Author(s)
Marino, Stephen A.
Title
Operation and calibration of the NPS Ultraviolet Imaging Spectrometer (NUVIS) in the
detection of sulfur dioxide plumes
Publisher
Monterey, California. Naval Postgraduate School
Issue Date
1999-12-01
URL
http://hdl.handle.net/10945/8773
This document was downloaded on May 04, 2015 at 23:43:18
MONTEREY CA
93943-5101
NAVAL POSTGRADUATE SCHOOL
Monterey, California
THESIS
OPERATION AND CALIBRATION OF THE NPS
ULTRAVIOLET IMAGING SPECTROMETER (NUVIS)
LN THE DETECTION OF SULFUR DIOXIDE PLUMES
by
Stephen A. Marino
December 1999
Thesis Advisers:
David
S.
Davis
Richard C. Olsen
Approved
for public release; distribution
is
unlimited.
REPORT DOCUMENTATION PAGE
Public reporting burden for this collection of information
is
estimated to average
1
Form Approved
OMB No.
0704-0188
hour per response, including the time for reviewing instruction, searching
existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information.
Send comments regarding
this
burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington headquarters Services,
Directorate for Information Operations and Reports. 1215 Jefferson Davis Highway, Suite 1204. Arlington, VA 22202-4302, and to the Office of Management
and Budget Paperwork Reduction Project (0704-0188) Washington DC 20503.
1.
AGENCY USE ONLY (Leave blank)
2.
REPORT DATE
3.
December 1999
REPORT TYPE AND DATES COVERED
Master's Thesis
TITLE AND SUBTITLE OPERATION AND CALIBRATION OF THENPS
ULTRA VIOLET IMAGING SPECTROMETER (NUVIS) IN THE DETECTION OF
SULFUR DIOXIDE PLUMES
4.
6.
5.
FUNDING NUMBERS
AUTHOR(S)
Marino, Stephen A.
7.
PERFORMING
ORGANIZATION REPORT
8.
PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)
Naval Postgraduate School
Monterey, CA 93943-5000
9.
NUMBER
SPONSORING / MONITORING AGENCY NAME(S) AND ADDRESS(ES)
10.
AGENCY REPORT NUMBER
N/A
11.
SPONSORING / MONITORING
SUPPLEMENTARY NOTES
The views expressed
in this thesis are
those of the author and do not reflect the official policy or position of the Department of
Defense or the U.S. Government.
12a.
DISTRIBUTION / AVAILABILITY STATEMENT
Approved
13.
for public release; distribution
is
12b.
DISTRIBUTION CODE: A
unlimited.
ABSTRACT (maximum 200 words)
The Naval Postgraduate School's Ultraviolet Imaging Spectrometer (NUVIS) is a hyperspectral sensor with a spectral
response of 300 to 375 nanometers. This thesis research concentrates on the operation and calibration of NUVIS in the
detection of effluent sulfur dioxide (S0 2 ) plumes. NUVIS is capable of detecting and quantifying S0 2 emissions in the
form of effluent smokestack plumes by exploiting S0 2 's unique UV absorption signature. Laboratory comparison UV
spectra of S0 2 were recorded and used to calculate curves of growth for four different S0 2 spectral features. Laboratory
results were employed to analyze field data taken of a coal-burning power plant. Analysis of this plume data yielded a
mean plume S0 2 mixing
assessment of
NUVIS
365 ± 200 ppm,
ratio of
indicates that
its
in
lower limit for
agreement with the
S0 2
in situ stack
value of 400 ppm.
detection in typical field applications
is
Further
approximately 70
ppm.
14.
SUBJECT TERMS
Ultraviolet, Hyperspectral, Spectral Imaging, Spectrometer, Sulfur Dioxide, Pollution,
Remote
15.
Sensing, Environmental Monitoring
17.
SECURITY CLASSIFICATION OF
REPORT
Unclassified
NUMBER OF
PAGES
104
SECURITY CLASSIFICATION
16.
PRICE CODE
LIMITATION OF
19. SECURITY
CLASSIFICATION OF
20.
OF THIS PAGE
Unclassified
ABSTRACT
UL
18.
ABSTRACT
Unclassified
NSN
7540-01-280-5500
Standard Form 298 (Rev. 2-89)
ANSI Std. 239-18
Prescribed by
11
DUDLEY KNOX LIBRARY
N ^VAL POSTGRADUATE
SCHOOL
B
MONTEREY, CA 93943-5101
Approved
for public release; distribution
is
unlimited
OPERATION AND CALIBRATION OF THE NPS ULTRAVIOLET IMAGING
SPECTROMETER (NUVIS) IN THE DETECTION OF SULFUR DIOXIDE
PLUMES
Stephen A. Marino
Lieutenant, United States
B.S., University
Submitted
Navy
of Pittsburgh, 1991
in partial fulfillment
of the
requirements for the degree of
MASTER OF SCIENCE IN APPLIED PHYSICS
from the
NAVAL POSTGRADUATE SCHOOL
December 1999
DUDLFY KNOX LIBRARY
POSTGRADUATE SCHOOL
MONTEREY CA 93943-5101
ABSTRACT
The Naval Postgraduate School's
Ultraviolet
Imaging Spectrometer (NUVIS)
a hyperspectral sensor with a spectral response of 300 to 375 nanometers.
thesis research concentrates
on the operation and calibration of
detection of effluent sulfur dioxide (SO2) plumes.
and quantifying SO2 emissions
exploiting S02's unique
spectra of
different
UV
in the
absorption signature.
spectral features.
value of 400 ppm.
SO2
ratio
in the
capable of detecting
Laboratory comparison
to calculate curves
UV
of growth for four
Laboratory results were employed to analyze field
data taken of a coal-burning power plant.
mean plume SO2 mixing
is
This
form of effluent smokestack plumes by
SO2 were recorded and used
SO2
NUVIS
NUVIS
is
Analysis of this plume data yielded a
of 365 ± 200 ppm, in agreement with the
Further assessment of
detection in typical field applications
NUVIS
is
indicates that
its
approximately 70 ppm.
in situ stack
lower limit for
VI
TABLE OF CONTENTS
INTRODUCTION
I.
MOTIVATION AND OBJECTIVES
OUTLINE
A.
B.
II.
1
1
4
BACKGROUND
REMOTE SENSING AND SPECTRAL IMAGING
A.
ULTRAVIOLET REMOTE SENSING
B.
1.
2.
C.
1.
2.
D.
III.
The Concept
Earth's Atmosphere and
7
7
12
12
Ultraviolet
Wavelengths
13
SULFUR DIOXIDE
15
An Atmospheric Pollutant
An Ultraviolet Absorber
ABSORPTION SPECTROSCOPY
15
18
20
1.
Introduction
20
2.
Terminology
3.
The Curve of Growth
22
26
NUVIS
29
A.
BACKGROUND
29
B.
OPTICAL, OPTOMECHANICAL AND ELECTRONIC HARDWARE
THE COMPUTER, DATA ACQUISITION AND CONTROL SYSTEMS
29
C.
D.
DATA ACQUISITION, CONTROL AND DATA DISPLAY SOFTWARE
35
IV.
LABORATORY CALIBRATIONS OF NUVIS
32
41
A.
PURPOSE
41
B.
WAVELENGTH CALIBRATION
EXPERIMENTAL OVERVIEW
41
C.
43
LABORATORY MEASUREMENTS OF S0
D.
1.
Gas Test
45
2.
Calibration Source and Reflective Screen
49
3.
Test Cell Preparation
51
4.
Data Acqusition
52
5.
First Results
.
GENERATING EXPERIMENTAL CURVES OF GROWTH
E.
53
55
1.
Approach
55
2.
Experimental Results
56
3.
Analysis of Experimental Data
59
APPLICATION OF LABORATORY CALIBRATIONS TO FIELD DATA
V.
44
2
Cells
63
A.
FIELD EXPERIMENTS
63
B.
RESULTS
67
VI.
CONCLUSIONS AND RECOMMENDATIONS
vii
75
APPENDIX A. NUVIS FIELD CHECKLIST
79
APPENDIX
B.
NUVIS OPERATIONAL CHECKLIST
81
APPENDIX
C:
RANGE EFFECTS AND NUVIS ANOMALIES
87
LIST OF
REFERENCES
91
INITIAL DISTRIBUTION LIST
93
Vlll
ACKNOWLEDGEMENTS
The author would
motivator in
my
Stephen Finney,
analysis.
thank Dr. David Geary whose enthusiasm was a major
decision to begin this
USAF,
Thanks
to
during this research.
love, devotion
like to
work with NUVIS.
Special thanks to Major
for assisting in the 1999 field experiments
NPS
staff
and subsequent data
members Jay Adeff and Don Snyder
Finally, the author
would
and patience.
IX
like to
for their support
thank his wife, Michelle, for her
INTRODUCTION
I.
MOTIVATION AND OBJECTIVES
A.
Remote sensing instruments
are critical to the areas of technical intelligence, the
remote sensing of hazards, and support to military operations. In an
capabilities available to the
pursued a
we have
came, in
Analysis
remote sensing community as well as
used portion of the electromagnetic spectrum.
little
part,
Over
wavelengths.
we have
the last three years,
Our motivation
from the success of the school's Middle Ultraviolet SpecTrograph
of Nitrogen
Spectrometer
enhance the
to the military,
designed and constructed an ultraviolet hyperspectral imager.
instrument.
This
Gases
resulted
(NUVIS),
NUVIS
is
a
in.
(MUSTANG),
the
rocket-borne
research
UltraViolet
Imaging
NPS
imager
operates
first
that
for
atmospheric
development of the
hyperspectral
one of the
a
at
near-ultraviolet
such instruments of its kind: a portable, ground-
based, hyperspectral imager that operates at ultraviolet
to
effort to
(UV) wavelengths: nominally 300
375 nanometers (nm).
From
the onset,
NUVIS
designers recognized that ultraviolet remote sensing
could be used as an important tool in effluent analysis. Through the efforts of Dr. David
Cleary,
LT Todd Hooks and Major Andrew MacMannis, NUVIS was
school, and fabrication of the instrument
was
to focus research
effluent plume.
communities.
was completed
on the instrument's
Such
ability to
in the fall
designed
at the
of 1997. The next step
analyze and characterize a chemical
capabilities are of interest to both the civilian
and military
As
testing
initial
was about
to get
underway, a major question arose: Could
gaseous effluents emitted from smoke stacks be detected, imaged, identified, and
quantified using hyperspectral remote sensing techniques at
thesis research set out to
answer
that question
capability to perform such a mission.
dioxide
(S0 2 ), was
Numerous
using
and
UV
wavelengths?
This
to further define the instrument's
In support of these efforts, the trace gas, sulfur
targeted for study.
studies have focused
spectroscopic
monitoring
on the topic of remote sensing of trace gases
Supporting
techniques.
literature
indicates
that
considerable attention has been placed on the study and techniques of infrared remote
sensing. Until
little
attention.
now, remote sensing
at ultraviolet
wavelengths has received compartively
This study sets out to challenge such thinking by exploring
new ways
to
exploit the ultraviolet region of the electromagnetic spectrum utilizing the recently
developed technique of hyperspectral imaging.
Although
ultraviolet
ozone absorption than do
wavelengths suffer more from atmospheric scattering and
others,
study
this
will
show
that
demonstrated dual-use imaging and spectroscopic capabilities and
at short range, the effluent trace
determine whether effluent
longer distances using
UV
S0
2
gas
S0
2
.
With
this initial success,
NUVIS
is
it
has
already
capable of detecting,
became necessary
to
could reliably be identified, measured and quantized over
remote sensing.
An
approach was taken such that laboratory
experiments, calibrations, and operational field testing would lead to an evaluation of
NUVTSTs
capabilities to detect the trace gas
and the
utility
of such technology.
This thesis project focuses on calibrating the spectrometer for the detection of
gaseous sulfur dioxide. Laboratory experiments were conducted
of S0 2 to measure changes
SO2
is
in its absorption
at
known
concentrations
spectrum as a function of gas concentration.
a triatomic molecule that has a set of strong absorption signatures between 300
and 325 nm.
The
number of
detailed spectrum across this spectral region exhibits a
For each observed
identifiable absorption lines.
spectral absorption, called
its
equivalent width,
line,
may be
an associated measure of
easily measured.
its
Calibration
curves of the lines' equivalent widths versus gas concentration, called curves of growth,
can then be generated.
During
analysis,
four
this research,
six sets
of calibration data were analyzed.
absorption
individual
lines
were
identified
for
From
equivalent
this
width
measurements. Four curves of growth were successfully generated from these laboratory
experiments.
These calibration curves were then applied to
attempt to quantify effluent
It
all
S0 2
should be noted that
of the instrument's
initial
the process of leaving.
field data in
an
concentrations.
my
involvement with
designers had either
Because of
this,
NUVIS came
moved on from
at a
time when nearly
the school or were in
a period existed where Dr. Cleary had done a
considerable amount of work with the instrument himself, and the majority of this
had gone unpublished.
initial
work
Because Hooks and MacMannis were unable to complete
laboratory and field testing of the instrument, a detailed operating procedure had not yet
been developed. This
thesis will not only present the
work involved with generating
the
S0 2
curves of growth but will also outline the basic operating procedures for the
Such documentation had not existed
instrument.
prior to this study.
an attempt to maintain continuity with the work done by Hooks and
In
MacMannis,
this thesis also includes certain portions
as background material. This
is
due to the
in part
undergone certain improvements from
inclusion of such material
In
1998,
is
its
of Dr. Geary's unpublished work
original configuration.
experiments
were
preliminary spectroscopy and spectral imagery of effluent
accompanying
data analysis have
resultant
detection,
analysis
is
appropriate, the
all
conducted
and resulted
S0 2 -producing
targets.
calibration, a follow-on field experiment,
been accomplished
in
and characterization with respect
involved in this study
Where
system has
clearly indicated.
ground-based field
thesis research, the
NUV1S
fact that the current
a follow-on to the
initial
to
This
and the
support of chemical plume
The work
sulfur dioxide.
experimental research initiated in 1998.
