United States
Environmental Protection
Agency
Environmental Monitoring
Systems Laboratory
P.O. Box 15027
Las Vegas NV 89114
July
Research and Development
Landsat Estuarine
Water Quality Assess-
ment of Silviculture
and Dredging Activities
-------
LANDSAT ESTUARINE WATER QUALITY ASSESSMENT
OF SILVICULTURE AND DREDGING ACTIVITIES
by
John M. Hill
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada 89114
Project Officer
John A. Eckert
Advanced Monitoring Systems Division
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada 89114
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA—89114
-------
DISCLAIMER
This report has been reviewed by the Environmental Monitoring Systems
Laboratory-Las Vegas, U.S. Environmental Protection Agency, and approved for
publication. Mention of trade names or commercial products does not consti-
tute endorsement or recommendation for use.
ii
-------
FOREWORD
Protection of the environment requires effective regulatory actions
based on sound technical and scientific data. The data must include the
quantitative description and linking of pollutant sources, transport
mechanisms, interactions, and resulting effects on man and his environment.
Because of the complexities involved, assessment of exposure to specific
pollutants in the environment requires a total systems approach that
transcends the media of air, water, and land. The Environmental Monitoring
Systems Laboratory at Las Vegas contributes to the formation and enhancement
of a sound monitoring-data base for exposure assessment through programs
designed to:
t develop and optimize systems and strategies for moni-
toring pollutants and their impact on the environment
t demonstrate new monitoring systems and technologies
by applying them to fulfill special monitoring needs
of the Agency's operating programs
This report describes the application of Landsat multispectral scanning to
estuarine water quality with specific reference to dredging and silviculture
practices. The synoptic view, orbital characteristics, and remote sensors of
Landsat make it a valuable tool for the assessment and monitoring of
environmental quality parameters in and around estuaries.
Environmental Monitoring Systems Laboratory
Las Vegas
111
-------
,/
ABSTRACT.
The objective of this project was to define and demonstrate the role of
Landsat multispectral scanner (MSS) data in monitoring environmental impact in
estuaries ~nd their associated drainage basins. Florida State University
(FSU) has been collecting water quality data from Apalachicola Bay, Florida
(East Bay in particul arr almost biweekly since 1972. Landsat-1was 1 aunched
in 1972, thereby creating an invaluable and almost irreplaceable data base for
correlative land-use/water quality investigations. .
Landsat-1 data were examined to determine their use in detecting and
delineating temporal distributions of water color and land-use categories in
the bay system. Water color distributions were easily discriminated. Landsat
data were successfully used to monitor the effects of man-made and natural'
structures (i.e., holes in barrier islands, channels, islands (bridges), and
oyster bars) on the hydrodynamics of the bay. The distributions of water
types {i.e., acidic swamp/forest runoff (a form of nonpoint source pollution),
turbid river water, clear Gulf water) were monitored under numerous.
environmental 'conditions. Land-use categories, with an emphasis on
silviculture activities (i .e., clear-cutting, pine plantations, swamp/forest
communities) were also detected and delineated. The primary source of.
nonpoint source pollution in East Bay appears to be the lumber industry.
Trends in Landsat derived land-use activities followed trends in the
improvement .(recovery) of water quality in East Bay. The collection of tandem
surface truth data would have greatly improved the quantitative aspects of
this project. The use of Landsat data proved to be a more accurate method to
monitor water color distributions and associated water patterns than the
.presently accepted traditional water sampling schemes. However, the most
accurate presentation of water patterns would be derived if tandem water
quality samples were required at the same time as the satellite overpass.
These water quality data could be used as calibration points during the
classification stages of the Landsat imagery.
The information derived from this research is being used by the University
of Florida in the planning stages of their water quality sampling program and
will eventually be used to help construct and validate hydrodynamic models of
the East Bay drainage basin and Apalachicola Bay. . The transfer of this
technology to other estuaries around the nation (i .e.,' Lake Pontchartrain,
Louisiana and the Chesapeake Bay, Maryland) should be encouraged.
This report covers the period from July 1,1977 to July 21, 1978, when the
project was completed.
iv
-------
CONTENTS
Fore'll()rd. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Abs tract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fig ure s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tables. . . . . . . . . . . ... . . . . . . . . . . . . . . . . . . .
List of Abbreviations and Symbols. . . . . . . . . . . . . . . . . .
Acknowl edgment. . . . . . . . . . . . . . . . . . . . . . . . . . . .
l.
2.
3.
4.
Introduction. . . . . . . . . . . . . . . . . . . . . . . . .
Conclusions. . . . . . . . . . . . . . . . . . . . . . . . .
Recommendations. . . . . . . . . . . . . . . . . . . . . . .
Background. . . . . . . . . . . . . . . . . . . . . . . . . .
Monitoring turbidity qualitatively with Landsat. . . . . .
Monitoring turbidity quantitatively w;thLandsat. . . . . .
Geographic Study Area. . . . . . . . . . . . . . . . . . . .
Materials and Methods. . . . . . . . . . . . . . . . . . . .
Re s u 1 t s . . . . . . . . . . . . . . . . . . . . . . . . . . .
Temporal water color (type) distributions. . . . . . . . .
Discrimination of land-use activities. . . . . . . . . . .
Temporal land-use distributions. . . . . . . . . . . . . .
Signature extension. . . . . . . . . . . . . . . . . . . .
Reliability of classification. . . . . . . . . . . . . . .
5.
6.
7.
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ap'pe nd; x. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
v
Page
i i ;
iv
vi
x
xi
xii
1
4
5
7
16
17
24
37
51
62
75
87
89
91
10r
. 108
-------
Number
10.
11.
12.
Pathways of various components of light from water
that is received by a multispectral scanner. . . . . . . . .
Percent refl ectance for water with red sed iment. . .
FIGURES
1.
2.
3.
,
Percent re fl ectance for water wi th black sed iment
(Rio Grande Valley and Texas Blackland). . . . .
4.
Percent reflectance for pond water with algae
(Chickashe, Oklahoma). . . . . . . . . . . . . .
5.
Calibration curve for sediment load versus film
density in median California coastal water.
. . . . . . . .
6.
Comparison of radiance signatures of York and
Rappahannock Rivers. . . . . . . . . . . . .
. . . .
7.
Total particles versus image brightness of accurately
processed positive composites of Landsat Channels
2 and 3) . . . . . . . . . . . . . . . . . . . . . .
8.
The Apalachicola Drainage system including
the major rivers that contribute to the
Apalachicola Bay system. . . . . . . . . . . .
U.S. Geological Survey map of Apalachicola Bay,
Florida and vicinity. . . . . . . . . . . . .
. . . .., .
. . . . . .
. . . . . . 15
. . . . 20
. . . . 22
Page
10
13
14
19
. . .. . . . 26
9.
. . . . . . . 27
Landsat image (1516-15421) of Apalachtcola Bay,
Florida and vicinity with area of interest
outl i ned . . . . . . . . . . . . . . . . . . .
. 0 . . . . .
Classification of the harvesting of shellfish in
Apalachicola Bay (Area A=prohibited; Area.
B=conditionally approved. . . . . . . . . . .
Landsat MSS scanning arrangement.
. . . . . . .
. . . . . 0 . .
vi
......
28
31
38
-------
Number
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
FIGURES (Continued)
Landsat ground coverage pattern. .
. . . . . . . .
.....
Components of the EPA Data Analysi sSystem. .
. . . . . . . .
Simplified flow diagram of the functional steps' of
Landsat digital analysis. . . . . . . . . . . .
. . . . .
Map of FSU water quality sampling stations in the
Apalachicola Bay system. . . . . . . . . . . . . .
Standard water color table for classified images..
. . . .
. . . .
Computer-derived, four channel, spectral plots of
swamp/forest runoff training fiel ds, Image 5 . . ~ . . . .
Computer-derived, four channel, spectral plots of
diluted swamp/forest runoff training fields,
1m age 5. . . . . . . . . . . . . . . . . . . . .
. . . . .
Computer-derived, four channel, spectral plots of
highly turbid Apalachicola River plume training
fields, Image 5. . . . . . . . . . . . . . . . .
. . . . .
Computer-derived, four channel, spectral plots of
a type of turbid Apalachicola Bay water
training fields,' Image 5 . . . . . . . . . . . .
. . . . .
Computer-derived, four channel, spectral plots of
a type of moderately turbid St. George
Sound water training fields, Image 5 . . . . . . .
Computer-derived, four channel, spettral plots of
shallow turbid Gulf water training fields, Image 5
. . . .
. . . .
Computer-derived, four channel, spectral plots of
clear, deep Gulf water training fields, Image 5.
Classifted Landsat image (17 February 1973) of
water color distributions in Apalachicola Bay,
Florida under low wind and ebb tide conditions
. . . . .
. . . . . .
Classified Landsat image (13 April 1973) of
water color distributions in Apalachicola Bay,
Florida under low wind and .flood tide conditions
.....
vii
Page
39
41
42
50
53
55
56
57
58
59
60
A1
64
67
-------
Number
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
FIGURES (Continued)
Classified Landsat image (21 December 1973) of
water color distributions in Apalachicola Bay,
Florida under moderate northeast wind and
probable ebb tide conditions. . . . . . . . .
Classified Landsat image (3 March 1974) of
water color distributions in Apalachicola Bay,
Florida under low wind and beginning flood
tide conditions. . . . . .. . . . . . . . . .
Classified Landsat image (23 October 1974) of
water color distributions in ApalachicolaBay,
Florida under strong northeast wi nd conditions
Classified Landsat image (26 February 1975) of
water color distributions in' Apalachicola Bay,
Florida under low wind and near ebb slack
tide condition~. . . . . . . . . .. . . . . .
Classified Landsat image (20 July 1975) of
watercolor distributions in Apalachicola Bay,
Florida under much less than optimal atmospheric
conditions. . . . . . . . . . . . . . . . .. .
Classified Landsat image (19 August 1976) of
water color distributions in Apalachicola Bay,
Florida under strong northeast wind conditions.
Classified Landsat images of land-use activities
in East Bay drainage basin (A-17 February 1973;
B-13 April 1973; C-21 December 1973;
0-23 October 1974) . . . . . . . . . . . . . . .
Classified Landsat images of land-use activities
in East Bay drainage basin (E-26 February 1975;
F~19 August 1976). . . . . . . . . . . . . . . .
. . . . . .
. . 0 . D .
......
. . . . . .
000.0
. 0 . . .
. . . . .
. . . . .
Computer-derived, four channel, spectral plots of all
marsh training fields from Image 5 . . . . . . . . . .'. .
Computer-derived, four channel, spectral plots of
all swamp/forest training fields from Image 5. .
37. . Computer-derived, four channel, spectral plots of
all sand, mudbank, and road training fields from
Image 5. . . . . . . . . . . . .
.8 . 0 . . . . .
viii
.....
.....
Page
69
70
72
73
.76
77
79
80
82
83
84
-------
Number
38.
39.
40.
41.
42.
43.
44.
FIGURES (Continued)
Stages of clear-cutting and vegetation. . . . . .
.....
Computer-derived, four channel, spectral plots of
all cut and revegetated training fields
from Image 5 . . . . . . . . . . . . . . . . . .
.....
Computer-derived SCORECARD for Image 5
(21 December 1973) indicating the percentage
of original training field pixels assigned to
each of the classes under consideration. . . . . . . . . .
pH versus color of surface water in A'pafckhicol a Bay, .
Florida on 17 March 1976 . . . . . . . . . . . . . . . . .
Dissolved oxygen versus color of surface water in
Apalachicola Bay, Florida on 17 March 1976 . . .
. . . . .
pH distribution from simple linear extrapolations
of water quality data, February 1975 . . . . . .
. . '. . .
Water color map from simple linear extrapolations
of water quality data, February 1975 . . . . . .
. . . . .
ix
Paae .
-
86
92
93
98
98
99
100
-------
Number
TABLES
1.
Wavelength of maximum transmission and the
percentage of light of this color transmitted
in five different types of ocean. . . . . . .
2.
Summary of preferred Landsat spectral channels
(1, 2, 3 and 4) for turbidity related results
derived from the literature by number of projects.
List of Landsat computer compatible tapes
received for the Apalachicola Bay Project. . . .
3.
4.
Landsat four channel water class count means from
all classified images. -Channels 1, 2, and 3 are
on a compressed count value scale of 0 to 127 and
channel 4 ranges from 0 to 64. . . . . . . . . . .
Average monthl y f1 ows for the Apa.l achicol a
River at Blountsto~m for period 1961-1976. .
s.
6.
Percent and hectares (acreage) of swamp/forest and
dilute runoff for all images in East Bay,
Apalachicola Bay, and St. George Sound. . . . .
Landsat four channel land class means. (in counts)
from all classified images. . . . . . . . . . .
7.
8.
Percentages and hectares (acreage) of subminor,
minor and major watersheds in East Bay
drainage basin. . . . . . . . . . . . . . . . .
9.
Percentages and amounts of Landsat and lumber
company-derived clear-cutting data from East Bay
drainage basin (32,645 hectares) . . . . . . . .
x
. . . . . . .
. . . . .
. . . . . .
0....
. 0 . . . . 0 .
. . . . . .
. . . . . .
......
. . . . . .
Page
11
23
40
62
65
66
78
90
96
-------
ACSC
APHA
ASP
CCT
cfs
CPU
DAS
DCS
DRI
EEL
EPA
FSU
FWPCA
Ha
10
IPA
JTU
m
m3/s
mg/l
MPN
MSS
N
NASA
NEPA
nm
NPS
RBV
UF
~g/l
USGS
USPHS
U1M
W
0/00
LIST OF ABBREVIA TrONS Arm SYMBOLS
Area of Critical State Concern
American Public Health Association
American Society of Photogrammetry
Computer Cornpat i b 1 e Ta pe
cubic feet per second
Cobalt Platinum Units
Data Analysi s System
Data Collection System .
Developments of Regional Impact
Environmentally Endangered Lands
Environmental Protection Agency
Florida State University
Federal Water Pollution Control Act
hectares
Identification
Interagency Personnel Act
Jackson Turbidity Units
meters
cubic meters per second
milligrams per liter
Most Probable Number
Multispectral Scanner
North
National Aeronautics and Space Administration
National Environmental Pol icy Act.
nanometers
Nonpoint Source Pollution
Return Beam Vi di con
University of Florida
micrograms per liter
United States Geological Survey
United States Public Health Service
Un iversal Transverse Mercator
West
parts per thousand
xi
-------
ACKNOWLEDGMENTS
I wish to express my appreciation to Mr. John D. Koutsandreas (EPA), Mr.
Gene Coker (EPA), and Dr. Robert J. Livingston (FSU) for conceiving and,
encouraging this research. I ~sh to sincerely thank Dr. Merrill H. Sweet,
Dr. Leo Berner, Jr., Dr. John W. Rouse, Jr., Dr. William J. Clark, and Dr.
George L. Huebner, for their guidance during this research and for suggestions
on the publication of results. I owe particular thanks to Dr. Bruce J.
Blanchard and Dr. Richard W. Newton for the support and use of facilities at
the Remote Sensing Center, Texas A&M University during preparation' of this
report. I thank Mr. Steve Graham for his assistance in the interpretation of
data related to estuarine dynamics. The assistance and interest in the.
environment of Mr. Mark Starnes, Mr. John Curry, Buckeye Cellulose
Corporation, is sincerely appreciated.
This project was supported by the EPA's Environmental Monitoring Systems
Laboratory, Las Vegas, Nevada, under an Interagency Personnel Act (IPA)
appointment with the Remote Sensing Center of Texas A&M University, College
Station, Texas.
xii
-------
SECTION 1
I NTRODUC TI ON
Estuaries are an extremely valuable natural resource of national
importance. Modi fi ed from Pritchard (1967), an estuary "i s a semi -enclosed
coastal body of water which has a free connection with the open sea; it is
thus strongly affected by tidal action, and within it sea water is mixed (and
usually measurably diluted) with fresh water from land drainage" (Odum 1971).
Odum (1971) further states that estuaries are thought of as being transition
zones between fresh and marine habitats, but many of their most important
physical and biological attributes are not transitional, but unique. The uses
and abuses of estuarine systems by man need to be understood, monitored, and
wisely managed. Estuaries serve as nursery grounds for they provide
protection and abundant food. The preservation of estuaries could thusly be
justified solely on the basis of the numerous commercial and sport fisheries
that exist, at least during portions of their life cycles, in these areas. .
Man is encroaching on those systems that are generally more productive than
either the fresh-water drai nage or the sea (Odum 1971).
The multiple use effects of human beings in most cases inevitably decrease
the water quality along the coastal zone. Due to unsupervised development,
prime estuaries have been destroyed and closed due to potential health
hazards. There is no doubt that human interaction and settlement has
generally focused near estuaries. Pollution related to urbanization, sewage,
sedimentation, thermal and chemical contamination, construction changes
resulting in hydrodynamic modification (i.e., bulkheading, dredging, filling),
and land-use (watershed) activities generating nonpo;nt sources of pollution
are present in and around the coastal zone. Estuaries can no longer be
considered as a sump for societies' wastes. The "pressure" is on the nation's
estuaries and, as the impact spreads, the repercussions increase. The added
rapid and apparently unending increase in a search for energy sources wi 11
also continue to impact the estuaries. Rarely is any consideration made as to
what gives present day man the right to deny the beauties and benefits of
estuaries to future generations.
Institutions and agencies involved in the study and management of the
nation's coastal zone are faced with a large-scale monitoring problem. The
decl i ne in quantity and qual ity of estuari ne resources necessitates the.
development of best management practices for these resources. Scientists .and
management need timely, accurate inventories of impacts on the estuarine
environment. These data have been traditionally collected through more time
consuming, expensive and often inacturate traditional sampling procedures,
primarily field crews in boats and on land. Under present funding and
1
-------
.staffing levels, it is impossible to rapidly and accurately monitor and access
large coastal areas. Remote sensing techniques can often be applied to the
needs of 1 arge scale moni tori ng programs i n coastal envi ronments. Remote
sensing techniques offer many advantages over traditional sampling methods.
These advantages include a synoptic view; an expansion of the spectral range
and resolution of the human eye; a permanent and instantaneous record of a
particular point in time; a quantitative capability to determine size, shape,
and position of objects due to the precise geometry and digitization of some
images; a near real-time capability; and an inexpensive monitoring technique
by compari son with more traditional sampl i ng schemes.
The objectives of this research program were to (I) evaluate the
. capability of landsat multispectral scanner data to monitor water color
distributions, which in turn can also represent baseline coastal environmental
data, (2) identify and acquire temporal land-use information, with an emphasis
on forest management practices, which have a high probability of being related
to the amount and distribution of nonpoint source runoff and associated water
quality in a bay system, (3) determine if there is a relationship between land
use and water quality in the bay system, and (4) recommend how the results of
this study can be applied to aid various local, state, and federal agencies in
the development of monitoring programs leading to the establishment of best
management practices, including predictive environmental modeling, for
estuarine resources.
Not until the 1960's did Florida begin to realize that it had sacrificed
many of its natural marine resources while developing coastal resources.
largely because of unsupervised development, many prime estuaries have either
been destroyed due to siltation, pollution, or filling, or have been closed to
the activities of man due to potential health hazards. Apalachicola Bay,
Florida, being representative of lagoon type barrier island systems from Maine
to Texas, was selected as the test site for this research. Apalachicola Bay, .
located in Florida's "panhandle," is one of the most pristine and economically
valuable estuaries in the state. Biotic communities in and around the bay,
having survived the rigors of centuries of variation in environmental
parameters, are now being threatened. The primary land-use activity in the
basin is silviculture. The major problems within the bay are related to
transportation features such as bridges and related islands, intercoastal
waterways, and cuts in barrier islands.
Starting in March of 1972, FSU conducted a project to determine important
relationships between the Apalachicola River and the bay system for the
purpose of utilizing this information as a basis for regional planning and
resource management. The project was generated in response to numerous upland
developments in the Apalachicola Drainage System. The project, under the
direction of Dr. Robert J. livingston(FSU}, can be divided into two portions:
(I) biological studies to determine certain fundamental trophic relationships
of the bay system and its interaction with the river, and (2) an impact
analysis to be made based on short- and long-term floristic and faunal changes
in the bay~ Landsat-l was also launched in 1972, thereby creating an
excellent data base to monitoring synoptic, temporal changes in land-use
2
-------
and water color distribution in the bay system. These two data bases also
provide an excellent opportunity to compare the accuracy of data derived from
traditional water-type distribution and land-use sampling procedures with that
of landsat-derived data.
'\'.."
'r . ',~/~?!.J:~:~~
r. . .,~ ........,t
.-" -~ 1
."",,\.,.
, , '"~
.-,..: ./", ".f
~ -.,,-.
{. . ~ ~.;~ ~
. ,';'. 4;f,<"'.,
..:.-.;....\.
3
-------
SECTION 2
CONCLUSIONS
This research demonstrated that landsat can provide temporal cause and
effect information relating to land-use changes and water quality. Water
types, based on water color, were easily discriminated at different times of
the year and under varying environmental conditions (i.e., winds, tides, river
stages).
Water patterns, which ~re nearly impossible to acquire under more
traditional sampling schemes, were readily discerned. Inferences as to
current direction, offshore water movements including reversals, location of
gyres and areas of mixing, a real extent of river plumes, and the sources and
dispersal sites of sediment and nonpoint source pollution (swamp and forest
runoff) are possible. These patterns were also used to observe the effect of
manmade structures, such as channels, islands and cuts in barrier islands, on
current patterns in the bay. Commercial and sport fisheries could also use
these seasonal landsat images to effectively locate schools of fish under
varying environmental conditions, if such data were ever available on a
real-time basis.
Probably the greatest value of these landsat-derived distributions of
water classes is their use to obtain a seasonal overview of an area to be
investigated. This would assist in the most advantageous placement of
sampling stations, which would in turn provide the capability to correctly and
more accurately extrapolate data spatially over large areas from a minimal
number of sampling points. While only partially confirmed, apparent trends in
land-use changes (forest resource management practices) and related water
quality were obtained. In the past, ecologists have literally had the problem
of "not being able to see the forest for the trees." landsat, with its
synoptic characteristics, is an excellent way to view an entire ecosystem.
Regulatory and management agencies can use Landsat data to plan and initiate
land-use management activities that best take bay communities into
consideration. The water data can be used as a tool in the selection of the
most environmentally beneficial sites for such facilities as drainage canals,
pipelines channels, housing, industrial and recreational facilities, and power
and sewage plants. .
4
-------
SECTION 3
RECOMMENDATIONS
Several recommendations are advanced. It is believed that a more accurate
land-use classification could be achieved if the research had been finely
tuned by using Landsat scenes that were closer together (i.e., 1 to 2 months
apart). These additional scenes were not acquired due to a combination of
funding and time limitations and an insufficient knowledge of the seasonal
spectral signatures of target features at the start of the project. The
acquisition of tandem aerial photographs and ground truth data at the time of
a satellite overpass would also increase the classification of land-use
practices.
Future Landsat-derived water color distributions would also be enhanced
with the classification of additional Landsat scenes. More scenes would'
substantiate and even delineate more hydrodynamic features that exist in. the
bay under varying environmental c~nditions (i.e., tides, winds, river flow)
. that may have not been in effect. during this investigation.
The major difficulty with respect to the identification of the same class
in different scenes is the dynamic nature of the environment. Spectral
signatures change with varying atmospheric conditions. Future Landsat
investigations should include the acquisition of surface truth in the fonm of
spectral signatures of specific land and/or water classes. This information
could then be used to establish correction factors for such parameters as sea
state, sun angle, and atmospheric haze. These parameters were not considered
in this particular research effort, primarily because of computer software
limitations. Several of these correction software packages exist at numerous
digital data processing institutions around the country and should be acquired
and incorporated into EPA's Digital Analysis System (DAS). The use of these
correction factors in combination with ground truth samples collected at the
time of the satellite overpass would greatly improve the quantitative accuracy
of the Landsat classification of water types and land-use categories.
Correlations between Landsat data and other water quality parameters (i.e.,
suspended solids, turbidity, pH, and dissolved oxygen) could also be evaluated
using tandem water quality data and the above mentioned correction procedures.
This research should also be expanded to derive correlations between basin
land-use practices and water quality (nonpoint source pOllution) in the
neighboring estuary. This phase may be accomplished if Landsat-derived data
can be incorporated into predictive environmental modeling schemes.
5
.,.
:. l~:' .~J
A.~
.t..-
. .-., ..
..If- ....
"":.0;"
. . """,
. ' . 'f '. .....~
,~:.tŁ;
.;.: ~
-------
Since silviculture practices were easily identified and delineated, future
studies should include the development of change detection algorithms -- that
is to say, programs that wi 11 enable the investigator to quanti fy the amount
of land-use change between various images.
These data could also be used to solve problems presently plaguing
environmental modelers. It has not only been extremely difficult to acquire
data as input to produce an accurate predictive enviromental quality model; .
most models are never verified. Landsat data are grided, synoptic, and
instantaneous and are, therefore, amenable as input to both the initial
development and final verification stages of so-called real-time water quality
and hydrodynamic numerical models (Graham et al. 1978b). These land-use
results for portions of the New River are being used by the University of
Florida (UF) to calibrate an existing basin water quality model. Once the
model is working, Landsat-derived land-use data from Tate's Hell Swamp could
and should be used as input to the basin model, which will in turn run (drive)
a hydraulic model of Apalachicola Bay. .
. UF is presently refining a hydrodynamic model of the Bay. These landsat
derived water color distributions should be used in two phases of this bay
modeling effort. First, the water boundary information should be reformated
and used as input data to refine and construct the model. Second, the model
should be verified with the landsat data. If the model were run for the same
time as a satellite overpass, the model ouput and the landsat-derived water
color distributions could be scaled and overlayed for comparison purposes. If
the two. agree, the model is to a large degree verified. landsat data have
. been used for models related to land use, but this is apparently the first
attempt to use such data to verify a hydraulic model of an estuary. It is
highly recommended that this particular phase of this research, namely model
construction and verification, be continued.
