United States
Environmental Protection
Agency
BPA-600 3 80 07b
July 1980
Research and Development
xvEPA
A Remote Sensing
Technique to Monitor
Cladophora in the
Great Lakes
LIBRARY
v
U.S.
EDISOH, M.Jk O881?
i- ..-. - .
-------
RESEARCH REPORTING SERIES
Research reports, of the Office of Research and Deve'ooment U S E v. ronmenta!
Protection Agon; •/ nave been grouped Tito nine series These nine broad cate-
gories werp established to facilitate further development and application of en-
vironments tecrinolcgv' Elimination of trad:ttonai grouping was consciously
planned *o *os'f ' 'f'^nology trans'er ana -\ max'miim interface m related fields
The nme serif •-, -v-
"J SoCiOOl On, '^HC
6 S^ienC;" and Techmia1 Assi ss'^cn; Reports ^S"
7 Interag-'poy Energy-Environment Research ana Development
8 Suf'c -.11 Repcr;s
Tnis report nac o-on assigned to trie ECOLOC3ICAL RESEARCH series This series
descrioes r,-^,earcr' on the effects of polk^on on humans plant and anim.-ii spe-
Cies a"d md^^r.^l1- Problems are assessed 'or the r lon,> ana shc:rt term mflu
em es investigations include tormatio0 transport a >d parhway studies to d-'ter-
niir'e the iate •• ' no'l jtants and their etfee*s This worK |)rov les tpe technical bbs s
for sett|rifj s'acl.'>ras fo ni:riir>i 79 undesiraijk' 'hang^s i'i ii>.'ng '.rqamsms m the
aqua'1:, 'err^str a1 and atfiosphenc, env'rr)r>men;c
onal Teohiiica! Infomic-
-------
EPA-600/3-80-075
July 1980
A REMOTE SENSING TECHNIQUE TO MONITOR
CLADOPHOPA IN THE GREAT LAKES
by
Fred J. Tanis
Environmental Research Institute
of Michigan
Ann Arbor, Michigan 48107
Grant No. R803611
Project Officer
Michael D. Mullin
Large Lakes Research Station
Environmental Research Laboratory - Duluth
Grosse He, Michigan 48138
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
DULUTH, MINNESOTA 55804
-------
DISCLAIMER
This report has been reviewed by the Environmental Research Laboratory-
Duluth, U. S. Environmental Protection Agency, and approved for publication.
Approval does not signify that the contents necessarily reflect the views and
policies of the U.S. Environmental Protection Agency, nor does mention of
trade names on commercial products constitute endorsement or recommendation
for use.
11
-------
FOREWORD
Our nation's Great Lakes are vital for all animals and plants, yet
our diverse uses of water—for recreation, food, energy, transportation,
and industry—may physically and chemically alter these lakes. The presence
of the nuisance alga Cladophora along parts of the Great Lakes' shoreline
is one of the negative consequences of excess inputs of pollutants into
the lakes and their contributing rivers and streams.
This report demonstrates the usefulness of airborne remote sensing
to monitor the amount and distribution of shoreline algae problems.
Michael D. Mullin, Ph.D.
Project Officer
Large Lakes Research Station
Environmental Protection Agency-Duluth
Grosse lie, Michigan
iii
-------
ABSTRACT
The feasibility of using an airborne multispectral scanner to monitor
shoreline algae problems has been demonstrated. Multispectral data were col-
lected at two sites on the U.S. Lake Ontario shoreline. Computer generated
color maps were produced to show spatial distribution of Cladophova in the
nearshore zone and to estimate standing crop.
Ground truth data as unit samples of bottom vegetation were collected at
several locations which were marked in the imagery by surface floats.
A depth invariant model based upon principal component analysis was used
to process seven passive bands between 0.46 and 0.70 ym. Spectral features of
were related to measured standing crop.
This report was submitted by the Environmental Research Institute of
Michigan in fulfillment of Grant No. R803611011 under the sponsorship of the
U.S. Environmental Protection Agency. All work was completed by October, 1977.