OUTLINE
B.
This thesis
is
divided into six chapters and three appendices.
the necessary background material
on
Chapter
the topics of remote sensing
II
provides
and hyperspectral
imaging, ultraviolet remote sensing, sulfur dioxide and absorption spectroscopy.
NUVIS
instrument will be presented in Chapter
calibration
S0 2
in
III.
The
Chapter IV details the laboratory
of NUVIS and includes resultant curves of growth, one for each of the four
absorption lines selected for analysis.
Field data
is
briefly analyzed
application of these laboratory calibrations developed in Chapter IV.
by a simple
This process and
the
accompanying
results are presented in
are contained in Chapter VI.
field operation.
Appendix
B
Appendix
Chapter V. Conclusions and recommendations
A provides
a
recommended
checklist for
briefly details the system operating procedure.
provides additional data that should be considered in future reasearch.
NUVIS
Appendix C
THIS PAGE INTENTIONALLY LEFT
BLANK
BACKGROUND
II.
REMOTE SENSING AND SPECTRAL IMAGING
A.
Remote sensing
at a
a generic term used to describe
is
distance from an object to be observed.
is
carried
If the source
If the source
remote sensing (Lenoble,
is
a
man-made
lists
some
emitter, then the process
of these waves
1993).
It
wavelengths (200-400nm) with which
Table 2.1
Typically the information being
by electromagnetic waves, although other means are sometimes used.
of these waves
remote sensing.
of the measurements made
involves collecting information about an
It
object without being in direct contact with that object.
collected
all
is
is
NUVIS
is
sensing
dubbed active
known
natural, the process is
passive remote
is
at
as passive
near-ultraviolet
most concerned. As a quick reference,
typical wavelength, frequency
and wavenumber ranges for various
regions of the electromagnetic spectrum.
Name
Wavelengths
Frequencies
(nm)
(Hz)
8
13
Radio
~10 -10
Microwave
~10 7 -10 8
Infrared
~10 3 -10 7
700-400
Visible
4
~3x 10 -3x 10
-3xl0 9 -3x 10 10
~3x 10 10 -3x 10 14
4
-400-50
-7.5 x 10' -6.0x 10
XRay
Gamma Ray
-50-0.1
-(0.6-300) x 10
<0.1
To3x
2. 1.
10
1
)
10' -0.1
0.1-1
1-10
14
Ultraviolet
Table
(cm"
6
9
-(4.3-7.1) x 10
Wavenumbers
4
(1.4-2.4)
15
16
18
x 10
(2.5-20) x 10
4
4
(0.2-100) x 10
>10
6
8
Typical wavelengths, frequencies and wavenumbers for various regions of the
electromagnetic spectrum.
Over the
last
decade, a
number of instrument platforms have been developed
the express purpose of optical remote sensing.
number of
field operations, ranging
based applications.
quickly
These platforms have been employed
from space-based and air-based missions
to
for
in a
ground-
Spectral imagers are one such instrument family, and they have
become workhorses of the remote sensing community.
Spectral imagers measure
both the 2-D image produced by a scene and the spectral content of each picture element
(pixel)
of the image.
These spectral imagers are
bands they record. Figure
2.1 provides a brief
The motivation behind
classified
by the width of the spectral
overview of such
the development of
classifications.
NUVIS was
a perceived need to
explore the near-UV wavelength region with an operational hyperspectral imager.
Such
an instrument would be capable of collecting data from tens to hundreds of bands
much
at
a
smaller bandwidth resolution than that of a conventional multispectral sensor.
WAVELENGTH (ran)
400
700
nil
MULTISPECTRAL
4000
2000
1000
6000
14000
aitifgii
Spectral identification of major features,
Wide Bandwidth
ije.,
trees, grass, roads
Moderate Spectral Resolution
HYPERSPECTRAL
Spectral discrimination by species
Narrow Bandwidth
High Spectral Resolution
materials, and environmental conditions
ULTRASPECRAL
Very Narrow Bandwidth
Very High Spectral Resolution
Figure
2. 1.
Spectral imager classifications.
Identification
and discrimination of subtle
spectral details of materials, vapors and aerosols
From
Multispectral Users Guide, 1995.
NUVIS
Related research that began in 1996 resulted in the
hyperspectral imager with a spectral bandwidth of 300 to 375
instrument of today, a
nm
at
a resolution of
approximately 0.3 nm. (Geary, 1998)
A
hyperspectral imager produces a spectral image of a scene, which
displayed in a format
known
use a two-dimensional array
a
CCD) and
typically
Imaging spectrometers typically
as a hyperspectral cube.
(e.g.,
is
produce the three-dimensional data cube
containing two spatial dimensions and a third spectral dimension. The third dimension
is
obtained by scanning either the spectral dimension or one of the two spatial dimensions.
NUVIS
is
the latter type.
illustrated in Figure 2.2.
A
sample of NUVIS output
A
large
number of images
demonstrated in the image cube
is
are created
from contiguous, rather
than disjoint regions of the spectrum. This technique typically yields finer resolution than
is
possible with a multispectral imager and in doing so, provides
to the data analyst.
energy
at a specific
It is
from
this series
much more
of images, each measuring the level of incident
wavelength, that a combined image
is
formed.
Today, countless airborne hyperspectral sensors are employed
2.2 provides a
summary of many of these
relative non-existence of
UV
information
instruments.
hyperspectral imagers.
in the field.
Table
This table also demonstrates the
Thus,
NUVIS
based, passive remote sensor capable of hyperspectral imaging at
is
a portable, ground
UV
wavelengths; a
prototype instrument that could someday possibly be employed as an airborne sensor.
Figure
2. 2.
NUVIS
hyperspcctral cube.
Image
illustrates the
two
dimension comprising the image cube.
10
spatial
and one spectral
Sensor
(Agency)
AAS (ASTER)
Number
Spectral
Bandwidth
IFOV
FOV
of Bands
Range(nm)
FWHM(nm)
(mrad)
(deg)
760-850
1
20
3000-5000
8000-12000
AHS.Daedelus
48
AIS-1
128
(DAIS-2815)(GER)
3
90
600-700
1.0,2.5,
or 5.0
Data
•
Period of
Operation
28.8,65
Image
Since(s)
,
Cube
1991
(IC)
440-12700
200
20-1500
2.5
86
IC
1994
990-2100
9.3
1.91
3.7
IC
1982-85
10.6
2.05
7.3
IC
1986-87
or 104
1200-2400
128
AIS-2
800-1600
1200-2400
(NASA/JPL)
1-286
450-900
AMSS
32
(Geoscan)
8
A1SA
1.56-9.36
1
21
IC
SI 993
490-1090
170-240
2.1x3.0
92
IC
SI 985
2020-8500-
430-440
12K
550-590
(KarelsilvaOy)
6
Sensor
Number
Spectral
Bandwidth
IFOV
FOV
(Agency)
of Bands
Range(nm)
atFWHM
(mrad)
(deg)
Data
Period of
Operation
(nm)
ARES
ASAS
(AIP)
Upgraded
ASAS
(NASA/GSFC)
AVIRIS (JPL)
CASI
75
2000-6300
25-70
1.17
3x3
IC
29
455-873
15
0.80
25
IC
62
400-1060
11.5
0.80
25
224
380-2500
9.7-12.0
1
30
IC
288
400-1000
650
1.3,1.6
37.8
Profile
44.7
Image
80deg
IC
SI 993
1992
19
nominal
CIS
64
400-1040
10
1.2x3.6
(China)
24
2000-2480
3530-3940
20
410
1000
1.2x1.8
(Itres
Research)
1
S1985
1987-91
S1991
S1987
SI 990
1.2x1.2
2
10.5K-12.5K
CHRISS,SAIC
40
430-860
11
0.05
10
IC
AHIS,(SAIC)
DAJS-7915
288
440-880
3
1.0
11.5
IC
1994
32
10-16
3.3,2.5,
64-78
IC
SI 994
8
400-1010
1500-1780
36
or5.0
32
1970-2450
3
78
IC
SI 994
5
±45
IC
SI 994
1.3
70
Profile
1984-90
(GER/DLR)
3000-5000
36
2000
6
8.7K-127K
600
1
DAJS-16115
(GER)
DA1S-3715
(GER)
76
400-1000
8
32
32
1000-1800
25
2000-2500
16
6
3000-5000
333
12
8K-12K
333
2
400-1000
stereo
32
360-1000
20
1000
1
1
2175-2350
3000-5000
1
8K-12K
2000
4000
430-805
2.5
2
>288
FLI/PMI
1000-2000
1.2x1.2
50
Image
(Moniteq)
FTVHSI
256
440-1150
67cm-'
0.8
15
IC
1996
90
IC
SI 986
(Kestrel)
24
400-1000
25
2.5,3.3,
Scanner
4
1500-2000
125
4.5
(GER)
29
2000-2500
17.2
6
8K-12.5K
750
GER-63
Channel
Continued on next page.
11
or
Sensor
(Agency)
Number
Spectral
BandwidthF
IFOV
FOV
of Bands
Range(nm)
YVHM(nm)
(mrad)
(deg)
HYDICE
206
400-2500
(NRL/ER1M)
ISM
(DESPA/IAS/OPS)
64
64
MAS
Data
Period of
Operation
7.6-14.9
0.5
8.94
IC
1994
800-1600
12.5
3.3x11.7
40
IC
S1991
1600-3200
25.0
50
547-14.5K.
31-517
2.5
85.92
IC
SI 992
MISI(RIT)
60
400-1000
10
2
±45
IC
S1996
MIV1S
20
433-833
20
2.0
70
IC
1993
(Daedalus)
8
1150-1550
50
64
2000-2500
10
8
90
90
2500-7000
25-70
0.5
1.3
IC
1989
6K-14.5K.
60-1400
ROSIS (MBB/
DLR/GKSS)
SFSI (CCRS)
SMIFTS
84
430-830
4-12
16
IC
SI 993
115
1200-2400
10.4
75
1000-5200
lOOcm"
(U. of Hawaii)
35
3200-5200
50cnr'
128
430-850
460-880
(Daedalus)
MUSIC
(Lockheed)
8
2K-12 7K
400-500
0.56
(<2400)
TRWIS-A
TRWIS-B
90
80
TRWIS-II
TRWIS-III
384
(32)
0.4
9.4
IC
SI 994
06
6.0
IC
SI 993
3.3
1.0
15
IC
1990
4.8
1.0
15
IC
1991
1500-2500
12
0.5/1.0
7.5/15
IC
1992
300-2500
5/6.25
0.9
15
IC
1995
1
(TRW)
IC
Hybrid VIFIS
(U. of
Dundee)
WIS-FDU
WIS-VNIR
WIS-SWIR
(Hughes SRBC)
Table
30
30
440-640
10-14
1
31.5
620-890
14-18
1
31.5
64
400-1030
1.36
129.265
400-1000
0.66
81+90
1000-2500
0.66
2. 2.
Summary of hyperspectral
IC
Test 1994
10&15
IC
Test 1992
19.1
IC
1995
12.0
IC
1995
airborne imaging spectrometers.
From Kramer, 1996.
B.
ULTRAVIOLET REMOTE SENSING
The Concept
1.
Passive remote sensing at ultraviolet wavelengths
spectral distribution of the solar irradiance at
in Figure 2.3.
ultraviolet
From
this
wavelengths
wavelengths.
diagram,
is
NUVIS was
it
relatively
is
ground
level.
is
quite dependent
This distribution
on the
is illustrated
easy to see that the solar irradiance available at
low compared
to
that
at
visible
and infrared
designed to operate in the daytime, using ambient scattered
UV radiation in the 300 to 400nm spectral band.
12
25
Energy curve
2
for
blackbody
at
6000° K
Solar irradiance curve outside atmosphere
Solar irradiance curve at sea level for a zenith angle of 0°
15
2
1.0
0.5
_UV/
200 400
600
2000
1600
800 1000
2600
3000
Wavelength (nm)
Figure
Spectral distribution of the solar irradiance (flux). Notice that the magnitude
2. 3.
of the available flux
absorption,
with
its
which
off rapidly
below 400nm. This decline
From
will be discussed later.
is
due, in part, to ozone
Finlayson-Pitts
& Pitts,
earth's atmosphere can be divided into a
own
approximate
unique characteristics.
behavior
number of atmospheric
layers each
Figure 2.4 illustrates these layers and provides the
of temperature
characteristic of the earth's
atmosphere
over
is
nonreactive major component gases (N 2
,
that
2,
these
it is
regions.
generally a
One
2.3.
13
distinguishing
homogeneous mixture of
noble gases) superimposed upon variable
concentrations of minor volatile and/or chemically reactive gases,
Table
1986.
Earth's Atmosphere and Ultraviolet Wavelengths
2.
The
falls
as
indicated
in
130
at
110
4>
u
VI
IB
o
"
E2-
_
*,
—
The rrri o s phe re
zL
90
" o
d
l.
o
*"*
.
-
-
S
-
-
A
Mesosph«r«
^"~^^^
>
V
<u
"
*i
—
50
-
Stratosphere
jS
30
e
Figure
Name
/
2. 4.
"
Oxygen / (Permanent)
Argon / (Permanent)
(Variable)
Carbon Dioxide
/
(Variable)
Z
layers.
(K)
From Lenoble,
T993.
% by volume
Formula
Molec. Mass
N
o
2
28.0134
78.084
2
31.9988
20.9476
Ar
39.948
0.934
H,0
18.0160
0-7
C0
44.00995
0.01-0.1
2
Ozone / (Variable)
Neon / (Permanent)
Helium / (Per. .anent)
Methane / (Permanent)
o,
47.9982
0-0.01
Ne
He
20.183
0.001818
4.0026
0.000524
16.04303
0.0002
Sulfur Dioxide
SO,
64.064
0-0.0001
Hj
2.01594
0.00005
46.0055
0-0 000002
Hydrogen
/
(Permanent)
Nitrogen Dioxide
Table
Since
CH
(Variable)
/
/
2. 3.