Landsat has been demonstrated to be of use in the detection and
quantification of possible causes and effects of large-scale environmental
degradation in a coastal area. Once detected, higher resolution techniques
(i.e., high resolution photography, ground truth teams) should be used to
study particular aspects of the problem in greater detail. Lastly, since
Apalachicola Bay ;s fairly representative of most lagoon type estaur;es along
the entire Gulf and East Coasts of the United States, the use of landsat can
and should 'be extended to similar environmental areas around the nation. It
is reasonable to assume that this technique could be used at least throughout
the entire Southeast and most probably in areas along the East Coast (i .e.,
the Chesapeake Bay, Maryland).
6
-------
SECTION 4
BACKGROUND
Just a little more than a century ago investigators interested in
monitoring the earth's resources had to rely entirely upon direct
on-the-ground observation to collect the desired information, (American
Society of Photogrametry (ASP) 1975b). Managers of coastal resources, because
of the difficult nature of the water medium, have often had to utilize hard to
acquire and increasingly expensive ships to collect their data. Research
vessels can cost $5,000 per day to operate. Other problems arise when water
samples are collected over large areas. Because water is such a dynamic
medium, itis often physically impossible to collect adequate water quality
data in a sufficiently short period of time. When one ship is utilized to
collect data, e.g., from a river plume entering an estuary, the cruise track
followed often does not provide the necessary data needed to contour and study
various water boundaries, because the sampling stations were located without a
synoptic knowledge of the area being sampled. Coastal resource managers are
continually looking for new techniques which can help them economically
collect meaningful data. .
Whether observed on the local, regional, national or global scale, the
human demand on most of the earth's resources is rapidly depleating these
resources both in terms of quantity and quality. The situation necessitates
the development of best management practices for these resources. Management
must be provided with timely, accurate inventories so that it will have an
. accurate up-to-date knowl edge on the. amount and qual ity of each resource
within a particular geographic area of responsibility. Almost invariably such
inventories can best be accomplished by the utilization of various remote
sensing techniques (i.e., by obtaining photography and related data from
aircraft and/or spacecraft during periodic overpasses) (ASP 1975b).
Remote Sensing is, as defined in the recent Manual of Remote Sensinq
(ASP 1975a),
...the measurement or acquisition of information of some property
of an object or phenomenon, by a recording device that is not in
physical or intimate contact with the object or phenomenon under
study, e.g. the utilization at a distance (as from aircraft,
spacecraft or ship) of any device and its attendant display for
gathering info nnat ; on pertinent to the envi ro nment, such as
measurements of force fields, electromagnetic radiation, or
acoustic energy.
7
..".
" ..";-..
--
.
'"
: ,
.'f04'"
,,,
....... ~
. ,.~. "
.~ :. ~
". ~
. . .."
.... .~:.:
~
-------
Remote sensing systems, borne by satellites and aircraft offer investigators
many advantages over traditional sampl i ng procedures. The advantages as
outlined by James and Shwartz (1972) include:
1.
Synoptic overview. An observer on a ship or on the earth's surface
is very limited in what he can see about the surrounding area.
Satellite or aircraft acquired imagery enables the obseryer.to
increase his field of view to detect relevant features or conditions
extending over hundreds of square kilometers.
Spectral range and resolution. These systems extend the spectral
limitations of the human eye, which are normally about 400 to 700 nm.
2.
3.
Spatial resolution. Many remote sensors have a greater spatial"
resolution than the human eye. The human eye resolves approximately
5 lines per millimeter while images can often contain several hundred
lines per millimeter. .
4.
Time record. Images provide an instantaneous record at a particular
point in time. This particular advantage has a special value in the
study of dynamic water systems.
guantitative capability. These sensors are all energy sensors. The
quantity of light reflected from the water surface is usually
recorded in some manner. Whether it be computer sensor images, they
can be quantitatively measured with such devices as computers or
densitometer~. The relatively precise geometry of most images (i.e.,
vertical aerial photographic and Landsat data permit the accurate
determination of size, shape, and position of ohjects.
5.
6.
Digitization. The proper digitization of the data creates a grid
which can often be incorporated into existing data bases. These data
may then have potential as direct input to environmental quality
models.
7.
Near real-time capability. Near real-time images have been received
onboard operational research vessels (8arker et al. 1975; Hill and
. Dillion 1976). This capability enables researchers to conserve
valuable fuel and time by quickly locating the area of interest.
Prevailing weather patterns surrounding the study area may also be
readily observed and/or predicted.
Inexpensive. Once a reliable relationship has been established
between the imagery and the surface truth, it is cheaper to acquire
information from imagery as opposed to the alternative of expensive
and time consuming field sampling teams and ships, especially when
conducting seasonal, or long-term investigations.
Before conducting the most basic of remote sensing studies, the effects of
varying atmospheric conditions on incoming solar radiate should be understood.
8.
8
-------
There are various regions of selective absorption which depend on the
compositon of the ai r, primarily water vapor, carbon dioxide and ozone, and
the particular wavelength of light. For wavelengths less than 290 nm, ozone
and oxygen are responsible for absorbing ultraviolet radiation. Water vapor
and carbon dioxide ar-e primarily responsible for absorbing energy in the near
infrared (Holter 1967). The atmospheric path length through which this energy
must travel also affects its attenuation. James (1970) and Elterman and
Toolin (1965) provide details on the subject of atmospheric attenuation.
Molecular scattering is of major importance under clear dry atmospheric
conditions. The scattering by very small particles of molecul ar dimensions is
inversely proprotional to the fourth power of the wavelength (Jensen 1968)
and, therefore, affects shorter wavelengths more than longer ones. As
particle sizes in the atmosphere increase, the wavelength of maximum scatter
becomes less sensitive and extends beyond the blue and progresses into the
green, yellow, and other regions of the spectrum (James 1970). As particles
increase from molecular to aerosol size, the color of the sky changes toward a
cloud white. For these reasons, the near infrared is less affected by
atmospheric haze.
Once the sun's energy has passed through the atmosphere and reaches the
surface of the water, two. things happen (Figure 1). Some of the radiation is.
refl ected (s urface refl ectance) and some is transmitted through the ai r-water
interference, a portion of which is further refracted, transmitted, scattered,
and/or absorbed in the water. If various constituents of water are to be
investigated, that portion of scattered light which is directed upwards and
which passes through the sea-air interface (volume reflectance) must be
measured. Volume reflectance is difficult to measure with existing wideband
Landsat multispectral scanners, James (1970) has described ways to subtract
out its undesirable, additional effects.
The intensity and spectral character of radiation is modified by
scattering and absorption as it travels through water. In a turbid medium the
scattering is caused by the reflection and diffraction of 1 ight rays by small
particles of suspended matter and colloidal solutions. . Just as in the
atmosphere, if the size of the particles is small compared to the wavelength
of light, the intensity of the scattering is inversely proportional to the
wavelength to the nth power, where the exponent n decreases with increasing
particle size from a value of four for pure water to near zero for coarse
suspended matter (Jensen 1968). Therefore, for solutions containing small
particles, the blue light has the maximum scatter, and for solutions with
large particles all colors are scattered approximately the same amount (James
1970). Work conducted by Tyler and Richardson (1958) indicates that light
scattering in vario~s solutions is directly related to the concentration of
the contaminant. Jerlov (1968) states that in cases of scatttering by large
particles, the intensity of scattered light is proportional to the particle
surface area that ;s exposed to the incident beam. The intensity of the
scattered light is proportional to the particle surface area that is exposed
to the incident beam. The intensity of the scattered light is proportional to
the turbidity if the particle size is uniform. Szekielda and Curran (1973)
found that light penetration in water is affected by plankton and dissolved
9
-
"
~
;
..
..
't :J
-------
J s~an~er I
e
Water
.A.A..A..A.
A..A.A.A
...............
~
..A.AAA
~
I-'
a
~
$~
d'~ $(\
">~. "'$0'
~..s
'"
~
........
........
~t~~1~~$$t~~~~1~~~~~1~~~i~i~~~~~~}~:~~~;:g~~~~~~t{~~t3~.I\~~~~1~~t~~~~J~~~~~~i~~~}~~~~s
Figure
1.
Pathways of various components of light from water that
is received by a multispectral scanner.
~,
-------
and suspended matter. Therefore, the composition of backscattered light from
below the air-sea interface is determined by the nature of some of the
constituents in the water.
Other investigators stress the ;mport~nce of th~ effect of dissolved and
particulate substances on water color. Kalle (1937, 1938, 1939) and Jerlov
(1951a, 1951b, 1953) indicated that in coastal waters the spectral change fro~
blue to green is a result of selective absorption by gelbstoffe, yellow
. organic particulates, and possibly detritus produced by an aging community of
phytoplankton in the water.
Ramsey (1968) states that the deep blue color of the oceans can be
attributed to the selective scattering of light by small particles and water
molecules. As productivity increases, the blue gradually changes to green.
There is a decrease in the reflection coefficient in the blue region. This is
due to an increase in yellow substances that are directly related to the
organic material in water and is often dependent on the level of productivity.
.Blue water is, therefore, not very productive. Jerlov (1948) categorizes
various marine water types from blue ocean waters to more productive inshore
waters according to the transmission characteristics of visible light (Table
1). .
TABLE 1. WAVELENGTH OF MAXIMUM TRANSMISSION AND THE
PERCENTAGE OF lIGHT OF THIS COLOR TRANSMMITTED IN
FIVE DIFFERENT TYPES OF OCEAN (Source: Jerlov 1948)
Wavelength of Percent
Types of Maximum Tran smi ss ion
Ocean Water Transmission (nm) per f'1eter
Clearest oceans 470 98.1
Average coastal 475 89.0
Clearest coastal 500 88.6
'
Average coastal 550 72.4
Average inshore 600 60.8
Yentsch and Owen (1975) conducted researchi n an attempt to demonstrate
how gelbstoffe and chlorophyll vary with respect to one another and to try to
explain the spectral characteristics of the particulates. These
investigations showed that the amount of yellow substance was inversely
related to salinity. The pattern of distribution, therefore, closely
represented the ratio between river and coastal ocean waters. Their study
11
-------
was conducted in July when production was nitrogen-limited near the Merrimack
River. Fresh water was associated with enrichment. Since fresh water and
yellow substance are correlated, yellow substance is also associated with
enrichment. Their study results indicate a positive correlation between -
yellow substance and chloropyhyll.
A summary of the interrelationships between the absorbers phytoplankton
and yellow substances) that Yentsch and Owen (1975) studied is as follows:
1)
There appears to be a posit ive corre 1 at ion between phytopl ankton
chlorophyll (band intensity at 670 nm) and the "particul ate yellows"
and IIdisso1 ved yell OWs".
There is a positive correlation between the "particulate yellows" and
IIdissolved yel1owsll.
2)
3)
The source of the "particulateyellows" is not known. They could
arise from the metabolic activities of the phytoplankton, or they may
be a particulate form introduced with river water.
They found that in waters containing
and yellow substances, the back scattered
only by a loss of short wavelengths, but
to scattering.
high concentrations of phytoplankton
light spectra were categorized not
long wavelengths were accentuated due
Still ather researchers have concentrated their research toward the
detection of various sediment types and ge1bstoffe. Blanchard and Leamer
(1973) utilized a spectral radiometer to measure radiation in the visible and
near-infrared portion of the spectrum in an attempt to examine: 1) different
concentrations of red, black, and gray clay particles in water, and 2) several
samples of natural pond water containing sediments and chlorophyll bearing
algae. They found that reflectance (attenuation) curves in the near infrared
(750 nm - 3000 nm) showed little change with changing sediment concentrations.
Reflectance (attenuation) curves in the visible region (400 - 700 nm) were
found to be sensitive to low concentrations of less than ZOO ppn suspended
solids (down to 75 ppm). This depended on the color and source of the
~ediments (Figures 2 and 3).
In studyi n9 four f1 ood detention reservoi rs conta i ni n9 green algae,
Blanchard and Leamer (1973) observed a peak attenuation at 570 nm and a dip at
690 nm (Figure 4). They suggest the possibility of utilizing techniques that
radio wavelengths to detect chlorophyll while simultaneously measuring- low
concentrations of suspended particulates. Although these relationships would
be most useful if a narrow band multi-channel sensor were utilized, they
indicate that sediments and algae can be monitored with sensors that operate
in the visible and near infrared regions of the spectrum. -
Yentsch and Owen (1975) made measurements which verify that the color of
water depends, to a large extent, upon the depth of which light penetrates the
water column. More light of shorter wavelengths is backscattered in areas of
12
-------
50
,........ 40 0 PPM
t::
~o
U...-j
Z.jJ
<:t:(1j
E--<:J
Ut::
~<:> 30
~.jJ
u-..jJ
~(1j
..... 0:::
w >--
E--<..--1
Z..--1
~(1j 20
U:J
0:::.jJ
~u 60 PPM
p"'(1j
'-'
10
o
I+- Landsat Green 41......
Landsat Red
400
450
500
550
600
650
Wavelength (nm)
I Landsat
-++- IR ~
700
750
Figure 2.
Percent refl ectance for water wi th red sediment (source;
Blanchard and Leamer 1973).
,I:.."';.: );.~. '.~;,
-------
.....
.j:::.
r--.
!::
~o
U.r-i
Z.j.J
-,
f-orl
Zrl
~tO
U;:j
r::r::.j.J
~u
P-.tO
'--'
50
40
30
20
--
--
-
-
-
-
10
o
400
450
87 PPM
9 PPM
It- Landsat Green +
500
550 600
Wavelength (nm)
-Rio Grande Valley
--Texas Backland
17 PPM
66 PPM
.......
.......
--'
~........
\
J Landsat
Landsat Red ~ IR -t
650
700
750
Figure 3. Percent reflectance for water with hlack sediment (Rio Grande
Valley and Texas Blackland) (source; Blanchard and leanrer 1973).
-------
s
r"'"'\ 4
I::
~o
U'M
z~
...
E-<,.....,
z,....., 20
~ c-j
U;:J
o::~
~u
c-.CIj
'-'
10
o
400
~ Landsat Green
L Landsat
Landsat Red -+tr IR ~
450
500
550 600
Wavelength (nm)
650
700
750
Figure 4.
Percent reflectance for pond water with algae (Chickashe,
Oklahoma) (source; Blanchard and Leamer 1973).
-------
low transparency than in waters of high transparency. Since longer
wavelengths strongly absorbed by water, any return in these wavelengths is due
to scatters near or on the surface. Their observations also show that
phytoplankton not only absorb blue wavelengths, but also strongly backscatter
red wavelengths. They advocate the use of wood band sensors similar to
Landsat multi spectral scanners to develop a general scheme to di sti ngui sh
yellow substances from phytoplankton. This separation .capability would enable.
researchers to better study pollution related parameters (i.e., specific
speci es of al gae). They encourage the uses of Landsat in the study of
selected parameters related to coastal eutrophication and water quality
mapping.
MONITORING TURBIDITY QUALITATIVELY WITH LANDSAT
Landsats-1 and 2 MSS's have broad spectral .channels \'A1ich were primarily
selected for monitoring the terrestrial environment. Landsat has, however,
been established as an acceptable tool for the detection, estimation, mapping,
and monitoring of selected water quality parameters~
Fleicher (1973) and Bowker et al. (1973) utilized Landsat digital data to
map turbidity patterns which indicated the flow of water masses in the
Chesapeake Bay, Maryland. Other investigators have used Landsat to
distinguish and monitor circulation patterns and changes in the relative
sediment load of discharging rivers. Such studies have been conducted by
Anderson et al. (1973) in coastal areas around Alaska, Ruggles (1973) in Long
Island Sound, Pluhowski (1973) for the Walland Canal and the Genessee and
Ousego Rivers, and Kelmas (1973) in Delaware Bay.
Fisheries information related to turbidity has also been derived from
Landsat data. ~1aughan et al. (1973) and Kemmerer and Benigno (1973) found
Landsat was a useful tool in providing information significant to the
harvesting of menhoden. Wright et al. (1973) utilized Landsat to plot areas
of (probable) fish concentration of commercial importance.
Pollution has also been monitored usi~g the synoptic view of Landsat.
Lind et al. (1973) monitored a discharge plume from a major paper mill along
the shore of Lake Cham~lain. Scherz et al. (1973) used Landsat to monitor
turbidity patterns around the water intake for the cities of Duluth and
Superior. Stumpf and Strong (1975) charted surface current patterns of
southern Lake Superior and Lind et al. (1973) utilized Landsat imagery to map
turbidity patterns in Lake Champlain. Klemas (1973) utilized Landsat data to
observe water boundaries and fronts (regions of high horizontal gradient with
associated horizontal convergence). From a pollution standpont, surface
slicks and foam collected in areas along frontal convergence zones near
boundaries contain concentrations of Cr, Cu, Fe, Hg, Pb, and Zn which are
higher by two to four orders of magnitude than mean concentrations in ocean
water (Szeki el da et al. 1972), Maul and Gordon (1973) observed rel ationshi ps
between Landsat radiance values and watercolor gradients across oceanic
fronts. Mairs et al. (1973) successfully collected the necessary information
on offshore waste disposal and estuarine flushing dynamics along New Jersey's
coast using Landsat data. Landsat data also proved useful in locating most
major sewage effluent sources in the Potomac River-Estuary (Schubert and Mac
Leod 1973).
16
-------
The distribution of nonpoint source (NPS) pollution have also been
observed by several investigators utilizing Landsat data. Carlson (1974)
identified and described principal sources of turbid water such as rivers. He
also determined that secondary sources, (e.g., landslides, beach cliff and
headland erosion, reworking of the bottom by storm waves, plankton blooms, and
waste effluents) can be observed, but are often masked by high river flows.
Lepleyet al. (1973) determined the distribution and flow of water masses
at four depth intervals by analyzing Landsat imagery through the use of
optical models of classes of vertical oceanographic profiles for the Northern
Gulf of Mexico. In a study of Kansas lakes, Osborne and f4arzo1f(l972)
utilized Landsat to" monitor turbidity which also correlated well with total
suspended solids. Other investigators have examined the capability of Landsat
to monitor various water types. Pirie and Steller (1973) detected the source
and movements of sediments in the nearshore and offshore zones of California.
An effort to correlate Landsat data ~nth measurements of turbidity and
transmittance from various water types in Lake Superior was conducted by
Stortz and Sydor (1974). Landsat was found to be useful in determining
general turbidity values. They demonstrated correlation between turbidity and
suspended solids. Landsat data were used to map various water types in the
Gulf of Carpentaria, Australia, by Teleki et al. (1973).
Landsat images of the Beaufort Sea, Alaska were used to distinguish three
water types with the following characteristics (Barnes and Reimnitz 1973):
1.
Turbid river runoff - low salinity (0-10) parts per thousand higher
tanperatures(1-110C), and high turbidity (>15 1 ight attenuation
coeffi c i ent GO: ).
2.
Melt from pack ice - salinities from 5-15 parts per thousand
tanperatures from 0-2°C and low light attenuation «3«).
3.
"Oceanic water" - sal inities from 25-30 parts per thousand, cold
water temperatures (1°C) and relatively clear «~«).
Landsat imagery was found to be especially useful in monitoring coastal
processes " in these often very remote areas!,
MONITORING TURBIDITY QUANTITATIVELY WITH LANDSAT
Of equal interest to this research are the efforts of investigators to
quantitatively estimate turbidity and related water quality parameters using
Landsat data. Egan (1974) in experiments near St. Thomas Island found that
Landsat data correlated well with surface turbidity while turbidity integrated
over depth showed a weak correlation with Landsat data. Chlorophylls a and c,
total chlorophyll and total carotenoids revealed an even weaker correlation.-
The lower limit of remote detection of turbidity was found to be about 0.2 JTU
(Egan 1974). Egan believes that ". . . turbidity, or equivalently, the.
backscattered radiance in the blue, green, or red is the optical parameter
that must be related to biology."
17
-------
Concentrations of suspended sediment load up to 2 mg/l, in the Ventura
Harbor-Anacapa Island area, showed a close correlation to variations in
Landsat film density along the survey line (Steller and Pirie 1974). The
radiance increased for sediment loads greater than 2 mg/l (Figure 5)..
Williamson and Grabau (1973) round that spectral IIsignaturesll of suspended
materials in the York and Rappahannock Rivers were not as distinct as they had
previously expected. In Figure 6 the correlation bands for thes~ two rivers
are pl aced on top of each other. For concentrations below 15 mg/l, the
correlation bands overlap so greatly that discrimination between sediment
types was clearly impractical. Channel 3 did provide some differences for
suspended material concentrations between 15 mg/1 and 25 mg/l.
Wright and Sharma (1973), in studies of Alaskan coastal circulation
patterns, found that photographic transparencies of Channell best delineate
(upon visual interpolation) suspended loads in the concentration range between
2 to 20 mg/l , and Channel 4 was best for concentrations over 1,000 mg/l .
Channels 2 and 3 were good for estimating intermediate conc.entrations. Color
.density sliced transparencies used to produce enhanced prints correlated well
with suspended load measurements. .
In a study conducted using Landsat-l to monitor two Kansas reservoirs,
Channels 2 and 3 were found to best correlate with suspended load and sunlight
penetration depth (Yarger et al. 1972). Channel 2 exhibited good
correlations, but these varied with atmospheric conditions. These variations
were however, not well explained in the article. Channel 4, although poorly
correlated, revealed a brighter return, apparently over the other three
c.hannels, for suspended loads >100 PP11. Correlations deteriorated for
concentrations greater than 100 ppm. Hunter (1973), in studies along the
Texas coast, found that Landsat wa s capable of reveal i ng di fferences as 1 ittl e
as 0.7 mg/l tn concentration of suspended particulate matter in surface
waters. Sampled concentrations of suspended matter at the surface ranged from
0.2 to 1.8 mg/l .
A more definitive quantitative estimate of suspended particles was
conducted in the New York Bight by Yost et al. (1973). They developed the
following method to predict what they termed the absolute value of total
suspended particles using Landsat composites of Channels 2 and 3 which were
precisely made using the step wedge supplied on the imagery. It must be
remembered that digital data, which was not considered in this project, yields
even better estimates of suspended material than does photographic data.
Positive images were made from negatives to bring out maximum water detail
(Yost et al. 1973). The black-and-white image brightnesses of Channels 1,2,
3 and 4 were obtai ned and a regressi on anal ysi s performed wi th respect to an
average extinction coefficient, total cell counts, and chlorophyll a.
Following this analysis, Channels 1 and 2 were found to best predict the
extinction coefficient. Positive images of Channels 2 and 3 provided the best
estimates of .tota1 suspended particles. The images were then density sl iced
and put on a video display to construct a chart of water characteristics being
analyzed. The brightness of the projected reprocessed positive image was then
measured at each sampling station for each of the four individual MSS
channels. The relationships between screen brightness in each channel and
18
o~
-------
""
::.:::
, [/)
H
A
-
lD
H '
::r::
u
u
.~
[/)
o
v
; So n Ftr 111 sco
(" Calibration curve
_I 5800 A Band
Brightness versus
1/.. sediment load
II .
II
. V
/ ..
. II
. /
. /
2
4
6
8
10
12
14
1
5
10
50
SEDIMENT, LOAD
(Mg/1)
100
200
1000
Figure 5. Calibration curve for sediment load versus film density in
median California coastal water (source; Pirie and Steller 1973).
, .
.5
n
;x:.
t-<
H
td
. 4 ~
t-3
H
o
Z
t::I
.3 t'I1
Z
en
H
t-3
--<
. 2
-------
..
1.4
1.2 Channell
1.0
5 10 15 20 25 30
SUSPENDED MATERIAL (Mg/1)
0.8
,-.,
Q::;
U)
I
N
e
u 0.6
........ Channel 2
~
e
'--'
~
u 0.4
z
< 5 10 15 20 25 30
t-t
~
~ SUSPENDED MATERIAL (Mg /1)
0.5
0.3 Channel 3
--
--
--
--
---- ---
.--- ----
--
0.1
5
10 15 20
SUSPENDED MATERIAL (Mg/1)
25
30
Figure 6. Comparison of radiance signatures of York and
Rappahannock Rivers (source; Williamson and Grabau 1973).
. 20
-------
total suspended particles were plotted and a linear regression \'ias developed.
for each. These models showed that in the case of single channel images,
little relationship is.evident between total particles and image brightness.
However, combined positive images of Channels 2 and 3 demonstrated a
significant relationship (Figure 7). By using more precise controls on the
processing of the positive Landsat images through calibration using the step
wedge provided on each image, a quantitative relationship between image
brightness and total particles can be derived. This same technique was used
to calculate relative extinction coefficients for water from Channels 1 and 2,
but atmospheric affects apparently caused variability in image characteristics
which prevented the acquisition of absolute measurements.
In sum, these investigations demonstrated that in selected geographic
areas, Landsat MSS data can be used to detect and often monitor such coastal
zone features as water masses, river plumes, surface circulation patterns,
current blockage; water types, sediment sources, sediment dispersal sites,
upwellings and pollu'ti6'ti{effluents and water boundaries). The MSS can also
be used to estimate water quality parameters such as secchi depth; extinction
coefficients, suspended solids, sediment load, and turbidity.
Because of the uniqueness of each and every coastal area, the usefulness
of Landsat data in conjunction with water quality data must be evaluated for
calibration purposes on an individual basis. Landsat, as with other remote
sensi ng devices, does not al ways. provide the necessary data from di fferent
geographic areas, under all environmental conditions, during each season of
the year. Very few of the above mentioned Landsat applications have been
accepted and util ized by the "truell user community; in this case, managers of
coastal resources.
Table 2 summarizes Landsat channels identified by some investigators for
the detection and monitoring of turbidity and turbidity related patterns.
Photographic transparencies of Channels 1 and 2 were utilized for qualitative
studies of plumes, water types, and circulation patterns. Although not
stressed in the literature, it was assumed that digital data of Channels 1 and
2 have also been used to at least qualitatively, if not, quantitatively, study
these parameters.
