REMOTE SENSING APPLICATIONS
FOR ACID DEPOSITION
-------
July 1988
REMOTE SENSING APPLICATIONS FOR ACID DEPOSITION
by
Lynn K. Fenstermaker
Environmental Research Center
University of Nevada, Las Vegas
Las Vegas, Nevada 89154
Cooperative Agreement No. CR814002 01
Project Officer
Thomas H. Mace, Ph.D.
Advanced Monitoring Systems Division
Remote and Air Monitoring Branch
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada 89193-3478
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA 89193-3478
-------
NOTICE
The information in this document has been funded (wholly or in part)
by the United States Environmental Protection Agency under cooperative
agreement number CR814002-01 to the Environmental Research Center,
University of Nevada, Las Vegas. It has been subjected to Agency review
and approved for publication.
Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
-------
ABSTRACT
This report presents manuscripts and summaries resulting from a
special session sponsored by the U.S. Environmental Protection Agency. The
session entitled Remote Sensing Applications for Add Deposition was held
March 18 and 19, 1988, in conjunction with the Annual American Congress on
Surveying and Mapping, and American Society for Photogrammetry and Remote
Sensing (ACSM/ASPRS) Convention in St. Louis, Missouri. Presentations and
panel discussions centered around the use of remote sensing technology for
the assessment of acid deposition impacts to vegetation, building
materials, and surface waters. The utilization of passive and active
systems such as photography, multispectral scanners, and laser
fluorosensing were reported for acid deposition assessments. The consensus
of the participants at the close of the session was that remote sensing is
an invaluable tool for large scale or regional acid deposition assessments,
and that this tool is not being utilized by those who could benefit the
most, i.e., process scientists and modelers. Some recommendations for
improving the utilization of remotely acquired data included organization
of a conference/workshop for remote sensing and process scientists, and
initiation of basic research projects which require cooperation and
coordination between remote sensing and process scientists.
-------
CONTENTS
Page
Abstract iii
Acknowledgment . . . v
1. Introduction 1
2. Conclusions and Recommendations 4
3. Section 1 - Summaries of Panel Discussions 8
Panel discussion on remote sensing applications
for vegetative damage assessment 9
Panel discussion on building materials effects
assessment 13
Panel discussion on remote sensing applications
for assessment of water quality degradation due
to acid deposition 17
Panel discussion on new applications of remote
sensing technology to acid deposition assessments .... 21
4. Section 2 - Session Manuscripts 26
Remote detection of acid mist and ozone effects
on conifer and broadleaved vegetation . 27
A preliminary evaluation of the use of TM imagery
for the study of forest decline in the Southern
Appalachians 30
The use of quantitative remote sensing techniques
to assess forest decline damage in Vermont 37
Remote mapping and monitoring of forest damage 56
Use of high spectral resolution sensors to detect
air pollution injury in conifer forests 72
Result of an experiment using the enviro-pod
camera system to inventory building surface
materials in Cincinnati, Ohio 86
The detection of acid rain damage to building
stone using spectral reflectance measurements 102
Remote detection of dissolved organic matter (DOM),
aluminum, and hydrogen ion using laser-induced
fluorescence (LIF) 122
Application of remote sensing techniques for
estimating spatial variability of dry deposition
of acidic pollutants 149
IV
-------
ACKNOWLEDGMENTS
The assistance of the contributing authors and session moderators is
gratefully acknowledged. Their expertise and participation in the Remote
Sensing Applications for Acid Deposition special session were critical to
the success of this project. The willingness of ACSM and ASPRS to include
this session in their Annual Convention is also gratefully acknowledged.
Finally, the peer review of this document by Dr. John Estes of the
University of California, Santa Barbara, Ms. Janet Degner of the University
of Florida, Mr. Fred Luce of the Pennsylvania State University, and Dr.
Thomas Mace of the EPA Environmental Monitoring Systems Laboratory is
sincerely and appreciatively recognized.
-------
INTRODUCTION
A special session on Remote Sensing Applications for Acid Deposition
was organized by the Environmental Research Center, University of Nevada,
Las Vegas under cooperative agreement to the U.S. Environmental Protection
Agency's Environmental Monitoring Systems Laboratory, Las Vegas, Nevada.
The session was held in conjunction with the Annual American Congress on
Surveying and Mapping, and American Society for Photogrammetry and Remote
Sensing (ACSM/ASPRS) Convention in St. Louis, Missouri, March 14-19, 1988.
Papers discussing the use of remote sensing technology for acid deposition
assessments were presented Friday morning, March 18. A series of panel
discussions covering four topical areas, vegetation damage, surface water
quality, building materials deterioration, and new applications and
technologies, followed the presentations. The panel discussions occurred
Friday afternoon and Saturday morning, March 18 and 19. The complete
schedule for the two days is presented on the following two pages. Two of
the authors, D. Williams and D. Marks, to their regret had to cancel their
presentations. Mr. Williams, however, did submit a manuscript for
inclusion in this report. Two other researchers in addition to the session
presenters submitted manuscripts for this report due to their strong
interest in the topic area. They are Dr. Siamak Khorram et al from the
North Carolina State University and Dr. Walter Westman from the University
of California, Berkeley.
This report contains two major sections. Section 1 includes
summaries of the panel discussions prepared by each panel leader. The only
summary missing from this section is the summary of the wrap-up panel
discussion. The discussion from the wrap-up session is provided in the
Conclusions section. Section 2 contains the manuscripts written by the
session presenters and other researchers. The two sections are organized
in the order of the following subtopics: vegetation damage assessments,
building materials assessment, water quality, and new applications.
The objective of the session and this report is to provide the EPA
with an up-to-date survey on the use of remote sensing data for acid
deposition assessments, and to clearly demonstrate that remote sensing is a
viable and efficient tool for environmental assessments. The questions
which were answered by this session are: 1) what capabilities have been
developed within remote sensing which are viable for acid deposition
assessments; 2) what research is being performed at this time in this area;
3) where is more research and development needed; and 4) what role should
remote sensing technology play in future acid deposition assessments.
-------
SPECIALTY SESSION ON
REMOTE SENSING APPLICATIONS FOR ACID DEPOSITION
Chairman: Dr. Thomas H. Mace
Friday, March 18, Cervantes Convention Center
0900-0920 Results of an Experiment Using the Enviropod Camera
System to Inventory Building Surface Materials in
Cincinnati. OH.Ellefsen, R.A.*
0925-0945 The Detection of Acid Rain Damage to Building Stone
Using Spectral^Reflectance Measurements.
Kingston, M.J.
0950-1010 Remote Detection of Forest Damage. Rock, B.N.,
J.E. Vogelmann, and N.J. Defeo
1015-1035 The Use of Spectral Reflectance and Laser Induced
Fluorescence Measurements for Assessing Vegetation
Subjected to Acidic Deposition in Vermont.
Williams, D.L., D.W. Case, and E.W. Chappelle
1040-1100 Use of High Spectral Resolution Sensors (PIDAS and
AVIRIS) to Detect Air Pollution In.iurv in Conifer
Forests. Ustin S.L. , B. Curtiss, and S.N. Martens
1105-1125 Detection of Effects of Lake Acidification bv Remote
Laser Fluorosensinq. Philpot, W.D. and A. Vodacek
1130-1150 Snow Surface Energy Balance Calculations Over
Rugged Terrain. Marks, D. and J. Dozier
1155-1215 The Role of Remote Sensing Techniques in Estimating
the Spatial Variability of Dry Acidic Deposition.
Fisher, L.T. , R.T. McMillen, B. Levinson, and M. Hewitt
Denotes session speaker
1330-1430 Panel Discussion on Remote Sensing Applications for
Assessment of Vegetative Damage Due to Acid Deposition,
Panel Leader: Dr. John Brockhaus
1430-1530 Panel Discussion on Remote Sensing Applications for
Assessment of Water Quality Degradation Due to Acid
Deposition.
Panel Leader: Dr. Michael Bristow
-------
1545-1645 Panel Discussion on Remote Sensing Applications for
Assessment of Building Material Damage Due to Acid
Deposition.
Panel Leader: Dr. Richard Ellefsen
Saturday, March 19, Sheraton St. Louis Hotel
0830-1000 Panel Discussion on New Applications of Remote Sensing
Techniques for Acid Deposition Assessments.
Panel Leader: Dr. Lee Williams
1015-1130 Wrap-up Panel Discussion for the Specialty Session on
Remote Sensing Applications for Acid Deposition, and
a Summary of the Previous Panel Discussions.
Panel Leader: Dr. Thomas Mace
-------
CONCLUSIONS AND RECOMMENDATIONS
This section summarizes the key lessons learned during the two days
of the specialty session as discussed at the wrap-up panel discussion. The
participants clearly indicated that there is an enormous amount of remotely
acquired data available and forthcoming for environmental assessments.
Research to date has shown that it is possible to detect and map areas of
pollutant damage with passive and active remotely acquired data, and that
this data should be extremely valuable for assessment purposes. However,
due to a lack of knowledge and communication, process scientists (i.e.,
biologists, foresters, soil scientists, etc.), and modelers have not
incorporated this data into their inventories and models. It was the
strong belief and recommendation of each participant that the non-remote
sensing scientific community should be apprised and educated on the extent
and significance of this data. Full utilization of remote sensing data
would significantly advance modeling and assessment efforts.
Information exchange between the remote sensing community and process
scientists would actually benefit both groups. The remote sensing
scientists do not know the acid deposition processes well enough in some
areas to recommend an appropriate sensor and/or spectral range for damage
assessment. The process scientists do not understand the tremendous
sophistication and capabilities of remote sensing measurement technology.
The combined expertise of these two groups would advance the science of
damage assessment more rapidly and cost effectively than each group
conducting separate research. Costly field investigations could be limited
to a few diverse areas to provide a ground truth, i.e., accuracy
measurements, for remotely acquired data of regional scale. If remote
sensing data were iteratively incorporated into models, the models could
potentially improve significantly. Remote sensing could provide a timely
assessment of changes to a particular area of interest, and perhaps provide
decision-makers with advance warning of a potential adverse environmental
impact.
Several methods for providing this kind of interchange and initiation
of multidisciplinary research were suggested by the participants, and are
1isted below.
1. Make use of EPA Research Cooperatives to promote communication,
cooperation, and close interaction between remote sensing
scientists and process scientists, including modelers.
2. Recommend that when EPA or other agencies fund process oriented
research such as greenhouse studies they include remote sensing
(handheld spectrometer studies) as part of the study. Not only
-------
would this provide closer interaction between the remote
sensing community and process scientists, it would provide an
improved opportunity to identify which spectral bands afford
the best measure of a variable of interest.
3. Organize a conference for process scientists and remote sensing
scientists on acid deposition. This conference would be
different from any other in that initial discussions would be
scheduled to provide the audience with background information
to understand the papers which would be presented. At the close
of the paper presentations, a series of round table discussions
would provide an atmosphere for iterative exchanges of
information.
Canada has a conference of this type where the "users" meet with
system builders. It was suggested by some of the participants that the
National Acid Precipitation Assessment Program (NAPAP) should be approached
to organize this type of conference.
VEGETATIVE ASSESSMENTS
Remotely acquired data has been used to assess vegetative health
since the development of infrared photography during World War II (1), when
color infrared film was used to differentiate camouflaged targets from the
surrounding vegetation. The use of multispectral scanner data is also well
developed for vegetative assessments. The spatial mapping of damaged areas
and the detection of temporal changes are now routine analyses. Present
research is endeavoring to specifically identify the agents promoting
vegetative damage and map the areas impacted by a particular pollutant,
disease, or pest. For example, in the eastern U.S. it is relatively easy
to map areas with acid deposition-induced vegetation damage. However, in
the west, acid deposition impacts are difficult to map. There are two
primary reasons for this. One reason is that western vegetation has not
been exposed to acid deposition at the concentrations and duration that
eastern vegetation has, and therefore, very little damage is evident. The
other reason is that western vegetation may be impacted more severely by
ozone.
Another practical application of remote sensing is the mapping of
areas potentially sensitive to a particular pollutant, especially acid
deposition. Soils within a watershed play an important role in
neutralizing or buffering the impact of acid deposition. Remote sensing is
used to map native vegetation, and hence soil types. Soils with low
buffering capacities are not capable of neutralizing acid deposition due to
their lack of basic cations (nutrients) and already low pH. The vegetative
species which naturally habitate these soils are visually and spectrally
distinct. Therefore, it would be relatively easy and inexpensive to
produce maps of large areas depicting watersheds potentially sensitive to
acid deposition impacts.
-------
The primary recommendation from the panel to improve the application
of remote sensing to vegetative assessments is the promotion and funding of
basic research projects. The type of research needed would result in the
identification of distinct narrow spectral bands to measure particular
physical' processes and conditions. Some methods for achieving this include
the use of handheld spectrometers in greenhouse studies and access to
AVIRIS (advanced visible and infrared imaging spectrometer), an airborne
scanning system. In particular, the panel recommended the purchase of
handheld spectrometers for the EPA Research Cooperatives.
BUILDING MATERIALS EFFECTS
Initial experiments for inventorying building materials and assessing
building stone decay have been performed by EPA and U.S. Geological Survey.
The results of these experiments now need to be modified for and conveyed
to state and local officials as well as to the National Park Service, to
develop a routine monitoring program. Part of the modification should
include procedures for combining inventory and damage effects into one data
base. These data could then be used to assist in developing models which
will use prevailing weather patterns, and pollutant type and transport data
to predict potential decay of building materials. Other areas which the
panel believed to warrant further investigation include: experimentation
to transfer knowledge gained from U.S. Geological Survey spectrometer
research to facilitate potential usage of AVIRIS data for building material
assessment; and perform additional urban building material inventories to
facilitate extrapolation of the urban terrain zones developed by Ellefson
(2) to other cities.
WATER QUALITY EFFECTS
Research to date has shown that the analysis of remotely sensed data
can be related to optical properties of water. However, studies have not
conclusively shown high predictability of the acid deposition-related
parameters aluminum and pH from the analysis of remotely sensed data. The
remote sensing instrument which has been most effectively used to date is
the laser fluorosensor. Laser fluorosensor data have been successfully
correlated to surface water quality parameters such as dissolved organic
carbon, water clarity (optical attenuation), and chlorophyll a. Only
partial success has been made in correlating fluorescence data with pH and
aluminum. Research is being conducted in this area by EPA's Environmental
Monitoring Systems Laboratory-Las Vegas and the Cornell Laboratory for
Environmental Applications of Remote Sensing (CLEARS). Results of their
research to date clearly indicate that laser fluorosensor technology can
detect changes in surface water quality and could be effectively used to
"flag" lakes which are undergoing rapid change.
Passive scanners such as the Landsat's multispectral scanner (MSS)
and Thematic Mapper (TM) have also been used to detect and map differences
and changes in lakes and their quality. Passive scanners have the
advantageous ability of producing images which contain all points within a
-------
lake and all lakes within a defined scene. However, the use of passive
remotely sensed data requires more water samples for correlation
development, and these data typically have slightly lower coefficients of
determination (R2) values than the active systems. It has been suggested
that a combination of passive and active systems may improve existing
mapping by using the active systems to densify "ground truth" for the
passive (scanning) systems.
REFERENCES
1. Lillesand, T.M. and R.W. Kiefer. Remote Sensing and Image Interpret-
ation. John Wiley and Sons, Inc., New York, 1979. p. 55.
2. Ellefsen, R. Urban Terrain Zone Characteristics. U.S. Army Human
Engineering Laboratory, Aberdeen Proving Ground, Maryland, 1987.
358 pp.
-------
SECTION 1
SUMMARIES OF PANEL DISCUSSION
8
-------
PANEL DISCUSSION ON REMOTE SENSING APPLICATIONS
FOR VEGETATIVE DAMAGE ASSESSMENT
by: Siamak Khorram, Ph.D.; John A. Brockhaus, Ph.D., Computer Graphics
Department and Forestry Department, North Carolina State University,
Raleigh, North Carolina
INTRODUCTION
Satellite and airborne remote sensing systems are currently being
used to map the spatial patterns of forest decline in the United States and
Europe. The data provided by these systems cannot, however, be used to
determine the processes or mechanisms which have contributed to the
development of decline conditions. Previous and present remote sensing
research efforts have focused on monitoring the spatial distribution of
forest decline levels and not on attaining a basic understanding of the
physiological responses to stress.
Results from these studies tend to be very site specific. Thus, it
may not be possible to extend results to region wide applications where the
goal is to identify the response of vegetation to specific stress agents.
However, it may be possible to understand these mechanisms from remotely
sensed data if experiments are conducted which combine destructive
vegetative measurements with acquisition and analysis of high spectral
resolution data acquisition.
EXISTING REMOTE SENSING SYSTEMS
Color and color infrared aerial photography acquired from systems
such as the EPA's enviropod, the National High-Altitude Photography (NHAP)
program, and the U.S. Forest Service panoramic camera system have been used
in assessing existing levels of forest condition. Such imagery has been
successfully used to map vegetative mortality and stress patterns in
natural and agricultural environments. However, aerial photography alone
cannot always provide information detailing the types of stress occurring
within an ecosystem or give a previsual indication of stress.
Aircraft scanning systems with over 200 high spectral resolution wave
bands have been developed by the National Aeronautics and Space
Administration (NASA) and private firms. .Scanners of this type may provide
the data necessary to determine the physiological response of vegetation to
specific stress agents. Due to the narrow swath width of systems currently
in use, however, cost effective regional monitoring of vegetative damage is
-------
not feasible.
Satellite acquired remotely sensed data is available dating back to
1972, the Landsat multispectral scanner (MSS). This data could be used in
time series analysis of changing vegetative damage patterns. Assessments
of the impact of specific stress agents with MSS data is not possible due
to the coarse spatial and spectral resolution of the data.
The Landsat thematic mapper (TM) provides data of a finer spectral
and spatial resolution than the MSS. Data acquired from the TM is
currently being utilized to model changes in forest decline over time in
the spruce-fir ecosystems of the southern Appalachians. However, the TM
and MSS on Landsats 4 and 5 could fail at any time, leaving researchers
with a data gap as no follow-on U.S. satellites are currently planned to go
into operation until the early 1990s.
Alternate satellite remote sensing systems such as the French SPOT
and Japanese MOS satellites could fill this gap. However, results from
preliminary research efforts indicate that these systems do not provide the
spectral resolution required to conduct these types of investigations.
High spectral resolution satellite remote sensing systems are being
developed. These systems are being patterned off of existing aircraft
scanner systems developed by NASA. However, these systems are not planned
for deployment until the mid to late 1990s.
THE NEED FOR ADDITIONAL RESEARCH
There is a distinct need for research in which destructive vegetative
measurements are made in conjunction with the acquisition and analysis of
high spectral resolution data. These types of studies may allow scientists
to determine the physiological responses of vegetation to stress which are
being exhibited by changing spectral patterns.
Research of this nature is not presently being conducted because of
the high cost of obtaining high spectral resolution spectroradiometers
($20,0000-$30,000). This is unfortunate as numerous investigators involved
in EPA sponsored research cooperatives are conducting controlled vegetative
physiological response experiments. If a spectroradiometer were made
available to each cooperative, then high spectral resolution measurements
could be made in conjunction with these experiments in a very cost
effective manner. In this way, specific physiological and spectral
responses to various stress agents could be determined.
SUMMARY
At the present time, TM data and aerial photography are being used to
monitor and map existing and changing levels of vegetative damage. Imagery
of this type does not, however, provide insight into the physiological
10
-------
responses of vegetation to specific stress agents. Very high resolution
satellite remote sensing systems are planned for deployment in the mid
1990s. These systems will be capable of providing information useful in
inferring physiological responses to stress.
Before data from such advanced remote sensing systems can be used to
provide information concerning vegetative responses to stress, controlled
greenhouse and field experiments need to be performed. These efforts
should focus on destructive vegetative measurements in conjunction with
high spectral resolution data acquisition. Analysis of this data will
provide insight into the physiological responses to specific stress agents
that may be detected with remotely sensed data. Additionally, results from
this research will indicate where within the electromagnetic spectrum
researchers should concentrate their efforts in detecting vegetative
responses to stress. It is recommended that the EPA supply each of its
sponsored research cooperatives with a high spectral resolution
spectroradiometer to facilitate conducting this type of research.
11
-------
VEGETATIVE ASSESSMENT PANEL DISCUSSION PARTICIPANT LIST
Name
Sayed Heshmat
William Philpot
Edward Collins
Mike Bristow
Lee Williams
Richard Ellefsen
Lawrence T. Fisher
Margo J. Kingston
Michael V. Campbell
John Brockhaus
Jon Schneeberger
Mark Stranieri
Billy Fairless
Donald WiIbur
Susan Ustin
Glen Green
Sky Chamard
Brian Curtiss
Gerald Dildine
Tom Mace
Jon Beazley
Nancy DeFeo
Lynn Fenstermaker
Address
B156 NCEL, Univ. of Illinois at Urbana-Champaign, IL
CLEARS, Cornell Univ., Hollister Hall, Ithaca, NY 14853
1712 Silver SE, Albuquerque, NM 87106
EPA, Environmental Monitoring Systems Laboratory, P.O. Box 93478,
Las Vegas, NV 89193-3478
804 Dale Hall Tower. Univ. of Oklahoma, Norman, OK 73019
Geography Dept., San Jose State Univ., San Jose, CA 95192
Lockheed EMSCo, 1050 E. Flamingo Rd, Ste 129, Las Vegas, NV 89119
U.S. Geological Survey, Mail Stop 927, Reston,- VA 22092
North Carolina State Univ., Computer Graphics Center, Box 7106, Raleigh,
NC 27695
North Carolina State Univ., Forestry Dept., Box 8002, Raleigh, NC
27695-8002
National Geographic Society, 1145 17th St., Washington, DC 20036
North Carolina State Univ., Computer Graphics Center, Box 7106, Raleigh,
NC 27695
EPA Region VII
PA Dept of Transportation, Harrisburg, PA
Dept of Botany, University of California, Davis, CA 95616
Washington Univ., EPSc Dept., Box 1169, St. Louis, MO 63130
E. Coyote Enterprises, P.O. Box 10761, Eugene, OR 97440
University of Colorado-Boulder, CIRES, Box 449, Boulder, CO 80303
North Carolina Dept. of Transportation, Highway Bldg., Raleigh, NC 27611
EPA Environmental Monitoring Systems Laboratory, P.O. Box 93478,
Las Vegas, NV 89193-3478
330 Ponce St., Tallahassee, FL 32303
26 Chestnut St., Apt. 15, Wakefield, MA 01880
Environmental Research Center, University of Nevada, Las Vegas, 4505
South Maryland Parkway, Las Vegas, NV 89154
12
-------
PANEL DISCUSSION ON
BUILDING MATERIALS EFFECTS ASSESSMENT
By: Richard Ellefsen, Department of Geography, 1 Washington Square, San
Jose State University, San Jose, California 95152
The discussion was initiated by a review of the building inventory
work done to date. Shortcomings of early ground surveys in Cincinnati,
Pittsburgh, New Haven, and Portland (ME) were noted for building materials
inventorying. The use of air photo interpretation greatly improved the
inventory process.
In addition to the Cincinnati Enviro-Pod work presented in this
volume, a pilot study has been conducted for a part of downtown Baltimore,
and a study of 30 urban test sites in the South Coast Air Basin (Los
Angeles area) is currently underway. For these studies, the necessity of
using oblique aerial photography -- to see the side walls of buildings --
is fully recognized.
A problem encountered in using the Enviro-Pod imagery (even though it
employs a high resolution reconnaissance camera and high quality film
[Kodak Aero Ektachrome 2448]) is that with the 80 mm normal lens plus
adherence to an FAA-imposed restriction on minimum flying altitude over
cities (1,000 feet), the images do not permit manual identification of all
building materials by employing the usual identifiers of shape, size,
color, tone, and texture. Instead, the architectural/construction form of
the buildings are observed and compared with a key developed to delineate
the suites of wall materials for given types of buildings.
Discussion then turned to the possibilities of either improvement of
the photo equipment or exploration of other remote sensing means.
Experimentation with different combinations of fixed cameras and different
lens combinations would broaden the base of knowledge on the employment of
the texture of materials as a key aid in identification. A longer focal
length lens could bring the desired resolution but perhaps such a lens
would not be feasible to use with the panning camera system of the
Enviro-Pod. This type of lens, however, would considerably increase the
number of exposures to provide broad coverage. This, in turn, would
suggest altering the nature of the flying mission from one of "flying the
universe" -- with controlled overlap and sidelap along straight flight
lines -- to flying 360* circles around preselected sample areas. Achieving
this would necessitate sample selection in advance of taking the
photography, rather than after, thus in some senses limiting flexibility
13
-------
and negating the possibility of using a broad photography base for other
aspects of an inventory project or other related projects.
Discussion on resolution led to the common question of how much
detail oh building materials is required and how remotely sensed data
compares with those acquired directly on the ground. At these early stages
of acid precipitation assessment -- with damage function data from the
laboratory not yet complete and aggregate costs not yet determined -- the
optimum level of discreteness of the classification of building materials
is still not known. For instance, the value of being able to determine
whether a wall is painted concrete block or painted brick has not been
determined.
The relatively trivial problem of attaining the correct exposure for
Enviro-Pod imagery was addressed. The system, at present, requires a
preflight setting of the camera's aperture with no possibility of changing
it in flight. Mounting a control (probably in the same box as the
intervalometer) to be manipulated by the aerial photographer would be a
major improvement. Adopting an automatic exposure control would in all
likelihood be even better.
The possibility of employing some sort of non-chemical film
electronic sensor was explored by the group. Charge Coupled Devices
(CCD's) were suggested but resolution is probably lower than that of
photographic film. Some high spectral resolution scanners which may
perform well for inventories and damage assessments have recently been
introduced. An aspect of this discussion included a summary of research
which had been performed by the U.S. Geological Survey on the detection of
acid rain damage to building stone. Two building stones which were
examined in this study were limestone and marble. A hand-held spectrometer
successfully detected the decomposition of these carbonate stones to gypsum
as a result of acid deposition. It was proposed that this research be
expanded to other building materials and that the use of an airborne sensor
be investigated. It was also recommended that the assessment of building
decay utilizing spectrometer be implemented on a local basis within urban
areas and combined with inventory data. However, in an aircraft-carried
mode, a high spatial and spectral resolution scanner might present some
complex data management problems.
A part of the discussion focusing on extrapolation of existing
inventory studies to new urban areas brought out the point that definitive
work on the distribution of building materials in cities does not exist.
Steps have been taken in both the Cincinnati and the Los Angeles studies to
employ urban terrain zones, i.e., areas that are homogeneous in functional
and morphological characteristics, as a sampling frame for this very
purpose. If these zones replicate, as anticipated, and proportions of
different building materials remains fairly constant (with probably some
regional modification), quantitative knowledge about the nature of cities
will be gained.
Another area of general discussion was the possibility of making
significant measurements about building roof materials. The photo data
14
-------
base is rich but data requirements on type, distribution, and relationship
of roof materials to various types of buildings and urban terrain zones
have not yet been specified by the National Acid Precipitation Assessment
Program. Building roofs which directly intercept most of the acid
precipitation invite further study.
A final recommendation of the panel was to initiate the use of a
geographic information system (GIS) to better store, analyze, and produce
maps and reports. The use of GIS would also facilitate the manipulation of
data for inclusion in modeling activities.
In summary, building material areas that could profit from more
research are:
Experiment with other camera/film/filter combinations to
increase spatial resolution;
Examine possibilities of using obliquely aimed multispectral
scanners and other CCD instruments;
Experiment with exposures under various lighting conditions,
perhaps flying under a high cloud cover;
Resolve questions on the optimum level of building materials
classification system discreteness vis-a-vis remote sensing
opportunities and constraints;
Determine relationship of urban terrain zones and building
surface materials in general, adjusting for regional variations
should that prove necessary;
Explore the placement of building materials data into a
Geographic Information System for interaction with
meteorological models;
Determine floor space/ground space ratios per urban terrain
zone to gain a better knowledge of the proper significance of
all parts of the city; and
Experiment on the direct determination of building surface
deterioration using a hand-held spectrometer for other building
materials.
15
-------
BUILDING MATERIALS EFFECTS PANEL DISCUSSION
PARTICIPANT LIST
Name Address
Bill Philpot CLEARS, Cornell Univ., Hollister Hall, Ithaca. NY 14853
Mike Bristow EPA, Environmental Monitoring Systems Laboratory, P.O. Box 93478,
Las Vegas, NV 89193-3478
Lee Williams 804 Dale Hall Tower, Univ. of Oklahoma, Norman, OK 73019
Richard Ellefsen Geography Dept., San Jose State Univ., San Jose, CA 95192
Margo Kingston U.S. Geological Survey, MS-927, Reston, VA 22092
Jon Schneeberger National Geographic Society, 1145 17th St., Washington, DC 20036
Jon Beazley 330 Ponce St., Tallahassee, FL 32303
Tom Mace EPA, Environmental Monitoring Systems Laboratory, P.O. Box 93478,
Las Vegas, NV 89193-3478
Lynn Fenstermaker Environmental Research Center, University of Nevada, Las Vegas, 4505
South Maryland Pkwy Las Vegas, NV 89154
16
-------
PANEL DISCUSSION ON REMOTE SENSING APPLICATIONS
FOR ASSESSMENT OF WATER QUALITY DEGRADATION
DUE TO ACID DEPOSITION
By: Michael Bristow, EPA, Environmental Monitoring Systems Laboratory,
P.O. Box 93478, Las Vegas, NV 89193-3478
At the present time, it is not possible to remotely measure by direct
means the chemical and physical properties thought to influence or be
influenced by surface water acidification. Some of these parameters are
pH, total alkalinity, and concentrations of sulfate, nitrate, and
extractable aluminum, although a potential exists for measuring sulfate and
nitrate concentrations directly by remote laser vibrational Raman
spectroscopy.
Consequently, the approach adopted has been to examine those
physical, chemical or biological water quality parameters that can be
measured by remote sensing methods to see whether their behavior is, in
some way, influenced by lake water acidification. These methods can be
conveniently divided into passive and active categories.
DISCUSSION
Passive methods, either satellite or airborne, include photography,
videography and multispectral scanner (MSS) imagery and involve monitoring
the visible and near infrared (IR) solar radiation backscattered from the
water surface and column. MSS imagery, which provides spatially resolved
data in a number of visible and near IR spectral bands, can, via multiple
regression, be related to chlorophyll a, suspended sediments (water
clarity), dissolved organic carbon (DOC) and pH provided that suitable
"ground truth" data is available to calibrate the remotely sensed data.
Previous investigations funded by the Environmental Protection Agency have
focused on determining whether temporal and spatial changes in lake water
acidity are correlated with changes in phytoplankton concentration and
diversity, and changes in water color and clarity. The latter phenomenon
is thought to be influenced by reductions in DOC concentrations through
precipitation by increasing aluminum concentrations. Unfortunately, no
consistent trends or universal regression models have been found to exist
and the degree of correlation is often low. Consequently, it is difficult
to differentiate the influence of acidification from other effects such as
seasonal variability and long-term ecological trends unrelated to lake
water acidification. A positive note might be that an extensive (MSS) data
base for a given impacted region might, nevertheless, provide a baseline
from which to determine whether and by how much lake water acidification is
17
-------
influencing the limited number of remotely sensed variables about which
quantitative information may be acquired.
Active methods, specifically airborne laser profiling systems, are
able, through the process of laser-induced fluorescence, to measure the
concentrations of chlorophyll a and DOC, and the optical attentuation
coefficient (a specific indicator of water clarity). Although the
precision of the airborne laser fluorosensor data is generally much higher
than that for passive sensor data, it is still necessary to calibrate the
data against a limited concurrent ground truth data set using simple
(linear) regression methods. Again, interpretation of the active sensor
data as they relate to acid-deposition trends is subject to the same
limitations as those indicated above for the data acquired by passive
sensor systems.
In a further development, it has been established that the
fluorescence spectra, which provide the DOC data, are also influenced by
changes in aluminum concentration and pH through the process of
fluorescence quenching. However, the behavior of these-relationships are
not as yet sufficiently well defined or understood so that the data from
these spectra could be used to accurately predict either pH or aluminum
concentration.
In contrast to the above passive and active remote sensing
techniques, which provide data that is only indirectly related to lake
water acidity, laser induced Raman spectroscopy is known to be capable of
measuring the concentrations of a number of anions (SO,, NO,, C03, and POJ
that are directly related to acid deposition. The feasibility of using
Raman spectroscopy in this type of application has been investigated a
number of times over the last 20 years where it was shown that the Raman
emission intensity varied linearly with anion concentration. Putting this
application into practice has so far not been possible because the
intensity of the solar background and concurrent laser-induced fluorescence
signals completely dominate those of the anion Raman signals for the
concentrations prevailing in lakes affected by acid deposition. Extracting
these weak Raman signals from an intense noise background will require
sophisticated signal enhancement and noise suppression techniques.
CONCLUSIONS
The general consensus of the panel discussions are as follows:
1. Existing passive and active methods alone have, so far, failed
to demonstrate that they can accurately and consistently
measure either directly or indirectly parameters related to
lake water acidity, although measurement of other water quality
indicators have been demonstrated consistently;
2. It was recommended that a specific parameter or a suite of
parameters be defined encompassing both active and passive
18
-------
sensors that will provide a warning that specific lakes or
regions are undergoing changes due to acid deposition as
related to an existing baseline data set;
3. It was suggested that the potential of remote laser Raman
spectroscopy be investigated as a tool for use in measuring
lake water acidity parameters. Specifically, special emphasis
should be placed on investigating whether signal enhancement
and noise suppression techniques can make this method viable;
and
4. At the present time, liaison, communication, and coordination
between the engineers and scientists responsible for designing
and' operating existing remote sensing systems and those
ecologists, chemists and biologists responsible for conducting
the (in situ) water quality surveys is practically nonexistent.
Efforts must be made to bridge this gap and to determine
specific areas where present and future sensors can be of
direct benefit to those conducting surveys such as those
related to lake water acidity.
19
-------
WATER QUALITY PANEL DISCUSSION
PARTICIPANT LIST
Name
Address
Sky Chamard
Gerald Dildine
John Sherbert
Tom Mace
Jon Schneeberger
Margo Kingston
Lawrence T. Fisher
Richard Ellefsen
Lee Williams
Mike Bristow
Edward Collins
Bill Philpot
Lynn Fenstermaker
E. Coyote Enterprises, Box 10761, Eugene, OR 97440
North Carolina Department of Transportation, Raleigh, NC
North Carolina Department of Transportation, Raleigh, NC
EPA, Environmental Monitoring Systems Laboratory, P.O. Box 93478,
Las Vegas, NV 89193-3478
National Geographic Society, 1145 17th St., Washington, D.C. 20036
U.S. Geological Survey, Mail Stop 927, Reston, VA
Lockheed EMSCo, 1050 E. Flamingo Rd.. Las Vegas, NV 89119
Geography Dept., San Jose State Univ., San Jose, CA 95192
Geography Dept., Univ. of Oklahoma, Norman, OK 73019
EPA, Environmental Monitoring Systems Laboratory, P.O. Box 93478,
Las Vegas, NV 89193-3478
1712 Silver SE, Albuquerque, NM 87106
CLEARS, Cornell University, Ithaca. NY 14853
Environmental Research Center, University of Nevada-Las Vegas, 4505
South Maryland Pkwy. Las Vegas, NV 89154
20
-------
PANEL DISCUSSION ON
NEW APPLICATIONS OF REMOTE SENSING TECHNOLOGY
TO ACID DEPOSITION ASSESSMENTS
by: T.H. Lee Williams, Geosciences Remote Sensing Group, University of
Oklahoma, Norman, OK 73019
ABSTRACT
This section summarizes the panel discussion on new applications.
Remote sensing for acid deposition will make use of spaceborne, airborne
and ground-based sensors. New sensors are being developed for each
platform level. The NASA EOS program will provide comprehensive spaceborne
visible, infrared and microwave data sets commencing in the mid-1990's,
preceded by airborne sensor prototypes or simulators. The visible and
shortwave infrared imaging spectrometers are of great interest for
vegetation and building damage assessments. Airborne ultraviolet Lidars
provide opportunity for 2-D profiling of atmospheric pollutants. Practical
considerations include equipment size and weight to allow use of light
aircraft for economical operations. Ground-based narrow-band spectral
radiometers have potential for routine on-site assessment of building
materials damage. Commercial systems at reasonable prices are becoming
available. A significant area for remote sensing and CIS lies in providing
input to regional air pollutant transport models. The models themselves
will evolve to make use of the capabilities of remote sensing and GIS.
INTRODUCTION
This paper reports on the discussions held in the panel on new
applications. The panel ranged freely over sensors, models and monitoring,
and applications to vegetation, building materials and the atmosphere.
This section is organized by and summarizes these topics. It should not be
taken as a comprehensive review and assessment of new remote sensing
applications to acid deposition.
SENSORS
Acid deposition studies will make use of spaceborne, aircraft, and
ground-based remote sensors. Recent sensor developments and planned
missions in the 1990's will provide a wide range of possibilities.
21
-------
The NASA Earth Observation System (EOS) program will provide a series
of specialized sensors in the mid-1990's, mounted on free-flier and manned
space platforms. The actual configuration has not been finalized but the
sensors will cover the visible, shortwave and thermal infrared, and
microwave regions. Several of the sensors are of particular interest. The
High Resolution Imaging Spectrometer (HIRIS) will provide 30 meter
resolution imagery in 200 narrow spectral channels in the visible and
shortwave infrared. The spatial coverage will be non-continuous, but
targeted for specific project/problem areas. The sensor is pointable both
sideways and fore-aft. The latter allows increased view time on particular
targets, thus yielding an improved signal-to-noise ratio for subtle
spectral features (e.g. in water). Of interest in forestry studies is the
potential to provide canopy, geometry information through multiple view
angles of the same scene. The MODIS sensor complements HIRIS, and has 40
to 60 spectral channels, one-half to one kilometer resolution, and two-day
repeat time. MODIS will provide polarization information that may yield
information on the physical condition of vegetation. The multifrequency
multipolarization microwave sensors proposed for EOS may provide canopy
water stress and volumetric distribution information, as well as surface
roughness data for atmospheric transport models.
The EOS sensors are scheduled for the mid-1990's and beyond.
Aircraft prototypes of the sensors will be flown before then to provide
advance data for development of research and applications areas. In some
areas, e.g. vegetation studies, we are now producing more spectral data
than we currently understand how to use. Research is needed in the use of
these sensors specifically for acid deposition studies. In particular,
researchers need more access to AVIRIS (Advanced Visible and Infrared
Imaging Spectrometer) data for their study areas now in order to prepare
for the spaceborne HIRIS. These imaging spectrometers present massive data
handling and processing tasks. NASA is funding software development to
handle the data. One area of interest is in spectral band selection for
specific applications areas. Much basic lab and field research is required
to document and understand the spectral characteristics of surfaces
affected by acid deposition. The EOS program will provide multisensor data
covering the visible, infrared and microwave wavelength, and will spur much
necessary research on the analysis of these combined wide-spectrum data
sets.