NUVIS
4
N0
(Variable)
-
300
Atmospheric
(type)
-
,r
250
Nitrogen/ (Permanent)
"
a;
Troposphere
Temperature
/
o
E
<a
3
^—_^_
_
200
Water Vapor
a.
o
E
\Tropopause
10
Constituent
o.
:
Atmospheric constituents. From Durkee, 1999.
currently operates at ground level, the properties of ultraviolet
radiation in the lower troposphere are of paramount interest.
radiation can reach the troposphere,
it
must
14
first
Before any solar ultraviolet
pass through the various overlying
atmospheric
layers.
Far ultraviolet wavelengths are strongly absorbed at higher altitudes
(above 100km) by molecular and atomic oxygen and nitrogen. More importantly, a great
majority of
UV absorption is caused by oxygen in the mesosphere and upper stratosphere,
principally in the form of ozone.
NUVIS
(~300nm). Figure
2.5
These absorbers are active
summarizes these
ii
in the spectral
bandpass of
effects.
—
1
l
i
i
150
E
o
100
-
Z>
»-
50
<
i
i
50
100
150
WAVELENGTH
Figure
2. 5.
Depth of penetration of solar
attenuation of Me.
C.
II
i
i
1
250
200
300
(nm)
radiation.
Altitudes correspond to an
From Lenoble,
1993.
SULFUR DIOXIDE
1.
An Atmospheric
Pollutant
Absorption of ultraviolet radiation by atmospheric gases
cases of nitrogen, oxygen and ozone alone.
atmosphere that absorb
at
these shorter
not limited to the
There are a number of trace gases
(UV) wavelengths.
15
is
in the
Nitrogen dioxide (NO2)
and sulfur dioxide (S0 2 ) are two such examples. For the purpose of
work, only the
this
trace gas, sulfur dioxide, will be studied.
Sulfur
dioxide,
suffocating odor.
in the
It is
Figure
2.6,
extremely
is
a
colorless
fuels
major atmospheric pollutant. This trace gas
cycle illustrated in Figure 2.7.
pollutant
emitted by
fuel
and
is
in the smelting
characteristic
tract.
It is
formed
of sulfide ores and
only a small part of the
This cycle not only demonstrates that
much
S0
2
is
is
emitted as
A^
*
larger sulfur
is
a primary
2. 6.
S0 2
(Finlayson-Pitts and Pitts, 1986).
Sulfur (S)
Oxygen
(O)
Sulfur dioxide molecule
(SO : )
-Gother
PMs
]
~ C OS
J
^-
f
(
(
H 2 S }—I—f SQ 2 }— HSQ3 }— [ SQ 3
[
)
[n 2 SQ4 ]
(
JOlSj
t *
biogenic
Figure
volcanic
2. 7.
a
combustion (from coal burning power plants) but also
demostrates that most volcanic sulfur
Figure
which has a
eyes and respiratory
irritating to the
combustion of sulfur-containing
gas,
anthropogenic
Sulfur cycle demonstrating volcanic and
16
man-made sources of S0 2
.
Monitoring of sulfur dioxide has become a global concern as data are routinely
from a number of
collected
Monitoring
activity,
S0
SO
2
:
is
different
sensors
to
monitor
of interest for a number of reasons.
see Figure 2.8
ground
worldwide.
For years now, the Correlating Spectrometer (COSPEC), a non-
).
Such monitoring
is
to
monitor effluent
warn of a
rise in
2. 8.
COSPEC
instrument
at
17
2
SO, concentrations
level.
Figure
S0
important, especially in densely populated
regions, to aid in predicting volcanic activity and to
at
levels
In regions of high volcanic
imaging spectrometer, has been employed by vulcanologists
(
:
monitoring can provide yet another valuable tool to assist in the study of
volcanic activity.
levels
SO
a volcanic
site.
From USGS,
1999.
The Environmental Protection Agency (EPA) has long been
A
monitoring anthropogenic (man-made) sources of sulfur dioxide.
been placed on coal burning power
plants.
The
EPA
has established
interested
primary focus has
strict
guidelines for
allowable types and levels of emissions that these plants can produce.
effluent levels
the
power
from these sources can be accomplished
plant's
either
by sensors
stacks themselves (in situ measurements) or
in
Monitoring
installed within
by remote sensing
techniques.
With these
factors in mind,
NUVIS
designers set out to build an instrument
capable of detecting SO, emissions from both natural and
S0
Since typical
power plants
to
2
ppm
An
S0
2
effluent sources.
in the case
of regulated
NUVIS would
hundreds or even thousands of ppm for volcanic vents,
required to detect
2.
concentrations range from only tens of
man-made
be
over a very wide range of absorption column densities.
Ultraviolet
Absorber
Sulfur dioxide was chosen for study because the gas exhibits a strong absorption
signature in the near-ultraviolet (200-400nm).
Although
S0
2
absorbs to some extent over the entire 240 to
the region longward of
atmosphere.
This signature
300nm
This region
is
is
(see Figure 2.3).
A
Figure 2.9.
320nm wavelength
of most interest for measuring
S0
2
range,
concentrations in the
advantageous because the illumination source for doing
absorption spectroscopy in the field
level, this scattered radiation
is illustrated in
is
Rayleigh scattered solar
diminishes sharply
further advantage
is
at
UV
At ground
wavelengths shorter than about
that if in this region,
18
radiation.
S0
2
300nm
spectra can be
readily distinguished
from the spectra of other absorbers (see Figure
chose to concentrate on laboratory and
350
nm
wavelength range.
—
200
field
measurements of
A
Figure
2. 9.
S0 2
2.10).
Hence,
we
spectra in the 300 to
/nm
Gas phase absorption spectrum of S0 2
.
The
extinction coefficient sis
proportional to the absorptivity coefficient presented in section ID. 2.
From
Atkins, 1990.
PHOTOABSORPTION CROSS SECTIONS
T
Yoshino + Freemon O.
1
I
95K
SOLSTICE (scoied)
lO"
,-18
10"
lO"
lO"
O (Amoruso
et ol)
_i
i
i_
_i
240
Figure
2. 10.
i
i_
_i
260
i
L
j
280
Wavelength (nm)
»
30O
Photoabsorption cross sections of atmospheric
cross section data obtained from the Solar Stellar Irradiance
(SOLSTICE)
is
also displayed.
19
From
320
UV absorbers.
340
Solar
UV
Comparison Experiment
Strickland, 1999.
ABSORPTION SPECTROSCOPY
D.
Introduction
1.
Absorption spectroscopy
radiation from a continuum
is
the study of the wavelength-dependent absorption of
some chemical
source by
species of interest.
laboratory, this technique usually involves working with a calibrated
source.
In
NUVIS
Rayleigh scattered
The
UV
5800
proportional to
effect of these
source in the
I
striking
the atmosphere has
is
light
the ambient
approximately the spectral
blackbody (see Figure
4
where X denotes wavelength (Lenoble, 1993).
,
two natural phenomena
300nm wavelength
also utilizes this
background continuum source
K
A,
continuum
solar radiation, as discussed above.
solar irradiance
distribution of a
is
field applications, the
In the
region.
to
is
The
2.3).
The Raleigh
scattering cross section
The combined
provide a useful background continuum
COSPEC
instrument previously mentioned
continuum radiation source.
Understanding
how
molecular absorption spectroscopy
is
used to
measure
molecular concentrations requires a basic knowledge of the energetics of the molecules of
interest.
The
total
energy
E of a molecule
arises
from four contributions, represented by
the following equation:
E=E
E
rot
is
rot
+ E vib + E elec + E trans
the rotational energy of the free molecule about
its
(
center of mass.
E vib
is
2" 1
the
vibrational energy of the constituent atoms, typically in one of the molecule's vibrational
normal modes.
E
elec
is
the electronic energy analogous to that of the electronic energy of
20
atoms.
isolated
E
molecule's overall translational
the
is
trans
kinetic
energy.
In
spectroscopic applications, the rotational, vibrational and electronic energy levels are of
interest, since
energy
is
quantized for these
modes of free atoms and molecules. Because
of the bond energies, masses and moments of inertia of small molecules, transitions
between
rotational
states
are usually observed at infrared
Vibrational transitions are typically observed,
at
Electronic transitions generally occur at visible and
visible
and longer wavelengths.
and infrared wavelengths.
UV wavelengths.
(Lenoble, 1993)
Figure 2.11 provides a generalized representation of these energy levels for a
simple molecule.
characteristic
Like
many
UV absorption
Figure 2.11.
other molecules, gaseous
S0
2
exhibits
its
own
unique
spectrum. If an incident radiation field induces photons
Molecular energy curves for two electronic
states.
These curves
demonstrate the relationship between vibrational, rotational and electronic levels.
(This general depiction
is
not to scale.)
21
From Eisberg and
Resnick, 1985.
whose energies correspond
the molecule
may
to the difference
absorb a photon and
between two energy
jump
to a higher energy level.
wavelength-dependent absorption, or absorption spectrum, that
quantum
is
The
result is a
characteristic
of the
structure of the absorbing species.
Terminology
2.
Transmittance
is
the term used to describe the fraction of the energy of the
incident radiation that passes through an absorbing
beam of
levels of the molecule,
parallel radiation before
medium having
is
after
it
Figure 2.12
illustrates a
has passed through a layer of absorbing
a thickness of d and a concentration of c of an absorbing species.
to the interactions
intensity I
and
medium.
between the incident photons and the absorbing
attenuated to an exiting intensity / (Skoog, 1985).
particles,
Due
an incident
The transmittance
T
of
Absorbing medium
of concentration:
Path length:
Figure
2. 12.
*"
d
A parallel beam of incident radiation before and after
it
has passed through
a layer of absorbing medium.
the
medium
is
thus defined as the ratio
T = IIIn
22
(2.2)
A
Absorbance,
is
defined as
A=-log(T)=log(I
If the likelihood
(2.3)
of multiple absorptions and reemissions along the path
linear relationship
by Beer's
/I).
is
small, there
a
between absorbance and concentration of an absorbing species given
Law
A=a(X)dc
(Skoog, 1985).
The function a(X)
Law, the attenuation of a
is
(2.4)
called the absorptivity coefficient.
signal can be given
From
a is the absorption cross
Absorption
is
(2.5)
section and TV is the molecular
a quantity that differs from absorbance.
of the Einstein probability coefficient,
Bn
,
Beer's
by
I=I exp(-crNd),
where
is
but in practice,
number
It
density.
can be described in terms
it is
more often described
in
terms of an absorption coefficient, k(v). This parameter describes the fractional decrease
in flux density (intensity) at frequency
v per unit path length through the absorbing
medium:
-5I(v)=I(v) k(v) Sx,
where I(v)
is
incident
upon the layer of thickness
Sx.
(2.6)
Assuming
a
homogeneous
layer of
thickness d, integration yields
Id(v)=I (v)exp(-k(v)d),
23
(2.7)
where k(v) d
is
known
as the frequency-dependent optical depth, z(v).
increasing optical depth on an isolated absorption line profile
(Thome
et al.,
From
The
is illustrated in
effect
of
Figure 2.13.
1999)
a practical standpoint,
it
is
the effective area of an absorption line that gives
the most useful measure of spectral absorption that
absorption lines (curves) of Figure 2.13,
may be
is
taking place.
The areas under the
defined by the following equation,
"o
Figure
2. 13.
Hypothetical absorption line profiles demonstrating the effect that an
increase in optical depth has on a line shape. Lines
increasing optical depths, t(v)
which
is
.
a, b,
and c represent shapes
From Thorne
et. al.,
for
1999.
normalized to the incident flux density,
W=L
e
(l-exp(-k(v)j)dv.
24
(2.8)
Equation 2.8 equates well to transmittance through the following relation,
W = f Jl-T(v)]dv,
(2.9)
n
where
W
is
called the line's equivalent width, and T(v)
transmittance from equation 2.2.
spectral line
the
where the shaded area
continuum baseline,
Figure
Figure 2.14
2. 14.
is
the frequency dependent
further depicts the equivalent
width of
equivalent to the area between the line profile and
1(0).
Equivalent width measurement. Note that the shaded area
area between the line profile and continuum baseline, 1(0).
It
is
is
From Thorne
equal to the
et. al.,
1999.
should be noted that the actual absorption of an isolated spectral line can be
spread over a broad range of frequencies by natural, Doppler and pressure broadening
(Thorne
et al.,
1999).
Furthermore, the relation between the absorbance and I(v)
logarithmic, not linear, so equation 2.9 does not hold true for
individual equivalent line widths
is
all
cases.
thus a more detailed process than
overall trend of the absorption cross section of Equation 2.5.
25
However,
it
is
Measuring
measuring the
can be shown
that linearity applies as long as the absorbing line is not highly saturated
profile's center
utilized
widths
equivalent
using
nonsaturated
This
lines.
accomplished without correcting for any line shape effects and assumed
2.9
was applicable
see Thorne
et. al,
*
l(v)
at the
Therefore, for the purposes of this thesis research, Equation 2.9 was
).
calculating
in
(
to all equivalent width
that
was
Equation
measurements. For a more detailed approach,
1999.
The Curve of Growth
3.
In spectrochemical analysis by
means of absorption spectroscopy,
it
has become
standard practice to quantify the effects of an absorbing species abundance on the profile
of an absorption
line as
growth for a spectral
an empirical graph called a curve of growth.
line describes the
concentration of the absorbing species.
illustrated in Figure 2. 15.
known absorber
1-5
behavior of the equivalent width as a function of
It
normally takes the form of a log-log plot as
In practice, the equivalent widths are
concentrations in the laboratory.
measured
Figure
2. 15.
1.5
2.0
3.0
Typical curve of growth. (Also
curve.)
26
for a series
The equivalent width of lines found
r_
1.0
The curve of
known
5.0
as a calibration
of
in
field data
may be
subsequently compared to the curve of growth, which then yields the
molecular abundance of the
field sample.