Transparencies and digital data from Channels 1 and 2 v~re found to best
provide both qualitative and quantitative turbidity and suspended solids
information. Channels 3 and 4 were occasionally utilized to derive
concentrations of suspended solids when such concentrations were either
relatively high or surface oriented.
21
-------
1 New York
Bight
U) total particles
U) 9
~
Z
N E-<
::z:: 8
N 19
~
A::
I:Q 7
z
~
~ 6
A::
U
U)
S
4
8
12
]6
if
-tr
y:::5.05+.l6x
.
31 May 1973
. TOTAL PARTICLES ex108)
* 2S January 1973
20
24
28
Figure 7. Total particles versus image brightness of accurately processed
positive composites of Landsat Channels 2 and 3 (source; Vost et al. 1973).
-------
TABLE 2. SUMMARY OF PREFERRED LANDSAT SPECTRAL CHANNELS (1, 2, 3 AND 4)
FOR TURBIDITY RELATED RESULTS DER I VED FR(}1 THE LITERATURE BY
NUMBER OF PROJECTS
----9ualitative Results
Transparencies Digital Data
123 4 1 234
----9uantitative Results
Transparencies Digital Data
1 2 3 4 123 4
Turbidity 7 3 1 5 6
Sus pended 2 3 6 5 6
sediments
Pl ume s 3 4 3
Water types 3. 3
Ci rcul ati on 5 3
patterns
Po 11 uti on 2 3 1
5 8
637 1
2 3 5 5
23
-------
SECTION 5
GEOGRAPHIC STUDY AREA
Because of its long coastline, Florida is a prime example of a state which
could benefit from a monitoring program using remote sensing systems. Many
Florida agencies are interested in assessing the environmental impact of land
development occurring in rural areas, assessing the success pf environmental
recovery projects in wetland areas, and studying coastal zone eco10'gy in an
attempt to develop best management practices for their many natural resources.
Universities, industry, and local, State, and Federal agencies are also
collecting background water quality data in an effort to measure the impact of
such activities as ranching silviculture, dredging, and damming on the ecology
of estuarine systems. The U.S. Environmental Protection Agency is also
interested in the acquisition of baseline water quality and land-use
information.
The specific parameters measured depend upon the objective of the
monitoring program and the constraints relating to personnel, money, time, and
the availability of technology. Frequently estuarine monitoring programs
measure and/or delineate current patterns, as well as the distribution of
suspended sediments, nutrients, color, turbidity, pollution, and related
patterns.
This information can be used in the selection of nuclear power plant
sites, and the siting of submerged oil pipeline corridors, various other
mari ne, and waterfront developments. Sites for the future propagati on of new
marine resources can be selected using the monitoring data. The utility of
remote sensing techniques is accentuated when one considers the extreme
variability from one estuary to another. Each estuary should be approached on
an individual basis with such factors as latitude, drainage area, river flow,
offshore circulation, and depth taken into consideration (Livingston 1975),
while keeping in mind that delineated ecosystem boundaries can often be quite
difficult.
The geographic area selected as the study area for this research program
was the Apalachicola Bay System, Florida. Ninety percent of Florida's oysters
and ten percent of the nation's oysters come from this particu1ar'bay. The
bay system at present has a dependable source of comparatively pollution-free
productive fresh water in an area of minimal urban and industrial population
growth (Whitfield and Beaumariage 1975). The marshes surrounding the bay are
also still relatively unaffected by man's activities. The bay system harbors
several rare and endangered plants and animals (i .e., southern bald eagle,
24
-------
osprey, black bear, alligator, snapping turtle, southern sturgeon, in- fact the
highest species density of amphibians and reptiles in North America (north of
Mexico) is found in the upper Apalachicola River Basin (Kiester 1971). This
estuarine system also serves as a major nursery for penaeid shrimp, blue crabs
and various fishes such as spot, croaker, redfish, seatrout, flounder, mullet,
and sheepshead. The Apalachicola Bay ecosystem is very succeptable to
man-induced perturbations (i.e., upland development forest resource management
activities) which could reduce the populations of shellfish and finfish,
thereby adversly impacting bay associated industries.
The Apalachicola Drainage System (Figure 8) encompasses more than 50,000
square kilometers (19,500 square miles) in three states (Florida, Georgia, and
Alabama) (Graham, pers. comm. 1978). The system is composed of numerous
streams, creeks and four major river systems (Flint, Cattahoochee, Chi pol a,
and.Apalachicola). The Cattahoochee and Flint rivers fiow into Lake Seminole
created by the Jim Woodruff' Dam. Flow from the dam forms the Apalachicola
Ri ver, wh i ch in combi nat i on wi th the Ch ipo 1 a Ri ver, is the maj or source of
fresh water and nutrients received by the Apalachicola Bay System (Figures 9
and 10). The Apalachicola River represents the largest drainage in Florida,
with an average flow rate of 23,460 cubic feet per second (cfs) (U.S.
Geological Survey 1971). At present the drainage area is sparsely populated
and not very developed in terms of agricultural and industrial activity.
Livingston et al. (1974) state that the Apalachicola system is a multi-fold
complex of interlocking wetland habits that include river-stream-creek
associations, wooded and shrub swamplands, marshes, and an extensive
estuari ne-coastal area. They have found that the ri ver fl ow is a very
significant factor in the ecology of the bay system which in turn serves as an
interface between the fresh water upland areas and the Gulf of Mexico. At
present, the Apalachicola Drainage System is viewed as one of the largest
relatively unpolluted areas in the United States and there is a widespread
interest to keep it that way for future generations to enjoy.
<'.
'"
The Apalachicola Bay System (latitudes 290 35' N to 290 50' N; longitudes
840 401 W to 850 151 W) is a shallow coastal estuary bounded by a series of
barrier islands (Livingston et al. 1974). The total area of the bay is
approximately 549 square kilometers (212 square miles) and is composed of East
Bay, St. George Sound, Apalachicola Bay, Indian Lagoon, and St. Vincent Sound
(Figure 9). A stable non-tidal current system is prohibited due to the
enclosing barrier islands in conjunction with (seasonal) fresh water inflows.
At mean low water the bay depth averages 2.7 m. The bottom consists of a
quartz sand base covered by up to 20 m of sediment composed mainly of silt and
clay. There are only four connections with the Gulf of Mexico, Sikes' Cut (a
dredged pass), Indian Pass, West Pass, and St. George Sound.
. Dawson (19S5) in his description of the hydrography, of the bay stated
that during periods of high river discharge, or influx of fresh water
discharge, the bay is fresh while during other times of the year it is saline.
Surface salinites are also affected by wind speed and direction. Winds over
the bay are predominantly from the northwest during the winter months and from
the south during the summer (Jordan ~973). Dawson (1955) expected that
vertical stratification did not occur in the bay since it was so shallow
25
-------
Figure 8.
~
Chattahoochee
Fllnt
Lake Seminole
Apalachicola
The Apalachicola drainage system including the major rivers
that contribute to the Apalachicola Bay System
(source; Livingston et al. 1974).
26
-------
N
.......
-
D~~
SO
\
G~
G~p~ 2
\~1 ~MArlh
SI'> " /' '''aoJ.
/'
/'
",,/
~
-- -
~.
Figure 9.
u.s. Geological
Survey map of Apalachicola Bay. Florida and vicinity
(scale 1:250,000).
-------
N
(X)
-..----- -
- -..---------
Figure 10. Landsat image (1516-15421) of Apalachicola Bay,
Florida and vicinity with area of interest outlined. "
i,
-------
<....,.'-
and therefore rapidly mixed by wind action. Livingston et ale (1974) found
the bay to be well mixed during periods of strong wind, but Estabrook (1973)
found that under moderate wind conditions (great enough to drive surface
waters but not strong enough to mix the water column), the surface salinity is
uncoupled from bottom salinities. Plumes of high salinity water extend into
the bay along the bottom and surface (at times) from both West Pass and Sikes'
Pass. Surface water temperatures, wi th an annual range of monthly average
temperatures of about 16.5°C, usually parallel the air temperatures (Oawson
1955). The tidal range is from approximately 0.5 m to 1.0 m. Tides in the
bay are semidiurnal with diurnal inequality.
Water turbidity correlates strongly with river discharge except at times
when bottom sediments are suspended in the water column as a result of'wind
mixing (Livingston et ale 1974). Nutrients varied widely with salinity, the
highest values usually associated with the areas of fresh water. Average bay
nutrient values vary seasonally with the highest nutrient concentrations
occurring during high (winter) river discharge and corresponding low salinity.
The chlorophyll values for Apalachicola Bay range from 3 to17 mg .
chlorophyll a/m3 with bottom concentrations often exceeding surface values
(Livingston et.al. 1974). Minimal levels of chlorophyll a and productivity in
February are associated with high river discharge which flush the bay system.
The primary factor 1 imi ti ng the growt h of phytopl ankton is temperature.
Light, turbidity, nutrients, grazing, and flushing rates are additional
controlling factors. Livingston et al. (1974) postulated that decreased river
discharge in the summer months ;s associated with decreases in nutrient
concentrations. When compared to Tampa Bay (Sykes 1970) and Long Island Sound
(Ril ey 1956), Apal achicol a Bay is a very product ive area. Less than 7 percent
of the bay bottom is covered with submergent vegetation according to the
National Estuary Study (1970). High turbidity is probably the cause for the
relative paucity of the bottom vegetation. The economy of Frankl in County,
Florida is largely dependent on the commercial fisheries associated with
Apalachicola Bay. The major commercial organisms include oysters, shrimp,
blue crabs, and numerous finfishes (Livingston et ale 1974). At least
three-fourths of the commercial landings are due to species that spend all or
some part of their life history in ApalachicolaBay (Menzel and Cake 1969).
If anything harmful were to happen to the bay's resources, the consequences
would be disastrous to not only a multimillion dollar seafood industry, but
also an entire way of life in Franklin County, Florida.
Several geographical areas in the bay system must be monitored in order to
understand various ecological trends. Numerous research scientists are
presently monitoring indicators of environmental change within these areas of
suspected impact. Coliform indices are of value in measuring levels of
domestic wastes and, in some cases, industrial wastes. The U.S. Public Health
Serv)ce (USPHS) has set a limit most probable number (MPN) value of 70 for
Group Coliform organisms in approved oyster growing areas. Based on sampled
levels of coliform bacteria, the State has therefore classified Apalachicola
Bay into an area where oyster harvesting is prohibited and other areas where
29
-------
-----'11--
harvesting is conditionally approved based on the above standards (Figure 11).
Although.values of Coliform Group MPN have not changed significantly since the
1940.s, Livingston et al. (1974) believe that as population (people and
animals) pressure in the area increases great water quality and related
fisheries problems may arise. There appears to be a relationship between
river stage and coliform MPN (Livingston et al. 1974). The river is probably
the most significant factor in the distribution of coliform bacteria in the
bay. At present, a stationary circle with an approximate two mile radius is
drawn from the mouth of the river into the bay proper to delineate waters that
are prohibited to shellfish harvesting. The coliform sampling program as well
as the water classification program conducted by various state and local
agencies would be greatly enhanced if water color patterns and related river
plume boundaries could be remotely monitored and delineated.
Barrier island development is still another activity with the potential to
damage the bay ecosystem. St. George Island appears to have the most
important growth potential for tourism in Franklin County (Colberg et al.
1968). Livingston et al. (1974) state that although the impact of bridges in
the area remains unknown, these structures could contribute to changes in
current patterns and exchange rates. Because the island is narrow, there is
1 ittle room for the natural filtering. of human wastes and runoff. Remote
sensing techniques are capable of providing information which can be used to
site commercial and residential developments. Information of this type can be
used in planning efforts which are aimed at minimizing adverse environmental
impacts and preserving ecologically sensitive portions of an island such as.
marshes, dunes, and beaches.
Other land-use practices also need to be monitored. Ranching and farming
activities include clearing, plowing, cultivating, harrowing, discing,
dredging, dyking, and damming. Intensive forest management activities
involving ditching, draining, clearcutting, and replanting occur in the East
Bay drainage basin (Livingston et al. 1974). Fishermen in the local area have
described and complained about the low quality of the "black" water draining
from the East Bay drainage basin after hea~y rains. Although unverified by
scientific data, damage to shrimp, crabs, and gamefish has also been reported.
Livingston (1974) has observed in the laboratory that shrimp avoid this water
. characterized by low pH and altered physical and chemical properties.
However, runoff from undisturbed (stable) land, as well as recently clear-cut
areas, has no observed deleterious effect on the feeding habits of various
organisms (Hydroscience 1977). Silviculture practices actually slightly
increase the pH of runoff (Hydroscience 1977), but research should be
conducted into the determination of whether or not the key parameter to
monitor may be the amount of drainage channels rather than the acreage of
various land-use activities. Livingston (1974) also states that little is
known about possible related alterations in the salinity structure of the bay
and the long-term consequences resulting from the introduction of such
chemicals as tannins, humates and fertilizers. Large quantities of highly
colored low pH waters (sometimes with a pH of 4) have been observed draining
directly from these clear-cut areas into the bay system. The runoff not only
damages organisms which live in the East Bay System, but this type of land-use
could increase the level of eutrophication. The forests included in these
wetlands are believed to be a source of nutrients and detritus for the complex
30
-------
w
......
~
N
~i~o~e~e~s j
Gulf of Mexico
o
3
6
Figure 11. Classification of the harvesting of shellfish in Apalachicola Bay
(Area A = prohibited; Area B = conditionally approved) (source; Livingston et ale 1974).
Shellfish areas are delineated on the map located 1n the back pocket of th1s report.
-------
food webs of the adjoining streams, rivers, and estuaries. Woodall and
Wallace (1972) suggested that watershed vegetation may be a primary factor
determining species abundance and composition in aquatic systems. Livingston
et al. (1974) state that from the results of their studies of detrital-based
food webs in Apalachicola Bay, it is very possible that the above-mentioned
generalization might apply to this system.
The U.S. Army Corps of Engineers is presently conducting an extensive
program to maintain maximum depths along the Apalachicola River for
navigation. It is also conducting a study of a plan for damming portions of
the river (Livingston et a1. 1974). Such programs are expected to affect
salinity and nutrient levels in the Apa1achico1a Bay System. Baseline
environmental quality data are needed to answer questions refiecting the
general lack of knowledge and the effects of future dams on the Apa1achicola
Drainage System. Numerous questions need to be answered. For example, is
fresh water from the Apalachico1a River being diverted into the Gulf of Mexico
through Sikes' Cut, a channel through St. George Island? The diversion of
this fresh water could increase the salinity of St. Vincent Sound, resulting
in an influx of 'oyster predatorsnonnally found in regions of higher
sa1 inities (Menzel et ale 1957; Menzel et al. i1966). Oysters generally 1 ive
in waters with salinities of 15-25%0.
The Apalachicola Basin and Bay are potential sites for industrial
facilities as well as inexpensive transportation. Corps of Engineers'
activities raise various questions concerning local habitat destruction by
flooding, interruptions of migrations by anadromous fishes such as shad and
striped bass, reduced nutrient and detritus flow, and alterations of
temperature and salinity regimes in Apalachicola Bay (Livingston 1975). The
Corps has constructed and maintained three relatively significant underwater
channels within Apalachicola Bay. One channel is oriented;'n a north south
direction and approaches the Apalachicola airport. A second, the intercoastal
waterway, comes from the east, runs east-\~st, and turns north into the
Apalachicola River. The third crosses an oyster bed off Cat Point and runs in
a north-west/south-east direction (see map in pocket). The effects of such
channels upon the bay ecosystem is presently being rigorously investigated by
State and Federal agencies. Remote sensing techniques have the potential to
acquire needed water current information to help answer several of the above-
mentioned questions. .
The various factors that stress an estuarine system are generally
established (Odum 1970). What remains to be answered is just how these
factors'effect productivity and ecological relationships over long periods of
time. Livingston et a1. (1974) state that alterations, similar to those
mentioned above, have already had significant effects on many estuarine
systems in Florida. Despite attempts to manage the area, no integrated
management program has yet been implemented in the Apalachico1a Drainage
System. Due to the size of the Bay System, the synoptic view offered by
Landsat would greatly aid agencies in acquiring (at a minimum) a temporal
overview of estuarine processes of ecological importance. At present, several
long-term investigations of the Apa1achicola Bay System are being conducted. .
The Landsat information derived from this research has been requested by
32
-------
numerous regulatory agencies and universities (FSU and UF), within the state
of Florida, concurrently conducting related studies in and around the bay.
Livingston (1975) in an effort to develop a resource management program for
the Apalachicola Drainage System, outlined the following laws and
administrative regulations (along with responsible agencies) designed to
promote management and conservation of aquatic systems:
1.
Federal
A.
Rivers and Harbors Act of 1899 (33 U.S. Code, Sections 401, 403,
404, 406-417)
Applies to filling, excavating, or altering navigable waterways,
also regulates discharge of pollutants, refuse, and dredge spoils
into navigable waters. U.S. Army Corps of Engineers is ,
responsible for pennitting (in cooperation with Florida Board of
Trustees and Department of Pollution Control).
B. Federal Water Pollution Control Act (33 u.S. Code, Section 1141
et seq.) - amendments of 1972 (Title 33, U.S. Code, Section 1251
et seq.) .
Aims to restore and maintain chemical, physical, and biological
intergrity of all waters of U.S. Calls for elimination of
pollutant discharges. by 1985 and achievement of water quality
for protection and propagation of fish, shellfish, and wildlife
by 1983. Responsible agencies include U.S. Environmental
Protection Agency (EPA), U.S. Anny Corps of Engineers, U.S.
Coast Guard, with help from Florida Department of Pollution
Contro 1 . .
.,
"
C.
National Environmental Policy Act (NEPA) of 1969 (42 U.S. Code,
Sections 4332, 4344)
D.
Establishes environmental protection and restoration as national
policy with provisions for generation of environmental impact
statements concerning any actions of federal agencies that may
impinge on the environment. The Council on Environmental
Quality, established by NEPA, provides guidelines for such impact
statements. U.S. Environmental Protection Agency is primary
agency involved in enforcement, although most federal, state, and
local agencies operate within NEPA.
Marine Protection, Research and Sanctuaries Act of 1972 (33 U.s.
Code, Section 1401 et seq.)
Concerned with protection of oceans from pollutants discharged
from vessels including dredge spoils, chemicals, etc.
Responsible agencies include U.S. Environmental Protection Agency
and U.S. Army Corps of Engineers.
33
-------
2.
. C.
E.
Fish and Wildlife Coordination Act of 1958 (16 u.s. Code, Section
661-666C) "
Requires consideration of effects of work in navigable waters on
fish and wildlife. U.S. Army Corps of Engineers coordinates with
other federal and state agencies.
F.
Endangered Species Act of 1973 (Public Law 93-205)
Provides conservation measures for endangered and threatened
species. Administered by "U.S. Department of the Interior.
State
A.
Florida Air and Water Pollution Control Act (Chapter 403,011,
Florida statutes)
B.
Public policy to conserve quality of state air and waters,
provided that no wastes are discharged into water without
proper treatment, etc. Administered by the Florida
Department of Po 11 uti on Control wi thO hel p from the
Division of Health of the Florida Department of Health and
Rehabilitative S~rvices.
Florida Water Resources Act of 1972 (Chapter 373, (Florida
Statues)
Relating to all state waters (except with respect to water
quality), conservation and control programs for management
" and conservation of such related resources (fish,
wildlife, etc.). Utilization of surface and ground water,
prevention of damage by flooding, soil erosion, excessive
drainage, etc. Administered by Florida Department of
Natural Resources with delegation of powers to five re-
gional water management districts. Presently involved in
generation of a state water use plan.
Florida Environmental Land and Water Management Act of 1972
(Chapter 380, Florida Statutes) "
Establishment of an Area of Critical State Concern (ACSC)
program and the developments of regional impact (DRI) evaluation
process. Areas of critical concern qualify for" such designation
by having environmental, historical, or archeological importance,
or being affected by major development. The purpose is to
formulate state decisions establishing land and water management
policies for the guidance and coordination of local decisions
concerning growth and development. This does not apply to more
than 5 percent of the land of Florida as an ACSC, and
agricultural activities are exempt from its provisions. A DRI is
a report filled out by the developer according to specified
questions that are to be answered concerning the overall impact
34
-------
of the development on the region's environment, natural
resources, economy, etc. The Division of State Planning,
Department of Administration implements this act; review of
DRIls are considered by the appropriate regional planning agency
with the local government conferring final approval, approval
with conditions, or denial. The overall purpose of this act is
to promote the creation of principles to guide development on the
local level within specified state-sanctioned guidelines so that
any major development in a given area is compatible with the
local environment.
D.
Florida State Comprehensive Planning Act of 1972
(Chapter 23, Florida Statutes)
Provides plan for long-term guidance for staff growth by
establishing goals, objectives, and policies. This
includes coordination of planning efforts among local,
state, and federal agencies. Division of state planning
is responsible for implementation of this act.
Land Conservation Act of 1972 (Chapter 259, Florida Statutes)
E.
Environmentally Endangered Lands (EEL) Program based.
on analysis of available ecological information to determine
priorities of environmentally endagered land. An EEL plan
will be developed to guide the purchase by the state of
endangered 1 ands. In such purchases, .there is no emi nent
domain power to implement land acquisition; this precludes
identification and priority listing of endangered lands.
The choice between acquisition and regulation depends on
level of protection necessary to achieve the desired
environmental aims. Emphasis is on ecological significance,
the importance of submerged lands, and appropriate
evaluation. Administration is by the Department of Natural
Re sources wi th input from other state agenc i es and a panel
of experts on environmental and planning concerns. This
includes interagency planning and advisory committees with
final approval by the governor and cabinet.
Beach and Shore Preservation Act (Chapter 161, Florida Statutes)
F.
Provides for beach nourishment, erosion control, regulation
of coastal construction, and establishment of setback lines
along beaches. Administered by the Department of Natural
Resources.
Graham et al. (1978a) report that the Federal Water Pollution Control Act
(FWPCA) of 1972 (Public Law 92-500) requires both point (201) and nonpoint
(208) pollution control. The 208 section requires that basin water quality
plans must include nonpoint sources of pollution. In Florida, such planning
for silviculture practices is presently being coordinated by the Division of
35
-------
Forestry and the Department of Environmental Regulation and should meet the
deadline for completion. Section 208 (F) requires the use of:
(F) a process to (i) identify, if appropriate, agriculturally and
silviculturally related nonpoint sources of pollution. . ., and
(i i) set forth procedures and methods (i nc 1 udi ng 1 and-use
requirements) to control the extent feasible such sources.
The FWPCA also requires the investigation of the affect of drainage
operations on fresh water hydrology (Graham et ale 1978a). The FWPCA
therefore mandates the development and use of:
(1) a process to (i) identify, if appropriate, salt water
intrusion into rivers, lakes and estuaries resulting from
reduct ion of fresh water fl ow from any cause, industry,
irrigation. . ., and (ii) set forth procedures and
methods which are otherwise parts of water treatment
management pl an. .
Livingston (1975) stated that the Apalachicola system is an example of
what is occurring in estuaries all around the country. Confl icting interests
are competing for the use of terrestrial and aquatic resources. It was
anticipated that this research would aid in the development of remote sensing
t~chniques useful for long range planning and resource management programs
which require the acquisition of extensive synoptic scientific data. The
ultimate goal of an estuarine resource management program is to provide a
-specific management plan, based on objective scientific data, which allows for
the application of intelligent alternatives to a given local or regional
situation (Livingston 1975). Livingston (1975) stated that ". . . only in
this way can the often difficult decisions be made which concern,resource uses
in our estuaries."
In summary, Apalachicola Bay, Florida is an enviromentally and
economically valuable area. It is well worth preserving for the benefit of
future generations. An extensive and rare surface truth (both water and land)
data base has been acqui red in the Bay System si nce 1972. The base i ncl udes
detailed temporal water quality and land-use data. With such a unique surface
truth data base already in existence, this project was conducted in an effort
to utilize Landsat to temporally monitor land-use changes and water color
distributions in the Apalachicola Bay System, Florida. .
36
-------
SECTION 6
MATERIALS AND METHODS
In some instances, photography is gradually being replaced by digital
satellite and aircraft derived multispectral scanner data in an attempt to
better acquire quantitative results. A digital format allows the data to be
manipulated through utilization of a computer without degrading the quality of
data which occurs in the processing of several generations of film (Van Wie et
ale 1975). This section describes the Landsat system and procedures followed
for the analysis of Landsat digital data used in this study. A much more
detailed description of the analysis of Landsat digital data can be found in
Hill ( 1978) .
The greatest advance to data in the acquisition of earth-orbital imagery
undoubtedly occurred with the launch of Landsat-I, the first Earth Resources
Technology Satellite (ASP 1975a). Landsat-1 (formerly called ERTS, Earth
Resources Technology Satellite) was launched on July 23,1972 and was joined
by Landsat-2 on January 22, 1975. Each satellite is synchronized to pass over
the same point on the Earth every 18 days. The two satell ites are nine days
apart. Thus, i gnori ng the side overl ajJ coverage, the same area is covered
every nine days. Landsat, as it daily makes its way across the Earth, moves
in a 1.43 degree westerly direction which separates each path from another by
about 159 kilometers at the equator. The paths converge at the poles.
Landsat operates at an altitude of approximately 920 kilometers. It circles
the Earth every 103 minutes, completes 14 orbits for each of the 17 days and
13 on one day and can view the entire Earth every 18 days (National
Aeronautics and Space Administration (NASA) 1972).