The EOS program includes a variety of atmospheric sensors, but will
not provide detailed lower-atmosphere data. Aircraft systems will be an
important component of an acid deposition assessment program. The EPA is
developing an ultraviolet Differential Absorption Lidar (DIAL), for use in
the Regional Acid Deposition Model, which deals with air pollutant
transport. NASA has already developed a UV-DIAL for 03 measurements, and
is currently testing it in a DC-8 aircraft. NASA's experience with their
UV-DIAL has been advantageous to the development of the EPA UV-DIAL. The
EPA DIAL will give range-resolved 03 and S02 concentrations above or below
the plane, thus providing a 2-D profile of the atmosphere. Measuring N02
concentrations may be 'possible also. An important aspect of any aircraft
sensor is the equipment size and weight. If the sensor can be mounted in a
small plane of less than 12,500 Ib gross weight (e.g. a Twin Otter), then
22
-------
airplane costs are significantly lower as planes above 12,500 Ib require
two pilots and additional licensing.
Ground-based sensors will also play a significant role. They will be
used to calibrate airborne sensors, and aid in the development of sensors
and models. They will also be used for operational data acquisition such
as assessing building damage through spectral reflectance observations. A
number of field spectrometers and radiometers are becoming available. The
PIDAS (Portable Instantaneous Display and Analysis Spectrometer) provides
visible and shortwave infrared data with Inm and 5nm resolution in the
visible and shortwave infrared respectively. Other devices with a small
number of selectable narrow bandpasses are available at costs starting
around $10,000. The selection of specific spectral bands used in the
sensors for a particular application will be based on research done using
the PIDAS/AVIRIS spectrometers. For example, studies of damages to marble
building materials can be achieved using narrow bands in the shortwave
infrared. Relatively simple dual-beam ratioing radiometers can be
developed and used widely for on-site building damage assessment.
ATMOSPHERIC TRANSPORT MODELS AND MONITORING
Current observations of pollutants, e.g. S02, N02, 03, are based on
point samples. Most atmospheric transport models are regional in scale and
lack sufficient sample data for verification. We need data on the spatial
distribution of pollutants to validate and improve the models. It is not
feasible to collect the required volume of point-sample data using grab
samples. Remote sensing can and should play a role in providing the
necessary data on the spatial distribution of airborne pollutants and other
spatial parameters required by atmospheric transport models. However,
while current models lack sufficient sample data for verification, it is
also true that the spatial detail and volume of data derived from remote
sensing will overwhelm existing models. There will be an evolutionary
process whereby the models themselves will change to accommodate the
information available from remote sensing. Remote sensing can provide both
atmospheric data (e.g. from the EPA DIAL system) and relevant land surface
data such as elevation, slope, land cover, aerodynamic roughness (either
from land cover or directly from microwave measurements) and albedo,
incorporated into a geographic information system.
Remote sensing will play an important role in validating and
initializing atmospheric pollutant transport models. However, the current
models make assumptions that fail in complex terrains and near coastal
areas, which are both of great interest. Remote sensing will play a role
therefore as a primary means for monitoring the spatial distribution of
pollutants in complex areas where models fail, hence the need for airborne
sensors that can be flown on economical small aircraft. The UV Lidar
systems are promising, although there is disagreement over their practical
value. Further research needs to be conducted to make them fully
operational.
23
-------
VEGETATION STUDIES
The effects of acid deposition on vegetation have been documented
mostly in forestry, where the long-term effects and economic/environmental
impacts are most readily seen. Significant advances in forest studies may
result from the use of the multiple view angles of the fore-aft pointable
HIRIS sensor in determining canopy geometry. The mul tifrequency
multipolarization radars also offer significant potential for canopy
geometry and volumetric water distribution. Fumigation studies have been
done on agricultural crops producing changes in plant growth, senescence
and crop yield in response to atmospheric pollutants. However, these
studies in general did not include a remote sensing component. Detecting
the effects of acid deposition on annual crops is difficult as one cannot
observe cumulative impacts over many years, and it is difficult to
deconvolve the effects of pollution from the effects of weather, treatment,
and soils. Although large-area monitoring is difficult, remote sensing is
more feasible for local monitoring of the effects of strong point source
pollutants. Acid deposition may actually have a positive effect on certain
soils, e.g., alkaline soils. The areas most sensitive are those with
crystalline bedrock and thin or organic poor soils with minimal natural
buffering. The impacts of acid deposition on rangelands may be easier to
address, but grazing practices and soil buffering capacity may mask the
effects.
Wetlands are the primary native plant communities other than forested
areas that may be affected. Most work has been done on the East coast with
emphasis on salinity, trace metals, and agricultural runoff. The effects
of acid deposition on wetlands are unknown, but many wetlands soils are
acidic and may be sensitive to acid deposition.
BUILDING MATERIALS
Ground-based radiometers with selected narrow bandpasses are
appropriate for monitoring masonry. Work has been done on the effects of
acid deposition on metals and paints, but little is known about the impacts
on roofing materials. Basic research is needed before remote sensing
systems for monitoring roofing materials can be proposed.
24
-------
NEW APPLICATIONS PANEL DISCUSSION PARTICIPANT LIST
Name
Address
Mike Bristow
Richard Ellefsen
Lee Williams
Tom Mace
Susan Ustin
Brian Curtiss
Bill Philpot
Jon Beazley
Margot Kingston
Lynn Fenstermaker
EPA Environmental Monitoring Systems Laboratory, P.O. Box 93478,
Las Vegas, NV 89193-3478
San Jose State Univ., Geography Oept., San Jose, CA 95192
804 Dale Hall Tower, Univ. of Okla., Norman, OK 73019
EPA Environmental Monitoring Systems Laboratory, P.O. Box 93478,
Las Vegas, NV 89193-3478
University of CA-Davis. Dept. of Botany, Davis, CA 95616
University of Colorado, CIRES, Box 449. Boulder, CO 80303
CLEARS, Cornell University, Hollister Hall, Ithaca, NY 14853
330 Ponce St., Tallahassee, FL 32303
U.S. Geological Survey, MS-927, Reston, VA 22092
Environmental Research Center, University of Nevada-Las Vegas, 4505
South Maryland Parkway. Las Vegas. NV 89154
25
-------
SECTION 2
SESSION MANUSCRIPTS
26
-------
REMOTE DETECTION OF ACID MIST AND OZONE EFFECTS
ON CONIFER AND BROADLEAVED VEGETATION
By: Walter E. Westman, Environmental Policy Analysis Unit, Lawrence
Berkeley Laboratory, Bldg. 90-H, University of California, Berkeley, CA
f\ A TOrt
94720
ABSTRACT
A series of satellite, aircraft, field and laboratory measurements
were taken on two types of California vegetation in order to relate remote
images of vegetation stress to underlying changes in spectral reflectance
from canopies. The two ecosystem types under study were the yellow
pine-giant sequoia mixed conifer forests of the southern Sierra Nevada
(Pinus Jeffrey/, P. ponderosa, Sequoiadendron gigantea), and the coastal
sage scrub of the Santa Monica Mountains in the Los Angeles basin (Salvia
mellif era, Artemisia californica, Rhus laurina). In each case remote
imagery was obtained along a pollution gradient composed primarily of
ozone, with secondary influence by intermittent acid mist, including both
nitrogen and sulfur types. In order to study changes in leaf reflectance
in response to alterations in leaf chemistry, anatomy and moisture content
of foliage of these species under controlled conditions, two conifer
species (Jeffrey pine, giant sequoia) were exposed intermittently to four
lavels of acid mist and three levels of ozone in factorial combination in
fumigation chambers over a two-year period. Reflectance measurements were
taken with a Collins VIRIS instrument. Ten species of coastal sage scrub
were also exposed in fumigation chambers to multiple levels of ozone and
sulfur dioxide, alone and in combination.
Changes in leaf chemistry varied by species and pollutant treatment
(1,2), raising the possibility that chemical "signatures" in foliage could
be used to differentiate the effects of different pollutants in a pollutant
mixture in the field. The potential for this application was illustrated
in the Santa Monica Mountains (2), where it was shown that foliar chemical
changes along the pollution gradient were those expected by a predominant
influence of ozone, with secondary influence by acid mist. In the southern
Sierra Nevada, chemical changes were not significant along the pollution
gradient, with the possible exception of a mild "fertilization" effect due
to acid mist (1).
The potential for remote detection of chemical features in foliage by
analysis of aircraft-borne Airborne Imaging Spectrometer (AIS) data was
examined along the pollution gradient in the Santa Monica Mountains for
coastal sage scrub by comparison of AIS spectra with field-derived spectra
27
-------
obtained with a Spectron SE-590, Collins VIRIS and Barringer Refspec Ha
(3,4). It was concluded that since the stretching frequencies of key
organic bonds such as -OH are found in several of the major plant compounds
(cellulose, starch, sugar inter alia), the differentiation of particular
chemical compounds that are themselves highly correlated in concentration
is not readily amenable to stepwise regression analysis techniques with the
modest sample sizes in our data. Much larger sample sizes are needed to
unravel the multiple influences on AIS spectra than those obtained to date,
particularly in view of the large number of potential predictor wavelengths
in relation to sample spectra (3,5).
The use of broad-banded remotely-obtained data in detecting
pollution-induced stress was also examined at both sites using Thematic
Mapper (TM) band data from aircraft (simulator) and satellite (LANDSAT)
(4,6). The changes in TM reflectance along the pollution gradients were
interpreted in relation to data from laboratory experiments. In addition
to comparing TM changes from fumigated plants (conifer seedlings), an
artificial air-drying experiment with red pine needles was conducted to
examine changes in TM bands with moisture loss (7). Reasons for change in
leaf reflectance upon drying were studied by quantitative measurements of
change in needle tissue cross-sectional areas, water content, and spectral
reflectance during the course of water loss. Denaturation of chlorophyll
pigments upon water loss appeared most influential on TM band 3; this
phenomenon was also capable of inducing a blue shift of the red edge of the
near-infrared plateau of the spectral reflectance curve. Reduced water
absorption combined with shrinkage of cellulose micelles in cell walls
seemed best able to account for changes in TM band 4. In TM band 5, the
loss of water absorption predominated in effect on reflectance. Air
pollutants were able to mimic some of these reflectance changes by their
effect on stomatal closure and hence water loss, and by inhibition of cell
expansion consequent upon water stress, leading to cell stunting -- with
effects on reflectance comparable to cell shrinkage (7).
The use of Thematic Mapper data over the two pollution gradients in
the field revealed that natural variations in canopy closure, with
subsequent exposure of understory elements, were sufficient to cause
changes in spectral reflectance that could obscure differences due to
visible foliar injury symptoms observed in the field. Thematic Mapper band
data are therefore more likely to be successful in distinguishing pollution
injury from background variation when homogeneous communities with closed
canopies are subjected to more severe pollution-induced structural and/or
compositional change than currently occurs at those southern California
sites examined in this analysis (4).
28
-------
REFERENCES
1. Westman, W.E., and Temple, P.J. Acid Mist and Ozone Effects on Leaf
Chemistry of Two Western Conifer Species. Environ. Pollution: in
review, 1988.
2. Vlestman, W.E. Detecting Early Signs of Regional Air Pollution Injury
to Coastal Sage Scrub. In: G.M. Woodwell (ed.), Biotic
Impoverishment: Changes in the Structure and Function of Natural
Communities under Chronic Disturbance. Cambridge Univ. Press, New
York. In press, 1988.
3. Price, C.V., and Westman, W.E. Toward Detecting California Shrubland
Canopy Chemistry with AIS Data. In: G. Vane (ed.), Proc. 3rd
Airborne Imaging Spectrometer Data Workshop. Jet Propulsion
Laboratory, Pasadena. JPL Publ., 87-30, 1987. pp. 91.
4. Westman, W.E., and Price, C.V. Detecting Air Pollution Stress in
Southern California Vegetation Using LANDSAT Thematic Mapper Band
Data. Photogr. Engr. Rem. Sens.: in review, 1988.
5. Westman, W.E., and Price, C.V. Remote Detection of Air Pollution
Stress to Vegetation: Laboratory-level Studies. In Proc. IEEE Intl.
Geoscience Rem. Sens. Symp., Vol 1, IEEE, New York, 1987. pp. 451.
6. Westman, W.E. Monitoring the Environment by Remote Sensing. Trends
in Ecology and Evol., 2:333, 1987.
7. Westman, W.E., and Price, C.V. Spectral Changes in Conifers Under
Air Pollution and Water Stress: Experimental Studies. IEEE Trans.
Geosci. Rem. Sens., 26:11, 1988.
29
-------
A PRELIMINARY EVALUATION OF THE USE OF TM IMAGERY FOR
THE STUDY OF FOREST DECLINE IN
THE SOUTHERN APPALACHIANS
By: Siamak Khorram, John A. Brockhaus, and William W. Cure, Department of
Forestry and Computer Graphics Center, North Carolina State University,
Raleigh, NC 27695-7106; Robert I. Bruck, Departments of Plant Pathology
and Forestry, North Carolina State University, Raleigh, NC 27695-7616.
ABSTRACT
The apparent decline of the remnant spruce-fir forests of the
southern Appalachians has stimulated a considerable research effort there
to document the health of these forests and to develop hypotheses as to
possible causes. The rugged terrain and heterogeneity of community types
within the southern Appalachians make a remote sensing approach to forest
assessment highly desirable and cost effective. For a preliminary study,
two Thematic Mapper (TM) images, one in the fall of 1984 and the other in
the fall of 1986, were obtained of the Black Mountains in North Carolina,
including Mt. Mitchell, the highest point in the eastern U.S. Correlations
were obtained between digital values and ratings of tree decline from
permanent plots established throughout the range. Two methods were
compared for locating the plots and retrieving digital values from the
images. One used map coordinates obtained from the investigators who had
set up the plots and for the other, an individual familiar with the plots
located them visually on the image. Correlations between the resulting
sets of digital values and ratings of tree decline within the plots were
very different. Since both methods were subjective and the results so
different, neither set of findings can be considered definitive. Thus
before remote sensing can be used for assessing stand conditions within
these mountainous forests, techniques must be developed for accurately
locating the permanent field plots.
INTRODUCTION
The apparent decline of the remnant spruce-fir stands in the southern
Appalachians has stimulated a considerable research effort there to
document the health of these forests and to develop hypotheses as to
possible causes. Characterization of stand conditions centers around
observations made in permanent plots selected to represent the widest
possible range in slope, aspect, elevation, and soil conditions (1).
Repeat observations are expected to provide a database for assessing
30
-------
vegetative changes over the years. Field observations of this type,
however, are both time consuming and costly and at best, cover but a small
portion of the ecosystem (2). Given the difficult terrain and the wide
variety in microclimate, assessments of the entire ecosystem by field crews
would be impractical.
Recent advances in technology, however, have made available satellite
and airborne scanners with such improvements in spectral and spatial
resolution that the study of ecosystems as spatially diverse as the
spruce-fir stands in the southern Appalachians should now be possible with
remotely-sensed data (3). Since satellite scanners provide repetitive
coverage of the same ground areas, the possibility also exists for temporal
studies of changes in stand conditions.
A preliminary study was conducted to assess the feasibility of using
remotely-sensed data to survey spruce-fir forests in the southern
Appalachians with two TM images of the Black Mountains in North Carolina,
one from 1984 and the other from 1986. The Black Mountains include Mt.
Mitchell, the highest peak in the eastern U.S. Density values in the six
reflective bands of the TM images, corresponding to the locations of
permanent field plots, were compared with data characterizing stand
conditions within the plots. The objectives were to determine the degree
of correlation between the digital data and various measures of forest
decline and to formulate a modeling approach for a larger study of
ecosystem changes from 1984 to 1987. Of particular concern, given the
abrupt changes in community types, were the effects on the data sets to be
used for modeling. (The models stratify the forests by decline class.) Of
concern were the different methods of locating the permanent study plots on
the images. Two methods of plot location were compared.
METHODS
SITE CHARACTERIZATION DATA
The Black Mountains are a north-south ridge approximately 17 km in
length. Data from permanent plots established there by investigators from
North Carolina State University (NCSU) and Virginia Polytechnic Institute
and State University (VPI) were used in calculating correlations with
density values from the TM scenes. [Work by the VPI group was supported by
the Spruce-Fir Cooperative of the Forest Response Program.] Plots
established by the NCSU workers were 1 ha in circular projection, those by
VPI, 400 mz (7 of the former and 10 of the latter were used in the analysis
with the 1986 scene; only data from the 7 NCSU plots were available in
1984). Pertinent data included plot elevation, aspect and percent slope as
well as tree species and decline ratings. Four decline classes were used:
class 1, 0 to 10% defoliated; class 2, 11 to 50% defoliated; class 3, 51 to
90% defoliated; and class 4, 91 to 100% defoliated (standing dead).
31
-------
IMAGE PROCESSING
Landsat 5 TM scenes were acquired October 3, 1984, and September 7,
1986. Both were collected with less than 10% cloud cover. Selection of
subscenes of the Black Mountain study site was required for comparisons of
changes in forest stand conditions. A distinctive surface feature
northwest of the study site was used to define the upper left corner of the
1984 subscene. Image x,y coordinates of this feature were converted to
latitude and longitude using transformation algorithms and the ancillary
geographic coordinate data for the TM scene. Latitude and longitude for
this point were then converted to image x,y coordinates for the 1986 TM
scene using that scene's ancillary geographic data. The location of this
point in the two TM scenes was then checked visually to confirm that the
upper left pixels for each subscene were identical. Then, relative to the
upper left pixel, subscenes were defined by determining the number of scan
lines and pixels per scan line necessary to cover the study site. Each TM
subscene was then transformed to a UTM map projection using algorithms
within the Image Analysis System (IAS).
LOCATION OF STUDY PLOTS ON THE IMAGES
Two methods were evaluated for locating the field plots with the 1986
scene. For the first, 7.5' USGS maps were provided by the VPI group
showing the location of the plots they had established. The maps were
annotated by the NCSU workers to include their 7 plots. Latitude-longitude
for these points were used to retrieve density values in the six TM bands.
Alternatively, one of the NCSU investigators familiar with the Black
Mountain study area (R. Bruck) visually located the plots on the images and
density values were obtained corresponding to these points.
Two separate data sets were constructed, each containing the same set
of decline ratings but with density values obtained by one of the two
methods of plot location. Also, two values were used to represent the
densities in each TM band for each plot, one corresponding to the single
pixel at plot center and the other calculated as the average of a 3 by 3
kernel around the plot center. Correlations were estimated with procedures
within PC-SAS. The arcsine-square root transformation was first applied to
all percent-type variables.
RESULTS AND DISCUSSION
In 1984, decline ratings were low with more than 85% of the trees
rated as class 1 or 2. By 1986, only about 70% of the trees were still in
these two classes and 20% were rated as standing dead (class 4).
Correlations were significant between the decline ratings and plot
elevation both years but especially with the healthy trees in class 1
(r=-0.773 for the 1984 data set and -0.714 in 1986). Correlations between
decline ratings and plot aspect were generally in excess of |0.4| although
the relation was not significant; percent slope was not correlated with
decline. As models developed with these data will be used for classifying
TM scenes, regardless of the method of locating the plots and obtaining the
32
-------
digital values, they will thus have to take elevation and aspect into
account. Also, pooling data across years will be necessary to obtain a
wide enough range in the response variables to adequately characterize the
relation between the image and forest data.
The relationship between the field and image data was very sensitive
to the method of locating the plots. Correlations were much higher,
regardless of method, between the class 1 (healthy) and class 4 (dead)
ratings and the digital data than with either of the two intermediate
decline groups. This would indicate that the ability to discern forest
stand conditions in these mountains with TM data could be limited to
quantifying levels of mortality. [Table 1 lists correlation coefficients
between the TM digital values and field data for decline classes 1 and 4
resulting from the two methods of plot location.]
Table 1. Correlation Coefficients (r) Between Decline Classifications and
TM Digital Values for the Two Methods of Locating the Plots on the Images*.
Plot Location Method
Map Data Visual
TM Band*
1
IK
2
2K
3
3K
4
4K
5
5K
7
7K
Class 1
0.58*
0.33
0.36
0.28
0.28
0.24
0.34
0.37
0.23
0.22
0.14
0.08
Class 4
-0.47
-0.39
-0.37
-0.34
-0.47
-0.39
-0.30
-0.36
-0.25
-0.30
-0.20
-0.19
Class 1
-0.57*
-0.46
-0.46
-0.33
-0.60*
-0.43
0.09
0.26
-0.29
-0.21
-0.44
-0.41
Class 4
0.84**
0.70**
0.73**
0.61**
0.73**
0.66**
0.14
0.05
0.50*
0.49*
0.68**
0.66**
p<0.05 , *; p<0.01, **
Band 1, 450 to 520 nm; Band 2, 520 to 600 nm;
Band 3, 630 to 690 nm; Band 4, 760 to 900 nm;
Band 5, 1550 to 1750 nm; Band 7, 2080 to 2350 nm;
a band number followed by a "K" designates the mean density value for
a 3x3 cluster of pixels around plot center.
33
-------
The visual method produced a data set in which the correlations were
consistently higher than with that based upon map data. Not only were the
correlations produced by the two methods different in absolute values, even
the signs were often opposite. For example, data from the visual method
yielded a model which predicted a percent decline in class 4 which would
increase with increasing digital values. However, when estimated with a
model from data located with the map coordinates, the percent decline would
decrease with increasing digital values.
Thus in these mountains, where a stand of healthy spruce or fir
(class 1) can be but a short distance from one with a high proportion of
standing dead (class 4), accurate location of field plots on the images
will be critical to developing models predicting stand conditions with TM
digital data. A displacement of one or two 30-m pixels can radically alter
relations among variables (see Figure 1). Averaging pixel values around
the plot center would reduce some of the effects of stand heterogeneity,
but it also weakens correlations and, unless the plot centers are reliably
located, includes portions of the forest not associated, with the study
plots. Also, both methods were subjective. Location of the plots on the
maps was approximate, and neither method could be independently verified.
While the results from the visual approach appeared promising, they are
thus suspect and it could be expected that deletion or addition of a few
plots would alter the correlations subsantially.
A quantitative and definitive method of establishing the plot centers
is essential for modeling TM and forest data. Only then will there be
enough confidence in the data sets to begin addressing important questions
concerning the adequacy of plot size and the true effects of elevation and
aspect on stand conditions. Potential techniques for accomplishing this
include the use of NAVSAT devices which utilize satellite tracking systems
for the location of ground positions. An alternative approach would
involve the location of field plots on large-scale aerial photography
followed by the transfer of these locations to orthophoto quads.
CONCLUSIONS
Relying solely on presently established permanent field plots is
questionable considering existing plot size, TM pixel size, and r values.
Field plots of 3x3 pixels, selected randomly within homogeneous areas
at all damage levels, are needed for model development.
Statistical models for successful monitoring of changes in forest
conditions should be applicable to multidate normalized TM data.
34
-------
Image Coordinates
X
609
610
611
612
613
321 322 323 324 325
53
55
62
80
83
60
56
62
78
87
66
75
77
82
82
71
85
85
86
88
97
91
87
89
93
Figure 1. TM band 4 density values for a 5x5 cluster of pixels in the
spruce-fir forests of the Black Mountains.
35
-------
Accurate relocation of existing field plots by devices such as NAVSAT
could prove useful in finding the pixel blocks representing their spectral
characteristics.
Aerial photographs of appropriate
developing sampling strategies for field
assessing decline conditions.
scale could prove
data collection as
valuable in
well as for
Preliminary results with the two Landsat-TM scenes analyzed in this
study indicate that TM-based models can potentially predict the changes in
forest conditions subject to the above considerations.
REFERENCES
Bruck, R.I., and W.P. Robarge. Observations of Boreal Montane Forest
Decline in the Southern Appalachians. Aquatic Effects Task Group (F)
Peer Review - Research Summaries, 1984. pp. 425-433.
Bruck, R.I
Brockhaus,
Decline in
Carolina -
., W.P. Robarge, S.
A. McDaniel, and P.
the Boreal Montane
An integrated Forest
Khorram, W. Cure, S. Modena, J.
Smithson. Observations of Forest
Ecosystem of Mt. Mitchell, North
Response Approach. Proceedings of
the U.S.-F.R.G. Symposium on Forest Decline, Burlington, VT, Oct.
19-24, 1987. U.S. Forest Service Technical Publication, Broomall,
PA., 1987.
Rock, B.N.,
Hoshizaki.
36:439-445,
J.E. Vogelmann, D.L. Williams,
Remote Detection of Forest
1986.
A.F. Vogelmann, and T.
Damage. BioScience
36
-------
THE USE OF QUANTITATIVE REMOTE SENSING TECHNIQUES
TO ASSESS FOREST DECLINE DAMAGE IN VERMONT
Darrel L. Williams, Physical Scientist, NASA/Goddard Space Flight Center,
Earth Resources Branch/Code 623, Greenbelt, Maryland 20771, (301) 286-8860;
David W. Case, Senior Member of the Technical Staff, Science Applications
Research, Lanham, Maryland 20706, (301) 286-4864; and Emmett W. Chappelle,
Photobiologist, NASA/Goddard Space Flight Center, Greenbelt, Maryland
20771, (301) 286-6638
ABSTRACT
Since 1984, members of Goddard's Earth Resources Branch have conducted
research in cooperation with other institutions to determine if
quantitative remote sensing techniques can be employed to detect and assess
damage in spruce/fir forest stands located in the north-central portion of
Vermont. The main causal agents for the stressed condition of these forest
stands are believed to be acidic deposition and ozone. Both passive
spectral reflectance and/or active laser induced fluorescence (LIF)
measurements were made during the 1984-1987 growing seasons at the canopy,
branch, or needle level. The samples for these measurements were collected
from twelve red spruce stands which were selected to represent a range of
health conditions from low to high damage, or from seedlings reared under
different acidic and heavy metal regimes in a greenhouse. Analyses of
these data indicate that both reflectance and fluorescence spectra are
useful for differentiating between low and medium to high forest damage
classes. The intent of this paper is to provide a brief, narrative
description of Goddard's overall study effort in order to highlight the
various remote sensing techniques that were employed, and to discuss the
key observations and results that have been obtained to date.
INTRODUCTION
Since the 1960's, researchers have noted a decline of red spruce
(Picea rubens Sarg.) forests in the high elevation areas of the
northeastern United States. Researchers investigating this phenomenon have
suggested many causes for the decline of these forest areas, such as
periods of drought, insect outbreaks, and air pollutants, such as acidic
deposition (both wet and dry) and ozone. The principle causal agents for
this forest decline phenomenon are still being debated, but many scientists
37
-------
believe that the air pollutants listed above are primarily responsible for
the stressed condition of these forests (1, 2).
Site Selection and Study Objectives
One area where substantial dieback of red spruce stands has occurred
is on Camels Hump Mountain which is located in the high peaks region of
Vermont. In 1984, remote sensing specialists at the Goddard Space Flight
Center (GSFC) in Greenbelt, Maryland, the Jet Propulsion Laboratory (JPL)
in Pasadena, California, and researchers from the Universities of Vermont
and Maryland, initiated a cooperative project to assess the utility of
quantitative remote sensing techniques for monitoring stressed forest
areas. Because of the decline of red spruce stands located on the slopes
of Camels Hump Mountain, and the existence of a historical data base for
these stands, this area was chosen as the primary study site. The overall
objectives of this cooperative project were:
1. To determine if quantitative remote sensing techniques could be used
to differentiate low, medium and high damage red spruce stands; and,
if so,
2. Could these stands be monitored over time using second generation
satellite data, such as that provided by the Landsat Thematic Mapper
(TM), to determine the rate of decline and to assess the areal extent
of the problem?
Institutional Roles and Funding
GSFC researchers focused their research activities primarily on
Objective #1, while the larger JPL contingent addressed both objectives.
The Universities of Vermont and Maryland provided both field and laboratory
support during these investigations. Funding for this research was
provided by the Land Processes Branch within the Earth Sciences
Applications Division, Office of Space Science and Applications, NASA
Headquarters, Washington, D.C.
Period of Performance and Type of Data Collected
During the latter part of the 1984 - 1987 growing seasons, GSFC
researchers collected various types of remote sensing data for 12 red
spruce stands located on and near Camels Hump Mountain. Two techniques
were employed to obtain "passive" spectral reflectance data of these
stands, or representative components of these stands (i.e., branches): (a)
in situ canopy spectral reflectance data were acquired by hovering over
each stand in a helicopter which was equipped with a spectro-optical system
consisting of a high spectral resolution spectrometer, a broad-band
radiometer, 35 mm flight research cameras, and a color video camera; and
(b) lab-based spectral reflectance data were acquired for branch samples
from these stands using the same set of instruments and a hemispherical
illumination system developed for this project. These spectral instruments
provide coverage of the visible, near infrared (NIR) and short wave
38
-------
infrared (SWIR) regions of the electromagnetic spectrum (EMS), between
approximately 0.4 to 2.5 microns.
Laser induced fluorescence (LIF) data, an "active" form of remote
sensing, were also obtained in the lab for branch samples collected from
these stands. Both types of data (i.e., "passive" spectral reflectance and
"active" LIF) were also acquired for red spruce seedlings which were
germinated and raised in a controlled greenhouse environment. These
seedlings were treated with various combinations of acidic solutions (pH 3,
4, and 5) and heavy metal concentrations (Al, Cu, Pb, and Zn) to ascertain
how their reflectance and fluorescence characteristics changed as a
function of the treatment received.
Status of Data Analyses
Analyses of the spectral data obtained for these spruce stands have
yielded results which indicate that distinct differences in the reflectance
characteristics for healthy versus damaged sites do exist -- an air of
cautious optimism is warranted. However, as with any research where one is
pushing into uncharted territory, each new answer often leads one into a
new series of questions which need to be addressed to fully understand the
earlier results. Thus, more work is needed before definitive statements
can be made, particularly with regard to assessing damage over widespread
geographical regions using data collected solely by spaceborne sensors.
A number of publications have already appeared in (or been submitted
to) the scientific literature which discuss various aspects of this
cooperative project (3,4,5,6,7,8,9,10,11,12). In the remainder of this
report, we provide a brief description of the study area(s) and summarize
the highlights of the in situ and laboratory-based research conducted by
Goddard scientists.
STUDY AREA
As previously noted, data were collected for 12 red spruce stands.
Five of these stands were located on Camels Hump, with the remaining seven
sites located approximately 15 - 30 kilometers south of Camels Hump. A few
of these sites were located in the northern portion of the Green Mountain
National Forest.
The development of understory vegetation at these sites varied
depending on the terrain and overstory conditions. Stands found on
high-elevation, steep slopes, or where dieback was prevalent, had an
understory composed of ferns, young spruce trees, and a substantial amount
of young hardwoods that were able to compete due to the increased
filtration of sunlight to the forest floor. Several sites also contained a
small percentage of balsam fir. Exposed rock and soil were also
commonplace at these sites. For the sites located at lower elevations, or
on flatter terrain, the overstory conditions were generally healthier and
the understory was mainly composed of ferns and young spruce and fir
seedlings.
39
-------
The 12 sites were chosen to represent a gradient of stress conditions
from low to high damage. In the summer of 1984, each stand was visited and
assigned a damage rating based on the amount of spruce foliar damage
readily apparent (see Table 1 later in report). Because this rating system
did not .involve any direct quantitative measure of plant vigor (such as the
rate and efficiency of photosynthesis), it is more appropriate to group the
sites into three broad categories (i.e., low, medium and high damage sites)
when making comparisons between sites.
SPECTRAL DATA COLLECTION
Spectron Engineerings' (SE) 590 spectroradiometer and the Barnes
12-1000 Modular Multiband Radiometer (MMR)1 were the two instruments used
to collect the spectral reflectance data for this project. The SE590 is a
high resolution spectroradiometer with 252 data detectors (channels), each
with a spectral resolution of approximately .0025 microns (/im) or 2.5
nanometers (nm). The SE590 has a spectral sensitivity ranging from 0.37 to
1.113 /im. The MMR is a broad-band radiometer with eight discrete data
channels that duplicate the Landsat TM instrument in spectral coverage,
plus one additional channel in the NIR which covers the 1.15 to 1.30 jtm
region. The techniques which were employed to collect data using these
instruments are described in the following sections of this paper.
In situ Data
In situ spectra of the spruce canopies were collected by mounting the
spectrometers on a helicopter and hovering over each stand. In addition to
the two spectrometers, two 35 mm flight research cameras (with normal and
telephoto lenses) and a color video camera were included on the instrument
package. All five devices were mounted on the nose of the helicopter as
shown in Figure 1, and were triggered simultaneously during data
acquisition. Both the SE590 and the MMR were fitted with lenses having a
1° instantaneous field of view (IFOV). This IFOV, coupled with a nominal
hovering altitude of 300 meters (1000 feet) above the ground, resulted in a
ground sample resolution of approximately 5.5 meters (18 feet).
The results derived by analyzing the in situ spectra collected in 1984
were promising. Figure 2 is a plot of mean spectral reflectance for a high
and medium damage site located on Camels Hump Mountain. These data were
obtained under ideal solar illumination conditions during the afternoon of
August 17, 1984. Of interest is the difference in the location of the
point of minimum reflectance (maximum absorption) at approximately 670 nm.
1. The reference to a manufacturer does not imply endorsement by the
National Aeronautics and Space Administration (NASA).
40
-------
Figure 1. NASA Helicopter used to obtain in situ spectra. Inset shows
the instrumentation mounted on the nose of the helicopter.
This area of the reflectance curve is referred to as the chlorophyll
absorption region, and its location within the EMS has been found to be
quite consistent for nearly all green vegetation. The point of minimum
reflectance for the high damage site was at 658 nm, whereas the point of
minimum reflectance for the moderate damage site was at 669 nm; a
Terence of approximately 11 nm towards the shorter wavelengths of the
This significant shift in the position of the chlorophyll absorption
•eature is referred to as the "blue shift," because the absorption feature
towards the shorter "blue" wavelength portion of the EMS. In
geobotanical remote sensing studies, this blue shift phenomenon has been
Pressed forest vegetation growing in areas where the soil was
known to contain high concentrations of heavy metals (13). Thus, it is
worth noting that over an 11 year period, researchers documented an
increase in the concentrations of copper, lead, and zinc in soils from
Camels Hump Mountain of 32 percent, 95 percent, and 48 percent,
respectively (14).
41
-------
40
35
30
25-
(D
O
OB
20
10 -
Plot of Mean Reflectance ± One Standard Deviation
Camels Hump High Damage Site
(Elevation - 945m; % Spruce Damage = 76.0)
Plot of Mean Reflectance ± One Standard Deviation
Camels Hump Medium Damage Site
(Elevation - 840m; % Spruce Damage = 34.9)
Point of Minimum Reflectance
is 0.669um, with a Mean
Reflectance of 2.34%
Point of Minimum
Reflectance is
0.658um, with a Mean
Reflectance of 2.88%
0.6 0.7 0.8
Wavelength (urn)
Figure 2. Mean spectral reflectance curves for a high damage site
versus a medium damage site found on Camels Hump. Note the
point of minimum reflectance differs between the two sites by
11 nanometers (0.658 /im for the high damage site versus 0.669
Urn for the medium damage site).
The blue shift phenomenon was also observed in ground-based spectral
data collected by JPL scientists coincident with the 1984 overflights.
They found a 9 nm shift in spectral reflectance curves of needle samples
taken from high versus low-to-medium damage sites. These researchers also
took Scholander pressure bomb data of branch samples from these sites. The
pressure bomb data indicated that the high damage sites were under greater
water stress than the low-to-medium damage sites. Differences in canopy
water status were also seen in SWIR reflectance data taken by these
researchers (5), as well as in the SWIR data acquired from the helicopter
platform using the Barnes MMR radiometer.
Based on the success of the 1984 field data collection, attempts were
made in 1985 and 1986 to obtain in situ canopy reflectance data using the
42
-------
helicopter system. For both years the weather conditions were not optimal
(persistent cloudy-to-overcast conditions) during the two weeks in late
August and early September when the collection periods were scheduled.
Hence, no additional canopy reflectance data were collected to corroborate
the "blue shift" or "canopy dryness" phenomenon seen in the in situ data
collected in August 1984.
Laboratory Measurements of Branch Samples
Due to weather constraints (i.e., the lack of cloud-free days), and
the desire to more precisely compare spectral reflectance data between
sites or for the same site from one year to another, a transportable source
of artificial illumination was devised and built at Goddard between the
1984 and 1985 collection periods (see Figures 3a and 3b). This
illumination system permitted one to take spectral reflectance data
in-doors under consistent, reproducible illumination and viewing angle
conditions. For a detailed description of the transportable hemispherical
illumination system (THIS) that was developed, the reader should consult
Williams and Wood (11) and Williams et a7. (12).
Beginning in 1985, the "THIS" illumination system was transported to
Vermont annually and used to obtain spectral reflectance data of branch
samples collected from each of the 12 study sites using the following
procedures. For each site, three to five branches were excised- from each
of five randomly selected trees. Branches were collected from limbs that
were at (or nearly at) the canopy level of the stand. The samples were
immediately placed in plastic bags that contained a moistened paper towel,
and the bags were sealed. The sealed bags were placed in coolers, on ice,
and transported back to the lab where they are placed in a walk-in cooler.
The spectral reflectance data were acquired within 24 to 48 hours of when
the branches were excised from the trees.
When the spectral data were being taken, the branch samples for a
given site were randomly selected from the bags and stacked on a board that
was covered with an opaque canvas cloth which had a nominal reflectance of
five percent across the visible, NIR, and SWIR regions of the EMS. The
branches were stacked on the board to insure that the canvas cloth was
completely covered and so that they created a uniform layer of target
material. See Williams et a/., (12) for a more complete discussion of
lab-based data acquisition.
As previously noted, the THIS illumination system permits one to make
measurements under reproducible illumination and viewing geometry
conditions. Thus, this facilitated the direct comparison of spectral
reflectance data between and within sites for the same year or from year to
year. Comparing 1985 reflectance data against 1986 data has proven to be
particularly interesting. For the months of May, June, July, and August of
1985, the average precipitation was normal. However, no measurable
precipitation fell during the first 15 days of August -- the period just
prior to when the branch samples were collected. For the same four months
of 1986, the average rainfall was 39%, 21%, 32%, and 50% above normal,
respectively, with no periods of longer than 5 days without rain.
43
-------
(a)
(b)
RANSPORTABLE HEMISPHERICAL
ILLUMINATION SYSTEM (THIS)
30" ALUMINUM HEMISPHERE
SUPPORT STRUCTURE
HEMISPHERE APERTURE (12")
ADJUSTABLE TRIPOD COLUMN/HEAD
POWER SUPPLIES
BARIUM SULFATE REFERENCE PANEL
OPTICAL HEAD FOR SE-590
BARNES MMR RADIOMETER
9.
10.
11.
12.
13.
14.
15.
16.
35 MM CAMERA
STORAGE CASE FOR HEMISPHERE
STORAGE CASE FOR SUPPORT STRUCTURE
P.C. FOR DATASTORAGE/ MANIPULATION
SE-590 CONTROLLER UNIT
DATA STORAGE DEVICE FOR MMR
BACK-DROP PANEL/ 5% REFLECTANCE
MUFFIN FAN FOR COOLING
Figure 3.
a) Photograph of THIS, with the various component parts labeled
and identified, b) Photograph illustrating the flexibility of
allows one to mount the hemisphere vertically or
itally depending on the needs of the researcher.
l this PerV°d. in 1986» °zone concentrations measured at the
International Airport in Burlington, Vermont, located within 15
' 60 ppb) were significantly less in 1986 (i e
37 hours in 1986 vs. 209 hours in 1985; Rock et al . (7)) Since
researchers have theorized that both drought stress and pnotooxidant
pollutants may be two key factors responsible for forest decline. It Is
plausible to assume that reduced exposure to ozone, coupled with moist
summer conditions, may be conducive to hardy growth in the spruce forests
44
-------
in Vermont. Could the favorable growing conditions encountered in 1986
promote a recovery of the highly damaged spruce sites being monitored for
the project? Based on the mean reflectance spectral curves for selected
sites for 1985 versus 1986 shown in Figures 4 a-c, the answer would seem to
be "yes."