Remarkably, curve of growth analysis has been found
to different line profile shapes
(Thome
et. al.,
1999). Hence,
a wide variety of spectroscopic instruments and conditions.
line shapes,
Figure
Gaussian and Lorentzian, are illustrated
2. 16.
Gaussian and Lorentzian
profiles.
From Thorn
in
to
it is
be relatively insensitive
of general
Two commonly
Figure
2. 16.
utility
across
encountered
This study will only
Areas under each curve are the same.
et.al.,
1999.
involve calculations based on Gaussian line profiles that
may be
represented by curves
having the generalized form:
F(x)=A
exp{-[(x-
27
AtfAf/21
(2.
10)
where F(x)
is
the profile,
A
is
the height of the Gaussian,
and A 2 is the width of the Gaussian. Equation
calibration of the
NUVIS
2.
A
}
is
the center of the Gaussian
10 will be utilized in the curve of growth
instrument and in the analysis of field data presented in
Chapters IV and V.
28
NUVIS
III.
BACKGROUND
A.
NUVIS was
NPS
designed, developed and constructed by former
Professor Dr. David Cleary and students,
MacMannis (MacMannis,
1997).
Todd Hooks (Hooks, 1997) and Andrew
the
is
It
Associate
latest
spectroscopic instruments developed by Dr. Cleary and
a
in
is
succession
of ultraviolet
the first to incorporate spectral
imaging capabilities (Cleary, 1998). For a more detailed account of some of this
work done on
UV instruments such as MUSTANG and DUUVIS, see Johnson,
In principle, the
NUVIS
instrument of today
developed by Hooks and MacMannis.
is
identical to the instrument first
Designed and
.
1996.
HARDWARE
OPTICAL, OPTOMECHANICAL AND ELECTRONIC
B.
earlier
built
to
test
the
utility
hyperspectral imaging at ultraviolet wavelengths, the instrument's optical design
classic
Rowland
scanner.
As
it
circle spectrograph
appears in Figure 3.1,
and are summarized
its
imaging design follows
NUVIS
is
in
The instrument
wavelengths.
of a push broom
a portable instrument capable of field
Table
3.1.
itself is
comprised of nine basic components, which are
Figure 3.2. Light enters the instrument through a
UV
that
a
Instrumental parameters remain as originally detailed by Dr. Cleary
operation with ease.
in
and
is
of
It
filter
window, designed
then encounters a scanning mirror assembly.
comprised of three sub-components:
1) the
mirror
29
itself,
2) a stepper
illustrated
to pass
only
This assembly
motor
driver,
is
which
mm M'Ssrtm
Figure 3.1.
NUVIS
instrument.
NUVIS PARAMETERS
Optical Design
Rowland
Imaging Design
Push Broom Scanner
Spectral
Response
300
Spectral Resolution
Full Field
to
Circle Spectrograph
375nm
0.3 run
of View
20° x 6° (350mrad x lOOmrad)
Bands (Spexels)
~ lmrad
640
Positions
480
Lines (available)
351
Mode of Operation/
SNAP/ 640x480x1 data array
SCAN/ image cube (640 x 480
Angular Resolution
DATA
output
Table
3.1.
NUVIS
>ZZZ
parameters.
30
From
x
ZZZ)
Cleary, 1998.
Figure 3.2.
NUVIS
components. Top: Schematic of NUVIS:
scanning mirror assembly, 3) baffles, 4) telescope mirror, 5)
grating, 7)
image
intensifier, 8)
CCD camera,
1) filter
slit,
window,
and 9) cable connectors.
Bottom: Photo of NUVIS instrument components, (with cover removed)
31
2)
6) diffraction
and 3) the
rotates the mirror,
Light reflected by the scanning mirror strikes a
the angular position of the mirror.
which focuses the beam onto the entrance
telescope mirror,
Rowland
optimized for
image
From
circle.
The
grating.
which provides feedback information about
shaft encoder,
the
slit,
image
intensifier.
operation, optically amplifies the signal
intensifier" s
output
is
which
is
C.
intensifier,
The
.
lum by 13um
pixels.
NTSC
lines.
1
30 frames per second
is
(fps) analog
video
limited to an eight bit grayscale and a
(Hooks, 1997)
THE COMPUTER, DATA ACQUISITION AND CONTROL SYSTEMS
The next major component of
The 233
MHz PENTIUM
Pro
PC
includes
equipped with standard 3.5 floppy and
256Mb Ram and
CD-ROM
is
an
NUVIS
data
supporting peripherals.
a 3.0
GB
drives as well as a
is
the
is
1
hard drive.
It is
GB
JAZ
internal
processed into digital form by a
PC
grabber interface card. Interfaces to the stepping motor controller and
shaft position encoder are provided
monitor
system
CPU, monitor and
The video output from NUVIS's camera
IMAQ-1408 frame
NUVIS
the total
acquisition and control computer, including the
drive.
The image
(CCD)
conventional
by 480
on the
coupled to a Pulnix TM-745e charge coupled device
through a standard RS-170 interface and
resolution of 640
located
by a factor of nearly 10 6
video camera. The camera has 768 by 494 pixel resolution with
The camera output
is
light travels to a field-flattened spherical diffraction
diffracted light then goes to an
UV
slit,
NEC
Multisync
LCD
by DAC-02 and PIO-24ISA/EISA
15 inch flat screen display.
external views of the computer system.
32
cards.
Figure 3.3
The
illustrates
Lh'U, Monitor, Keyboard, ana mouse. Mote
that all
still
components are
tsacK panel 01
c^u
witn network nuo
DAC-02 PIO-24ISA/EISA &
PC IMAQ-1408 frame grabber cards
easily operated while
attached.
inside the customized shipping case.
,
the
displayed from the bottom up.
Figure 3.3.
NUVIS
control computer.
Control of the scanning mirror drive mirror drive-is provided by the
stepper motor controller that
NUVIS
system deployed
is illustrated in
in the field,
involved with on-site operations.
A
Figure 3.4.
demonstrating
detailed account
required for field operation can be found in the
Figure 3.4.
Max-410
33
NUVIS
MAX-410
Figure 3.5 shows the complete
some of
the logistics and setups
of suggested system components
field checklist
stepper controller.
of Appendix A.
Left:
Volcanic Site
Figure 3.5.
NUVIS
Right:
system components and
34
Power Plant
Site
field logistics.
DATA ACQUISITION, CONTROL AND DATA DISPLAY SOFTWARE
D.
As NUVIS was being develpoed,
for instrument control, data acquisition
several key software routines
and data
display.
were developed
Various modules of
this
software were written in Microsoft Visual Basic and Interactive Data Language (IDL),
which
is
a Research Systems, Inc. (RSI) product.
in the following references:
in
Appendix
Hooks, 1997; Aitken, 1997; and Using IDL, 1997 as well as
B.
In essence, the acquisition of spectral
algorithmic loop.
interest falls
light
may be found
Specifics of the software
The scanning mirror
is
on the instrument's entrance
from the
slit
across the camera.
NUVIS
is
a simple
rotated so that one edge of the target scene of
slit.
The
diffraction grating then disperses the
Because the Rowland
stigmatic (image-preserving), the resultant
along the dispersion axis as a function of
image data with
2D image
ID scene
circle spectrometer is
consists of spectral information
height along the orthogonal axis.
In
other words, an image frame consists of a set of spectra for each scene pixel focused on
the entrance
strip
is
slit.
Each of these scene
pixels,
however, belongs to only a narrow vertical
of the whole target scene. So, the camera frame
is
stored, then the scanning mirror
rotated slightly to position the next vertical strip of the scene onto the exit
frame acquisition procedure
is
repeated.
This sequence of operation
is
slit,
and the
repeated until
data from the entire scene has been acquired. Subsequent off-line processing can then be
used to generate a hyperspectral cube data
set
35
and to display the data
in various formats.
Sample output from a
This particular set
is
Two
deployment
is
provided in Figures 3.6 through 3.8.
drawn from data taken of an
The image scene includes
some
field
industrial coal burning
three operating smokestacks.
level of sulfur dioxide emissions, as well as a
of the stacks have modern to contain
third (dirty) stack emits
S0
2
at levels
S0 2
(Further detail of this field collection
number of other
be found
The black and white image of Figure
effluent products.
scrubbers that reduce emission levels.
on the order of ten times the
may
plant.
All three stacks are producing
This particular collection was conducted on
scrubbed stacks.
power
3.6
in
is
site at
ppm
The
values of the
a range of 1.5 km.
Chapter V.)
produced directly from the sensor's
recorded digital signals through a transpose, rotation and resizing of the data array.
Although the stacks are quite
visible, this
format does not provide enough information to
allow for the isolation and identification of SO, signatures in the data.
Figure 3.6.
NUVIS
IDL
black and white image.
false-color
This image was produced by the
procedure, Bw_roughimage.pro.
36
A
From Cleary,1998.
image can also be created to survey the
This was done for the power plant work,
data.
using color assignments:
Red
:
350-375nm,
Green: 325-350nm,
Blue
This approach results
in the false
:
300-325nm.
color image of Figure 3.7, in which potential regions of
sulfur dioxide appear yellow in color.
Figure 3.7.
NUVIS
false-color image.
"color_image.pro". Regions of
Field operation of
data
in
NUVIS
S0 2
This
is
appear yellow
necessitates a
to
allow
this.
modeled absorption cross section of
illustrated in Figure 3.8.
It
in color.
means of obtaining quick looks
order to ascertain whether or not the instrument
procedure was developed
IDL procedure,
From Cleary, 1998.
a product of the
is
functioning properly.
performs a least-squares
SCX
fitting
Output displays from
this
37
An IDL
of the data to a
procedure are
This corresponds to 200 different spatial (altitude)
(wavelength) slices of the hyperspectral image cube.
at the
vs. spectral
For presentation purposes, data
midway through
processing has been interrupted
windows presented
the output
of Figures 3.6 and
that
actively displayed
presented and
is
3.7.
from from
the scene at line
to the data analyst.
Window
During the execution of
right to
left.
A
1
number 99
shows a display similar
this procedure,
full false-color
to illustrate
processing
image of the scene
to
is
is first
then actively processed, scan by scan (line by line) within the window.
The display labeled
as
Window
2 represents the raw data or digital
number
that is
recorded for each individual pixel in the spatial (altitude) vs. spectral (wavelength) slice
corresponding to line number 99.
(plane),
The raw data displayed on
corresponds to the line dividing the color and black and white images in the
the
This line marks the point
left.
Within
display.
opposite
is
Window
2, the
at
This raw display
during system operation. Feedback
of the same data
scan
sliced to produce such a
The
is
is
at a
time
the only real-time feedback provided
provided through the
later in analysis
Thus, Figure 3.8 also
process
presented to the user
is
NUVIS
displayed by the Visual Basic instrument control program
and two.
window on
darker pixels represent lower levels of detection.
2 also parallels the actual raw data display that
during instrument operation.
is
right,
true for the brighter pixels.
Window
which
which the image cube was
the
is
operational
NUVIS.EXE.
presented in the lower
illustrates the representative
window,
A view
windows one
output at the 150"' vertical
line.
In the field, the data acquisition
and control
PC
can be networked to another
computer, typically a laptop PC. The various IDL data analysis and display procedures
38
can then be run on the laptop while the host
instrument.
PC
is
otherwise occupied with running the
This provides a quasi- real- time means of verifying
NUVIS's
operation
in the
field.
Window
1
Window
I
Figure 3.8.
for the
99
th
SO^ algorithm
vertical line scan
"assay_image.pro."
the 150
th
output.
Upper Windows demonstrate output
of the image scene during execution of the algorithm,
Lower windows demonstrate processing some
scan position.
In
Y)
displays
each case, altitude
direction.
From
39
is
displayed
Cleary, 1998.
time later at
in the vertical (spatial-
THIS PAGE INTENTIONALLY LEFT
40
BLANK
LABORATORY CALIBRATIONS OF NUVIS
IV.
PURPOSE
A.
In order for
NUVIS
to
be applied to
field
measurements of
of calibrations must be performed, either beforehand or on
First,
site
S0
2
plumes, two types
during field deployment.
an approximate spectral wavelength scale must be established, so that spectral
signatures of interest can be reliably and reproducibly identified.
Second, the relevant
curves of growth for the signatures must be available so that the instrument can be
employed
for reliable
and credible spectrochemical abundance measurements.
This
chapter summarizes the laboratory calibrations that were performed in these contexts.
The next chapter discusses applications of the
B.
calibrations to actual field data.
WAVELENGTH CALIBRATION
Before continuing onto generation of the curves of growth,
discuss the association of instrument spectral samples (sometimes
wavelength.
it
is
necessary to
known
as spexels) with
This phase of calibration was conducted by Dr. Cleary, just prior to
involvement with the
NUVIS
project.
first
my
Results of the wavelength calibration were saved
MB data file entitled wl_cal.dat.
in a 1.2
The
source.
calibration
was accomplished using a standard spectroscopic emission
The source used was a platinum
(Pt)
hollow cathode lamp, whose characterisitics
and calibration are traceable to the National
(NIST).
This source was chosen because
emissions in the 300 to
line
400nm wavelength
41
Institute
it
of Standards and Technology
exhibited a
range.
number of
useful spectral
This particular lamp
is
used in
conjunction with a Harrison 6522
set at
400
volts
15mA. Nine
DC
power supply. For operation, the supply
and 0mA. The current
is
is initially
gradually increased to the operational level of
Pt emission lines were identified and utilized in the calibration.
These lines
are listed in Table 4.1.
Wavelength
Line
Table 4.1
.
Line
Wavelength
Wavelength
Line
504.26
337.82
359.36
506.47
340.81
369.42
54.54
352.05
372.71
Nine platinum emission
lines
chosen for the wavelength calibration.
FromCleary, 1998.
As with any
diffraction grating instrument, the grating equation applies. (For
further explanation see both
MacMannis, 1997 and Hecht, 1998.)