Landsat satellites are equipped with a Return Beam Vidicon (RBV) Camera, a
Multispectral Scanner (MSS), and a Data Collection System (DCS). This
research deals solely with the MSS. The MSS uses an oscillating mirror to
continuously scan perpendicular to the path of the satellite. The system
receives data from four spectral channels simultaneously with each sweep of
the mirror. . The four channels are in the visible-and rear infrared spectral
channels from 0.5 to 1.1 micrometers (,500-600 nm, 600-700 nm, 700-800 nm, and
800-1100 nm). It is 1 imited to a dayl ight operational mode. The swath width
is approximatley 185 kilometers. .
The six optical fibers for each of the four channels are arranged in a 6 x
4 matrix located in the focused area of the scanner's telescope (Figure 12).
Light from the Earth is conducted to an individual detector, uniquely
sensitive to a specific spectral band, through an optical fiber. An image
37
-------
6 detectors
per channel
(24 total)
} 6 lines scan
I
Direction
of flight
Figure 12.
Landsat MSS scanninq arrangement
(source; NASA 1972):
- -
of a line across the swath is swept across the fiber each time the mirror
scans. A video signal is produced at the scanner output for each of 24
channels. These electronic signals are sa~pled, digitized, and formulated
into a serial digital data stream. Figure 13 is an example of the ground
coverage pattern derived from Landsat.
Landsat data are initially recorded in an analog format when received by a
ground station. The data are then reformated into a digital format.
Investigators have the option of receiving the digital data on
computer-compatible tapes (CCT) or photographic images. This study used the
Landsat digital data recorded on CCTls. All of the Landsat images in this
report, with the exception of Figuire 10 which was reproduced from a color
composite film positive, were generated from the digital data.
Upon receiving a complete list of sampling dates from FSU, the next stage
was to determine the dates vmen Landsats 1 and 2 viewed the same area. Tne
Browse Facility at the Remote Sensing Center of Texas A~1 University was
utilized to determine these dates. Before selection of the best overpnss
38
-------
W
1.0
~
Equator-159km
Figure 13.
N+l, day M+l
N+l, day M
Orbit N, day M+l
Orbit N, day M
landsat ground covera~e pattern (source; NASA 1972).
-------
dates, several factors were considered to assure that the selected images
would provide the most val uable information. It was necessary to acquire
images from specific times of the year in order to view such occurrences as
high and low river flow, high and low flow of runoff from the East Bay
drainage basin, various tidal stages, varying wind conditions, and seasonal
vegetation changes on land. It was unfortunate, but water sampling dates did
not coincide with 'those of the historical Landsat images. Taking into
consideration the above-mentioned factors, Landsat CCT's were ordered in sets
of two for ease of handling (Table 3).
TABLE 3.
LIST OF LANDSAT COMPUTER COMPATIBLE TAPES RECEIVED
FOR THE APALACHICOLA BAY PROJECT
Image Scene Landsa t
Number Date 1. D. Number 1 or 2
1*+ 17 February 1973 1209-15394 1
2* 17 February 1973 1209-15401 1
3*+ 13 April 1973 1264-15454 1
4 19 May 1973 1300-15452 1
5*+ 21 December 1973 1516-15421 1
6* 3 Ma rch 1974 1588-15402 1
7*+ 23 October 1974 1822-15333 1
8*+ 26 February 1975 1948-15282 . 1
9* 20 July 1975 5092-15204 1
10 11 December 1975 5236-15120 1
11 12 February 1976 2386-15303 2
12*+ 19 August 1976 5488-14542 1
* Fully analyzed Landsat computer-compatible tapes.
+ Analyzed for land-use practices.
Once a pair of Goddard tapes are recieved at the Enviromental Protection
Agency's (EPA) Environmental Monitoring Systems Laboratory (EMSL), they are
taken to the Data Analysis System (DAS) (Figure 14) for analysis. The DAS
system is a low-cost data analysis system for processing multispectral
aircraft and satellite scanner data. The tapes are immediately reviewed for
such information as proper format, damage, cloud cover, data compression, scan
lines and elements encompassing the area of interest, and which of the four
files comprising anyone entire image, contain the study site. In order to
save valuable processing time, a copy (one file per tape) is then made of only
the pertinent data segments by skipping a predetermined number of san lines.
The functional steps of the "supervi sed II processi ng of these Landsat r~ss
digital data are summarized in Figure 15. The MSS computer-compatible tapes
are preprocessed before image analysis is initiated. Preprocessing entails
destripping, an attempt to balance differences in detector responses due to
varying gains, and converting the data into a format that is compatible with
subsequent DAS programs.
40
-------
-- -~-
-- ~_.
----
DATA ANALYSIS SYSTEM
OPERATOR'S TERMINAL
AND CARD READER
9-TRACK MAGNETIC
TAPE DRIVES
PLAYBACK SYSTEM AND
CENTRAL COMPUTER
.. --
Figure 14.
Components of the EPA Data Analysis System.
41
-- -
-------
I~--
MSS
COMPUTER
COMPATIBLE
TAPE
,
DATA
PR E-PR OCESSIN G
TV DISPLAY IMAGE ANALYSIS HARD COpy
. Image -Image Enhancement . B & W Film
. Statistics 8 Classification 8 Electrostatic Plots
. Statistics
,
POST ANALYSIS
PROCESSING
8 Color Assignment
8 Annotation Overlays
8 Rectify to Map Base
TV DISPLAY HARD COPY
.Color/B & W Image IMAGE LINE PRINTER
. Statistics -Color/B & W Film 8 Statistics
. Electrostatic Plot
Fi gure 15.
Simplified flow diagram of the functional steps
of Landsat digital analysis.
42
-------
The next phase is to create a tape that enhances desired features for the
sole purpose of picking training fields. The Display Tape Program (DAPIOS)
contrast stretches and 1 eve 1 sl ices the raw data to generate an enhanced
image. Channels 2 and 4 were used for the generation of this image. Two
separate and different display tapes were produced, for it was necessary to
(1) separate land from water in order to enhance land features, and (2) to
separate water from land in order to enhance water types. Both of these tapes
were used for the selection of training fields.
The next stage utilizes the Comtal (television screen) and the Comtal
Varian Image Processing System (CVIPS). This phase of analysis, involving the
selection of training fields, is probably the most critical phase of the
entire operation, for it is only from class spectral signatures, derived from
statistical comparisons of the individual training fields, that the computer
decides how to classify every pixel or picture element, in the entire area of
interest. A training field is a small area of a known land-use or highly
suspected water type, selected from a display tape, Which is to represent the
spectral signature of a particular field of interest during classification.
Before even sitting do~~ at the computer, the goals of the project are
reviewed in detail, ground truth materials are gathered and studied, and
potential training fields are selected. Since aircraft imagery was not
available for the dates selected, it was decided to utilize USGS maps,
National Ocean Survey maps, and maps supplied by the Buckeye Cellulose
Corporation, as the sole source of ground truth information. Qifferent
criteria for the selection of training fields are followed, depending on the
value of a particular class to the goals of the research. For classes such as
clear-cut and swamp/forest runoff, it is necessary for the values in the
training field to be relatively homogeneous when compared to the rather
variable, but encompassing training fields such as marshes, cities, and
natural forests. A goal was to classify all forested areas as forest, but not
to actually break them up into forest types.
It is desirable to have the training fields of specific interest unifonnly
distributed over a scene, but this may not be always possible in nature. It
is al~o helpful in most cases to include some noise within a training field so
as not to drastically limit the range of a spectral signature.. Through
experience, it was determined that a sample size smaller than 15 pixels is not
statistically representative of a specific spectral signature (except in the
case of rivers and roads where 15 pixels are often difficult to acquire). It
is also best, for statistical purposes, to select sampJe sizes that are
approximately the same si ze. At least three samples are selected for each
candidate class, except in the water where obvious differences and boundaries
existed. Generally, land and water class spectral signatures determined from
one tape are transferable to the neighboring tape, but it is procedure to at
least view the adjoining tape within the Landsat scene, and in most cases,
select several new training fields. At the end of this process, the image is
reviewed with a computer graphic overlay of all the training fields. This is
to insure that the entire area and all different areas of specific interest
have been sampled, because the addition of training fields at a later date is
time consuming and difficult.
43
-------
Before processing begins, disk files under a specific qualifier (name) are
created in order to store the data handled during most of the computer phases.
Once experience is gained, it is most convenient to store the data area of
interest on any of six scene files, since the entire area of interest is not
viewed at anyone time. The next scene is restored when the adjoining set of
training fields are to be selected. The training fields are either selected
by automatically drawing a box of a preselected size or by using the cursor to
draw an irregular polygon around the area of interest.
Training fields for a particular class are chosen at the same time so as
to simpl ify 1 ater stages of analysis. It is a practice to only sampl e areas
that can be identified with a high degree of confidence as a specific land-use
type on the accompanying ground truth maps.
The water enhanced display tape is thoroughly examined. Again, all
necessary scenes are stored before the water training fields were selected. A
color table (STEP 26), di spl aying a specific color for a particul ar count from
o to 63, is selected which best separates the various water types. For the
most satisfactory classification results over the very low counts in the
water, the training fields include a lot of neighboring noise. Such noise is
created by mixing of water types, striping due to detectors and even sea
state. Including the noise widens the range of the specific class just enough
to include most of the noise, thereby smoothing out the final classification.
If mixing zones between water types are as distinct as the bordering water
types the boundary sampled and is generally made a separate class. If the
boundary is not distinct (mixed) it is split statistically by widening the
thresholds between the two mixing water classes thereby forming a more
distinct boundary between even larger water types. In general, far fewer
training fields are selected for a certain water type than are selected for
landtypes. Up to 16 different types of water were identified. Even though
the water types are only often one count apart, they display better
separability qualities than land features which are often separated by many
counts in several channels.
Late in this research effort the capability to interactively compute, view
and store training field statistics was taught to EPA personnel during a NASA
Tra~sfer Technology Course. After the training fields are selected and stored
in a file on the disk (TSIN file), the Compute Statistics (CS) routine is
entered. This procedure was only performed on the last processed image, Image
5, due to time and resource limitations. The statistics of each training
field are reviewed even before they are stored in the TSIN file. Although it
takes approximately 1 minute to compute the statistics for one training field,
several useful training field statistics are generated due to this extra
processing. Approximately 100 training fields had to be selected for each of
the 8 images in order to discern land and water classes.
Once all of the training field statistics for all four channels are
computed and stored in a statistics file (STAT file), the histogram program
(HISTOS) is run. This program presents a compact, yet comprehensive, printout
of the relevant statistic~ of a training field. This process saves a lot of
paper if used properly in comparison with several of the other s~atistical
44
-------
programs. Other products generated from the computed statistics are four
channel spectral plots of either particular classes of similar training fields
(i.e., all marsh training fields) or individual training fields. This is also
another valuable tool used to group training fields into classes. These tapes
are printed from a separate plot tape on the STATOS printer/plotter.
Lastly, the SCORECARD program is run to give a rather biased estimate of
the accuracy of the final classification. This program merely gives a
percentage of the classes into which each original training field was finally
classified.
. Returning again to the main flow of data analysis, all of the coordinates
for the training fields are stored in a file on the disk. The next program
(ISOFLD) uses the training field coordinates on the disk file to locate and
isolate the individual training field raw data, in all four channels, from the
reformatted DAS tape.
At this stage the pattern recognition routines are utilized with the end
goal of grouping training fields into specific classes with unique spectral
signatures. The specific program is called PATREC STATS, and it enables the
analyst to calculate the following statistics on either individual training
fields or classes; mean vectors, covariance matrices, correlation matrices,
histograms, spectral plots and divergence. The four different modes of
operation in the statistics package are ALLS, EDITS, POOLS, and DIVERG.
The ALLS program is first run to derive statistics on each and every
separate training field. After observing the mean, standard deviation and
spectral plot for each of the four channel s, the trai ni ng fi e1 di s fi rst
determined to be a good sample (i.e., with a relatively normal distribution)
and second, the field is compared to all other training fields and placed into
s imi 1 ar spectral groups termed "cl asses". An added aid to trai ni ng fi el d
grouping is an EDITS run which not only produces statistics on each training.
field but also produces an interfielddivergence matrix which compares a
particular training field with every other training field.
After grouping of individual training fields into classes, the POOLS mode
is run. With a minimum amount of printout, statistical summaries of each
class, class spectral plots, and the interclass divergence matrix (in
conjunction with the class statistics as a double check) are used to compare
the closeness of classes to one another. It is inevitable that similar
classes would occur and a second and usually final POOLS is run. After being
satisfied that the divergence matrix has not pointed out any conflicting
classes, a signature tape (SIGTAP) is produced during a final run of the POOLS
program. The SIGTAP has on it the characteristic spectral signatures of all
land and water classes.
The ELIPSE program is then run to build .a look-up table so that the
pattern recognition can be accomplished on a pixe1-by-pixel basis (Eppler
1974). As compared to the technique of computing and comparing probabilities
during the process of c1assification, the table look-up technique saves a
45
-------
significant amount of computer time (Jones 1974). The ELIPSE program produces
a table tape (TABTAP) containing look-up tables for the classification
process.
The ASSIGN program actually performs the pattern recognition by using the
previously produced table look-up algorithm to classify the reformatted raw
- data (LANREF tape) and thereby classifying the entire image. To shorten the
process even more, this program first looks at the preceding pixel and, if
similar, places the next pixel in the same class. For the ASSIGN run, each
land class was usually weighted with a threshold (confidence level) of 13.28
(95%) and water classes had a threshold of 20.00 (99%). Water types had
higher confidence levels due to their consistent spectral uniqueness,
uniformity, and therefore, ease of separability. The classified data are
output to computer tape (CLSTAP).
I .
The results are first viewed at this stage. The classified image is
viewed on the COMTAL. In approximately 75 percent of the cases, at least part
of the image is either misclassified, unclassified, or classified in such a
manner as to create an undesirable result. Two procedures can then b~
followed depending on the problem. If part of the classified or unclassified
image is speckled, usually thresholds are lowered to incorporate these noisy
points into nearby classes. Often boundaries bet\veen classes, especially
areas of mixing between water types, can be absorbed in this matter. If the
unclassified data in the area of interest are occurring in large areas, it
necessitates the taking of new training fields from the unclassified area(s).
The new trai ni ng fi el d( s) is then added to the end of the ori gi nal ISOTAP by
simply skipping the appropriate number of old training fields and adding the
new field(s). The entire previously mentioned process from the running of the
ISOFLD program to the final ASSIGN run is repeated. This process is normally
conducted atl east twice before a satisfactory classified image was produced.
The classified image by itself is useless other than for display purposes
if it is not gographically corrected. In order to geographically correct a
Landsat image, it is referenced to a Universal Transverse Mercator (UTrA.)
coordinate data base (map). For this project, 1:250,000 USGS maps ~th an
overlayed UTH grid were used due to a lack of updated 7.5 or 15 minute maps.
The GEOREF CONSTANTS Program utilize a least squares regression technique.
to derive the constants used as input to the transformation equations that
position the pixels to correspond with the UTM base map. Two steps are
required to perform the CONSTANTS Program. Control points are selected from
the raw imagery in the first step. Display tapes are generally inaccurate for
such purposes. For accurate determination of specific intersections, the zoom
capability on the Comtal us used quite heavily. Scan lines and element values
for each control point are recorded. Corresponding Northings and Eastings are
calculated and recorded once the operator left the COMTAL. Channel 2 is best
for viewing features such as road intersections, and Channel 4 is best for
land-water boundaries depicting such features as river and creek intersections
and points of land encompassed by water. The Georeference Program (GEOREF)
corrects a one degree by one degree area. The best results are derived by
selecting accurately located points from at least all four corners of the area
of interest. This process often requires the incorporation of points from
adjoining tapes.
46
-------
The appropriate Northings, Eastings, scan lines and elements, along with a
point identification code (i .e., PI, P24, P8l), for each individual point are
fed into the GEOREF CONSTANTS Program. A new time-saving process was
developed (Levy, pers. comm., 1977) which cut down on the interactive
guesswork needed to throw out bad points. The GEOREF Program fits the lines
and elements .to a transformation using the corresponding UTM coordinates.
Each point is printed out by the program along with residuals and a 2 Sigma
error for the entire set, for both scan lines and elements. Points are then
interactively deleted and/or added until both scan lines and elements for as
many points as possible (15-20 for a good scene) either has a residual of 5 or
all were within the 2-Sigma error.
Points with a high degree of confidence were heavily weighted (by a factor
of 7) and those with low confidence were weighted with a 1. The heavier
weighted points are brought into range first, followed by an analysis of those
with a weight of 1. The constants are printed at the end of the interactive
GEOREF CONSTANTS run. These constants are fed into the GEOREF BUILDER Program
which actually geographically corrects the classified image.
If the image is to be processed on the stand alone film recorder and
eventually printed, the T-COLOR Program is executed. T-COLOR converts the
image to the desired final scale. T-COLOR tapes are produced for each image
at scales of 1:250,000 and 1:500,000. It was later discovered that at a scale
of 1:250,000 the entire image was too wide for the film recorder format. The
1:500,000 images were enlarged by two on a second run to acquire a scale of
1:250,000.
There are several reasons why so much time was spent geographically
correcting the data. Once all of the images are corrected, they are
registered with one another. Data such as acreage estimates from one scene
can be compared to all other scenes. The same basin coordinates, used to
derive acreage of various classes, are used for all sets of data. There is
also the overlay capability in which one data set, such as a USGS map, can
then be overlayed on any or all the classified data sets.
In order to acquire acreage of all classes within a particular selected
polygon(s), apex cooordinates (scan lines and elements) are selected.
Longitude and latitude for a particular polygon (basin) are converted to
Northings and Eastings and finally to scan lines and elements. The Defense
Mapping Agency (DMA) through a phone call and eventual computer printout
(mailed to EPA) , converted the latitude and longitude into Nbrthings and
Eastings. A second T-COLOR run is made to assign (group) colors on the
classified tape to the appropriate classes.
Multiple basins are then read into the Polygon Acreage Program (POLYAC)
from cards. The resulting printout produced, by basin, the following
information about each class: number of pixels, percent coverage over the
basin, acreage, and square miles. .
Aside from the Landsat data, it was necessary to acquire other supportive
data. A local lumber company provided the clear-cutting, replanting, ground
47
-------
cover and East Bay drainage basin information in the form of several land-use
maps (Hill 1978). One map indicated, on a year-by-year basis, the areas that
were clear-cut on company owned land located wholly in the drainage basin of
East Bay, Florida. A second map depicted the roads on company land along with
the year they were constructed. Most of the roads are paralleled by drainage
ditches.
The above-mentioned clear-cutting map was the source of land-use data
used to pick training fields in the East Bay drainage system. A training
field is a representative area, usually of limited size and of a known
surface material, which is used to train the computer to identify all other
areas having a like or similar spectral signature. USGS 7.5 and 15 minute
topography maps. The 1 umber company derived cl ear-cutti ng map wa s used as
guides to sample generalized land cover types (i.e., cities, beaches,
marshes). According to the lumber company forest resource manager, they
purge their clear-cutting information file about every three years. The
maps they provided were. constructed using any remaining information. The .
only infonnation they retain for extended periods is the time and locations
of ~eplantings. This information was not received in time for training field
selection. Had it been available, revegetated training areas might have been
broken down into finer categories. . .
A third map was also provided which outlined all of the smaller individual
drainage basins comprising the East Bay drainage system. Input coordinates
delineating the boundary of the East Bay drainage basin were taken from this
map and input to the computer for the purpose of acquiring acreage estimates
of land-use types within the basin. A table of clear-cutting acreage by
months within each plot accompanied this lumber company map, and was the
primary information used to verify the results of the computer classification.
FSU, through the efforts of Dr. Robert J. Livingston, is currently.
conducting its seventh year of water quality monitoring in the estuary.
Approximately 15 stations are sampled for some 15 pysicochemical parameters
as frequently as once a week. Studies are also being conducted concerning
assemb 1 ages of benthic i nfauna and epibenthi c fi shes and i nverbrates. The
station data are acquired in the following manner (Livingston and Woodsum
1976): .
Physical-chemical data (by area, station, date, time of day, and depth)
- color - measured using an American Public Health Association
(APHA) platinum-cobalt standard test .
- turbidity - analyzed using Hach Model 2100A laboratory
turbidimeter (z2% of scale) and expressed in Jackson Turbidity
Units (JTU) .
- Secchi disc depth - measured with a standard Secchi disc
48
-------
- salinity - measured with a Beckman salinometer (to.3%o)
calibrated by a Bissett~Berman Model 6230 laboratory salinometer
(t.003%o) and rechecked prior to sampling with a 47 ohm
resistor (tlO%o)
- pH ~ determined with a Leeds and Northrop porta~le pH meter
- river fl ow - taken at Bl ountstown, Florida and provided by the
U.S. Army Corps of Engineers (Mobile, Alabama)
- rainfall - provided by the National Oceanic and Atmospheric.
Administration (Environmental Data Service, Apalachicola, Florida)
FSU supplied most of the pertinent water quality data, including station
monthly means and standard deviations, and computer derived grey maps of pH
and color. This information was requested because this study started out as
a quantitative turbidity investigation, but was changed because of the lack of
concurrent water truth .data with the Landsat overpasses. Water qual ity data
was collected, at best, two to three days earlier or later than anyone
particular overpass. This research therefore, turned into a qualitative
water color study and a more quantitative land-use study, conducted to relate
land-use to water quality. Figure 16 is a map depicting all of Livingston's
water quality sampling stations.
Eight Landsat scenes were available to study seasonal fluctuations both.
in water and on 1and (Table 3). It would have been preferred to have tandem
water quality data for the overpasses. Existing monthly means were, however,
sufficient for at least identifying monitoring the distributions of water.
classes. While land-use information was not the best, it was quite sufficient
to monitor temporal land-use changes in the East Bay drainage basin.
49
-------
Tate's Hell
Swamp
lJ1
a
Apalachicola River
. Data Provided
Kilometers
~
I
6
\'1>\0(\
~eo(o.e
c.,\.
b. Data Not Provided
Gulf of Mexico
. . .
o 3
.
6
Study
Area
Figure 16.
Map of FSU water quality sampling stations in the Apalachicola Bay
System (provided by R. J. Livingston. FSU).
,..
-------
SEcn ON 7
RESULTS
Quantitative correlations between the landsat channels and water quality
parameters, such as turbidity, were not possible because the water samples and
measurements were collected several days on either side of the historical
landsat imagery. Since water is such a dynamic medium, samples must be taken
at very short time intervals on either side of an overpass. Through the
author1s experience, a maximum of one hour on either side of an overpass is
usually sufficient.
During the investigation of water color distributions, it was discovered
that runoff from the East Bay drainage basin could be easily identified by its
spectral response in the four landsat channels. There was only one major
change related to land-use activity in the drainage basin; silviculture. As a
result of these findings the research was expand~d to utilize Landsat data to
monitor changes in land cover due to silviculture practices. FSU started its
water quality sampling program in 1972, the same year landsat was launched.
FSU was, therefore, studying the result of a problem (declining water quality)
and Landsat could be used in this investigation to monitor the probable source
of the problem, namely, silviculture activities around the bay.
DISCRIMINATION OF ESTUARINE WATER COLOR DISTRIBUTIONS
As previously mentioned, direct quantitative correlations with water
quality data were not possible because the historical imagery was acquired
several days earlier or later than the water quality sampl ing dates. The
objective of this portion of research was to demonstrate the capabilty to
dfscern various types of water according to their perspective colors in
selected Landsat spectral channels.
The areas of specific interest were East Bay followed by Apalachicola Bay,
St. Vincent Sound, and St. George Sound. The parameters monitored were the
Landsat-derived four channel ~pectral reflectance responses of all water types
within an image. As mentioned earlier, the detectors receive reflected energy
and integrate these values over a given area, into single reflectance values
(counts) within the four wavelength ranges of the detector. Spectral
reflectance signatures (spectral plots) were derived by plotting count values
(reflectance) versus wavelength. From Landsat four channel statistical (Table
look-up Algorithms; Eppler 1974) correlations and group and individual
training field spectral plots, water types were found to be fairly easy to
discriminate from one another. Training fields for water types were selected
51
o
-------
on the basis of spectral signature geographic location, and available
historical water quality data.
A total of sixteen water classes were distinguishable on the basis of
their spectral characteristics. The Landsat spatial resolution of 0.44
hectares did not greatly affect the classification process, because the water
classes were sufficiently large, areally speaking, to provide a good
integration of reflectance over any particular pixel (picture element).
Striping, due to differences between detectors, was partially corrected
through use of a NASA supplied destriping program. In water, where the
difference between classes was often one count or less, the destriping program
did not do as good a job. The striping, although not usually noticeable in
Apalachicola Bay, is very evident in imagery of the Gulf of Mexico.
Hopefully, this problem will be corrected by NASA in data derived from future
Landsat satellites.
The color assignments for the water types are as follows (Figure 17):
- Tan, orange, browns, and greens:
waters presumably high in turbidity.
- Blues and greys:
water classes with a likelihood of low turbidity.
- Deep red: suspected, s\'/amp/forest runoff from the East Bay drainage
basin; believed to be highly colored, not very turbid, and fairly
acidic.
- Salmon:
believed to be dilute swamp/forest runoff.
Figure 17 is an aid to the interpretation of the classified Landsat images.
Visual comparisons between FSU's water quality data (Appendix A) and the water
color distributions revealed an approximately 90 percent agreement between
areas colored red and salmon "lith the high \'Iater color values. There was a
one-to-one correspondence between turbidity values and color distribution in
at least half of the images. Measurements of pH were often not available for
comparison, but the water labeled as s\'Iamp/forest runoff in the images was
located in the same area as highly colored swamp/forest runoff was observed
and sampled by both Livingston (1978) and Hydroscience, Inc. (1977).
. Livingston (pers. comm. 1978) stated that water color most closely
correlated with rainfall in East Bay and the flow rate of the A.palachicola
River. Proximity to the mouth of the Apalachicola River, also played a large
role in the classification and coloring of water types. Hater types of the
same color in different geographic areas may vary in water chemistry, but have
s imil ar spectral refl ectance characteri st ics. The color codi ng of water types
and land-use types was done on an image-to-image basis because individual
spectral signatures for the same class were quite different from scene to
scene. This difference largely was due to varying atmospheric conditions.