IV
-------
CONTENTS
Foreword ill
Abstract iv
Figures vi
Acknowledgments vii
1. Conclusions 1
2. Recommendations 2
3. Introduction 3
4. Methods 4
5. Ground Truth Program 12
6. Results and Discussion 17
References 27
-------
FIGURES
Number Page
1 Location of Lake Ontario Cladophora monitoring sites 5
2 Geometry of airborne scanning 6
3 Development of depth invariant signatures 9
4 Surface float 12
5 Spectral bottom reflectances measured at Hamlin Beach,
New York on June 15, 1977 16
6 Projection of training sets on Cladophora principal
components C^ and C~ 19
7 Relationship between standing crop and the the
multispectral Cladophora measure 20
8 Cladophora recognition map, distribution and standing
crop (g dry wt/m^), site A, Hamlin Beach, New York 21
9 Aerial photograph, site A, Hamlin Beach, New York 22
10 Cladophora recognition map, distribution and standing
crop (g dry wt/m^), site B, Hilton Beach, New York 23
11 Aerial photograph, site B, Hilton Beach, New York 24
12 Shoreline profile, Cladophora standing crop (10-* kg dry
wt/m), site B, Hilton Beach, New York 25
VI
-------
ACKNOWLEDGMENTS
Several individuals contributed to the project. In particular, Mr.
Robert A. Shuchman and Mr. Carl Davis of ERIM and Dr. Warren Flint and Mr.
Joseph Hichar of the Great Lake Laboratory SUNY at Buffalo are acknowledged
for their help with the field work. In addition, Mr. Kenneth Knorr is acknow-
ledged with sincere thanks for his help with the multispectral processing.
Thanks are also extended to Ms. Nancy Moon and Ms. Evelyn Wrabel for secre-
tarial support.
-------
SECTION 1
CONCLUSIONS
Results from the airborne multispectral data collection and analysis
program on the Lake Ontario shoreline near Rochester, New York has led to the
following conclusions:
(1) The airborne method of collecting information on the extent and
growth of Cladophora in the nearshore zone has been demonstrated
as a practical method of data acquisition over extended stretches
of shoreline.
(2) Recognition of Cladophora. biomass from aircraft imagery requires
gathering Cladophora samples at known locations. Sampling locations
can be marked with highly reflective floats which are visible in the
multispectral imagery.
(3) The use of multispectral digital instrumentation allows development
of automatic recognition techniques. Results can be displayed as
high resolution maps for selected areas and to compile standing
crop estimates over large reaches of shoreline.
(4) Since both areal distribution and standing crop measurements are
possible, a method now exists for monitoring Cladophora growth on
a multiyear basis.
-------
SECTION 2
RECOMMENDATIONS
The airborne multispectral data collection and analysis program on the
Lake Ontario shoreline has led to the following recommendations and suggested
applications:
(1) Aircraft remote sensing should become a standard accepted technique
to monitor or to map the distribution and biomass of benthic algal
communities for the Great Lakes.
(2) Based upon these results and previous work completed in 1974, a
complete Cladophora survey and monitoring program can be initiated
for the Great Lakes (or portions thereof) leading to the establish-
ment of a systematic historical data base.
(3) Specific influences of local sources of nutrient or various environ-
mental contaminants on the growth and extent of Cladophora should be
monitored using this aerial technique with multiyear surveys.
Improvements in local water quality as a result of implementation
of remedial programs can be expected to affect Cladcphora growth and
therefore are observable in the airborne surveys.
-------
SECTION 3
INTRODUCTION
The filamenteous alga Cladophova GlomeTata grows abundantly along a num-
ber of Great Lake shorelines. This benthic form grows selectively on hard
surfaces and rocky substrates and appears to be the dominant species.
Cladophora has long been an annoying problem for shoreline property owners,
boaters, and users of public beach facilities. After an early and rapid
period of growth, filaments are easily broken during summer storms. For the
shoreline property owner subsequent decomposition of large masses of Cladophora
produces obnoxious conditions. In addition, accumulated algae can force the
closing of public recreational facilities.
Influx of nutrients from shoreline activities plays an essential role in
the nutrition of the benthic algae. The location of Cladophopa beds is fre-
quently a good biological indicator of a nutrient pollution source. Changes
in existing beds may indicate corresponding changes in the local cultural
nutrient sources.