This means that the
actual wavelength of a spectral feature should be a sinusoidal function of
(spectral sample) in the
measured data frame.
It
the sinusoid could be adequately approximated
range. This behavior
is
summarized
in
its
position
was found experimentally, however,
that
across the 300 to 400
nm
by a
linear
fit
Equation 4.1
Y=(0.118)x + 300.775,
where x
is
the spexel number,
spexe! and
Y
is
m the spectral dispersion,
(4.1)
b the wavelength of the '"zeroth"
the corresponding wavelength in nanometers.
illustrated in Figure 4.1.
This calibration was applied in
all
This behavior
instances
necessary to associate wavelength to a corresponding spectral sample.
42
where
It
it
is
also
became
also provided a
simple means for calibrating the wavelength scales of spectra measured with
NUVIS,
provided that the optomechanical configuration of the instrument remained stable.
Wavelength Calibration
200
300
600
4C
Spectral Sam pie
Figure 4. 1
.
Wavelength calibration associating band number/spexel
to
wavelength.
C.
EXPERIMENTAL OVERVIEW
As
stated earlier, only absorption lines existing longwards
for laboratory
measurement because absorption
shorter wavelengths
were
lines that
of 305nm were chosen
appeared in the field data at
either saturated or subject to severe noise because
decreasing background illumination.
of rapidly
Figure 4.2, vividly demonstrates the relatively
closely spaced group of absorption lines that were targeted for calibration.
In the end,
four absorption lines were identified as suitable for equivalent width measurements in the
calibration.
The
spectral
sample numbers and wavelengths corresponding to the centers
of each of these absorption lines are
listed in
Table 4.2. Because
43
NUVIS was
expected to
operate over a wide range of column densities, the decision was
on the instrument's lower
made
to first concentrate
sensitivity limits.
Measured S02 Absorption Lines
c
.2
0.8
%
0.6
-
X>
<
«
0.4
Is
13
0.2
-*
200
100
400
300
Spectral
500
600
Sample Number
Figure 4.2. Example of SCVs closely spaced absorption
plot,
lower spectral sample numbers equate to higher
and higher numbers equate
Measured
700
to
In this
lines.
UV wavelengths
lower ones. See Figure
4. 1.
1
2
3
4
535
554
572
590
313
310.9
308.8
306.5
Absorption Line
#
Spectral Sample
Number
Center
Wavelength
(ran)
Table
D.
4.2.
Absorption lines targeted for measurement.
LABORATORY MEASUREMENTS OF SO,
There are two contraints on the gas concentrations
calibration studies.
First,
that are useful for laboratory
the experimental noise floor sets a practical lower limit to the
instrument's sensitivity. That
is,
weak
spectral features that are buried in the noise
44
must
simply go unmeasured.
This translates into a lower useful limit to
used in the laboratory. Second, as discussed
earlier,
S0
2
concentrations
highly saturated spectral features do
not provide reliable measurements of spectrochemical abundance, and so should be
This sets a practical upper limit on laboratory gas concentrations.
avoided.
S0 column
weighing such considerations,
six
summarized
4.3.
They
are
in
Table
abundances were decided upon for study.
2
Column abundance
Sample
(oc
concentration)
[(#molecules/m
Table 4.3. The six
After
5.33
3A
3A
7.48
4B
4B
2.25 x 10
5.61 x 10
4.24
)
m)]
xlO 21
1A
2B
1.07
3
21
xlO 21
xlO 22
22
xlO 22
test cell concentrations utilized in
generating the curves of
growth.
Gas Test
1.
S0 was
2
UV
placed in transparent test cells of various lengths and illuminated with
continuum radiation
1.55 meter long
Cells
PVC
to
produce laboratory spectra. The
tube with quartz
windows
first test cell
(see Figure 4.3).
employed was a
Once evacuated with a
conventional fore pump, the cylinder could be filled through a pressure regulator with
0.11
%
sulfur dioxide in an inert dilution gas
pressure gage
Figure 4.4.
(
to
760
mm Hg
)
that
(N 2 ).
were both
For various reasons, the 1.55
m
45
The valve manifold and Matheson
utilized in this process are illustrated in
cell
proved
to
be cumbersome, prone to
leaks,
and too long for producing low column abundances of
test cell
S0
2
was designed.
Figure 4.8. Pressure gage and valve manifold.
Figure 4.4.
1
.55
m gas test cell #1
46
.
Therefore, another
The next
test
cell
was considerably shorter and more
potentially be utilized not only in the laboratory but also in the field.
design was more portable than the
Even though
this
abandoned for two reasons:
1)
Figure 4.5.
entrance
(filter
The
window) of NUVIS.
test cell
2)
The
It is
illustrated in
first, it
was quickly
could not reliably be coupled to the optical
cell
and could potentially damage the instrument's
and could
portable,
combination was mechanically unstable
filter
window.
Figure 4.5. Gas test cell #2.
To remedy
obtain a
use
these differences with the first
more enriched mixture of
much
shorter gas test cells.
sulfur dioxide
two gas
(Matheson
test cells,
1
0%
it
was decided
to
and
to
mixture in
air)
Four inch diameter (10.16cm) by 3.25cm inner path
length quartz cells were purchased from Weiss Scientific Glass Blowing
47
Company. They
are
shown
in
Figure 4.6. These smaller
be mounted easily
and
field use.
at
NUVIS's
had a mechanical advantage: they can
optical entrance filter
window, permitting easy laboratory
A minor difficulty with their use was discovered, however.
window diameter
is
only 8.77cm once the
aluminum housing/mounting.
instrument's field of view.
regions and a
cells also
new
set
cell
is
installed
in the
This cause slight vignetting at the
During thesis research, care was taken
of six inch diameter
development. In Figure 4.7 a layout of the
Figure 4.6. Four inch gas
test cells.
test cells
filter
Their available
shop fabricated
corners
to
of the
exclude these
and associated mountings are under
window and
the gas test cells
is
shown.
Photos taken before and after installation.
2.873-
r
03.450
Figure 4.7. Test
cell
comparison. Dotted line represents
NUVIS
filter
window. Circles represent available cross sections associated with the 4 inch
(inner circle) and 6 inch (outer circle) cells respectively. The reduced cross
sections are due to the cell mounts themselves.
48
The
filter
window
is
represented by a dotted line and the two circles (of diameter 3.45
and 5.475 inches, respectively) represent the available cross sections provided by the four
new
inch and
was
six inch cells.
2.
Calibration Source and Reflective Screen
During
initial calibration
utilized as a
(uW/cm
2
nm).
measurements
EG&G Gamma
maximum
These source
in the laboratory.
A
a
NIST
tracable
of this source
is
was on the order of
to replace the
UV source and was utilized throughout all
deuterium lamp.
calibrations.
-
50cm
#
i
•
j
«
A
2-
«,
v
s
2
M
C
i
4>
U
i
V
=
2
"
a
1
f
c
83
J
•
§
i
t
15
\
*
*
i
1
A
s
i
i
«
1
§
-
',
a-
*
\
A
S
"S
£
2 0.5
A
-
A*
i
A*
i
-
200
A A
250
300
350
400
450
500
Wavelength (nm)
Figure 4.8. Spectral irradiance of FEL calibration source over operational
NUVIS
wavelengths.
49
It
The spectrum
illustrated in Figure 4.8.
£>
i
375 run
1000 Watt quartz halogen tungsten filament lamp
Spectral Irradiance of F-309 at
Z.D
to
proved to be too low for practical absorption
(FEL-309 IR) from Optronics Laboratories was chosen
is
300
In the
specific output irradiance
levels
deuturium lamp
Scientific
background continuum illumination source.
spectral region, this lamp's
10" 2
attempts, an
The lamp was mounted
which was used
to
maximize
in
an enclosed box with a controllable output aperture,
the optical fluence
on a
diffuse reflective screen, described
below. The lamp was powered by a Hewlett-Packard 6030-A power supply.
drive current
was
set to 8.0
amps.
Maximum
The lamp required a 15 minute warm up period
(according to the manufacturer) to achieve steady state operation before measurements
could be taken.
The
from the
light
FEL lamp was
scattered
by a reflective screen assumed
an ideal diffuse Lambertian surface (see Hecht, 1998).
diffuser
(
with greater than
see Figure 4.9
It
).
97%
reflectivity
in
was placed 1 50 cm from
the
the
wavelength region of
FEL and
Plot
—
o
u
OS
L_
S
rz
-j
360
340
380
Wavelength (nm)
Figure 4.9. Screen reflectance values.
50
interest
approximately normal to the
0.995
320
be
The screen was a Spectralon
Spectralon Reflectance Target Calibration
300
to
400
lamp-screen axis.
axis.
The NUVIS instrument was oriented
The geometry of the setup
is
this
normal
illustrated in Figure 4.10.
Figure 4.10. Experimental set up. Instrument
from screen-lamp
45 degrees from
at
axis.
UV
source
is
is
placed at 45 degree angle
enclosed by the box and
is
positioned for normal incidence to the screen.
3.
Test Cell Preparation
Calibration cells were prepared by
vacuum was drawn on
exhaust hood. The
background
cell
first
connecting them to the valve manifold.
the cell using a mechanical fore
gas pressure
was monitored by
calibrations, evacuated cells could be used.
51
the
pump
A
vented through a gas
Matheson pressure gage. For
If the cell
was
to
be
filled
with
sulfur dioxide, the appropriate concentration
was introduced by using a standard gas
regulator while monitoring the pressure.
The
entire setup (see Figure 4.11)
marked pressures
in increments of
5mm
had some inherent limitations. The gage only
of mercury
(mm
Hg).
This did not permit very
precise determination of cell gas pressure and caused uncertainties in gas concentration,
particularly at
low pressure.
Figure 4.11. Cell preparation setup.
Data Acqusition
4.
Once
instrument.
the lab
first
a cell
was prepared,
Because flourescent
it
was mounted
lights exhibit well
was darkened before measurements were
tested to determine
using the
VIDEO
in front
of the
filter
known emissions
taken.
window of
the
in the ultraviolet,
A range of instrument gains
was
where sensor saturation occurred. This was accomplished by
feature displayed in the
NUVIS
52
control
window.
An image
intensifier
gain of 7.0 was set in the control window.
the
number of images per
SNAP
mode,
in
SNAP
first
Spectrum were then recorded using the
to 100.
which the scan mirror
Measurements were obtained by
is
not rotated.
setting
NUVIS
In addition to gas spectra at each
concentration, spectra of empty cells were recorded for comparison purposes.
spectra were saved to a
GB
Jaz cartridge as a
30MB NUVIS
first
experiment, fifteen different spectra were recorded.
spectra from experiment #1 are illustrated in Figure 4.12.
spectra calculated by averaging over
averaging over
500
100 images
all
are displayed.
As
to
all
480 pixels
enhance the experimental
discussed earlier, absorption
cell.
The
are
results
spectrum exhibited identifiable
the sensitivity of
NUVIS was
20
order of approximately 10 m"
S0
illustrated
2
in
is
ratio.
53
Only
spectral
samples >
.
the corresponding spectrum of the
Figure
insufficient to detect
.
image column and then
characterized by the ratio / / 1
4.13.
spectral absorption features.
2
Three representative
These plots represent mean
in the vertical
Hence each S0 2 spectrum was then divided by
evacuated
file.
First Results
5.
For the
1
The
S0
2
at
Neither of these
initial
This demonstrated that
column abundances on
the
Experiment #1
Sample Spectra
-*—
35
i
*
•
^jIBt
"*
Vacuum Cell
11% S02 @ 760mm Hg
0.11% S02 @ 680mm Hg
^<
30
«•
^
i
^"*
25
£
E
Z 20
w
C»
Q
15
.
10
5
580
560
Spectral
Sample Number
Figure 4.12. Sample spectra of experiment #1 including measurements from
one evacuated
cell
and two SO,
test cells.
Relative Absorption Spectra
(Experiment #1
:
4" Test Cell)
18
0.16
"
S02@ 680mm Hg
0.11% S02@ 760mm Hg
0.11%
520
540
580
560
600
620
Sample Number
Figure 4.13. Relative absorption spectra of experiment # 1
Spectral
54
64
GENERATING EXPERIMENTAL CURVES OF GROWTH
E.
Approach
1.
A
S0
2
second
set
of experiments was undertaken with higher concentrations of
Various concentrations were tried until unambiguous detection of
.
was
features
the cells
spectral
2
Given the limitations of the pressure gage, described above, and the
assured.
of only
availability
S0
0.1
1%
was somewhat
and
10% S0 2
restricted.
But,
mixtures, the range of reliable gas abundances in
when
data from these abundances were combined
with those from earlier measurements using the 1.55
m
cell,
enough data points
for the
generation of calibration curves of growth were achieved. The details are summarized in
Table
The second experiment was
4.4.
concentration in an attempt to see the
cell
first
concentration meant utilizing the
be backfilled with
10% SO,
to
increase
signs of absorption line formation.
10% S0 2
as the test source.
imposed by the Matheson gage setup (Figure
restrictions
cell to
designed
initially
only as low as 50
test
cell
Increasing
Unfortunately, the
4.4) limited the four inch test
mm
Hg.
This was the lowest
achievable concentration with the existing pressure regulator, gage and valve manifold.
Details of
growth
all
cell
parameters and concentrations utilized in generating the curves of
may be found
Sample
in
Cell
4.4.
Source
Pressure
(mm Hg)
Path length (m)
(%S0 2 )
Number
1A
2B
3A
4A
5B
6B
Table
Table
4.4.
4"
10
50
1.55m
4"
0.11
100
1.55
10
70
.0325
.0325
4"
10
100
.0325
1.55m
0.11
400
1.55
1.55m
0.11
760
1.55
Cell parameters utilized in generating the curves of growth.
55
Experimental Results
2.
The
S0 2
resultant
raw spectra of experiment #2 are
spectra are ratioed with
spectra of Figure 4.15
their corresponding
are generated.
marked
A
(new) and
B
evacuated
cell
spectra, the ratio
Geary
data sets) in Figure 4.16.