Colors assigned to the classes in one image do not necessarily represent the
same types of water in another image. The two runoff water classes, the river
plume, and the waters of the Gulf of Mexico would not necessarily represent
similar water types from image to image. Conjunctive water quality data would
also have helped to alleviate color- differences betvJeen possibly similar water
52
-------
I .
.
Figure 17.
... -- - - - -- --
Standard water color table for classified images.
Acidic Runoff
Diluted Acidic Runoff
Bay Water
1
Bay Water
2
Bay Water
3
Bay Water
4
Bay Water
5
Bay Water
6
Bay Water
7
Bay Water. 8
Bay Water
9
Bay Water 10
Bay WatE:r 11
Gulf Water 1
Gulf Water 2
Gulf Water 3
53
-------
types. Each image should be viewed independently in terms of the
procedure. Unlike regular color photographs, the colors in these
interactively selected and assiqned by the author while operating
computer.
color coding
images ~'/ere
the
Water color distributions were compared ~th 1976 National Ocean Survey
nautical charts and found to indicate bottom features only in the case of mud
fiats lining the delta of the Apalachicola River. These flats were observed
by the author at low tide. Bottom reflectance was, therefore, not of concern
to this project. There was also a need to determine whether sun angle had any
significant effect on the spectral refl ectance of water. Spectral response
versus classes were plotted and compared with images from opposing seasons.
Although the data were not substantial, due to the mean contributing variable,
it was expected that refl ectance woul d decrease with low. sun angl e in the
winter. Summer scenes would have high sun angle and higher spectral
reflectance responses. Comparison was difficult, for the same relative water
types were not present in most images. General observations revealed only
sl ightly higher reflectance values for the August 1976 scene in Channel 2.
Winter and spring data overlapped, and in a few instances overlaped the summer
data. Due to the small range (spectral resblution) in Channel 4 data, there
was no apparent diffference between data from different seasons. The effects
of the atmosphere were, without a doubt, present but were not subtracted from
each image before processing. .
. The water type of interest was swamp/forest runoff. This runoff was
easily distinguishable from other types when observed at the ground level
. because it was characteristically fresh, highly colored (coffee brown), acidic
(pH 4-7) water. It was always found to be significantly different spectrally
from other water types. As expected, in all four channels the swamp/forest
runoff had low channel means, compared to other water types (Figures 18-24).
Channell means were the lowest of all water types (19.33, Table 4).
The water type termed "dilute swamp/forest runoff" had sl ightly higher
channel means than the sit/amp/forest runoff in Channels 1, 2 and 3. Their
responses were generally the same in Channel 4 (Figure 19, Table 4). The
swamp/forest runoff was not very turbid and had a deep coffee color, while the
dilute swamp/forest runoff had a higher refl ectance because it had been
diluted with the brighter, highly turbid water of the Apalachicola River.
I
I
I
I
,
,
Reflectance generally increases with increasing levels of turbidity.
Figures 20, 21 and 22 illustrated that the river and bay waters had the
highest spectral reflectances of all other classified water types. Monthly
station means of water quality data for each image are presented in Appendix
A. While such means were derived data collected on one to several different
water sampling dates within the month, they are still indicative of general
bay conditions at any particular station. The type and areal extent of each
Landsat-derived bay water type is in agreement with average monthly water
quality measurements of Secchi depth, color, turbidity, and salinity. The pH
values do not follow the water type as closely. Landsat-measured spectral
54
-------
ACID
C
o
U
N
T
S
WAT
WATR01
4
01
02
CHANNELS
Figure 18. Computer-derived, four channel, spectral plots
of acidic runoff training fields, Image 5.
55
-------
Figure 19.
DLACID
c
o
U
N
T
5
01
WATR09
'"
02
CHANNELS
Computer-derived, four channel, spectral plots of diluted
swamp/forest runoff training fields, Image 5.
56
-------
PLUME1
c
o
U
.N
T
5
01
'212
CHANNELS
Figure 20. Computer-derived, four channel, spectral plots
of highly turbid Apalachicola River plume training
fields, Image'S. .
57
-------
BAV002
c
o
U
N
T
S
01
02
CHANNELS
Figure 21. Computer-derived, four channel, spectral plots of a type
of turbid Apalachicola Bay water training fields, Image 5.
58
-------
EBAY01
C
o
U
N
T
S
01
02
CHANNELS
Fi gure 22. Computer-derived, four channel, spectral plots of a type
of moderately turbid St. George Sound water training fields, Image 5.
59
-------
GULF01
C
o
U
N
T
S
01
02
CHANNELS
Figure 23. Computer-derived, four channel, spectral plots
of shallow turbid Gulf water training fields, Image 5.
60
-------
Figure 24.
GULF04
c
o
U
N
T
S
01
02
CHANNELS
Computer-derived, four channel, spectral plots of cl ear,
deep Gulf water training fields, Image 5.
61
-------
TABLE 4. LANDSAT FOUR CHANNEL WATER CLASS COUNT MEANS FROM ALL CLASSIFIED
IMAGES (Channels 1, 2 and 3 are on a compressed count value
scale of a to 127 and channel 4 ranges from 0 to 64).
C1 ass Channel 1 Channel 2 Channel 3 Channel 4
Swamp runoff 19.33 12.60 7.91 1.40
Oil ute runoff 22.04 15.25 9.34 1.40
Bay ~~ater 1 28.67 23.46 12.42 2.11
Bay water 2 29.82 22.50 12.30 2.15
Bay water 3 29.31 23.70 12.53 . 2.22
Bay wa ter 4 29.34 23.56 11.90 1.63
Bay water 5 26.79 19.37 10.15 1.64
Bay wa ter 5 28.01 26.24 8.72 0.86
Bay water 7 25.84 16.56 7.57 0.98
Bay water 8 21.85 12.67 5.54 0.49
Bay water 9 20.53 11 . 82 5.25 0.27
Bay wa ter 10*
Bay water 11 25.78 16.60 9.54 2.37
Gu1 f water 1 28.78 15.56 7.39 1.12
Gu1 f water 2 24.44 12.05 5.97 0.83
Gu1 f water 3 23.73 12.21 6.05 0.88
* Only occurred in one Landsat image.
responses appear to corre1 ate best wi th water color and turbidity, fo 11 owed by
Secch i depth.
FSU did not collect water quality information from the Gulf side of the
barrier is1 ands. It was, however, noted that as distance increase's from the
influence of the coast (islands) and approaches deeper areas, the percent
reflectance of water drops off dramatically in Channel 2 and moderately in
Channel 4. Count values for water in Channel 4 were always low as expected
(Figures 23 and 24). It is very likely that both turbidity and spectral
reflectance decreased with distance from the coast. Shallow coastal water
types were also identified and suspected of having relatively high levels of
turbidity.
TEMPORAL WATER COLOR (TYPE) DISTRIBUTIONS
After the water types were delineated and color-coded, the spatial
distributions of the classes were examined in terms of the seasonal existing
conditions at the particular time of image acquisition. An attempt was also
made to identify possible environmental problem areas. For the interpretation
of these images, it was important to keep in mind that Landsat, with minimal
penetration capabilitie~ into turbid coastal water, was only viewing surface
62
-------
water features. Features such as salt wedges characteristic of this estuary
could be inferred. However, the bay system is usually considered to be
vertically mixed.
The oldest image, 17 February 1973 (Figure 25), indicated relatively low
wind conditions, primarily because water patterns were well defined. Being
winter, the runoff was at an expected hig'h. The flooding of the river has
been found to be' out of phase with rain and associated high amounts of: 'rulJoff
in the East Bay drainage basin (Livingston 1975). Generally, due to ,,-
geographical distances and related atmospheric conditions, when it rains
during the summer months to the north in the upper reaches of the Apalachicola
, River, land in the vicinity of East Bay is dry. During the winter the
Apalachicola River is usually at low flow, but runoff is heavy from the East
Bay basin. Table 5 has average monthly river flows, collected by the USGS,
which also supported this statement (Graham et al. 1978b).
As in all observed images derived from r.1SS data, coll ected under low wi nd
conditions, the swamp/forest runoff was held against the eastern shore of East
Bay by ,the hydraulic head (pressure) produced by the easterly flow of river'
water from the Apalachicola River delta. The apparent primary sources of
swamp/forest runoff are Cash and West Bayous. Swamp/forest runoff. .
(Livingston, pers. comm; 1978) was also apparently being extruded, in an
easterly direction, from the mouth of the Carrabelle River. There is probably
less runoff present around the Carrabelle River than is shown in the
classified image, because water that was not acidic was misclassified due to
the striping effect.
At this time the swamp/forest runoff covered 16.34 percent or 1,120
hectares (2,765 acres) of East Bay (Table 6). Once the runoff reached Cat
Point it was diverted into St. George Sound. There are two probable causes
for this phenomenon. First, the net velocity of Apalachicola River water is
in general toward the east, and second, there is an underwater channel,
Bulkhead Shoals Channel, (see map in back pocket) which diverted water toward
the south.
Water types within the bays and sounds appeared to be uniformly
distributed. The barrier islands obstruct water leaving the Apalachicola
River, forcing it to the east, west or south. Surface water is observed
escaping to the south through Sikes' Cut, a channel in St. George Island.
This suggests that river water which entered Apalachicola Bay is quickly
forced into the Gulf by way of Sikes' Cut. The plume extruded from Sikes' Cut
is readily apparent in this image. Water quality data at Station 1B (Figure
16), just inside the cut, 1 ended credence to the hypothesis that fresh.,
nutrient laden .water is lost to the Gulf of Mexico through Sikes' Cut.
Station 18 had a high monthly mean surface nitrate/nitrogen value of 191 jJg/l,
an orthophosphate val ue of 8 jJg/l, an ammoni a readi ng of 16 J,lg/l, and alow
salinity ~aluE! of 2.3%0. This image was acquired during an ebb tid'e;....',
. /:" ~;~ <,' . -:.: ~~~-'>"
There is no appearance in thi s image (17 February 1973) of any. ',::,..
diversionary effect upon current patterns due to the Gulf Intercoastal
Waterway (see map in back pocket). There is, however, a visible boundary
63
-------
0'\
.p..
Fi gure 25. C1 ass ifi eo Landsat image (17 February 1973) of water color di str; but ions
in Aralachicola Bay, Florida under 10\'1 \'Jind and ebb tide conditions (source; 1.lil1 197R).
-------
TABLE 5. AVERAGE MONTHLY FLOWS FOR THE APALACHICOLA RIVER AT
BLOUNTSTOWN FOR PERIOD 1961-1976 (source; Graham et al. 1978b)
Mean
Oi scha rge Ratio to Mean
Month m3s-1 Annual Discharge
October 387 .55
November 391 .56
December 617 .88
Janaury 985 1.40
February 1,131 1.60
March 1,209 1.72 .
Apri 1 1,082 1.54
May 687 .97
June 580 .82
July 505 .71
August 496 .70
Se ptember 392 .56
X 705 1.00
between two water classes formed where a channel, running north and south,
extends from the center of the western side of Apalachicola Bay toward
Apalachicola Municipal Airport. Such a boundary may indicate that aquatic
communities in St. Vincent Sound are isolated to some as of yet undertermined
extent from the nutrient laden waters of the Apalachicola River. Sufficient
water quality data were not available for Station 1, but Station 2 data
indicated high nutrient levels. Apalachicola Bay water extended to Dog
Island at the time of image acquisition.
The second scene, 13 April 1973 (Figure 26), illustrates the same
. locations of swamp/forest runoff, namely West and Cash Bayous and the
Carrabelle River. Runoff from the Carrabelle River is traced toward the east.
The striping is less pronounced, at least for the bay water types. There is
no apparent swamp/forest runoff in Round Bay. The river's hydraulic head is
still keeping the runoff against the peninsula of East Point. This image
depicts a flood tide situation. It is plausible that the runoff is ponding up
in East Bay due to the .effect of the incoming tide.
. There were no apparent water boundaries caused by the presence of the
Intercoastal Waterway which transverses southward across Apalachicola Bay from
the river mouth. The river water has evidently f1 owed down the waterway in a
southerly di rection until it was pushed into an easterly di rection by more
influential bay currents. Apparently water leaves Apalachicola Bay, enters
St. George Sound, and extends to the area just north of Dog Island. There was
evidence that any water escape into the Gulf through East Pass between St.
George Island and Dog Island.
65
-------
TABLE 6. PERCENT AND HECTARES (ACREAGE) OF SWAMP/FOREST AND DILUTE RUNOFF
. FOR ALL IMAGES IN EAST BAY, APALACHICOLA BAY AND ST. GEORGE SOUND
Area and Image Numbers
Water Type 1 3 5 6 7 8 12
East Bay
6.26 6.83 14.66 4.67 13.40 9.47 11.94
Swamp 428.90 468. 18 1004.80 326.02 918.54 664.20 818.50
runoff (1059) (1156) (2481) (805) (2268) (1640) (2021)
10.08 4.06 12.73 .00 22.00 3.45 18.11
Oil uted 690.93 278.24 872.78 .00 1508..22 236.12 1241. 73
runoff (1706) (687) (2155) (O) (3724) (583) (3066)
0'1 Apalachicola Bay
0'1
06 .02 1.34 .12 . .04 .02 .04
Swamp 81 .00 3.64 304.56 28.35 . 8.50 5.26 8.91
runoff (200) (9) . (752) (70) (21) (13) (22)
1.17 .29 4.29 .00 .93 .29 1.49
Diluted 265.28 66.02 974.84 .00 211.00 64.80 339.98
runoff (655) (163) (2407) (0) (521) (160) (837)
St. George Sound
.03 .07 1. 72 .54 .61 .00 .01
Swamp 40.50 12.15 315.09 98.42 110.97 .00 .Hl
runoff (100) (30) .( 778) (243) (274) (0) (2)
.39 .72 6.21 .00 18.25 .47 8.42
Oil uted 48.60 132.44 1136.02 .00 3341. 25 86.67 1541. 43
runoff (120) (327) (2805) (0) (8250) . (214) (3806)
-------
0'1
.....
- --
Figure 26. Classified Landsat irnage (13 April 1973) of \'iater color distributions
in Apalachicola Bay. Florida under low wind and flood tide conditions (source; Hill 1978).
-------
The classification product generated from MSS-data acquired 21 December
1973 (Figure 27) depicts the distributions of water types at a time of low
flow in the Apalachicola River and moderate runoff from the East Bay drainage
basin. Areas color-coded red outside the study area cannot be positively
identified as swamp/forest runoff, but they may be highly suspected as such.
The necessary water truth is lacking. The dilute sw~mp/forest runoff in this
scene was apparently driven by the wind against the western shore of East Bay.
Shoal Bayou, Alligator Bayou and numerous smaller inlets are affected. Wind
broke up the many boundaries, which normally occur when there is little wind.
While swamp/forest runoff covers a greater extent of East Bay, once the
swamp/forest runoff enters Apalachicola Bay it is transported into St. George
Sound. The p'lume of the Carrabelle River is also still oriented in aiL"-- .,-
easterly d"irection. Apalachicola Bay waters enter St. George Sound and-:-'pass
again to th~~vicinity of Dog Island. Some water is moving out into the-Gulf
instead of the area north of Dog Island. An apparently exposed spoil bank is
present paralleling the Intercoastal Waterway just south of the town of
Apalachicola (Figure 27). St. Vincent Point is more exposed than in previous
scenes, indicating a low tidal condition.
Lines of foam are identified and misclassified as land. The foam lines,
convergence or water boundary conditions, are located on the Gulf side of West
Pass and off Cape St. George. The foam near West Pass implies that bay water,
once it leaves West Pass, moves westward along St. Vincent Island. Lines of
foam and debris often occur between fresh water and the more dense salt water.
In this case, the less dense fresh water is from Apalachicola Bay and the
saltier water is that of the Gulf of ~~xico. This image depicts a situation
in which an ebb tide is beginning to occur. Shoreline currents, on the Gulf
side of the islands west of Cape St. George, appear to be moving eastward.
Tidal records indicated that the tide was beginning to flood the bay at
the time the image of 3 March 1974 was acquired (Figure 28). The fiow of the
Apalachicola River at Blount~town was 923 m3/s, but the average for the
previous 10 days was 1,400 m /s (Graham et al. 1978b). Local runoff in
either East Bay or near the mouth of the Carrabelle River. One type of
highly turbid water is occurring in the major portions of East Bay,
Apalachicola Bay, and St. Vincent Sound. There is a dispersed plume being
emitted-from Sikes' Cut. Water has moved into St. George Sound from the bay
and is paralleling the mainland. Water distributions in the image suggest
that an influx of Gulf water through East Pass is forcing this water against
the mainland. As in previous images there is evidence of a large area of
water mixing (swirling) just to the east of the toll bridge going to St.
George Island in St. George Sound. Again, it is credible that water leaving
Apalachicola Bay through West Pass moves westward. A sizeable sediment plume
is traveling southward from Cape St. George. A longshore current is the most
probable cause.of this plume's direction. No apparent direction of shore
line current~- is evident. Of environmental concern in the Apalachicola Bay
area is the ,effect of the island supporting the toll bridge from Cat Po'irit'to
St. George Island on the neighboring water bodies. Two channels exist, orie
north and the other south of the island. As inferred from this classified
image, surface water is evidentally transported into St. George Sound in a
path parallel to these channels. The result is the eddy located on the east
s ide of the i sl and in St. George Sound. Thi s part i cul ar event is apparent in
68
-------
m
l.O
Figure 27. Classified Landsat image (21 Decemher 1973) of water color distrihutions
in Apalachicola Bay, Florida under moderate northeast wind and prbbable ehb tide
conditions (source; Hill 1978).
-------
. '.
-....J
a
Figure 28. Classified Landsat ima~le (3 March 1974) of water color distributions in
Apalachicola RaYt Florida under low wind and beginning flood tide conditions
(source; Hill 1978).
-------
other images'.when wind conditions are low.
St. George '$o.und.
.'...,
The island influences mixing in
.. *;
, -
. '.
The extensive oyster bed protruding southward from St. Vincent Point could
conceivably offset the distribution of currents in the area of West Pass.
Di stri but ions of water type? in several Landsat scenes support thi s
hypothesis. Circulation patterns in St. Vincent Sound are also visible, but
have a low degree of organization. Detailed in situ current studies must be
. conducted to fully understand circulation patterns in both of these areas.
Another excellent example of the dramatic effect of wind on this shallow
bay is evident in the fifth image, 23 October 1974 (Figure 29). Although some
clouds are present in southwest corner of this scene and the eastern portion
of St. George Sound, general current patterns can still be examined. It is
not uncommon for the wind to be out of the northeast at this time of year..
The wind again distributed swamp/forest runoff into Round Bay, Shoal Bayou,
Alligator Bayou, and all other inlets along the western shore of East Bay.
This runoff covers 35.40 percent or 2,427 hectares (5,992 acres) of East Bay.
Cloud shadows over water appear as swamp/forest runoff in east St. George
Sound, because the results have a spectral signature that is nearly identical
to that of the runoff. It is, therefore, difficult to discern the direction'
of the plume from the Carrabelle River. Wind is known to have an intensive
mixing effect on this bay because it is only about four meters at its deepest
poi nt and has along fetch. Ingl e and Dawson (1953) emphasi ze that .
wi nd- i nduc ed flow c an have a greater effect on bay hyd rodyn ami cs than the. ti de
for short periods. Gorsline (1963) stated Apalachicola Bay tides are much
affected by winds and under strong wind conditions the astronomic tide could
be completely obscured. There are no visible homogeneities (brood areas that
have the same spectral signature) in the surface waters of the bay under the
wind conditions existing when the scene was recorded. Very little water
appears to be moving from the bay into St. George Sound. ~Jater from
Apalachicola Bay and St. Vincent Sound is being forced out of Sikes' Cut and
West Pass.
\~ater from St. George Sound is being blown into Apalachicola Bay, as
indicated in the image and by the surface salinity reading of 33.7%0 at
Station 1C. The boundaries which normally delimit the edges of various water
classes and/or underwater man-made structures are not visible, apparently
because they were totally broken up by wi nd action. The net flow of water
outside the islands is also in a southwesterly direction. .
Of the eight classified images, the scene acquired 26 February 1975
(Figure 30) best illustrates the boundaries between water types. Wind
conditions ~~re low, and as in almost all previously observed images with
water boundaries, water types and current patterns are easily discernible.
The river flow was 2,062 (about 2.9 times the annual mean flow), and the tide
was near ebb slack (Graham et al. 1978b). Swamp/forest runoff is protruding
from the East Bay drainage basin and the Carrabelle River. The runoff in East
Bay is again positioned against the East point peninsula. The western pQrtion
of East Bay,::~along with associated bays, bayous and inlets is free of .run9.ff.
The plume of'tb'eCarrabelle.River is moving in an easterly direction, 6n~'e.',
agai n enteri ng ,St. George So un d.' . ".
71
-------
~
I~
Figure 29.
Classified Landsat image (23 October 1974) of water color distributions
Apalachicola Ray, Florida under strong northeast wind conditions
(source; Hill 1978).
in
-------
......,
(.oJ
.-
Figure
in
30. Classified Landsat image (26 February 1975) of \mter color distributions
Apalachicola 8ay, Florida under low wind and near ebb slack tide conditions
(source; Hill 1978).
l_..... "'-
......--.
-------
'Water...type s ',In:' J\pa l-ach ko l.a .,Bay. and' ne.i ghp~Qri ng sounds .were uni fonn ly '.
di ?tri~bu.t'ed:; Al ~hough ..water' "q:ua l.:.ity' s ampl i ng . stat'i ons:' we're not adeq.u~te 1 y'
d i s'persed"t'o' tol retf,"aa:faJ. from "ea'ch" ofi.the' wa.ter 'types- (-.t-hey.'may' n'eve;rr'be') ,.
the water data make it possible to verify some of the inferences. The water
color-coded ~tan and dark and light brown indeed had the lowest Secchi
readings, the highest color (omitting swamp/forest runoff), the highest
turbidity, and the lowest salinity of the water types in the bay. The water
color-coded dark green represented a water type with high Secchi readings, low
color, low turbidity, and high salinity. The water color-coded gold generally'
had a higher Secchi reading, a bit higher CPU (color) reading, lower
turbidity, and higher salinity than the water color-coded dark green
(Livingston, pers. comm.; 1978). Gulf waters probably had even lower color
and turbidity, but also had higher salinity and Secchi depths than all the
other water types. These data demonstrate that the water at the mouth of the
Apalachicola River is the most turbid and sediment laden and has the highest
reflectivity of all the bay waters. As the distance from the Bay mouth
increases, water quality parameters change in the following,manner; turbidity
decreases, Secchi depth increases, color decreases, salinity increases, and
spectral reflectance decreases.
The area of swirling (mixing) is again recognizable in the western portion
of St. George Sound. The bay water is moving eastward and has reached the
area just north of Dog Island. Apparently, very little water from St. George
Sound is moving through East Pass and into the Gulf. As is observed in other
images, the greatest amount of water leaving Apalachicola Bay travels by way
of West Pass. Si kes I Cut agai n contributes to the loss of bay water to the
Gulf. Monthly mean water quality data indicate that for February the water at
Stations 1A and 1B was relatively high in nutrients. Surface orthophosphates
were 2.2 ~g/l and 3.7 ~g/l respectively while surface nitrate/nitrogens were
59 ~g/l and 117 ~g/l, respectively. Shoreline currents at Cape St. George
cau~e a sediment plume to be expelled into the Gulf. Water ejected from West
Pass is movi ng to the west. .
Several water boundaries were perhaps caused by the presence of man-made
underwater structures. Even if the locations of the underwater structures
were not known through other sources, their locations could be established
through close scrutiny of water type boundaries found on the classified image.
Bulkhead Shoals Channel (see map in back pocket), oriented parallel to East
Point between Godley's Bluff and Cat Point, is diverting the swamp/forest
runoff into St. George Sound. A boundary is visible over the approximate
location of the channel. The boundary which is visible in the vicinity of the
Gulf Intercoastal Waterway evidently indicates that the river water heading
south toward St. George Island is diverted to the east and west by the
. causeway. This interpretation of the image requires substantiation by further
water sampling investigations which will better define the hydrodynamics of
the area. The third, and most striking boundary occurs just south of Green
Point. It was first thought that the aforementioned channel leading the
airport from the bay had caused this phenomenon. However, the channel is not
in the immediate vicinity. Less dense fresh water from the river probably met
the more saline waters of St. Vincent Sound, resulting in this particular
boundary. Such a- boundary should be further investigated, for it is highly
74
-------
likely thatlarg~ oyster and other aquatic communities in St. Vincent~,S,ound
are cut off fr,oni'the nutrient laden waters of the .ll.palachicola River' and,Bay
System under these river flow, wind, and tidal conditions. ,'.:"
The seventh image, 20 July 1975' (Figure 31), is indicative of the effects
of clouds, cloud shadows, and haze. Clouds obstructed most of the features in
this image. This is' a summer scene where, as expected, the river flow 't'las
low. There is a very small plume at the mouth of the Apalachicola River. A
slight boundary is noted south of Green Point and there is a plume off of Cape
St. George. The turbid waters are obviously confined by the barrier islands.
Swamp/forest runoff is probably present, but is masked by the effects of
weather. Clouds are a serious problem in gathering this type of remote
sensing information.
'Lastly, the image acquired 19 August 1976 (Figure 32) directly supports
, the observat.ions and inferences made from the 23 October 1974 image (Figure
25). The wind is again out of'the northeast. The distributions of the
swamp/forest runoff in both scenes duplicate one another. The swamp/forest
runoff transverses most of East Bay (30.05 percent; 2,060 hectares; Table 7).