Figures 4 a-c illustrate the mean spectral reflectance for branches
from a low damage site (Fig. 4a), a medium damage site (Fig. 4b), and a
high damage site (Fig. 4c), for August 1985 versus August 1986. In Figure
4a, there is very little difference in reflectance between the two years,
especially in the visible wavelengths. This is what one would expect
because a healthy site has nothing to recover from, therefore it can be
treated essentially as a "control." According to our laboratory notes,
branches from the moderate damage site, Fig. 4b, were noticeably chlorotic
or yellowish in 1985, but not in 1986. This is corroborated by the
spectral reflectance curves; notice the higher reflectance in 1985 along
the slope between the maximum reflectance at 545 nm and the chlorophyll
absorption well at 675 nm. This is indicative of yellowish (chlorotic)
vegetation. Inspection of the visible wavelengths in Fig. 4c also
indicates that more chlorotic conditions existed in 1985 than in 1986.
Again, this compares quite favorably with laboratory notes for these branch
samples.
The change in NIR spectral reflectance between 1985 and 1986 is the
most dominant feature of these plots, particularly for Figures 4b and 4c.
This may be indicative of stress due to a combination of the lack of
rainfall in early to mid August, 1985 and longer periods of exposure to
high levels of ozone. Spectral reflectance in the NIR region is dominated
by the intercellular structure of the leaves or needles. Researchers who
have worked with the spectral reflectance of stressed or dehydrated
vegetation have documented that NIR reflectance can first increase and then
decrease. Gausman (15) cites several works that have documented an
increase in the NIR reflectance of plant leaves as they have dried out.
Westman and Price (16) have shown that the NIR reflectance of pine needles
first increased, then decreased, as the moisture content of the needles
went from 100% to 48%. Below 48%, they noted a substantial decline in the
NIR reflectance. Both researchers have concluded that the change in NIR
reflectance is related to the change in the leaf cell structure (cell
volume, air spaces) as the leaf drys out, and that more work on the NIR
reflectance of dehydrating leaves is necessary.
As stated previously, no measurable precipitation was recorded during
the first 15 days of August, 1985, total precipitation for the month was
1.2 inches below the normal average, and several hours of moderately high
levels of ozone were recorded during this period. Based on these data, and
the fact that August is one of the warmest months of the year, it would
seem that in 1985, vegetation in the region may have been under stress.
Conversely, in August 1986, the average monthly rainfall was approximately
1.95 inches above the 30 year average for the month, and the heavy
precipitation throughout the 1986 growing season resulted in chemically
"cleaner" rainfall and atmospheric conditions (7).
45
-------
—1865 Data
— 1M6Dota
—1085 Data
—1986 Data
OJO
0.5
0.6
Wavelength (/im)
Figure 4. Mean reflectance data for 1985 versus 1986 for (a) a low damage
site, (b) a medium damage site, and (c) a high damage site.
Thus, if the spruce trees of this region were experiencing an elevated
level of stress in August 1985 due to the adverse moisture and atmospheric
conditions which were known to have existed, this may have affected the
cell structure of their needles, leading to an increase in NIR reflectance.
Note in Figures 4 a-c, that all three damage levels experienced an increase
in NIR reflectance in 1985 compared to 1986. Notice also that the smallest
46
-------
difference in NIR reflectance occurred for the low damage site (Fig. 4a).
The low damage site was situated on fairly level terrain at a lower
elevation where the soil was deeper and moisture retention was greater than
when compared to the other two sites, which were located at higher
elevations, on steeper terrain, where the soils were generally very
shallow. Thus, the small interannual difference in the NIR reflectance for
the low damage site in comparison to the greater interannual differences
for the medium and high damage sites is consistent with the water stress
trends that one would expect. Again, more research and monitoring is
needed before definitive statements can be made.
LIF Measurements
Another approach which is offering promise as a means of detecting and
assessing the extent of damage suspected to be caused by atmospheric
deposition is the use of laser induced fluorescence measurements. This
method exploits changes in the fluorescence of constituent pigments in
plants to detect changes in the physiological status of the plant. The
primary pigment involved in these fluorescence changes is chlorophyll a.
The relationships between photosynthesis, chlorophyll concentration,
accessory pigments, and the magnitude of fluorescence at certain
wavelengths underlies the use of LIF for monitoring plant vigor. These
relationships are described in detail by Chappelle and Williams (4).
Briefly, however, there is an inverse relationship between photosynthetic
efficiency and the fluorescence of chlorophyll a and accessory pigments.
Thus, under conditions which adversely affect photosynthesis, there is an
increase in the magnitude of fluorescence at wavelengths corresponding to
compounds involved in photosynthesis. These changes in fluorescence may be
modulated, however, by changes in the concentration of these compounds.
Significant changes have been observed in the LIF spectra of vegetation
subjected to drought and nutrient stress (4).
The excitation source used in these studies was a Molectron UV-22
pulsed nitrogen laser emitting at 337 nm, pulsed at 30 Hz. The
fluorescence intensity was measured using a red sensitive gallium arsenide
photomultiplier with the signal being captured by the use. of a gated boxcar
integrator. The signal was fed by way of an A/D converter into a computer
where real time spectral data were generated and stored for subsequent
analysis.
LIF measurements were made on branch samples taken from 11 of the 12
red spruce sites described earlier. The spectra which were obtained were
typical for conifers in that fluorescence bands were observed at 440, 525,
and 740 nm. It was noted, however, that frequently the fluorescence
spectra of branches from high damage sites contained a band (or region of
higher response) at 685 nm (Figure 5). This band, which is normally
observed in herbaceous dicots and monocots, is the fluorescence maxima of
the chlorophyll a species most closely associated with photosystem II. It
is speculated that in healthy conifers, the transfer of electrons from this
species to the species which fluoresces at 740 nm is so efficient that none
of the excitation energy is dissipated as fluorescence, and, therefore, it
does not show up in the fluorescence spectra of healthy conifers.
47
-------
6000
4000
§ 2000
0
LL
DC
-2000
RED SPRUCE
HIGH DAMAGE
X
LOW DAMAGE
"350 400 450 500 550 600 650 700
WAVELENGTH IN NANOMETERS
750 800
Figure 5. Effect of acid rain damage on LIF spectra of certain red spruce
samples.
In addition, the LIF spectra were analyzed using a number of ratio
algorithms to determine those band ratios that best correlated with the
stress-index rankings. The fluorescent bands which related best to the
stress-index ranking of the various sites were at 440 nm and 525 nm, as
shown in Table 1. The data summarized in Table 1 indicate that all of
these bands or band combinations do an excellent job of differentiating the
higher damage sites from those with moderate-to-low damage. The apparent
sensitivity of the 440 nm band to changes in stand vigor has interesting
ramifications. We have found that tannic acid fluoresces in the vicinity
of 440 nm. It has been shown by Waring et a/., (17) that the
concentrations of tannic acid and lignin in vegetation increase as an
inverse function of nutrient availability in the presence of high light
flux. As greater damage was seen at the higher elevations where the trees
(in the absence of fog and clouds) would be subjected to a higher overall
light flux, as well as a greater percentage of UV radiation, the
accumulation of increased levels of tannin and lignin is possible.
However, an increase in the concentration of other compounds which
fluoresce in the 440 nm region,
be ruled out.
e.g., vitamin K, and plastoquinone, cannot
48
-------
Table 1. Comparison of Stress-index Value Ranking,
Rankings Based on LIF Criterion for Red Spruce
Elevation, and
Site Name
Rank
Stress-index
value1 Elev (m)
LIF Ranking by Band
440 nm 525 nm 490 nm2
Robert Frost
Rt. 73
Beaver Pond
CCC
South Tom
Swimming Hole
Cellar Hole
Banforth Ridge
Lower H.W.
Mt. Abraham
Upper H.W.
1
2
3
4
5
6
7
8
9
10
11
21
21
25
32
35
38
38
40
43
44
47
442
305
564
396
732
518
548
838
823
945
945
2
5
3
4
7
6
1
9
8
11
10
1
6
2
3
8
7
4
5
9
11
10
1
5
4
3
7
6
2
9
8
11
10
The stress-index value ranking of the sites was based on an assessment
of the percentage of dead branches within the live crown of -the spruce
trees at a given site. For comparative purposes, it may be more
appropriate to group the study sites into three broader categories: low
damage sites (i.e., stress indices < 30, ranks 1-3), medium damage
sites (i.e., stress indices > 30 but < 40, ranks 4-8), and high damage
sites (i.e., stress indices > 41, ranks 9 - 11). See Donnelly et a/.,
(18) for a more in-depth explanation of the derivation of the
stress-index value.
The 490 nm column does not represent the existence of a fluorescence
peak at 490 nm; it represents the RFI value at the isobestic of 440 and
525 nm, which occurred at 490 nm.
Recent studies have also shown a correlation between the relative
fluorescence intensity of the 440 and 525 nm bands to the rate of
photosynthesis. We believe that these "blue" fluorescent bands are due to
"Q" agents -- compounds which are involved in the electron transfer
occurring during photosynthesis. Investigations currently underway are
pointing to a relationship between these bands and certain bands seen in
passive reflectance spectra. These relationships may provide the basis for
the selection of reflectance bands in the visible, NIR and SWIR regions
which are highly correlated to photosynthetic efficiency.
49
-------
Spectral Reflectance Data of Spruce Seedlings
Another important aspect of the cooperative project was the
germination and rearing of red spruce seedlings under controlled greenhouse
conditions. It was felt that this was the only way to accurately document
the effects of different types of environmental stresses on the spectral
reflectance characteristics of red spruce. Since atmospheric deposition is
a regional phenomenon, there is no such thing as a true "control" forest
stand to make comparisons with. Thus, the collection of spectral data of
greenhouse-grown red spruce seedlings which were treated with aqueous
solutions containing prescribed amounts of specific heavy metals, at
prescribed pH levels, became an important aspect of the study.
The seedlings were raised for one year before application of the
aqueous solutions (to the potting medium) began. The four metals used in
solution were aluminum, copper, lead and zinc. They were applied at four
different concentrations, at three different pH levels (Table 2). The
different combinations of metals (4), metal concentrations (4), and
solution pH (3), plus a "control," resulted in 49 different treatments
There were 5 replicates of each treatment, and
of 7 seedlings. Thus, a total of 245 seedling
replicates per treatment) or 1715 seedlings (245
experiment. The treatments were applied over a 5
(i.e., 4x4x3, plus 1)
each replicate consisted
groups (49 treatments x 5
x 7) were included in the
month period. At the conclusion of the experiment, the seedlings were
harvested and a variety of morphological measurements were made on the
plants by the JPL research staff.
Table 2.
Summary of Experimental Design and Heavy Metal Solutions
Used in the Red Spruce Seedling Experiment
Experimental Design:
- 3 pH's (3, 4, 5)
- 4 metals (Al, Cu, Pb, Zn)
- 4 levels per metal (ppm)
- 5 replicates per treatment
- 7 plants per replicate
- plus, 5 control replicates
Heavy Metal Solution
Aluminum Chloride (A1C1J
Copper Sulfate (CuSOJ
Lead Nitrate (PbNO,)
Zinc Sulfate (ZnSCg
Concentration level in parts per million (ppm)
Lowest to Highest Amounts
0.003 0.03 3.0 300.0 (30.0)*
0.1 1.0 2.5 5.0 (3.0)*
0.2 2.0 20.0 200.0
0.01 1.0 10.0 100.0 (20.0)*
Concentrations in ppm were reduced to these levels at mid-point of
treatment period to lessen the probability of premature death.
50
-------
Spectral reflectance data were taken after three months of treatment,
in late June 1986, and again in late August 1986, after five months of
treatment. Both SE590 and MMR spectra were taken in the lab using the
hemispherical illumination system described earlier. In both June and
August, spectral data were collected of each treatment replicate by viewing
the side of the seedling "canopy" using the hemisphere mounted in the
vertical position to provide a horizontal beam of illumination (Figure 3b).
A total of twelve scans were made; the first two scans were of a barium
sulfate reference panel and the remaining ten scans were of the seedling
clump. After the third and sixth scans, the seedlings were shifted
slightly to the left or right to insure that a larger percentage of the
total seedling canopy was characterized. Mean reflectance and standard
deviation statistics were then derived.
In August, an additional set of spectral reflectance data were
obtained by pooling all 5 replicates of a given treatment together to
create a "miniature canopy" consisting of 35 seedlings. For these
measurements, the illumination and viewing angle were from above; e.g.,
Figure 3a.
Also during the August measurement period, LIF and photosynthesis data
were taken for seedlings randomly selected from a subset of the total
number of treatments. Due to time and funding constraints, these
measurements were made for 1 seedling from each of 3 replicates for 3 heavy
metals (Al, Cu, and Zn) at two levels per metal for both pH 3 and 4
solutions as follows: (1) Al applied at 0.003 mg/1 and 3.0 mg/1; (2) Cu
applied at 0.1 mg/1 and 2.5 mg/1; and (3) Zn applied at 0.01 mg/1 and 10.0
mg/1. For a more detailed summary of the experimental design and the
photosynthesis measurements, see Donnelly and Shane (19).
Only preliminary analyses of the seedling spectral reflectance and
LIF data have been performed at Goddard at this time, but more detailed
analyses are planned if the chemical analyses for the seedling tissue
material collected during the August harvest activity can be completed.2
However, results to date indicate that there are differences in the
spectral reflectance characteristics of seedling groups toxic level of
metal solution (Figure 6). The spectral data shown in Figures 6a and b are
for seedling groups that were given zinc sulfate (ZnSO,); the plants
represented in Fig. 6a received a dosage of 10 parts per million (ppm) in a
solution having a pH of 3, while the plants represented by Fig. 6b received
a dosage of 20 ppm in a solution having a pH of 4. It is readily apparent
from the spectral reflectance plots that ZnS04 at 20 ppm and a pH of 4 was
very toxic to the seedlings. In June 1986, after 3 months of treatment,
2. (Note: The chemical analyses for the entire seedling experiment, which
are needed to determine the extent to which the heavy metals were uptaken
by the various component parts of the seedlings, were to be performed by
members of the JPL research team, along with the summary of the
morphological parameters. However, funding for the chemical analysis phase
of the study has not been made available,- and it is impossible to make
reasonable conclusions as to the exact nature of the interactions between
metal uptake and spectral reflectance without this data.)
51
-------
many of the needles on the individual seedlings were chlorotic, and by
August, all of the needles were chlorotic or had fallen off the stems of
the plants. For the seedling group that received only half the amount of
ZnSO,, but in a more acidic solution (i.e., Fig. 6a), the treatment had an
adverse affect on seedling health, but not as dramatic as the higher
metal/less acidic dosage. The number of chlorotic needles had increased
between June and August, which is apparent due to the increase in
reflectance in the yellow-red region of the spectral reflectance plot for
the seedlings in August versus June (Fig. 6a). Similar trends were
observed for other treatments—a gradual deterioration in seedling health
was noticed, usually manifested by an increase in the number of chlorotic
needles or by the number of needles that fell from the stem of the plant.
These trends were recorded in the spectral reflectance data; most
reflectance plots had an increase in the yellow-red region, and a
"rounding" in the shoulder region associated with the rapid increase in
reflectance in going from the chlorophyll absorption region at 680 nm, to
the NIR region of the reflectance curve.
In lieu of a complete set of chemical analysis data, an attempt is
being made to correlate spectral reflectance and LIF measurements with
chemical analyses performed on the subset of seedlings included in the
LIF/photosynthesis experiment. Statistical analyses of the photosynthesis
data, which were acquired using an infrared gas analyzer, have shown that
net photosynthesis (PS), averaged over all treatments, was 2.74 /jmol/m2/s,
and that net PS of pH 4 seedlings (3.06 /imol/m2/s) was significantly higher
than net PS of pH 3 seedlings (2.42 /imol/m2/s). However, no significant
differences were observed in the photosynthetic rate between metals or
among treatment levels of a given metal at a given pH (19).
40-.
30-
20
10
— JucuDota
— Aug. Ma
40-
30
20-
10-
— JurwData
— Aug. Data
500 600 700 BOO
Wovdingth (nm)
BOO 1000
500
600
-H-
-f-
700 BOO
Wavelength (nm)
(b)
900
1000
Figure 6. (a) Plot to illustrate changes in spectral reflectance as a
function of the length of time since treatments were initiated
(3 months vs. 5 months); this treatment consisted of 10 ppm
zinc at pH 3; (b) Same as (a), but treatment consisted of 20
ppm zinc, at pH 4. This treatment had devastating effects on
plant health.
52
-------
SUMMARY
The NASA-funded cooperative research activity that focused on
assessing the utility of quantitative remote sensing techniques for
delineating and monitoring forest decline damage in Vermont believed to be
caused by atmospheric deposition has been very successful. Analyses of the
spectral, morphological, physiological, and chemical data collected by the
scientists involved in the project have yielded results which indicate that
distinct differences exist in the reflectance characteristics of healthy
versus medium-to-high damage classes. A great volume of data remains to be
analyzed, and funding to complete the critical chemical analyses of the
seedling tissue material is needed before the full value of the controlled
seedling experiment can be realized. However, based on results to date,
the future role of remote sensing in delineating, assessing, and monitoring
forest/vegetative decline damage over widespread geographical regions looks
promising, especially as new, improved sensors systems and platforms are
introduced.
Those interested in obtaining a more complete understanding of the
total project are encouraged to obtain and read the numerous publications
referenced in this report, and to contact directly the scientists involved
in this cooperative effort.
REFERENCES
1. Johnson, A.M. and T.C. Siccama. Acid Deposition and Forest Decline.
Environmental Science and Technology. 17:294, 1983.
2. Vogelmann, H.W., G. Badger, M. Bliss, and R.M. Klein. Forest Decline
on Camels Hump, Vermont. Bulletin of The Torrev Botanical Club.
112:274, 1985.
3. Case, D.W. and D.L. Williams. Obtaining Spectral Reflectance Factor
Measurements of Stressed Forest Vegetation. In: Proceedings of the
Fall Convention of The American Society of Photogrammetry and Remote
Sensing. American Society of Photogrammetry and Remote Sensing. Falls
Church, Virginia, 1987. p. 150.
4. Chappelle, E. and D.L. Williams. Laser Induced Fluorescence (LIF)
from Plant Foliage. IEEE Transactions On Geoscience and Remote
Sensing. 25:726, 1987.
5. Rock, B.N., D.L. Williams, and J.E. Vogelmann. Field and Airborne
Spectral Characterization of Suspected Acid Deposition Damage in Red
Spruce (Picea rubens) from Vermont. In: Proceedings of The Eleventh
53
-------
International Symposium on Machine Processing of Remotely Sensed
Data. Purdue University. Lafayette, Indiana, 1985. p.71.
6. Rock, B.N., J.E. Vogelmann., D.L. Williams, A.F. Vogelmann and T.
Hoshizaki. Remote Detection of Forest Damage. Bioscience. 36:439.
1986.
7. Rock, B.N., T. Hoshizaki, D.L. Williams., and R. Poirot. Evidence of
Red Spruce Recovery On Camels Hump. Submitted to Nature.
8. Vogelmann, J.E. and B.N. Rock. Assessing Forest Decline in
Coniferous Forests of Vermont Using NS-001 Thematic Mapper Simulator
Data. International Journal of Remote Sensing. 7:1303, 1986.
9. Williams, D.L., S.N. Coward, and C.L. Walthall. Collection of In Situ
Forest Canopy Spectra Using a Helicopter: A Discussion of Methodology
and Preliminary Results. In: Proceedings of the Tenth International
Symposium on Machine Processing of Remotely Sensed Data. Purdue
University, West Lafayette, Indiana, 1984. p. 94.
10. Williams D.L. Remote Sensing: A Tool in Acid Rain Research. Forest
Industries, 113:31, 1986.
11. Williams, D.L. and F.M. Wood, Jr. A Transportable Hemispherical
Illumination System for Making Reflectance Factor Measurements.
Remote Sensing of Environment. 23:131, 1987.
12. Williams, D.L., F.M. Wood, Jr., and D.W. Case. Acquisition of
Spectral Reflectance Data Using an Artificial Source of Hemispherical
Illumination. In: Proceedings of the SPIE, Vol. 924, Recent Advances
in Sensors, Radiometry, and Data Processing for Remote Sensing, 1988.
13. Collins, W.S., H. Chang, G. Gaines, F. Canney, and R. Ashley.
Airborne Biogeochemical Mapping of Hidden Mineral Deposits. Economic
Geology. 78:737, 1983.
14. Friedland, A.J., A.H. Johnson., and T.G. Siccama. Trace Metal
Content of the Forest Floor in The Green Mountains of Vermont:
Spatial and Temporal Patterns. Water. Air, and Soil Pollution,
21:161, 1984.
15. Gausman, H.W. Plant Leaf Optical Properties in Visible and Near
Infrared Light. Monograph No. 29, Texas Tech University, Austin,
Texas, 1985. 78 pp.
16. Westman, W.E. and C.V. Price. Spectral Changes in Conifers Subject
to Air Pollution and Water Stress: Experimental Studies. IEEE
Transactions on Geoscience and Remote Sensing. 26:11, 1988.
54
-------
17. Waring, R.H., A.H.S. McDonald, S. Larsson, T. Ericsson, A. Wiren, A.
Erisson, T. Lohammar. Differences in Chemical Composition of Plants
Grown at Constant Relative Growth Rates With Stable Mineral
Nutrition. Oecolgia (Berlin). 66:157, 1985.
18. Donnelly, J.R, J.B. Shane, D.R. Bergdahl, J.C. Clausen, R.A. Gregory,
and B. Wong. A Preliminary Assessment of Red Spruce Vigor as Related
to Physiographic Characteristics In Vermont. Northeastern Environ
mental Science. 4:18, 1985.
19. Donnelly, J.R. and J.B. Shane. Photosynthesis of red spruce
seedlings following application of solutions differing in pH and
heavy metal content. Final report submitted to NASA/GSFC under grant
NAG 5-791, 1987. 10 pp.
List of Acronyms
Al -
Cu -
EMS
GSFC
IFOV
JPL
LIF
aluminum
copper
- electromagnetic spectrum
- Goddard Space Flight Center
- instantaneous field-of-view
- Jet Propulsion Laboratory
laser-induced fluorescence
m - meter
MMR - modular multiband radiometer
NIR - near infrared
nm - nanometer
/urn - micron, or micrometer
Pb - lead
ppb - parts per billion
ppm - parts per million
PS - photosynthesis
SE - Spectron Engineering
SWIR - shortwave infrared
THIS - transportable hemispher-
ical illumination system
TM - Thematic Mapper
Zn - zinc
55
-------
REMOTE MAPPING AND MONITORING
OF FOREST DAMAGE
Barrett N. Rock, James E. Vogelmann, and Nancy J. Defeo, Institute for the
Study of Earth, Oceans, .and Space, Science and Engineering Research
Building, University of New Hampshire, Durham, New Hampshire 03824
ABSTRACT
Recent research has shown that remote sensing techniques are able to
accurately detect, quantify, map and monitor damage in conifer species. At
this time, however, remote sensing data cannot provide information
regarding the cause of damage. Several remote sensing studies currently
underway at the University of New Hampshire have been undertaken to supply
data which may relate to specific cause and effect issues. Results of
these studies indicate that:
1. A spatial pattern of inferred conifer damage exists in the
northeastern U.S. Based on Landsat Thematic Mapper data, the levels
of damage were found to be highest in the Adirondack Mountains in New
York, intermediate in the Green Mountains of Vermont and lowest in
the White Mountains of New Hampshire. This corresponds well with
spatial patterns of wet deposition pH which indicate lowest pH
(highest acidity) values in the Adirondacks, becoming less acidic
toward the east, with highest pH (low acidity) readings measured in
the White Mountains. Other pollutants, including ozone, may also
exhibit similar spatial patterns.
2. Near infrared (NIR) reflectance for forest stands in the Green
Mountains of Vermont changed between the years 1973 and 1984 as
measured by the Landsat Multispectral Scanner. Forest communities in
the transition zone had the largest decreases in near infrared
reflectance over the 11 year period. This decrease in NIR
reflectance is thought to be related to decreases in green leaf
biomass and increases in dead branches/trees that have occurred over
the time period. Based on extensive studies conducted on red spruce
(Picea rubens) from Camels Hump, Vermont, such decreases in NIR
reflectance are related to visual decline damage symptoms such as
foliar loss.
3. High-spectral resolution in situ data can be used to identify fine
absorption and reflectance features in forest community members
exhibiting various levels of morphological damage related to forest
56
-------
decline. However, specific causal agents such as sulfur and/or
nitrogen derived pollutants have not yet been correlated with
detailed reflectance measurements. It is hoped that in the future,
high-spectral resolution sensors will provide information regarding
specific cause and effects.
INTRODUCTION
The northeastern United States has been experiencing a decline in red
spruce and balsam fir since 1960 (1). At present, no specific causes have
been identified as responsible for the decline. The Forest Response
Program of the U.S.D.A. Forest Service has asked several specific questions
in order to better clarify cause and effect issues. The specific questions
for which remote sensing studies may provide input are as follows:
Are changes in growth and mortality in spruce-fir forests
in the eastern United States greater than can be attributed to
typical trends and levels of natural variability?
What spatial patterns, if any, exist in growth and
mortality changes in spruce-fir forests in the eastern United
States and how do these patterns relate to spatial patterns of
pollutant exposure?
What is the effect of sulfur and/or nitrogen derived
pollutants alone or in combination with oxidants on spruce and
fir morphology?
Previous forest damage and decline studies of Camels Hump in the
Green Mountains of Vermont have identified three components of a spectral
signature associated with decline in red spruce (Picea rubens): a blue
shift of the chlorophyll well/red edge; a drop in reflectance of the near
infrared (NIR) plateau; and a relative increase in the short wave infrared
(SWIR) reflectance values (2,3). A damage mapping technique was developed
which utilized a ratio of SWIR/NIR aircraft and satellite spectral bands
(3,4). This technique has been shown to be an extremely accurate means of
detecting, quantifying and monitoring forest damage in conifer stands in
both the northeastern and southeastern United States (5,6).
Presently, the University of New Hampshire is involved in several
remote sensing studies that build on this work. One study involves the use
of satellite data to determine the change in the amount of damage present
in the Green Mountains between 1973 and 1984. A second study examines
spatial patterns of damage which exist across the Adirondacks, Green
Mountains and White Mountains. A third study uses a high resolution
airborne sensor to look at spectral signatures characterizing various types
of damage. These studies are summarized below:
57
-------
CHANGE DETECTION STUDIES
Remote sensing investigations employing NS-001 Thematic Mapper
Simulator (TMS) and Landsat Thematic Mapper (TM) data (3,4,5,6) have shown
excellent correlations between ground-based estimates of conifer forest
damage and 1.65/0.83 micrometer band (TM 5/4) ratios. Figure 1 shows a
damage assessment image made using the TM band 5/band 4 ratio along with a
near infrared band (band 5) and a visible band (band 2), color coded red,
green, blue respectively. Red areas in the image represent damaged
deciduous zones. Numbered and lettered areas are sites for which ground
assessments have been made. Although images produced using this ratio are
extremely accurate in mapping and quantifying forest damage levels, it is
often difficult to ascertain what proportion of the damage detected is a
result of a general forest decline phenomenon ("unnatural" damage) and what
is attributed to "natural" conditions, such as those related to poor
growing conditions, ice and wind storms, and other natural stresses.
One can begin to address the question of what proportion of damage is
natural vs. unnatural by using multitemporal remote sensing data sets to
monitor forest condition through time. The following is a summary of a
study to evaluate the potential of using Landsat Multispectral Scanner
(MSS) data to detect long term reflectance changes indicative of
high-elevation coniferous forest health (7). Data from August 29, 1973
(Landsat 1) and August 21, 1984 (Landsat 5), from the Green Mountains of
Vermont were used in this study. Sun elevation was 48° for both data sets,
and solar azimuth was similar for both scenes (134° and 136° for the 1973
and 1984 data sets, respectively).
Multispectral Scanner data were computer-processed at the Jet
Propulsion Laboratory (Pasadena, CA) using the VICAR processing system
installed on a VAX 11/780 computer. Bands used in the study were centered
at 0.65 (0.60-0.70; MSS Band 5) and 0.95 (0.80-1.10; MSS Band 7)
micrometers. Following co-registration of portions of the data sets
including coverage of the Green Mountains, data sets were standardized by
use of 20 forested targets. These sites represented relatively mature
stands, most of which were located at low elevations, and were presumed to
have undergone minimal spectral change between 1973 and 1984. Sites were
field-checked in August of 1987 to verify that these areas had not been
logged or selectively thinned between 1973 and 1984. Standardization
targets included six coniferous sites and 14 deciduous sites. Mean digital
numbers were extracted from each site for the 0.95 /im band from both 1973
and 1984 data sets. The 1973 vs. 1984 values regressed against each other
yielded an r2 value of 0.971 for the 0.95 urn band. This indicates that an
essentially linear relationship exists between 1973 and 1984 data sets for
this band, and implies that the MSS band 7 spectral properties for these
standardization sites had not changed significantly during the time period.
58
-------
Figure 1. Damage assessment image made, using NS-001 Thematic Mapper
Simulator data. Numbered and lettered areas are study sites.
Red areas indicate heavy forest damage. Taken from Rock et
a7.(3)
59
-------
The linear regression equation derived from the relationship between
the 1973 vs. 1984 vegetation standardization targets was then used to
convert digital number values from 1984 data into units comparable in value
to 1973 data for the 0.95 ^m band. Following standardization of the 1984
data set for the 0.95 fj.m band, a difference image was produced by
subtracting 1984 from 1973 data sets, and adding an offset of 100 to
eliminate negative numbers. Pixels for which values were greater than 100
showed a decrease in reflectance from 1973 to 1984 relative to the 20
vegetation standardization targets. Pixels for which values were less than
100 showed an increase in reflectance between these dates, relative to the
standardization targets.
A three-color composite using the 1973 0.65 pm and 0.95 urn bands in
conjunction with the 0.95 Jim difference data set (Figure 2) in the order of
blue, green and red was produced. The 0.65 and 0.95 /im bands were linearly
stretched using standard methodology to enhance contrast, and the
difference data set was linearly stretched to enhance decreases in
reflectance between 1973 and 1984. This image not only indicates where
major reflectance decreases have occurred but also depicts topographic
relief. Areas of red or dark orange generally indicate where coniferous
areas decreased in reflectance, whereas yellow to light orange areas
indicate where deciduous vegetation decreased in reflectance, in relation
to the standardization targets. Areas that are green to blue showed either
no major near-infrared reflectance changes, or increases in reflectance.
Field and laboratory spectral data acquired for red spruce at the Camels
Hump study area and for Norway spruce in West Germany suggest that a
decrease in the near infrared reflectance accompanies an increase in
needle damage associated with forest decline (8,9,10). Within the montane
coniferous areas, decreases in reflectance were most apparent in the
transition zone forests on the western lower slopes, where balsam fir and
red spruce dominate. Near-infrared reflectance at the upper elevations,
where balsam fir dominated, was relatively unchanged. A general trend of
decreasing basal area and inferred biomass loss through time has been
documented for the montane forest on the west facing slopes of Camels Hump
(the northernmost mountain seen in Figure 2) (11,12). It is presumed that
this decrease in basal area and inferred loss of green leaf biomass and the
concomitant increase in amounts of dead branches/trees results in the
observed decreases in reflectance in the coniferous portions of the
difference image.
It should be noted that it has been found that lower reflectance in
the near-infrared implies lower amounts of biomass as estimated by leaf
area index (LAI) measurements for some species (13). However, it has not
been documented that lower levels of biomass (or LAI) correlate well with
near-infrared reflectance for conifer species (14,15). Therefore, at
present, it cannot be stated that decreases in the near-infrared
reflectance noted for much of the conifer zone at Camels Hump are directly
related to decreases in green leaf biomass, or with the increases in dead
branches and trees which accompany loss of biomass, or both.
60
-------
Figure 2. False color composite image from Multispectral Scanner (MSS)
data of a portion of the Green Mountains of Vermont using 1973
0.65 and 0.95 fim bands, and the 0.95 /im difference data set.
Areas of red, orange or yellow indicate where near infrared
reflectance has decreased from 1973 to 1984 in relation to 20
deciduous and coniferous targets. Taken from Vogelmann (7).
61
-------
The data sets being compared were acquired during approximately the
same time of year (late August), and thus potential problems due to
different solar angles and azimuths have been negated. However, it should
be noted that annual phenological differences due to rainfall and
temperature variations represent potential problems in multitemporal
studies. Phenological differences may be minimized, but not totally
eliminated as factors contributing to reflectance differences between data
sets, by selection of data from the same date from year to year. Rainfall
and temperature data for the areas and dates in question should be used in
order to determine if annual phenological differences are likely to be
major factors influencing the remotely sensed data sets being compared.
It is likely that the lower reflectance of the deciduous vegetation
in the higher elevations in 1984 vs. 1973 was due at least in part to
rainfall differences affecting green leaf biomass production. More
rainfall occurred during the growing season of 1973 as compared with that
of 1984. However, it is not felt that rainfall differences alone can
explain the reflectance differences noted in the high elevation coniferous
regions. Some of these areas showed reflectance changes (e.g., the west
facing transition zone) whereas adjacent regions (e.g., high elevation
areas dominate by balsam fir) that presumably were under similar climatic
conditions did not. The differences in rainfall might be expected to have
different effects on conifer leaf flush in 1984 vs. 1973. However, since
conifers retain their needles for several years, these first year needles
make up only a portion of the total conifer foliage influencing
reflectance. Such rainfall differences would not be expected to greatly
affect the phenologic state of the older needles. Thus, coniferous
vegetation would not be as susceptible to annual variation in rainfall
patterns as would deciduous vegetation.
At present, it is felt that the decrease in near-infrared reflectance
noted in the 1984 data set as compared to the 1973 data set for the
high-elevation coniferous regions is attributed to the general forest
decline process, being related to the increased levels of mortality and
decreased levels of green biomass that have been documented in this region.
SPATIAL STUDIES
The detection and quantification of spatial patterns of conifer
forest damage in the eastern United States may be done accurately and
objectively using remote sensing techniques. Remote sensing data can be
used to detect large, regional variations in forest condition that can then
be correlated with patterns of pollutant exposure, soil types, geology and
other factors that may affect the condition of forest communities.
Previous studies have found that the ratio of TM band 5 to band 4 is
strongly correlated with ground-based measurements of forest damage in the
northeastern United States. The higher the level of forest damage, the
higher the ratio value (3,4,5,6). A Thematic Mapper scene (that included
coverage of the Green Mountains and the Adirondack Mountains, acquired
August 4, 1984, and a second scene that included the Green Mountains and
62
-------
the White Mountains, acquired June 10, 1984) were used. From these data
sets, values of the damage assessment ratio (TM band 5/band 4) were
compared among the three mountain ranges.
The two TM scenes were standardized by calibrating pixel values based
on homogeneous ground targets and by using the same parameters to stretch
the band 5/band 4 ratio over the full 0-255 dynamic range. Coniferous
portions of the image were isolated from non-coniferous forest regions
using a method in which a mask was placed over all regions of the image
that did not correspond to coniferous forest. A complete description of
the method can be found in Vogelmann and Rock (6).
A damage rating scale was developed using the TM band 5/band 4 ratio
to assess relative damage levels of montane conifer areas among selected
mountains in the Green Mountains of Vermont and the White Mountains of New
Hampshire (6). The same procedure was used to assess relative damage
levels of conifer areas in the Adirondack Mountains. Low, medium, and
high damage study sites located on Camels Hump in the Green Mountains were
used as standards of reference. Damage levels for each of these reference
sites were determined by visually assessing percentage foliar loss at each
study site (4,6). Ranges of ratio values corresponding to low, medium and
high damage categories were defined, and numbers of conifer pixels falling
within each damage category were totaled. The level of damage for each
mountain was then summarized using the following equation:
Damage Rating= (100 - % Low Damage Pixels + % High Damage Pixels)
2
Table 1 is a summary of conifer damage for several high elevation
areas in the Adirondack Mountains, Green Mountains and White Mountains. It
should be made clear that these damage ratings are relative measures of
forest health based on the field work done at Camels Hump. Thus, a damage
rating does not correspond to percent mortality, but is merely a relative
measure that can be used to compare damage levels among individual
mountains.
63
-------
Table 1. Conifer damage in the Adirondack Mountains, Green Mountains and
White Mountains.
SITE
ELEVATION
(meters)
%LOW
DAMAGE
PIXELS
%MEDIUM
DAMAGE
PIXELS
%HIGH
DAMAGE
PIXELS
DAMAGE
RATING
ADIRONDACK
Whiteface Mt. 1484
High Peaks Area 1268
GREEN MOUNTAINS
Camels Hump 1244
Mt. Abraham 1260
Breadloaf Mt. 1165
WHITE MOUNTAINS
Mt. Moosilauke 1464
Lafayette Mt, 1585
8.6
2.3
26.5
25.6
37.9
72.3
63.9
12.7
6.8
20.7
25.6
24.6
16.8
19.0
78.7
90.4
52.8
48.9
37.5
10.9
17.0
85.1
94.3
63.2
61.7
49.8
19.3
26.6
It is apparent that there is a trend of decreasing damage from the
westernmost range (Adirondacks) to the easternmost range (the White
Mountains). It is also evident from the table and from field studies that
elevation, slope and aspect alone are not factors which account for the
relative levels of damage in coniferous forests in New York and New
England. The National Acid Precipitation Assessment Program sponsored
studies to determine the spatial patterns of wet deposition pH values in
North America. Results are shown in Figure 3. The pattern of pH values
that are found in the mountains of the northeastern United States correlate
with our damage assessment: The lowest pH values in North America are
approximately centered over the Adirondacks and pH values increase
(indicating less acidic conditions) in all directions from this area. The
lowest pH values (in the Adirondack Mountains) correspond to the highest
damage ratings, while higher pH values (in the White Mountains of New
Hampshire) correspond to the lower damage ratings of the three ranges
studied. The Green Mountains in Vermont, which have intermediate damage
levels, are inferred to have intermediate pH values.
64
-------
EXPLANATION
1 5.4 pH at sample site
•5.0- Line of approximately
equal pH value
Figure 3.
pH measurements for North America, measured in 1982 (16)
FLUORESCENCE LINE IMAGER STUDIES
To date, high-spectral resolution in situ and airborne sensor data
sets have been acquired for forest decline sites in the northeastern United
States (3,9), and the Federal Republic of Germany (8,10). Although these
high-resolution data sets provide a great deal of fine-spectral feature
information relating to specific symptoms of forest decline (chlorosis,
canopy dryness, and foliar loss), as yet, such symptoms have not been
related to exposure to specific pollutants such as sulfur and/or nitrogen
65
-------
compounds, either alone or in combination with oxidants. In order to
develop spectral signatures characteristic of spruce response to specific
pollutant exposure, high-resolution in situ spectral assessment studies
must be conducted in association with controlled-exposure experiments. As
a means of portraying the power of airborne high-spectral resolution data
in assessing types of forest damage, work presented in detail elsewhere (9)
is cited below.