Little is known about the distribution and biomass of benthic algae in
the Great Lakes. The International Joint Commission (IJC) committee for
surveillance has recommended that a CladophoTa monitoring program be developed.
Aircraft remote sensing technology provides an ideal method for collecting the
necessary information to assemble a Great Lakes Cladophova data base.
The Environmental Research Institute of Michigan (ERIM) has been involved
in a Cladophopa sensing program since the beginning of the IFYGL program in
1972. Under a research grant for the U.S. EPA (Grant No. 800778) ERIM studies
reported by Wezernak, Lyzenga, and Polcyn [1] in 1974 have shown how airborne
data can be used to estimate mass and extent of bottom growths.
The present study has been directed to further refine and develop multi-
spectral scanner techniques which allow automatic recognition of distribution
and biomass parameters when compiled with ground truth information. While
color aerial photography can be used to identify fields of benthic algae, it
cannot be easily processed quantitatively. The inherent problem in photo-
graphic analysis is correcting for changes due to water depth; which often
coincide with changes in algal density. Development on the other hand, of
depth-invariant digital techniques, for multispectral scanner data, can con-
veniently map benthic algae areal extent and biomass.
-------
SECTION 4
METHODS
Data were collected using the Environmental Research Institute of
Michigan (ERIM) remote sensing aircraft (C47). Multispectral data collection
was initiated along the U.S. shoreline on June 15, 1977 from altitudes of 300
meters and 500 meters. Data were collected at two sites, Hamlin Beach (Site
A) and Hilton Beach (Site B), New York, both of which were previously examined
in the 1972 flights (Figure 1).
MULTISPECTRAL SENSOR SYSTEM
The ERIM M-8 active/passive multispectral scanner was used for remote
sensing data collection [2]. Use of this instrument permits simultaneous
data collection in nine passive spectral bands over a wavelength range from
0.44 to 1.5 ym. Additional detector positions are available for spectral
bands in the ultraviolet and the infrared. The active data were collected
using a nitrogen-dye pulsed laser system operating at 0.52 ym.
What is recorded is time history of the returned laser pulse, including
surface reflection, scattering from water column, and bottom reflection.
These components are separated in the time history, permitting independent
analysis.
Pulsed laser data were collected on the June 15, 1977 flight at 300
meters altitude. These data could be potentially used in a Cladophora moni-
toring program to obtain a series of water depths (through extraction of the
time difference between surface and bottom returns) for a single bottom type.
Such information is useful to establish depth invariant signals where water
depth data cannot be collected by survey crews. For the present program these
laser data have not been processed.
Signals from passive detectors (as well as laser return pulse time his-
tory) are recorded on high density digital tape (HDDT) from subsequent image
reconstruction and data processing are performed. Data from reference lamps
and the solar irradiance sensors are also recorded. The scanner is positioned
in the aircraft so as to provide continuous scanning perpendicular to the
flight line as shown in Figure 2. As the aircraft advances, the rotating
mirror scans the scene over a 90° field of view continuously so as to assemble
an image line by line.
-------
Figure 1. Location of Lake Ontario Cladophora monitoring sites.
-------
•4-
Flight
Direction
Rot
Scan
^M
Radi;
IF'Irrtrrmir'sl . . .. .. . T*»«A
ating
Mirror x
ition O
Telescope
N<7f=De
*••«•-
Recorder
Lector
Aircraft Skin
From Ground
—Total Field of View
Ground
Resolution
Element
Figure 2. Geometry of airborne scanning.
-------
The scanner imagery is stabilized electronically about the roll axis of
the aircraft. No correction is made for aircraft pitch and yaw motion although
the average yaw angle of the aircraft is recorded and the gross yaw correction
can be made during data processing. The aircraft is flown at a constant pre-
selected ground speed and altitude to establish the imagery scale factor. The
geometrical accuracy of the imagery can be improved in data processing by
trial and error in the current experimental system or by more accurate sensing
of aircraft motion for an operational system.
In addition to data collected by the multispectral scanner, 70 mm black
and white panchromatic and color photographic records were collected along the
flight line.