(old) data respectively.
(Experiment #2)
520
540
560
Spectral
Figure
4. 14.
Raw
600
580
620
640
Sample Number
spectra from experiment #2, including measurements
from one evacuated and three SOt
56
test cells.
The
They have been adjusted
Raw Spectra
500
these
These spectra are plotted along with three
additional absorption spectra (calculated from the
data sets are
When
illustrated in Figure 4.14.
Relative Absorption
Spectrum
(Ratio Spectra from Experiment #2)
09
10%SO2@50mmHg
10%SO2@70mmHg
OS
0.7
10% SO2@100mmHg
c
5
Aft
I
s
l
;
6
l_
o
«
05
<
Q)
4VA 7
0.4
>
2
03
^v\
0)
.«.
02
^-»-*^/"
f
-^V^
500
520
540
560
Spectral
Figure
4. 15.
580
600
620
640
Sample Number
Relative absorption spectra from experiment #2.
Relative Absorption Spectra
(Ratio Spectra from
350
400
450
500
Spectral
Figure
4. 16.
Combined Data
550
Sets)
600
Sample Number
Relative absorption spectra generated from combined data
57
to the
same baseline
fluctuations.
by
the
It
five spectral
was
to
compensate
for
also discovered that the
sample numbers from the
now
had
instrument
instrumental
new
four
(Alpha data
sets)
new
demonstrate
To compensate, bands
were uniformly adjusted by
data sets prior to adjustment (one
this effect.
The adjusted
deployments and had
field
five
corresponding absorption lines of the old spectra (Beta data
three of the
lamp
This was not too surprising, since
major
experienced great deal of handling during shipment.
spectra
and
effects
drift
data set's spectral features were shifted
earlier data.
completed
gain
is
spexels
to
in the
new
match the
Figure 4.16 displays
sets).
annotated with an arrow) to
spectra are illustrated in Figure 4. 17.
Adjusted Relative Absorption Spectra
(for
500
combined data
540
520
560
600
Spectral Sample
Figure
4. 17.
sets)
Adjusted relative absorption spectra.
account for wavelength
shifts in the instrument.
details
on
cell
concentrations
58
620
640
Number
New data
sets adjusted to
Please refer back to Table 4.4 for
1A through 6B.
Analysis of Experimental Data
3.
The IDL proceedure GAUSSFIT was
utilized to
fit
individual Gaussian profiles to
each of four absortion lines (centered about spectral sample numbers 535, 554, 572 and
590) in each of the six data
sets.
The
resultant
are illustrated in Figures 4.18 through
Their equivalent widths were calculated by integrating each Gaussian (24 in
4.23.
accordance with Equation
line,
six
different equivalent width
recorded, one for each SO, concentration.
measurements were
Equivalent widths were then plotted against
A least squares
concentration and resulted in the generation of four curves of growth.
nonlinear
fit
has been performed to each of the four sets of measurements
sample numbers 535, 554, 472, and 590, respectively) using Mathematica
modeled
models
A
calibration curves.
is
all) in
Table 4.5 provides details for each of these 24 absorption
2.9.
For each absorption
lines.
cell
fits
summary of
to
(at spectral
produce four
the resultant equations for each of the four
provided in Table 4.6 and curves of growth are plotted in log-log scale format
in Figures 4.24.
S0 2 Column Abundance
(#molecules/m
Nominal
X:
306.5
2
)
5.33
5.61
7.48
10.7
22.5
42.4
0.382
0.561
0.761
1.12
1.64
2.57
Sample #572
0.375
0.396
0.576
0.673
1.06
1.40
Sample #554
0.250
0.303
0.422
0.739
0.902
1.68
Sample #535
0.211
0.281
0.384
0.505
1.05
1.39
xlO 21
Equivalent Width of
Spectral Feature in
nm
at:
Sample #590
nm
308.8
nm
310.9
nm
313
nm
Table 4.5. Line equivalent widths for measued absorption ines as a function of SO,
column abundances.
59
Gaussian
(Spectral
Fits
Sample Numbers = Wavelengths)
0.9
1A
08
e
O
— 1AFIT
07
a
5
06
§
05
V
B
.-;-„
%
S
or
'A
0-3
'
•
02
ai
Spectral Sample
Figure
4. 18.
Number
Spectral
Concentration 1A
Figure
4. 19.
Sample Number
Concentration
2B
1
09
09
3A
08
4A
08
— 3A FIT
07
i
°7
— 4AFIT
.
a
06
foe
i
05
§
a
>
8
04
5
i
01
05
A
o
=
0"
1
03
^y
' saA-'^
Q2
SB
540
590
Spectral
600
530
\
7
Q1
640
620
500
Sample Number
V
520
5S0
510
Spectral
Figure 4.20. Concentration
\
02
V
01
/
/V W
n
3A
5S0
600
620
640
Sample Number
Figure 4.21. Concentration
4A
-1
Q9
5B
r\
'
'
'
/
V
6B
07
S
06
a
1
6BFIT
A1
<
V"
<
i
05
7
>
ffl
tutor
*
i
'
1
•
ru V
08
if
/ \
5BFIT
•
I
04
03
02
01
Figure 4.22. Concentration
5B
/
/
J/
Figure 4.23. Concentration
60
6B
\4>
Sample Number
Equation for Modeled Curves of Growth
of Spectral
Feature
535
W= -28.234 + 0.5685
554
W=-3\A22 +
572
fF= -23.975 + 0.4867
590
W= -48.372 + 0.9754 In
In
(g)
0.6268 In
(
In
g)
(g)
( g)
Table 4.6. Equations for modeled curves of growth.
(W is the
equivalent width in
nm
and g is the column abundance
Curves of Growth
for
in
m" : )
SO.
3.0
-
Spectral
Sample § 535
1.0
0.3
m^\
3.0
i
1
Spectral
1
i
1.0
c
V
o
i
Sample § 554
•
E
c
0.3
i
y
3.0
i
i
i
Spectral
i
i
Sample
jf
572
1.0
cr
UJ
0.3
i
3.0
1.0
i
i
Spectral
i
i
|
i
i
Sample § 590
^
_
_
0.3
i
5x1f/
1
i
i
i
i
|
IxlO22
'
2X1022
i
5X1022
S02 Column Abundance (m~ 3 )
Figure 4.24. Laboratory generated curves of growth for SO,
61
THIS PAGE INTENTIONALLY LEFT
62
BLANK
APPLICATION OF LABORATORY CALIBRATIONS TO FIELD DATA
V.
As
characterize and to quantify
analysis instrument.
Chapter IV
NUVIS's
The completion of the laboratory
now permit us to
remote sensing spectrochemical
capabilities as a
calibration procedures described in
apply these qualitative calibrations to real field data.
FIELD EXPERIMENTS
A.
Data described
in this chapter resulted
NUVIS. Two
field collections
thesis project.
The
of
been the need to
stated earlier, the motivation for this thesis research has
power
plant.
The
were conducted
third field collection
this thesis research.
from three separate
field
in 1998, just prior to the inception
was conducted
in April
of this
of 1999 in direct support
Each of these experiments was conducted on
site at
plant provided three operational smokestacks that
remote measurement through clear desert
experiments with
a coal-firing
were available for
air.
During the 1998 collections, only two of the three operational stacks had
operational
S0
2
scrubbers.
Thus, the third stack provided a
signature than the other stacks.
higher
S0 plume
2
Direct in situ measurements provided by plant managers
indicated that the clean stacks emitted
ppm).
much
S0
2
at levels in the tens
of ppm (approximately 40
Similar measurements indicated that the unscrubbed stack operated with
SO,
emissions in the hundreds of ppm (approximately 100 to 400 ppm). Photos of the power
plant
may
be found in Figure
5.1
.
There are actually
six stacks in the
image scene. Three
of the six were recently built because of mandated reductions in pollutant emissions.
(Cleary. 1998)
63
Figure 5.1. Coal burning power plant. Photos were taken during 1998
1999
but in
During the most recent on
had just brought the
S0
emitted
The
2
at
site field
final set
or below 40
ppm
of scrubbers on-line.
This meant that
2)
and
1998
three stacks
concentration.
acquisition of experimental field data followed a procedure
The source of background continuum
was measured, was
to
all
1)
much
NUVIS was
natural sky scattered
UV
radiation, against
solar radiation.
be ratioed with a clear sky spectrum in order to quantify
shows a representative sky spectrum recorder
which
S0
like that
operated in
hyperspectral imaging mode, providing spectral information for each
its full
in
measurements, in the spring of 1999, the power plant
used in the laboratory, but with the following modifications:
pixel.
(left)
Only two of the three scrubbers were in operation
1999 scrubbers were on-line in all three stacks.
1998 photo courtesy of Cleary, 1998.
(right) collections.
2
2D image
absorption
Hence, each spectrum had
S0 2
absorption.
Figure 5.2
in the field.
Figure 5.3 shows a corresponding spectrum for a region of an image including the
effluent
plume from a smokestack.
It is
evident that absorption features exist that are not
64
present in the clear sky spectrum.
This
spectra are plotted jointly. Figure 5.5
The
plot.
typical
By
S0
2
is
particularly obvious in Figure 5.4,
shows the
ratio
where both
of the plume and clear sky spectra.
spectral signatures are evident at the high spectral
fitting profiles to these signatures, calculating
sample side of the
each profile's equivalent width,
and comparing those with the calibration curves of growth, effective column abundances
of S0 2 can, in principle, be determined.
Sky Data
100
200
300
Spectral Sample
400
Number
Figure 5.2. Sky spectrum recorded by
65
500
(Spexel)
NUVIS.
600
Plume Data
Spectral
Sample Number (Spexel)
Figure 5.3. Spectrum from smokestack plume.
Sky to Plume Comparison
300
350
400
450
Spectral
500
550
600
Sample Number
Figure 5.4. Sky and plume spectra comparison.
66
Ratio of Plume to Sky Spectrum
a
.2
o
<
_>
'•(—
p4
100
400
200
300
500
Spectral
Sample Number
600
Figure 5.5. Plume-to-sky ratio spectrum. Values range from 0-1.0
corresponding to 0- 1 00% absorption.
B.
RESULTS
In order to test the calibration procedure, representative sky and
interest
plume regions of
(ROIs) were selected for the "dirty" stack from one of the 1998 data
particular set
was obtained
illustrates the regions
at
a range of approximately 0.2
of interest selected for analysis.
km
from the
stack.
sets.
This
Figure 5.6
Figure 5.7 illustrates the
mean
spectra for each region, and Figure 5.8 displays the resultant ratio absorption spectrum for
the plume,
NUVIS
which has been labeled
as stack #1.
These
results validated the
had the capability of detecting sulfur dioxide plumes
of Figure 5.7
is
at short range.
claim that
The image
produced through the Regions of Interest algorithms of the Environment
67
for
in
Visualizing
Images (ENVI) software.
ENVI
Additional
analysis
further detailed
is
Appendix C.
Stack #1 with ROI
(at 0.2km)
|
Sky (1000
ROI Mean
pts)
250
Plume (1000
|
200
1
a
4i~L
iftjt
pts)
[JJr\gI
Spectra
H^^Q
I^Mn^f
J|
i
\
^
150
—
100
Sk v
H Jl
\
2
50
-
100
200
300
Spectral
Figure 5.7.
(see
400
500
600
Sample Number
ROI mean
spectra
ROI's of Figure
5.6)
Relative Plume Absorption
500
520
540
Spectral
560
580
Sample Number
Figure 5.6. Regions of
Interest (ROI's) used to
extract sky and
Figure 5.8. Relative plume absorption,
plume
(see Figure 5.6)
spectra.
68
600
Applying the same Gaussian curve
fitting
the lab data resulted in the curve
measurements were made
for
numbers 535, 554 and 572,
fit
techniques to this spectrum as were used for
illustrated
in
each of the absorption
Figure 5.9.
Equivalent width
lines centered at spectral
sample
respectively.
Stack #1 Gaussian
Fit (at 535, 554,
572)
to.
i
o.
>
'-5
It
»
10
20
30
40
50
Relative Spectral Sample
60
70
80
90
10
Number
Figure 5.9. Stack #1 Gaussian curve
fit.
This resulted in equivalent widths of 1.80, 1.40, and 2.03 nm, respectively.
laboratory curves of growth were then used to deduce the corresponding
column abundances.
plume
The
S0 2
Results from this application of the calibration curves to the field
data are summarized in Table 5.1.
69
Spectral
Table
Sample
Equivalent Width:
W
Column Abundance:
(nm)
535
1.80
8.88
554
1.40
3.42 x 10
572
2.03
1.60 x 10
5.1.
(m"
q
2
Number
)
xlO 22
22
23
Resultant column abundances of stack #1 from three spectral features.
These are the effective column abundances
layer of gas of uniform
number
density
for the plume,
n and thickness L,
in
assuming that
it
is
a
which case the column
abundance would be simply
q=nl.
However, the
So the actual
the
actual
plume geometry
is
approximately cylindrical, rather than layer-like.
effective absorption path length,
plume of diameter D, would be
(5-1)
the
assuming a uniform effluent density across
mean
thickness of a cylinder of that diameter,
as illustrated in Figure 5.10 and in Equation 5-2
where the
effective path length
STACK PATH LENGTH
NUVIS
*^
Figure 5.10
Plume geometry.
70
is
L eff
,
Leff
L
~
2
2
-
V7
x2(h
_xD
(5-2)
D.2
| dx
-D2
Therefore, for a measured column abundance q across a plume of diameter D, the
number
density of
SO :
in the
plume
will be
n=
Of course,
this
assumes
to the cylinder axis
measurement
is
that the
the horizontal, then
=
4^
T~ ~7T
(in other
words, in a horizontal direction). If the
at close range, sighting at
L e^must be
(5-3)
spectrum was measured along a direction perpendicular
of the vertical plume
performed
c
an elevation angle 6 with respect to
projected along a slant path, giving
nD
from which
it
follows that the measured
S0 2 number density
^=
4ccos£
t
7lD
71
will be
(5 _5)
Spectral
Table
6.75
554
2.60 x 10
572
1.22x10"
S0 2 number densities
5.2.