An airborne imaging spectrometer, known as the Fluorescence Line
Imager (FLI), and owned and operated by Moniteq, Ltd., Toronto, Canada1,
has been used to detect reflectance features associated with the
chlorophyll well/red edge blue shift characteristic of in situ spectral
measurements (9). These spectral fine features are not detected by
broad-band sensor systems such as the TM or TMS, but rather require the
high-spectral resolution capabilities of imaging spectrometers such as the
FLI and NASA's Airborne Visible Infrared Imaging Spectrometer (AVIRIS).
Figure 4 presents FLI data acquired for a low and a high damage site
on Camels Hump (sites 1 and 7 respectively, Figure 1). Both raw and
normalized plots of counts verses wavelength are presented and the blue
shift as well as the drop in NIR reflectance are readily seen. In
addition, spectral reflectance in the visible green and red regions of the
electromagnetic spectrum (0.50- 0.69 urn) characteristic of chlorosis are
also seen.
Using red edge parameters, a false color image of the FLI flight line
is presented in Figure 5, compared with a similar portion of the TMS image
for the same area. A comparison of the two images in Figure 5 suggests
that the broad-band TMS data are detecting generic damage in both red
spruce in the transition zone forest (lower elevation) and balsam fir
(higher elevation) in the conifer forest zone on Camels Hump. The FLI
image appears to be mapping only the damage, based on red edge parameters,
occurring in the transition zone red spruce. Winter damage and fir wave
damage is known to occur in the upper elevation, fir-dominated conifer
forests above the transition zone on Camels Hump. It has also been shown
that the balsam fir on the mountain has undergone a less severe (although
statistically significant) decline in vigor and biomass than has the red
spruce (12). This suggests that the use of both sensor systems (TM/TMS and
FLI) may provide information which allows separation of different damage
types: forest decline damage in red spruce and winter damage/fir-wave
damage in balsam fir.
1. Reference to specific manufacturers is for clarity and does not
constitute endorsement of product by NASA or the University of New
Hampshire.
66
-------
Plot of Count« Vter
961
Figure 4. FLI data acquired for low and high damage sites on Camels Hump,
Vermont. Red, yellow, and orange spectra represent high damage
areas; white, pink, blue, and green spectra are taken from low
damage sites. Modified from Rock et al. (9).
67
-------
Figure 5. False color image of the Camels Hump, Vermont, sites made using
red edge parameters from FLI data, compared to a damage
assessment image of the same area using Thematic Mapper
Simulator data. Red and orange areas in the images represent
damaged forest, while blue and green areas represent healthier
forests. In both images, the summit of Camels Hump is shown as
a bright right-angle outcrop on the left edge.
68
-------
High-spectral resolution remote sensing systems currently available
on airborne platforms will eventually be available on orbital platforms
such as the NASA/ESA Earth Observing System (EOS). Once in situ spectral
signatures have been identified which are diagnostic of specific pollutant
damage/exposure, such airborne and/or spaceborne sensor systems may provide
forest assessment capabilities which will relate specific spectral
signatures (effects) to specific causal agents so that direct cause and
effect relationships may be remotely detected and monitored on a regional/
global scale.
SUMMARY
Each remote sensing system has limitations based on spatial/spectral
resolution, band placement, software availability, and any of a number of
additional complicating factors. Used alone, each sensor can only be used
to solve a portion of the forest damage and decline/atmospheric pollutant
puzzle. When data are acquired with many sensors over the same region,
cause and effect issues can be addressed more readily.
In the above studies we found that the changes in health in certain
portions of the spruce-fir forests in the eastern United States were
greater than can be attributed to typical trends and natural variability.
Also, a distinct spatial pattern of greater damage in the Adirondack
Mountains, decreasing to the east has been detected and this pattern of
damage corresponds to spatial patterns of wet deposition pH values.
Finally, current research indicates that spectral signatures characteristic
of damage exist, and these can be used to identify various damage symptoms.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
REFERENCES
U.S. Department of Agriculture, Forest Service. Cooperative Survey
of Red Spruce and Balsam Fir Decline and Mortality in New York,
Vermont and New Hampshire, 1984. Broomall, PA: U.S. Dept. of
Agriculture, Forest Service, Northeastern area, 1985. 53 pp.
Rock, B.N., Williams, D.L. and Vogelmann, J.E. Field and Airborne
Spectral Characterization of Suspected Acid Deposition Damage in Red
Spruce (Picea rubens) from Vermont. Proceedings of the llth
International Symposium on Machine Processing of Remotely Sensed
Data, Purdue University, West Lafayette, IN, 1985. pp. 71-81.
69
-------
3. Rock, B.N., Vogelmann, J.E., Williams, D.L., Vogelmann, A.F., and
Hoshizaki, T. Remote Detection of Forest Damage. BioScience,
36:439-445, 1986b.
4. Vogelmann, J.E. and Rock, B.N. Assessing Forest Decline in
Coniferous Forests of Vermont Using NS-001 Thematic Mapper Simulator
Data. Int. J. Remote Sensing, 7:1303-1321, 1986.
5. Rock, B.N., Defeo, N.J., and Vogelmann, J.E. Vegetation Survey Pilot
Study: Detection and Quantification of Forest Decline Damage using
Remote Sensing Techniques. Final report to the USDA Forest Service,
Jet Propulsion Laboratory Document D-4669, Pasadena, California,
1987. 30 pp, appendices.
6. Vogelmann, J.E., and Rock, B.N. Assessing Forest Damage in High
Elevation Coniferous Forests in Vermont and New Hampshire Using
Landsat Thematic Mapper Data. Remote Sens. Environ., 24:227-246,
1988.
7. Vogelmann, J.E. Detection of Forest Change in the Green Mountains of
Vermont Using Multispectral Scanner Data. Int. J. Remote Sensing, 9,
in press, 1988.
8. Rock, B.N., Hoshizaki, T., Lichtenthaler, H., and Schmuck, G.
Comparison of In Situ Spectral Measurements of Forest Decline
Symptoms in Vermont (USA) and the Schwarzwald (FRG). Proc. of
Intern. Geosci. and Remote Sensing Symposium (IGARSS '86), IEEE
86CH2268-1, IEEE, New York, Vol. 3:1667-1572, 1986a.
9. Rock, B.N., Hoshizaki, T., and Miller, J.R. Comparison of In Situ
and Airborne Spectral Measurements of the Blue Shift Associated with
Forest Decline. Remote Sens. Environ., 24:109-127, 1988.
10. Herrmann, K., Rock, B.N., Ammer, U., and Paley, H.N. Preliminary
Assessment of Airborne Imaging Spectrometer and Airborne Thematic
Data Acquired for Forest Decline Areas in the Federal Republic of
Germany. Remote Sens. Environ., 24:129-149, 1988.
11. Vogelmann, H.W., Bliss, M., Badger, G., and Klein, R.M. Forest
Decline on Camels Hump, Vermont. Bull. Torrey Bot. Club,
112:274-287, 1985.
12. Vogelmann, H.W., Perkins, T., Badger, G. and Klein, R.M. A 21-year
Record of Forest Decline on Camels Hump, Vermont. Eur. J. For. Path:
in press, 1988.
13. Wiegland, C.L., Richardson, A.J., and Kanemasu, E.T. Leaf Area Index
Estimates for Wheat From Landsat and Their Implications for
Evapotranspiration and Crop Modeling. Agron. J., 71:336-342, 1979.
70
-------
14. Peterson, D.L., Spanner, M.A., Running, S.W., and Teuber, K.B.
Relationship of Thematic Mapper Simulator Data to Leaf Area Index of
Temperate Coniferous Forests. Remote Sens. Environ., 22:323-341,
1987.
15. Franklin, J. Thematic Mapper Analysis of Coniferous Forest Structure
and Composition. Int. J. Remote Sensing, 7:1287-1301, 1986.
16. NAPAP (National Acid Precipitation Assessment Program). Annual
Report to the President and Congress, Washington, D.C. 1983.
71
-------
USE OF HIGH SPECTRAL RESOLUTION SENSORS TO DETECT
AIR POLLUTION INJURY IN CONIFER FORESTS
by Susan L. Ustin and Scott N. Martens, Department of Botany, University of
California, Davis, CA 95616; Brian Curtiss, CIRES, University of Colorado,
Boulder, CO 80309; and Vern C. Vanderbilt, NASA Ames Research Center,
Moffett Field, CA 94035
Although the information in this document has been funded wholly or in part
by the United States Environmental Protection Agency under Grant
0R-814274-01-0 and contract 07B0008NTEX to S.L. Ustin, it does not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
ABSTRACT
Spaceborne or ground-based high spectral resolution imaging sensors
have the potential for non-destructive in situ monitoring of growth and
other physiological processes. Leaf biochemical properties have
absorptance features in the visible and reflected infrared spectrum that
may be used to assess photosynthetic capacity and carbon allocation into
metabolic and structural pathways. Spectral features related to leaf
pigments were shown to change in conifer species exposed to ozone under
controlled fumigation and under environmental conditions. Changes in the
width and depth of the chlorophyll absorption feature around 680 nm results
in a blue-shift of the "red edge" in canopy spectra. This change is highly
correlated with total needle chlorophyll concentration and the change is
sufficient for detection by airborne sensors; e.g., AVIRIS. Although other
sources of environmental variation cause shifts in this spectral region
they do not appear to duplicate changes resulting from chlorosis.
INTRODUCTION
The ability to detect and monitor vegetation response to a wide range
of anthropogenic pollutants is of considerable global significance.
Repeated surveys on regional to global scales are only possible through the
use of satellite or aircraft sensor technology. Use of this technology for
mapping the aerial extent of forest disturbance has been well documented.
However, development of methods for the early detection of pollutant Injury
are less well developed (1). In part, this limitation has been due to the
inability of relatively coarse spatial and spectral resolution scanners to
detect subtle changes in ecosystem processes and functioning evident before
changes in leaf area or community structure occur. Such changes may be
72
-------
spectrally apparent through changes In leaf chemistry resulting from
altered metabolic processes or carbon allocation. Spectroscopic assays are
the primary analytical method used in laboratory research for
identification and quantification of plant pigments and other biochemical
components of leaves. It has recently become feasible to acquire high
resolution spectra non-destructively under in situ environmental conditions
and several instruments are now available for either field-based or
airborne research.
The next generation of satellite sensors will include a high
resolution imaging spectrometer (HIRIS), having high spatial and spectral
resolution in the visible and reflected infrared wavelength region of the
solar spectrum. This sensor, or the current airborne instrument, the
Advanced Visible and Infrared Imaging Spectrometer (AVIRIS), is capable of
providing detailed information about the physiological condition of forest
ecosystems and of mapping spatial patterns associated with plant stresses.
HIRIS will acquire some 196 spectral bands, and AVIRIS acquires 224 bands,
in the 400 to 2500 nm region at sampling intervals of 10 nm. In addition
to these sensors, the EOS space platform in the late 1990s will include
other instruments of interest for the early detection of forest injury,
such as the moderate resolution MODIS-N, with frequent overpasses and
capability for measuring polarized reflectance, SAR with multifrequency
multipolarization microwave. Thus, there is considerable promise for
developing spectrally based assessments of pollutant injury to forested
ecosystems.
One would like to link the remotely sensed spectral measurement to a
mechanistically based model for predicting changes in productivity or
growth. Although canopy changes related to water content have profound
implications on the carbon budget we have chosen to direct our focus to the
detection and quantification of plant pigments and methods for their use in
evaluating vegetation stress in image spectra. The premature needle
senescence and chlorosis following exposures to ozone or other atmospheric
pollutants is expected to have a cumulative effect on lowered productivity
in forests. Although, understanding the fine structural spectral changes
associated with stress or those associated with particular stress agents is
limited (1). Nonetheless, research on leaf spectral properties suggest
that physiologically important processes related to plant productivity may
be detectable. It has been shown that biomass production is linearly
related to the interception of photosynthetically active radiation (400-700
nm) by the canopies of crop types and forests (2,3,4,5). This relationship
has been used by Sellers to develop theoretical models for predicting
photosynthesis and water transport processes from canopy reflectance
characteristics (6,7). Other advanced radiative transfer models may be
used as inversion algorithms to obtain biophysical parameters of canopies
(e.g., 8,9).
Ozone is a strong oxidant directly injuring cell membranes and is
toxic at low exposures. It is regionally distributed and clearly
implicated in forest decline (10). Needle mottle-chlorosis is symptomatic
for ozone injury in conifers. Chloroplast injury is an early metabolic
response to ozone exposure. Thompson et al. (11) report granulation of the
73
-------
chloroplast stroma are the first observed cellular anatomical changes
following fumigation in beans. General disruption of chloroplast function,
including loss of thylakoid membranes and grana, proceeds other anatomical
cellular changes. Good correlations between ozone exposures and changes in
reflectance have been reported (12,13,14,15).
LEAF SPECTRAL PROPERTIES
The general shape of reflectance curves for green leaves are similar
for all species (16,17). Absorption features in the visible spectrum are
dominated by the presence of photosynthetic pigments, the chlorophylls,
xanthophylls, and carotenes (Figure 1). The major absorptions for
xanthophylls and carotenes occur in the UV but include absorptions in the
400-500 nm range, while chlorophyll a and b have several absorption
features in the 600-700 nm region (18). Such absorption differences
suggest that it may be possible to detect specific photosynthetic pigments
using high spectral resolution instruments and quantify their
concentrations and relative proportions. Previous studies have shown good
correlations between leaf chlorophyll or nitrogen and reflectance from
spectral bands in the 550-700 nm range (19,20,21,22). Difference spectra
between ozone-fumigated and control conifer seedling canopies showed that
the most significant changes occurred in this wavelength region of the
400-2500 nm spectrum (13). Changes in leaf chlorophyll concentration alter
the depth and width of these features, resulting in apparent wavelength
shifts in the position of the "red edge" feature near 710 nm (23,24). As
the chlorophyll absorption bandwidth narrows, the "red edge" appears to
shift toward shorter wavelengths. Such "blue shifts" have been reported
for tree species exposed to acidic deposition (25) and ozone (13,14,26).
Lignin and cellulose have biochemical absorption features that
provide additional fine structure in leaf spectra (Figure 1). Starch,
protein and nitrogen have absorptions in the infrared and may be
identifiable in high resolution spectra (27,28,29). Since changes in the
proportion of lignin and nitrogen are related to carbon partitioning into
structural and metabolic pathways, such information may be useful in
analyses of nutrient cycling, energetics and productivity of the canopy
(28,30). Further, the presence of cellulose features, evident in dry plant
tissue, may be informative about the phenological condition and the
presence of non-photosynthetic tissue in the canopy.
Leaves have no major absorption features in the near infrared (NIR,
700-1200 nm) and reflectance in this region is controlled by the multiple
scattering of photons at water-air interfaces in the cell walls (31).
Nonetheless, epidermal modifications can cause changes in reflectance (16)
or changes in the proportion of polarized reflectance from leaf surfaces
(32). Canopy architecture, due to variation in the distribution and
angular position of canopy leaf and stem elements, has a major effect on
NIR reflectance from tree canopies (33). Broad-band reflectance in the red
and NIR have been used to estimate variation in leaf area index in conifer
stands (34) and water content in the NIR/IR (25).
74
-------
2200 2400
Figure 1. Reflectance curve for a typical green
position for absorptance features for
important compounds are indicated.
leaf. Wavelength
some metabolically
DETECTING SPECTRAL CHANGE LINKED TO PRODUCTIVITY PROCESSES
Clearly, leaf chlorophyll content limits maximum photosynthetic
capacity. The environmental factors (light, temperature, nutrients, water,
etc.) which limit photosynthetic rates also regulate chlorophyll synthesis
and maintenance (35). In equilibrated systems, the capacity for energy
capture is proportional to that for CO, fixation, resulting in linear
relationships between leaf chlorophyll ancf nitrogen concentrations (36,37).
Plant stress results in decreased chlorophyll concentrations and increased
chlorophyll a/b ratios (38), apparently due to the association of
chlorophyll b with the light harvesting complex of photosystem II. Under
conditions of stress, increased carotenoid pigments have been noted,
possibly providing a mechanism for protection from photooxidation of
photosynthetic reaction centers under high irriadiance (35,39). Carotenoid
and xanthophyll pigments have been shown to increase linearly with maximum
chlorophyll fluorescence for a number of species (40).
Figure 2 shows changes observed in the visible spectrum for seedling
canopies of Lodgepole pine, a moderately ozone-sensitive species, after
five months growth in open top chambers under a simulated ambient seasonal
75
-------
ozone exposure (13,41,42) and in ozone-filtered clean air. Four major
spectral features are observed between 560 and 680 nm, a region where
extracted chlorophyll a and b have several absorptions. Various complexed
forms of chlorophyll a and b exist in vivo so the precise pigment
identification of the absorption features, which are generally red-shifted
relative to extracted preparations, is uncertain (43,44). The spectral
changes between control and ozone fumigated conifers is indicative of a
significant chlorosis under this treatment. There is a general increase in
reflectance throughout the 560 to 680 nm spectral region (Fig. 1). This
results in a slope increase over this wavelength interval and a slight
decrease in the bandwidths of the absorption features.
10
9 -
Lodgepole pine
3 -
2 -
540 560
580
600 620 640
WAVELENGTH ( ntn )
660 680 700
Figure 2. Mean reflectance spectra of ozone fumigated (n=18) and control
(n= 9) first year Lodgepole pine seedlings measured at nadir
with the Portable Instantaneous Display and Analysis
Spectrometer (PIDAS). Fumigations followed simulated ambient
regime (39). Twenty-five seedlings were tightly grouped in a
tray to form a closed "canopy" for measurements. Measurements
were made using two 75W power regulated floodlamps and
calibrated using a Fiberfrax standard.
Ponderosa pine, an ozone-sensitive species, shows even greater
increases in reflectance in this spectral region although specific pigment
absorption features are less distinct (Figure 3). There is a 283% increase
in reflectance of ozone-exposed seedlings compared to nonfumigated
seedlings at 680 nm. On this figure, one observes a blue wavelength shift
of about 10 nm at the "red edge" due to the bandwidth narrowing of the
76
-------
chlorophyll absorption feature. The wavelength position of the "red edge"
inflection on the curves can be better observed when the first derivative
of reflectance is plotted (Figure 4).
ozone treated
control
— control
500
550 600 650
WAVELENGTH ( nm )
700
750
Figure 3.
Mean Ponderosa pine reflectance spectra for ozone fumigated and
control treatments as described in Fig. 2. Additional spectra
illustrate a typical forest soil spectrum and a mixed
soil-control spectrum synthesized to approximate the shape of
the fumigated spectra.
Similar changes are observed on needles obtained from forest stands
in the southern Sierra Nevada exposed to atmospheric ozone pollution. The
spectra in Figure 5 were obtained from measurements on needle whorls at a
site (#12) having moderate exposures to ozone. It shows a 148% increase in
reflectance at 670 nm between first-year needles (i.e, current year), which
do not exhibit visible ozone injury, and fourth year-needles, which show
some chlorotic injury. At this site, no sampled branches had more than
four years of needles although Ponderosa pine retains healthy needles for
six to ten years (45). Chlorophyll measurements were made on needle
samples from this site and another site having somewhat less visible ozone
injury. Chlorophyll a and b concentrations are highest during the second
year. During the first year chlorophyll concentrations are similar at both
sites but lower at the site having higher visual ozone injury in other
years; differences are greatest in the fourth year (Table 1). Scatterplots
of total chlorophyll, chlorophyll a, and b concentrations and reflectance
at each nm wavelength in the 475-750 nm spectral region were examined.
77
-------
UJ
It
UJ
>
4-1
n
Pondercsa pine
treated + soil; 88: 12
control
670 680 690 700 710 720 730 740 750 760 770
WAVELENGTH ( nm )
Figure 4. The first derivative of reflectance for data shown in Figure 3.
60
50-
40-
30-
u
UJ
20-
10-
Site 12
Year 1
Year 4
' /
//
I
I
450 500 550 600 650
WAVELENGTH ( nm
700
750
BOO
Figure 5. Mean spectra of first and fourth year needle whorls of
Ponderosa pine (n=60) from a site in the southern Sierra Nevada
exhibiting moderate visual symptoms of ozone injury.
78
-------
Correlation coefficients ranged between 0.1 and 0.5 (n=117) due to
the many conjugated forms of chlorophyll in the leaf having somewhat
different absorption spectra. However, a high correlation with little
scatter was obtained when the wavelength of the half bandwidth of the first
derivative maxima is plotted against total chlorophyll concentration
(Figure 6). A 0.93 regression coefficient is also obtained when the
derivative wavelength is plotted against chlorophyll a but with a lower
slope (y=-4.001 104+ 59.257x). The change in slope is due to a concurrent
change in proportion of a and b chlorophylls as total chlorophyll declines.
When the ratio of reflectance at 645/670 nm is plotted against the
chlorophyll b/a ratio a significant (p<0.05; n=117) correlation is obtained
although the sensitivity is low and with considerable scatter in the data
(r=0.36; y=0.035 + 0.235x).
2800
2600
'X 2400
§< 2200
3 2000
U 1800
OS
•4-1
o
H
1600
1400
1200
y = - 6.246e+4 + 91.9988x R = 0.93
696 698 700 702 704 706 708 710
Wavelength (nm) of Red Edge Inflection
Figure 6. The regression relationship between total needle chlorophyll
concentration (ug chlorophyll/mg dry weight) and the wavelength
of the first derivative maxima (bandwidth at half height).
Data show the mean chlorophyll per year from two sites in the
southern Sierra Nevada having moderate to low visual symptoms
of ozone injury.
The blue shift seen in the first derivative maxima for the sites
shown is sufficient to be detectable by remotely sensed scanners with the
bandwidth resolution comparable to those of AVIRIS or of the proposed HIRIS
data. Under conditions of incomplete crown closure, typical of Ponderosa
pine forests, it is questionable whether a 1-2 band shift is sufficient for
a detectable image change. However, more severely ozone affected sites in
the southern Sierra Nevada had fewer years needles and exhibited larger
percentages of visually estimated needle chlorosis than the results
presented here. Thus, spectral trends in AVIRIS images are expected to
exceed changes reported here.
79
-------
MODELING SPECTRAL CHANGES
A number of authors have found a shift in the wavelength position of
the "red edge" under conditions where canopy chlorophyll concentration
changed (15,23,24,46). Nonetheless, we questioned whether other sources of
environmental variance could produce a similar blue wavelength shift.
Shutt et a7. (47) and Vandeibilt et al. (32) proposed that changes in
canopy geometry could produce similar effects independent of physiological
condition. Vanderbilt et si. (48) found canopy orientation changes due to
wind could produce either a blue or red shift as a result of changing the
proportion of polarized reflected light, since it is reflected by the
cuticle before entering the mesophyll tissue. We have modeled two
conditions expected to occur in remotely sensed AVIRIS data of typical
Ponderosa pine forests: pixel spectra containing mixtures of soil and tree
canopy and pixels having backscattered illumination from the surroundings.
Table 1. Mean chlorophyll a, b, and total chlorophyll (ug chlorophyll/nig
dry weight) from Ponderosa pine needles collected from twenty
trees at two sites in the southern Sierra Nevada exposed to low
or moderate seasonal ozone concentrations.
Dry Weight:
Chlorophyll A
whorl 1 whorl 2 whorl 3 whorl 4 all whorls
Site 14
mean 1,525.6 1,911.4 1,634.5 1,625.0 1,663.2
std. dev. 261.6 377.3 455.1 335.5 284.4
Site 12
mean 1,614.5 1,799.3 1,548.1 1,243.3 1,597.0
std. dev. 426.9 555.0 612.3 615.6 430.6
Chlorophyll B
whorl 1 whorl 2 whorl 3 whorl 4 all whorls
523.2
122.7
525.1
186.5
668.8
162.6
577.2
207.0
607.0
196.7
538.7
254.2
641.2
171.9
443.4
217.8
594.6
137.2
530.5
185.2
Total Chlorophyll
,jrfhorl 1 whorl 2 whorl 3 whorl 4 all whorls
2,048.8
380.7
2,139.6
602.6
3.9%
2,580.1
536.1
2,376.5
755.5
2.4%
2,241.4
646.9
2,086.8
865.2
3.1%
2,266.2
503.0
1,531.3
917.2
1.2%
2,257.9
418.7
2,127.5
613.0
2.7%
80
-------
In the first case, we produced a simulated spectrum resembling the
ozone fumigation spectrum of Ponderosa pine from a mixture of red forest
soil and the unfumigated Ponderosa pine spectra (Figure 4). Although it
resembles the fumigation spectrum in albedo, the first derivative of the
mixed spectrum did not exhibit a blue shift (Figure 5). Furthermore, there
is no known natural substance having the required spectral properties to
produce such a shift. In the second case, we found that backscattered
light onto a surface altered the reflectance spectrum of that surface (49)
but that the wavelength position of the first derivative maxima would be
slightly red-shifted (<5 nm) under such mixed sources of illumination.
Thus, these models support the possibility of remotely detecting canopy
reflectance and quantifying subtle changes related to chlorophyll
concentration under conditions of incomplete canopy closure.
CONCLUSIONS
Leaf and canopy spectra exhibit absorption features related to
photosynthetic pigments and other biochemical compounds useful for remotely
monitoring changes in growth and productivity of forest stands. Foliage
exposed to ozone under controlled fumigations and under forest conditions
show significant changes in reflectance, indicative of the loss of
photosynthetic pigments. In particular, decreases in needle chlorophyll
concentration results in increased reflectance and band narrowing of
absorption features in the 550-700 nm wavelength region. Although band
depth of these features is linearly related to chlorophyll a and b
concentrations and significantly correlated at a number of wavelengths, the
conjugated forms of in vivo chlorophyll result in considerable scatter in
direct wavelength comparisons. Changes in the band-width of the long
wavelength forms of chlorophyll is readily detectable in the first
derivative of the spectrum. The wavelength maxima of the derivative
spectrum is highly correlated with chlorophyll concentration.
Modeling of AVIRIS spectral change arising from two sources of mixed
spectra which are expected in forests having incomplete crown cover, were
examined. In the first example, the pixel spectrum results from linear
mixtures of soil and canopy spectra; in the second example, a pixel
spectrum results from a surface receiving linear mixtures of multiple
sources of illumination, direct solar, diffuse, and backscattered off
surrounding surfaces of other source materials. Neither case produced a
blue shifted derivative maxima like that resulting from decreased needle
chlorophyll concentration. Although these results support the possibility
of remotely measuring such changes with high spectral resolution aircraft
or satellite sensors, they also point out expected difficulties of the
direct assessment of spectral changes without careful, and possibly,
complex data analyses of AVIRIS images.
81
-------
REFERENCES CITED
1.
2.
3.
6.
7.
8.
10.
11.
12.
Jackson, R.D. Remote Sensing of Biotic and Abiotic Plant Stress.
Ann. Rev. Phytopathol., 24:265-287, 1986.
Monteith, J.L. In: C.B.
Limiting Plant Productivity.
Johnson (Ed.) Physiological Processes
Butterworths, London, 1981. p. 23-38.
Ross, J. The Radiation Regime and Architecture of Plant Stands. In:
H. Lieth (Ed.) Tasks for Vegetation Sciences 3. W. Junk Publ.
Boston, 1981. 391 p.
Jarvis, P.G., and J.W. Leverenz. Productivity of Temperate,
Deciduous and Evergreen Forests. In: O.L. Lange, P.S. Nobel, C.B.
Osmond, and H. Ziegler (Eds.), Encyclopedia Plant Physiology,
Physiological Plant Ecology IV NS. Ecosystem Processes: Mineral
Cycling, Productivity and Man's Influence. Springer Verlag, New
York, 1983. p. 234-280.
Linder, S. Potential and Actual Production in Australian Forest
Stands. In: J.J. Landsberg, and W. Parsons (Ed.), Research for
Forest Management. CSIRO, Melburn, Australia, 1986. p. 11-35.
Sellers, P.J.
Int. J. Remote
Canopy Reflectance, Photosynthesis and Transpiration.
Sens., 6:1335-1372, 1985.
Sellers, P.J. Canopy Reflectance, Photosynthesis and Transpiration.
II, The Role of Biophysics in the Linearity of Their
Interdependence. Remote Sensing of Environ., 21:143-183, 1987.
Norman, J.M.
Barnfield and
Environment of
pp. 149-277.
Modelling the Complete Crop Canopy. In: B.J.
J.F. Gerber (Eds.) Modification of the Aerial
Crops, ASAE Monograph 2. St. Joseph, MI, 1979.
Meyers, T.P., and K.T. Paw U. Modelling the Plant Canopy
Micrometeorology with Higher-order Closure Principles. Agric. For.
Meterol., 41:143-163, 1987.
McLauglin, S.B.
Review. J. APCA,
Effects of Air Pollution on
35:512-534, 1985.
Forests: A Critical
Thomson, W.W., W.M. Dugger, and R.L. Palmer. Effects of Ozone on the
Fine Structure of the Palisade Parenchyma Cells of Bean Leaves. Can.
J. Bot., 44:1677-1682, 1966.
Runcekles, V.C., and H.M. Resh. The Assessment of Chronic Ozone
Injury to Leaves by Reflectance Spectrophotometry. Atmos. Environ.,
9:447-452, 1975.
82
-------
13. Ustin, S.L., and B. Curtiss. Spectral Characteristics of Ozone
Treated Conifers. Final Report EPA contract #7B0008NTEX, 1987.
14. Westman, W.E., and C.V. Price. Remote Detection of Air Pollution
Stress to Vegetation: Laboratory-level Studies. Proc. Intl. Geosci.
and Remote Sens. IGARRS '87, 1:451-456, 1987.
15. Westman, W.E., and C.V. Price. Spectral Changes in Conifers
Subjected to Air Pollution and Water Stress. IEEE Trans. Geosci. and
Remote Sens., 26:11-21, 1988.
16. Gates, D.M. Biophysical Ecology. Springer-Verlag. New York. 1980.
611 p.
17. Goetz, A.F.H., B.N. Rock, and L.C. Rowan. Remote Sensing for
Exploration: An Overview. Econ. Geol., 78:573-590, 1983.
18. Goedheer, J.C. Visible Absorption and Fluorescence of Chlorophyll
and Its Aggregates in Solution. In: L.P. Vernon and G.R. Seely
(Eds.) The Chlorophylls. Academic Press, 1966. p. 147-185.
19. Benedict, H.M. and R. Swidler. Non-destructive Method for Estimating
Chlorophyll Concentration of Leaves. Sci., 133:2015-2016, 1961.
20. Thomas, J.R. and G.F. Oether. Estimating Nitrogen Content of Sweet
Pepper Leaves by Reflectance Measurement. Agron. J., 64:11-13, 1972.
21. Gausman, H.W., D.E. Escobar, and R.R. Rodriguez. Discriminating
Among Plant Nutrient Deficiencies With Reflectance Measurements.
Proc. 4th. Biennial Workshop on Aerial Color Photography. Am. Soc.
Photogramm., 39:13-27, 1973.
22. Tsay, M.L., D.H. Gjerstad, and G.R. Glover. Tree Leaf Reflectance:
A Promising Technique to Rapidly Determine Nitrogen and Chlorophyll
Content. Can. 0. For. Res., 12:788-792, 1982.
23. Collins, W. Remote Sensing of Crop Type and Maturity. Photogramm.
Eng. and Remote Sens, 44:43-55, 1978.
24. Chang, S.H. and W. Collins. Confirmation of the Airborne
Biogeophysical Mineral Exploration Technique Using Laboratory
Methods. Econ. Geol., 78:723-736, 1983.
25. Rock, B.N., J.E. Vogelmann, D.L. Williams, A.F. Vogelmann, and T.
Hoshizaki. Remote Detection of Forest Damage. BioSci., 36:439-445,
1986.
26. Ustin, S.L. and B. Curtiss. Spectral Characteristics of Ozone
Treated Conifers. MS in preparation for submission to Oecologia.
1988.
83
-------
27. Spanner, M.A., D.L. Peterson, W. Acevedo and P. Matson. High
Resolution Spectrometry of Leaf and Canopy Chemistry for
Biogeochemical Cycling. Proc. of the Airborne Imaging Spectrometer
Data Analysis Workshop, April 8-10, 1985. JPL Publ., 85-41, 92-99,
1985.
28. Peterson, D.L., J.D. Aber, P.A. Matson, D.H. Card, N. Swanberg, C.
Wessman, and M. Spanner. Remote Sensing of Forest Canopy and Leaf
Biochemical Contents. Remote Sens. Environ., 24:85-108, 1987.
29. Wessman, C.A., J.D. Aber, and D.L. Peterson. Estimating Key Forest
Ecosystem Parameters Through Remote Sensing. Proc. IGARSS '87,
2:1189-1193, 1987.
30. Running, S.W., J.D. Aber, D.L. Peterson, P.A. Matson, and P.M.
Vitousek. A Simulation Model Integrating Carbon, Water and Nitrogen
Cycles in Forests. Proc. Symp. IUFRO Whole Plant Physiology. Oak
Ridge, TN, 1985.
31. Gates, D.M., H.J. Keegan, H.C. Schleter, and V.R. Weidner. Spectral
Properties of Plants. Appl. Optics, 4:11-20, 1965.
32. Vanderbilt, V.C., L. Grant, L.L. Biehl, and B.F. Robinson. Specular,
Diffuse, and Polarized Light Scattered by Two Wheat Canopies. Appl.
Optics, 24:2408-2418, 1985.
33. Jackson, R.D. and P.J. Pinter. Spectral Response of Architecturally
Different Wheat Canopy. Remote Sens, of Environ., 20:43-56, 1986.
34. Running, S.W., D.L. Peterson, M.A. Spanner, and K.B. Teuber. Remote
Sensing of Coniferous Forest Leaf Area. Ecol., 67:273-276, 1986.
35. Kramer, P.J. and T.T. Kozlowski. Physiology of Woody Plants.
Academic Press. New York, 1979. Slip.
36. Field, C., and H.A. Mooney. Leaf Age and Seasonal Effects on Light,
Water and Nitrogen Use Efficiency in a California Shrub. Oecologia,
56:348-355, 1983.
37. Field, C., J. Merino, and H.A. Mooney. Compromises Between Water Use
Efficiency and Nitrogen Use Efficiency in Five Species of California
Evergreens. Oecologia, 60:384-389, 1983.
38. Salisbury, F.B. and C.W. Ross. Plant Physiology, 3rd ed. Wadsworth
Publ. Co. Belmont, CA., 1985. 540 p.
39. Levitt, J. Responses of Plants to Environmental Stresses II. Water,
Radiation, Salt, and Other Stresses. 2nd. Ed. Academic Press. New
York. 1980. 607p.
84
-------
40. Demmig, B., K. Winter, A. Kruger, and F. Czygan. Photoinhibition and
Zeaxanthin Formation in Intact Leaves. Plant Physiol., 84:218-224,
1987.
41. Hogsett, W.E., D.T. Tingey, and S.R. Holman. A Programmable Exposure
Control System for Determination of the Effects of Exposure Regimes
on Plant Growth. Atmos. Environ., 19:1135-1145, 1985.
42. Hogsett, W.E. and D.T. Tingey. Sensitivity of Important Western
Conifer Species to S02 and Seasonal Interaction of Acid Fog and
Ozone. EPA Forest Response Program Annual Meeting. Project Status
Reports, 11:271-278, 22-26 February 1988. Corpus Christi, TX.
43. Brown, J.S. Forms of Chlorophyll In Vivo. Ann. Rev. Plant Physiol.,
23:73-86, 1972.
44. Junge, W. Physical Aspects of Light Harvesting, Electron Transport
and Electrochemical Potential Generation in Photosynthesis of Green
Plants. In: A. Trebst and M. Avron, (Eds.) Encyclopedia of Plant
Physiology, Photosynthesis I. NS. Photosynthetic Electron Transport
and Photophosporylation. 1977. p.69-92.
45. Driscoe, D.M. Evaluation of Ozone Injury to Ponderosa and Jeffrey
Pines in Yosemite National Park, California. Final Report
CX-0001-4-0058 Air Quality Division, National Park Service, U.S.D.I.
1987. 166p.
46. Rock, B.N., T. Hoshizaki, and J.R. Miller. Comparison of In Situ and
Airborne Spectral Measurements of the Blue Shift Associated with
Forest Decline. Remote Sens, of Environment, 24:109-127, 1988.
47. Shutt, J.B., R.R. Rowland, and W.H. Heartly. A Laboratory
Investigation of a Physical Mechanism for the Extended Infrared
Absorption ('Red Shift') in Wheat. Int. J. Remote Sens., 5:92-102,
1984.
48. Vanderbilt, V.C., S.L. Ustin and J. Clark. Canopy Geometry Changes,
Due to Wind, Cause Red Edge Spectral Shift. IGARSS '88. Intl. Soc.
Geosci. and Remote Sens. Symp. 1988. (abstract)
49. Curtiss, B. and S.L. Ustin. The Characterization of Sources of
Illumination in a Ponderosa Pine (Pinus ponderosa) Forest Community
Using the Portable Instantaneous Display and Analysis Spectrometer.
SPIE Technical Symposium on Optics, Electro-Optics, and Sensors.
Orlando, FL 4-8 April 1988. (in press).
85
-------
RESULT OF AN EXPERIMENT USING THE ENVIRO-POD CAMERA SYSTEM TO
INVENTORY BUILDING SURFACE MATERIALS IN CINCINNATI, OHIO
R.A. Ellefsen, San Jose State University, Department of Geography, San
Jose, California 95192
INTRODUCTION
Oblique aerial photography, taken by the Environmental Protection
Agency's Enviro-Pod camera system, was employed in this project to
inventory building surface materials in Cincinnati, Ohio. The method
employed in the project - conducted by Lockheed Engineering and Management
Services Company, Inc. (LEMSCO) of Las Vegas, Nevada, for EPA's
Environmental Monitoring Systems Laboratory at the University of Nevada,
Las Vegas - was developed in a pilot study in Baltimore (1) supported by
the EPA's Vint Hill Farms, Virginia laboratory and by NASA's High Altitude
Program at Ames Laboratory, Moffett Field, in Sunnyvale, California. The
method is currently being refined in a study of the South Coast Air Basin
(greater Los Angeles area) conducted by members of the Geography Department
at San Jose State University for the California Air Resources Board.
The need to seek an alternative inventory method to field observation
was identified during formal review of the National Acid Precipitation
Assessment Program (NAPAP) materials survey, a cooperative venture of the
U.S. Army Corps of Engineers, the Environmental Protection Agency, the U.S.
Geological Survey, and the Department of Energy. Project workers
encountered the problem of significant variability of land uses within what
were assumed to be homogeneous urban land uses of: single family
residences; multiple family residences; industrial and commercial classes;
and Central Business Districts. Many of the selected sample structures in
the multiple family designation were, if fact, single family residences.