Cladophora RECOGNITION BY REMOTE SENSING
The radiant energy received by an optical remote sensing instrumental
system oriented toward a water body consists of components of scattered radia-
tion from the intervening atmosphere and reflected radiation from the scene.
At depths (and wavelengths) where radiation penetrates to the bottom, the
latter component includes bottom reflectance, volume reflectance from the
water column, and reflectance from the water surface. In the ERIM multi-
spectral scanner, incident radiation is measured by a sun sensor and reflected
or emanating radiation is measured by the scanner system. Utilization of this
information to describe bottom features which result in radiation changes
requires a model or models which account for the important interactions and
their effects. Since the radiation reflected by bottom features must pass
through the intervening water column, the amount of radiation reaching the
sensor at a given wavelength is dependent on the volume attenuation character-
istics of the water. Attenuation is a function of the thickness of the water
layer and the absorption and scattering properties of the water.
In practice, the sensor output is recorded as voltage (V) on magnetic
tape. Neglecting atmospheric effects, the bottom-reflected signal in a given
spectral channel may be written as:
17 TTC. _L i -(sec 8 + sec )a,z
VA = VSX + kApAe X
where V, = voltage received
VS-, = surface reflectance
A
k, = constant which incorporates solar irradiances and scanner
characteristics
p, = bottom reflectance
9 = viewing angle (from vertical)
-------
Since water depth is variable, the signal received from an identical
bottom feature will vary with depth. From an operational standpoint, a depth
invariant model is required which in effect permits removal of the signal
because of overlying waters and water surface effects. One approach to the
problem is the use of a ratio of two spectral bands with similar water atten-
uation coefficients. Writing the above expression in terms of two spectral
bands results in:
V - VS p -(sec 6 + sec cf>) (a - a )z
where — = k .
J£ n -J
In cases where a1 is equal to a. ,
V - VS
•
-------
Depth Invariant
Hyperplane
Principal Depth
Component
Multispectral Data
x
n
Figure 3. Development of depth invariant signatures.
-------
used to derive the direction cosines for the depth vector from data with a
uniform bottom type. In order to analyze the projected feature space for
Cladophor>a spectral variants, data from the ground truth derived training
sets are first converted to depth invariant images by projecting along the
depth vector. The principal components of the transformed training set in
turn yield orthogonal directions of maximum variation. The objective is to
find the greatest spectral variants in the reflectance data and relate the
values of that variant to the measured Cladophora standing crop.
A rationale for this approach is as follows. The standing crop per unit
area will vary with the number of filaments per unit area and the nominal
filament length. As these parameters change, so does the amount of chlorophyll
a. and transparent cell wall material which will in turn change the spectral
characteristics of the bottom reflected radiance. The spectral measure of
Cladophora standing crop developed for the present data, is the vector pro-
jection of the spectral signature on the principal Cladophora spectral variant.
WEATHER AND SEA STATE REQUIREMENTS
Three related environmental conditions - sea state, solar radiation con-
ditions, and cloud cover - limit the observation of Cladophopa by aircraft.
Onshore waves will bring littoral bottom sediments into suspension causing
increased water turbidity. In Lake Ontario, a 10 mph onshore wind will
severely increase water turbidity in a matter of hours. Once suspended by
wave action, particulates may take one or two days to settle, again producing
acceptable water transparency.
During mid-summer, the southshore of Lake Ontario is frequently impacted
with windy and stormy weather. The frequently turbulent sea state combined
with debris from open lake productivity and broken Cladophora filaments pro-
duces nearly continuous turbid conditions. This was exactly the situation
when a Cladophora reconnaissance flight was attempted in early August 1975.
Conditions turned out to be sufficiently poor so that the collected multi-
spectral data could not be processed for Cladophova distribution.
Besides sea state conditions, cloud and solar radiation conditions are
important to obtain useful imagery. A thirty or forty percent cloud cover is
considered the maximum tolerable. Overcast conditions reduce the incident
sunlight so that insufficient contrast and the signal to noise ratio become
problems. The June 1977 flight conditions were considered ideal. Two days
of calm weather preceded the day of flyover. On that day the sea state was
calm and skies were clear. Secchi transparency was 4.0 meters.
