v>EPA
PO
EPA -'
T78
Remote Monitoring of
Coal Strip Mine
Rehabilitation
Interagency
Energy-Environment
Research
and Development
Program Report
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development. US. Environmental
Protection Agency, have been grouped into nine series These nine broad categories
were established to facilitate further development and application of environmental
technology. Elimination of traditional grouping was consciously planned to foster
technology transfer and a maximum interface in related fields. The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4 Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7 Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the INTERAGENCY ENERGY—ENVIRONMENT
RESEARCH AND DEVELOPMENT series Reports in this series result from the effort
funded under the 17-agency Federal Energy/Environment Research and Development
Program. These studies relate to EPA'S mission to protect the public health and welfare
from adverse effects of pollutants associated with energy systems The goal of the Pro-
gram is to assure the rapid development of domestic energy supplies in an environ-
mentally-compatible manner by providing the necessary environmental data and
control technology. Investigations include analyses of the transport of energy-related
pollutants and their health and ecological effects; assessments of. and development of.
control technologies for energy systems; and integrated assessments of a wide range
of energy-related environmental issues.
This document is available to the public through the National Technical Information
Service, Springfield. Virginia 22161
-------
EPA-600/7-78-149
July 1978
REMOTE MONITORING OF COAL STRIP MINE REHABILITATION
by
James E. Anderson
National Aeronautics and Space Administration
Earth Resources Laboratory
Slidell, Louisiana 70458
and
Charles E. Tanner
Lockheed Electronics Company, Inc.
Las Vegas, Nevada 89114
Under Contract 68-03-2636
G. J. D'Alessio
Project Officer
Western Energy/Environment Monitoring Study
U.S. Environmental Protection Agency
Washington, D.C. 20460
ENVIRONMENTAL MONITORING AND SUPPORT 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 and Support
Laboratory-Las Vegas, 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 recom-
mendation for use.
ii
-------
FOREWORD
Protection of the environment requires effective regulatory actions which
are based on sound technical and scientific information. This information must
include the quantitative description and linking of pollutant sources, trans-
port mechanisms, interactions, and resulting effects on man and his environment.
Because of the complexities involved, assessment of specific pollutants in the
environment requires a total systems approach which transcends the media of air,
water, and land. The Environmental Monitoring and Support Laboratory-Las Vegas
contributes to the formation and enhancement of a sound monitoring data base
for exposure assessment through programs designed to:
• develop and optimize systems and strategies for moni-
toring pollutants and their impact on the environment
• demonstrate new monitoring systems and technologies by
applying them to fulfill special monitoring needs of
the Agency's operating programs
This report discusses and evaluates the utility of aircraft multispectral
scanner data as a tool for monitoring surface strip mining operations in the
northern Great Plains and the western portion of the United States. It also
demonstrates the feasibility of using Landsat data as a regional planning tool
for pre-mining environmental impact evaluation. Procedures and techniques used
in processing digital data and interpreting aerial photographs of selected
strip mines are also demonstrated. This report should be of interest to those
State and Federal agencies who have the responsibility of enforcing the recent-
ly passed Surface Mining and Reclamation Act of 1977.
GeorgcB Morgan
Director
Environmental Monitoring and Support Laboratory
Las Vegas
iii
-------
ABSTRACT
The U.S. Environmental Protection Agency (EPA) and the National Aeronautics
and Space Administration (NASA) entered into a five-year, EPA-funded project to
transfer hardware and software technology for processing remotely sensed digital
data obtained from aircraft or satellite platforms.
This project was divided into three phases. Phase I was an 18-month task.
During this time, state-of-the-art technology for processing aircraft-acquired
multispectral scanner (MSS) data was transferred to the Environmental Monitoring
and Support Laboratory in Las Vegas, Nevada (EMSL-LV). Also, Landsat and air-
craft multispectral scanner data and photographic data over coal strip mines in
the Western United States were analyzed using basic pattern recognition tech-
niques refined and/or developed at the NASA/Earth Resources Laboratory (ERL)
in Slidell, Louisiana.
This report discusses the accomplishments of the Phase I operations and
also compares the results of manual photo-interpretation and automated data
analysis conducted during this phase. Also included in this report are the
results of a feasibility study to utilize Landsat data for performing a regional
land-cover classification of a portion of the Powder River Basin area in north-
eastern Wyoming where thpre are numerous active coal strip mines.
Phase II of the joint project will be of 18 months duration. EMSL-LV will
use the system for monitoring selected coal strip mines in the west, and ERL
will pursue research tasks identified during Phase I.
Phase III will be 2 years in duration during which EMSL-LV will test the
system in an operational mode with continued software development and assistance
from NASA/ERL.
iv
-------
TABLE OF CONTENTS
Foreword ..............................
Abstract .............................. iv
Figures .............................. vii
Tables ............................... viii
English-to-Metric Conversion Table ................. ix
1. Introduction ....................... 1
2. Conclusions ....................... 3
3. Recommendations ..................... 4
A. Data Acquisition ..................... 5
Aircraft Data .................... 5
Ground Truth .................... 7
5. Data Reduction and Processing Procedures ......... 8
Procedures for Aircraft Multispectral Scanner Data .... 8
Conventional Analysis Procedures ............. 11
6. Results of the Phase I Operations ............ 14
Technology Transfer to the Environmental Monitoring
and Support Laboratory-Las Vegas ............. 14
Automated and Manual Analysis of Aircraft-Acquired
Data Over Selected Coal Strip Mines ........... 15
Belle Ayr Coal Strip Mine .............. 15
Black Mesa Coal Strip Mine ............. 21
Dave Johnston Coal Strip Mine ............ 25
Navajo Coal Strip Mine ............... 30
Regional Application of Landsat Multispectral
Scanner Data ....................... 37
Comparative Analysis of the Automated and Manual
Classification Results .................. 40
7. Summary
-------
TABLE OF CONTENTS (Continued)
Page
Bibliography 44
A. Appendix A - Data Acquisition Systems 46
Airborne Multispectral Scanner .... 46
Aerial Camera 46
Landsat Multispectral Scanner .... 49
B. Appendix B - Data Reduction and Processing
Procedures 51
Preprocessing and Data
Transformation 51
Pattern Recognition 51
C. Appendix C - Scientific Names of Vegetative Types
Identified in Phase I Operations 58
vi
-------
FIGURES
Number Page
1 Locations of Mine Sites Surveyed in the NASA/ERL Phase I
Operation 6
2 Data Analysis System (DAS) 9
3 Multispectral Scanner Data Reduction and Processing Flow. . 10
4 Classification Results (Non-Mine-Related Classes) for the
Belle Ayr Coal Strip Mine, Campbell County, Wyoming ... 16
5 Classification Results (Mine-Related Classes) for the
Belle Ayr Coal Strip Mine, Campbell County, Wyoming ... 17
6 Photo-Interpreted Land-Cover Map of the Belle Ayr Coal
Strip Mine 18
7 Color-Coded Land-Cover Classification Map of Aircraft-
Acquired MSS Data for the Black Mesa Coal Strip Mine . . 22
8 Photo-Interpreted Land-Cover Map of the Black Mesa Coal
Strip Mine 23
9 Color-Coded Land-Cover Classification Map of Aircraft-
Acquired MSS Data for the Dave Johnston Coal Strip
Mine, Converse County, Wyoming 26
10 Photo-Interpreted Land-Cover Map of the Dave Johnston
Coal Strip Mine (Section A) 27
11 Photo-Interpreted Land-Cover Map of the Dave Johnston
Coal Strip Mine (Section B) 28
12 Color-Coded Land-Cover Classification Map of Aircraft-
Acquired MSS Data for the Navajo Coal Strip Mine .... 31
13 Photo-Interpreted Land-Cover Map of the Navajo Coal
Strip Mine (Section A) 32
14 Photo-Interpreted Land-Cover Map of the Navajo Coal
Strip Mine (Section B) 33
15 Photo-Interpreted Land-Cover Map of the Navajo Coal
Strip Mine (Section C) 34
16 Photo-Interpreted Land-Cover Map of the Navajo Coal
Strip Mine (Section D) 35
17 Landsat Land-Cover Classification of Powder River Basin
Area, Wyoming 38
vii
-------
TABLES
Number Page
1 Land-Cover Classification Hierarchy for Coal Strip
Mine Monitoring 12
2 Acreage Statistics for Mine-Related Classes Derived
from Aircraft-Acquired MSS Data for the Belle Ayr
Coal Strip Mine 19
3 Acreage Statistics for Non-Mine-Related Classes
Derived from Aircraft-Acquired MSS Data for the
Belle Ayr Coal Strip Mine 20
4 Acreage Statistics from the Aircraft-Acquired MSS
Data for the Black Mesa Coal Strip Mine 24
5 Acreage Statistics Derived from Aircraft-Acquired
MSS Data for the Dave Johnston Coal Strip Mine 29
6 Acreages Derived from Aircraft-Acquired MSS Data
for the Navajo Coal Strip Mine 36
7 Acreage Statistics for the NASA/ERL Landsat Classi-
fication 39
8 Chi-Square Contingency Table (2 x 9) for the Black
Mesa Coal Strip Mine 41
9 Chi-Square Contingency Table (2x8) for the Dave
Johnston Coal Strip Mine 41
10 Chi-Square Contingency Table (2 x 8) for the Navajo
Coal Strip Mine 41
11 Chi-Square Contingency Table (2 x 8) for the Belle
Ayr Coal Strip Mine 41
viii
-------
ENGLISH-to-METRIC CONVERSION TABLE
English Unit
1 Foot
1 Inch
1 Inch
1 Mile
1 Square Foot
1 Square Foot
1 Acre
Metric
0.3048 Meters
2.54 Centimeters
25.40 Millimeters
1.609 Kilometers
929.03 Square Centimeters
0.0929 Square Meters
0.404 Hectare
ix
-------
SECTION 1
INTRODUCTION
The U. S. Environmental Protection Agency (EPA) and the National Aeronau-
tics and Space Administration (NASA) entered into a five-year, EPA-funded pro-
ject in the summer of 1975. The purpose of this interagency project was to
transfer hardware and software techniques for processing remotely sensed digi-
tal data from NASA to EPA in order that EPA could develop the capability of
establishing and maintaining a fully operational remote sensing monitoring
system. This system would not only incorporate maximum utilization of EPA's
present aircraft capabilities, but would allow for the integration of currently
available and proposed Landsat multispectral scanner (MSS) data.
