strip Mines. EPA, Environmental Monitoring Systems Laboratory, P. O.
Box 15027, Las Vegas, Nevada 89114 (in publication).
Presents the techniques and detailed procedures developed at EPA's
Environmental Monitoring Systems Laboratory, Las Vegas, for processing
aircraft multispectral scanner data using an interactive computer system.
Updated Overhead Multispectral Scanner Monitoring of Western
Energy Coal Strip Mines. EPA, Environmental Monitoring Systems
Laboratory, P. O. Box 15027, Las Vegas, Nevada 89114 (in publication).
Presents the updated data processing procedure developed during
Phase II I operati ons 'of the Western Energy Project and describes the
upgraded capabilities of the EPA Data Analysis System.
28
-------
I'
f
II
,I
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO. 12. 3. RECIPIENT'S ACCESSION NO.
EPA-600/4-80-05l
4. TITLE AND SUBTITLE 5. REPORT DATE
SUMMARY OF THE WESTERN ENERGY OVERHEAD November 1980
MONITORING PROJECT 6. PERFORMING ORGANIZATION CDDE
7. AUTHORIS) 8. PERFORMING ORGANIZATION REPORT NO.
Paul Ishikawa, Jr.
9. PERFORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM ELEMENT NO.
Lockheed Engineering and Management
Servi ces Company, ~nc. 11. CONTRACT/GRANT NO.
Remote Sensing Laboratory
Las Vegas, Nevada' 89114 EPA 68-03-2636
12. SPONSORING AGENCY NAME AND ADDRESS 13. TYPE OF REPORT AND PERIOD COVERED
U.S. Envi ronmental Protection Agency--Las Vegas, NV Final (7-1-75/10-31-79)
Office of Research-and Development 14. SPONSORING AGENCY COOE .
Envi ronmental Monitoring Systems Laboratory
Las Veqas, Nevada 89114 EPA/600/07
15. SUPPLEMENTARY NOTES
G. A. Shelton, Project Officer, Advanced Monitoring Systems Division,
Environmental Monitoring Systems Laboratory, Las Vegas, Nevada 89114
16. ABSTRACT
The Environmental Protection Agency and the National Aeronautics and Space
Administration entered into a five-year overhead monitoring project in June, 1975. The
purpose of this joint project was to transfer, from NASA to EPA, hardware and software
technology for processing remotely 'sensed digital data and to assist EPA in developing
and maintaining an operational remote sensing monitoring system. The overa 11 object i ve
was to define, develop, and demonstrate operational remote sensing techniques to
rapidly moni tor, in a cost-effecti ve manner, the success with whi ch an energy-rel ated
extraction site has been, or is being, rehabilitated. This report discusses the
technology transfer that has successfully taken place. It also describes the remote
sensing monitoring system EPA has established and concludes that this system can be
used to effectively monitor surface mining reclamation activities on a case-by-case
basis but is less suitable for production-type monitoring requiring the quick
turnaround processing of data for hundreds of energy-related extraction sites.
17. -
KEY WORDS AND DOCUMENT ANAL YSIS
a. DESCRIPTORS b.IDENTIFIERS/OPEN ENDED TERMS c. COSA TI Field/Group
Computer hardware/software 62A,B
Remote sensing 68C
Energy research 97G
Mul t i spect ra 1 scanning
Aerial photography
18. DISTRIBUTION STATEMENT 19. SECURITY CLASS (This Report) 21. NO. OF PAGES
UNCLASSIFIED 36
20. SECURITY CLASS (This page) 22. PRICE
RELEASE TO PUBLIC UNCLASSIFIED
EPA Form 2220-1 (Rev. 4-77)
PREVIOUS EDITION IS OBSOLETE
------- |