DATA PROCESSING
Multispectral scanner data collected over Lake Ontario were processed for
Cladophora recognition with the previously described digital techniques. Air-
craft high density tapes were first converted to computer compatible tapes
(CCT's) during which corrections were made for scan angle and surface reflec-
tance effects.
10
-------
For each of the two shoreline sites, line prints were made for selected
channels. Within each scene, pixels were determined which are co-located with
ground truth sampling stations and fathometer transects.
For this experiment sampling station positions were made visible in the
multispectral imagery through the use of station marker floats. Two types of
training sets were abstracted from the data: signals from areas with uniform
sandy bottom type with varying water depth and signals from areas immediately
surrounding each ground truth station. The first set was used to establish
multispectral variants for water depth while the second set was used to define
the relationship between CladophoPa standing crop and the multispectral signa-
ture. This latter relationship was obtained with computer algorithms which
included deriving a depth invariant subset and performing principal component
analysis. The resultant is a spectral vector which was used to describe the
variation observed in Cladophora standing crop.
11
-------
SECTION 5
GROUND TRUTH PROGRAM
Ground truth data consisting primarily of Cladophora bottom samples and
water depth information were collected at two sites on Lake Ontario east of
Rochester, New York. These two sites were included in the 1972 IFYGL Lake
Ontario Cladaphora flights (EPA Grant 660/3-74-028) [1]. The aircraft flights
were made on June 15, 1977. Field sampling work was conducted from June 14
through June 16, 1977. Staff members from the Great Lakes Laboratory, SUNY
at Buffalo assisted in the bottom survey.
FIELD METHODS
Samples were collected at specific stations at each site located in dif-
ferent bottom types and Cladophora densities. Stations were located by sur-
face float or on transects through two or more such floats. The floats con-
sisted of four inflatable panels in a cross pattern (see Figure 4). The cross
pattern was distinctive and could be easily recognized in the resulting
imagery. The panels were leathered with four suspended anchors, one on the
end of each float and one central bottom anchor to fix the float position.
This float design was inexpensive, easily assembled, expendable, and produced
a high reflectance. Durability was not required since these floats were used
only under good weather conditions required for the aircraft mission. Bottle
buoys were used to mark locations at other times.
Five marker floats were employed at each site. A total of twenty one
samples were collected between the two sites. At each station samples were
gathered by cropping all plant material within a 0.17m2 wire hoop.
In addition to the bottom samples, water depth and other information
were recorded. At many of the stations, underwater photographs were taken
which provided valuable additional information on bottom conditions. Samples
were returned to ERIM and subsequently analyzed for wet, dry, and free-ash
weight according to Standard Methods [4].
FIELD DATA
Cladaphora ground truth information is summarized in Tables 1 and 2.
Besides these data light attenuation, transparency, and spectral reflectance
measurements of bottom materials were made using an ISCO Model SR spectrometer.
Percent reflectance values over the useful spectral wavelengths for the various
12
-------
bottom types are shown in Figure 5. The fact that these values are quite
different for each bottom type suggests that bottom classification can be
accomplished with remote sensing techniques.
Figure 4. Surface float.