Both agencies have designated laboratories through which the five-year
plan is being implemented. The NASA/Earth Resources Laboratory (ERL) located
in Slidell, Louisiana, and the EPA/Environmental Monitoring and Support labora-
tory in Las Vegas, Nevada (EMSL-LV), are the two agency laboratories involved.
Unless otherwise indicated, "ERL" and "EMSL-LV" are used to refer to these two
laboratories respectively.
The overall objective of this five-year project is to define, develop,
and demonstrate operational remote sensing techniques to rapidly monitor, in
a cost-effective manner, the success with which an energy-related extration
site has been, or is being rehabilitated to a state suitable for its intended
or previous land usage. The specific objectives are 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 classi-
fication with the photo-interpretation results
The two agencies agreed to divide the five-year project into three distinct
phases based on operational requirements. Phase I, which began in July of 1975,
encompased an 18-month period during which the following activities were con-
ducted :
1
-------
(1) A Data Analysis System (DAS) was defined and assembled at ERL, checked
out, and in January 1977, was transferred to EMSL-LV.
(2) EMSL-LV personnel received training at ERL in the use of the DAS for
processing aircraft and Landsat acquired multispectral scanner (MSS)
digital data.
(3) ERL defined an airborne multispectral scanner Data Collection System for
installation into an EMSL-LV aircraft.
(4) A NASA aircraft equipped with a multispectral scanner and camera was
used to acquire scanner data and aerial photography over selected coal
strip mines in western United States.
(5) Vegetation/land-cover classification of the selected mine sites was
produced at ERL using the scanner data and computer-implemented pattern
recognition techniques, and at EMSL-LV through use of the aerial photo-
graphy and manual photo-interpretation techniques.
(6) One set of Landsat acquired MSS digital data was processed at ERL to
gain insight into approaches and techniques for producing land-cover
classifications for regional monitoring.
This report focuses on the results of activities (5) and (6). However,
technical details of the data acquisition, data processing, and analytical
procedures are addressed in the text and in the appendices. The particular
objective was to develop and demonstrate the techniques and procedures for
deriving vegetation/land-cover classifications of coal strip mine sites and
their near environs for the purpose of subsequent monitoring of the rehabilitation
of mined areas. The results are analyzed with respect to the vegetation/land-
cover classes that were spectrally separable and the areal extent of each class.
In addition, a comparison of results from multispectral scanner data and aerial
photography is made.
Phase II of the project was begun in January of 1977, and will last for
18 months. During this phase, EMSL-LV will use the system on a shakedown
basis for monitoring purposes. Also during this time, ERL (utilizing a
similar system) will investigate those problems specifically defined by
EMSL-LV which require additional research. New techniques developed to
solve these problems will be transferred to EMSL-LV at the end of this phase.
During Phase III, EMSL-LV will test the system in an operational mode
for 2 years. ERL will provide continued assistance in the use of the system
and supply additional software programs as they are developed during Phases II
and III.
-------
SECTION 2
CONCLUSIONS
The results of Phase I analyses indicate the feasibility of using aircraft-
acquired multispectral scanner data for monitoring and evaluating coal strip
mine rehabilitation activities. These results support the following specific
conclusions:
Multispectral scanner data acquired by an aircraft at an
altitude of approximately 12,000 feet (3,660 meters) appears
to possess sufficient spatial resolution (element size) and
spectral resolution (band location and width) for vegetation/
land-cover classifications of western coal strip mine sites
and their near environs.
A chi-square analysis showed no significant differences in the
results of the land-cover mapping achieved through computer-
implemented techniques and the land-cover mapping resulting from
conventional photo-interpretation methods for the land cover
classes included in the analysis.
Adequate details of land-cover features can be obtained at
12,000 feet (3,660 meters). It is further concluded that
the additional details gained from low-altitude coverage
would not compensate for the necessary increase in data
processing.
Results of the preliminary land-cover classifications of a
large area using Landsat data were sufficiently encouraging
to proceed with a more detailed study of the use of Landsat
MSS data for regional analysis.
-------
SECTION 3
RECOMMENDATIONS
Based on the results and conclusions from Phase I Operations, the follow-
ing recommendations are proposed for consideration in Phase II of the project:
Define and conduct a task for use of Landsat data for
regional analysis in Phase II of project
Investigate the geometric correction of aircraft-acquired
MSS data, in order that the land-cover/vegetation maps
derived from this data can be fit to topographic maps of
specific scales
Develop techniques for reducing ground truth efforts
required for monitoring after a baseline land-cover/vege-
tation classification has been produced
Continue acquiring and processing aircraft-acquired MSS
data for the purpose of establishing operational proce-
dures
-------
SECTION 4
DATA ACQUISITION
AIRCRAFT DATA
Data acquisition was scheduled to occur during the months when the chlor-
phyll content of vegetation was at its maximum at the selected coal strip mine
sites (Figure 1). This "peak-green window" was determined by making inquiries
of the reclamation specialist on each of the mines being studied. From these
contacts, it was ascertained that Montana, Wyoming and Colorado have peak-green
conditions that can occur from the last week of June through the first two weeks
of July. Arizona and New Mexico peak-green conditions occur in late September.
The following aircraft data collection sensors and parameters were required
for each flight line flown:
(1) Multispectral scanner data collection at altitudes of 3,000,
6,000, and 12,000 feet (914, 1,828, and 3,657 meters) above
mean terrain elevation
(2) Photographic coverage at altitudes of 1,000, 3,000, 6,000,
and 12,000 feet above mean terrain elevation (305, 914, 1,828,
and 3,657 meters), using both color and color-infrared films
with 60% forward overlap
Flight lines were drawn on partial and full-frame color-infrared film en-
largements (40-by-40-inch or 1.02-meter format) and also black-and-white full-
frame enlargements produced by EMSL-LV from NASA high-altitude aerial photo-
graphy acquired prior to the low-altitude data collection. (For a more detailed
breakdown of flights flown for the Western Energy Project, see the "Western
Energy/Environment Monitoring Atlas", EPA 600/7-77-047a published in May, 1977).
These photographic enlargements, annotated with required flight lines,
were placed in a pilot's package, which included other items, such as start/
stop coordinates (latitude-longitude), for the flight lines, and topographic
maps showing various recognizable land surface features. The pilots used the
information contained in the packet to program the flight schedule. All air-
craft data were collected by the Data Acquisition System on board NASA/Johnson
Space Center's NP3A aircraft. After aircraft data collection was completed for
each mine site, the film was processed by EMSL-LV and duplicate positive trans-
parencies and prints were supplied to ERL for their analysis. The MSS data
were processed at ERL. (See Appendix A for description of data acquisition
systems).