Several residences were also found in the livelihood/industrial class areas
and in Central Business Districts. In a LEMSCO random sampling of forty
structures (ten from each of the four land-use zones) in Cincinnati,
twenty-seven were found to be single family houses. These discrepancies
are traceable to the sampling frame, a combination of census tract data
(weak determiners of urban functional zones) and the U.S. Geological
Survey's Land Use Data and Analysis series of maps compiled at the
relatively small scale (for urban purposes) of 1:250,000. The minimum
mapping unit of 10 acres, imposed by the scale, was often too coarse to
account for the fine gYain existing in urban areas.
The use of the interpretation of finely detailed low-altitude
Enviro-Pod color transparency photographs for this project takes advantage
86
-------
of air photo interpretation's inherent values of: (1) being synoptic in
coverage; (2) allowing the interpreter in the laboratory to have a long
"dwell" observation time over the selected sample "target," (3) eliminating
interference met by field observers (suspicious people, angry dogs,
difficult weather, heavy traffic, etc.); and (4) the ability to make
photographic mensuration (measurements of wall surfaces minus windows and
doors). Taking aerial photographs from at least three compass directions
of all buildings further faciliated the process of surface materials
identification; field observers, on the ground, could often see only the
front and part of the sides of buildings visited.
The area selected for study was a large part of the total Greater
Cincinnati Metropolitan Region; because of difficulty in obtaining
supporting tabular and map data, Enviro-Pod photography taken of the area
south of the Ohio River in Kentucky was not interpreted. A total of some
614 square kilometers was included in the study. The territory extends
from the Ohio River on the south to the northern end of the contiguously
built-up area lying to the north of Interstate 275 and from Interstate 275
on the east side to a north-south line centering approximately on the
junction of Interstates 275 and 74.
METHODS
DATA ACQUISITION
The physical limitations of the Enviro-Pod system cause some
problems. First among these is the inflexibility of the camera system.
It allows changing of the camera's aperture only prior to installing the
Enviro-Pod on the aircraft and not while airborne. If interior remote
aperture control were possible during the mission, adjustment could be made
for variations in lighting conditions and surface feature reflectance.
Even better than manual control would be an automatic exposure device on
the camera; these are generally available for aerial camera systems.
Developing one for the Enviro-Pod system could probably be done fairly
simply.
An alternative to external aperture control would be to search for
optimal compromise exposures. The most common exposure problem in
Cincinnati was over exposure of bright, highly reflective areas; e.g., new
industrial areas. Reducing exposures for these areas would probably be
preferable to setting the exposure for dark areas (wooded), areas where
increased backlighting would illuminate the required detail. Another
possibility would be to plan flight lines so that as many of the bright
areas as possible could be photographed on the same flight line(s).
A final suggestion would be to fly the photo missions under a high
cloud cover. With proper aperture adjustment, photography under these
conditions produces images with good color saturation while at the same
time minimizing the loss of detail in building shadow areas. Obviously,
keeping a crew waiting for such optimal conditions would be questionable.
87
-------
Another camera system modification would be to experiment with longer
focal length lenses mounted on a different camera; the current standard
Enviro-Pod lens used is the 80 mm lens. While use of longer focal length
lenses'would increase image detail - and thus improve texture - the
resulting reduced ground coverage, however, would require more exposures.
An alternative strategy would be to photograph only the test sites at large
scale.
PROCEDURE
Basic steps in the .existing procedure should be continued, i.e.,
visiting the study city, taking ground reconnaissance and hand-held aerial
photographs, delimiting the urban terrain zones, taking the Enviro-Pod
images and then training and advising photo interpreters. Delimitation of
the urban terrain zones could be improved by using good high quality
vertical photographs and 1\ minute USGS quadrangle maps rather than the low
resolution index photos that were used in this project. Resulting urban
terrain zone boundaries would be sharpened, resulting in a better base for
sample selection.
A significant part of any further inventory study would be full
consideration of the floor space of the buildings in the urban terrain
zones rather than just the total ground surface space. A more realistic
numerical indicator of the relative importance of one urban terrain zone to
another could be determined. For instance, knowing the floor space of a
Central Business District - as compared, for example, to an industrial
district - would be of great value in calculating total wall surfaces.
An estimate - based on the taking of a few samples in Cincinnati -
was made of the amount of floor space encountered in the various terrain
zones. First, the ground space area covered by a building was calculated.
The multiplication of surface space - by multiple floors - was then taken
into account. This figure was then multiplied by the ground surface to
obtain a floor space total for each urban terrain zone. The results (seen
on Table 1) demonstrate clearly the proportionately high amount of floor
space (and thus, by extension, outer wall surface space) when the density
and height of buildings in a terrain zone is considered. For instance, the
small ground surface area of the Central Business District, with its large
number of closely spaced tall structures accounts for only 0.6 percent of
the total ground surface space but forms 7.7 percent of the total floor
space. Conversely, close-set detached single family residences form 33.6
percent of the total ground space but the proportion of floor space is only
23.1 percent.
If photo interpretation is used in another city, the process of
developing reliable indices for floor space should be integrated into the
total procedure. Floor space proportions will probably replicate from city
to city. At the very least, variances will narrow upon applying the method
over a wide number of cities.
88
-------
EXPERIMENT PROCEDURES
The procedure employed in the study consisted of the distinct steps
of: (1) data gathering; (2) data manipulation; (3)sample selection; (4)
photo interpretation; (5) mensuration; and (6) data aggregation.
TABLE 1. TOTAL FLOOR AREA COMPARED TO GROUND AREA OF MAJOR
URBAN TERRAIN ZONES.
Urban Terrain Zone
High Density Housing (Dc3)
Low Density Housing (Do3)
Rail -Related Indus/Stor. (Dc4)
Low Density Indus/Stor. (Do4)
Administrative/Cultural (Do6)
Shopping Center (Dol)
Core Area (Al and Del)
Urb. Ter. Zone
Ground Area As
A Percent of
Total Ground Area
33.6
18.3
15.6
9.2
8.8
3.4
0.6
Urb. Ter. Zone
Floor Area As A
Percent of
Tot. Floor Area
23.1
4.2
25.1
6.3
14.1
2.3
7.7
DATA GATHERING
Three distinct operations were involved: a ground reconnaissance;
taking hand-held 35mm photographs of selected test areas; and the aerial
acquisition of Enviro-Pod images. Each served a special purpose in the
project and each was an integral part of the whole.
Ground Reconnaissance
Five days were spent in the Greater Cincinnati area examining in
detail examples of the various urban terrain zones (explained below) that
comprise the metropolitan area. Ground photographs were taken of
representative scenes within the zones and of numerous example buildings
for the purpose of serving as ground truth when compared with aerial views
of the same scenes. These were essential in the training of the LEMSCO
photo interpreters. Building construction type and wall materials were
noted. Examples of areas investigated were: Covington and Newport,
Kentucky; Cincinnati's Central Business District; several industrial areas
along Mill Creek; older housing near the University of Cincinnati; the
campus itself; commercial and industrial areas in Norwood; shopping centers
near the edge of the city; the older communities of Cheviot and Mount
Healthy; business parks near the interstates in the northern part of the
city; new truck-related industrial parks in several areas; and medium and
89
-------
high income level residential areas throughout the metropolitan area
(working in corporate Cincinnati alone would omit large sectors of Greater
Cincinnati).
Hand-Held Aerial Photography
Selected urban terrain zone sites that had been visited on the ground.
were photographed from a light aircraft. This mission, accomplished in two
days, yielded several rolls of imagery; shooting the photographs through an
open window, rather than through plexi-glass, ensured clarity. The flights
were made at the FAA-required minimum altitude of 1,000 feet (above the
ground) and taken with a 28-210 zoom lens. Various zoom focal lengths were
employed to meet different requirements. The sites - that had previously
been marked on a street map - were first located from the air and then
photographed from different perspectives ranging from near vertical to
shallow oblique. Care was taken to insure that structure walls were
clearly photographed.
Enviro-Pod Photography
The process of taking Enviro-Pod imagery consisted of four steps:
flight line planning; mounting and operation of the camera system; flying
the mission; and processing of the film. First, flight lines were planned
(and drawn on the map) to provide coverage of all of the Greater Cincinnati
area from at least three directions and with optimum side lap and forward
lap. Following available guides, plus experience gained in taking
Enviro-Pod imagery over 35 sites in the Los Angeles area, flight lines were
drawn at approximately 1.5 kilometers apart. The intervalometer (a timing
device to expose film at set intervals) was set at 6.5 seconds. To achieve
the desired degree of overlap (neither too much nor too little) with these
settings required that aircraft maintain a steady speed of 90 knots and an
altitude of 1,000 feet.
A total of sixteen north-south flight lines and 22 east-west flight
lines were drawn. The former were approximately 32 kilometers in length
and the latter approximately 30 kilometers; east-west lines were a little
shorter in the extreme southern part of the map; north-south lines were
shorter on the western side of the city.
The flying mission extended over a period of seven days. Actual
flying time was restricted to the optimum hours (for maximum light and
minimum shadows) of between 1000 and 1500 hours each day; heavy cloud cover
precluded flying for certain periods.
In practice, either three or four passes (either E-W or N-S) could be
flown with a singe roll of 200-foct long film (each roll permitting
approximately 300 exposures). After exposing a roll, the crew returned to
the base (Hamilton Air Field, Hamilton, OH) to remove the camera from the
pod (attached to the bottom of the aircraft) and to mount another
pre-loaded camera. A total of four separate cameras were used in the
operation. Fast turnaround time is desirable, especially when cloud
conditions can change quickly.
90
-------
Navigation was facilitated by noting on the first pass some
significant landmarks for the next (and parallel) flight line. This next
flight line was then "dry run" on the return to the starting location to
spot landmarks that would serve during the live flight. Cross winds
required the pilot to compensate by "crabbing" slightly into the wind in
order to remain on course; no observable difficulties are obvious on the
film as the angle was relatively shallow.
While the Enviro-Pod camera system can take both vertical and oblique
photographs, only the oblique was used; vertical photography is available
from general sources. The film product, for each of the fourteen rolls
exposed, was a series of 300 exposures per roll at a size of 70 by 200
millimeters. The film used was Kodak Aero Ektrachrome 2448 (full color
transparency format). Resolution quality is high; for instance, people who
were on the street are easily observed (see Figure 1, an example of one of
the ±4,000 images acquired of the Cincinnati metropolitan area). The
exposure settings chosen for the camera were a compromise. Only extreme
differences in surface lighting and reflectivity as between wooded areas
on the dark side and highly reflective parking lots, light-colored
buildings and rooftops caused some photos locally to be either
underexposed or overexposed.
Figure 1. Enviro-Pod color photograph of an area east and north of
downtown Cincinnati, OH, taken from 1,000 feet above ground
level at a depression angle of 45 degrees.
Photo Data Manipulation
After completion of the photo mission film was sent for development
to HAS Images, Incorporated in Dayton, OH. Technicians there developed a
test strip of each roll to determine processing time and then proceeded to
develop each roll.
LEMSCO identified each exposure for each roll with a unique number.
Individual photos were then matched to each of the test sites; three to
four photos of the test site were sometimes available. Exposures were
91
-------
assembled for each test site to provide the photo interpreter with views
from as many different perspectives as possible.
Sample Selection
Selection of specific sites for interpretation consisted of two
phases: the delimitation of urban terrain zones followed by the random
selection of blocks, and then particular buildings within the zones.
Urban Terrain Zone Delimitation
Urban terrain zones (2), defined as areas that are homogeneous in
both functional and morphological characteristics (Table 2), were delimited
for the entire metropolitan area. This procedure was accomplished using a
paper Diazo copy of a small black and white series of photos designed to
serve as an index to large-scale photos.
TABLE 2. URBAN TERRAIN ZONE CLASSIFICATION SYSTEM
Attached Buildings Classes
Al - Commercial offices, retail; core area; mostly high-rise; mass and framed construction
A2 - Apartments/hotels; near core area, medium-rise; brick construction common
A3 - Apartments and abutted-wall houses; adjacent to core; low-rise, brick construction
common
A4 - Industrial/storage; near core area; medium rise; mass and framed construction
A5 - Commercial ribbons; on arterials; low to medium rise; brick construction common
Detached Classes
Close-set Buildings (forming at least 75 percent street frontage)
Del - Commercial office; often in core redevelopment areas; high rise; light-clad framed
Dc2 - Residential apartments; widely distributed in city; mass and framed construction
Dc3 - Residential single-family houses; widely distributed; mass and framed construction
Dc4 - Industrial/storage; linear building pattern; railroad or dock related; low rise; mass
and framed
Dc5 - Commercial offices; (Outer City) locations; high-rise; light-clad framed construction
Detached Building Classes
Open-Set Buildings (forming less than 75 percent street frontage)
Dol - Shopping centers; beyond core; low-rise; mass and framed construction
Do2 - Apartments (usually planned units); widely distributed; low to medium rise; framed
construction
Do3 - Single-family houses; away from city core; low rise; frame-construction most common
Oo4 - Industrial/storage; in peripheral industrial parks; truck related; low rise; mass and
framed
Do5 - Modern commercial ribbons; on major arterials; low-rise; mass and framed construction
Do6 - Administrative/cultural; widely distributed; low to medium rise; mass and framed
construction
The terrain zone delimitation that resulted consists of several
hundred individual polygons; each urban terrain zone has several
representatives over the total study area. For example, the category of
92
-------
detached, close-set residential land occupies a great deal of territory but
is not all contiguous. An example of an urban terrain zone class
demonstrating a wide spatial distribution is administrative/cultural (such
land uses as schools with their buildings and grounds). By contrast, the
urban terrain zone Al (attached 1) and Del (detached, close-set 1)
comprising central business districts, is encountered only in Cincinnati's
downtown and in the larger of the outlying, but new subsumed, older towns
that are today an integral part of the metropolitan area.
Most of the urban terrain zones appear in Cincinnati. As elsewhere,
some are more commonly encountered than others. The residential zones are
the largest occupiers of surface space with the Dc2, Dc3, Do2, and Do3
collectively accounting for some 55 percent of the total ground space in
the metropolitan area. As might be expected in a city of the age of
Cincinnati (where so much growth and development took place in the decades
from the end of the Civil War to the 1920's) a high proportion of the
residential land use consists of close-set, but detached houses. The area
in attached housing (unlike the pattern seen in older Philadelphia and
Baltimore) is quite small. Large areas of post World War II housing tracts
at the edge of the metropolitan area are of the open-set variety, as
expressed in the 17 percent figure recorded for urban terrain zone class
Do3.
Sample Block Selection--
A total of ten locations from within each of the fourteen urban
terrain zone types was deemed to be a sufficiently large sample. These
sites, or individual blocks or their equivalents, were selected from the
photo-map on which the urban terrain zones had previously been delimited.
Interpretation of Enviro-Pod Imagery--
Heavy use has been made in the interpretation process of supporting,
or ancillary information. Reliance on the traditional air photo
interpretation principles of object recognition alone (3), i.e., shape,
size, photographic tone and color, pattern, shadow, topographic location,
and texture, is inadequate considering the need to distinguish particular
types of building materials (such as painted brick), especially given the
scale provided by the Enviro-Pod camera system (the U.S. Air Force's KA 85
with a fixed standard 80 mm lens) and a flying altitude of 1,000 feet.
Because it was realized as a result of parallel work in Baltimore and
Los Angeles that simplistic use of the traditional photo interpretation
principles is inadequate for the levels of discreteness of building
materials sought in this project, contextual information on building
construction types had to be employed. Earlier studies (1,4) had developed
principles on a variety of building characteristics ranging from wall type
and thickness to surface materials. Throughout, the generalization was
made that major classes of buildings had distinctive suites of
characteristics, not only the obvious architectural style but building
materials that were common to the time of construction, and the dictates of
both construction requirements and surface ornamentation to meet desirable
styling.
93
-------
Accordingly, heavy dependence has been placed on training the photo
interpreters in the rudiments of building construction types and their
related surface materials. Stressed in a workshop conducted in LEMSCO's
photo-interpretation facility, was the need to understand forms of building
construction. With this background knowledge, photo interpreters were
directed to adopt the procedure of first identifying the building
construction type and then to follow an elimination or branching key
devised especially for the project (Figure 2).
The key is designed to take the photo interpreter from a general view
of the urban terrain zone, in which the sample(s) belong, to the specific
building surface material. Thus, moving through the key from top to
bottom, the interpreter first notes the urban terrain zone type involved.
He/she then notes the type(s) of building construction commonly found in
that type of urban terrain zone (the variance is predominantly attributable
to age of the zone, the related type of building construction, and the
intended function).
With these anticipated buildings types in mind, the photo interpreter
then proceeds to the "Branching Key" part of the chart. Determination is
first made of the basic styple of the building, either mass or framed.
Information to the right side of the terms aids in the identification
process. The route then proceeds to another common physical characteristic
(low or high) in the case of framed buildings. From that juncture point it
moves to yet another level of identification, one that is in part function,
e.g., houses versus industrial/storage, or to heavy cladding versus light
cladding in the case of tall frame buildings.
Having gone through all these determinations, the photo interpreter
then uses a checklist of the characteristics indicated for each of these
sub-types. These vary by the presence of such identifying features as
parapets on the roof to typical architectural features such as pediments,
shafts, and capitals.
At this point, the photo interpreter examines the list of possible
building surface materials to be expected for the type (and sub-type) of
the buildings already identified. Reference is made in this part of the
process to the traditional photo interpretation keys of, for example, color
and texture.
The type of material is then noted for all walls of the sample
buildings under observation and noted on the cadastral maps. These maps
indicate the exact location and size of buildings on ownership parcels.
The scale is large enough (at 1" = 100' or 1:1200) to have sufficient space
to encode the type of building material identified within the building's
perimeter.
Mensuration--
Mensuration (literally measurement) of the building surface materials
was achieved through reference to the outlines of the buildings (from
cadastral maps) and to the established scale of the oblique Enviro-Pod
photographs. In the process, the lengths of the walls and building height
94
-------
GUIDE TO BUILDING MATERIALS INTERPRETATION
I Not* Urban Terrain Zone Typo
ID
en
Not* luHdlng Surfaca Malarial. Poaalbla
For Each Urban TarraM Zona
1
A I/DC i
MM
tent
FwiMil Hwvy CM
Tarn COM
COMTM*
BTM*
fencrm
MMM
1
A
MottryMni
»•*
aVldl
brnthMn.
2
d L*ttCM
A
HOOK,
Br«
3
htw
*
A
Min
frM*
FlWMdf
BTK*
Concra
' Mttti
4
•wry CM
I
A
MM
•nek
Dmctw
MMI
Wood!
S
Frmd
0
I
w
»
*
El
Mfc
90d
BCD
D
Itbal
Rt
Cone*
fVwiwd
frKk
Frinwd
Maul
T
C4
Mttodt)
HMvyCM
LifMCM
o
Ffwiwd
5KHW
Concr
Ural
CDmp
cS
L«MCM
Ml
0
MM
Concr
Frwntd
BricT
MM*I
Go rope
]
>1
«• land
«hlCM
N«
D
Frttmd L
Compo
Mtood
Siucco
>a
jM CM
man
D
FiwiMd
Br«*
Ston
ld*i
StUM
Br«*
>3
O«
MM
Conoi
Frwiwd
Maut
»^i
I
r4
«MCM
tronu)
D
ktoM
Cc«i
FrwrM
D»
«MCHJ
D
Uut
9te
Co
Fram
•r
Mb
Co
>•
rtamf
*a
dt
Ml
ncriM
I Photo Intarpratatlon Brandling «•»
| FRAMED | —
FRAMED Gwvil CtartCMrmia
'E*v>' tttidiraM of Mitt
Both MN*nd to* buiWmti
LOW SUILOWaS
ttfl • •tertot)
TALL BUILDINGS
(0 «t«rt«* •n4 «*••)
COfnirweiat O«m»Mia
Urft riwp dupWy v
SKT
Figure 2. Guide to building materials interpretation.
-------
were measured and then multiplied to acquire wall area; account was taken
of variations encountered at points on the oblique photographs away from
the center. Venting (doors and windows) was measured and the collective
area subtracted from the wall total to yield the area of surface material.
A detailed account of procedures used in mensuration, all aspects of the
photo interpretation process, and data manipulation is given in a recent
LEMSCO report (5).
Data Aggregation--
Wall surface totals for each building were placed into groups
according to Urban Terrain Zones, building construction type, and other
characteristics.
RESULTS AND DISCUSSION
Amounts and proportions of wall surface materials were recorded and
aggregated into various packages. In the most straightforward mode, wall
surfaces were recorded by type for each of the urban terrain zones. In the
aggregate - as might be anticipated considering the general appearance and
age of Cincinnati - brick was the dominant surface material, forming some
59.3 percent of the total (see Table 3). Concrete followed at a little
over a quarter of the total (26.7 persent). At the low end of the scale
were: concrete block (9.0 percent); wood (2.5 percent); and metal (2.3
percent).
TABLE 3. BUILDING MATERIALS: PROPORTIONS
ALL CINCINNATI SAMPLE AREAS
Building Material
Brick
Concrete
Concrete Block
Wood
Metal
Area (m2)
1,143,095
515,113
174,277
49,703
44,715
Percent of Total
59.3
26.7
9.0
2.5
2.3
Total
1,926,903
99.8
In response to a key question posed by EPA scientists regarding the
amount of painted surface of each material, the aggregate figures (Table 4)
show that: as expected, 99.1 percent of the wood was painted; concrete
block was almost always painted (96.8 percent); and poured concrete was
painted more often Jthan not (76.5 percent). Metal was in a similar
situation being painted 72.2 percent of the time. Only brick was
infrequently painted, at 8.9 percent of the total of all brick surfaces.
96
-------
TABLE 4. BUILDING MATERIALS: PRECENT PAINTED
ALL CINCINNATI SAMPLE AREAS
Building Material Percent Painted
Brick 8.9
Concrete 75.5
Concrete Block 96.8
Wood 99.1
Metal 72.2
Brick, owing to its natural decorative nature, was only seen in field
observation to be painted to improve appearances of older structures;
commercial buildings were common instances.
The high incidence of brick as a building material in Cincinnati led
to questioning its spatial distribution within the city's urban terrain
zones. As seen in Table 5, a very high occurrence is found in older areas
such as the core and its periphery and in fairly high proportions
elsewhere. The core periphery (Urban Terrain Zone A2), with a total of
98.4 percent brick, is an old section lying north of the heavily modified
Central Business District and composed largely of old three- to five-story
brick apartments, some commercial buildings, and some large brick churches.
The core area itself has a high proportion of brick surfaces (at 88.1
percent). Much of the brick here is cladding for framed high-rise
buildings. Remaining urban terrain zones with high proportions of brick
consist of high density apartments (Dc2) at 81.6 percent; low density
(generally newer) apartment areas at 75.5 percent; old core periphery
industrial/storage areas (A4) at 72.5 percent; old string street commercial
areas (A5) with their succession of attached low-rise brick stores at 72.2
percent; the very large low density single-family housing (Do3) at 67.1
percent; administrative/cultural areas (Do6), many of which are school and
church buildings, at 60.0 percent; and the old high-density, single-family
housing (Dc3) at 52.8 percent. The remaining urban terrain zones ranged
from 9.9 to 27.0 percent brick construction.
97
-------
TABLE 5. URBAN TERRAIN ZONES WITH HIGH
PROPORTIONS OF BRICK SURFACES
Urban Terrain Zone Percent Brick
Core Periphery (A2) 98.4
Core Area (Al and Del) 88.1
High Density Apartments (Al and Del) 81.6
Low Density Apartments (Do2) 75.5
Old Core Periphery Industry (A4) 72.5
Old String Streets (A5) 72.2
Low Density Housing (Do3) 67.1
Administrative/Cultural (Do6) 60.0
High Density Housing (Dc3) 52.8
Others 9.9 to 27.0
The spatial distribution patterns exhibited here are probably
replicated in other eastern U.S. cities, and should be of value when
invoked in the extrapolation and modeling process. Some adjustments for
heavy usage of local building materials would be required. For instance,
the use of local conifer softwoods as a home building material is apparent
in such areas as the Pacific Northwest, the South, and northern New
England.
Concrete surface materials, on the other hand, are found to dominate
in industrial and commercial areas, and particularly those that are of
recent vintage. As Table 6 indicates, Shopping Center (Dol) urban terrain
zones are mostly concrete with 87.3 of the total; the recent low density
industrial parks (Do4) at the edge of the metropolitan area (and
particularly near the by-pass interstate highway to the north (1-275) show
proportions of concrete (57.0 percent). The older railroad-related areas
have some concrete buildings (37.3 percent of total wall surfaces), while
the core periphery Industrial/Storage areas (A4) have only 21.2 percent.
98
-------
TABLE 6. CONCRETE SURFACE MATERIALS IN INDUSTRIAL/STORAGE ZONES.
Urban Terrain Zones Percent Concrete
Shopping Centers (Dol) 87.3
Low Density Industrial/Storage (Do4) 57.0
Railroad Related Industrial/Storage (Dc4) 37.3
Core Periphery Industrial/Storage (A4) 21.2
CONCLUSIONS
Considering the difficulties faced in the inventorying process in
earlier NAPAP work, the principal purpose of this study was to investigate
the feasibility of employing reconnaissance quality oblique aerial
photography to identify and measure building surface materials. Focus,
accordingly, is on evaluation of the soundness and viability of the method
and the quality of the product.
EVALUATION OF SOUNDNESS AND VIABILITY
A number of advantages of the method are readily enumerated. The
first is the advantage of having a synoptic view. The photo interpreter
takes advantage of the "view from above," one in which he/she can see not
only the "target" sample buildings but their settings and surrounding
buildings to interpret and classify the data. For another, air photo
interpretation is a well developed discipline, one with extensive support
in the literature on methods of interpretation and mensuration. Yet
another is the physical ease of interpreting the photography in a
laboratory setting; without need to regard the weather, the interpreters
can comfortably use sophisticated equipment to examine the scene "below."
The ability to "dwell" over the target scale is extremely valuable,
especially when compared to the known frustrations of site field visits.
Also, the cost of a site visit is negated. This translates into making far
more observations and measurements for more sample sites than would ever be
possible from field visits alone.
Another advantage is the ease of making measurements from the
photographs. The photo interpreter can readily make measurements on-the
photographs through use of magnifying loupes equipped with reticles and/or
by following other established methods. Measurement accuracy at the scale
of the Enviro-Pod photos - nominally 1:5,000 - was acceptable to EPA for
this study. Cross-checking against cadastral maps of the area can serve to
confirm measurements.
A final advantage is that all sides of every building can be seen
through use of photography taken from different directions. This is
especially advantageous when compared with field observation where often a
view of the rear of a building is denied. Further, though not part of this
99
-------
inventory, the aerial photography provides information on the character of
the roofs. Especially in the case of flat roofs, this information is
virtually impossible to gain from the field.
Some negatives are present, however. Of the greatest significance is
the inability of photo interpreting small features on buildings such as
gutters, downspouts, and window sashes. The value of attaining such
information has to be set against the advantages of using photo
interpretation. Or, consideration could be given to combining aerial
photography with field visits. Measurements of wall lengths and materials
surfaces could be made on the photographs, and small details from field
observation could be added for selected sample structures.
Costs of acquiring the photography, training for the photo
interpreters, and the labor involved can be assessed only in relation to
the total costs of the field observation method; cost-effectiveness studies
of the two approaches have not been conducted. In favor of air photo
interpretation, it should be noted that costs go down per unit while the
cost of visits to each sample building by a field team stays essentially
the same per unit regardless of how many sites are visited.
RECOMMENDATIONS
The most general recommendation is to urge that the method reported
here be employed for other cities in the U.S. where inventorying of surface
materials is contemplated. The method appears to be sound and the
procedures sufficiently developed to the point where it is nearly ready to
go operational; some improvements could undoubtedly be made in the course
of studying a second trial city.
The steps to be followed in use of the method for other cities would
be: (1) performing a ground reconnaissance; (2) acquiring hand-held 35mm
aerial photography (to be used in training and in ground-truthing); (3)
delimiting the urban terrain zones; (4) finding reliable floor-space
indices; (5) measuring urban terrain zones; (6) acquiring the Enviro-Pod
imagery; (7) interpreting that imagery; (8) making photo mensurations of
wall space; (9) considering use of the large amount of information
available in the imagery on roof characteristics; (10) placing the
information into a Geographic Information System; (11) interacting that
information with such data as meteorological measurements including
prevailing wind direction, pollution sources, etc.; and (12) developing a
model that would broadly serve the goals of NAPAP.
Specific recommendations stemming from lessons learned in the project
are largely technical and can be divided into those dealing first with data
acquisition and then with procedure.
100
-------
REFERENCES
1. Ellefsen, R., and B. Coffland. Using Enviropod Low Altitude Imagery
to Inventory Building Surface Materials for an Acid Rain Study: A
Baltimore Example. In Proceedings of the 1987 ASPRS-ACSM Fall
Convention, American Society for Photogrammetry and Remote Sensing,
Falls Church, Virginia, 1987. pp. 170-176.
2. Ellefsen, R. Urban Terrain Zone Characteristics. U.S. Army Human
Engineering Laboratory, Aberdeen Proving Ground, Maryland, 1987. 358
pp.
3. Avery, I.E., and G. Berlin. Interpretation of Aerial Photographs.
Burgess Publishing, Minneapolis, Minnesota, 1985. 554 pp.
4. Ellefsen, R., A. Carlson, and B. Thein. Urban Terrain Analysis.
U.S. Army Human Engineering Laboratory, Aberdeen Proving Ground,
Maryland, 1981. 255 pp.
5. Finkbeiner, M.A., R. Ellefsen, J. Engels, D. Williams, L. Ogiela, and
J. Teberg. Use of Oblique Aerial Photography to Develop an Inventory
of Building Surface Materials. Internal Report. U.S. Environmental
Protection Agency, Las Vegas, Nevada, 1988.
BIBLIOGRAPHY
Ellefsen, R. B. Coffland, and G. Orr. 1977. Urban Building
Characteristics, Setting and Structure of Building Types in Selected
World Cities. Naval Surface Weapons Center, Dahlgren, Virginia. 370
PP-
Engels, J.L. 1987. Verification of the U.S. Army Corps of Engineers
Building Surfaces Inventory. Internal Report. U.S. Environmental
Protection Agency, Las Vegas, Nevada.
Life Systems, Inc. 1986. Development of Extrapolation Procedures for a
Materials Distribution Data Base. Meeting Summary Report. Internal
Report. U.S. Environmental Protection Agency, Washington, D.C.
Lipfert, F.W., and M.R. Torpey. 1984. Methods for Materials Inventorying
in High-Rise Center Business Districts. Internal Report. Brookhaven
National Laboratory, Upton, New York.
Merry, C.J., and P.J. LaPotin. 1986. A Description of the Building
Materials Data Base for Cincinnati, Ohio. U.S. Army Corps of
Engineers, Hanover, New Hampshire.
101
-------
THE DETECTION OF ACID RAIN DAMAGE TO BUILDING
STONE USING SPECTRAL REFLECTANCE MEASUREMENTS
by: Marguerite J. Kingston, U.S. Geological Survey, Reston, Virginia
22092
The work described 'in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
ABSTRACT
Diffuse reflectance spectroscopy in the near-infrared (NIR) spectral
region is a useful method for the nondestructive measurement of the
mineralogical composition of stone surfaces. These measurements may be
used to monitor progressive mineralogical changes on building stone
surfaces linked to the deterioration of the stone due to atmospheric
pollution. In this study, gypsum accumulation on limestone and marble was
determined by NIR reflectance measurements. The precipitation of gypsum on
stone surfaces is a result of the interaction between atmospherically
derived H2S04 with carbonate.
Field data have been collected annually at four tests sites in the
eastern United States which were installed in late 1984 under the National
Acid Precipitation Assessment Program (NAPAP). The test sites are located
near meteorological monitoring stations to achieve correlation between
environmental conditions and material degradation. At each site, NIR
spectra were recorded with a portable spectrometer which scans the 0.4- to
2.5-/zm wavelength region. Results of these field studies indicate that no
measurable gypsum accumulated on the boldly exposed upper surfaces, but
varying amounts of gypsum were concentrated on the sheltered lower surfaces
of both limestone and marble briquettes at each site. Most gypsum
accumulated on the under surface of briquettes exposed in Washington, D.C.,
and the least occurrence of gypsum was measured at the Newcomb, New York,
site in Adirondack Park.
ACKNOWLEDGMENTS
This research was funded by the National Park Service as part of the
National Acid Precipitation Assessment Program. The author is grateful to
102
-------
David Brickey and Malcolm Ross (both of the USGS) for their many useful
suggestions during a critical review of the manuscript, to Arthur Youngdahl
(Argonne National Laboratory) for providing essential chemical analytical
data, and to Shelvie Burks who was responsible for typing.
INTRODUCTION
The National Acid Precipitation Assessment Program (NAPAP) has been
authorized by Congress under the Acid Precipitation Act to include a study
of the effects of acidic deposition on materials' surfaces in the United
States. The National Park Service has the lead role under NAPAP to survey
the damage to materials (e.g., stone, metal, and paint) from acid
deposition whose cause traverses state boundaries.
One way to assess damage to building and monument stone is to expose
these materials to the environment at representative sites (1). In 1983
four field test sites for exposing dimension stone were selected in the
eastern United States, at Research Triangle Park (RTP) near Raleigh, North
Carolina, the roof of the West End Library in downtown Washington, D.C.,
the Environmental Measurements Laboratory at Chester, New Jersey, and the
Huntington Wildlife Forest near Newcomb, New York. The field study was
designed to determine initial, seasonal, and long-term effects of ambient
acidic deposition and gaseous pollution on the weathering of economically
important building stone (2). The building stone employed at the sites,
the Salem Limestone of Indiana and the "Royal" variety of Shelburne Marble
of Vermont were selected because of their historical importance, their
long-term availability, and their presumed susceptibility to damage within
a short-term study period (3,4). Vermont marble was selected not only
because it is the most common polished marble but also because of its
cultural and economic importance as an ornamental stone (4). The stones
were quarried and commercially finished in June and December 1983. Both
marble and limestone were cut into slabs (5.1 by 30.5 by 61.0 cm) to be
used for rain run-off experiments and briquettes (5.1 by 7.5 by 8.5 cm) for
multiple studies.
Mineralogical characterization using X-ray powder diffraction
analysis, light and electron optical techniques and energy dispersive X-ray
chemical analysis was carried out by McGee (5) and Ross (3) of the USGS.
The gray-colored Salem Limestone is texturally homogenous and is comprised
predominantly of oolitic fossil debris. The very white Shelburne Marble is
also calcitic with blue-grey streaks of inclusions which contain phlogopite
with minor dolomite, chlorite, and muscovite.
Each of the four field sites includes an established environmental
monitoring station with the capability of measuring SO,, N0x, and 03
precipitation chemistry and intensity, wind, and temperature parameters
(6). At each site, two adjacent south facing sample racks hold an array of
72 limestone and 72 marble briquettes which are slanted upward at a
30-degree angle from the horizontal. Slabs are mounted on adjacent racks
and also oriented to slant 30° upward. Initial plans for the briquette
103
-------
measurements included: color change and weight loss, surface roughness and
recession, mineralogical and chemical profiling, surface chemistry, and
scanning laser acoustic monitoring of surface attenuation, as well as
visible and near-infrared spectral reflectance measurements. It is the
intent of the materials effects research group to synthesize the results of
all these studies in order to determine a dose response curve which will
describe the kinetics of carbonate building stone degradation as a function
of the ambient environment at the four test sites (7).
Scientists in Europe and more recently in the United States have
recognized the role of pollutants such as S02 and acid rain in contributing
to the accelerated weathering of carbonate stone buildings and monuments
without a complete understanding of the mechanics of the degradation
processes (8,9). It has long been known that gypsum is the product of a
complex series of reactions involving sulfur dioxide, water and carbonate.
Gypsum has a solubility product of 2.4 x 10"5, several times greater than
that of calcite which is 4.8 x 10"9. If gypsum forms as a result of SO, or
acid rain pollution on carbonate stone and is subsequently exposea to
precipitation, it will dissolve and be washed away after formation on
stone, resulting in weight loss and textural changes (10). In Britain, an
on-site exposure experiment demonstrated that the accumulation of gypsum on
limestone samples protected from direct rainfall was proportional to the
weight loss of unprotected stone at the site (11). The authors concluded
that gypsum accumulation on protected stone surfaces is a better indicator
of stone deterioration than are measurements of ambient S02 levels in the
atmosphere.
This paper will describe the preliminary results of a three-year
study of the use of NIR reflectance spectroscopy for monitoring gypsum
accumulation on the surface of limestone and marble briquettes exposed at
the four test sites. A purpose of this study was to demonstrate the
feasibility of this nondestructive in situ technique for determining the
rate of materials degradation as a function of environmental pollutants.
SPECTRAL REFLECTANCE
Diffuse reflectance spectroscopy in the visible to near-infrared
(VIS-NIR) region is a rapid method for the in situ and nondestructive
measurement of the mineralogical composition of stone surfaces. Spectral
reflectance measures that component of radiation which penetrates at least
to 40 urn into the medium and undergoes multiple scattering at the surface
of individual mineral particles. Spectral features are produced when
photons have passed through a volume of the material and have been absorbed
at certain wavelengths before being scattered or refracted. In the VIS-NIR
region, spectral absorption features are produced by two mechanisms,
electronic processes and by overtone and combination tones of fundamental
vibrational processes (12). The intensity of absorption bands of a mixture
of minerals may be related to the relative proportions of the minerals
composing the stone surface (13). Variations in particle size, shape and
packing density also affect absorption band intensity. Of interest to this
104
-------
study are the absorption features related to the vibrational modes of the
C03, OH and HOH groups.
In the NIR, carbonate spectra (Figure 1) are characterized by five
distinct bands which derive from bending and stretching modes of the C-0
bond (14). The Salem Limestone displays these, with intense bands at 2.35
and 2.5 urn, and weak features at 1.88, 2.0, and 2.16 /zm. Weak features
near 1.94 and 1.4 /jm due to OH stretching modes and HOH bending modes can
be attributed to fluid inclusions within the carbonate mineral grains. The
spectrum of the Shelburne Marble (Figure 2) exhibits absorption bands at
the same wavelengths as those in the Salem Limestone but they are more
intense. Gypsum also displays characteristic features (Figure 3) in the
VIS-NIR by a prominent series of intense absorption bands due to the water
of hydration, including a multiple band centered near 1.94 urn.
100 r
SALEM LIMESTONE
50
0.7
0,9
1,5 1.7
WAVELENGTH,/i m
1.9
2.3
2.5
Figure 1. Visible to near-infrared spectrum of freshly quarried Salem
Limestones showing absorption features due to carbonate,
hydroxyl, and water.
105
-------
100
90
80
S70
o
ui 60
CL
50
£30
20
10
0
SHELBURNE MARBLE
0,5
1.0 1,5
WAVELENGTH,
2.0
2.5
Figure 2. Visible to near-infrared spectrum of freshly quarried shelburne
Marble showing absorption features due to carbonate, hydroxyl,
and water.
106
-------
1001-
GYPSUM
1.0
.21
1.43
1.64 1.86
WAVELENGTH
2.07 2.29 2.50
Figure 3. Visible to near-infrared spectrum of gypsum showing absorption
features due to overtones and combination tones of the water
fundamental.
107
-------
To establish baseline spectral reflectance data, visible and
near-infrared (0.4 to 2.5 urn), measurements were collected and stored for
each limestone and marble briquette before the samples had been exposed at
the four test sites. An additional set of measurements was made at each of
the test sites shortly after the briquettes were mounted on the racks.