Round Bay, Shoal -Bayou, All igator Bayou, and all of the otber- inJ~ts. on the
western shore of East Bay are inundated with runoff. Again the runoff from
the Carrabelle River is dispersed in all directions, as opposed to its,
sometime easterly flow. The usual water boundaries are not visible i,n the bay
and the 1 i kel i hood is great that most of the bay water is bei ng forced through
West Pass. There is no readily discernible surface plume being emitted from
Sikes' Cut. The h'ater is well mixed as expected off Cape St. George. The
nearshore Gulf Waters appear to be flowing in a southwesterly direction.
Water from St. George Sound has pushed into Apalachicola Bay. ' These
observations are qualitatively supported by the water quality data.
This portion of the study demonstrates the value of this technology to
identify, delineate, and monitor the distribution of water types in an
estuarine system. Field personnel at the Florida Department of Natural
Resources Oyster Sanitation Station, are responsible for monitoring coliform
bacteria and other pollution in Apalachicola Bay. They have corroborated the
water patterns in these Landsat images, and the Landsat data verify their
theories on coliform distributions. Such theories were a result of ten, plus,
years of experience and sampling in the bay. The coastal engineers at the UF
have already begun to utilize these Landsat data in the initial planning
stages of a large scale hydrodynamic modeling program for the East Bay
d ra i nage basi n and bay property. FSU personnel stated that the Landsat
information will give tremendous spatial resolution to their hydrographic
measurooents and numerical model of Apalachicola Bay, Florida (Graham pers.
c omm.; 1978). '
DISCRIMINATION OF LAND-USE ACTIVITIES
This sectjon describes the results of the effort to use Landsat data to
detect land-u~~e activities with an anphasis on silviculture practices/: ,This
research was :c.onducted to detect and monitor potential envi ronmenta': ;i[!iP.acts
of land-use"cH'ange in the East Bay drainage basin on water quality in;,ji+ie bay
75
-------
~
0'\
';;:t:,~\ii;~!I~~fY <,:,~~~ij ~iz~:~J~~.
.' C ,:\ 'f, '~'if~'d-' n>;,Ci,J".yr' ..;;(.;..:~::" ':, . ""." d'''{'~ '."
, -'c~~~~ .~nt.~,~~~~i~i~li~\,:/, ~:::':::':" " . ~~".""'~'"" ~...
. ~~.)--r ~... ~ \~ ~ ~ .~ ~l ,.,...,_...,:..;r,."',i.,;,?,.~,~.;::','.~...,r.,':': -<-. Y'*' .
, ~, ,I"~V',. "". ",''-',,~. ",,'."a ---'t' ,_., ~ - ':"-" - ..ij;"''''>o",'':'' , ,,:" ~:,.~"-::~;.i'
, ,.;{iiii;:;~'if;'j,{f»~,~i" :}" ',' /~;,,(~\,:;'f~ 't~.
""~.. ""..~ .. .4.; -" - " . c'-" ~-Łi.":/,.o <'o""'~~~;? ~~~~
".. ~Jf_~.r,.~,~r.~:~.:~ -~~::,~)'~~~E",/"'};~> ;,~
"~"'::L""'-- .' ''',.,,-~ ' -
- ;~ ~;~~~-:(fi,~~'<#> - -
'/ :." ~. ,,/~~(t[~f':?~~~J~~:'J' . "
-*t~~!;~~~~ ,~~ i~~;:' ..
,~, ",.-"~~tC',I,':',:J~ .~~.,/ f!::v.' c~:'X");"'\;:;'" .
.: ,~,:.;~~!\:.":--~'.'r;.'.'-'/~~'%.''; ,~:~~~' ,0/ .~. "..,., ~:; ~ ..~.-' f~' 1~$r.~~~i-
. ','" "-.:c~. >...;"" ,,",'- ".:~'fu~:{Y~i(~J.:t,:~\;~
,[ "'~. :'l~:-..~ ~'h1:~I~"'S:S-">:" ~"-:;.',;:.:.;. 'L;'.;,..;}~':~a~~\f!:'~
";~{!"{;J~:~!~~,~~i*~;~' .
.r 'I OI,i~.. ' , ~. - ~,
j~_. f
.~: Ł~:~~
.,
., ."... "..,.
. . ~ ".: ~~~y
-",,~~ ,
~_r~,
,~_:~
.
';',.
.'
----. --
Figure 31. Classified Landsat image (20 July 1975) of water color distributions
in Apalachicola Bay~ Florida under much less than optimal abnospheric conditions
(source; Hill 1978).
-------
"
......
"I
Figure 32. Classified Landsat image (19 August 1976) of water color distributions I
in Apalachicola Bay, Florida under strong northeast wind conditions (source; Hill 1978).
l-.- ,
"
----.........-.
-------
'.' " TABLE 7., LANDSAT FOUR ,CHANNEL Lf\ND CLAS~ MEANS
',.' ,. (IN..COUNT,~) F~OM 'ALL CLASSLFIE'O' IMAGES :....,~ ' '..
. .. ., hi ;..~..t.' '..;.~~-,;~,~f" - ...' . .' ~ .~~. ~...'.r" " .; ..:-":C~,~'~ . I.. I"
. , ''':'~ '" _." --~.,.. "' : -,.. '. . . . ,,- '.. . .'t .
.,'
",<; ,.~;~~~1~,'.'.:
Class
. ~. ~." " ~...; .. I -" . '... .' . . .
: t'..., "." ,"d.
':; r ': ....: :. ," .~':J' -. .... .. ""-"".
Channell Channel 2
. ~ . ...
", ~. ,,'It . .. /',,',,:"-:;.".:'" . ." ". '.,"'"
Chah'nel . 3 Channel 4'" "
. . . . .
Urban
Ma rs h
Natural
Sand
Agricul ture
Bank
Revegetated
Cl ea r-c ut
35.16 21. 07 39.77 18.77
24.21 18.48 26.43 13.32
22.05 15.21 25.70 13.80
60.76 65 . 62 65.21 30.28
36.15 35.64 42.15 20.25
21.85 17.82 17.38 7.51
20.80 15.65 23.27 15.14
22.18 17.74 17~18 9.86
system. As previously mentioned, the lumber company which owns most of the
land in the East Bay drainage basin suppl ied the bulk of the ground truth,
information used in this segment of the study. Six Landsat-l scenes (Figures,
33 and 34) were analyzed to produce the results in this section (Table 3).
The parameters monitored for this study were the Landsat-1 derived four
channel spectral reflectance responses of selected land-use types within a
particular scene. The major land categories (classes) of interest in the
East Bay drainage basin are clear-cut (within one year), revegetated,
swamp/forests, marshes, and roads. As previously mentioned, fourteen final
land categories were selected. Figure 34 is the land color-code table for all
land classified images.
This section describes the detectability of individual land-use classes'
and also presents a temporal interpretation of the classified images. This i~
followed by an assessment of the reliability of said data.
Urban areas in this particcular geographic area were found to have uniqu~
bimodal spectral signatures which made them easy to discriminate from all the:
other classes. Small rural towns, such as Apalachicola, often demonstrate
this characteristic signature (Stone, pers. comm., 1977).
"
Next to clouds, sand beaches had the highest spectral reflectance.
:.
Average values for sand in Channels 1, 2 and 3 were 60.76, 65.62 and 65.21, .
respectively. Sand and urban classes were combined into one class since they:
both had highly refl ective spectral signatures and represented non-vegetated,', .
areas. ":
, .. .
There was very little agriculture in the study area.' Small plots may
exist, but the area is not conducive to traditional farming practices,
especially in the East Bay drainage basin. There is, however, an area of
specific interest to several regul atory agencies and universities along the
78
-------
Figure 33. Classified Landsat images of land-use activities in
East Bay drainage basin (A - 17 February 1973; B - 13 April 1973;
C - 21 December 1973; D.- 23 October 1974 (source; Hill 1978).
79
-------
-
.
.
Unclassified
Urban
Agriculture
Marsh 1
Marsh 2
Marsh 3
Marsh 4
Marsh 5
Revegetated
Clear-cut
Swamp/forest 1
Swamp/forest 2
SKamp/forcst 3
~llId banks
Figure 34. Classified Landsat images of land-use activities in
East Bay drainage basin (E - 26 February 1975; F - 19 August 1976).
Included is the land-use color table for classified images
(so urc e ; Hi 11 1978).
80
-------
Apalachicola River. This is the M and K Ranch located on the west bank of the
Apalachicola River. Although not of direct interest to this project, the 'area
was sampled. This particular land-use activity demonstrated high count values
in Channels 1, 2 and 3. The area was found to have a unique spectral. .
signature and was assigned a maroon color. The ranch was classified in three
scenes; 13 April 1973, 21 December 1973 anmd 26 February 1975. On 21 December
1973, the ranch class conflicted with the urban and sand classes. For this
date, the ranch was put into the urban and sand class and was colored white.
The ranch was classified, but it was cropped from the final image because of
picture size limitations, and it was of no particular importance to this
particular project.
Ma rshes and swamp/ forest categori es were very d i fficul t to di sc rimi nate
from each other. Marshes are interspersed with swamp/forest communities, at
least along shorelines of East Bay Seasonal characteristics, such as die-back
and browning of vegetation, often hindered separation. The computer-derived
four channel spectral plots (signatures) were of assistance in determining
their separability. Figures 35 and 36 clearly demonstrate the spectral
similarities of marsh and swamp/forest classes. The closeness of spectral
means in all four channels of marsh and swamp/forest communities makes
separation very di-fficult. Best marsh discriminations v.'ere achieved for the
image 13 April 1973 (Figure 33-B). The marsh classes were separated from each
other primarily by their different appearance on the image generated from the
display tape and an examination of-their spectral statistic. Vegetation may
have been starting to IIcome alivell at this time. In this image it is easy to
discern brackish marshes (color-coded 1 ight green) lining the bayous along the
delta. These areas were verified as marshes through USGS maps. Plant species
unique to the flood plain of the Apalachicola River were also found to be
spectrally unique in the spring.
During the classification of land-use features a unique class appeared in
a few images at the tips of the Apalachicola River delta. This feature is
very pronounced in the 21 December 1973 image (Figure 33-C). Tidal data and
nautical charts indicate that these features are either a product of depth
penetration and the return of energy from arefl ective bottom through several
centimeters of turbid water or are exposed mud flats. The low count value in
Channel 4, 7.51, indicated that this feature was probably water. .
Roads often formed excellent boundaries around the fields of interest and
frequently were the only points of reference in the study area. Some 500
miles of road lie within the lumber company's land in the vicinity of the East
Bay drainage basin. The roads in the study area are usually composed of a
highly reflective material. They may, however, be masked by overhanging trees
or, due to the low spatial resolution of the satellite, may be
indistinguish~ble from neighboring vegetation (i.e., marshes, revegetated
areas). Roads may also have appeared as a marsh class, for the channels
.lining the roads often generate marsh communities (Figures 35 and 37).
Most roads are paralleled by drainage channels often 2 to 3 meters in
width and approximately 2 meters deep. The channels drain swampy areas to
provide suitable growing conditions for pine seedlings. The road and channel
81
-------
MARSH
.-. --......-
- ~...
. - - - J
I
I
~
c 1
o
u I
N
T
S
-1
I
MRSH06
MRSH03
01
4
CHANNELS
Figure 35. Computer-derived, four channel, spectral
plots of all marsh training fields from Image 5.
82
-------
NATRL
c
o
U
N
T
S
NA11.01
NA11.03
01
CHANNELS
Fi gure 36. Computer-derived, four channel, spectral plots of
all swamp/forest training fields from Image 5.
83
-------
SAROBA
C
o
U
N
T
S
01
2
CHANNELS
03
SAND01
ROAOO3
ROAD01
ROAD02
BANK03
04
Fi gure 37. Computer-derived, four channel, spect ral plots
of all sand, mud bank, and road training fields from ImageS.
84
-------
system is oriented primarily in a north-south direction, with the net water
movement being to the south.
The various stages of clear-cutting are sho~m in Figure 38. After the
construction of access roads and associated drainage channels, the underbrush,
such as, Titi, is cut down to gain access to natural timber. Once the natural.
stand of timber is removed, the slash, stumps, and brush are plowed into the
soil. When a plowed field is dry enough to support pine trees, seedlings are
planted on the mounds formed by plowing. A natural grass-type ground cover is
usually established within six months after plowing. The trees mature and are
ready for harvesting in 25 to 30 years. For most of the area under
investigation it took more than six years to establish the resemblances of a
climax community (Figure 38). However, current forest management practices
prevent the e~tab1ishment of a natural c1 imax community on the plantations.
The clear-cut and revegetated areas \~re easily separated and classified
due to their spectral differences. Mean spectral plots (signatures) of these
two classes (Table 7) are similar in Channels 1 and 2, but differ
significantly in Channels 3 and especially 4. Count values from the recently
cl~ar-cut areas are very low (9.86) in the near infrared (Channel 4) due to
the lack of ground cover. - However, when the revegetated fields began to
resemble natural communities, it became difficult to separate the classes on
the basis of their spectra. A few fields, due to seasonal variations and
similar stages of succession, were sometimes classified as natural, and many
months later, again resembled a revegetated field. The areas clear-cut in
1973 and 1974 were classified as clear-cut in this image, because they did not
have time to develop established ground cover. .
The Information Transfer Laboratory at NASA's Goddard Space Fl ight Center,
in a cooperative project with the Weyerhauser Company (Williams and Haver
1976), attempted to use Landsat data as suppl ementa1 input to a forest
inventory system. Their study area was in North Carolina. They experienced
difficulty in classifying clear-cut areas ~th Landsat data. Their data had a
low 54 percent of agreement with-air photointerpretation data. The main
reason for this difficulty was the fact that ground cover had established
itself in the span of over one year bet\~en Landsat scenes. By this time, the
clear-cut fields had the spectral signature of pine plantations. Winter
scenes were found to be good for outlining hardwood and pine forest canopies.
Spectral signatures ~lere also extracted for clear-cut and replanted areas, but
these broad categories could not be statistically divided into finer
categories such as age, vigor or density. By using a combination of winter
and summer data, a total of 8 channels with an emphasis on Channels 3 and 4,
they were able to break the replanted fields into three crown closure
categories:
closed c~nopy pine - 100 percent canopy
partial pine canopy closure - some exposed ground vegetation
- open pine canopy - 50 percent canopy, 50 percent ground vegetation
85
-------
A. Newly cleared
C. Six months after planting
E. Six years after
planting
I.
I
B. Plowed
,-
~.
- --..
.~- . '~: .:~~~" . 'c. .
-------
Revegetated areas \~re not broken into finer categories due to the lack of
detailed information available to the author at the beginning of the research
in this EPA report. Such information is avairable and, as in the Goddard
investigation, it may be possible to further break the revegetated class into
finer subcategories. This research monitored changes in land-use activities,
primarily forest management practices, in the East Bay drainage basin.
TEt~PORAL LAND-USE D ISTR IBUTIONS
Classified Landsat images are discussed in this section. An
interpretation of temporal land-use changes in marshes, swamp/forests and
other natural areas are discussed, but the emphasis is on the silviculture
activities in the East Bay drainage basin.
Clouds are present in the northwest corner of the 17 February 1973 image
(Figure 33-A), but fortunately they did not obscure the area of interest.'
Marsh communities along the Apalachicola River and the swamp/forests within
the East Bay drainage basin are not readily separated on the basis of their
spectra. This is probably due to a combination of spectral similarities of
vegetation types during the winter and possibly the low sun angle causing
merged signatures that at other times of the year are .separaol e. .
This scene indicates that extensive recent clear-cutting has occurred
north of East Bayou. It is also inferred that the middle to upper reaches of
the basin were clear-cut at an earlier date (1971) because the majority of the
area is classified as revegetated. Lumber company data suppo~ted this
finding. A few areas known to have been replanted are classified as marsh.
Grasses present during the first stages of replanting are evidentally
spectrally similar to marsh grasses. Again, this similarity is probably due
to the winter die back of the vegetation.
It is likely that as the trees in the replanted fields mature and develop
a canopy, spectral reflectances increase, especially in Channel 4. The older
more mature pine plantations no longer overlapped spectrally with the marsh
classes.
April appears to be a prime time for the discrimination of various marsh
communities. The second classified scene, 13 April 1973 (Figure 29-B) depicts
marsh species (color-coded light green) lining the delta. Due to their
geographic location, these marsh species are probably more salt tolerant.
Further up river along the Apalachicola flood plain, fresh water plant
communities are color-coded a bright yellow. Due to minimal ground truth for
the delta, one area which is erroneously labeled clear-cut remains
unexplained. Roads are relatively well defined because their highly reflective
sandy surface results in a spectral signature markedly different from that of
the bordering lush vegetation.
It is apparent that in the two months which lapsed bet'Neen the acquisition
of the above scenes, the major lumber company successfully completed
clear-cutting the timbered areas that were partially cut in the first image.
87
-------
Timber harvesting in swampy areas is dependent upon the weather conditions.
Therefore, only portions of'fields may be clear-cut at anyone time. Numerous
areas are either very thick (i.e., stands of Titi) or too swampy to risk
getting heavy equipment stuck. These areas are left alone. There is a
speckled pattern of misclassified pixels over the entire image which
represents an overlap between revegetated and marsh communities. A greening
of vegetation in both areas at the same time resulting in similar spectral
signature could have caused such confusion.
It is inferred that ground cover is quick to invade and grow in clear-cut
fields during the spring. This cover, however, died back in the winter,
returning the fields to a condition which spectrally resembles the clear-cut
a rea s ~
Winter is not the optimum time to spectrally discriminate marsh from
swamp/forest communities. In the 21 December 1973 scene (Figure 33-C), the
roads are masked by dying vegetation and are classified as marshes and
swamp/forests. Further ground truth is needed to verify the brown delta
features in terms of emerged mud flats, submerged mud flats or marsh grasses.
. Ve ry 1 i ttl e new cutting is evi dent in th i s image. Lumber company data
verify that relatively little cutting occurred at this tiille in 1973. Several
clear-cut fields from the 13 April 1973 (Figure 33-8) image now resemble
revegetated fields. Lumber company officials state that ground cover is
normally well established within six months after a clear-cutting operation.
This research revealed that under the temporal resolution of the available
Landsat scenes, the clear-cut fields acquired ground cover within at least 8
to 10 month s.
The marsh communities in the delta proper are still spectrally unique in
October, but as demonstrated in the 23 October 1974 image (Figure 33-0), it
was very difficult to discriminate marsh areas from swamp/forest communities.
Portions of the delta marsh are again classified as clear-cut. Confusing
classifications certainly warrant careful ground truth investigations and
perhaps ~th more time a reclassification of this area would result in a
better separation.
Most of the area classified as'clear-cut in the previous image, 21
December 1973, are classified as revegetated in this scene. A few areas were
labeled clear-cut, but these areas did not appear in the data provided by the
lumber company. These few discrepancies observed between lumber company and
Landsat data may lie in the data supplied by the lumber company, the
classification technique, or in the actual field conditions. These particul ar
fields may not have been replanted and/or possessed poor growing conditions.
. The same areas of concern gave similar conflicting results in the 21 Oecember
1973 image. In the fi fth image, 26 February 1975 (Fi gure 34-E), the marsh
species along the Apalachicola River are relatively well defined. Clear-
cutting in the East Bay drainage basin has apparently ceased. Areas of recent
cutting in the previous scene (23 October 1974) still appear as clear-cut, but
revealed minor amounts of regrowth in each.
88
-------
The last of the images, 19 August 1976 (Figure 34-F) provides evidence of
resumed cutting in small areas just north of West Bayou. This land is owned
by one of the other four lumber companies in the area.
Landsat-d~rived acreage values for the basin are within those supp1 ied by
the lumber company (Table 8). Acreages computed from the images is moderately
higher than that supplied by the lumber company~ because this particular
company owns most, but not all, of the drainage basin. Other lumber companies
are also operating in the basin area. Livingston1s data implies that \'~ater
qualtiy was at its worst level in 1974 \'mich coincides with the peak of
clear-cutting activities during 1972 through 1974. The Landsat classified
images verify 1974 as the year of heaviest cutting during this period.
Livingston (pers. comm.) has reported that East Bay is presently in a state of
recovery. He greatly encouraged the study of more recent Landsat images for
the purpose of monitoring a continued recovery in water quality, as well as
related land-use activities.
As the research effort to utilize Landsat data to temporally monitor
land-use changes in the East Bay drainage basin progressed, significant
changes were also observed to the east along the Carrabe1le River. The land
along the- Carraberle-River Was classified' using' spectral signatures derived '
from ground truthed training fields in the East Bay drainage basin. This is
being a form of signature extensions. As with the East Bay drainage basin~
s il vi culture acti vit i es were the only ma n-i nduced 1 and-use changes apparent
along the Carrabe1le River. The northern extremes of this river lie \'/ithin
the relatively undisturbed Apalachicola National Forest.
, It is inferred from the 17 February 1973 image (Figure 33-A) that large
areas of revegetated land border both sides of the Carrabelle River. The
purple area along the southern reaches of the river, just north of the to\'In of
Carrabe11e~ are misclassified as marshes. The 13 April 1973 image (Figure
33-B) suggests that no new cutting has occurred along the Carrabelle River~
except'perhaps for a small area east of the town of Carrabelle. More cutting
is observed in the 21 December 1973 scene (Figure 33-C). Large forested
sections were cut between 21 December 1973 and 23 October 1974 (Figures 33-C
and 33-0). Just to the east of the northern most portton of the cutting in
the Ea st Bay d rai nage basi n ~ 1 umber .company act ivi ti es have pi cked up in the
New River drainage basin. Only minimal new cutting occurred during the
following four months based on the changes noted between the 23 October 1974
image (Figure 33-0) and the 26 February 1975 image (Figure 34-E). The 19
August 1976 image (Fig~re 34-F) discloses extensive accelerated clear-cutting
operations along the west side of the New River. Fishermen have reported a
recent decline in catches within the area of St. George Sound affected by the
Carrabe11e River plume (Livingston, pers. comm.; 1977). The reports of these
fi shermen certai n1y wa rrant invest i gati on.' La nd-use act i vi ti es are very
probably linked to water quality in East Bay and Apalachicola Bay, Florida.
SIGNATURE EXTENSION
The magnit ude of thi s project generated thoughts of the poss i b i1 i ty of
signature extension over time. That is to say, if the spectral signature
89
-------
TABLE 8. PERCENT AND HECTARES (ACREAGE) OF SUBMINOR, MINOR AND MAJOR WATERSHEDS
IN EAST BAY DRAINAGE BASIN
5ubmi no r ~1i no r Ma j 0 r
Watershed Wa tershed Percent of East Watershed
Watershed (Ha Ac re s) (Ha Acres) Bay Watershed (Ha Ac re s )
Cash Bayou 11786 (29101)
Bea r Creek 802 (1981) 2.50
Cash Creek 10984 (27120) 34.25
Cash Creek (small) 2293 (5662)
Sand Bank Creek 2162 (5338)
Rake Creek 2792 (6895)
High Bluff 2941 (7262)
Cash Creek ditch 179 (443)
Cash Creek ditch #2 616 (1520)
lO Cash Bayou ditch 820 (2025) 2.56
a
West l3ayou 16114 (39788)
Whiskey George Creek 16114 (39788) 50.25
Whiskey Geo. Cr. (small) 6295 (15544)
So. Juniper Creek 657 (1623)
Doyle Creek 3576 (8830)
TO\'Je r Road ditch 4992 (12325)
Hest Bayou (small) 594 (1466)
Round Bay
Montgomery 510ugh and 1959 (4838) 6.11 1959 (4838)
Saltwater Creek
Sand Beach Branch 1389 (3430) 4.33 1389 (3430)
Dot Gri d \vatershed Grand Total 100.00 32069 (79182)
Crnnputer-generated acreage Grand Total 32645 (80604)
-------
of a specific class is identified, it can be used to classify that feature in
all other scenes. This turned out to be unreliable. While reasonable mean
spectral signatures are acquired for all classes (Tables 4 and 7), the
individual signatures were quite different from scene to scene or date to
date. Corrections related to sea state, atmospheric, and sun angle conditions
can help alleviate most in-scene differences, but are as yet not incorporated
into the DAS software. Land and water are readily separated from one another
in different scenes using the previously described supervised classification
technique. Due to the relative ease of separation of specific classes in this
study, research should be encouraged to produce techniques (i .e., unsupervised
classification, which is presently underway at several research centers) that
would make classifications, in the near future, an assembly line process.
RELIABILITY OF CLASSIFICATION
It is a .difficult task to determine the reliability of classification
demonstrated in this report. The project was conducted without the aid of
aerial photography and presented the opportunity to determine the near stand
alone (with a minimum of ground truth data) capability, if any, of using
Landsat data to accurately detect and monitor desired land and water features.
This -would have been best accomplished-by a comparativeustudy where aerial
photographs were available during all stages of Landsat classification.
As mentioned by Williams and Haver (1976), human variables due to the
interpreter always enter this type of study. Such variables are "fatiguell,
the interpreter's ability to detect gradual changes in color or cover type and
the ability to make consistent decisions. Still another unavoidable problem
exists because Landsat averaged conditions over each pixe1 (0.44 hectares; 1.1
acres). Wi 11 i ams and Haver (1976) stated that compari sons shoul d be made in.
terms of percentage of "agreement" and not percent "correct".
EPAls desire was simply to use Landsat to synoptically, and quantitatively
where pO$sible, detennine if there was environmental damage in the area. This
goal was achieved, but first several procedures had to be followed to
detennine the reliability of the Landsat classifications and associated
acreage data. The DAS system has various programs that aid the investigator
in this verification process. One such program is tenned SCORECARD. This.
program assigned to each of the possible classes gives the percentages of
pixels which are in a specific original training field. SCORECARD is,
however, biased because it decides how well an original training field chosen
by the interpreter was classified. In other words, the investigator pretty
well knew what it was before the classification process began. The program
SCORECARD was only run on the third set of data (due to time limitations) to
determine the reliability of the classification. Figure 40 is an abbreviated
example of the product generated from SCORECARD. Sand classes (numbered 2 and
4) were classified 96 and 100 percent. Mud banks along the river were 82
percent. Marsh classes (numbered 5, 7, 12, and 16) varied with percents of
100, 98, 41, and 45, with the latter two values resulting in spectral
similarities with the other marsh classes. Swamp/forest classes (numbered 6,
13, 14, 15, and 17) were classified with the following percentages: 0, 94, 55,
61, and 68. As expected, these classes are either in conflict with other.