-------
Figure 1. Locations of Mine Sites Surveyed in the NASA/ERL
Phase I Operation.
-------
GROUND TRUTH
The color-infrared enlargements were also used as a base from which train-
ing sample sites were chosen. A training sample site is an area of the Earth's
surface which appears to be homogeneous in composition when observed from above.
Examples include a coal seam, a revegetation plot, etc. Such areas are subse-
quently used to "train" the computer to recognize these and/or similar land-
cover categories, hence the term "training" sample site. Each training sample
site was chosen and located on the color-infrared enlargements, and assigned
a consecutive number or code. Prints of these aerial photographs were taken
into the field for the ground truth operation. (See Appendix B for details of
selecting training samples.)
-------
SECTION 5
DATA REDUCTION AND PROCESSING PROCEDURES
PROCEDURES FOR AIRCRAFT MULTISPECTRAL SCANNER DATA
Processing of the aircraft MSS data was accomplished through task-designed
software modules on the Data Analysis System (DAS) (Figure 2). These modules
are linked in such a manner that output from one module serves as input to
another. However, each module can be composed of several software programs
that accomplish one major task. Figure 3 illustrates a simplified flow dia-
gram which outlines the reduction and processing of data from an aircraft
multispectral scanner. There are only minor differences in processing air-
craft MSS data and Landsat MSS data.
Following data acquisition, the MSS data are decommutated and reformatted.
The decommutation/reformatting process converts the PCM (Pulse-Code Modulation)
format of the priginal aircraft data into a digital format compatible with the
computer system. The digital tape that is produced by the decommutation/re-
formatting procedure is checked for data anomalies such as, sun angle mani-
festations, missing data, recording problems, etc. Most anomalies can be rec-
tified by the application of software programs in the pre-processing and trans-
formation block as seen in Figure 3. A data quality check can be performed by
analyzing a hardcopy product (resulting from intermediate processing) or view-
ing the data on the Interactive Display System.
In the Pattern Recognition Modules a classification of the data set is
initiated.
The supervised approach to pattern recognition requires that the analyst
know something about the scene, because he is responsible for "training" the
computer to recognize the various land-cover categories that are of interest.
This is accomplished by outlining areas with known physical characteristics
and then by computing statistics (means, variance, correlation matrix, etc.)
for each training sample and grouping the samples into land-cover classes.
(For details, see Appendix B).
The Pattern Recognition Modules were written to accommodate a maximum of
4 channels of digital data; therefore, it is necessary to run the Channel
Selection Program to determine the 4 best channels for use in classifying each
data set. Once the 4 channels have been determined, the analyst can begin
classifying the data, using the approach described in Appendix B.
-------
OPERATORS TERMINAL
AND CARD READER
9 TRACK MAGNETIC
TAPE DRIVES
PLAYBACK SYSTEM
AND CENTRAL COMPUTER
INTERACTIVE DISPLAY SYSTEM
COLOR FILM RECORDER
Figure 2. Data Analysis System (DAS)
-------
fSENSOR \
PREPROCESSING
DATA
TRANSFORMATION
HARD COPY
OUTPUT
DEVICE
COMPUTER
COMPATIBLE
TAPE
PATTERN RECOGNITION
• TRAINING SAMPLE SELECTION
•STATISTICS COMPUTATIONS
•CHANNEL SELECTION
•CLASSIFICATION
ACREAGE
COMPILATIONS
THEME
INVENTORY
ETC.
Figure 3. Multispectral Scanner Data Reduction and
Processing Flow (Simplified).
10
-------
CONVENTIONAL ANALYSIS PROCEDURES
Interpretation of aerial photography is a well-developed art. Experienced
photo-interpreters can detect and identify man-made and natural features of the
Earth's surface by means of shape, form, color, texture, pattern, and shadow.
Objects can also be detected by indirect means, such as deducting their pre-
sence from the recognition of objects with which they are normally associated.
The photo analysis of the coal strip mine aerial photography was performed
using a zoom-stereomicroscope, a light table, and the classification hierarchy
developed by EMSL-LV personnel for analysis of coal strip mining areas (Table
1). First, appropriate stereo-pairs of each coal strip mine were located within
the roll of duplicate positive transparencies. Next, transparent overlay ma-
terial was placed over one of the frames of imagery, and the features of in-
terest were outlined and annotated on this overlay. When the frame of imagery
was completely interpreted, the overlay was sent to the Cartographic Laboratory
for scribing. Scribing produces a "clean" product, e.g., lines are of equal
width, numbers are the same size, etc.
11
-------
TABLE 1. LAND-COVER CLASSIFICATION HIERARCHY FOR COAL STRIP MINE MONITORING
00 - Active Mine Features
01 - Advance cut
02 - Topsoil stockpile including embankment
03 - Stripping bench, high wall, 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 reservoirs
24 - Water courses, including canals
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
(Continued on next page)
12
-------
TABLE 1 (Continued)
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 - Institutional
64 - Industrial
65 - Transportation, communications and utilities
66 - Resource extraction
13
-------
SECTION 6
RESULTS OF THE PHASE I OPERATIONS
Phase I operations for the first 18 months have been concluded and all of
the specific objectives have been addressed by ERL and/or EMSL-LV. For con-
venience, the specific objectives are listed below:
Transfer the data analysis system and its associated
software technology and maintenance requirements tech-
nology 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
Classification with the photo-interpretation results
TECHNOLOGY TRANSFER TO THE ENVIRONMENTAL MONITORING AND SUPPORT LABORATORY-
LAS VEGAS
The first objective was accomplished in two separate stages. Stage 1, a
technology transfer session for EMSL-LV personnel, was conducted at ERL during
the second week in October and again during the first week in December (1976).
Both sessions lasted 3 weeks. During the first session, EMSL-LV personnel at-
tended lectures which detailed the software and hardware aspects of the system
and processed canned data sets under ERL contractor supervision. Session 2
began with a 3-day review and refamiliarization of the system. The remaining
12 days were used to perform a supervised classification of a strip mine data
set. This task included all the reformatting steps in the data reduction and
processing flow required to produce a hardcopy color print of the classifica-
tion results.
Stage 2 was completed in January of 1977 when personnel from ERL delivered,
setup, and performed the necessary tests to make the EMSL-LV DAS a separate and
operational system.
14
-------
AUTOMATED AND MANUAL ANALYSIS OF AIRCRAFT-ACQUIRED DATA OVER SELECTED COAL
STRIP MINES
To accomplish this objective, ERL personnel processed aircraft-acquired
MSS data over selected coal strip mines or lease areas while EMSL-LV personnel
analyzed the aerial photography corresponding to the MSS coverage. Four MSS
data sets, one mine per set, were acceptable in data quality. Four additional
data sets were not accepted due to tape recorder or scan angle/sun angle pro-
blems. The results of the supervised classification approach, photo-interpre-
tations, and descriptive information about the mines are presented on the
following pages.
BELLE AYR COAL STRIP MINE
The Belle Ayr Mine is located near Gillette (Campbell County), Wyoming,
(the extreme northeastern corner of the state). Belle Ayr is operated by AMAX
Coal Company and produces over 275,000 tons of coal per month. This coal is
transported to the Public Service Companies of Pueblo, Denver, and Boulder,
Colorado, and the Kansas City Power and Light Company of Burlington, Iowa.
Vegetation of this area is of the sagebrush-steppe type. This vegetation
type is dominated by sagebrush in open grassland containing wheatgrasses, ne-
edle-grasses and needle-and-threadgrasses on silt to silty clay-loam soils.
(See Appendix C for scientific names of vegetation types). This type occurs
in northeastern Wyoming and most of the major species have moderately good
suitability for rehabiliation, but are relatively unavailable for that purpose.
Eight channels of aircraft-acquired MSS data were decommutated/reformatted,
and a subset of the four best channels was used to generate the classification
maps (Figure 4 and Figure 5). The supervised approach to automatic data pro-
cessing of MSS data was used for the Belle Ayr Mine MSS data set. Figure 6
presents the results of the EMSL-LV photo-interpretation of approximately the
same area. The classification hierarchy for this photo analysis is presented
in Table 1 (located in Section 5 of this report).
To facilitate an image-to-image evaluation of the results, the analyst
conducted two computer runs to separate mine-related from non-mine-related
classes as seen in Tables 2 and 3. In general, the classification of non-
mine-related features agreed with the photo-interpreted results and ground
checks. Note the olive green area in the bottom right hand portion of the
image and how it compares with the photo results. Note also, the amount and
location of bare soil areas on each image. Even the diverted Caballo Creek
is classified. Finally, note the distribution of the natural vegetation and
how the classified image compares with the photo analysis. The classification
of the mine-related features also agreed with the photo-interpreted results
and ground checks. Note the identification of the buildings comprising the
mine facility (white area) and the areal extent of the active mining pit.