13
-------
TABLE 1. CLADOPHORA GROUND TRUTH DATA
SITE A, HAMLIN BEACH STATE PARK, NEW YORK
Date: 15 June 1977
Float Water
stations depth (m)
Al
A2
A3
A4
A5
Transect
stations
ATI
AT2
ATA
ATS
AT6
ATS
3.3
2.6
1.8
1.2
4.1
6.1
4.6
2.9
2.4
1.8
1.2
Wet weight
Description (g/m2)
Cobbles, patchy growth, 92
medium density, 2-3"
filaments, 40% cover
Boulders interspaced 300
with sand, healthy
growth, medium density,
3-4" filaments, 45%
cover
Medium cobbles, medium 460
density, bleached end
filaments, 5-7" filaments
95% cover
Small to medium cobbles, 2100
dense growth, bleached
end filaments, 8-15"
filaments, more than
95% cover
Small to medium cobbles, 140
patchy-healthy growth,
medium density, 1-2"
filaments, 50% cover,
high sediment content
Sand
1% cover <10
40% cover 250
Poor conditions, 30% cover 32
Healthy, 90% cover 670
Healthy, more than 95% cover 940
Dry weight
(a/m2)
46.7
95.6
134.0
503.0
82.1
0.7
111.0
14.3
167.0
212.0
14
-------
TABLE 2. CLADOPHORA GROUND TRUTH DATA
SITE B, HILTON BEACH, HILTON, NEW YORK
Date: 16 June 1977
Float
stations
Bl
B2
B3
B4
B5
Transect
stations
BT1
BT2
BT4
BT5
BT7
Water
depth (m)
3.2
2.7
Wet weight
2.8
1.8
1.7
4.0
3.7
2.7
2.1
2.1
Description
Medium cobbles, stubby-
patchy growths, 1-2"
filaments, 15% cover
Sandy patch surrounded
by Cladophora, sparse
to dense growth, 75%
cover
Large cobbles, dense
growth, 4-12" filaments,
90% cover
Large flat rocks and
small cobbles, dense
growth, 9-12" filaments,
more than 95% cover
Medium cobbles, dense
growth, 4-8" filaments,
more than 95% cover
20
High
Sediment
Content
520
660
Healthy, 1" filaments,
40% cover
Pebbles and large
boulders, 1-3" filaments,
60% cover
Sand in large patch
More than 95% cover
30% cover
830
138
141
580
66
Dry weight
(g/m2)
9.87
71.10
117.00
166.00
153.0
65.1
36.4
83.9
23.7
15
-------
g
o
20
10
9
8
7
• Sand
O Rocky Bottom
• Cladophora
c
ri
o
o o
1
1
.45
.50 .55 .60
Spectral Wavelength (/j m)
.65
.70
Figure 5. Spectral bottom reflectances measured at Hamlin
Beach, New York on June 15, 1977.
16
-------
SECTION 6
RESULTS AND DISCUSSION
Aircraft multispectral and ground sampling data were collected under
excellent conditions which occurred on June 15, 1977. A total of six flight
passes were made over the two sites during the times of minimum sun glint.
Of all the data, only the morning flight data were free of any solar glint
problem. The 500 meter altitude data were selected for processing at both
sites because it produced a good swath width coverage of 1000 meters. A
2,100 scan line (2.1 km) section located in the vicinity of the Cladophora
sampling data was assembled for analysis.
A total of 21 Cladophova samples were available to calibrate the two
sites. A 200 point area (15 meters x 15 meters) surrounding each ground truth
site was averaged over each of seven multispectral channels. The resulting
Cladophora matrix was 21 by 7 values. Channels selected for analysis included:
Band 1:
Band 2:
Band 3:
Band 4:
Band 5:
Band 6:
0.46
0.48
0.50
0.52
0.55
0.58
- 0.49ym
- 0.52jim
- 0.54ym
- 0.57ym
- 0.60ym
- 0 . 64ym
Band 7: 0.62 - 0.70ym.
The sand bottom multispectral set (sand) which was generated from portions of
five transect sets contained 48 values. Each value represents an average over
3x3 pixels. Depth range for the sand set was one to five meters, which
covers the depth range at which Ctadop'hox'a. were found.
Principal component analysis (PCA) of the sand set produced essentially a
single component of data in the direction of multispectral variation due to
water depth. The eigenvector for the first component was (0.3787, 0.3786,
0.3690, 0.3840, 0.3729, 0.3790). The corresponding eigenvector accounted for
95.6% of the total variation in the sand data set.
Because of the exponential nature of light attenuation in water, the
linear depth vector is only approximate. Absolute errors were estimated to be
minimal at shallow depths and not to exceed 5% of any single multispectral
channel at a depth of five meters. Since the major portion of Cladophora
biomass is located at depths much less than five meters, errors resulting from
the linear assumption are not considered to be great.