However,
number
parts per million
:1
plume S0 2 abundance
number of molecules per cubic meter of plume
(i.e.
monitoring devices
(ppm) of stack
in the stacks
express
effluent, a standard unit
convert our spectrochemical measurements to those units
S0 2
gas).
of
in the fields
So,
ratio.
we must
a meaningful comparison
if
in
concentrations in
of measure
chemical and environmental engineering, called the gas mixing
to
)
with geometric scaling.
analysis just discussed yielded the
density
in situ pollution
xlO
(m"
21
535
The spectrochemical
the form of a
S0 2 Number Density
Sample Number
is
be made.
Images indicate that cylindrical plumes
rise
a considerable distance above the
tops of the stacks without changing their overall shapes or diameters.
This implies that
the gas pressure in the flow must be approximately isobaric with the surrounding cool
This means
above the
that, for the
comparitively short section of plume observed immediately
flues, Bernoulli's
equation (Halliday and Resnick, 1997) demands that
Pair
where
P
air
is
the
air.
hydrostatic
=
Pplume + 0.5 Ppfome v~plume
pressure
(5-6)
,
of the surrounding cool
corresponding hydrostatic pressure within the
72
warm plume, ppuime
is
air.
the
Pp ume
i
is
the
mass density of
the
plume
gas,
and vplume
is
plume gas flow
the
velocity with respect to the surrounding
air.
The
actual
situation
Bernoulli's equation
P
good approximation, somewhat simpler than
a
to
is,
would imply. Since
=
air
1
5
.0x1
Pa,
pplume = pair =
3
.3kg m" and vpUtme
1
the magnitude of the hydrodynamic term in Bernouilli's equation
0.5
This
is
much
smaller that
P
air
be the sum of the
constituents,
thermodynamic equilibrium
and Resnick,
1
997),
at
PSo2
and
that
is
and
for
calculate
-
.
With
that
Tphltne
is
,
we have
S0 2
,
other
me must
and other effluent
mixed and
well
is
in
Therefore,
air.
(nair ^air
I
pplume)
in hand, the
n S02
itso2
,
constituents,
10~23 J/K),
and
and the
T^
is
the
follows that
(5" 1 0)
>
this
formula
may be used
measured plume gas mixing
nS02/n other
73
"
it
(5-9)
,
plume
Boltzman's constant (1.38 x
number
i„
a
the ideal gas law (Halliday
i
each spectrochemical measurement of
n other
plume
S0 2
Tp ume + n other k Tplume= n air k T^
absolute temperature of the surrounding
Mother
So,
the above results, requires that
,
k
in the analysis.
of the
other
absolute temperature
when combined with
respectively,
P
the
where n So2, n olher n air are the number densities of
air,
(5-8)
i
nso2 k
surrounding
roughly
is
Ppiume equals Pa T to a good approximation. Pp
Assuming
respectively.
(5-7)
,
2
pressures
partial
1
1
ppbme v pUime^ 65Pa.
and so may be neglected
quasi-hydrostatic system in which
= 0ms"
to
ratio
(5-11)
may be
estimated and compared with the in situ
Spectral
n
Sample
S0 2
measurements. Here are the
nS02/ n other
]
ther(m~'
results:
Mixing Ratio (ppm)
Number
535
1.967 xlO
554
1.968 xlO
572
1.967
Assumed
Values:
5.3.
3.43 xlO
25
xlO
1.32
25
Mixing
So the average
S0 2
25
m"
3
ratio is
measurements which are near 400 ppm.
the
4
xlO"
132
620
T
* air
T
'plume
300K
410K
365± 200ppm, where
"air' air * plume
1.968 xl0
from observations of stack#l
the standard deviation about the average.
work
343
4
ratio results
mixing
-4
6.20 xlO"
flair
2.69 xlO
Table
25
at 0.2
m~
3
km.
the uncertainty
This agrees very well with the
25
is
simply
in situ direct
Because we have only three measurements to
with, the statistics are not overwhelmingly compelling, but the consistency between
two independent measurements
is still
quite promising.
74
CONCLUSIONS AND RECOMMENDATIONS
VI.
This
thesis
detecting and quantifying
effort entailed
of
S0
2
demonstrated
research
S0
2
conclusively
at different
First,
now
laboratory comparison
S0
2
spectral features.
S0
2
in
from a coal-burning power
plant.
spectra
The curves
Direct in situ
one of the plant's effluent plumes was known
400 ppm. Detailed analysis of NUVIS data from the plume yielded
ratio
UV
Second, the results of the laboratory studies were
to analyze real field data
chemical measurements of the
mixing
of
be used for any subsequent spectrochemical measurements of SO,
column abundances by NUVIS.
employed
capable
These spectra were used to calculate
concentrations.
least-squares fitted curves of growth for four different
of growth can
is
emissions in the form of effluent smokestack plumes. This
two separate phases of research.
were recorded
NUVIS
that
a"
to
be
mean plume SO,
of 365 ± 200 ppm, in excellent agreement with the in situ value.
Based upon an assessment of the signal-to-noise
ratios in representative field data,
20
2
and the laboratory observation thatS0 2 column abundances must be > 10
m" for
NUVIS
to detect the
S0
2
signatures that
lower mixing ratio limit for
S0 2
we have
detection in the
approximately 70 ppm. This indicates that
S0
2
NUVIS
we
field.
can estimate the practical
We
is
to
S0
2
estimate that
it
is
should not have detected quantifiable
emissions during the 1999 observations of the power plant,
plumes were chemically scrubbed down
That
analyzed,
mixing
ratios
when
on
all
of the plant's
the order of 40
ppm.
indeed the case, further demonstrating the consistency of the experimental and
analytical procedures developed during this thesis research project.
75
A
number of
steps could be taken to
enhance the calibration and analysis
techniques outlined in this thesis and to improve the overall sensitivity of the
hyperspectral remote sensing instrumentation.
be performed. 2)
range of
S0
2
A new set of calibration curves
concentrations.
location for each scene
upon
when
3) Repetitive
A new
1)
UV
wavelength calibration should
should be developed, involving a wider
measurements should be made
in the field in order to
at
each
check for consistency and to improve
the signal-to-noise in the data. 4) Further develop and extend the analysis processes
presented in Chapters
IV and
V
to include additional spectral features, thereby providing
a larger statistical set of measurements.
devices to
measure
temperature,
GPS
As
all
applicable
5)
NUVIS
should be augmented with peripheral
environmental
factors
(i.e.
a follow-on to this thesis work,
quantify the
more rigorous
analysis should be conducted
column
densities
ENVI
on
software to more
through a more refined application of
this
This process should be further developed through continued field and
calibration.
laboratory experimentation and should also involve corrections for
scattering
pressure,
generated position) in conjunction with each field measurement.
the existing field data utilizing the algorithms provided by the
reliably
barometric
and the contributions due
to other
UV
foreground
chemical effluents.
This study has provided an account of
NUVIS
progression to date.
It
also
demonstrated that sulfur dioxide does exhibit a characteristic ultraviolet absorption
signature.
It
also indicated that the existing instrument
signatures at concentrations of tens of ppm at long ranges (see
76
was unable
Appendix
to
C).
detect
S0
2
The sensor's resolving power
is
limited by the eight bit processing capability of
the detector, a Pulnix
CCD
Appendix C) can most
likely be attributed to optical
mirror and diffraction grating.
most
likely
camera.
Instrument anomalies (also briefly detailed in
components such as the scanning
The anomalies experienced with the processing PC can
be attributed to the instrument's software.
In either case, further testing is
required to fully determine the extent of such anomalies.
Further improvements to the existing instrument could be
made
to correct
of these problems but a more beneficial approach involves construction of a
instrument, incorporating the lessons learned from
resolving
power and
NUVIS.
It
operational anomalies and glitches.
new
should have higher
greater sensitivity utilizing a digital twelve/sixteen bit
more robust processing power and an improved
many
electronic design that
CCD
camera,
minimizes spurious
We recommend that further improvements to NUVIS
be made, and that a new, more capable instrument be developed and tested with the
ultimate goal of
someday
fielding a similar airborne
77
UV spectral imager.
THIS
PAGE INTENTIONALLY LEFT BLANK
78
APPENDIX A. NUVIS FIELD CHECKLIST
Field experiments involve a great deal of logistics to accomplish a successful mission.
Ensuring that
NUVIS
and
support components arrive
all its
concerns during such operations. The following checklist
for
NUVIS
mission
field operation.
The
list is
hand may require modifications.
at
designated packing/shipping material:
field
deployment of NUVIS.
recommended
•
CASE
1)
by no means
i.e.
at the site is
is
provided as a material guide
all-inclusive and,
foam, bubble wrap etc..) are dedicated to the
Each shipping case
will first
be
listed
#1:
CPU
(IDL
HASP may remain
Max-410&
[Notes:
one
in place).
Foam
crate.
1)
fit
It is
has been specifically cut to specification for packing purposes.
nicely into place to allow shipping of the
recommended
that only the
•
CASE
1)
installed).
tripod (with wrench).
#3
Spare PC.
79
Foam
pieces are
CPU, monitor and Max-410 remain
foam be removed and
#2
NUVIS
foam
JS auxiliary connector terminator.
the crate during field operation.]
CASE
followed by the
contents of each crate:
labeled and
•
depending on the
Five shipping crates (including pre-
2) Monitor (monitor sits firmly against right wall of the shipping crate with
3)
only one of many
that these
components remain
in
in
CASE #4
1)
25 foot
NUVIS
NUVIS
2) 8 foot
umbilical cable.
umbilical cable.
3) Extension cords (quantity
is
mission dependent).
4) Category 5 cables, blue in color (2 jumpers, 2 network).
5)
Monitor
6)
COM1
7)
UPS.
8)
Power
cable.
to
Max-410
interface cable.
strip.
9) Surge suppressor.
10)
Mouse.
1
Tool
1)
kit (screwdrivers, utility knife,
wire cutters,
pliers,
Allen wrenches,
etc..
.).
12) Miscellaneous items (paper towels, rope, lens paper, gloves, plastic, etc.).
CASE #5
1)
Spare stepper motor.
2) Spare Pulnix
3) Spare
camera
& Lens.
AZ encoder.
4) Spare computer keyboard.
5) Spare mouse.
6) Spare
JAZ
7) Spare
IDL HASP.
8)
JAZ
drive.
Cartridges.
9) Diagnostic tools, Fluke, meter
10) Repair items (tape, cable
1
.
1)
ties,
and
oscilliscope.
soldering iron, solder, sponge, spare wire, etc.).
Documentation (software documentation, Keithly,
12)
Backup
13)
Gas
US
Digital
discs (operating system, software, hard drive, etc.).
test cells.
14) Spare camera.
80
and Max-410
docs.).
APPENDIX B. NUVIS OPERATIONAL CHECKLIST
SETTING UP THE SYSTEM
Remove tripod from case and set in place. Use key (stored at the base of the
move instrument locking mechanism from its slot by unlocking the setscrew.
#1
tripod) to
Remove
#2
instrument from case and slide instrument into place on the tripod
mounting plate. Once instrument
screw. (Key may be restowed.)
Remove
#3
protective cover
is
from
cover not be removed until system
protection to the
filter
in place, reset locking
is
optical sight.
It is
mechanism and
recommended
that the
fully operational. This will provide
window
maximum
window.
Ensure computer system components are accessible. Currently
#4
retighten set-
it is
recommended
computer components and stepper motor remain in the carry case with only the
two access covers removed. This provides maximum protection and ease of mobility
while all components remain readily accessible.
that the
Connect the
#5
UPS
to the appropriate source
of power (115
VAC outlet or field
generator). Connect the computer, monitor, Ethernet hub and stepper motor controller
power cables to each component and to the UPS. Ensure that the voltage switch on the
computer
is set
to
1
15 volts.
#6
Connect monitor cable, mouse and keyboard
#7
Now connect the remaining cables including NUVIS
category five cables. Connections are to be
NUVIS
instrument as well as to the
made
to the
PC.
umbilical cables and blue
to the varios cable
connectors on the
AMS 410 and rear panel of the PC.
Figure B.l
illustrates these connections.
Figure B.l.
NUVIS
umbilical, instrument,
81
PC and AMS-41U
cable connections.
GETTING UP AND RUNNING
(You should now be ready
Turn on
#1
to
power up
the monitor and
the system.)
PC. (To turn on
PC
ensure switch on back panel
is set
to
the on position and then push, turn and release blue switch on the front panel of the PC.)
#2
Wait
for the
(Windows 95)
Once
#3
is
is
display and
indication that the operating system
initial
some
functioning. This will take
this is
If networking
BIOS
time.
complete, the standard 'Enter Network Password' screen will appear.
desired, enter appropriate password, otherwise press 'escape.'
case, the standard
Windows
In either
screen with short-cut icons will next appear.
DO NOT LAUNCH THE NUVIS.EXE PROGRAM
UNTIL COMPLETING STEPS SIX THROUGH EIGHT:
#6
Turn on the stepper motor
#7
Once
this is
cover from the
controller.
accomplished and you are ready
NUVIS window.
studs, install the cell,
(If the
and secure the
4inch
cell in
to take data,
remove
the protective
test cell is required, install the
mounting
place with the stud nuts.) Align the
instrument as required.
The NUVIS.EXE program
#8
before launching this program.
is
memory
intensive. Close all other applications
It is
actually
recommended
that
NO OTHER
APPLICATIONS BE OPENED PRIOR TO LAUNCHING NUVIS.EXE.
applications had been previously opened,
This will dedicate the
restart.
are
now
ready for
NUVIS
PC
to
it
might be wise
NUVIS
to shut
operation only.
Once
down
If a
the
this is
number of
PC and
complete, you
operation.