Finally note, in the upper portion of the image, the sharply defined "advance-
cut" areas.
15
-------
ALFALFA WHEAT GRASS
AGRICULTURAL CROP/PASTURE
HARVESTED CROPS
FALLOW FIELDS
DISKED SOILS (DRY)
DISKED SOILS (MOIST)
WATER
BARE SOILS
NEEDLE and THREAD/BROME/MISC GRASSES
GRAZED NATURAL VEGETATION
NATURAL VEGETATION
UNCATEGORIZED
o 100 1000 1100
• All *">
Figure 4. Classification Results (Non-Mine Related Classes)
for the Belle Ayr Coal Strip Mine, Campbell
County, Wyoming.
16
-------
COAL SEAM
MINE PIT
MINE PIT WAILS
EXPOSED SOILS (RAILROAD LOOP>
SCRAPED TOPSOIl
PILED TOPSOIL
PILED OVERBURDEN
BUILDING
REVEGETATION PLOTS
WATER
UNCATEGORIZED
Figure 5. Classification Results (Mine-Related Classes)
for the Belle Ayr Coal Strip Mine, Campbell
County, Wyoming.
17
-------
oo
Figure 6. Photo-Interpreted Land-Cover Map of the Belle Ayr Coal Strip
Mine. (See Table 1 for a key to EMSL-LV Land-Cover numbering
sys t em.)
-------
TABLE 2. ACREAGE STATISTICS FOR MINE-RELATED CLASSES DERIVED FROM
AIRCRAFT-ACQUIRED MSS DATA FOR THE BELLE AYR COAL STRIP MINE
Class Name Acres
Coal Seam 11
Mine Pit 19
Mine Pit Walls 21
Exposed Soils (Railroad Loop) 20
Scraped Topsoil 47
Piled Topsoil 13
Piled Overburden 9
Building 41
Revegetation Plots 4
Water 2
Uncategorized * 797
* Includes non-mine-related classes on P. 20
19
-------
TABLE 3. ACREAGE STATISTICS FOR NON-MINE-RELATED CLASSES DERIVED FROM
AIRCRAFT-ACQUIRED MSS DATA FOR THE BELLE AYR COAL STRIP MINE
Class Name Acres
Alfalfa/Wheat grasses 4
Agricultural Crop 8
Pasture 1
Harvested Crops 1
Fallow Fields 15
Disked Soils (Dry) 3
Disked Soils (Moist) 41
Water 77
Bare Soils 155
Grazed Natural Vegetation 18
Natural Vegetation 55
Uncategorized * 606
* Includes mine-related classes on P. 19
20
-------
BLACK MESA COAL STRIP MINE
The Black Mesa Coal Strip Mine is located in the northeastern corner of
the State of Arizona, in Navajo and Apache counties. It is operated by Peabody
Coal Company headquartered in St. Louis, Missouri. Black Mesa is in the 500,000-
and-over tonnage class and strips in excess of 20,000 tons per day. The total
tonnage mined in 1974 (which is the last year statistics are available) was
3,933,493 tons. Coal from this mine is transported via pipeline to power plants
outside the local area. Average elevation on the lease area is approximately
6,500 feet (1,980 meters), placing it in the pinyon pine/juniper vegetation
zone.
Dominant vegetation types within the lease area are:
Pinyon Pine Snakeweed
Juniper Sagebrush
Alkali Sacatone Blue Gramma
Greasewood Winterfat
Rabbit Brush
Eight channels of data were also decommutated and reformatted for this mine
and the four best channel subset was used in generating the classification
map (Figure 7). For comparative purposes, the photo-interpreted image is pre-
sented in Figure 8. Note the amount of detail that was produced from classi-
fying the digital data, e.g., revegetated areas (red on the classified image)
and the relative distribution of the pinyon/juniper association. Also, note
how well the location and distribution of the piled overburden on the two
images compare. The acreage software program available on the DAS eliminates
the need to perform dot counts or use of a planimeter to calculate acreages.
Table 4 was developed through the use of such software and is reasonably ac-
curate when all input parameters are known and entered correctly.
21
-------
•
sons
IOOSI COAI
Figure 7. Color-Coded Land-Cover Classification Map of
Aircraft-Acquired MSS Data for the Black Mesa
Coal Strip Mine.
22
-------
Figure 8. Photo-Interpreted Land-Cover Map of the Black Mesa Coal
Strip Mine. (See Table 1 for a key to EMSL-LV land-cover
numbering system.)
-------
TABLE 4. ACREAGE STATISTICS DERIVED FROM THE AIRCRAFT-ACQUIRED
MSS DATA FOR THE BLACK MESA COAL STRIP MINE
Land Cover Category Acres
Natural Vegetation:
Pinyon Pine/Juniper 584
Rabbit Brush/Sagebrush/Native Grass 1,234
Active Mine Features:
Roads, Exposed Soils 238
Exposed Soils 4
Coal 52
Loose Coal, Mine Pit, Coal-Covered Soil 81
Leveled Overburden 200
Piled Overburden 122
Revegetated Areas 172
24
-------
DAVE JOHNSTON COAL STRIP MINE
The Dave Johnson Coal Strip Mine is located in the north-east central
portion of the State of Wyoming in Converse County. The Dave Johnston Mine
is operated by Pacific Power and Light Company and is termed a "captive mine"
because it ships coal by rail to the Dave Johnson Power Plant located in nearby
Glenrock, Wyoming.
The mine has two coal seams that are being mined at the present time.
The youngest, the Badger seam, is approximately 16 feet in thickness, and the
oldest, the School seam, is roughly 37 feet in thickness. The average depth
of the active pit is about 140 feet with a maximum depth of approximately 180
feet.
Vegetation is of the short-grass prairie type. This type primarily occurs
on dry prairies on shallow soils in southeastern Montana and northeastern
Wyoming. Dominant species are gramma wheatgrasses and various needlegrasses.
The species that characterize this type have moderately poor suitability and
fair availability for rehabilitation.
Aircraft-acquired MSS data were processed in the same manner as the data
from the previous mines. Processing results for this mine, based on field-
verified training samples, produced a classification map which contained about
60 percent uncategorized data. Using the aerial photography as a source, sam-
ples were selected for ground-cover types that were uncategorized, and the
classification was re-run. However, 2.7% of the data which were scattered
over small unrelated areas remained uncategorized (black areas on the classi-
fied image). This indicated that, in terms of land-cover classes, no real
deficiency existed. A visual examination of the classified mine scene indi-
cates that general land-cover categories were classified reasonably well, in-
cluding both mine-related features and surrounding natural land-cover types.
Figure 9 depicts the computer classification results and Table 5 summarizes
the acreage computed for each class. Photo-interpretation results are pre-
sented in Figures 10 and 11.
25
-------
COAL SEAM
RECONTOURED OVERBURDEN
SCRAPED/EXPOSED OVERBURDEN
PIT PONDING
EXPOSED SOILS
INERT MATERIALS
REVEGETATION AREAS
CHEATGRASS/BROME (HIGH DENSITY)
GRAMMA/SAGEBRUSH/BROME
SAGEBRUSH/BROME/NEEDLE and THREAD GRASS
MEDIUM DENSITY NATURAL VEGETATION
UNCATEGORIZED
fV>
mni
VCAII l»»ftOII««!l> I l>411
Figure 9. Color-Coded Land-Cover Classification Map of
Aircraft-Acquired MSS Data for the Dave Johnston
Coal Strip Mine, Converse County, Wyoming.
26
-------
Figure 10. Photo-Interpreted Land-Cover Map of the Dave Johnston Coal
Strip Mine, Section A. (See Table 1 for a key to EMSL-LV
land-cover hierarchy system.)
-------
CO
oo
Figure 11. Photo-Interpreted Land-Cover Map of the Dave Johnston Coal
Strip Mine, Section B. (See Table 1 for a key to EMSL-LV
land-cover hierarchy system.)