91 -
-------
- ...
CT-RVG
. .
C
o
U
N
T
S
01
CHANNELS
RVEG01
CCUTe1
03
04
Fi gure 39. Computer-derived) four channel) spectral plots
of all cut and revegetated training fields from Image 5.
92
-------
+-
+-
p ['" c ~ i'il (Lli S S .T F r F !) Ji': T r, C L ~ ~ S
+
I:
;>
3
~
c::;.
7
q 1 (\ 1 1
1213 l/J 15 1~ 17 1~
!.i
':
++-+-+-++-+++++-++-+++-+++++++++-+-+++-++-+-++++-+++-+-+-+-++-++++-++-+-++++-+++-
I +-
;:: ..
::< ..
II +
5 +-
" +-
7 +-
~ +-
Q +-
t(\ +
- , 1 --+-
Ii? +
1 3 +
1 LJ +-
1" +-
1 ~ +
1 7 +
1 ~ +-
1° +
? (, +-
21 +-
22 +-
;q +
;:>u +-
25 +
"Ph +
27 +-
21' +-
;>q ..
30 +-
';( 1 +-
~2 +
3~ +
~IJ +
7;5 +-
y' .~
,h
~2 ? I: f-
100,
1 (1<)
c; ') c;n
qp. 2
2 iJ I 1 1 C) 1. (I 1 10 51
7 1 6':; 3 <1 1 ~ 2 7 1
1 ? ~ 1" ?c;; 1;; :V
2 q~
? IJ 7 g tq LIt
1 '14 1 ?
? p, Cj"= 1 c;
- ..'
;:,::.. h 1 5 c:;
-
(' 1 '5 1Jl! 1 115
? ;; ;:> (j bH ~2
10 gO
? 2
7
3h ..
+-..+++++-+++-++++++-+++..++++++++..++..+++-++++~+++++++++++++..++-+
+- r) 2:~ II C; S 7 .'I, 9 1 () 1 1 t .2 1 3 1 4 , C; 1" 1 7 1 f:
Figure 40. Computer-derived SCORECARD for Image 5 (21 December 1973)
indicating the percentage of original training field pixels assigned
to each of the classes under consideration.
93
-------
* * pt, I.( T T ,,; Ii n F
RV(A;.(O **
19 20 21 ?2 ?~ 24 2~ ?~ ?7 ?R 2<1 30 31 32 33 3a 35 3~ 37
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
3
27 29 1 Q 1 '3 1 1
2 SQ 7 4 18 q 1
? 93
5 75 15 '5
2 9b 2
1 3 1 7?- 2 3 4
2 n '12
I3CJ 7 3 1
4 B? f. 1 5
7 ,~2 7 lJ
p 75 1 7
q9
3 'p 5
1 () 4 "( 75 8
1 3 1 ~ ~n b
2 1 1!! 2 ~ 73
q R4
4 q 1
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
19 20 21 22 23 24 2S2~ 27 2P 2g 30 31 32 33 34 35 36 37
3 c; 7 5. P r J ,\; T S COP Q E C T I") " I T (} F
55&2.
TOT;\L POIl'JTS.
6ll.('5 PEP r:E(\jT
ACCUR,\CY.
AVERAGE ACCUR6CY RY CCiSS
72.P3 PEP. cniT
Figure 40 (Continued).
94
-------
swamp/forest classes or marsh classes. Neither case had a great effect on the
acreage estimates of the drainage basin.
Of specific interest, are the revegetated and clear-cut classes. The
classification percentages for the three revegetated classes (numbered 8, 9,
and 18) are 11, 65, and 90. Thi s fi rst and rather low cl ass ificat i on is not a
great problem when it was observed that this first revegetated class is mainly
merged with the third revegetated class, and to a small extent, with a
swamp/forest class. The second revegetated class is also competing with
swamp/forest classes. This overlap was an expected response after viewing
the overlapping character of the class spectral plots. The sole clear-cut
class (numbered 11) has 98 percent of its original pixels classified into the
expected or predicted ~lass (i.e., clear-cut).
The average land class accuracy was 61.00 percent. These results,
although biased, are quite good. Hhen land and water classes v/ere combined,
3,575 pixels from a total of 5,582 were correctly classified for an overall
percent accuracy of 64.05 percent. The average accuracy by cl ass for the
third image is a respectable 72.83 percent. This means that certain
individual classes were easily discriminated while others were more difficult,
resulting in a .lower overall accuracy but a fairly high individual class
accuracy.
A final check on the reliability of the land classification products was
made through a comparison of acreage estimates derived from the Landsat
classifications with those provided by the lumber company (Table 9). Acreages
. computed from the classified Landsat images are not in direct agreement with
that provided by the lumber company (Table 9), but the trend in cutting
activities was- apparent from both the lumber company and Landsat data.
Several lumber companies are active in the basin and the Landsat values may
include several of the areas under their control. Table 8 shows close
agreement between the East Bay drainage basin acreage provided by the company
(a dot grid estimate of the map provided by the lumber company) and the
computer-generated basin acreage figures. However, more detailed ground truth
must be incorporated into any follow-up investigation to develop more accurate
estimates of silviculture activities.
It is far more difficult to determine the relaibility of the water
classification because there are no concurrent water truth data to correlate
with the '(later types. Three i ndi rect methods were used to determi ne the
accuracy of these classes. First, the biased SCORECARD program was run on
water classes from the third image (21 December 1973). Second, a comparison
of monthly mean station water quality data was made with -\'Iater color in the
images. Third, acreage estimates of individual water masses (i.e., East Bay,
Apalachicola Bay) derived from the images were compared with acreage estimates
acquired by using a standard stratified random dot grid over a 1:250,000 map
of the area. These acreages agreed, indicating that at least a good land-
water boundary had been selected. In general, water types were much easier to
derive final classifications for than were the land classes. In this study,
spectral signatures of water, although nearly always one count apart, were
often more distinct than land signatures. It is felt that the boundaries
95
-------
TABLE 9. PERCENTAGES AND AMOUNTS OF lANDSAT AND lUMBER COMPANY-DERIVED
CLEAR-CUTTING DATA FOR EAST BAY DRAINAGE BASIN (32,645 hectares)
landsat derived revegetated data are also provided.
19 August
1976
landsat Lumber Company Landsat
,Cl ea r-Cut Cl ea r-Cut Revegetated
Estimate Estimate Estimate
8.96 * 7.48 15.49
2925 ** 2443 5057
(7223)+ (6033) (12486)
10.31 6.24 20.20
3364 2039 6595
( 830 7) (5034) (16284)
6.03 3.00 14.52
1968 963 4740
(4860) (2378) (11705)
4.43 1.80 15.59
1446 586 5091
(3571) (1447) (12570)
5.12 1.63 19.08
1671 532 6228
(4127) (1314) (15379)
7.35 1.67 20.36
2459 546 6646
(6071 ) (1348) (16411)
Date of
Image
17 February 1973
13 April
1973
21 December 1973
23 October 1974
26 February 1975
* Percent of East Bay drainage basin.
**Area in hectares
+ Area in acres
associated with water types were indeed true features of the bay at the time
of the overpass. Another reason for the ease of discrimination of water types
was probably that the individual water types are often relatively homogeneous
over large areas. land training fields, on the other hand, were not
homogeneous because of the numerous features (i .e., trees, brush, grass,
water) that contribute to a mass spectral reflectance from a land community.
SCORECARD was also run 'for the water types in the third image, 21 December
1973 (Figure 40).
96
-------
The water types were classified to an average accuracy by type of 80.05 .
percent. This SCORECARD illustrated that it was easier to classify water than
1 and.
The distribution of water types in this investigation was corroborated by
other water quality investigations in Apalachicola Bay. Long-term water
quality data (post 1972) collected by Livingston et al. (1974), agree with the
location of several of the water types identified in the Landsat images. A
survey conducted by Hydroscience, Inc. (1977) on 17 March 1976 showed the same
general location of swamp/forest runoff as in several of the Landsat scenes
under similar environmental conditions. Hydroscience also produced tidally
assumed surface water color maps (one at high and one at low tide) which
indicate swamp/forest runoff in the same general area of the bay as evidenced
from the satellite images. The Landsat-derived water patterns also correspond
with the Hydroscience output of a bay water color model (Hydroscience 1977).
Landsat is used in this study to monitor water color which can be used to
infer other water quality parameters relating to specific types of water.
The MSS cannot directly sense salinity, pH, and dissolved oxygen.
. However, in this geographical setting, ph and disssolved oxygen (00) values
corrrelate well with water color (Figures 41 and 42). As tidally averaged
surface values of pH and 00 decrease, color increases. Other water quality
parameters (i .e., Secchi depth) could have been considered, but Hydroscience
chose to concentrate on pH and DO. To the point where mixing and dilution
have a great influence on water types, inferences of values of other
associated water quality parameters remain valid Graham et al. 1978).
Figures 43 and 44 are indicative of the water quality maps available at
the start of this project. These maps were constructed from a simple linear
extrapolation between water quality data collected at known water sampling
stations. This is a very poor, but traditional way, to derive the
distribution of water types within the bay. Landsat produced a synoptic,
grided, instantaneous image of the entire bay. The water quality maps as they
are presently being derived, produce an inadequate and incorrect picture of
water patterns in the bay. Never under numerous environmental conditions did
the water patterns in the bay, as represented ~n FSU.s maps, closely resemble
those in the Landsat images. Landsat images have been shown to represent a
more accurate synoptic representation of water color patterns in the bay.
However, the most accurate presentation would be derived if boats were
acquiring water quality samples at the same time as the satellite overpass.
These water quality data could then be used as calibration points during the
classification stages of the Landsat imagery.
The results of this investigation have made researchers aware of a more
accurate. method to. map water color distributions in Apalachicola Bay, Florida.
It has also opened to them new ideas concerning "'/ater transport mechanisms in
the bay. The transferability of this technique has already been proposed for
other estuarine systems around the nation (i.e., Lake Pontchartrain,
Louisiana, and Chesapeake Bay, Maryland).
97
-------
7 X
Q)
01
0
-
Q)
> X
~ x~
o
"'U
.... x
I
\I) 6
+-
c:
:)
:r:
Q.
5
pH v s. Co lor of Su doc e
Water
on 3/17/76
*
L mixed-source water
x
x
x
100
200
300
CPU Color Units - Tidal Average
Figure 41. pH versus color of surface water in Apalachicola Bay,
Florida on 17 r~arch 1976 (source; Graham et ale 1978b).
Q)
C1
o
v 8
>
~
o
"'U
~ 7
I
"
C1
E 6
o
o
Q)
u
~ 5
....
:J
V>
9
o
o
x
xx
DO vs. Color of Surface
x
x
Water on 3/17/76
X
Locean
4
x
x
x
I:i:)- two measurements
t upland runoff
100 200
Surface Color - CPU Units
300
Figure 42. Dissolved oxygen versus color of surface water in
Apalachicola Bay, Florida on 17 March 1976
(source; Graham et al. 1978b).
98
-------
lD
lD
'----+----I----+----~----+---_7_---+----~----+----S----+----&----+----7----+----8----+----9----+----1----+----.
T ... T
T
I
I
+
I
!
I
1
i
I
I
.
I
I
I
T
Z
I
I
i
.
I
T
I
I
3
T
T
J
T
+
I
I
I
1
~
I
r
!
.:U.19 ..1.
... .......&.00
..... . ......
....... ......
'.&i~*I.............111
".".."'7.19."
""."..".....'
1:11111111111111:'
I""..""'.'"
.................
l'I:::III~ia*II:I.
.................
APALACHTCI1 LA ..................
:)IH~ ....... EAST BAY..
... ,., .......................
... ..... ........................
.. ".&.&9"........'............'
..: 11........'.1:"7.~0.1.'.'
.. . .......... . ..... ... EAST ..
."'."'.7.'0'.'.""""'...' prINT ...
ADALAC~JcnLA.".."."""."".."""'.'. ......
.............................. .1..'.
.....,..,......... ... ......... .. ....
............................................
... ..................,...............,...""......,."..
.,.... "."""""""'7.lq"'IZ""""""""r.'I'1'11'1'1'"
.................................. .................... ....
[[[I.."'.",.,.,..
[[[
"""""""',.,"",""""1""'.,.,.,.,':'.,........................
..............,........ ....................... ............................
"""."."".7.~~..".""..,.,.,."".",..,."",.I'..'..".""..""'."
[[[
.:::::::::::::::::::::::::1::1::::::::::11::::1::::::.:::::1111:::1:1111::::111111.
[[[
..............0[[[
"":""".""":""011.'...':'."...".".".':".'.':"1":.:""1:."""
..... ............. I"'. ...... ...........~mo'l~ I..'" .. I. . .... ...,...
. ST. ~1"C~~T"I'I".I""."""II".""'."""W~III~.I~C11~1.11.1"""."'7.50""'."'"
I !~LANr """.""'."""""'...J"I"ICII.III09~~~8~~III"SI....""."".",""..'"
1 """"I""""I"""'.""'11111~9qq99Aq~9A89qA9al1'8"1~"."':'."'."'.'
1 ...................... .p.LAc~r~rLI RAY xxxXxxt~~9a9AeRII""""" ........
I ".'.""'.'."."""'.."'I'119A9~~XYXXXXXXYXXXAXA9988111"".'..".".'
~ """'."'."""".'.""'II.q9~~YyxxYXXXXXXXXXXAX9Aqaql...'..".,,'"
T """'."""".".'..".IIIA98~XYXXY~XX(XXAYXXXXxXX9A9811.'I~'II"'.
I '."""..""""""."'..'9q9QX~XyyxtvYtVXXvxXXXtXXQq881..IIIIII11
T '."."..""""""'.'.IDoee~AYXXXXXXYxXXvXXXxvXXvXXge8q.~~
T ,.",......t""""'."'.~leij~XXXXXXXXX(XXXxxx++,XXXXXXe8G8AA
. ""'.3~'.""'I""""'I~~AeexxxxxxXVXvXXYX.+.+++.xxX7.6qAij ST. GfOPGE
I '.""".'."""""""""'I~A8exxXXXxvXYYXXVXY',,+.'+++XXXXXA8 TSLANO
T "'."""""""""""'I.'I.A9xxXXAXX~XYvXXxXv..+......'XXXX
I ... "'I"""'..."""""i~eeexxxYAVXXYXx.xxxx..+...6.0O+X
J ''''''''.,11' ,.I"""ulllesqx xx XXXXvXXX x xY Y X +". . . '" .'
& Wl ST "",."" ....... "11J1I9"S XXX X XXv xx xx 7 .tJ~ x .... .. ..
T ~AS~ ..I..'..'......I.liIAQ6XXXxXYXYXXXXxxxxx",..
T '.I""."."'.~Q.epxxxxxv.xxxxXX~XXYX++'
J """"..""»E8AAYXXXXXXXXXXXXXXXXX +.
I ,.....1. ....~~eeexxxxxxXXXXXxxx +
+ ..........1:I1II8ae) XXXXVtxn SYI([S CUT +
; """I~Ae9AXXXXXXY I
I "UJII"SSXXXu T
-------
.-.
a
a
._---.----1----.----2----t----~----~----4----+----r,----.----b----+----1----+----~----.----q----.----1----.--_e.
I ... t
T III" IT
r 1'10.00 ... I
I III' ""'310.00 I
. 1111" .........
I ...111.. ...... I
I 15 0 . 0 QI .11 .
I APAL~I..HICOLA RA~ SYSIE'1 IIUllllllllllllllllt1ll I
I 1111..1..130.001111 I
1 1II199&999QIIIIII 1
t .18898&9989981.mlm T
I 8A999Q99999gellllll I
I 89999999999991111 t
I 98ee999999Y9999911 I
. 999898810~.~089~ee .
I A89989998A8999888 !
I 99a89998989Ra989'1 I
t APALACHJr.O LA 8111!88ge8e9898A9889 I
r ~JV<~ 9999998 nr.'1LA8eQ9geflQQ{jlj99994>.8"99"99999QeIl98 9ge9Q9 T
J 'I'!99899899geQA8119A98eeQ9989PAX XXXXXX I
1 9"!99Q999999999999998e8e8'199XXXXXX XXXXXXX I
I AA999Agea99B9999ge899989998XXXXXX.1(XXXXXXXXXX I
l ~XX XXXXXAXXX89999889699999999998G8899XXXXXXXXXXXXXXXXXXX. l
I ..XXXA xxxxxxxxxXX~XXXXI0~. 0089999Cj998"9'1'18aXXXXXxxXXXXXXXX'"'''' I
""AvXXXXX)XXXXXXXXXXXXXXXXX9a99999'19989998XXXXXXXXXXxxx........... t
! ....... xxx XXX~XX~X x x Xx XXXv xxx x X X 811899811AA998XXXX XXX x XXXX'" ........... I
I """""XXXXXXXXXXXXXAXXXXXXAXXXX99'199Cj99XXXXXXXXXXXXX"""""""" J
. .......... ...X XXXH XXXXXXXXX XX xxx xx XAX988eXXxX XXXvXXXXX' ...... """'''. .. .
I """""""'XXXXXxvXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"""'."""'....... J
I .. .... .. .. .. .. .. .. t .. .. .... 7 r;. 00)( xx '( x x X A X X XX)'X X X X X X X X X X X. '( )t':( 'I( X xy x x x x.. .... .. .. ........ t.. .... . . . . . . . . . . . t
r .. t . .. t . .. . .. .... t . .. .. .. .. .+ X X X X)( ;( XX x.(..... x XX xx x): xx 'I( X'.( X)' X'( XX X '()eX XX" .... ...... .. .. t.. ... . . . . . . . . . . . . . I
I .. ... tt4-', ++ ...t...... t.++XX)t'XXXXX)("'XX'XX,(XX)(XXXX~X)()(XXX)(X)(+"""..t..'."......... ....... I
". .t.... .t....... tt .+..)(XXX,(XX)('(X,(X'(XX'J(X)(''''~XX)(XXXX)(XX........+t. t........... ........ 4
I t + .. .... ........ .. .... .. .... .. .. "".A X X X '( xxx ~ XX xx x x '( '( XX)( X xx X XX)( t....... 'to"" t.. t" . . . . . . .. . . . . .. . . . . . . . . I
I .. .. .. .. ...... .. .. t .. .. .. .. .... t .. + 1.)".)( XX X X X xx XX): Y :x. X x J(X xx X t ...... .. .. .......... .. t.. .. . . . . . . . . . . . . . . . . . . . . . . I
I . .... .... .. t t .. .. .. .. .... .. ..+ t .. '() x x. x x x '(' x X AX Y 1.. X X). 'J(.. ...... .. .......... ............ . . . . . . . . . . . . . . . . . . . . . . . . . . I
I .. .... .. .... t.. .. t" + .. . t . ... ++. . +1. XX X xxx X.. +-. ........ ...... t.". t . .4. . . . . .. . . . . . . . . . . . . .. . .. . . . .. t
.. 'S T. v I ~IC f N" .. ... ... . . . .. .... .. +- .. .. ... ..... ... ..... . + ... t. .. t ... t.. .. ... . +-+........ . . . . . . . . . . . . . . . . '+ 0. 00. . . . . . . . . . . ..
t I SLAU 0 .. .... +- . t". +-... .. . ............. ..... . t. .... . ...... .. t... ... . . . . . . .:. . . .. . . . . .. . . . .. . .. . .. . .. .. .' I
T '..t.. ...t ............+.... .+tf.t+................. .... ... ........ ...... ...... ..... t
I ........................ ,..\DALAC~IC('lLA OAY ..................................... I
I . +.. . ... .... + t....... .. ....... .......... .. .... ... +. . . . . . . . . . . . . . . . . . .. . . . . .. . . . .. . . I
:J + ..+........ ...t....... ...t................... ...... ... ... ..... .... ........ 5
r . .... ... .... ....+-. ............ ......+-..... +..... .... ...... .... .....,. ...... I
T .......... +-. +... .......... +... t..."."..." + +".... .... .. ... ... ... ...... . ... I
r . +- +..................... t t. + +....................... +-. .. ... . .... . "'" ... I
I .....+.....t..........+........++.....+.+....t..................... I
. .+."').00....."...+...+..+..+..........+..++...........40.00.. Sf. GEOQ:GE .
I +.. .. . .. . + . . .. . t f ... .. . . . . . .. ... .. . .. +-.. . . . . .. .. . +- .. +-. ... . . . . . . . . . . . . . . . . . . . . . t Ii) L A NO ' I
I f........'...'........+........+++-......"++..+-.+-... ... ..... ..... t
1 +.. ...++..........+....++.++...+.+."'++.++"'.+.+.+.....",0.00.. I
I ...........+.....+.........+............................... I
h Wf5T .t........t..+.t.........++++........fiU.OOt........ f)
T PAS~ .e.........................+-...................... I
I .......................................... I
I ...+.+.t.+..+......."..'+...+....++... .. t
i ....:::::::::::::::::::::::::+..+ SYKES CUT !
I ....+.i....i......". I
I ...............+ I
'----'----1----'----2----'----~----'----~----'----5----'----&----'----1----'----8----'----'1----'----1----'----.
!'YJ
Figure
44. Hater color map from simpl e 1 inear extra pol ations of \oJater
quality elata, February 1975 (provided by Livingston lQ78).
-------
REFERENCES
American Society of Photogrammetry. 1975a. R. G. Reeves, A. Anson, and O.
Landen (eds.). .rA.anua 1. of Remote Sensi ng. 1. Keuffe 1 and Es ser Company.
I.
I
1975b. R. G. Reeves, A. Anson, andD. Landen (eds.).
Remote Sensing. 2. Keuffel and Esser Comapny.
~1a n u a 1 0 f
Anderson, D. M., J. L. McKim, W. K. Crowder, R. K. Haugen, L. W. Gatto, and T.
L. Marlor. 1973. Applications of ERTS-l imagery to terrestrial and
marine environmental analyses in Alaska. Proc. 3rd Earth Resour. Technol.
Satellite-I. IB: 1575-1606.
Barker, J., C. BO-hn, L. Stuart, and J. H-ill.
processing, a near real-time application.
Survey Symp. lC: 2063-2074.
1"975-. Landsat digital data
Proc. NASA/Earth Resour.
Barnes, P. W. and E. Reimnitz. 1973. New insights into the influence of ice
on the coastal marine environment of the Beaufort Sea, Alaska. Proc.
Symp. Significant Results Obtained from Earth Resour. Technol.
Satell ite-1. IB: 1307 -1314.
Bl anchard, B. J., and R. W. Leamer. 1973. Spectral refl ectance of water
containing suspended sediment. 17th Proc. of Symp. Remote Sensing and
Water Resour. r~anagement. Am. Water Resour. Assoc. 1: 339-347.
Bowker, D. L, P. Fleischer, 1. A. Gosink, W. J. Hanna, and J. Ludwick.
Corre 1 ati on of ERTS mul t i spect ra 1 imagery wi th suspended matter and
chlorophyll in Lower Chesapeake Bay. Proc. Symp. Significant Results
Obtained from Earth Resour. Technol. Satellite-I. lB: 1291-1298.
1973.
Carlson, P. R. 1974. Surface currents along the California coast observed on
ERTS imagery. Proc. 9th Int. Symp. Remote Sensing Environ. 2: 1279-1288.
Colberg, M. ~., T. S. Dietrich, and D. M. Windham.
economic values of Apalachicola Bay, Florida.
Poll. Cont. Admi n. Contract 14-12-117.
1968. The social and
Final Rept. to Fed. Water
'Dawson, C. E.
Florida.
1955. A contribution to the hydrography of Apalachicola Bay,
Publ. Tex. Inst. Mar. Sci. 4: 15-35.
101
-------
Egan, W. G. 1974. Boundaries of ERTS and aircraft data within which useful
water quality information can be obtained. Proc. 9th Int. Symp. Remote
Sensing Environ. 2: 1319-1343.
Elterman, l., and R. B. Tool in. 1965.
Geophysics and Space Environment.
Force Cambridge Res. laboratories.
Atmospheric optics.
Shea l. Valley (ed).
7 . 1-.7 . 36.
In: Handbook of
Cambridge. Air
Eppler, W. G. 1974. An improved version of the Table look-up Algorithm for
pattern recognition. Proc. 9th Int. Symp. Remote Sensing Environ. Univ.
Michigan. 793-812.
Estabrook, R. H. 1973.
Bay. M. S. Thesis.
Phytoplankton Ecology and Hydrography of Apalachicola
Florida State Univ.
Fleicher, P. 1973. Investigation to relate chlorophyll and sediment control
in waters of lower Chesapeake Bay to ERTS-l imagery. Semi-Annual Rept.
Goddard Space Fl ight Center. N 73-20385.
Gorsline, D. S. 1963. Oceanography of Apalachicola Bay. In: T. Clements
(ed.), Essays on Marine Geology in Honor of K. O. Emery. Univ. S.
Carolina Press.
Graham, D. S., K. DeCosta, and B. A. Christensen. 1978a. Stormwater runoff
in the Apalachicola Estuary. Florida Sea Grant Final Report. HY-7802.
, J. M. Hill, and B.A. Christensen. 1978b. Verification of an
estuarine model for Apalachicola Bay, Florida. Proc. ASCE Symp. on Model
Verification.(In press).
Hill, J. M. 1978. landsat Assessment of Estuarine Water Quality with
Specific Reference to Coastal land-Use. PhD Thesis. Texas A&M Univ.
pp. 209.