17
-------
Using the above sand eigenvector a depth invariant subset of the
Cladophor>a data set (CLAD), was derived with matrix rotation and projection
methods. The derived set could then be examined using PCA for principal spec-
tral features which were depth-invariant. Eigenvectors C-i and €2 are corres-
ponding eigenvalues obtained from analysis of the depth invariant CLAD training
set were:
C±: (0.3431, 0.4027, 0.3829, -0.3951, -0.3924, -0.4025, -0.3172) and
C2: (0.5737, -0.0768, -0.2962, -0.1186, -0.2990, -0.0048, -0.6882);
Percent of total variation explained by C-^ = 87.0%;
Percent of total variation explained by C2 = 9.2%;
By all other components 3.8%.
Thus, the depth invariant Cladophora data (CLAD) can be practically repre-
sented by a single component C. Projecting all of the SAND/CLAD data onto the
C-^/C2 plane shows how the data can be spectrally separated (see Figure 6).
Furthermore, the CLAD data which project furthest along the GX vector from C2
had the greatest biomass. CladophoPO. samples which resulted in a projection
close to C2 had low biomass and sparse cover. The separation dimensionally
between the sand and Cladophora data allows the use of a discriminant based
upon [C^,C23 which can separate these two basic bottom types. The C^ vector
alone becomes in this instance a good measure of the amount of Cladophora
present.
The relationship between observed standing crop as grams dry weight per
square meter and the derived multispectral measure, component vector Cl, is
shown in Figure 7. While there is much scatter, especially for the low values
of standing crop, a definite linear trend appears. The position of several of
the data points as plotted in Figure 7 were found to be suspect after closer
examination. These were: A5, B2 Cladophox>a samples have high sediment con-
tent; BT7 located on edge on rapdily changing bed of C'Ladophora; BT4, AT2, ATI
samples from sandy and rocky areas with very little Cladop'hor'a present.
A linear regression analysis was performed on the data shown in Figure 7
after deleting the above suspect data. This analysis produced a regression
correlation coefficient of 0.94 and a linear relationship of: Standing crop =
1.81 + 1.46 (C]_, D) (g dry weight/m^), where (C]_, D) is the vector projection
of GI on the multispectral data vector D.
By projecting the entire data set along C-. and classifying the sandy bot-
tom areas, the original seven dimensional multispectral data set was reduced
to a single variable which describes the amount of Cladophora present. Using
this derived Cladaphova measure of standing crop the multispectral data sets
for each of the two sites were converted into single channel digital maps.
Those maps were then in turn displayed on ERIM's MIDAS computer system by
level slicing the signal into distinct ranges of standing crop. Figures 8
through 12 show the resulting maps and corresponding aerial photographs. Using
site A as a validation test map and B as a calibration site, map values of pre-
dicted biomass compare quite well with that of field samples.
18
-------
DECISION
BOUNDS RY
SAND/CLAD Discriminant
-12.5 O.o 12.5 25.o 37.5 50.o
Cl
Legend
• Cladophora
• Sand
Figure 6. Projection of training sets on Cladophora
Principal Components C-^ and ^2-
19
-------
1000
100
§•
tl
O
bl>
e
10
0
ATS
BT7 (Trans.)
^
0
Bl
•a
5
4-J
CO
AT2
10 100 1000
Synthetic Channel (C..,D)
Figure 7. Relationship betweeen standing crop
and the multispectral Cladophora measure.
20
-------
6
•M
00
^^
cx
O
O
00
•H
tfl
C
O
•H
•H
S-l
4-1
to
•H
P.
c
O
•H
tl ^
« S
00 ><
O
O 5
a) 55
PH A
O O
^ rt
SX 0)
0)
a)
4J
-H
-------
Figure 9. Aerial photograph, Site A, Hamlin Beach, New York.
22
-------
-------
Figure 11. Aerial photograph, Site B, Hilton Beach, New York.
24
-------
I I I f I I f
o
o
co
o
o
r-
o
o
co
o
o
IT3
o
o
Tf
o
o
to
0)
-4_>
d>
M
5
I
"a!
o
»*•"
w
o
o
-------
Corresponding Cladophora maps produced from the 20 June 1972 data as
ratio imagery can be found as Figure 9, page 23 and Figure 10, page 24 in
the 1974 IFYGL Cladophora Report [1].
Comparison of the photographs with the computer maps indicates that very
good recognition of Cladophora resulted from using this automatic technique.