-------
TABLE 5. ACREAGE STATISTICS DERIVED FROM AIRCRAFT-ACQUIRED MSS DATA
FOR THE DAVE JOHNSTON COAL STRIP MINE
Class Name Acres
Coal Seam 116
Recontoured Spoils 481
Scraped/Exposed Overburden 83
Pit Ponds 11
Exposed Soils 1,907
Inert Materials 315
Revegetation 504
Cheatgrass/Brome (High Density) 291
Gramma/Sagebrush/Brome 2,369
Sagebrush/Brome/Needle and Threadgrasses 2,213
Medium Density Natural Vegetation 642
Uncategorized 247
29
-------
NAVAJO COAL STRIP MINE
The Navajo Mine is located in San Juan County, New Mexico, in the extreme
northwestern corner of the State. The mine is also located in the largest and
best known coal field in the San Juan Basin, the Navajo field, which extends
about 35 miles from the San Juan River southward along and to the east of the
outcrop of the Fruitland formation. The Navajo Mine is operated by Utah In-
ternational, which holds the leases for the northern two-thirds of the field
while El Paso Natural Gas Company holds the southern leases.
The coal is subbituminous B to A rank with a slight decrease in quality
southward owing to an increase in ash content. In 1974, this mine produced
6,955,000 tons of coal. The preceeding year, it produced over 7.3 million tons.
Vegetation in the San Juan Basin is limited to those species found in
grassland associations. There is a considerable amount of the various grama
species, galleta, snakeweed, sacatone and Mormon tea within the lease area of
the mine. This is by far, the longest mine studied in Phase I.
The same processing procedures were used to select a subset of four chan-
nels for use in supervised classification runs. The results of the 62-category
classification are presented in Figure 12. This classification required 6.5
hours of computer processing time. These 62 categories had to be merged into
6 broad classes that correspond to Level I of the hierarchy in Table 1.
For comparative purposes, the photo-interpreted images are presented in
Figures 13 through 16. Note in Figure 16 the branching of the San Juan River
at the top of the photo and how it was resolved and classified on the digital
data set. Also, note, the amount and extent of the recontoured areas on the
photo-interpreted images and how well they correspond with the classified images.
Table 6 summarizes the results of the classification. By far the greatest num-
ber of acres fall into the natural vegetation class. The general category of
mine-related features ranks second in size.
30
-------
OVERBURDEN
LEVELED OVIRBURDEN
COAL COVERED SOIL
EXPOSED COAl SfAM
OVERBURDEN SOILS
NATURAL VEGETATION
BUSH TREE VEGETATION
NATURAL VEGETATION
NATURAL VEGETATION
NATURAL VEGETATION
AGRONOMIC
BARE SOIL
Figure 12.
Color-Coded Land-Cover Classification Map of
Aircraft-Acquired MSS Data for the Navajo Coal
Strip Mine.
31
-------
OJ
Figure 13. Photo-Interpreted Land-Cover
Mine, Section A. (See Table
cover hierarchy system.)
Map of the Navajo Coal Strip
1 for a key to EMSL-LV land-
-------
U)
CO
Figure 14.
Photo-Interpreted Land-Cover Map of the Navajo Coal Strip
Mine, Section B. (See Table 1 for a key to EMSL-LV land-
cover hierarchy system.)
-------
u>
Figure 15.
Photo-Interpreted Land-Cover Map of the Navajo Coal Strip
Mine, Section C. (See Table 1 for a key to EMSL-LV land-
cover hierarchy system.)
-------
Ul
Figure 16. Photo-Interpreted Land-Cover
Mine, Section D. (See Table
cover hierarchy system.)
Map of the Navajo Coal Strip
1 for a key to EMSL-LV land-
-------
TABLE 6. ACREAGES DERIVED FROM AIRCRAFT-ACQUIRED MSS DATA
FOR THE NAVAJO COAL STRIP MINE
Class Acres
Water 94
Agronomic 1,780
Natural Vegetation 11,449
Mine-Related:
Coal 238
Overburden 1,853
Coal-Covered Soils, Haul Roads, Exposed Soils 866
36
-------
REGIONAL APPLICATION OF LANDSAT MULTISPECTRAL SCANNER DATA
The use of Landsat data was incorporated into Phase I of this project to
gain insight into approaches and techniques for producing land-cover classi-
fications which are suited to regional monitoring and analysis of energy-
related activities.
Several features associated with Landsat digital data serve to enhance
the use of such data for the "regional" type of analysis. First, costs
associated with computer processing of data covering large areas are relatively
low when compared to the corresponding aircraft data-derived cost figures.
Second, since the satellite system has a regular, established coverage date,
repetitive coverage for change detection is possible. And finally, the data
can be registered (matched) to map coordinates, which permits the integration
of non-satellite ancillary information, e.g., soil type, elevation, geology,
etc., via digitizing apparatus.
The area, selected for this study, approximately 1 degree of latitude by
1 degree of longitude, was outlined on the Gillette 1:250,000 quadrangle.
This area was selected becuase of the presence of a vast number of active
coal strip mining operations in the alluvial valley of the Powder River
Basin.
Both a "supervised" approach as previously described, and an "unsupervised"
approach were used in data processing. The unsupervised approach differs
from the supervised approach mainly in that every element of data is examined
by the computer for inclusion in or initiation of a land-cover class. In the
unsupervised approach, the investigator can set several statistical limits in
such a manner that signatures are developed for a large number of spectral
classes. In the supervised approach, the classes are limited to those land-
cover categories for which training samples were selected. Ground truth data
were not acquired specifically for processing Landsat multispectral scanner
data. Therefore, the same training sample sites that had been verified in
the immediate vicinity of a mine during the summer missions of 1976 were
used. Computer-compatible tapes corresponding to Landsat Scene 5415-16333
(June 7. 1976) were selected and processed.
The land-cover classification was geographically rectified to the Univer-
sal Transverse Mercator projection, scaled to 1:250,000 and placed on that
section of the Gillette quadrangle corresponding to the 1 degree latitude by
1 degree longitude area of interest. The resulting map product was photographi-
cally reduced for inclusion in this report (Figure 17). An acreage compilation
by land-cover category is shown in Table 7. The results obtained appear to
correspond closely to major land-cover patterns known to exist in the area.
Individual mine sites are detectable when colors are assigned to the combined
classes of exposed coal seams, exposed soil and overburden, and exposed
topsoil.
37
-------
00
5
i
IWVSA
MA1A Hf IA1IM •
Figure 17. Landsat Land-Cover Classification of Powder River Basin
Area, Wyoming.
-------
TABLE 7. ACREAGE STATISTICS FOR THE NASA/ERL LANDSAT CLASSIFICATION
Land-Cover Category Acres
Water 6,152
Exposed Coal Seam 500
Exposed Soil/Overburden 22,108
Exposed Topsoil 3,047
Near Bare Soil (Natural) 82,161
Very Sparse Natural Vegetation 563,397
Sparse Natural Vegetation 276,822
Moderately Dense Intermittent Stream Vegetation 259,233
Moderately Dense Natural Vegetation 620,073
Dense Natural Vegetation 679,079
Dense River/Streambed Vegetation 21,839
Agricultural/Natural Vegetation 163,132
Agricultural/Exposed Soil 70,091
Exposed Soil (Natural) Inert Materials 29,183
Clouds/Cloud Shadows i'650
Uncategorized 43,171
39
-------
COMPARATIVE ANALYSIS OF THE AUTOMATED AND MANUAL CLASSIFICATION RESULTS
In order to assess the overall value of our classified multispectral
products, a decision was made to compare them with the photo-interpretation
results. It was fully understood that the photo-interpreted data are not
absolute, but the process does take into account various parameters/indica-
tors that a computer cannot consider because of programming limitations.
The photo-interpreted aerial photographs, simultaneously acquired with
the multispectral scanner data were used as the reference for the evaluation
of each class. Random check points on the photography were selected and
then located on the computer-derived land-cover classification map. A
running total of points, agreement vs. disagreement per class, was recorded
for each mine the the chi-square value was then computed.
According to MacDonald (1966), a measure of the discrepancy existing
between a set of "k" observed frequencies (Oi, 02-..0^) and a corresponding
set of "k" expected frequencies (e^, e2..-ejc) is given by:
+ (02 - e2)2 + ... (Ok - ek)2 = k (Oi - 6i)2
Z
e;L e2 ek i=l 6i
where :
0^ = ith observed frequency
ei = ith expected frequency
If x2 (as calculated from the above equation) is equal to zero, observed
and expected frequencies agree exactly; if x2 is greater than the critical
value in the x2 table, a discrepancy exists. The larger the difference
between x2 an^ tne critical value (assuming x2> critical value) , the greater
the discrepancy between the observed and expected frequencies. If x2 is
found to be larger in magnitude than the critical value, the null hypothesis
being tested (i.e. a discrepancy exists) must be rejected, and it is therefore
concluded there is no significant difference between observed and expected
frequencies (i.e. the alternate hypothesis) .
