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

-------