Hill, J. M., and T. M. Dillion. 1976.. A unique and effective oceanographic
surface truth monitoring program for correlations with remotely sensed
sateilite and aircraft imagery. Texas Engineering Expt. Sta., Texas A&M
Univ., Tech. Bul., No. 76-2.
Ho lte r, M. R. 1967.
6: 376-383.
Infrared and multispectral sensing.
Bioscience.
Hunter, R. E. 1973. Distribution and Movement of suspended sediment in the
Gulf of Mexico of the Texas coast. Proc. Symp. Significant Results
Obtained from Earth Resour. Technol. Satellite-I. IB: 1341-1348.
Hydroscience, Inc. 1977. The effects of forest management on the water
quality and aquatic biota nf Apalachicola Bay, Florida. Buckeye Cellulose
Corp. Rept. Perry, F. L.
Ingle, R. M., and C. E. Dawson.
State Board of Conservation.
1953. A survey of Apalachicola Bay.
Division of Oyster Culture.
F10rida
102
-------
James, W. P. 1970. Airphoto Analysis of Waste Dispersion for Ocean Outfalls.
PhO. Thesis. Oregon State Univ. p.152.
, and M. O. Schwebel. 1972. Application of remote sensors to water
qual ity management in the coastal area. Technical Report RSC-38. Remote
Sensing Center. Texas A&M Univ. p.138.
Jensen, N. 1968. Optical and Photographic Reconnaissance Systems.
John Wiley and Sons, Inc.
N.Y.
Jerlov, N. G. 1948. Optical studies of ocean waters.
Deep-Sea Expedi t ion.. 3 (1) .
Rept. of the Swedish
1951a. Optical studies of ocean waters.
Deep-Sea Expedition. 3(1): 3-57.
Repts. of the Swedish
. 1951b.
(3): 122-218.
Optical measurements of particle distribution.
Tellus 3
" . h 1953'd Jnf1uenc~ Qf sU~Renqed and dissoJved matter on the
transparency of sea water. Tellus 5 (1): 59-63.
Jerl OV, N. G.
Compa ny .
Jordan, C. L. 1973. Cl imate. In; J. 1. Jones, et al. (eds.).
Knowledge of the Eastern Gulf of Mexico. State Univ. Syst.
Oceanogr. IIA. 1-IIA.22.
1968.
Optical Oceanography.
Amsterdam, Elsevier Publishing
A Summary of
Florida Inst.
Kalle, K. 1937.. Meerskundliche chemische utersuchungen mit hilfe des
zeisschen pulfrich photometers. Ann. Hydr. Mar. Meteor. 6; 276-282.
1938.
66: 1-3. "
Zum problem der meereswasserfarbe.
Ann. Hyd"r. Mar. Meteor.
. 1939. Die farbe des meeres.
et Proc. Verb. 109; 98-105.
Cons. Perm. Int. Explor. r1er. Rapp.
Kemmerer, A. J., and J. A. Benigno. 1973. Relationships between remotely
sensed fisheries distribution information and selected oceanographic
parameters in the Mississippi Sound. Proc. Symp. Significant Results
Obtained from Earth Resour. Technol. Satellite-I. 1B: 1685-1696.
Kiester, A. R.
reptil es.
1971. Species density of North American amphibians and
Syst. Zool. 20: 127-137.
Klemas, V. 1973. Satellite studies of turbidity, waste disposal plumes and
pollution concentrating water boundaries. Proc. 2nd Conf. Environ. Qual.
Sensors. 7; 57-87. "
103
-------
Lepley, l. K., G. Calderon, and J. R. Hendrickson. 1973. Oceanographic
mapping of structure and dynamics of the northern Gulf of California by
the use of spectral modeling and ERTS-l. Proc. Symp. Significant Results
Obtained from Earth Resour. Technol. Satellite-I. 18: 1349-1356.
Lind, A.O., E. B. Henson, and J. Pelton. 1973. Environmental study of ERTS-l
imagery: lake Champl ai n and Vermont. Proc. Symp. Si gni fi cant Resul ts
Obtained from Earth Resour. Technol. Satell ite-I. 1A: 643-650.
livingston, R. J. 1974. Field and laboratory studies concerning the effects
of various pollutants on estuarine and coastal organisms with application
to the management of the Apalachicola Bay System (North Florida, U.S.A.). .
Florida Sea Grant Report. R/EM-1
1975. Resource management and estuarine function with application
to the Apalachicola Drainage System. Estuarine Pollution Control and.
Assessment. 1: 3-17..
. 1978. Short and long term effects of forestry operations on \vater
quality and biota of the Apalachicola Estuary (North Florida, U.S.A.).
Florida Sea Grant Program. Final Rept. (Unpublished).
, R. L. Iverson, R. H. Estabrook, V. E. Keys, and J. Taylor, Jr.
1974. Major features of the Apalachicola. Bay System: physiography,
biota, and resource management. Florida Sci. 37(4): 245-271.
, and G. C. Woodsum. 1976. Special program for ecological science
( SPECS) : Summary of capabil it i es. (Unpub 1 i shed) .
Mairs, R. L., F. J. Wobber, D. Garofalo, and R. Yunghans. 1973. Application
of ERTS-1 data to the protection and management of New Jersey.s coastal
environment. Proc. Symp. on Significant Results Obtained from the Earth
Resour. Technol. Satellite-I. 1A: 629-634.
Maughan, P. ~1., A. D. Marmelstein, and O. R. Temple. 1973. Application of
ERTS-1 imagery to the harvest model of the U. S. Menhaden industry. Proc.
Symp. on Significant Results Obtained from the Earth Resour. Technol.
Satell ite-I. 1B: 1405-1412.
Maul, G. A., and H. R. Gordon. 1973. Relationships behleen ERTS radiances
and gradients across oceanic fronts. Proc. 3rd Earth Resour. Technol.
. Satell ite-1 Symp. IB: 1279-1308.
Menzel, R. W., and E. W. Cake, Jr. 1969. Identification and analysis of the
biological value of Apalachicola Bay, Florida. Rept. to Federal Water
Pollution Control Administration. Contract. 14-12-191.
, N. C. Hulings, and R. R. Hathaway. 1957. Causes of oyster
depletion in St. Vincent Bar, Apalachicola Bay, Florida. . Proc. Natl.
Shellfish Assoc. 48: 66-71.
lC4
-------
, , and 1966. Oyster abundance in Apalachicola Bay,
Florida, in relation to biotic associations influenced by salinity and
other factors. Gulf Res. Rept. 2(2): 73-96.
National Aeronautics and Space Administration.
Technology Satellite Data Users Handbook.
Center. 71504249.
1972.
~,!ASA .
NASA Earth Resources
Goodard Space Flight
National Estuary Study.' 1970. U.S. Dept. Int. 'Fish v/ild. 'Serv., Bur. Sport
Fish. Wild.. and Bur. Comm. Fish. ~Jashington, D.C. v-3.
Odum, W. E. 1970.' Insidious alteration of the estuarine environment.
Am. Fish. Soc. 836-847.
Trans
I
i
Od urn, E . P. 19 71 .
Ph il ad e 1 ph i a.
Fundamentals of Ecology.
W. B. Saunders Co.,
Osborne, J. A., and G. R. Marzolf. 1972.
. photosynthesis in turbid reservoirs.
. --. . Resources - Re~~ar_ch Jnst.
Pirie, D. M., and D~ D. Steller. 1973. California coastal processes study.
Proc. 3rdEarth Resour. Technol. Satell ite-1 Symp. 1B: 1413-1446.
Effect of spectral composition of
Contribution No. 106. Kansas Water
Pluhowski, E. J. 1973. Remote sensing of turbidity plumes in Lake Ontario.
Proc. Symp. Significant Results Obtained from Earth Resour. Technol.
Satellite-I. 1A: 837-846.
Pritchard, D. W. 1967. ~/hat is ahestuary: physical viewpoint~ In:
Estuaries. G. H. Lauff, ed. . Amer. Assoc. Adv. Sci. Publ. No. 83.
Washington, D.C. 3-5. .
Ramsey, R. C. 1968. St udy of the remote measurement of oceancol or.
Rept. NASA Contract NASW-1658.
Fi na 1
Riley, G'. A. 1956. Oceanography of Long Island Sound.
Production and utilization of organic matter. Bull.
Coll. 15: 324-344.
1952-1954. IX.
Bingham Oceanogr.
Ruggles~ F~ H. 1973. Plume development in Long Island Sound observed by
remote sensing. Proc. Symp. Significant Results Obtained from ERJS-1.
1B: 1299-1304.
Schertz, J. P., J. F. Van Domelen, and S. A. Kooster. 1973. Aerial and, .
satellite photography-a valuable tool for water quality investigations.
Proc. Am. Soc. Photogramm. 2: 883-905.
Schubert, J. S., and N. H. Macleod. 1973. Digital analysis of Potomac River
Basin ERTS imagery: sedimentation levels at the Potomac-Anacostia
confluence and strip mining in Allegheny County, ~laryland. Proc. Symp.
Significant Results Obtained from Earth Resour. Technol. Satellite-i. 1:
659-664.
1G5
-------
Steller, D. D., and D. M. Pirie. 1974. California nearshore processes
ERTS-I. Proc. 9th Int; Symp. Remote Sensing Environ. 2: 1251-1278.
Stortz, K., and M. Sydor. 1974. Remote sensing of western Lake Superior.
Proc. 9th Int. Symp. Remote Sensing Environ. 1: 933-937.
Stumpf, H. G. and A. E. Strong. 1975. Surface correl ati on in the Great Lakes
as observed by Landsat-1 August 1972 to December 1973: Southern Lake
Mi ch igan. Proc. NASA Earth Resour. Survey Symp. lC: 1973-1988.
Sykes, J. E. 1970. Report of the Bureau of Commercial Fisheries Biological
Laboratory. St. Petersburg Beach, Florida. U.S. Fish. \.Iild. Servo Circ.
22: 187-217.
Szekielda, K. H., and R. J. Curran. 1973. Biomass in the upwell ing areas
along the northwest coast of Africa as viewed with ERTS-l. Symp.
Significant Results Obtained from Earth Resour. Technol. Satellite-I.
1: 1385-1402. .
Szekielda, K. H., S. L. Kupfennan, V. Klemas, and D. F. Polis. 1972. Element
enrichment in organic films and foam associated with aquatic frontal
systems. J. Geophy. Res. 77 (27): 5278-5282.
Teleki, P. G., G. A. Rabchevsky, and J. W. White. 1973. On the nearshore
circulation of the Gulf of the Carpentaria, Australia-a study in uses of
satellite imagery (ERTS) in remotely accessible areas. Proc. Am. Soc.
Photogramm. 2: 717-736.
Tyler, J. E., and W. H. Richardson. 1958. Nephelometer for volume scattering
function in situ. J. Optc. Soc. Am. 48: 354-357.
u.S. Geological Survey.
Washington, D. C.
1971.
Environmental atlas of the United States.
Van Wie, P., M. Stein, E. Puccinelli, and B. Fields.
rectification system preliminary documentation.
Flight Center.
1975. Landsat cigital
NASA Goddard Space
Whitfield, W. K., Jr., and D. S. Beaumariage.
Apalachicola Bay: past, present, future.
Apalachicola Drainage System. 1: 130-140.
Williams, D: L., and G. F. Haver. 1976. Forest land management by satellite:
Landsat-derived infonnation as input to a forest inventory system.
Intralab Project 75-1. NASA Goddard Space Flight Center.
1975. Shellfish management in
Proc. of the Conf. on the
Williamson, A. N., and W. E. Grabau. 1973. Sediment concentration mapping in
tidal estuaries. Proc. 3rd Earth Resour. Technol. Satellite-1 Symp. 1B:
1347-1377 .
Woodall, W. R., Jr., and J. 8. Wallace.
small southern Appalachian streams.
1972. The benthic fauna in four
Am. ~1idl. Nat. 88: 393-407.
lC:::
-------
Wright, F. F., and G. D. Sharma. 1973. ERTS studies of Alaskan coastal
circulation. Proc. Am. Soc. of Photogramm. 2: 872-882.
Wright, F. F., G. O. Sharma, and O. C. Burbank. 1973. ERTS-l observations of
sea surface circulation and sediment transport, Cook Inlet, Al~ska. Prac.
Symp. Significant Results Obtained from Earth Resour. Technol.
Satellite-1. 1: 1315-1322.
Yarger, H. L., J. C. Coiner, G. W. James, L. M. Magnuson, J. R. McCauley, and
G. R. Marzolf. 1972. ERTS-l Study of Reservoirs in Kansas. Proc. 8th
Int. Symp. Remote Sensing of Environ. 2: 1477-1490.
Yen t sc h, C.. S., a nd W. P. Owe n . 1975 .
NASA Res. Rept. NASA-CR-139145.
Ocean color a three component system.
Yost, E. F., R. Hollman, J. Alexander, and R. Nuzzi. 1973. An
i nterdi sc i pl i nary study of ,the, estuari ne and coastal oceanography of
Island Sound and adjacent New York coastal waters. Proc. 3rd Earth
Resour. Technol. Satellite-I. Symp. 1B: 1607-1618.
81 ack
In
-------
APPEND IX
SAMPLING STATION LANDSAT IMAGE COLOR
AND RELATED MEAN MONTHLY IN SITU DATA
Water quality data were not collected at the time of the satellite
overpasses for the historic images, 1973 to 1976, studies in this research.
direct correlation between satellite and in situ water quality data was,
therefore, impossible.
A
The location of various water classes appeared to be fairly consistent
from year to year when under simil ar environmental conditions- (wind, tide,
river flow, etc.). A comparison of monthly station means for pertinent water
quality data with the color of a water type derived from the classified
Landsat image at the same station during the same month was conducted. This
comparison (Tables A-I, A-8) at least partially, confirmed that the various
water classes, the acidic swamp runoff in particular, had been properly
identified for they occured in areas similar to those observed by surface
investigations from research vessels.
The water quality parameters under investigation were Secchi depth, water
color, turbidity, salinity, and pH. Occasionally water quality data existed
for a sampl i ng station that was not in the area encompassed by the image. The
image color v/ould in this case be labeled "Off Image" for the color could not
be determined. The sampling stations are located on the map (see map in
pocket). Water quality data were provided by Or. Livingston (FSU) under
Florida Sea Grant and U.S. EPA funded projects.
1~3
-------
TABLE A-I. COLOR OF SAMPLING STATIONS ON LANDSAT IMAGE,
17 FEBRUARY 1973, AND RELATED IN SITU MEAN MONTHLY STATION
DATA, FEBRUARY 1973, FOR ALL PERTINENT AVAILABLE STATIONS
(Livingston, pers. comm., 1978)
Secchi 1.~ate r
Sampl i ng Image Depth Color Turbidity Salinity
Station Co lor (m) (PCU) (JTU) (0/00) pH
001 Off Image 0.2 180 205 0
002 White 0.6 105 38' 0
003 Li ght Brown 0.3 130 -60
004 Light Brown 0.2 155 88 0
005 Da rk Brown 0.2 160 90 0
006 Da rk Brown 0.3 100 52 0
007 Wh it e 0.4 110 37 0
01A Off Image 0.7 0 27 5.3
01B Green 0.3 75 40 2.3
01C Gree n 1.3 0 16 5.8
OlE Green 2.7
01F Off Image 3.2
02A Green 0 32 2.7
028 Off Image 22 4.2
05A Sa 1 man 0.1 370 160 0
lCS
-------
TABLE A-2. COLOR OF SMIPLING STATIONS ON LANDSAT IMAGE,
13 APRIL 1973, AND RELATED IN" SITU MEAN MONTHLY STATION
DATA, APRIL 1973, FOR ALL PERTINENT AVAILABLE STATIONS
(Livingston, pers. comm., 1978)
Secchi Water
Sampl i ng Image De pt h Co lor Turbidity Salinity
Station Co lor (m) (PCU) (JTU) (0/00) pH
002 Dark Brown 0.4 90 43 0
004 Tan 0.4 55 44 0.8
005 Tan 0.5 90 58 0.5
006 Tan 0.4 85 56 0
01B Da rk Brown 1.4 17 9 19.5
OlC Green 0.8 8 7 26.8
010 Off Image 12.9
OlE Tan 21.8
01F Off Image 28.6
02A Green 7 21.8
02B Off Image 5 12.9
05A Sa 1 mo n 0.3 75 82 0
110
-------
TABLE A-3. COLOR OF SAtA,PLING STATIONS ON LANDSA T I~1AGE
21 DECEMBER 1973, AND RELATED IN SITU MEAN MONTHLY STATION
DATA, DECEMBER 1973, FOR ALL PERTINENT AVAILABLE STATIONS
(Livingston, pers. comm., 1978)
Secchi Wate r
Sampl i ng Image Depth Color Turbi dity Salinity
Station Co lor (m) (PCU) (JTU) (0/00) pH
001 Da rk Green 1.1 15 5 5.2
002 Wh i t e
004 White 0.9 18 5 2.9
005 Salmon . .0.8 25 10 0
007 Wh i te 4
alA White 1.3 9 5 21.8
01B li ght Green 1.5 10 5 20.1
01C Whi te 1.4 12 8 18.4
02A Wh i te
02B Green
05A Red 1.0 35 5 2.9
III
-------
TABLE A-4. COLOR OF SAMPLING STATIONS ON LANDSAT IMAGE,
3 MARCH 1974, AND RELATED IN SITU MEAN MONTHLY STATION
DATA, MARCH 1974, FOR ALL PERTINENT AVAILABLE STATIONS
(Livingston, pers. comm., 1978)
Secchi \~ater
Sampl i ng Image De pth Color Turbi d ity Salinity
Station Co lor (m) ( PCU) (JTU) (0/00) pH
001 Wh i te 0.8 20 17 10.3
002 Wh i t e 0.6 30 8 0
004 White 0.8 40 7
005 Wh i t e 0.7 40 22 2.3
007 Wh ite 0.8 50 11 0
01B ~'Jh it e 0.9 20 4 11.5
01C White 0.9 20 7 2.3
05A Wh i t e 0.6 50 17 0
112
-------
TABLE A-5. COLOR OF S~~PLING STATIONS ON LANDSAT IMAGE,
23 OCTOBER 1974, AND RELATED IN SITU MEAN MONTHLY STATION
OATA, OCTOBER 1974, FOR ALL PERTINENT AVAILABLE STATIONS
(Livingston, pers. comm., 1978)
Secchi Water
Sampling Image Depth Co lor Turbidi ty Sa 1 in ity
Station Co lor (m) (PCU) (JTU) (0/00) pH
001 White 1.6 20 1 23.0
002 White 1.6 20 2 10.1
003 ~Jh ite 1.1 15 2 11.8
004 Sa 1 man 1.0 40 2 11.2
005 Salmon 1.0 25 3 10.1
006 Sa 1 mo n 1.2 10 3 24.0 6.9
007 IAh i te
01A Gal d 1.7 0 1 25.7
01B White 2.2 20 1 25.2
01C . White 1.9 10 1 33.7
OlE White 1.0 30 1 28.4
01X Gal d 1.3 22 1 25.9
02A Gold
028 ~/h it e
04A Salmon 1.1 5 1
05A Salmon 1.2 20 2. 27.3
058 Red 1.0 55 2 24.2
05C Red 1.0 55 2 25.7
1 ' -
... ~ ...~
-------
TABLE A-6. COLOR OF SAMPLING STATIONS ON LANDSAT IMAGE,
26 FEBRUARY 1975, AND RELATED IN SITU MEAN MONTHLY STATION
DATA, FEBRUARY 1975, FOR ALL PERTINENT AVAILABLE STATIONS
(Livingston, pers. comm., 1978)
Secchi Wate r
Sampl i ng Image De p t h Co lor Turbidity Salinity
Station Color (m) (PCU) (JTU) (0/00) pH
001 Deep Blue 0.9 75 3 5.2 7.4
002 White 0.8 100 10 2.0 7.2
003 White 0.5 110 16 2.5 7.3
004 Whi te 0.6 105 16 3.0 7.5
005 Wh i t e 0.6 100 16 3.0 7.1
006 Li ght Brown 0.3 120 18 3.0 6.7
007 Wh i t e
01A Gal d 1.1 60 2 7.9 7.3
01B Gal d 1.4 60 1 9.0 7.8
01C Green 1.1 40 2 10.6 7.5
01Ł Green 0.6 40 2 9.f> 7.7
01X Green 1.1 30 2 12.8 8.0
02A Dark Brown
02B Gal d
04A Light Brown 0.3 150 23 3.0 6.1
05A Light Brown 0.4 130 20 2.5 7.2
05B Red 0.3 190 5 3.0 6.2
05C Red 0.3 285 5 3.0 6.0
l' .1
...'
-------
TABLE A-7. COLOR OF SAMPLING STATIONS ON LANDSAT iMAGE,
20 JULY 1975, AND RELATED IN SITU MEAN MONTHLY STATION
DATA, JULY 1975, FOR ALL PERTINENT AVAILABLE STATIONS
(Livingston, pers. comm., 1978)
Secchi Wa te r
Sampling Image De pth Co 1 or Turbidity Salinity
Stat ion Co lor (m) (PCU) (JTU) (0/00) pH
001 li ght Brown
002 Dark Brown
003 Da rk Brm't'Tl 0.7 50 6 7.9 8.0
004 Da rk Brown 0.4 55 15 14.4 8.2
005 Wh it e 0.9 110- 12 12. J - 8.0 -
006 Da rk Brown 0.4 55 14 10.1 7.9
007 vJh it e
01A Gal d 1.1 35 5 17.8
01B Go 1 d 1.1 - 33 8 30.2 7.8
01C li ght Brown 10 15 8.2
OlE Light Brown 10 25 8.1
01X Light Brown 1.2 22 6 20.6 7.9
02A Da rk Brown
02B Green
04A Dark Brown 1.5 6 2 2.1 7.9
05A Gal d 0.8 60 15 8.5 8.0
05B Da rk Brown 0.8 117 4 1.3 6.3
05C Black 0.6 148 2 2.1 6.7
1 . r-
.Ll::
-------
TABLE A-8. COLOR OF SAMPLING STATIONS ON LANDSAT IMAGE,
19 AUGUST 1976, AND RELATED IN SITU MEAN MONTHLY STATION
DATA, AUGUST 1976, FOR ALL PERTINENT AVAILABLE STATIONS
(livingston, pers. comm., 1978)
Secchi \~ater
Sampl i ng Image. De pth Color Turbidity Salinity
Station Co lor (m) (PCU) (JTU) (0/00) pH
001 Wh i te 1.0 5
002 Wh i t e 0.6 2 13 4.0 0
003 \oih i t e 0.6 20 " 15 12.2 G.9
004 Dark 8rown 1.2 20 7 16.6 7.9
005 Sa 1 mo n 0.6 20 11 15.0 7.6
006 Sa 1 mo n 0.5 10 15 3.5 8.1
01A Green 1.3 9 18.0
01B "Gol d 2.0 4 30.8
01C Da rk Brown 1.2 11 32.5
DIE \~h it e 0.6 10 31.5
OlX Wh ite 1.3 10 31.6
02A Wh it e
04A Red 0.9 14 8 4.6 8.6
048 Red 0.8 25 6 4.5 8.9
05A Red 0.7 30 17 12.2 7.5
058 Red 0.8 9 9 5.1 7.4
05C Red 0.8 9 9 5.8 7.6
ll~
-------
TECHNICAL REPORT DATA
{Please read l.'lSrr~crions 011 the rel'erse before complering}
1. REPORT NO. 12. 3. RECIPIENT"S ACCESSION NO.
.
4. TITLE AND SUBTITLE 5. REPORT DATE
LANDSAT ESTUARINE WATER QUALITY ASSESSMENT OF
SILVICULTURE AND DREDGING ACTIVITIES 6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S) 8. PERFORMING ORGANIZATION REPORT NO.
J. M. Hill
9. PERFORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM ELEMENT NO.
Environmental Monitoring Systems Laboratory 1BD613
Office of Research and Development 11. CONTRACT/GRANT NO.
U.S. Environmental Protection Agency
Las Vegas, NV 89114 NjA
12. SPONSORING AGENCY NAME AND ADDRESS 13. TYPE OF REPORT AND PERIOD COVERED
U.S. Environmental Protection Agency--Las Vegas, NV
Office of Research and Development 14. SPONSORING AGENCY CODE
Envi ronmental Monitori ng. Systems. Laboratory
Las Veaas. Nevada 89114 EPA/600/07
15. SUPPLEMENTARY NOTES
~he author's present address is: College of Engineering, Louisiana State University
Baton Rouqe. Louisiana 79803
16. ABSTRACT
This report describes the application of Landsat multispectral scanning to
estuarine water quality, with specific reference to dredging and silviculture
practices.
Water quality data collected biweekly since 1972 in the Apalachicola, Bay,
Florida, by Florida State University, and Landsat data covering the same
geographical area were used as data base for these correlative investigations.
The research indicates that Landsat can provide temporal cause and effect
information relating to land-use changes and water quality. ~later types, based on
water color, were easily discriminated at different times of the year and under
varying environmental conditions. Water patterns, which are nearly impossible to
acquire under traditional sampling schemes, were readily discerned.
Among other advantages, Landsat-derived distributions of water classes provide a
seasonal overview of an area, enabling the most advantageous placement of sampling
stations, and, as a consequence, provide the capability to more accurately
extrapolate data spatially over large areas from a minimal number of sampling
points.
17. -
KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS b.IDENTIFJERS/OPEN ENDED TERMS C. COSA TI Field/Group
water qual ity Apalachicola Bay, Flori da 48G
bays and estuaries Landsat 680
silviculture
multispectral scanning
18. DISTRIBUTION STATEMENT 19. SE~j't~t[Ass-iPH:D'iS Reportj 21. NO. OF PAGES
RELEASE TO PUBLIC 20. S!:CU?ITY CLASS (This pagl:?) 122. PRICE
UNCLASSIFIED
EPA Form 2220-1 (Rev. 4-77)
PREVIOUS EDITION 15 OeSOLETE
------- |