The following chi-square contingency tables and associated x2 values
have been derived from data check points located (as previously mentioned)
on the Black Mesa, Dave Johnston, Navajo, and Belle Ayr MSS data sets.
For the Black Mesa Mine (Table 8), the critical value X2.99 and ^95
for 8 degrees of freedom, is 20.1 and 15.5 respectively, and 7.86 (the
calculated x2 value) <22.0 and 15.5. Hence, it is concluded that the
photo- interpretation results at the 0.05 or at the 0.01 levels of signifi-
cance are not significantly different from the MSS classification results.
40
-------
TABLE 8. CHI-SQUARE CONTINGENCY TABLE (2 x 9) FOR THE BLACK MESA COAL
STRIP MINE
Land Use Classification
Observed Frequency (MSS)
Expected Frequency (Photo)
X2 = 7.86
32 41 07 32 04
13231
33333
V = 8 degrees of freedom
05 08 41 33
1120
3333
TABLE 9. CHI-SQUARE CONTINGENCY TABLE (2 x 8) FOR THE DAVE JOHNSTON COAL
STRIP MINE
Land Use Classification
Observed Frequency (MSS)
Expected Frequency (Photo)
X2 = 5.37
32 41 07 33 04
42304
55515
V = 7 degrees of freedom
08 22/23 09
334
545
TABLE 10. CHI-SQUARE
STRIP MINE
CONTINGENCY TABLE (2x8) FOR THE NAVAJO COAL
Land Use Classification
Observed Frequency (MSS)
Expected Frequency (Photo)
X2 = 9.54
41 07 32 34 04/06
1262 2
1575 3
V = 7 degrees of freedom
TABLE 11. CHI-SQUARE CONTINGENCY TABLE (2 x 8) FOR
STRIP MINE
Land Use Classification
Observed Frequency (MSS)
Expected Frequency (Photo)
X2 = 10.25
32 50 56 05/04/03
500 3
533 3
V = 7 degrees of freedom
22/23 50 02/03/04/08
34 4
4 10 6
THE BELLE AYR COAL
6 02 22/23 66
121 2
144 4
41
-------
The critical value of x2 QQ and at 7 degrees of freedom for the
• "y •yj
remaining mines (Tables 9, 10, and 11) is 18.5 and 14.1 respectively, and
none of the computed x2 values are greater in magnitude than these critical
values. Therefore, it is concluded that the observed automatic classifica-
tion results do not differ significantly from the photo-interpreted aerial
photographs at the 0.05 or at the 0.01 levels of significance.
-------
SECTION 7
SUMMARY
Phase I operations for the Western Energy Overhead Monitoring Project
were successfully completed. The Environmental Monitoring and Support Labora-
tory in Las Vegas, Nevada, received a Data Analysis System and its personnel
were trained to operate and maintain the system.
Aerial photography and aircraft scanner data over selected coal strip
mines in the Western United States were processed and analyzed to determine
the feasibility of each for monitoring purposes. Also, Landsat MSS data were
processed on the DAS to determine their utility in regional analysis of coal
strip mining operations.
Finally, the results of the interpretation of aerial photography and the
processed aircraft scanner data were compared to identify apparent weak areas
in either approach and to determine if either approach was superior. The x2
(chi-square) test demonstrated that there were no significant differences in
the results obtained using computer implemented techniques and those obtained
using conventional aerial photo-interpretation techniques.
43
-------
SECTION 8
BIBLIOGRAPHY
Cashion, K. D. Method of Improving Scanning Radiometer Data Degraded by
Atmospheric, Attenuation and Sun Angle Effects. Unpublished Report,
1976. NASA, Earth Resources Laboratory, 1010 Cause Blvd., Slidell, LA
70458.
EPA, "Western Energy/Environment Monitoring Atlas". EPA 600/7-77-047a, May,
1977. EPA, Environmental Monitoring and Support Laboratory, P. 0.
Box 15027, Las Vegas, NV 89114.
Hrabal, G. C., and Algranti, J. S. NASA Airborne Instrumentation Research
Program Mission 324 Post Mission Report, 1975. NASA, Lyndon B. Johnson
Space Center, Code CC, Houston, IX 77058. (Cover letter dated December
15, 1975.)
Jones, C. Documentation for Program ELIPSE, Unpublished Software Documentation
1975. NASA, Earth Resources Laboratory, 1010 Cause Blvd., Slidell, LA
70458.
Jones, W. C. Implementation of an Advanced Table Look-Up Classifier for
Large Area Land-Use Classification. NASA/ERL Report No. 115, 1974.
NASA, Earth Resources Laboratory, 1010 Cause Blvd., Slidell, LA 70458.
Jones, W. C. CHOICE: 36-Band Feature Selection Software with Applications
to Multispectral Pattern Recognition, NASA/ERL Report #59, 1976. NASA,
Earth Resources Laboratory, 1010 Cause Blvd., Slidell, LA 70458.
Joyce, A. T., Ph.D. Procedures for Gathering Ground Truth Information for a
Supervised Approach to a Computer-Implemented Land-Cover Classification
of Landsat-Acquired Multispectral Scanner Data. Report No. 163, 1977.
NASA/Earth Resources Laboratory, 1010 Cause Blvd., Slidell, LA 70458.
Kerley, W. G. Documentation for Program ASSIGN, Unpublished Software Documenta-
tion, 1976. NASA, Earth Resources Laboratory, 1010 Cause Blvd., Slidell,
LA 70458.
MacDonald, P. Mathematics and Statistics for Scientists and Engineers, 1966,
pp. 255-256. D. van Nostrand Company, Ltd., 358 Kensington High St.,
London, W14 England.
44
-------
NASA: Earth Resources Technology Satellite Data Users Handbook, Doc. 7154D4249,
1972, NASA Goddard Space Flight Center, Greenbelt, MD 20071.
NASA: Specifications for Airborne Multispectral Scanner, NASA Contract
NAS9-15173, Exhibit "A", 1975, Lyndon B. Johnson Space Center, Houston,
TX 77058.
NASA: Earth Resources Laboratory: ERL-DAS Tape Format Manual, 1975. NASA,
Earth Resources Laboratory, 1010 Cause Blvd., Slidell, LA 70458.
National Academy of Sciences: Rehabilitation Potential of Western Coal
Lands - A Report to the Energy Policy Project of the Ford Foundation,
1974, 198 pp. Ballinger Publishing Co., Cambridge, MA.
Packer, Paul E. Rehabilitation Potentials and Limitations of Surface-Mined
Land in the Northern Great Plains. USDA Forest Service General Technical
Report INT-14, 1974, 44 pp. Intermountain Forest and Range Experiment
Station, Ogden, UT 84401.
Pearson, R. Program LANCOR, Unpublished Software Documentation, 1976. NASA,
Earth Resources Laboratory, 1010 Cause Blvd., Slidell, LA 70458.
Pendleton, T. W. Digital Overlaying of the Universal Transverse Mercator
Grid with Landsat-Data-Derived Products, NASA/ERL Report #160, 1976.
NASA, Earth Resources Laboratory, 1010 Cause Blvd., Slidell, LA 70458.
Spectra-Physics: Geodolite (3A) Laser Distance Measuring Instrument Data
Sheet, 1969, 4 pp. Spectra-Physics, 1250 W. Middlefield Road, Mt. View,
CA 94040.
Whitley, S. W. Low-Cost Data Analysis Systems for Processing Multispectal
Scanner Data. NASA TR R-467, 1976. NASA, Earth Resources Laboratory,
1010 Cause Blvd., Slidell, LA 70458.
45
-------
APPENDIX A
DATA ACQUISITION SYSTEMS
Effective application of data collected by the multispectral scanner or
aerial camera is dependent upon understanding the instruments involved.
Therefore, a brief discussion follows to familiarize the reader with both of
these systems.