Measurements were made with a portable spectrometer designed and
constructed by Geophysical Environmental Research Inc. (GER)1. The
instrument uses a grating monochrometer design to make high resolution
measurements in the 0.4-2.5 /zm range. The detectors are controlled by a
battery operated electronics unit containing a microprocessor. Data may be
plotted in the field and is temporarily stored on cassettes. These files
are later transferred to a mini-computer for data analysis and storage.
The initial experimental protocol was to make several spectral
measurements on the exposed upper surface of the briquettes. However, in
the fall of 1984, it was noted that a tan stain had developed within the
center of the downward facing and protected under surfaces of some of the
briquettes at the North Carolina site. An absorption feature near 1.94 /zm
appeared or increased in depth in the spectra measured from the limestone
under-surface. Henceforth, the under-surfaces as well as the
upper-surfaces of the briquettes were measured at each site. Spectral
measurements were collected at each of the four sites in 1985 and 1986. An
artificial light source was used (even on cloud-free days) so that the
absorption bands near 1.9 urn would not be obscured by atmospheric
attenuation. The spectral reflectance curve for the upper surface of the
briquettes has remained unchanged during this study.
CALIBRATION
Contemporaneously with the initial measurements of the briquettes at
the test sites, we conducted laboratory experiments at the U.S. Geological
Survey to determine the composite reflectance spectra of known increments
of gypsum present on limestone and marble surfaces (15,16). The band depth
intensity (also defined as band strength) of the absorption feature
centered near 1.94 urn increased with the chemically determined gypsum
concentration (Figures 4a and 4b). Band depth (BD), a dimensionless unit,
which is proportional to the amount of radiation absorbed, was defined by
Clark and Rousch (13) as
BD - 1 - Rb/Rc
where Rb is the percent reflectance at the band center and Re is the
percent reflectance of the spectral continuum adjacent to the band. The
percent reflectance at 1.8 fan was determined for the spectral continuum.
1. Any use of trademarks or trade names in this publication is for
descriptive purposes only and does not constitute endorsement by the U.S.
Geological Survey
108
-------
100 r-
/SALEM LIMESTONE
\
1.5
1.7
1.9 2.1
WAVELENGTH (|im)
2.3
Figure 4a. Near infrared spectra of fresh Salem Limestone, of pure gypsum
and of limestone plus gypsum. Note adsorption bands centered
near 1.94 jum.
100 r
80
i-
O
cr
60
UJ
o
40
UJ
20
.^MELBOURNE MARBLE
_L
JL
1.5 1.7 1.9 2.1
WAVELENGTH (pm)
2.3
2.5
Figure 4b. Near infrared spectra of fresh Shelburne Marble, of pure gypsum
and of marble plus gypsum. Note absorption bands centered near
1.94 fj.m.
109
-------
The experiments showed that the spectral response of this gypsum
water-of-hydration band near 1.94 /jm is linear. This linearity is the
result of the relatively low absorption coefficient of the 1.94 urn band
which makes it less prone to saturation (16). These laboratory studies
demonstrated the potential for semi-quantitative determinations of gypsum
accumulation on limestone and marble buildings and monuments (17).
In order to develop a realistic calibration curve relating spectral
band strength to gypsum accumulation under actual environmental conditions,
spectral measurements were collected on samples designated for chemical
analysis at the Argonne National Labs. These spectral measurements were
made on 12 limestone and 12 marble briquettes which had been exposed at the
four test sites for up to two years. Several measurements were collected
on the under surface as well as on the upper surface of these briquettes.
Mean band strength was calculated by subtracting the average band depth at
1.94 /zm of the non-sulfated upper surface spectra from band percent
reflectance of the lower surface spectra. Samples were shipped to the
Argonne National Labs after spectral characterization for chemical analysis
where a 250 /zm thick layer of the outer surface was "shaved" from each
briquette, powdered, and analyzed by ion chromatography (Youngdahl,
personal communication). Values for sulfate are reported in ug/cm2. Note
that gypsum and sulfate are used interchangeably to relate stone sulfation,
since no other sulfate occurs on a briquette.
Mean band strength was plotted against sulfate (Figures 5a and 5b).
The correlation coefficient and linear regression equation was calculated
for both the marble and limestone data sets using the BMDP statistical
software program. Despite the low frequency of points plotted (only 12
briquettes of each stone type were available for this study), the
correlation is high enough to be acceptable as a calibration curve relating
sulfate concentration to band strength at 1.94 /zm. The regression equation
was used for quantitative determination of the sulfate accumulation from
spectral measurements made at the four test sites in 1985 and 1986. Some
differences between individual briquettes were noted. Sulfate did not
accumulate as heavily on briquettes located on the lower corners of the
racks, compare Figures 6a and 6b.
RESULTS
There was no evidence of gypsum accumulation on the fully exposed
upper surface of the briquettes measured by spectral reflectance at all
four sites during the exposure period. Results of the NIR spectral data
were corroborated by both chemical and mineralogical studies. However,
gypsum accumulated within the "spot" which continued to develop in size and
deepen in color on the protected under-surface of the briquettes (Figures 6
and 7). The presence of gypsum was confirmed by chemical analysis (where
it was analyzed as' sulfate ion by ion chromatography as previously
described) (Youngdahl, personal communication), by X-ray analysis and
scanning electron microscopy (SEM) (Ross and McGee, personal communication)
as well as by the increased band depth at 1.94 /im in the NIR spectral
110
-------
1200 r-
1000
800
600
13
4OO
200
MEAN VALUES OF LIMESTONE
1.94 fim BAND VS SULFATE DATA
N = 12
R = .983
P< .001
MEAN S. D.
X 183.50 139.45
Y 464.O8 335.88
REGRESSION LINE
Y= 3.8484 + 2.508I-X
40 120 200 280
1.94/im BAND STRENGTH
360
Figure 5a. Linear regression relating the 1.94 fim band strength of
limestone plus gypsum to the sulfate concentration. The
correlation coefficient is 0.983.
MEAN VALUES OF MARBLE
1.94 [im BAND VS SULFATE DATA
300 -
250
e 200
i
ui
150
_
D
CO
100
50
J_
N= 12
R= .924
P<.OOI
i I
MEAN S. 0.
X 70.417 41.071
Y 140.50 76.476
REGRESSION LINE
Y= 19.387 -H.7I99«X
J_
J_
_L
10 30 50 70 90 110
1.94 ^m BAND STRENGTH
J
130 150
Figure 5t>. Linear regression relating the 1.94 /zm band strength of marble
plus gypsum to the sulfate concentration. the correlation
coefficient is 0.924.
Ill
-------
Figure 6a.
Figure 6b.
SPECTRA OF UNDER SURFACE OF LIMESTONE BRIQUETTE
Exposed two years at the New Jersey site
100 r
80
60
LJ
O
2 40
o
UJ
K 20
Sulfate, 1156/ig/cm
1.0
1.19
1.38
1.56 1.75
WAVELENGTH
1.94
2.13
2.31
2.50
Average spectra of the under surface of a limestone briquette
located at Chester, NJ., for 2 years. Briquette was emplaced
in center of rack. Note the deep adsorption band centered near
1.94 /un. Sulfate from regression line in figure 5a.
SPECTRA OF UNDER SURFACE OF LIMESTONE BRIQUETTE
Exposed two years at the New Jersey site
100 r
80
LJ
O
-------
SPECTRA OF UNDER SURFACE OF MARBLE BRIQUETTE
Exposed two years at the District of Columbia site
100 r
o
oc
CA
60
UJ
o
40
20
Sultate, 419/ig/cm2
1.0 1.19 1.38 1.56 1.75 1.94
WAVELENGTH (/im)
2.13
2.31
2.50
Figure 7a. Average spectra of the under surface of a marble briquette
located at Newcomb, NY, for 2 years. Note the -adsorption
feature at 1.94 /jm. Sulfate calculated from regression line in
figure 5b.
SPECTRA OF UNDER SURFACE OF MARBLE BRIQUETTE
Exposed two years at the New York site
100 r
1.19
1.38
1.56 1.75
WAVELENGTH
.94
2.13
2.31 2.50
Figure 7b. Average spectra of the under.surface of a marble briquette
located at Washington, DC, for 2 years. Note the deep
adsorption band centered near 1.94/wn. Sulfate calculated from
regression line in figure 5b.
113
-------
measurements. Gypsum was not detected by these methods on the
under-surface area surrounding this spot. As described by SEM analysis,
the gypsum that accumulated in the spot varies from thinly bladed to nearly
blocky crystals sometimes forming a dense mat on the underlying calcite.
Black specs described as being morphologically and compositionally similar
to carbonaceous fly ash were occasionally dispersed on the gypsum.
Particles such as these trapped within the gypsum blades may account for
the tan to brown color associated with the "spot." No morphological
difference was noted between gypsum on the limestone and on the marble.
Also, the SEM data of the upper, exposed limestone briquette surfaces
describes a "frosting" of fine white powder on some of the calcite grains
not seen on unexposed limestone which may be reprecipitated calcite. There
was not abnormal spectral response to this textural alteration in the upper
briquette surfaces.
The sulfate concentration developed within the spot area was
calculated from the average band strength measured at the four test sites
using the regression equations derived from the calibration curves for
limestone and marble (Figures 8a and 8b).
Funds were available for in situ measurements in 1987 at the North
Carolina site only, but laboratory measurements were made on briquettes
from the other sites since they had been shipped to the USGS for
mineralogical analysis.
The increase of gypsum on marble was nearly linear at the test sites
over the three-year period. The heaviest gypsum accumulation on both
marble and limestone occurred at the urban Washington, D.C., site. At the
rural New York site, the small amounts of sulfate measured on the
under-surfaces ranged near the lower limit of detection by NIR spectral
reflectance as determined by laboratory measurements.
The decrease in gypsum accumulation on limestone for 1987 as recorded
by NIR spectral data may be the result of differences in data acquisition.
The 1987 samples were packed and shipped to the USGS and some of the gypsum
on the outer surface could have been lost during this handling, resulting
in lower values than might have been rendered in situ. Also, only one
briquette each of limestone and marble was measured for the 1987 data,
whereas several briquettes were measured at each test site in 1985 and
1986.
The pattern of gypsum accumulation, DC > NJ > NC > NY, complements
some of the air quality and chemical data recorded at each of the test
sites, Tables 1 and 2.
114
-------
AVERAGE SULFATE ACCUMULATION
ON UNDER SURFACE OF LIMESTONE BRIQUETTES
1200 r
1100 -
1000 -
900 -
North Carolina
New Jersey
New York
District of Columbia
1985
1986
1987
Figure 8a. Sulfate accumulation calculated from spectral .reflection
measurements of under surface of limestone briquettes at each
of the four test sites. Sulfate calculated from regression
data in figure 5a.
800
700
N£ 600
u
"£•500
uj-400
£300
w 200
100
0
AVERAGE SULFATE ACCUMULATION
ON UNDER SURFACE OF MARBLE BRIQUETTES
1
2
— North Carolina
New Jersey
New York
District of Columbia
1985
1986
1987
Figure 8b. Sulfate accumulation calculated from spectral reflection
measurements of under surface'of marble briquettes at each of
the four test sites. Sulfate calculated from regression data
in figure 5b.
115
-------
Table 1. Atmospheric Gas Chemistry at the Four NAPAP Monitoring Sites.
Site
RTP, NC
Washington
D.C.
Chester, NJ
Newcomb, NY
Average
Range
Average
Range
Average
Range
Average
Range
S02 ppbv
3
0-12
10
4-22
6
2-13
2
1-5
N02 ppbv
14
8-20
31
22-43
14
9-22
2
0-7
NO ppbv
7
1-18
20
7-87
9
3-20
1
0-2
03 ppbv
25
8-44
19
2-34
32
13-48
30
21-39
Table 2. Average* Rain Chemistry at Four NAPAP' Monitoring Sites.
Measurements are made on wet bucket samples. Values are weighted for the
volume of rain collected for each measurement period. Ion concentrations
are in microequivalents per liter.
RTP, N.C.
Washington, D.C.
Chester, N.J.
Newcomb, N.Y.
Ca++
uequ/L
2.7
18.3
5.3
5.9
N03-
uequ/L
20.1
33.8
28.3
24.1
*
so4=
uequ/L
43.8
86.0
56.1
43.5
H*
uequ/L
42.7
72.4
57.5
47.9
PH
4.37
4.14
4.24
4.32
"Averages are for the period June 1984 through July 1985
The results of spectral reflectance measurements in this study
demonstrated that the largest concentration of gypsum was accumulated on
the lower surfaces of limestone and marble briquettes at the urban site,
Washington, D.C. (Figure 8). The highest ambient levels of SO,, NO,, and NO
were recorded at that meteorological station. Comparative results at the
two suburban sites were mixed. Throughout the three-year study period
(Figures 8a and 8b), slightly more gypsum was measured on the marble lower
surface at RTP, N.C., but more gypsum accumulated on the limestone at the
Chester, N.J., site. Levels of atmospheric gas were only slightly higher
at the New Jersey site than in North Carolina.
Ozone concentration recorded at Newcomb, N.Y. (Table 1), was
relatively higher than that at the other three sites, possibly because
there is little atmospheric S02
destroy ozone.
NO,, and NO to enter into reactions which
116
-------
Rain was collected for chemical measurements in Aerochrome Meters
"wet bucket" collectors which open at the onset of a rain event, and remain
open only during the rain episode so that the chemical composition is not
contaminated by dry deposition (18).
The highest concentration of major ions dissolved in rain water was
measured at Washington, D.C., where the average pH of 4.14 was the lowest
of the four sites (Table 2). Ion concentrations were roughly equivalent at
the three other sites. However, the rainfall with lowest acidity was
recorded at RTP, N.C., not Newcomb, N.Y., where gypsum was consistently
measured in the lowest range. We need more data to understand the relative
significance of S02 and N0x levels, as well as rain pH, in causing the
accumulation of gypsum on building stone surfaces. Either gaseous or
aqueous phase S02 may be oxidized to sulfuric acid through diverse and
complex chemical reactions which may take place within thin liquid films on
surfaces. Carbonaceous material such as soot may act as a catalyst for
such a reaction (19). Concentrations of S02, N0x, ozone, and rain pH are
highly interdependent and involve an understanding of atmospheric chemistry
of extreme complexity which is not the subject of this paper.
Despite the small size of this data set (4 test sites and 3 years of
briquette exposure), regression coefficients were derived to determine the
correlation between the various environmental parameters and the sulfate
accumulation (Tables 3 and 4).
Table 3. Correlation coefficients (r) between gypsum on under-surface of
limestone briquettes and pollutant concentration of the atmosphere and rain
at the four test sites. Gypsum concentrations are determined from spectral
reflectance measurements (N=4).
1985 1986 1987
S°7
N0?
NO
°?
pH*
H**
vs.
vs.
vs.
vs.
vs.
vs.
Gypsum
Gypsum
Gypsum
Gypsum
Gypsum
Gypsum
0
0
0
0
0
0
.942
.999
.995
.789
.794
.827
0
0
0
0
0
0
.875
.931
.919
.587
.690
.712
0
0
0
0
0
0
.943
.967
.969
.646
.800
.817
* Wet bucket measurements.
The correlation between gypsum accumulation as measured by NIR
spectral reflectance and NOX concentrations over the three-year period was
remarkably high for both limestone and marble briquettes. The correlation
coefficient (r) dropped below 0.800 for the 1987 series of marble
measurements vs. SO. concentration, but otherwise there was good
correlation (Table 4).
117
-------
Table 4. Correlation coefficients (r) between gypsum on under-surface of
marble briquettes and pollutant concentration of the atmosphere and rain at
the four test sites. Gypsum concentrations are determined from spectral
reflectance measurements (N=4).
1985 1986 1987
SO,
ml
NO
°3
pH*
H**
vs.
vs.
vs.
vs.
vs.
vs.
Gypsum
Gypsum
Gypsum
Gypsum
Gypsum
Gypsum
0
0
.0
0
0
0
.821
.972
.938
.891
.618
.794
0
0
0
0
0
0
.912
.920
.887
.976
.583
.642
0.705
0.878
0.827
0.659
0.447
0.485
* Wet Bucket Measurements
Johansson et al. (20) reported results of experiments which showed
that the presence of N02 significantly influenced the rate of S02 uptake on
the surface of marble. The NO, catalyzes the otherwise slow reaction step,
whereby the intermediate calcium sulfite is oxidized to gypsum. Further
experiments are needed to understand the relationship of stone
deterioration to N02 at concentrations found in urban areas. Results of
studies of tombstone deterioration as reported by Husar et a/. (21)
concluded that there was a linear relationship between stone erosion rates
and S02 concentrations.
Correlation between pH alone and gypsum was not good in this data set
although the site of the lowest pH was also the site of the greatest gypsum
accumulation. An observation to be drawn from this attempt to correlate
gypsum accumulation and environmental parameters is that there is a
complexity and interdependence of all these factors.
CONCLUSIONS
Diffuse spectral reflectance is a useful technique for measuring the
mineralogical composition of rocks as it measures the interaction of
electromagnetic radiation with materials. The patterns of absorption bands
recorded when diffuse light is reflected by a sample are characteristic of
minerals. When minerals are mixed, as in the combination of gypsum and
calcite, changes in the depth of certain diagnostic bands such as the water
of hydration band near 1.94 /zm are proportional to the amount of gypsum
present. We have shown that it is possible to quantitatively measure the
accumulation of gypsum on calcite by spectral reflectance measurements in
the NIR.
118
-------
Gypsum does accumulate on those surfaces of limestone and marble
which are protected from direct rainfall. It also forms on the boldly
exposed surfaces of carbonate building stone but is washed away because of
its high solubility. It has been observed in Europe as well as in the
United States (Doe, personal communication) that encrustations, dark in
color and often black, form on parts of buildings or statuary that may be
protected from direct rainfall. Analysis of NIR spectral measurements made
on parts of a balustrade in the rear of the Pan American Building in
Washington, D.C., which had these encrustations, showed that the crusted
surface was a combination of calcium carbonate and gypsum. The
concentration of gypsum calculated by NIR data was 520 pg/cm.. The black
color was probably soot trapped within the gypsum blades as we have seen
occur in the test briquettes.
It has not been determined yet that gypsum accumulations which
develop on the protected surface of a briquette is complementary to weight
loss and surface recession, but preliminary results suggest that this is
true (22). Calcium ion, nitrate, and sulfate are highest in the rain
run-off of the limestone and marble slabs at Washington, D.C. At all four
sites, the average rate of surface recession for skyward surfaces of
briquettes is near 15 /im/yr. Weight loss ranges from 0.4 gm/yr for marble
and 1.8 gm/yr for limestone. Quantitative measurement of gypsum then would
be a good indicator of dissolution and surface recession of carbonate stone
buildings and monuments. The most significant damage may not be due to the
erosion of calcium carbonate but rather the destruction of stone through
precipitation and recrystallization of salts. The molar volume of gypsum
molecules is greater than that of carbonate molecules which would encourage
stress and weaken mineral grains, resulting in splitting, cracking, and
eventually spall ing of outer layers. Most important, the spectral
reflectance technique is fast, nondestructive, and measurements are made in
situ. Data storage and retrieval are also simple.
What are the upper limits of gypsum detection
reflectance? Will gypsum continue to accumulate on
briquette under surfaces until the calcite is completely
a spectrum is recorded for pure gypsum?
removed by humid conditions, such as fog
questions remain, but could be answered
briquettes at the four test sites.
Or will some
, heavy rain,
by continued
by NIR spectral
the spot of the
obscured, so that
.of the gypsum be
or snow? These
NIR study of the
REFERENCES
1. Jaynes, S.M., and R.U. Cooke. Stone Weathering in Southeast England.
Atmospheric Environment, 21:1601-1622, 1987.
2. Reddy, M.M., S.I. Sherwood, and B.R. Doe. Limestone and Marble
Dissolution by Acid Rain: An On-site Weathering Experiment. In
Materials Degradation caused by Acid Rain, Robert Baboian, Ed., ACS
Symposium Series 318, American Chemical Society, Washington, D.C.,
1986, pp. 226-238.
119
-------
3. Ross, M., and L. Knab. Selection, Procurement, and Description of
Salem Limestone Samples Used to Study Effects of Acid Rain. National
Bureau of Standards Report NBSIR 84-2905, 1984, 17 pp.
4. Ross, M. Description, Selection, and Procurement of Shelburne Marble
Samples Used to Study Effects of Acid Rain. U.S. Geological Survey
Open-fil Report 85-594, 1985, 15 pp.
5. McGee, E.S. Mineralogical Characterization of the Shelburne Marble:
a Vermont Marble Test Stone Used to Study the Effects of Acid Rain.
U.S. Geological Survey Open-File Report 87-447, 1987, 19 pp.
6. Sherwood, S.I., and B.R. Doe. Acid Rain Stone Test Sites. EOS,
Transactions, American Geophysical Union, 65:1210, 1984.
7. Flinn, D.R., S.D. Cramer, and J.P. Carter. Field Exposure Study for
Determining the Effects of Acid Deposition on the Corrosion and
Deterioration of Materials - Description of Program and Preliminary
Results. Durability of Building Materials, 3:147-175, 1985.
8. Pearce, F. Acid Eats Into Britain's Stone Heritage. New Scientist,
26:26-27, 1985.
9. del Monte, M., and 0. Vittori. Air Pollution and Stone Decay: The
Case of Venice. Endeavor, New Series, 9:117-122, 1985.
10. Frediani, P., G. Menchi, and U. Matteoli. Gypsum on Works of Art in
Marble: Determination by Infrared Spectrpscopy, Proc. 3rd Intr.
Congress on the deterioration and preservation of stones, 1979. pp.
195-203.
11. Bultin, R.N., R.U. Cooke, S.M. Janes, and A.S. Sharpe. Research on
Limestone Decay in the United Kingdom. In: Proceedings of 5th
International Congress on Deterioration and Conservation of Stone, G.
Felix, Ed., Presses Polytechniques Ramandes, Lausanne, Switzerland,
1985. p. 537.
12. Hunt, G.R., and J.W. Salisbury. Visible and Near-infrared Spectra of
Minerals and Rocks: II. Carbonates, Modern Geology, 2:195-205,
1971.
13. Clark, R.N., and T.L. Roush. Reflectance Spectrpscopy: Quantitative
Analysis Techniques for Remote Sensing Applications, Jour, of
Geophysical Res., 89, no. B7, pp. 6329-6340, 1984.
14. Gaffey, S.J. Spectral Reflectance of Carbonate Minerals in the
Visible and Near Infrared (0.35-2.55 micros): Calcite, Aragonite and
Dolomite. Am. Mineral, 71:151-162, 1986.
120
-------
15. Kingston, M.J., and C.M. Ager. A Spectral Reflectance Method to
Measure Acid Deposition Effects on Building Stone, in: Proceedings
of the American Society of Photogramrnetry Annual Meeting, March
11-16, Washington, D.C., 1985. p. 871.
16. Eastes, J.W., and J.W. Salisbury. Spectral Properties of Sulfated
Limestone and Marble: Implications for In Situ, Assessment of
Atmospheric Pollution Damage to Carbonate Rock Building Materials.
Applied Spectroscopy, 40:954-959, 1986.
17. Kingston, M.J., and L.C. Rowan. Application of Near-infrared
Spectral Reflectance Measurements for Detection of Acid Damage to
Building Stones. Proceedings Symposium of the 8th ICOMOS General
Assembly, Oct. 7-15, 1987, Washington, D.C. 1987.
18. Reddy, M.M. Acid-rain Damage to Carbonate Stone: A Preliminary
Quantitative Assessment Based on the Aqueous Geochemistry of Rainfall
Run-off. U.S. Geological Survey Water-Resources Investigations
Report 87-4016, 1987. 27 pp.
19. Brodzinsky, R., S.G. Chang, S.S. Markowitz, and T. Novakov.
Kinetics and Mechanism for the Catalytic Oxidation of Sulfur Dioxide
on Carbon in Aqueous Suspensions. Journal of Physical Chemistry,
84:3354-3358, 1980.
20. Johansson, L.G., 0. Lindquist, and R.E. Mangio. Corrosion of
Calcareous Stones in Humid Air Containing S02 and N02. in: Air
Pollution and Conservation: Safeguarding Our Architectural Heritage,
Jan Rosvall, Ed., Swedish Institute of Classical Studies in Rome,
Symposium, Oct. 14-17, 1986.
21. Husar, R.B., D.E. Patterson, and N.S. Baer. 1985. Deterioration of
Marble - A Retrospective Analysis of Tombstone Measurements in the
New York City Area. U.S. Environmental Protection Agency report DW
14930338-01-1, 23 p.
22. Reddy, M.M., and C.A. Youngdahl. Acid Rain and Weathering Damage to
Carbonate Building Stone: Results of Material Loss Measurements.
Corrosion 87, paper no. 415, March 9-13, 1987, meeting, San
Francisco, CA, 1987. 7pp.
121
-------
REMOTE DETECTION OF DISSOLVED ORGANIC MATTER (DOM), ALUMINUM,
AND HYDROGEN ION USING LASER-INDUCED FLUORESCENCE (LIF)
by W.D. Philpot and A. Vodacek, Cornell Laboratory for Environmental,
Applications of Remote Sensing (CLEARS), Hoi lister Hall, Cornell
University, Ithaca, New York 14853
ACKNOWLEDGEMENTS
This work was supported by the Environmental Protection Agency
through cooperative agreement #CR-813039-01. Graduate student support was
also provided for some sampling activities by the Edna Bailey Sussman Fund
through Cornell University. There are also several individuals whose
assistance was important to the successful completion of the project. In
particular, the authors would like to thank the EPA Project Officer, Dr.
Michael Bristow for his assistance and guidance throughout the project.
The EPA/Lockheed personnel were extremely cooperative and helpful in
providing samples from the Michigan and Wisconsin lakes. We would also
like to thank Dr. Ellis Loew who assisted in the calibration of the
fluorescence detection system by providing both equipment and expertise.
ABSTRACT
Fluorescence of dissolved organic matter (DOM) in lake water is
affected by many factors, among them the concentration of trace metals, pH,
temperature and the composition of the DOM itself. An empirical study was
conducted seeking a specific relationship between the intensity and
spectral distribution of DOM fluorescence and the concentration of aluminum
and hydrogen ion in lake water. Spectra for water samples from forty-nine
lakes in northern Michigan and northern Wisconsin were collected and
analyzed with respect to their water chemistry.
Trends that had been apparent in earlier laboratory tests were also
apparent in the lake water data, but simple, general relationships between
fluorescence spectra and water chemistry were not apparent in the overall
results. The data are consistent with the hypothesis that the changes in
fluorescence are due mainly to reactions of the humic acid portion of the
DOM as opposed to fulvic acid. Hence, reaction of DOM with various cations
(Al*3, Fe*3, H+, Ca+2) tends to both reduce the total fluorescence and to
shift the fluorescence to shorter wavelengths. The effects are strongly
related to the specific cation involved.
122
-------
By itself, laser induced fluorescence of dissolved organic matter
will not be useful as an indicator of hydrogen ion concentration (pH);
however, total fluorescence intensity is generally related to dissolved
organic carbon (DOC) concentration (presuming a constant composition of
DOM) and the changes in fluorescence spectra are directly related to
changes in water chemistry (the strong correlation with organic aluminum
per unit DOC (AiyDOC) is a good example of this). DOM fluorescence should
be particularly useful as a survey tool in conjunction with limited water
chemistry or as a means of monitoring change in a lake whose initial
chemistry and fluorescence properties have been defined.
INTRODUCTION
Several substances present in natural waters fluoresce, and their
fluorescence spectra are often distinctive. Since the intensity of
fluorescence is usually proportional to the concentration of the
fluorescing material, it is possible that laser induced fluorescence (LIF)
could be used to detect and measure the amount of the fluorescing
substances. LIF would thus be appropriate for airborne monitoring for
chlorophyll, oil, dissolved organic matter (DOM), or any other fluorescing
materials.
Non-fluorescing substances may also be detectable as a result of
their interaction with fluorescent material. For instance, Cornell
researchers (1) established a link between the fluorescence of dissolved
organic matter (DOM) and the concentration of aluminum in lake water
samples. Aluminum was detectable because it reduced (quenched) the
intensity and altered the spectral character of the fluoresced radiation.
This suggested the feasibility of remotely monitoring aluminum, a trace
metal which is toxic to fish at high concentrations and which has been
identified as an important parameter of acidified lakes.
Fluorescence of DOM will also be altered by other trace metals (iron,
copper), pH and temperature (2). Similar interactions are likely to occur
between any fluorescing substance (e.g., oil, chlorophyll) and other water
quality parameters. Potentially, any material that fluoresces or
measurably alters the fluorescence of another substance will be susceptible
to detection by LIF. The appropriate technology is available, although in
many cases research is needed to: (1) better understand the absorption and
fluorescence characteristics of specific substances in water; (2)
characterize the alteration of the fluorescence by other water quality
parameters; and (3) define the radiative transfer problem well enough to
develop an effective detection and monitoring system.
The overall goal of the present study was to explore the relationship
of the quenching and spectral shift of DOM fluorescence with the
concentration of aluminum and hydrogen ion in lake water. If a clear
relationship could be demonstrated, then laser-induced fluorescence would
be useful as a tool for remote monitoring of acidified lakes and/or trace
metal concentrations. Quenching effects have been observed in both
123
-------
synthetic samples and in selected natural water samples taken from a wide
range of lake types in the Adirondacks (3,4). Initial results indicated
that the quenching effects were measurable and that the degree of quenching
correlated with the concentration of aluminum (3). Spectral alteration of
the fluorescence (associated with the quenching) also correlated with pH
and DOM concentration of the water samples.
The particular purpose of this project was to determine (1) if
fluorescence quenching due to the presence of trace metals would be
spectrally distinguishable from changes in fluorescence due to changes in
other parameters, especially DOM concentration or pH, and (2) if LIF is a
technique that is applicable to lakes over a broader geographic range than
had yet been investigated (i.e., outside the Adirondacks.) The proposed
fluorescence analyses were to be conducted under laboratory conditions;
however, since the intent was to use LIF for remote sensing, the first
phase of the project was to design and assemble a laser fluorosensing
system which would be usable both in the laboratory and in the field.
After this was done, fluorescence spectra were collected for water samples
taken from forty-nine lakes in northern Michigan and northern Wisconsin
(Figure 1). Results of the spectral and chemical analyses were used to
examine the response of the spectral fluorescence to changes in water
chemistry. Finally, the results from the Michigan/Wisconsin lake study
were compared to earlier results for Adirondack lakes.
PROCEDURES
SAMPLING AND SHIPPING PROCEDURES
Lake water samples were collected by an EPA NAPAP survey team 1.5
meters below the lake surface using a Van Dorn type sampler. An aliquot of
the sample was transferred to a Nalgene brand 125 ml high density, linear
polyethylene, amber bottle. Bottles were labelled with the National
Surface Water Survey (NSWS) identifier, the date and time of the sample
collection, and the sample type (5). Upon arrival at Cornell the samples
were analyzed for fluorescence as soon as possible. If the analysis could
be done the same day the samples arrived the samples were allowed to warm
to room temperature. When the fluorescence measurements could not be made
the same day, samples were stored in the dark at 4°C. Several of the
samples were delayed in shipment. In a few instances the packing ice had
melted and the samples were warm. The warming was recorded, but the
samples were handled exactly the same as any other samples.
Chemical analysis of the lake samples was provided by the
Environmental Protection Agency (EPA) (6). The measurements most important
to the fluorescence analysis were pH, DOC, total reactive Al, organic
reactive Al, total Al, total Fe, and total Ca. The data used here were
taken from the "preliminary" data set. Only small changes were expected in
the final, verified data.
124
-------
Figure 1. Location of lakes in northern Michigan and Wisconsin from which
water samples were collected for this study. These lakes were
sampled as part of the National Surface Water Survey (NSWS),
Phase II.
125
-------
pH MEASUREMENTS
Sample pH was acquired directly from the sample bottle, shortly after
completion of the fluorescence procedure. The pH was measured with a
Corning 150 pH/Ion meter with automatic temperature compensation and an
Orion-Ross' 8104 combination pH electrode.
FLUORESCENCE MEASUREMENTS
The laser-induced fluorescence apparatus acquired for this study was
chosen for its flexibility and ease of operation. (The same equipment used
in this laboratory study can be used for later field studies.) The
laboratory configuration is illustrated in Figure 2. A pulsed
Photochemical Research Associates (PRA) model LN 103 nitrogen gas (N2)
laser was used. The N2 laser wavelength is 337 nm with 400 KU peak power,
70 microjoules pulse energy, and 300 ps pulse width. Pulses from the
Ultra-violet (UV) laser excite fluorescence in a water sample held in a 1x1
cm quartz glass cuvette. Quartz glass is required for effective
transmission of UV light. Fluorescence and scattered light from the water
sample pass first through a cutoff filter and then through a cylindrical
lens which focuses the light on the entrance slit of the spectrograph. The
cutoff filter is a liquid filter held in a quartz cuvette and inserted
between the sample cuvette and the spectrograph entrance slit. The filter
is a 2 g/L solution of 2,7-dimethyl-3,6-diazacyclohepta-l,6-diene
perchlorate (DDDP), a solution which provides a sharp cutoff for light
below 350 nm (7).
SPECTROGRAPH
1
DETECTOR
gate
pulse
PULSER
transfer
lens i—
DDDP i
filter '
quartz
cuvette
trigger
pulse
CONTROLLER
frequency
generator
NITROGEN (N2)
LASER
COMPUTER
Figure 2. Diagram of the laboratory laser system used in the
laser-induced fluorescence study. The design of the system and
selection of components was intended to allow the system to be
used in the field as well as in the laboratory.
126
-------
The spectrograph is an Instruments SA model 320, with interchangeable
gratings, a holographically ruled 1200 groove/mm grating used for high
resolution and a 148 groove/mm grating used for a wide spectral window.
The 148 groove/mm grating was used for all experiments reported here.
Spectrally dispersed light from the spectrograph is detected with a
Princeton Instruments (PI) multi-channel plate, intensified, diode array
with 700 active elements. A PI controller and pulser provide the firing
pulse for the laser as well as the start and stop pulses for the detector,
all under the control of PI, menu-driven software. A PC'S LIMITED AT is
used as the host microcomputer for the PI system.
Each spectrum was corrected for instrument sensitivity and
configuration. Two types of calibration were necessary, a wavelength
calibration to determine the spectral position of each element in the
photodiode array, and an intensity calibration to correct for the varying
response of the spectrograph and the photodiode array at different
wavelengths. The wavelength calibration was obtained by determining the
diode position for seven known mercury lines in a standard fluorescent lamp
spectrum. The spectral intensity calibration was corrected by input of
correction factors derived from a quartz-tungsten standard source with a
known spectral intensity.
Collection of the fluorescence data involved two facets of sample
manipulation, sample preparation and transfer of the sample to the cuvette.
The procedures used were designed to minimize alteration of the lake sample
fluorescence characteristics between the acquisition of the sample and
collection of the fluorescence spectrum. Sample preparation consisted
simply of warming the bottles to room temperature while loosely capped,
thus the samples were open to the atmosphere prior to being transferred to
the cuvette used in the fluorescence procedure.
With the spectrograph aligned to keep the laser line off the
intensifier, the fluorescent light spectrum used for the wavelength
calibration was collected using the detector's continuous wave (CW) mode.
A wavelength calibration spectrum was collected for each day that lake
fluorescence was obtained. Switching to the gate mode, the timing was
adjusted to insure detection of the entire laser-induced fluorescence
pulse: the gate timing was adjusted to maximize the intensity of Raman
pulse relative to the fluorescence, and the integration time was adjusted
to use a high percent of the photodiode array dynamic range. A background
spectrum was then collected, followed by two fluorescence spectra. The
background spectrum was then subtracted from the fluorescence spectra.
As a last step, the spectra were normalized by the water Raman signal
(8). A typical, spectrally corrected, Raman normalized fluorescence
spectrum of DOM in lake water is illustrated in Figure 3. The laser
wavelength, shown at 337 nm, is normally filtered out by the DDDP filter.
Raman scattering by water occurs centered at 381 nm; the dashed line
represents the best fit straight line to the fluorescence curve. The total
Raman signal is taken to be the area under the curve and above the dashed
line. A linear interpolation based on a least squares fit to points on
either side of the Raman peak (361-370 nm and 390-399 nm) was used in
127
-------
calculating the dashed line. Raman normalization of the entire spectrum
consists of dividing the intensity at each wavelength by the total Raman
signal.
Typical Fluorescence Spectrum
255
245
351
457 562
Nanometers
665
767
Figure 3. A typical, spectrally corrected, Raman normalized fluorescence
spectrum of DOM in lake water.
There is little variation in spectral detail among the normalized
fluorescence spectra. The primary differences -- a change in overall
intensity, a shift of the gross distribution of intensity and a change in
the relative intensity on either side of the fluorescence peak -- can be
effectively represented by only a few parameters. Those chosen for this
study were: 1) the total intensity between 370 nm and 590 nm, Ftot; 2)
fluorescence at the Raman peak, F3.0, computed by integrating over the
fluorescence curve between 370 nm ana 390 nm; 3) the mean wavelength of the
fluorescence over the range 370-590 nm, Fm; 4) the standard deviation of
the fluorescence over the range 370-590 nm (essentially a measure of the
bandwidth of the fluorescence); and 5) the ratio of fluorescence FR =
F /F380. These parameters are generally reasonable, but were chosen
128
-------
heuristically for this study and are probably not entirely independent.
Indeed, it is not clear which (if any) of the selected parameters best
characterize the spectrum. A summary of the parameterized fluorescence
data is presented in Table 1.
RESULTS AND DISCUSSION
In preliminary work proceeding this study, Vodacek and Philpot (1)
established that, when aluminum was added to a water sample containing DOM,
the fluorescence was quenched and the spectrum shifted slightly to the
blue. Similar changes in fluorescence occurred in response to changes in
pH, temperature and the concentration of other trace metals (2). These
quenching effects were then observed both in controlled laboratory
experiments with prepared samples and in fluorescence spectra of over 50
unaltered lake water samples.
Initial results indicated that the quenching effects would be
measurable. However, it was also apparent that chemical parameters
affecting DOM fluorescence were interdependent, and that distinguishing
among their effects could be difficult. Thus, the research question was
whether or not fluorescence data alone would be sufficient to serve as an
indicator of aluminum and/or hydrogen ion in lake water. As this was an
empirical study, the approach was to look for a statistically, meaningful
predictor for aluminum and pH using the several parameters derived from the
fluorescence spectra (Section III.C.3).
pH PREDICTION
Since, from the preliminary studies, it appeared that fluorescence
from water samples might serve as an indicator of lake pH, the first step
in analysis of the data was to attempt to predict pH based on one or more
of the fluorescence parameters. One of the most distinct relationships in
the earlier work was the correlation of pH with total fluorescence. A plot
of total fluorescence versus pH for aliquots of a prepared water sample
containing humic material for which the pH had been adjusted in the
laboratory (2) is shown in Figure 4. A similar figure for samples from the
Michigan/Wisconsin lakes is shown in Figure 5. In the earlier laboratory
experiment (Figure 4), samples with pH > 5.0 showed little correlation with
fluorescence. However, there appeared to be a distinct cutoff at about 530
units of fluorescence intensity which corresponded roughly to a pH of 5.0.