INSTRUMENT OPERATION WITH NUVIS.EXE
#1
Launch
#2
Upon
the
NUVIS.EXE
program.
execution of the program, four windows will appear on the monitor prior to
the presentation of the actual operational window. These
B.l.
82
windows
are detailed in Table
SEQUENCE
WINDOW STATEMENT
DESCRIPTION/
1)
IsMax-410Ready?
This
Mirror
OK.
Once
REQUIRED ACTION
2)
is
aligned.
is
As
a reminder.
long as
Max-410
previous step was completed, the
is
on, enter
Max-410
ensured communications with the instrument and
centered the mirror to position zero. This statement
confirms that action. Enter
OK.
OK
3)
Blank
Blank Window. Enter
4)
Based on memory, # frames
This indicates the number of images that
XXX.
is
(i.e.
613)
in
RAM.
number
If this
application
proceeding,
was most
#3
Once
.
Window
is
is
likely previously
available.
sequence upon execution of the
the above sequence
is
Off
;
Mirror Position
-ooqgo
j»
Scan Settings
OQ&;y
start
«»
Step
'A
ilool
;
'
Tl
ages./
Snap
Figure B.2.
NUVIS
operational
83
Otherwise, Enter
nearly complete.
.
t
opened. Before
OK.
window of Figure B.2
Photomuit. C3ain
ocr& ;v s ,,
be stored
NUVIS.EXE program.
completed, the operational
obtained and preparation for an actual data collection
may
below 400, another
best to restart the sequence to ensure
it is
enough memory
Table B. 1
is
Window.
is
any problems are encountered prior to this point, it is best to exit the program and
restart the process. If memory was initially low, this may require a complete system
(If
shutdown and
sequence, the
program screen locks up during portion of the startup
command to end the task may not be available. If this occurs,
If the
restart.
'ctrl-alt-del'
do not attempt to use this command a second time to restart your computer. This method
of rebooting (once the NUVIS.EXE program is executed) may cause damage to certain
Windows operating files. As a last resort, if the program locks up, try to turn the system
power off manually and
restart the system.
Normal system shutdown
of this procedure. Likewise, when exiting the
is
detailed in the
NUVIS.EXE
program, it is best
perform a proper system shutdown every time. This requires using the EXIT button
appearing in the operational control window.
final steps
to
#4
Set the Photomultiplier Gain.
begin. Gains
may be
from
set
to 10
It is best to start at a lower value (5.0 or below)
by keyboard entry or by using the arrows and
to
mouse.
Assuming
#5
that the instrument is pointed at the appropriate target, a quick-look at
the sensor's active digital signal
may be
obtained by activating the
brighter pixels correspond to higher digital signals. Keeping in
should not be saturated, the image intensifier should be adjusted
signal
the
is
displayed.
VIDEO
keep in mind
#6
that the screen display
an appropriate
does not represent the image scene directly as
Prior to collecting data, ensure that there
JAZ
cartridge
is
is
actually inverted
from the
Two
initially
operational
is
enough memory available on the
installed for data collection.
Mirror Position (center position) may be
number should
#8
until
itself.
hardrive and/or ensure that a
#7
The
intensifier
viewed from behind the instrument. The screen display
scene
button.
that the detector
must also first be turned ON prior to selecting
of the two modes described in the steps that follow. Also,
The image
button or either
VIDEO
mind
set
through the active window. This
be zero since the mirror was centered during the startup sequence.
modes
exist
with
NUVIS:
(1)
SNAP
and
(2)
SCAN. The
following information should be reviewed before selecting either one of these modes:
(1)
SNAP:
Images/Snap. This setting determines the successive number of SNAPS to be performed.
NUVIS SNAP,
SCAN:
During a
(2)
the mirror remains centered on the position of step #7.
84
The
Start
and Stop
settings only apply to the
SCAN function.
Setting mirror Start
and
Stop positions automatically resets the number of steps through which the mirror will
scan.
The number of steps can only be controlled by the
maximum
failures
may
of 35 1 steps
due
to
memory
A NUVIS
at
may
be utilized but steps in excess of 200
settings.
A
cause system
also be selected. This defines the
number of SNAPS
to
each repositioning of the mirror
"SNAP"
simply a 640 x 480 byte array of data achieved through the
is
Each individual
instrument control program.
hard drive or internal
obtained.
and stop
allocation faults.
The number of Images/Step may
be performed
start
One frame
JAZ
SNAP may
drive and will register as a
simply a two-dimensional
is
set
be written to either the PC's
300KB
file if
only one frame
is
of data representing the spectral
(wavelength) and spatial (vertical) components of the representative scene. Although this
frame of data
is
not really an image, the operational
SNAP
Selecting multiple images per
NUVIS window refers to
it
as such.
100), will increase the size of the data file as
(i.e.
100 different 640x480 byte arrays are written, one by one, to the either of the available
mode of operation during an
view of the instrument is uniformly
focused on a standard gas test cell and not a true image scene. These multiple SNAPS
may be averaged during the analysis process to improve signal to noise. A file containing
100 "images per SNAP" is normally 30MB in size. With these settings, a data file
smaller than 30MB indicates a problem with that particular collection.
drives as a complete data
when
instrument calibration
A NUVIS "SCAN"
being targeted.
window
(see
By
is
file.
This
is
the preferred
the entire field of
a bit different.
It is
the preferred
selecting the mirror stop
Appendix B)
and
the operator sets the
number of
through which the mirror will scan. The mirror, which
during startup,
scan.
same
step
It
is
moved
may
also be set
start/stop values
positions.
(i.e.
number of steps
by the operator.
scan line numbers)
at
Approximately
300KB
position zero
and begins to
magnitude of the
The number of images per
Typical settings for a routine
of -50 to +50, which
CPU
(i.e.
centered
until reaching the
stop position) at the end of the scan.
SCAN
include
corresponds to 101 different steps or mirror
of data storage
demonstrated by each of these operational modes,
requiring that the
steps
is first
is
instrument control
to the initial offset position (i.e. start position)
rotates through the specified
offset position
method when an image scene
start positions in the
is
required
NUVIS
for
each
step.
As
data files are quite large
be fully dedicated to the reading/writing of this data once a data
collection has been initiated.
NUVIS.EXE program that
is
This
is
a major restriction imposed by the current
written in Visual Basic.
85
#9
Select the appropriate mode:
#10
Once
the
mode
SNAP
or
SCAN.
selected and executed, the reading and writing of sensor data
is
amount of time (30 to 120 seconds in most cases). The program
which to store the data. The customary file extension used in
a '.nuv' extension. The file will be saved in JPEG format.
will take a considerable
will ask for a file
all
NUVIS
#11
files is
It is
command
highly
is
the market.
is
in
recommended
issued.
and
limitations
to
name
It is
that
best to be patient, as the
not as streamlined as
As
program has
that only
own
inherent
software applications on
one function
at a
time
failures.
previously mentioned, to exit the system use the
The TEST ARC and
its
many of the commercial
System operators should ensure
reduce the possibility of program
#12
one operation be fully completed before the next
EXIT button.
LOAD IMAGE buttons are not functional.
86
is
requested
APPENDIX C: RANGE EFFECTS AND NUVIS ANOMALIES
Data presented
burning power plant.
A
principle
were taken from the 1998
in this section
Figure C.l illustrates the scene in a principle components format.
components transform on bands 500
640 reduces the data
to
which highlights the differences between the clean and
the
plumes
is
field trip to the coal
done with eight regions of
illustrated in Figure C.2.
The
interest at a
dirty plumes.
A
to this
image,
brief analysis of
range of 1.5 km. These regions are
ratioed relative absorption spectra are displayed in Figure
C.3, demonstrating not only the effects of range on the detected spectra but also providing
a comparison of the signatures of the scrubbed and unscrubbed stacks at this range.
A
technique.
very preliminary
Because
the
analysis of these data
absorption
lines
unambiguously detected by the instrument
improve the detection of
S0
2
signatures.
at
for
was conducted using a matched
the
clean
range, this approach
The matched
filter
spectra for both the sulfur dioxide and the sky background.
are displayed in Figures C.4 and C.5.
indicated
by the dark regions
unscrubbed stack
matching
in the
at a
in
the
range of 1.5km.
modeled sky spectrum
stacks
were
was
not
A
being
tried in order to
technique utilized in scene
The
results
of this technique
In Figure C.4, apparent regions of
display.
filter
S0
2
are
This particular scene includes the
similar matching technique resulted in pixel
as illustrated
87
by the
lighter regions
of Figure C.5.
3 Stacks - 1.5 km
Principal Components
Cyan: PC1
Regions of Interest
at 1.5
Red PC2
km
Stack 1A
Stack 1B
Stack 1D
Stack 2A
Stack 2B
Stack 3A
Stack 3B
Sky
Figure C.l. Principle components image
Figure C.2. Principle component
taken from bands 500-640 of the original
with regions of interest superimposed.
data
set.
Principal
component
1
is
Color codes correspond
encoded
in
as cyan (green/blue guns), principal
compent
2
is
encoded
as red.
88
Figure C.3.
to the
1
image
mean
spectra
540
500
580
Spectral
620
Sample Number
Figure C.3 Relative Absorption Spectra for regions of interest at 1.5 km.
Figure C.4.
S0
2
matched
filter.
Scale
indicates relative match.
Figure C.5. Sky matched
filter.
89
THIS PAGE INTENTIONALLY LEFT
90
BLANK
OF REFERENCES
LIST
Aitken, P., Visual Basic 5
Programming Explorer, The
Atkins, P.W.. Physical Chemistry, Fourth Edition,
York,
New York,
Coriolis Group, Inc., 1997.
W.H. Freeman and Company,
New
1990.
Cleary, D.D., Private Conversation, (1998).
Durkee, P.A.,
"MR 3480 - Thermodynamic and
Radiative Processes in the Atmosphere",
http://www.met.nps.naw.mil/~durkee/MR3480/MR3480.htm 1999.
,
Eisberg, R., and Resnick, R.,
Particles,
Quantum Physics of Atoms, Molecules,
Second Edition, John Wiley and Sons,
Finlayson-Pitts,
B.J.,
Pitts,
J.N.
Jr.,
Atmospheric
Experimental Techniques, John Wiley and Sons,
Haliday, D., Resnick, R., Walker,
John Wiley and Sons,
Hecht,
Optics,
E.,
J.,
Chemistry:
1985.
Fundamentals
New York, New York,
and
and
1986.
Fundamentals of Physics Extended, Fifth Edition,
New York, New York,
Third
Solids, Nuclei,
New York, New York,
Edition,
1997.
Addison- Wesley
Longman
Inc.,
Reading,
Massachusetts, 1998.
Hooks, T.A., "Design Construction and Operation of the Naval Postgraduate School's
Ultraviolet
Imaging Spectrometer (NUVIS)", Master's Thesis, Naval Postgraduate
School, Monterey, California, (1997).
Johnson, E.O., "Design Development and Testing of an Ultraviolet Hyperspectral
Imager", Master's Thesis, Naval Postgraduate School, Monterey, California, (1996).
Kramer,
H.J.,
Observations of Earth and
Its
Environment, Third Enlarged Edition,
Springer-Verlay, Berlin, Germany, 1996.
Lenoble,
J.,
Atmospherics Radiative Transfer, A. Deepak Publishing, Hampton, Virginia,
1993.
MacMannis, A.R., "The Design of the Naval Postgraduate School's Ultraviolet Imaging
Spectrometer (NUVIS)", Master's Thesis, Naval Postgraduate School, Monterey,
California, (1997).
91
Multispectral Users Guide, Department of Defense, August, 1995.
Skoog, D.A., Principles of Instrumental Analysis, Third Edition, Saunders College
Publishing, Philadelphia, Pennsylvania, 1985.
Strickland, D.J.,
and Bishop,
Computational Physics
J.,
Remote Sensing of Tropospheric
S0 2 from
Space,
Inc., Fairfax, Virginia, 1999.
and Johansson, S., Spectrophysics: Principles and Applications,
Spnnger-Verlay, Berlin, Germany, 1999.
Thome,
A., Litzen, U.,
United States Geological Survey(USGS), "Description:
http://vulcan.wr.usgs.gov/Monitoring.
Using IDL, IDL Version
5.0,
.
.
COSPEC
scriptions/description
Research Systems
92
Inc.,
Measurements",
COSPEC.html
Boulder, Colorado, 1997.
1999.
INITIAL DISTRIBUTION LIST
1.
Defense Technical Information Center
8725 John J. Kingman Road., Ste 0944
Fort Belvoir,
2
VA 22060-6128
Dudley Knox Library
Naval Postgraduate School
411 DeyerRd.
Monterey, C A 93943-5 101
2
David S. Davis Code PH/DV
Department of Physics
Naval Postgraduate School
Monterey, CA 93943-5101
2
4.
Richard C. Olsen Code PH/OS
Department of Physics
Naval Postgraduate School
Monterey, CA 93943-5 101
5
5.
Chairman, Code
2.
3.
PH
2
Department of Physics
Naval Postgraduate School
Monterey, CA 93943-5101
6.
Curncular Officer, Code 34
1
Engineering and Technology
Naval Postgraduate School
Monterey, CA 93943-5107
7.
Tamir Elias
1
U.S. Geological Survey
Hawaiian Volcano Observatory
Hawaii National Park, HI 96718
8.
Willie Williams
1
14675 Lee Road
Chantilly,
VA 20151-1715
93
9.
Lieutenant Michael Rigo
1
4675 Lee Road
Chantilly,VA 20151-1715
10.
Greg Pavlin
1
SITAC
1
1781 Lee Jackson Memorial Highway, Suite 500
Fairfax,
11.
VA 22033-3309
Mr. Stephen A. Marino
2
3215 Concord Drive
New Castle, PA
16105
12.
Major Stephen Finney
346 Summerwood Lane
Castle Rock, CO 80104
13.
Dr.
.
.2
David D. Cleary
th
2901 W. 96 Street
55431
Bloomington,
2
MN
94
60
^2225
G/02 22527-200
>m
d
P