Obviously the accuracy of the maps decreases with decreasing standing crop.
Values below 50 grams per square meter are not considered reliable.
All of the sand/Cladophora boundaries nearshore are well defined. In
deep water the map shows mixed lower levels of standing crop as the Cladophora
beds thin out with decreasing cover and filament length. At depths greater
than five meters little Cladophora was observed on the rocky bottom probably
because of the low light level condition.
It is not practical to image all Cladophora digital data that would be
generated in any large survey using this technique. The detailed spatial
distribution of Cladophora can be obtained readily from aerial color photo-
graphy. The photography serves as a good permanent record of location and
qualitative extent of the Cladophora beds except where depth obscures the
standing crop. The quantification of the amount of Cladophora present is best
obtained by processing the multispectral data. Standing crop estimates per
unit of shoreline can then be calculated by summing over the Cladophora data
fields. In the present work the estimated standing crop was summed over each
scan line of the Hilton Beach data. The result is a Cladophora standing crop
shoreline profile which shows the total standing crop per meter of shoreline
(See Figure 12). The flexibility of the high resolution digitally processed
data will allow estimation of standing crop over any desired spatial unit.
While high resolution is needed to estimate standing crop results can
obviously be displayed with much coarser resolution.
In large scale survey programs where results can be displayed as low
resolution shoreline profiles incremental aircraft survey costs are expected
to be in the range of $50 to $100 per linear mile. Additional cost would be
expected for ground sampling and overall program planning and management.
26
-------
REFERENCES
1. Wezernak, C. T., D. R. Lyzenga, and F. C. Polcyn. Cladophora Distribu-
tion in Lake Ontario (IFYGL), U. S. Environmental Protection Agency,
EPA 660/3-74-028. December 1974.
2. Hasell, P. G., L. M. Peterson, F. J. Thomson, E. A. Work, and F. J.
Kriegler. Active and Passive Multispectral Scanner for Earth Resources
Applications 2, ERIM, Ann Arbor, Michigan, Report 115800-49-F. June
1977, 93 pgs.
3. Lyzenga, D. R. Passive Remote Sensing Techniques for Mapping Water Depth
and Bottom Features, Journal of OPTICS. (In press.)
4. Standard Methods for the Examination of Water and Wastewater, 14th
Edition. 1975.
27
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO
EPA-600/3-80-075
3 RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
A REMOTE SENSING TECHNIQUE TO MONITOR CLADOPHORA
IN THE GREAT LAKES
5. REPORT DATE
July 1980 issuing date
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Fred J. Tanis
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Environmental Research Institute of Michigan
Infrared and Optics Laboratory
Ann Arbor, Michigan 48107
10. PROGRAM ELEMENT NO.
1BA769
11. CONTRACT/GRANT NO.
Grant No. R803611
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Research Laboratory-Duluth
Office of Research and Development
U.S. Environmental Protection Agency
Duluth, Minnesota 55804
13. TYPE OF REPORT AND PERIOD COVERED
Final 1975-1977
14. SPONSORING AGENCY CODE
EPA/600/03
15. SUPPLEMENTARY NOTES
16. ABS, RACT
The feasibility of using an airborne multispectral scanner to monitor shoreline
algae problems has been demonstrated. Multispectral data were collected at two
sites on the U.S. Lake Ontario shoreline. Computer generated color maps were
produced to show spatial distribution of Cladophora in the nearshore zone and
to estimate standing crop.
Ground truth data as a unit samples of bottom vegetation were collected at
several locations which were marked in the imagery by surface floats.
A depth invariant model based upon principal component analysis was used
to process seven passive bands between 0.46 and 0.70 micrometers. Spectral
features of Cladophora were related to measured standing crop.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDEDTERMS
c. COSATI Field/Group
Remote Sensing
Monitors
Detectors
Eutrophication
Cladophora
Lake Ontario
Imagery
08/H
13. DISTRIBUTION STATEMENT
Release to Public
19 SECURITY CLASS (This Report)
UNCLASSIFIED
21 NO. OF PAGES
36
20. SECURITY CLASS (Tliis page/
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
28
USGPO 661-099 9/80
------- |