Significantly, there is a clear decrease in fluorescence intensity for pH <
5.0. The correlation of total intensity with pH for pH < 5.0 indicates
that, at high enough concentrations, the hydrogen Ion Itself will quench
fluorescence.
It is more difficult to interpret the Michigan/Wisconsin data (Figure
5); the scatter in the data is too great for a clear distinction between
acidic and non-acidic lakes. Hydrogen ion quenching is probably still
occurring, but its effect is masked by other parameters which also affect
total fluorescence or spectral variation in natural water samples.
129
-------
LAKE
ID
LAKE WATER CHEMISTRY
pH
Total Organic
Reactive Reactive Total Total
DOC Aluminum Aluminum Aluminum Calcium
(ng/l)
AVERAGED FLUORESCENCE DATA*
Mean Standard
wavelength Deviation
(nm) (nm)
f380
rtot
1016R
1022R
1035R
1038R
1039R
1040R
1041R
1042R
1047R
104801
104802
104803
1048R
1052R
1061R
1064R
1066R
2004R
2007R
2024R
2038R
2044R
2049R
2055R
2061D1
206102
206103
2061R
2074R
2075R
2078R
207901
207902
2079F
2079R
2082R
2090ft
2098R
2100R
3007R
3008R
300901
300902
3009D3
3009R
3012R
3013R
3020R
3023R
3027R
3028R
3030R
3031R
3034R
3037R
3051R
30SSR
3056R
3057R
30S8R
3071R
55085
55103
55111
55141
55171
55211
55223
55242
55271
6.23
.80
.93
.39
.96
.85
.91
5.19
4.69
4.38
4.39
4.37
4.38
4.74
4.96
4.65
4.56
6.39
51
93
34
09
94
66
54
54
5.56
5.53
6.17
93
75
12
09
06
98
72
24
81
90
6.29
6.46
95
97
85
92
81
87
82
43
8.06
8.14
7.99
8.44
4.83
7.33
7.15
6.74
5.77
6.78
06
OS
04
03
17
05
06
5.01
5.04
9.3
4.0
7.8
2.0
2.4
3.2
2.4
3.4
0.9
0.3
0.3
0.3
0.4
3.4
2.9
1.7
1.2
7.7
2.9
S.1
7.3
3.0
4.0
3.6
21.8
21.1
21.3
21.8
7.2
4.5
2.2
4.4
4.1
3.8
3.7
3.8
3.4
6.9
6.2
4.7
6.3
4.6
4.5
5.4
4.6
12.2
6.2
11.5
9.1
4.2
14.8
3.2
7.9
4.6
3.7
2.3
1.9
6.6
5.6
4.0
5.6
3.2
3.1
3.4
3.2
3.1
3.0
3.2
3.5
3.2
39.2
20.4
37.6
30.8
64.0
57.5
38.7
30.8
22
204
204
206
206
19
19
24
77
16.0
27.6
43.6
14.8
17.3
31.7
63.3
97.3
92.8
92.8
95
16
15
31
14
14
16
12
27
34
29
84
50
.1
.0
.9
.7
.5
.5
.7
.4
.5
.4
.3
.0
.6
17.1
17.6
26.4
66.1
56.0
43.0
30.5
26.1
40.9
25.7
47.7
22.1
33.2
25.6
12.6
16.7
33.2
18.4
27.8
132.6
130.
128.
135.
132.
135.
133.5
134.0
129.5
19.5
9.0
27.3
8.2
24.8
32.6
32.6
17.1
15.5
13.3
13.3
13.3
15.8
13.3
23.9
13.
18.
15.
27.9
37.4
16.6
21.4
19.5
24.4
71.0
76.4
73.7
76.4
15.9
20.9
22.7
17.2
17.2
19.8
17.2
22
25
29
44
18
19.1
21.
16.
14.0
16.6
34.3
34.2
40.8
25.8
17
36
18
24
20
13.8
14.2
13.8
17.2
22.2
15.3
18.5
41.4
45.0
35.8
30.5
33.7
41.2
37.1
44.6
36.4
92.0
37.4
105.9
NO
NO
36.3
NO
80.8
NO
NO
NO
ND
NO
26.8
40.
53.
49.0
25
42.6
21.8
NO
16.1
18.8
18.1
555.0
533.9
562.0
555.0
41.4
ND
ND
18.9
ND
ND
ND
NO
34.0
14.5
78.2
ND
NO
NO
ND
86.6
ND
251.8
114.9
889.4
130.5
21.4
1177.2
16.3
ND
ND
ND
ND
ND
140.8
980.7
17.8
NO
1840.0
843.6
407.5
ND
625.7
843.0
599.0
616.6
1198.0
1200.0
1249.0
1213.0
1174.0
562.8
443.2
345.6
695.
1 2443.0
418.5
1615.0
1864.0
444.7
598.7
ND
2575.0
2541.0
2625.0
2575.0
1295.0
626.0
592.9
1077.0
1058.0
1146.0
1045.0
936.1
665.4
2257.0
1101.0
2376.0
2728.0
12440.0
16840.0
14670.0
15790.0
3766.0
753.4
2043.0
12710.0
17570.0
9212.0
867.3
37380.0
21630.0
22450.0
501.7
13770.0
8573.0
3072.0
673.8
3705.0
488.32
484.56
488.79
485.14
480.15
485.52
485.19
485.25
480.70
481.11
478.82
470.29
478.99
483.93
485.50
483.61
479.26
488.20
488.50
490.06
489.59
486.62
485.71
487.87
492.74
492.78
492.76
492.70
486.86
485.13
484.28
488.67
488.75
489.03
489.33
487.78
487.95
488.42
484.99
489.72
488.42
488.57
488.23
488.50
488.96
490.56
487.40
491.02
490.75
485.15
492.49
485.03
487.92
486.87
487.69
484.65
488.57
489.29
487.00
486.47
485.79
480.53
480.58
480.85
483.13
481.02
482.13
481.11
481.06
481.08
51.76
53.61
50.88
53.78
54.02
52.38
54.11
52.52
57.46
62.18
62.58
56.73
62.03
52.95
52.31
53.28
54.69
52.96
51.54
51.49
52.36
52.53
52.62
53.54
50.14
49. 79
50.01
49.85
53.06
52.62
54.00
53.06
52.85
52.99
53.26
52.51
52.61
51.58
51.69
51.93
52.21
52.09
52.08
52.08
52.58
50.90
51.89
50.70
52.45
54.84
50.92
53.67
53.02
52.89
53.82
52.92
53.40
52.93
51.58
53.83
54.35
52.52
53.16
52.98
52.79
53.03
53.52
52.90
52.96
53.01
10.55
1.54
5.47
0.91
1.42
2.84
1.02
1.91
0.30
0.18
0.14
0.18
0.18
1.49
1.17
0.70
0.82
9.88
3.01
1.49
1.83
2.65
52.76
61.47
60.02
64.50
3.47
1.94
0.56
1.92
2.04
1.92
1.89
1.82
1.95
3.02
9.46
2.28
3.11
3.66
3.68
3.50
2.84
10.34
5.06
8.16
6.65
0.97
24.80
1.46
4.52
2.06
1.35
1.38
1.05
2.53
4.92
1.84
0.73
5.39
6.34
75
97
03
77
49
71
5.69
456.68
44.80
251.26
26.15
34.00
98.41
28.84
62.02
5.36
2.21
1.58
2.05
2.20
44.07
40.64
19.64
16.84
123.94
62.97
475.92
121.33
54.32
60.98
85.20
3504.60
4471.18
4063.51
4817.52
122.87
64.29
15.26
72.05
78.13
73.76
70.01
67.89
69.45
129.13
351.71
102.89
120.31
147.98
145.83
138.53
108.71
555.82
199.60
431.33
292.53
25.64
1393.24
44.99
166.75
69.77
44.89
42.22
38.07
100.80
204.29
57.78
20.95
157.03
176.49
132.42
161.15
145.57
134.22
127.86
133.18
157.27
- Two or more fluorescence spectra were averaged for each water sample.
NO - The concentration was below the detection limit of the measurement technique.
Table 1. Summary of the chemical analyses (unverified data sets)
fluorescence data for the Michigan/Wisconsin lake samples.
43.29
29.02
45.97
28.63
23.93
34.65
28.33
32.55
17.84
12.41
11.03
11.40
11.97
29.60
34.59
28.14
20.64
37.12
40.59
48.17
40.34
36.40
33.30
32.20
66.43
72.73
67.70
74.69
35.45
33.13
27.47
37.49
38.23
38.45
37.05
37.24
35.57
42.80
37.16
45.20
38.67
40.45
39.59
39.63
38.26
53.76
39.41
52.88
44.02
26.55
56.18
30.85
36.86
33.79
33.38
30.64
36.26
39.77
41.55
31.44
28.75
29.11
27.83
27.91
32.44
28.92
28.15
28.45
28.30
27.65
and
130
-------
-
£ 650-
c
0)
j:
luorescence
01 O
Ol O
0 0
, 1 : , , , !
li-
CU
to 500-
0)
oc
15
° 450-
4
Figure 4
HS Sample
o Lower pH with Nitric Acid
+ Raise pH with Sodium Hydroxide
x Lower pH with Nitric Acid
„
• • • • L- ?"****
» • »*• »
0 w * +
• "
0 + « + +
o
e
o o
o
00 »
a +
*•
0*
V
| — i — i — r™ r y f i i i | i i i i | i i i " | • • • • | • • • •
0 4.5 5.0 5.5 6.0 6.5 7.0
pH
Variation of the total fluorescence as the pH of water of a
pond water sample was adjusted under laboratory conditions (2).
Inclusion of other fluorescence parameters produced no significant
improvement in correlations with pH. The best overall predictor found in
this study for pH was obtained with a multiple linear regression, using as
independent variables the mean fluorescence wavelength, fluorescence ratio
and standard deviation to predict pH. The coefficient for the standard
deviation was not significant at the 95% confidence level (Student's
T-test) and it was dropped from the regression. Using only the mean
wavelength and fluorescence ratio the best correlation with pH had an
R2 = 0.35 for predicted versus actual pH; standard error of estimate • 0.92
pH units (Figure 6).
With no clear, general relationship between fluorescence and pH, it
was necessary to consider other interactions. In addition to the pH
effects, environmental factors such as temperature and light attenuation as
well as other chemical parameters such as Al, Fe, and Ca will affect
fluorescence data. Compensation for these environmental factors is
relatively straightforward and can probably be accomplished using remote
data (2). However, the chemical parameters affecting DOM fluorescence can
be interdependent, and separating their effects will be more difficult.
Effects of the chemical parameters are discussed in the following sections.
131
-------
E
c
o
O)
CO
1
o
CO
>
*^
'(A
C
0)
+^
_c
0)
0
c
0)
luoresc
LL
12 -
11 -
10 -
9 -
8 -
7 -
6 -
5 -
4 -
3 -
2 -
1 -
-
n
D
n
D
D
D
D
D
n n n Q
rP Q
•^
nOD
n D n
D° ° o
D n
O dP CD Cbtfi
n n n n n ^ ^ n
° on * o
G n
n D
I "n — • 1 1 i i i '
45678
pH
Figure 5. Variation of the total fluorescence with the pH of water for
the Michigan/Wisconsin lake samples.
DEPENDENCE ON DOM CONCENTRATION AND COMPOSITION
One obvious difficulty with using DOM fluorescence as an indicator of
pH is that fluorescence is dependent upon both the concentration and
composition of DOM. Correlation of fluorescence intensity with the DOM
concentration is high enough that it has been used effectively as a measure
of DOM concentration (9). Furthermore, several researchers have noted that
the fluorescence is somewhat dependent on the composition of DOM (9,10,11).
Finally, the spectral distribution of fluorescence may be altered by
concentrational quenching and differential absorption by the DOM itself
(2).
132
-------
pH
Figure 6. Results of a multiple linear regression to predict pH. The
only significant variates in the prediction were the mean
fluorescence wavelength and the fluorescence ratio.
If the mechanism by which pH quenches fluorescence differs from that
governing concentrational quenching or compositional effects of DOM, then
it may be possible to discriminate between the two by first removing any
dependence on *OM concentration from the data; any pH dependence should
then be more obvious in the residual data structure. Given a series of
water samples for which the only difference was the concentration of DOM,
(i.e. the pH, trace metal concentration and composition of DOM are all
constant), one might expect a plot of the mean wavelength and the total
fluorescence to be similar to that in Figure 7 (2). Figure 8 is a plot of
fluorescence intensity versus mean wavelength for a subset of the
Michigan/Wisconsin lakes, excluding two lakes for which the fluorescence
intensity was exceptionally high. Those lakes were excluded in order to
better display the spread in the majority of the data. The solid line in
Figure 8 represents the loci of fluorescence parameters for samples for
which only concentration is changing. The curve is merely representative
133
-------
of the trend seen in the laboratory data (Figure 7) and does not represent
a specific model. If the fluorescence parameters always varied in a
regular way with changes in concentration of DOM, then deviations from such
a curve, should be an indication of the presence of other chemical
constituents that might affect the fluorescence.
c
-------
W
'E
3
e
c
o
CO
I
o
CO
o
c
to
o
(A
03
k.
o
12
11 -
10 -
9 -
n pH < 5.6
+ 5.6 < pH < 6.0
o pH > 6.0
470
,
474
1
478
482
Mean (nm)
486
490
494
Figure 8. Mean fluorescence wavelength vs. fluorescence intensity for
three pH ranges of the Michigan/Wisconsin lake samples,
excluding two lakes for which the fluorescence intensity was
exceptionally high. The solid line represents the expected
loci of samples for which only the concentration of DOM is
varying.
Both these attempts to predict pH suffer from the same interferences.
First, there does not appear to be a spectral or intensity response in the
fluorescence to changes in pH between about pH 6 to pH 8 (Figure 4).
Second, much of the scatter appears to be real, i.e., due to quenching and
precipitation effects of A1Q, Fe, and Ca, rather than a signal processing
or noise problem.
The prediction of pH can be improved -- or at least made less
confusing -- by use of some a priori assumptions about these other chemical
effects. For example, alkalinity class maps such as those given in the
NSWS Phase I were used to eliminate lakes from this analysis with a high
135
-------
calcium concentration from the regression. This was done under the
assumption that the calcium would cause precipitation of a portion of the
DOM, thus shifting the fluorescence spectrum, resulting in errors in the pH
prediction. The regression was performed with the mean fluorescence as the
only independent variable since neither the fluorescence ratio nor standard
deviation coefficients were significant at the 95% level (Figure 9). There
was a small improvement of the correlation (R2 = 0.41 for predicted versus
actual pH), and the standard error of the estimate decreased by almost half
(standard error of estimate = 0.48 pH units). The use of this a priori
information is limited in that not all lakes within an alkalinity class may
actually have that alkalinity. Similar a priori generalizations may be
difficult to make with iron and aluminum since it would require a rather
extensive knowledge of the soil chemistry of the watershed.
I
a
•o
a>
*•>
o
0)
i.
a.
pH
Figure 9. Predicted vs. actual pH for the Michigan/Wisconsin lake water
samples; samples from lakes in regions with high soil
alkalinity have been eliminated from the data set.
136
-------
ALUMINUM PREDICTION
Figure 10 presents fluorescence data for the Michigan/Wisconsin lakes
showing Al quenching of DOM fluorescence. The fluorescence quenching is
measured by a shift in the mean fluorescence wavelength with increasing
concentration of the organically bound metal. The measure of Al used --
organic aluminum, Al (mg/1) normalized to the DOC (mg/1) -- is an attempt
to account for the increasing number of reaction (and fluorescence) sites
as DOM concentration increases and an acknowledgement that only the
organically bound aluminum will quench fluorescence.
The effects of Al on fluorescence are further complicated by changes
of pH since the reactio°n of the metals with DOM is pH dependent. This pH
dependent reaction of A1o is illustrated in Figure lla, using the Michigan
lakes data for organic aluminum, Al (mg/L) normalized to total reactive
aluminum, Al (mg/L) plotted versus pH. Although these data are quite noisy
-- none of the ratios of AiyAl should exceed 1.0 -- the general trend in
the data, illustrated by the curved, solid line, is still apparent. The
peak in Al-DOM interaction at around pH 6 is similar to the results of a
model of Al-DOM binding presented previously (3) (Figure lib). Note that
the ratio of organic aluminum to total reactive aluminum can be the same at
two pH levels, indicating identical aluminum quenching and shift in the
mean fluorescence.
Using the fluorescence data to predict organic Al normalised to DOC
with the fluorescence mean, standard deviation and intensity ratio resulted
in much better correlations than those obtained for pH (Figure 12). The R2
was 0.89 for predicted versus actual, and the standard error of estimate
was .0032 Al /DOC units. All coefficients were significant at the 95%
level. This0 high correlation is strongly dependent on the presence of
several samples from a single lake with a high AiyDOC concentration.
These samples lie in line with the trend in the rest of the data and there
is no reason to suspect the chemistry or fluorescence value; however,
since there are no data at intermediate levels, the correlation is heavily
weighted by these samples. Without them the correlation would be less
convincing but still believable (R2 = 0.66; Figure 13). A better
predictive capacity than that for pH occurs because the measure of organic
Al normalized to DOC combines information on several water chemistry
parameters that determine the fluorescence. But there are still some
difficulties with a simple interpretation of the data. First, excess Al is
not predictable. Once the aluminum concentration exceeds the number of
possible binding sites on the DOM, the excess aluminum has no quenching
effect to cause a shift of the mean fluorescence and precipitation may
occur. This level of aluminum does occur for some lakes and would not be
predictable based on the DOM fluorescence of those lakes. Second, the
organic aluminum to DOM ratio does not contain information about calcium
and iron, which can also cause spectral changes. For example, when the
high alkalinity lakes (lake IDs 2B3-XXX) are excluded from the regression
(but including the outlier) the R2 increases to 0.91 with a standard error
of the estimate of .0034 AiyDOM units.
137
-------
0.04 -
0.035 -
0.03 -
.± 0.025 -
0
§ 0.02 ^
0.015 -
O
I 0.01 -|
^
O
0.005 -
n
D
D
nn
470
474
478 482 486
Mean Fluorescence (nm)
490
494
Figure 10. Spectral shift in the mean fluorescence wavelength with
increasing concentration of organically bound aluminum per unit
concentration of DOC.
CALCIUM
As these data were analyzed, it became apparent that the lake water
samples which were high in calcium, Ca, were somewhat unique. In spite of
the fact that calcium is not known to quench fluorescence, the Michigan
lake data appear to show just such an effect; in a plot of total Ca versus
mean fluorescence (Figure 14), a shift of the mean to shorter wavelengths
is apparent at high concentrations of Ca. Since calcium is known to cause
precipitation of DOM (12), the shift in mean wavelength is probably due to
the removal of the more reactive, longer fluorescence wavelength DOM. As
with the organic Al, there is not a unique fluorescence mean for each Ca
concentration.
138
-------
O)
l_
O
8
(O
'o
O
'E
re
en
L.
O
I
-
9~
•
*-
•
7 -
.
•
6_
5-
4~
-
Humic Substance Samples
+ Al-organic (left axis)
x % quenched (right axis)
+ *
f
X
t
^
X *
•
X
-»-
i i i i 1 — i i i — i — [~~i — r~i — i — | — n i i | i i i i | ' ' • ' [""" l~~1 r~
- tf u
-30
T3
0)
£
U
-20 a
D
CJ
:
-10
.
-0
3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
pH
Figure 11. pH dependence of DOM fluorescence quenching by organic
aluminum, a) Michigan/Wisconsin lake data.
b) Prepared samples and model results.
139
-------
0.045
0.01 0.02 0.03
AI(org)/DOC(mg/L)/(mg/L)
0.04
Figure 12. Predicted vs. actual organic
Michigan/Wisconsin lake data.
aluminum to DOC for the
DOM FLUORESCENCE AND WATER CHEMISTRY: A HYPOTHESIS
The variability in DOM fluorescence characteristics is quite complex
and cannot be ascribed solely to pH or to aluminum concentration. On the
other hand, fluorescence is clearly related to water chemistry and -- if
the chemical relationships were known -- could probably serve as an
indicator of changes in water chemistry. It is proposed that the observed
spectral changes in DOM fluorescence resulting from cation reactions are
due to preferential reduction of fluorescence from the humic acid portion
of the DOM either by quenching or precipitation of the humics. Presuming
that the humic acid portion of DOM fluoresces at the relatively longer
wavelengths, then reaction of DOM with various cations will reduce
fluorescence and shift the fluorescence to shorter wavelengths.
140
-------
o
o
o
0>
*-•
o
T)
0)
0.002
0.004
0.006 0.008 0.01 0.012
Alo/DOC
0.014
T T
0.016 0.018
Figure 13. Predicted vs. actual organic aluminum to DOC for the
Michigan/Wisconsin lake data; the one lake which appears to be
an outlier in Figure 12 was removed for this analysis.
Based on the results of the present study and related results in
published research, a general description of the factors affecting and
altering DOM fluorescence has been developed. DOM is a mixture of organic
molecules with inherent variability in reactivity and fluorescence.
Aromatic rings are the sites of most of the observed fluorescence and the
addition of functional groups onto the aromatic structure generally shifts
the fluorescence to longer wavelengths (13). The sharing of electrons
between the functional groups and the aromatic structure can provide a
means for cation reactions of DOM to affect DOM fluorescence, since the
functional groups, mainly phenolic and carboxyl groups, serve as cation
reaction sites (14). Aromatic rings and the associated functional groups
occur with varying frequency among the many organic molecules that make up
DOM. To some extent, the efficiency and spectral character of the
fluorescence of different types of DOM can be attributed to this
variability. DOM is sometimes divided into two broad categories: humic and
141
-------
fulvic acid. Hypothetical structures for typical humic and fulvic acid
molecules are shown in Figure 15. Typically, fulvic acid fluoresces more
intensely per unit DOC than humic acid. Fluorescence from humic acid is
also shifted toward the longer (redder) wavelengths relative to that from
fulvic acid.
40
V>
•o
C
to
35 -
30 -
O
t 20 H
O)
3
CO
O
15 -
10 -
5 -
470
474
—T
478
"h M1"-'
1
482
a
a
486
490
494
Mean Fluorescence (nm)
Figure 14. Shift in the mean fluorescence with increasing concentration of
calcium.
Complexation of a cation with a functional group which is attached
directly to an aromatic ring may quench fluorescence by de-excitation of
the shared electron. Complexation with a functional group not directly
attached to an aromatic ring is less likely to alter fluorescence. Since
more of the functional groups in humic acids are attached directly to
aromatic rings, fluorescence quenching is more likely to occur with humics
than with fulvics.
142
-------
OH COOH
HOOC
CH2OH
'CH2-COOH
can OH
CH2-C-"C^OH
0 COOH
Type structure of fulvic acid
as proposed by Buffle in 1977,
(HC-OH),
(Sugor)
COOH
!
COOH
COOH
Hypothetical structure of humic acid showing free and bound phenolic
OH groups, quinone structures, oxygen as bridge units, and carboxyls
variously placed on the aromatic ring.
Figure 15. Hypothetical structure of typical humic and fulvic acids.
Fulvic acid fluorescence is stronger per unit DOC and somewhat
more blue than humic acid fluorescence.
143
-------
Reaction of the functional groups with Al, and Fe decreases
(quenches) the total fluorescence and alters the DOM fluorescence spectrum.
The decrease in fluorescence (quenching) is thought to be due to
deexcitation of the electron; the presence of the metal enhances transfer
through nonradiative states of energy which would otherwise be converted
into fluorescence (13). Fluorescence quenching by aluminum was apparent in
Figures 9 and 10. Quenching by iron could not be observed since the data
set lacked a measure of organic Fe, but several studies have shown Fe does
quench fluorescence (1,15). At very high concentrations the metals may
initiate precipitation of the higher molecular weight humic acids (15,16).
Calcium does not appear to directly affect fluorescence to the degree
iron or aluminum can, in part because it is less tightly bound (15,17).
But, reaction of DOM with Ca*2 can cause precipitation of higher molecular
weight DOM, i.e. humic acid, from solution (12). Thus, the fluorescence
spectrum is altered as in Figure 14.
Evidence also indicates that the hydrogen ion, H+, alters DOM
fluorescence both by direct quenching and by precipitation of DOM
(2,13,14,15,18; and Figure 5).
Since the fluorescence of humic acids is more susceptible to
quenching than fulvic acids and humic acids are also more likely to
precipitate than fulvic acids, all cations will have essentially the same
qualitative effect on the fluorescence whether by direct quenching or
through precipitation. Although the mechanisms may differ, the overall
effect of all the cations, (Al*3, Fe*3, Ca*2 and H+) will be similar: a
reduction of fluorescence and a spectral shift toward the blue.
In summary, DOM exhibits a continuum of structures and spectral
ranges of fluorescence. Observed spectral changes in DOM fluorescence
because of cation reactions may not be due to spectral shifts in the
fluorescence of individual molecules, but rather the preferential reduction
of fluorescence from the humic acid portion of the DOM. If the humic acid
portion of DOM fluoresces at the relatively longer wavelengths of the DOM
fluorescence continuum, then reaction of DOM with various cations shifts
the DOM fluorescence to shorter wavelengths. Since all the cations compete
for the same reaction sites it will be difficult to distinguish among them
based only on their effects on DOM fluorescence.
SUMMARY AND CONCLUSIONS
Trends suggesting a direct connection between pH or aluminum and DOM
fluorescence had been apparent in previous laboratory tests. Similar
trends were apparent in the lake water spectra in this study, but a simple,
predictive relationship between fluorescence spectra and pH was not found.
In fact, attempts to predict pH based solely on fluorescence parameters
were quite poor. On the other hand, the correlation between fluorescence
parameters and organically bound aluminum per unit DOC (AiyDOC) was good.
144
-------
These results are consistent with a rather complex picture of DOM
fluorescence. DOM is a collection of organic molecules which vary in their
fluorescence efficiency per unit DOC and in the spectral composition of
their fluorescence. Differences in fluorescence may be the result of
quenching of fluorescence, natural variability in the composition of DOM,
or preferential precipitation of high molecular weight DOM (humic acid).
Specifically, aluminum, Al+3, and iron, Fe*3, both quench fluorescence. The
hydrogen ion, H+, alters fluorescence both by quenching and by
precipitation of high molecular weight DOM (humic acid), and calcium, Ca*2,
alters fluorescence by selective precipitation of high molecular weight
DOM. Unfortunately, the diverse factors affecting fluorescence all appear
to have a similar effect; all reduce the total fluorescence and shift the
fluorescence toward the blue.
Quenching and a gross spectral shift of the fluorescence were the
only significant effects observed in this study. The typical DOM
fluorescence spectrum is not complicated by any strong absorption or
emission lines, and none are introduced by any of the parameters considered
in this study. Thus, by themselves, DOM fluorescence spectra are not
sufficiently detailed to distinguish among more than a few measures of
water chemistry.
The most convincing relationship found in this study was between
fluorescence parameters and Al /DOC. This appears to be a fairly general
relationship. The correlation of fluorescence intensity with DOM
concentration is probably also robust when the composition of DOM is
constant and quenching effects are not excessive.
Remote fluorescence data would be more generally useful for
monitoring spatial or temporal change in order to extrapolate from a
limited number of water samples. A change in fluorescence that was
consistent with a suspected change in pH and detected in lakes observed
over a period of time could be used to determine whether or not to sample
that lake. Fluorescence data would also be useful in the more standard
remote sensing application of extrapolating from a limited number of water
samples to characterize a larger area. The complexity observed in the
forty-nine lakes included in this study would not likely occur in any
single lake. Thus, changes in fluorescence would be more easily attributed
to a specific cause.
The primary difficulty in extending these results farther lies with
the relatively poor understanding of DOM fluorescence properties and how
they are affected by various cations. DOM fluorescence might be a useful
analytic tool, but applications will be hampered by the lack of clear
understanding of the DOM properties which affect fluorescence. We have
presented a theory that is supported by the literature but requires some
specific experimental results to strengthen it. The variablity of DOM
structure and reactivity is important in understanding both intensity and
spectral fluctuations. Of particular interest is the frequency of
occurrence of the fluorescence centers and the position and relative
frequency of the functional groups.
145
-------
Attention should also be given to the use of other forms of remote
spectroscopy. Raman scattering, for example, is very specific to the
scattering molecule and could be induced and observed remotely with
basically the same type of hardware used in this study. In order to
address lake acidification directly, relationships with pH need to be
developed since H* itself is not Raman active. It might be feasible to
attempt remote measurement of S04"2, N03, or other significant ions.
The conclusions may be summarized as follows:
1. Spectral fluorescence of dissolved organic matter (DOM) is directly
related to water chemistry. Specifically, it is controlled by the
concentration and composition of DOM, and the reaction of DOM with
various cations (Al+3, Fe*3, H+, Ca+2).
2. Although both pH and aluminum alter DOM fluorescence, fluorescence
alone is not adequate for detecting low pH or high aluminum
concentrations in a wide population of lakes.
3. The ratio of organic aluminum to DOC (AiyDOC) does appear to be
generally predictable.
4. Changes in fluorescence parameters with time should be indicative of
chemical changes; e.g., LIF would be useful as an independent
monitoring tool to detect change and select those lakes requiring
more detailed study.
5. Applications will be hampered by the lack of clear understanding of
the DOM and the environmental properties that affect DOM
fluorescence. We have presented a theory that is supported by the
literature but requires some specific experimental results to
strengthen it.
7. Attention should also be given to the use of other forms of remote
spectroscopy, particularly Raman scattering, which is very specific
to the scattering molecule.
REFERENCES
1. Vodacek, A., and W.D. Philpot. Use of Induced Fluorescence
Measurements to Assess Aluminum-organic Interactions in Acidified
Lakes. Proceedings: 51st Annual Meeting, American Society of
Photogrammetry, 1985.
2. Vodacek, A., and W.D. Philpot. Environmental Effects of
Laser-induced Fluorescence Spectra of Natural Waters. Remote Sensing
of Environment, 21:83-95, 1987.
146
-------
3. Vodacek, A. Laser Fluorosensing for Remote Detection of Dissolved
Organic Carbon and Aluminum in Water. M.S. Thesis, Cornell
University, 1985.
4. Vertucci, F.A., and A. Vodacek. The Remote Sensing of Lake
Acidification Using Laser Fluorosensing. Proceedings: 51st Annual
Meeting, American Society of Photogrammetry, 1985. 793-801 pp.
5. EPA. NSWS Stream Survey, Field Training and Operations Manual, 1986.
6. EPA. Characteristics of Lakes in the Eastern United States.
EPA/600/4-86/007a: Vol. 1, Population descriptions and
Physico-Chemical Relationships, 1986. 136 pp.
7. Bristow, M., and D. Nielsen. Remote Monitoring of Organic Carbon in
Surface Waters. Report No. EPA-600/4-81-001, NTIS PB81-168965,
Environmental Monitoring Systems Laboratory, Las Vegas, NV, 1981. 83
pp.
8. Bristow, M., D.Nielsen, D.Bundy, and R.Furtek. Use of Water Raman
Emission to Correct Airborne Laser Fluorosensor Data for Effects of
Water Optical Attenuation. Applied Optics, 20:2889-2906, 1981.
9. Bristow, M.P.F., D.H.Bundy, C.M.Edmonds, P.E.Ponto, B.E.Frey, and
L.F.Small. Airborne Laser Fluorosensor Survey of the Columbia and
Snake Rivers: Simultaneous Measurements of Chlorophyll, Dissolved
Organics and Optical Attenuation. International Journal of Remote
Sensing, 6:1707-1734, 1985.
10. Stewart, A.J. and R.G.Wetzel. Fluorescence/Absorbance Ratios. A
Molecular-weight Tracer of Dissolved Organic Matter. Limnology and
Oceanography, 25:559-563, 1980.
11. Laane, R.W.P.M, and L. Koole. The Relation Between Fluorescence and
Dissolved Organic Carbon in the Ems-Doll art Estuary and the Western
Wadden Sea. Netherlands Journal of Sea Research, 15:217-227, 1982.
12. Stewart, A.J., and R.G.Wetzel. Asymmetrical Relationships Between
Absorbance, Fluorescence, and Dissolved Organic Carbon. Limnology
and Oceanography, 26:590-597, 1981.
13. Wehry, E.L. Practical Fluorescence: Theory. Methods, and Technique.
Guilbault & Guilbault, eds., Marcel Dekker, New York, 1973.
79-136 pp.
14. Stumm W., and J.J. Morgan. Aquatic Chemistry. 2nd edition,
Wiley-Interscience, New York, 1981.
15. Willey, J.D. The Effect of Seawater Magnesium on Natural
Fluorescence During Estuarine Mixing, and Implications for Tracer
Applications. Marine Chemistry, 15:19-45, 1984.
147
-------
16. Willey, J.D., and L.P. Atkinson. Natural Fluorescence as a Tracer
for Distinguishing Between Piedmont and Coastal Plain River Water in
Nearshore Waters of Georgia and North Carolina. Estuarine, Coastal
and Shelf Science, 14:49-59, 1982.
17. Ryan, O.K., and J.H. Weber. Copper (II) Complexing Capacities of
Natural Waters by Fluorescence Quenching. Environmental Science and
Technology, 16:866-872, 1982.
18. Laane, R.W.P.M. Influence of pH on the Fluorescence of Dissolved
Organic Matter. Marine Chemistry, 11:395-401, 1982.
148
-------
APPLICATION OF REMOTE SENSING TECHNIQUES FOR ESTIMATING SPATIAL
VARIABILITY OF DRY DEPOSITION OF ACIDIC POLLUTANTS
by: Lawrence T. Fisher and Mark V. 01 sen, Lockheed Engineering and
Management Services Company, Inc., Las Vegas, NV; Robert T. McMillen,
National Oceanic and Atmospheric Administration, Oak Ridge, TN; Barbara
Levinson, U.S. EPA Office of Acid Deposition, Environmental Monitoring and
Quality Assurance, Washington, D.C.; and Mason J. Hewitt III, U.S. EPA
Environmental Monitoring Systems Laboratory, Las Vegas, NV
NOTICE
Although the research described in this article has been funded
wholly or in part by the U.S. Environmental Protection Agency through
contract 68-03-3245 to Lockheed Engineering and Management Services
Company, Inc., it has not been subjected to Agency review and therefore
does not necessarily reflect the views of the Agency and no official
endorsement should be inferred.
ABSTRACT
Dry deposition of gases and particulates constitutes a significant
part of the total acid deposition to the environment, but measurements are
difficult to obtain. Dry deposition atmospheric trace gases may be
calculated from concentration measurements and modeled deposition
velocities, Flux = concentration (c) x deposition velocity (Vd). The Vd is
known to be influenced by, and often controlled by, the local meteorology,
terrain features, and vegetation type, and these variables are used to
drive a Vd model currently being developed. The purpose of this project
was to study the variability of the Vd across an 80x80 kilometer (km) area
in central Pennsylvania with vegetation type and terrain being the sole
manipulated variables. Digital imaging processing techniques were applied
to analyze terrain data from digital elevation models and ground "cover
information was derived from Landsat satellite data. Both types of data
were available with ground resolutions of 30 meters. The area was
subdivided with Ikmxlkm cells and the model was calculated for each cell.
The variability of the Vd is now being analyzed.
149
-------
INTRODUCTION
DRY DEPOSITION OF ACIDIC POLLUTANTS
As a result of concerns that acid deposition may be accelerating
deterioration of natural ecosystems, materials (metals, paints, and
carbonate stone) and public health, Congress authorized the National Acid
Precipitation Assessment Program (NAPAP) in 1980. The research program has
focused on the emissions of the acid precursors, sulfur dioxide and the
nitrogen oxides, the processes that convert these gases into acids, the
monitoring of these substances as they deposit, and the effects they have
on the environment.
Acid deposition is often divided into its wet and dry components.
Wet deposition is measured on a routine basis by the analysis of
precipitation at the 150 sites of the National Trends Network. Monitoring
dry deposition is much more difficult. At present, the direct measurement
of dry deposition is performed at a few research sites; however, the
technique employed is inappropriate for routine monitoring. As a result,
the National Dry Deposition Network was designed to measure air
concentration and meteorological parameters from which flux could be
inferred.
The deposition flux is inferred by multiplying the air concentrations
by a variable called the dry deposition velocity (Vd). The deposition
velocity model inputs are meteorological, vegetation, and terrain data.
Because of the great spatial variability of the parameters that are used to
calculate deposition velocity, such as vegetation type and topography, the
extrapolation of deposition velocity from a point measurement to an areal
average carries with it large uncertainties.
Because the site measurements from the Dry Deposition Network will
ultimately be extended to areal estimates of total loading, the need to
quantify the uncertainty in the extrapolation estimates is critical. A
pilot study was therefore initiated to quantify the spatial variability of
the deposition velocity due to vegetation and terrain complexity across an
80x80 km grid in central Pennsylvania. Landsat Thematic Mapper (TM) Data
and U.S. Geological Survey (USGS) Digital Elevation Models were used as
inputs into the deposition velocity model.
TEST SITE
LOCATION, SIZE, AND CHARACTERISTICS
A test site approximately 83 kilometers square was selected in
central Pennsylvania, centered about 20 km east of State College (Figure
1). Morphologically, this area is dominated by Appalachian ridge and valley
structures, with ridges rising several hundred meters above valley floors
and oriented from southwest to northeast. Numerous valleys and glens are
150
-------
PENNSYLVANIA
Altoona
^Pittsburgh
College
AHarrisburg
Philadelphia
Figure I. Test site.
151
-------
incised into the ridges. Lower slopes are generally limestone, capped by
softer sandstones. Often, the sandstones have eroded to form shallow
valleys running along the tops of the ridges. The northern third of the
area includes the beginnings of the Allegheny Plateau, rising to a highland
area incised by river valleys several hundred meters deep. Elevations
range from about 120 meters in the Juniata River valley in the southeastern
portion of the test site to about 750 meters in the Plateau to the north
and west.
Valley bottoms are generally agricultural, with corn and pasture
crops predominating. Frequent woodlots, usually deciduous, also appear.
Ridges are almost entirely covered with deciduous forest, mostly oak. A
few hemlock appear in shaded, damp areas. Some open grassy areas exist,
usually in conjunction with brush and scrub.
The Allegheny Plateau is generally deciduous forest with some open
agricultural areas, usually pasture. There are frequent strip mines, many
abandoned and at least partially grown over with grass or brush.
Urban areas include State College and the Pennsylvania State
University, near the center of the test site, and the towns of Huntington,
Williamsburg, and Lewistown in the southern portion.
TEST SITE GRID
The test site boundary is rotated with respect to north because of
the design of the computer model being used. It is subdivided into 6,889
(83 x 83) cells each covering one square kilometer. Data products prepared
for this project were produced based on this system of grid cells.
RELEVANT DIGITAL DATA SETS
DATA REQUIREMENTS
To obtain information about terrain and land cover variability, data
with as much detail as possible were required. However, cost and time
constraints mandated that the data be obtained from already-existing
sources. Computer-compatible data were required, already in digital form.
Fortunately, such data exist in the form of Digital Elevation Models and
remotely-sensed images obtained from satellites. Both of these forms are
"raster based" data, in which data elements exist as lines of discrete data
points.