AIRBORNE MULTISPECTRAL SCANNER
The airborne multispectral scanner (MSS), Daedalus DS-1260, acquires
data at altitudes ranging from 500 to 20,000 feet (152 to 6,000 meters) above
ground level. This is an 11-band system designed to collect and record
radiant energy in the ultraviolet through the thermal infrared portions of
the electromagnetic spectrum (Table A-l). The scanner has a rotating mirror
that scans across the ground scene, perpendicular to the line of flight
(Figure A-l). Radiant energy from the ground surface is reflected through
focusing optics to a beam splitter which diverts the short wavelength radiation
(0.38 to 1.10 micrometers) to a 10-channel spectrometer and the thermal
infrared radiation (8 to 14 micrometers) to a solid-state detector. Electronic
signals from the 11 detectors are digitized and recorded as high density
digital data on an analog magnetic tape. The MSS scan rate is synchronized to
the aircraft ground speed and altitude, which results in scan-line contiguity
at nadir, thereby avoiding over- or under-scanning the ground scene. The
actual size (ground resolvable distance) of each element is a function of
three factors: (1) aircraft altitude above target level; (2) scanner instan-
taneous field of view (measured in milliradians); and (3) scan angle (that
angle, measured from the perpendicular, at which reflected electromagnetic
radiation is being received as the scanning mirror rotates). The scanner is
equipped with internal visible and thermal reference sources which provide
information for calibration of the data. The aircraft sensor tape is processed
on the ground-based DAS to display and create images of the surveyed scene.
AERIAL CAMERA
The camera used in the data acquisition phase of this study was a Wild-
Heerbrugg RC-8 metric camera. The RC-8 camera is a medium focal length (6-
inch), 9-inch square format sensor. Its primary use by NASA is as a ground
coverage documenting camera with a gyroscopically stabilized mount. The
film/filter combinations and exposure techniques were chosen to obtain
46
-------
SCAN UNE
Figure A-l. Multispectral scanner imaging characteristics
simplified
47
-------
TABLE A-l. MULTISPECTRAL SCANNER WAVELENGTH BANDS
Channel
1
2
3
4
5
6
7
8
9
10
11
Wavelength Band (ym)
0.38-0.42
0.42-0.45
0.45-0.50
0.50-0.55
0.55-0.60
0.60-0.65
0.65-0.70
0.70-0.79
0.80-0.89
0.92-1.10
8.00-14.00
Color/Spectrum
Ultraviolet
Blue
Blue
Green
Green
Red
Red
Near Infrared
Near Infrared
Near Infrared
Thermal Infrared
48
-------
photographic imagery in the desired spectra. Two cameras are available on
board NASA's NP3-A aircraft and are usually selected to provide overlapping
or contiguous simultaneous coverage using color, color-infrared, or black-
and-white film.
The following lists the camera specifications:
Type Square image (9-inch) using a 6-inch F/5.6
Universal Aviogon lens. Unit has no forward-
motion compensation.
Lens Focal lenght, 6-inch nominal, with Water-
house stops; apertures of F/5.6, 6.8, 8,
11, 16, 22, and 32. View angle 74°.
Shutter Continuous variable rotary; speed, from 1/100
to 1/700 second.
Cycle Interval One cycle each 3.5 seconds, maximum rate.
LANDSAT MULTISPECTRAL SCANNER
NASA's Earth Resources Technology Satellite (ERTS) Program presently
consists of two orbiting satellites, names Landsat-1 and Landsat-2, devoted
solely to remote sensing of the Earth's surface. Each Landsat satellite
operates in a circular, sun-synchronous, near-polar orbit at a constant
altitude of 571 statute miles (919 kilometers). The orbital path of each
satellite repreats every 18 days and the orbits of Landsat-1 and Landsat-2
are so synchronized that coverage over any point on Earth is repeated every 9
days.
The primary sensor on each satellite is the four-band MSS which collects
spectral data reflected from the Earth's surface in two visible bands (0.5-
0.6 micrometers and 0.6-0.7 micrometers) and two near-infrared bands (0.7-0.8
micrometers and 0.8-1.1 micrometers). The MSS is a line-scanning device
which uses an oscillating mirror to continuously scan perpendicular to the
direction of flight (Figure A-2). Six lines are scanned simultaneously in
each of the four spectral bands with each mirror sweep. The along-track
progression of the scan lines is provided by the forward motion of the
spacecraft. Reflected energy is sensed simultaneously by an array of six
detectors in each of the four spectral bands. These data are telemetered
directly to ground data acquisition stations or recorded on magnetic tape for
delayed transmittal to the ground. Finally these data are processed at the
NASA Goddard Space Flight Center to produce computer tapes and to convert the
digital data to image formats. Each Landsat scene covers an area 115 x 115
statute miles and each picture element (pixel) in the scene represents about
1.1 acres (2,332 square meters).
49
-------
U1
o
'vPPTICS
6 DETECTORS
PER BAND
(24 TOTAL)
NOTE ACTIVE SCAN IS
WEST TO EAST
ACTIVE
SCAN
DIRECTION
OF FLIGHT
SCAN MIRROR
*\ HELD OF VIEW • 11.56 DEGREES
185 km (100 nm)
6 LINES/SCAN/BAND
Figure A-2. Landsat Multispectral Scanner Imaging System (Simplified)
-------
APPENDIX B
DATA REDUCTION AND PROCESSING PROCEDURES
The flow of data processing within the Data Analysis System (DAS) is
graphically depicted in Figure 3. There are minor differences in processing
aircraft-acquired and Landsat digital data.
PREPROCESSING AND DATA TRANSFORMATION
Conventional digital computers such as the one used in the DAS cannot
read the data format produced by an aircraft multispectral scanner (MSS), so a
special device has been designed to convert aircraft MSS data into the format
expected by a conventional computer. Once this conversion (termed decommuta-
tion) is made, standard data processing techniques can be applied to all other
functions.
The subsequent reformatting process converts the decommutated aircraft data
into a format compatible with the software of the DAS. The digital tape that
is produced is screened to determine if problems are associated with the data,
such as sun angle/scan angle situations, and defective or missing data records.
These problems are handled by software in the preprocessing data transformation
block.
PATTERN RECOGNITION
Data processing and analysis begin when the software programs contained
in the pattern recognition block are implemented. It is within this block
that the data are classified. The four basic steps in pattern recognition are
implemented in accordance with the following guidelines.
Training Sample Selection
The first, and probably the most important step following data acquisition
is the selection of training samples from which statistics are computed and a
signature developed for a given class.
Selection of Training Samples
Aircraft-acquired color, color infrared, or black-and-white photography
are used in the preselection of training fields. Customarily, both the
photography and scanner data are collected simultaneously, but it is the
51
-------
photography that provides the analyst with first-look capability and the
capacity to use photo-interpretation techniques to locate and delineate
features of interest. To optimize data acquisition and at the same time make
it easier for the analyst to interpret, photography and digital data are
obtained simultaneously.
In the selection of training samples, the image analyst attempts to
distribute them evenly throughout the geographic area covered by the scanner.
A relatively even distribution of samples will in most cases:
1. Account for variations in ground-cover conditions in the area imaged
and,
2. Prompt field teams to observe a larger geographic area in which
additional fields may be selected as supplements or replacements of
invalidated fields. In the process of traveling between preselected
training samples, field personnel frequently observe features or
conditions which are unique enough to include them as additonal
training samples. By covering a wide area in the field, the potential
for adding representative samples is immeasurably increased.
Whether identified by photo-interpretation or field observation, every
training sample must be precisely located on the photographs. In fact, when-
ever possible, samples are located so that geographic features can help in
locating them on the photography, in the field, and on the screen of the DAS
TV display. These training samples may be delineated on translucent overlays
and annotated with pre-established coding from the various categories in the
land-cover hierarchy being used. Fiducial marks and distinctive geographic
features such as roadways, pipelines, and water bodies are traced for assistance
in overlay orientation.
In addition to surface conditions such as size, homogeneity, and uniformity,
anomalies in scanner data may invalidate one or more training samples. It
must be recognized that no single sample can be expected to adequately represent
every other area of similar surface material within the scanner field of view.
Field verification of any group of preselected samples may result in as much
as a 20% attrition rate. All samples identified should be plotted and coded
on maps, imagery, and overlays with the same precision and consideration as in
the preselection process.
Computer Location of Training Samples
Following the photo-interpretation, the analyst locates these same
training samples in the digital data. This task is accomplished through
the use of an Interactive Display System (IDS) (Figure 2). This device,
intergrated with the computer and described in detail by Whitley (1976),
premits a visual display of reformatted digital data on a cathode ray tube.
The data can be displayed in either a black-and-white or color mode. In this
manner, polygon-shaped training sample boundaries previously located on the
photography can be located through the use of a movable cursor. By this
method each corner of the sample is located and the corresponding scanline/element
coordinates are recorded by the computer. This process is repeated until all
52
-------
training samples have been located and the analyst is ready to compute statistics
on each training sample or group of training samples (classification category).
Statistical Computation
The next step is the computation of statistics associated with the digital
data contained in each training sample. This step can occur after each indivi-
dual training sample is located, or all samples can be chosen, stored, and the
statistics computed for each simultaneously, whichever is more convenient.