DIGITAL ELEVATION MODELS
Digital Elevation Models (OEMs) based on large scale (7.5 minute)
maps are a relatively new product offered by the U.S. Geological Survey
(USGS). The program is new and coverage nationally is far from complete.
However, there is fairly complete coverage of the test site for this
project.
152
-------
General Description
7.5 minute OEMs produced by the USGS are prepared either from
digitized contour maps or by scanning stereographic aerial photographs.
Each DEM covers an area bounded by a standard 7.5 minute topographic map
quadrangle. Elevations are supplied at 30-meter intervals along
south-to-north profiles separated by 30 meters. Approximately 400 profiles
are supplied for each 7.5 minute quadrangle.
Each elevation is presented relative to a base or datum elevation
(which may change from profile to profile) at a resolution of 1 meter.
Vertical accuracy for all of the OEMs used in this project is stated by
USGS to be within 7 meters Root Mean Square Error (RMSE) (1,2).
DEM data are supplied by the USGS on standard computer tapes. Each
DEM occupies about 1.25 million bytes or characters so a single tape can
hold data for 10 or more OEMs.
UTM Coordinates
The bounds of map quadrangles, and thus of OEMs, are set by the
graticule, or system of lines of latitude and longitude. DEM profiles,
however, are aligned with the Universal Transverse Mercator (UTM)
coordinate system. This system assigns a unique coordinate position
consisting of a zone number (1 to 60) and an easting and northing, in
meters, to every point on Earth whose latitude is between 75 degrees south
and 75 degrees north. (Polar latitudes are handled differently and will
not be considered here.)
UTM coordinates are defined by imposing a Cartesian (right-angled)
X-Y grid on a series of 60 maps of the Earth drafted using a Transverse
Mercator projection. Each map covers a zone 6 degrees wide. The zone
centers, or "central meridians" are 180 degrees West, 174 degrees West,
etc., for Zones 1, 2, etc. The maps are defined to have a scale slightly
larger than the Earth itself. A mathematically regular ellipsoid is chosen
for each zone to best approximate the true shape of the planet in that
zone. From the nature of the ellipsoid, the longitude of the central
meridian, and the scale, any latitude and longitude within the zone
possesses a unique X-Y value where X is negative west of the central
meridian and Y is negative south of the equator. An arbitrary "false
easting" of 500,000 meters is added to the X value so that all eastings are
positive. Similarly, a "false northing" of 10,000,000 meters is added to
southern hemisphere "Y" values.
The north or "Y" axis of the coordinate system grid in any zone is
perfectly aligned with the central meridian of each zone, and "X"
coordinates (lines of constant northing) are tangent to parallels at the
center. (All Mercator projections are "conformal", or angle-preserving, so
all meridians and parallels intersect at right angles.) Away from the zone
center, however, the UTM grid departs markedly from the graticule. At the
zone edges, the angle between the grid and a meridian or parallel is
approximately 1.7 degrees.
153
-------
The angle between the graticule and the UTM grid has some
implications for users of OEMs. First, the beginning and ending
coordinates of profiles differ from profile to profile, depending on where
they intersect the parallels or meridians bounding the quadrangle. This
can complicate the process of mosaicing multiple OEMs into a seamless
entity. Second and more serious, the change from one UTM coordinate system
to another introduces very significant complications if a zone boundary is
crossed. This indeed was the casa in this project.. .the 78th meridian
West, which is the boundary between UTM zones 17 and 18, passes directly
through the test site.
The processing and analysis problems imposed by the grid angles and
the crossing of UTM zones were solved in software and will be discussed
below.
Availability
DEM coverage of western and central Pennsylvania is reasonably
complete. Figure 2 shows OEMs that were available for this project.
Several quadrangles in the southwestern corner of the test site, the
extreme southeastern corner, and one quadrangle in the north edge were
missing. An inconvenient omission was the Julian quadrangle, in the west
center of the site. However, some data for this quadrangle, at reduced
accuracy and resolution, were obtained from the Pennsylvania State
University.
LANDSAT THEMATIC MAPPER
Land cover information was derived from the Landsat Earth observation
satellite system operated by NASA and EOSAT, Inc. The present satellite,
Landsat 5, is the latest of a series of orbiting platforms designed to
provide information about Earth's resources; predecessors have been
collecting data since 1972.
General Description
Orbital Characteristics--
Landsat 5 orbits in a nearly circular near-polar orbit at an altitude
of about 705 km. The orbit is sun synchronous, so that the local time of
equatorial crossing is constant with each orbit. It passes south on the
lighted side of the planet, returning north on the dark side. Each orbit
takes approximately 100 minutes. The satellite repeats its coverage over
any given area at intervals of 16 days.
Sensor System--
The primary sensor on Landsat is the Thematic Mapper (TM). This is a
scanning radiometer which sweeps across the satellite's direction of
motion, sensing reflected or emitted electromagnetic radiation in seven
wavelengths ranging from green to far infrared (3). Table 1 summarizes
these. As the satellite moves forward, scanned information builds up a
two-dimensional image of the Earth.
154
-------
41°00'
40°30'
78°00'
77030'
77°00'
41°00'
40°30'
78°00'
77°30'
77°00'
Figure 2. Digital elevation models available.
155
-------
TABLE 1. LANDSAT 5 THEMATIC MAPPER SPECTRAL CHANNELS
Wavelengths
Channel (Micrometers)
1
2
3
4
5
6
7
0.45- 0.52
0.52- 0.60
0.63- 0.69
0.76- 0.90
1.55- 1.75
10.40-12.50
2.08- 2.35
Scene Size and Resolution--
The scanning system views a swath of the Earth's surface 100 nautical
miles (185 km) wide. At any instant, radiation is received from an area
approximately 30 meters in extent for reflected bands and 120 meters for
the thermal infrared band (band 6). The scan rate and the sampling rate
during each scan are commensurate with this instantaneous field of view, so
that in the reflected bands each resulting sample is slightly less than 30
meters in size.
Data Delivery--
Radiation data are sampled, converted to digital form, and relayed to
Earth via another satellite system called the Tracking and Data Relay
Satellite System (TDRSS). Data are received at White Sands Missile Range,
New Mexico, forwarded to the the Goddard Spaceflight Center near
Washington, DC, and ultimately disseminated to users by EOSAT, Inc.
Data are organized in "scenes" 185 km square. Each scene consists of
2,983 scan lines, each 4,220 samples wide. Because of the very large
amount of data, data are provided on computer tapes organized as
quarter-scenes. A total of six computer tapes are needed for each scene.
Landsat TM data are usually delivered in a partially geometrically
corrected form. Effects of Earth's rotation are corrected by EOSAT by
adding appropriate numbers of zero samples before and after each scan line.
The data set is aligned with the sub-satellite ground track of the system,
which is about 13 degrees west of south at the latitude of the test site
for this project.
East-west extent of a scene is dictated by orbit and scanner
geometry. North-south extent is much more arbitrary, since data acquisition
can be continuous during an orbit. EOSAT has arbitrarily established a
series of "row numbers" across each orbit. Until recently, they would only
provide data sets centered along these rows. This policy has now been
changed, but it was in effect at the time data were obtained for this
156
-------
project. As a result, it was not possible to obtain Landsat data for the
north 15 percent of the test site, which lay north of the row boundary.
Availability
Time constraints on this project prevented any field activities until
late August 1987. Landsat coverage as near that date as possible would
have been desirable. However, no late-season data were available due to
cloud cover. Accordingly, a scene recorded on June 25, 1987 was obtained.
METHODOLOGY
DIGITAL ELEVATION MODEL ANALYSIS
For elevation analysis of a large area, it is impractical to process
individual OEMs. This is because of the difficulty in carrying consistent
analysis methods across the edges and onto to neighboring OEMs, a
difficulty which is compounded by the rotated boundaries of OEMs. It is
much better to mosaic multiple OEMs into a seamless whole.
DEM Mosaics From the Pennsylvania State University
Researchers at the Office for Remote Sensing of Ea'rth Resources
(ORSER) at the Pennsylvania State University had earlier undertaken a
project to form DEM mosaics of Centre County, Pennsylvania (the county in
which State College is located). Two mosaics had been produced. One, for
quadrangles in UTM Zone 17, west of the 78th meridian West, consisted of a
mosaic of 10 OEMs. The other, for the eastern part of the County in UTM
Zone 18 (east of 78 degrees longitude), included 24 quadrangles. The
extents of these mosaics are shown in Figure 3.
The Zone 18 mosaic also included data digitized at ORSER from
topographic maps for the otherwise unavailable Julian quadrangle. These
data were at lower resolution and accuracies than USGS OEMs.
Although DEM elevations are only supplied to the nearest meter above
some datum elevation, that datum is defined to a high degree of precision.
For that reason, and to incorporate possible future high-precision data,
each elevation in the ORSER mosaics was stored as decimeters (tenths of
meters). Through a cooperative arrangement with ORSER, these mosaics, on
computer tape, were made available to us in exchange for copies of
additional OEMs obtained from USGS for parts of the test site outside
Centre County.
Other OEMs
In addition to the 34 OEMs in the 2 mosaics, ORSER provided an
additional 11 that were not part of the mosaics. These were provided on
magnetic tape in a variety of formats and blocking factors.
157
-------
Ol
CO
N4.560
N4.540
4,520
UTM ZONE 17
COORDINATES
IN 1000's OF METERS
o
IN
N4.560 N4.560
-N4.540 N4.540-
N4.520 N4.520
N4.500
N4.480
N4.460
4-
O
CNJ
CM
O
»»•
CM
-I-
c\j
UJ
o
^O
CSJ
o
00
CM
UJ
o
00
C\J
O
o
UTM ZONE 18
COORDINATES IN 1000's OF METERS
Figure 3. UTM Zone 17 and Zone 18 digital elevation model mosaics.
o
eg
ro
N4.560
4.540N
N4.520
N4.500
N4.480
N4.460
-------
Ten additional OEMs were not available through ORSER. These were
obtained from the USGS National Cartographic Information Center.
DEM File Format
To facilitate DEM analysis, a disk file format was designed at EPA's
Environmental Monitoring Systems Laboratory, Las Vegas, Nevada (EMSL-LV) to
store DEM information in a fashion similar to that used for other
raster-based data sets. This file was organized as a direct access file
storing sequences of 16-bit elevation values at 30-meter intervals for
lines of constant northing. Each line begins on a 512-byte sector boundary
and occupies as many sectors as needed, with 256 16-bit elevations per
sector. Lines of constant northing follow one another in the file from
south to north.
The largest signed integer that can be stored in 16 bits is 32,767.
If elevations are stored in decimeter, the largest elevation storable would
be 3276.7 meters or about 10,000 feet. To allow for possible future needs
at high elevations, the file was designed to store values either in
decimeter or meters, indicated by a code value in a file header block.
Also, to allow for possible future projects in areas at or below sea level,
the file was designed so that missing data are represented by -9,999
instead of zero. Header data in the DEM file format store, minimum and
maximum eastings and northings, UTM zone number, and the m/dm elevation
code. Once the DEM file format was designed, two such files were created
and initialized. One was large enough to accommodate the ORSER UTM Zone 17
data set (and one additional quadrangle that was later obtained from USGS).
The second file, based on UTM Zone 18 coordinates, was designed to be large
enough to encompass the entire test site. Its bounding coordinates were
extended west across the zone boundary by about 0.26 degrees of longitude.
The bounds and size of this file are summarized in Table 2.
TABLE 2. BOUNDS AND SIZE OF OVERALL DEM MOSAIC FILE
Coordinate System: UTM Zone 18
Eastings: 212,910 meters E to 321,960 meters E
Northings: 4,457,460 meters N to 4,571,820 meters N
Number of northings (at 30 meter intervals): 3,813
Number of elevations per northing
(30-meter intervals): 3,636
Total Number of Elevations: 3813 x 3636 = 13,864,068
Number of Sectors per Northing: 15
Total File Size in Sectors (including header): 57,196
159
-------
Adding OEMs to the Mosaic
Assembling OEMs one-by-one into a mosaic using conventional image
processing programs is difficult because the boundaries of the OEMs are
generally rotated with respect to the coordinate system grid. This
operation was made still more difficult by a variety of tape formats and
blocking factors on various DEM tapes provided by USGS and ORSER.
Therefore a general-purpose program was written to deal with this
operation. It takes advantage of the large physical, and larger virtual
memory, of the Digital Equipment Corporation VAX 11/785 computer.
This program, MakeDEM, either creates and initializes a new DEM file
or operates with an existing file that has some data already mosaiced into
it. It reads an entire new DEM data set from tape into main memory, taking
into account the different formats that are available. These data include
information about the global UTM coordinates of the entire DEM. Using
these bounds, the corresponding portion of the DEM file is copied from disk
into main memory in a large 2-dimensional data array.. To simplify program
bookkeeping, data from disk are always read starting at sector boundaries,
where each sector stores 256 16-bit elevation samples across a line of
constant northing. A data array holding 512 northings each 768 (3 x 256)
elevations wide is sufficient for all possible DEM and sector geometries.
Once data for the appropriate portion of the disk file have been copied
into main memory, the DEM data can be processed profile by profile. The
tape data for each profile (by now resident in main memory) contain header
information giving starting easting and northing, number of elevations, and
datum elevation for the profile. These data are used to determine where to
place the new elevations in the memory array.
When data for an entire DEM has been added to the memory array, the
array is written back to the DEM disk file. Then the process is repeated
for the next DEM on tape, if any.
The UTM Zone Crossing Problem
It was essential to resample elevation data from the UTM Zone 17
mosaic into UTM Zone 18 coordinates to place all elevations for the test
site into a consistent coordinate system. To accomplish this, Program
UTMtoUTM was written. This program was constructed as a general purpose
routine and is not limited to this particular project. However, this
project will be used to explain its operation. The setting will be as
follows:
(1) The Zone 17 mosaic (the smaller of the two) was resampled to produce
elevations in the Zone 18 data file.
(2) The Zone 18 mosaic was constructed with sufficient "overlap space" on
its west side to accommodate the resampled data.
160
-------
UTM Projection Mathematics--
The mathematics to accomplish the projection of latitude and
longitude to Cartesian X, Y form and vice versa for UTM are presented by
John P. Snyder (4). Subroutines to transform latitude and longitude into
UTM eastings and northings (the "forward projection") and from zone,
easting, and northing to latitude and longitude (the "inverse projection")
were taken directly from that reference.
Mercator projections in general and Transverse Mercator projections
in particular are not especially complicated if a spherical Earth can be
assumed. In this case, Mercator projections can be calculated for virtually
the entire globe, although the Transverse Mercator case produces a world
map with very unusual shape. However, large scale mapping requires an
ellipsoidal globe. In this case, the Transverse Mercator mathematics
involve elliptic integrals which have no closed form solutions. Instead,
series approximations are used. These lose accuracy as distance from the
map's central meridian increases, but the errors are acceptable for about
four degrees, or about one degree beyond the edge of a UTM zone (4).
This implies that OEMs extending up to about a degree east or west of
a UTM zone boundary can be successfully resampled and placed in the
coordinate system of the adjoining zone.
General Method--
The process requires two DEM files:
File 1 contains elevation data for UTM Zone 17 (or in general, for
any zone).
File 2 is based on Zone 18 (or more generally for either zone
adjacent to that of File 1).
For each northing in File 2, we determine which points are across the
zone boundary of Zone 18. From the UTM Zone 18 coordinates, we calculate
latitude and longitude of the point, then find its UTM Zone 17 coordinates.
From this, we determine the four nearest samples in the Zone 17 file (File
1), do a bilinear interpolation to estimate the elevation at our point of
interest, and place that elevation into File 2.
Calculating the Zone Boundary Coordinates--
As each northing in File 2 is processed, it is necessary to calculate
exactly which multiple of 30 meters is the first that crosses the UTM zone
boundary. The boundary line, if plotted on a sketch of File 2 such as
Figure 3 (whose coordinate system is UTM for Zone 18) is nearly a straight
line but is not exactly so. The projection mathematics that relate
latitude and longitude to UTM easting and northing are nonlinear.
Therefore, an iterative binary search is used to determine the zone edge.
This operates as follows:
(1) At the northing in question, form a line joining eastings of
500,010 meters (the nearest multiple of 30 meters to the center
of the zone) and an easting that is either 100,020 meters (if
161
-------
the overlap region is at the west side of File 2) or 900,000
meters (if the overlap area is at the east side.) This easting
is well beyond the zone boundary for any possible latitude.
(2) Calculate the latitude and longitude of the midpoint of this
line using the inverse projection mathematics.
(3) If the midpoint is east of the zone edge, replace the west end
with the midpoint; else replace the east end.
(4) Repeat (2) and (3) until the line is reduced to two points that
are 30 meters apart. One point will lie on one side of the
zone boundary and the other on the opposite side.
The Resampling Algorithm--
First the corner eastings and northings of File 1 (in UTM Zone 17
coordinates) are inverse projected to latitude and longitude and then
forward projected to Zone 18 coordinates to determine what portion of File
2 must be processed. Also, as much data from File 1 is read into main
memory as memory space allows, starting at its southern edge.
Figure 4 shows the geometry involved. A typical Zone 18 northing is
shown in the overlap region (solid line). The dotted lines show the north
and south extremes of Zone 17 data encountered. For every such northing in
File 2 in the processing portion (in Zone 18 coordinates, and in steps of
30 meters),
(1) Find the easternmost 30-meter multiple that is west of the zone
boundary using the binary search outlined above. Also find the
westernmost 30-meter multiple that is at, or east of, the
western extreme of File 1.
(2) For each easting between these two extremes (in steps of 30
meters), calculate latitude and longitude with inverse
projection mathematics, and then calculate the corresponding
Zone 17 coordinates using the forward projection. Call this
coordinate [E17, N17].
162
-------
-^
Cl
E220
E240
03'
e
*
E260
E280
E300
E320
-f
-^
00
+
£ *• .*• *-
-y, 01 01 01
§ o e o
E740
1
-h t- T
4
|
H
4
s
h
j
*.
CT>
' *
•+
•:
•*•
+
-t-
z
^
*.
i |
i |
I i
i i
1 1
f I
1
1 i
1 1
II
*
'•>
;. i*';., '--,*- .-'"* •'-• •+• f-
* " V ' -
.t. J-t '^T'1 **•*'
T^*- v ^T^ •• -.% T "• i
-
• t 1 < . j
s
! *
„ - ; " v * s ,
-! -- 5 4* ?....... 4». -,.., ,j -H
F 7 Aft
-J
OB
e
£
E260
E280
E300
E320
S * .* *
- 01
2 0 0 0
o '-' '-'
Figure 4. Geometry of the UTM zone crossing resampling algorithm.
163
-------
In general, this point will lie at some point surrounded by four
elevations zO, zl, z2, and z3 in File 1. The geometry will look something
like this:
(
1 —
/
:
0 --
(SW
W)
V
'
) 2
V. jy I'l ^
12 E17
[x, y]
• +
'.Q 7
z3 (NE
/
- N17
30
!l (SE)
)
<
M
zO, zl, z2, and z3
are elevations at
30 M grid locations
surrounding point
[E17, N17]
Now subtract multiples of 30 from E17 and N17 and divide the remainders by
30 to obtain a normalized coordinate [x, y]. Use this and the elevations
of the four File 1 neighbors to perform a bilinear interpolation.
Interpolate first along the y
the y = 1 edge, joining z2 and z3:
0 edge, joining zO to zl, and along
zO
z2
Pa
Pb
zl
) X
1
z3
Pa = zO (1 - x) + zl x
Pb = z2 (1 - x) + z3 x
+-
0
-+
1
164
-------
and then interpolate along y between Pa and Pb.
1 + -__-___ pD
Pa (1 - y) + Pb y
Pa
REMARK: This interpolation could have been equally well done by
interpolating between zO and z2 (at x = 0) and between zl and
z3 (at x = 1), and then interpolating between these two values
along y. Expanding the interpolation formulas above, and those
from the alternative, results in each case in:
z = zO + (zl - zO)x + (2.2 - zO)y + (zO - zl - z2 + z3)xy
As this process continues, working north up the two files, the time
may come when the program attempts to reference a File 1 data value that is
north of the last data in memory. When this happens, it is detected by a
subroutine which interrupts processing while a new block of File 1 is read
into memory. This time-consuming step is minimized by making the File 1
data block as large as memory permits.
Contingencies--
For samples from File 1 that are just at the zone boundary, there may
not be four valid neighbors surrounding point [E17, N17]. One or more of
the neighbors may exist in File 1 but be at the "no data" flag value of
-9,999. In this case, the program averages whatever "good" values it can
find.
Results--
The UTM zone crossing program can produce seamless resampling to
carry OEMs across a UTM zone boundary. Figure 5 shows a portion of the
resulting UTM Zone 18 data file after Zone 17 data have been resampled into
it. This shows the canyon of the West Branch of the Susquehanna River
where it is crossed by the UTM zone boundary (the 78th meridian). As can
be seen, there are no discontinuities. Figure 6 shows the entire mosaic,
at a smaller scale.
Delivered Products
Initial Algorithm--
Once the DEM mosaic had been completed, analysis of elevation data
was straight-forward. The initial algorithm defined an Index of
Variability for each one square kilometer of the test area as follows:
165
-------
78°W
UTM ZONE 17 UTM ZONE 18
UTM ZONE 17
78° W
UTM ZONE 18
Figure 5. A portion of the resampled digital elevation model mosaic the
The UTM zone boundary (78 degrees west) is shown.
166
-------
Figure 6. Final digital elevation model mosaic.
167
-------
(1) Determine the UTM coordinates of the corners of the cell by
defining a 2-dimensional conformal transformation relating UTM
coordinates of test site corners with a 1 km coordinate system
aligned with the test site edges.
(2) Fill a 41x41 memory array with elevation data for a region
aligned with UTM coordinates that envelops the cell. (This
array size is sufficient to encompass the 1-km portion of the
grid for all possible geometries.)
(3) For every elevation sample in the enveloping region, test
whether that sample is interior to the cell. If so, then
(a) Fetch the elevation of that sample and its four
neighbors. (We assume the enveloping region and the
overall DEM data set is large enough that every point
interior to a cell has four well-defined neighbors and
that edge conditions do not need to be considered.)
(b) If the elevation of the point or of any of its neighbors
is negative, assume this represents missing data. Return
a variability value of -9.99; else
(c) Treat the five elevations as a cross:
v =
El eft
Calculate:
[ 2
[( [Eleft + Eright 1 )
[( [ ------- ....... E] )
Ehigh
E
Elow
[(
Xhor
Eright
2 ] 1/2
( [Ehigh + Elow ] ) ]
( [ - E] ) ]
( [ 2 ] ) ]
( ) ]
( Xhor ) ]
where:
Xhor is the horizontal spacing between samples, 30 meters.
( 2
( [Eleft + Eright - 2E]
= ( t ]
( [ 2Xhor ]
Accumulate a "point-in-cell counter"
2 )l/2
[Ehigh + Elow - 2E] )
[ 1 )
[ 2Xhor ]
(4) Calculate the Index of Variability for the cell to be the
average of all v's from (3c).
168
-------
This algorithm was implemented at EMSL-LV and the calculations were
performed and delivered to the National Oceanic and Atmospheric
Administration, Atmospheric Turbulence and Diffusion Division (NOAA/ATDD).
However, a systematic bias related in some way to individual OEMs
manifested itself. This is illustrated in Figure 7, which plots the
initial Index of Variability for 1 km grid cells. (Because the plot is
aligned with the grid, DEM boundaries appear to the rotated.) Missing
data, of course, are clear. The absence of variability around grid rows 37
to 45 and columns 18 to 27 is not unexpected—this is the Julian quadrangle
digitized from maps at ORSER; it does not have as much detail as the
others. However, all of the western quadrangles and several others clearly
show much lower variability than others. The reason for this is not known,
but it is surmised to result from different DEM production procedures.
Final Algorithms--
The Index of Variability algorithm was revised at NOAA/ATDD. It was
found that two new measures would produce improved indices of complexity.
The first was simply the standard deviation of the terrain height. The
second method calculates the normal to the four adjacent surfaces formed by
connecting lines between a terrain point and the four surrounding points.
The measure of complexity is the average angular difference between the
normals. This measure of complexity was selected because there is a known
problem with
the standard deviation in the case of a uniform inclined plane. In both
cases, however, a threshold could be chosen such that terrain with higher
indices was well correlated with areas of the terrain map which were
subjectively judged to be complex.
LANDSAT THEMATIC MAPPER ANALYSIS
DATA SET
A Landsat TM scene acquired on June 25, 1987 was. purchased from
EOSAT, Inc. About 10 percent of the scene was covered with cloud or cloud
shadow, mostly along the ridges in the southern half of the test area. The
northern extent of the scene began about 15 km south of the north edge of
the test site, due to the inflexible path/row policy which was then in
effect at EOSAT.
The extent of the data set is shown in Figure 8. This is a plot of
TM Channel 5, in the near infrared.
Desired Ground Cover Categories
Modelers and remote sensing specialists met to decide what land cover
categories were needed for the modeling exercise which would be reasonably
practical using remote sensing methodologies. The following categories
were selected:
169
-------
9-
0.00 0.01
grid shows centers of rodm 1 krn grid ce I I s
Index of Variability—Spatial Variability Test Site
(initial algorit hm)
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
Figure 7. Index of variability calculations using initial algorithm.
170
-------
Figure 8. Landsat thematic mapper, June 25, 1987. Channel 5.
171
-------
Agricultural Categories
Corn
Pasture, alfalfa, turf, and other grasses
Fallow land
Forest
Deciduous
Coniferous
Brush and Scrub
Urban, Roads, Pavement
Bare soils
Construction sites
Strip Mines
Rock slides
Water
Other agricultural categories were initially considered, but field
observations and consultations with people at the Pennsylvania State
University revealed that there are almost no crops other than corn and
pasture in the test site.
It was also decided to categorize each of the major categories as
percent-of-cell in one square kilometer cells as the final form of the
data.
Ground Verification Activities
Two hundred and twenty-two observation points were located on
1:250,000 scale maps of the test area. These were chosen to be at
locations which would be identifiable on the TM imagery and which could be
located easily in the field.
Each of these sites was visited during a 2-week period in late August
1987. The field crew consisted of graduate students from the Pennsylvania
State University, working under a subcontract with EMSL-LV, and two of the
authors (Levinson and McMillen). At each site, 35 mm photographs were
taken, usually in several directions. Field observations were recorded in
notebooks. Often sketches were made on 7.5-minute USGS topographic maps
giving approximate locations and sizes of agricultural fields, forest
stands, and of other pertinent information.
Classification
The land cover classification of TM data for this study was performed
using an unsupervised classification approach. A variety of analysis
programs and systems'were used to accomplish this. Some of the analysis
was performed on EMSL-LV's VAX 11/785 computer using ELAS software
developed by NASA at the National Space Technology Laboratory at Bay St.
Louis, Mississippi. Additional analysis used software developed at EMSL-LV
172
-------
operating on both the VAX 11/785 and a Concurrent Computer Corporation
Model 3210 computer, and an ERDAS image analysis system using an IBM PC/AT
computer.
Spectral Channel Selection--
The first step of the classification procedure was to select a subset
of TM data channels to be used for the classification. It was determined
that the thermal channel (TM channel 6) would not be useful for the land
cover analysis. Each of the six remaining TM channels were statistically
compared to each other and analyzed for correlation between channels. The
results of this analysis showed channels 1, 2, and 3 to be highly
correlated with each other but not with the remaining three channels.
Consequently, channels 3, 4, 5, and 7 were selected to perform the land
cover classification.
Unsupervised Classification--
The next procedure in the classification process was to develop a
list of spectral class signatures. This was accomplished using an
unsupervised signature selection algorithm called KLUSTER, developed at
EMSL-LV. This routine compiled a list of statistics describing unique
spectral classes found within the data set. This list consisted of means,
variances, and covariances for each of the four channels for the 63 classes
which were selected. These values were then supplied to a maximum
likelihood classifier which was used to classify each pixel of the data set
into one of the 63 spectral classes defined by KLUSTER. This algorithm
computed the probability of a pixel belonging to each class. The pixel was
then assigned to the class with the highest probability if that value was
greater than or equal to a threshold value set by the analyst. Any pixel
whose highest probability was lower than the threshold was left
unclassified. The result was a data file which contained 63 different
groups or spectral classes of pixels.
Georeferencing--
The next step in the analysis procedure was to geographically
reference the classified data. This procedure resampled the original data
file which was aligned with the satellite's orbital track and thus oriented
at an angle of approximately 13 degrees from north, into a new data file
aligned with UTM ground coordinates (easting and northing).
A total of 30 control points were selected for features, mostly road
intersections and bridges, which could be identified in both the digital
data and on USGS 7.5-minute topographic maps. Three of these were rejected
because of excessive residual error. The remaining 27 were used to
calculate coefficients of bivariate cubic polynomials relating the two
files, using the method of least squares. The cubic polynomials provided
the framework to locate nearest classified data points in the original data
file; these were then placed in the new file. This procedure resulted in a
geometrically corrected digital map of the classified data. The positional
accuracy of these data, as .expressed by the root mean square error, is less
than 12 meters in both the X (easting) and Y (northing) directions.
173
-------
Association of Land Cover Categories With Classes--
The final step in the classification procedure was to assign land
cover categories to classes in the classified data. This was accomplished
with the use of the field information and photographs obtained for the 222
ground verification sites described above. The initial categorization
attempts used multi-colored output plots of the classified data. These
plots were made in sections for the entire study area at an approximate
scale of 1:24,000. Each plot was produced with reference UTM grid tic
marks which were used to overlay and align the plot with each of the
7.5-minute topographic maps used and marked by field personnel. This
allowed the sample point locations to be transferred to the plots. The
field data and photographs were then analyzed for each point while
referring to the plots to determine which spectral class numbers
corresponded to the observed ground cover features. This technique
resulted in the assignment of some land cover information but was unable to
resolve many of the classes.
It was ascertained that some of the confusion stemmed from an
inability to tell the actual appearance of the ground.- This was largely
due to the fact that ground verification data were taken in late August,
near the end of the growing season, while the TM imagery was obtained in
late June, near its beginning.
A common and successful source of verification data for remote
sensing analysis is simultaneous acquisition of aerial photography.
Usually color infrared photographs are most useful because of the
information gleaned from high infrared reflectance of vegetation. In the
absense of such photography, a technique was developed by one of us (01 sen)
to use the satellite data itself as a simulation.
This technique made use of EMSL-LV's interactive image display
capabilities. It involved simultaneously displaying a color infrared
simulation of both the original TM data and the classified data on two
different image display devices located side-by-side. The color infrared
simulation of the raw data was accomplished using the three channel display
capability of the IBM PC/AT-based ERDAS image processing workstation.
Channels 3, 4, and 5 were read into the image memories of the ERDAS system
and displayed through the blue, green and red color guns, respectively.
This resulted in an image which had the appearance of a color infrared
photograph.
Another color infrared simulation was produced for the classified
data using a second display device and another computer. The relative
display intensity of red, green, and blue was computed from the Channels 3,
4, and 5 mean values generated by KLUSTER for each spectral class. The
equation used for this computation is listed below:
174
-------
[ (X)i - (Xmin)i]
I = 15 x [ ]
[(Xmax)i - (Xmin)i]
Where:
I = relative intensity (integer value between 0 and 15)
(X)i = class mean value for band i
(Xmin)i = minimum mean value for band i
(Xmax)i = maximum mean value for band i
i = band number (3, 4, or 5 in this case)
These values were used as the color display intensities for the blue,
green, and red color guns of the color monitor.
With the two simulations displayed side-by-side, the area surrounding
each of the training sites was displayed and the cursor of each device was
positioned to the UTM coordinate of that site. The color infrared display
of the raw data was used to identify ground cover features found in the
review of field data sheets, maps, and photographs. The color IR
simulation of the classified data was then compared to the display of the
raw data to discriminate the same features. The class values for those
features were read from the display device. Observations of the feature's
color in the simulated IR image were also made and noted.
This was used to help determine the state of growth for many of the
vegetation categories. In particular, it explained the confusion found in
the classified data between areas which were observed to be corn fields and
roads. In fact, in late June (at the time of satellite data acquisition)
corn plants were still small enough that many of the corn fields appeared
as dominantly bare soil and therefore were easily confused with other
highly reflective surfaces such as roads and barren lands. This
explanation was readily determined through the interpretation of the color
IR display.
Final Categorization
Discrimination between coniferous and deciduous forest proved
elusive. There are in fact few conifers in the test site, and separating
them from nearby deciduous trees was not possible. Differences in forest
classes appear to be due to slope, aspect, and scene illumination.
The final categorization provided data, as percent of one
square kilometer cell, for the eight categories shown in Table 3.
175
-------
TABLE 3. RESULTS OF LANDSAT TM CLASSIFICATION
Category Description Relative Frequency
0
1
2
3
4
5
6
7
Figures
categories for
Unclassified (including clouds)
Cloud Shadow
Deciduous Forest
Brush and Scrub
Corn
Grass and Forage Crops
Roads, Urban, Bare Soils
Water
9 and 10 show percent-of-cel 1 plots of each
the 83x83 1-km grid cells.
8.4 %
5.9 %
40.5 %
9.7 %
12.9 %
19.0 %
3.4 %
0.2 %
of the eight
Accuracy Assessment
One method to test whether an image map is of acceptable accuracy is
to select samples of map points, check the map classification against
ground data, and then make a statement about the true accuracy of the map.
Such a statement generally claims some minimum level of accuracy with some
high level of confidence, e.g., a minimum of 85 percent accuracy at the 95
percent confidence level. The sampling problem is then one of determining
the number (N) of map samples to be compared with the ground data, and an
allowable number of misclassifications (X) of these samples. After the
samples are determined, N map samples are selected and their
classifications are compared against the true field data. If X or fewer
points are misclassified, then the map is accepted as accurate at the
specified level of precision.
In any statistical test there is a probability or risk that
interpretation of the test results will lead to the wrong conclusion. The
probabilities associated with the two types of erroneous conclusions may be
termed Consumer Risk and Producer Risk. Aronoff (5) showed that the
consumer and producer risk could be correlated with traditional Type I and
II statistical error. Stated in another way, Consumer Risk is the
probability that a map of unacceptable accuracy will pass the accuracy
test; while the Producer Risk is the probability that a map of some
acceptable accuracy will be rejected (6).
The Spatial Analysis Laboratory (SAL) at EMSL-LV employs the Minimum
Accuracy Value (MAV)6 test for the accuracy assessment of all remote
sensing products. The MAV is the highest accuracy level for which the
observed number of misclassifications would constitute passing the accuracy
test at the user-specified consumer risk. In addition to reporting whether
a thematic map had passed or failed a specific accuracy test, a minimum
accuracy value should be calculated for the map and represent a statistical
measure of quality.
176
-------
„ -
»"
t '
,1-r*"
i > ** - •. • f ,
Cen ters of 1 KM Grid CeI Is
CATEGORY 0 (UNCLASS'FI ED)
Percent of Ce I I
MEfrftii50 60 '
100
s s s s s s
Centers of 1 KM Gr id CeIIs
CATEGORY 2 (FOREST)
Percent of CeI I
10 20 30 40 SO 60 7C
100
S S ? 8 S
i
Centers of 1 KM Grid Ce I Is
CATEGORY 1 (CLOUD SHADOW)
Percent of Ce I I
0 10 20
S % 3 $ $
i ., i l l
-
?
^j- 'ji'r
V-.,.'
It- T
J * .•& .:-
* i
Centers of 1 KM Gr id CeI Is
CATEGORY 3 (SCRUB/BRUSH)
-<••'
S 8
00
Figure 9. Percent-of-cell classifications: Categories 0 to 3.
177
-------
i r
S S S S S £
Centers of 1 KM Grid CeI I a
CATEGORY 4 (CORN)
Percent of CeI I
30 40 50 60 70
I I I 1 I
2 s a ? a 8 s
Centers of 1 KM GrId CeI Is
CATEGORY 6 (UR8AN/ROADS/BARE SOIL)
Percent of CeI I
100
100
3 8 8
i i i_
; % ' FujlF
:* *:'. v
£^/" -> -v
A' '- „••< «ft
C? t-1"
- > * - •!
i i r
g s s s s s
Centers of 1 KM Gr id Cei Is
CATEGORV 5 (GRASS/PASTURE)
Pe-cent of Ce!I
0 10
100
70-
50-
30-
20-
10-
sssssseg
i i i i i i i i
>*
1 1 1 ! 1 1 1 1
OOOOOOOQ
f-Nn*m9*m
^70
-50
-30
-20
-10
Centers of 1 KM Gr i d Ce I I s
CATEGORY 7 (WATER)
Percent of Ce I I
(
) 10 20 30 40 50 60 70 80 100
i - ^'^•{•^••••i
Figure 10. Percent-of-cell classifications: Categories 4 to 7.
178
-------
Given this background and the needs of the Spatial Variability
project, it was determined that a Consumer Risk of 10 percent or less with
a minimum accuracy level of 90 percent would be the objective. Using
values provided by Aronoff (6) the actual map accuracy was computed as:
Target Accuracy: 90.0%
Consumer Risk: 10.0%
Producer Risk: 10.1%
Number of Points Sampled (N): 175 Allowable
Misclassifications (X): 12
Observed Misclassifications: 15
Actual Map Accuracy: 88%
Even though the actual map accuracy fell short of the target accuracy
value, it was deemed not worth the effort to reclassify the map to achieve
only a 2-percent gain in accuracy. In fact, a reclassification effort
could raise the Consumer Risk value to an unacceptable level.
RESULTS
At EMSL-LV, this project has resulted in the development of several
capabilities: Multiple Digital Elevation Models can now be mosaiced
easily. OEMs can be be carried across UTM zone boundaries. This is a
capability that no one had before and it solves a problem that will
repeatedly occur as DEM mosaics become common. The use of simultaneous
simulated color IR from raw and classified Landsat data provides a new and
novel way to use Landsat as its own ground truth.
At NOAA/ATDD, terrain and land cover variability data sets are now
being evaluated. They have been incorporated into the ATDD model, and
although the results are at present very sketchy, they are adequate to
demonstrate the validity of the approach. Techniques have been developed
to successfully integrate data at high resolution (30 meters) into a 1-km
grid framework. This technique seems to provide a satisfactory measure of
complexity.
CONCLUSIONS
This project has demonstrated that terrain ?nd land cover can be
analytically manipulated to provide useful data for the deposition velocity
model. This is a new type of application of traditional remote sensing and
image processing methods.
179
-------
REFERENCES
1. U.S.G.S National Cartographic Information Center. Digital
Cartographic and Geographic Data. Pamphlet. 1985.
2. Elassel, A. A., and V. M. Caruso. Digital Elevation Models. In:
U.S.G.S. Circular 895-B, 1983.
3. Freden, S. C., and F. Gordon. Landsat Satellites. Chapter 12 of
Manual of Remote Sensing, 2nd Edition, Volume 1, Amer. Soc. of
Photogrammetry, 1983.
4. Snyder, J. P. Map Projections, A Working Manual. U.S.G.S.
Professional Paper 1395, 1987.
5. Aronoff, S. Classification Accuracy: A User's Approach. Photo.
Eng. and Remote Sensing, 48(8):1299-1307, 1982.
6. Aronoff, S. The Minimum Accuracy Value as an Index of Classification
Accuracy. Photo. Eng. and Remote Sensing, 51(1):99-11, 1985.
180
------- |