Regardless of the technique followed, the resultant statistics, e.g., mean,
mean vector, covariance matrix, and standard deviation, define the distribution
of training sample data in a multidimensional space (which in turn is defined
by the number of channels being used). These multispectral statistics charac-
terize the sample and are referred to by the term "spectral signature".
First, the statistics are computed for each training sample and the
statistical homogeneity of the sample population must be determined. This
means that the histogram of the data (for all channels) appears to be normally
distributed, with no "outliers" (data points located at some unusually large
distance from the mean) or bimodal tendencies (Figure B-l). Second, the
actual number of data points (elements) in each sample must be of sufficient
number to constitute a statistically "valid" population (generally greater
than 30 elements). By selecting the largest possible training sample (and
hence including the greatest number of elements), statistical tests conducted
at a later time can be made with greater levels of confidence.
Third, the analyst determines whether the coefficient of variation for
any and all channels is sufficiently low. Experience with various data sets
indicates that acceptable values for the coefficient of variation lie in the
vicinity of 10 to 15% for aircraft data. However, this value is by no means
considered absolute; the investigator should be prepared to change the values
of the coefficient of variation in response to the quality and type of data
being processed.
Fourth, the overall quality of the data must be considered. As previously
mentioned, outliers or bimodal tendencies (Figure B-l) would indicate potential
problems with training sample validity. Other problems include (but are not
restricted to) data value skipping observable as skips or gaps in the histogram
at regular intervals (Figure B-2), saturation (data values consistently occurring
at the extreme high limit of the signal for a particular channel), and pure
samples (no variance - all data have the same radiometric value - primarily a
problem associated with water). When all values are identical for a given
material, the statistical equation used in classification are invalid (matrix
cannot be inverted).
If none of these problems mentioned above are observed, the statistics
are stored for later use. If, however, problems are encountered, remedial
action should be taken, such as dropping the sample or relocating the sample
in the digital data and developing a new set of statistics. This latter
action is highly desirable in situations where the particular training sample
in question is the only one representing a particular land-cover type. If
53
-------
14+
13*
CHANNEL 2 HISTOGRAM
O
z
UJ
D
O
UJ
11 +
10 +
9 +
8 +
7*
b +
5 +
4 +
3+
2 +
1 +
-7
*
*
* * *
* * *
* ***
** *** *
****** *
******** *
• + «•<
18
COUNTS
CHANNEL 4 HISTOGRAM
JENCY
_>
O
UJ
u.
13+
12+
11 +
10 +
9+
8 +
7 +
6 +
5 +
4 +
3 +
2 +
1 +
*
*
*
*
*
*****
*****
*****
*******
• + + <
28
COUNTS
53
Figure B-l. Histograms of Data Illustrating Outliers (top)
and Bimodal Tendencies (bottom) of Scanner Data.
-------
u
a
u
a:
u.
14 +
13 +
12 +
11 +
10 +
-------
Channel Selection
The Environmental Monitoring and Support Laboratory-Las Vegas (EMSL-LV)
airborne scanner contains 11 channels. However, software developed for the
DAS restricts the number of input channels to the classifier to a maximum of
four. This is not a severe limitation, as four channels give almost the same
accuracy as a greater number, and computation speed is greatly improved by
limiting the number of channels.
For EMSL-LV purposes, the most desirable technique for reducing the
number of channels to four, utilizes the program SEPARATION. SEPARATION has
been designed to select from an input set of "n" channels, the four channels
that will provide the spectral separation of the training samples used as
inputs. Only one set of four channels will be selected for use in classifica-
tion. The statistics generated for the individual training samples are
utilized in a manner similar to that described by Jones (1976). The statistics
are then recomputed for the four chosen channels for each training sample.
Divergence analysis is conducted and groupings based on only four channels are
made.
Classification of Digital Data
Two software programs are required for classifying the data. The first
program (ELIPSE) computes, from the class statistics, the hyperellipsoidal
decision boundaries used for classifying the digital data. Each hyperellipsoidal
table partitions off a subspace of the n-dimensional space (where n = number
of channels) in which classification of the digital data is to occur. • The
program also prints out other parameters necessary to implement the second of
the two software programs.
Tables constructed by ELIPSE, plus the computer compatible tapes from
which the training samples were selected, are used as inputs to ASSIGN, the
second software program. Each element is located in n-dimensional space and
checked to see if it falls within the hyperellipsoidal subspace of any class.
If so, it is assigned (hence "ASSIGN") to the land-cover category represented
by that particular class. If the element does not fall within any class
subspace, it is assigned to the "uncategorized" group which is a group of
land-cover types for which no training sample sites were chosen. The next
element in the scan line is then considered. This process is continued until
all data have been classified.
The output from ASSIGN Is the classified data on magnetic tape, and line
printer output with statistics indicating the number of elements (and percent
of all elements) in each class (Table B-l). The tape serves two purposes.
First, it can be used to furnish acreage statistics, calculated by an acreage
software program. Second, the classified data tape Is used in viewing the
classified scene on the IPS display or in the production of hard copy products,
e.g., color or black-and-white film recordings (to desired scale), or electro-
static printer/plotter paper products. The IPS viewing is generally done
first to evaluate the classification results. Colors, at the discretion of
the investigator, can be assigned to each class. Colors are generally assigned
In such a manner that various intensities of one color are used for classes
representing varying conditions of a particular land-cover category.
56
-------
TABLE B-l. SUMMARIZATION OF CLASSIFICATION RESULTS BY CLASS (EXAMPLE)
Class
0
1
2
3
4
5
6
7
8
9
10
Pixel Count
6,037
1,830
37,046
19,946
53,302
52,673
3,533
1,101
3,838
1,223
3,640
Percent of Scene
3.28
.99
20.12
10.83
28.94
28.60
1.92
.60
2.08
.66
1.98
Acres
3,409
1,033
20,918
11,262
30,096
2-9,741
1,995
622
2,167
691
2,055
Square Miles
5.33
1.61
32.68
17.60
47.03
46.47
3.12
.97
3.39
1.08
3.21
TOTAL 184,169 100.0 103,989 162.50
57
-------
APPENDIX C
SCIENTIFIC NAMES OF VEGETATIVE TYPES IDENTIFIED IN PHASE I OPERATIONS
Common Name
Genus and Species
Alkali sacatone
Blue grama
Greasewood
Green needlegrass
Juniper
Mormon tea
Needle-and-threadgras s es
Pinyon pine
Rabbit brush
Sagebrush
Snakeweed
Wheatgrass
Winterfat
Sporobolus airvoides
Bouteloua gracilis
Sarcobatus vermiculatus
Stipa viridula
Juniperus sp.
Ephedra sp.
Stipa comata
Pinus edulis
Shrysothamnus sp.
Artemisia sp.
Gutierrezia sp.
Agropyron sp.
Eurotia lanata
58
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/7-78-149
2.
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
REMOTE MONITORING OF COAL STRIP MINE
REHABILITATION
5. REPORT DATE
July 1978
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
James E. Anderson
Charles E. Tanner
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Lockheed Electronics Company, Inc.
Remote Sensing Laboratory
Las Vegas, Nevada 89114
10. PROGRAM ELEMENT NO.
INE 625
11. CONTRACT/GRANT NO.
68-03-2636
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency-Las Vegas, NV
Office of Research and Development
Environmental Monitoring and Support Laboratory
Las Vegas, Nevada 89114
13. TYPE OF REPORT AND PERIOD COVERED
Final (7-1-75 / 12-31-76)
14. SPONSORING AGENCY CODE
EPA/600/07
15. SUPPLEMENTARY NOTES
G.J. D'Alessio, Project Officer, Western Energy/Environmental Monitoring
Study, U.S. Environmental Protection Agency, Washington, B.C. 20460
16. ABSTRACT
This report discusses the accomplishments of the Phase I Operations of the
EPA/NASA joint project and also compares the results of manual photo-interpretation
and automated data analysis conducted during this phase. Also included in this
report are the results of a feasibility study to utilize Landsat data for
performing a regional land-cover classification of a portion of the Powder
River Basin area in northeastern Wyoming, where there are numerous coal strip
mines.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Aerial Photography
Land Use
Photographic Reconnaissance
Photointerpretation
Spaceborne Photography
S t ereophotography
Automatic Data Processing
Multispectral Scanner
Ground Truth
Field Observations
09 B, F
08 I
14 E
20 F
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF F
72
kGES
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (R«v. 4-77) PREVIOUS EDITION is OBSOLETE
U.S. GOVERNMENT PRINTING OFFICE: 1978—786467
------- |